linux/mm/zsmalloc.c

2743 lines
65 KiB
C
Raw Normal View History

/*
* zsmalloc memory allocator
*
* Copyright (C) 2011 Nitin Gupta
* Copyright (C) 2012, 2013 Minchan Kim
*
* This code is released using a dual license strategy: BSD/GPL
* You can choose the license that better fits your requirements.
*
* Released under the terms of 3-clause BSD License
* Released under the terms of GNU General Public License Version 2.0
*/
/*
* Following is how we use various fields and flags of underlying
* struct page(s) to form a zspage.
*
* Usage of struct page fields:
* page->private: points to zspage
* page->index: links together all component pages of a zspage
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
* For the huge page, this is always 0, so we use this field
* to store handle.
* page->page_type: first object offset in a subpage of zspage
*
* Usage of struct page flags:
* PG_private: identifies the first component page
* PG_owner_priv_1: identifies the huge component page
*
*/
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
zsmalloc: replace per zpage lock with pool->migrate_lock The zsmalloc has used a bit for spin_lock in zpage handle to keep zpage object alive during several operations. However, it causes the problem for PREEMPT_RT as well as introducing too complicated. This patch replaces the bit spin_lock with pool->migrate_lock rwlock. It could make the code simple as well as zsmalloc work under PREEMPT_RT. The drawback is the pool->migrate_lock is bigger granuarity than per zpage lock so the contention would be higher than old when both IO-related operations(i.e., zsmalloc, zsfree, zs_[map|unmap]) and compaction(page/zpage migration) are going in parallel(*, the migrate_lock is rwlock and IO related functions are all read side lock so there is no contention). However, the write-side is fast enough(dominant overhead is just page copy) so it wouldn't affect much. If the lock granurity becomes more problem later, we could introduce table locks based on handle as a hash value. Link: https://lkml.kernel.org/r/20211115185909.3949505-9-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:14:13 +00:00
/*
* lock ordering:
* page_lock
zsmalloc: consolidate zs_pool's migrate_lock and size_class's locks Currently, zsmalloc has a hierarchy of locks, which includes a pool-level migrate_lock, and a lock for each size class. We have to obtain both locks in the hotpath in most cases anyway, except for zs_malloc. This exception will no longer exist when we introduce a LRU into the zs_pool for the new writeback functionality - we will need to obtain a pool-level lock to synchronize LRU handling even in zs_malloc. In preparation for zsmalloc writeback, consolidate these locks into a single pool-level lock, which drastically reduces the complexity of synchronization in zsmalloc. We have also benchmarked the lock consolidation to see the performance effect of this change on zram. First, we ran a synthetic FS workload on a server machine with 36 cores (same machine for all runs), using fs_mark -d ../zram1mnt -s 100000 -n 2500 -t 32 -k before and after for btrfs and ext4 on zram (FS usage is 80%). Here is the result (unit is file/second): With lock consolidation (btrfs): Average: 13520.2, Median: 13531.0, Stddev: 137.5961482019028 Without lock consolidation (btrfs): Average: 13487.2, Median: 13575.0, Stddev: 309.08283679298665 With lock consolidation (ext4): Average: 16824.4, Median: 16839.0, Stddev: 89.97388510006668 Without lock consolidation (ext4) Average: 16958.0, Median: 16986.0, Stddev: 194.7370021336469 As you can see, we observe a 0.3% regression for btrfs, and a 0.9% regression for ext4. This is a small, barely measurable difference in my opinion. For a more realistic scenario, we also tries building the kernel on zram. Here is the time it takes (in seconds): With lock consolidation (btrfs): real Average: 319.6, Median: 320.0, Stddev: 0.8944271909999159 user Average: 6894.2, Median: 6895.0, Stddev: 25.528415540334656 sys Average: 521.4, Median: 522.0, Stddev: 1.51657508881031 Without lock consolidation (btrfs): real Average: 319.8, Median: 320.0, Stddev: 0.8366600265340756 user Average: 6896.6, Median: 6899.0, Stddev: 16.04057355583023 sys Average: 520.6, Median: 521.0, Stddev: 1.140175425099138 With lock consolidation (ext4): real Average: 320.0, Median: 319.0, Stddev: 1.4142135623730951 user Average: 6896.8, Median: 6878.0, Stddev: 28.621670111997307 sys Average: 521.2, Median: 521.0, Stddev: 1.7888543819998317 Without lock consolidation (ext4) real Average: 319.6, Median: 319.0, Stddev: 0.8944271909999159 user Average: 6886.2, Median: 6887.0, Stddev: 16.93221781102523 sys Average: 520.4, Median: 520.0, Stddev: 1.140175425099138 The difference is entirely within the noise of a typical run on zram. This hardly justifies the complexity of maintaining both the pool lock and the class lock. In fact, for writeback, we would need to introduce yet another lock to prevent data races on the pool's LRU, further complicating the lock handling logic. IMHO, it is just better to collapse all of these into a single pool-level lock. Link: https://lkml.kernel.org/r/20221128191616.1261026-4-nphamcs@gmail.com Signed-off-by: Nhat Pham <nphamcs@gmail.com> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Acked-by: Minchan Kim <minchan@kernel.org> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-11-28 19:16:12 +00:00
* pool->lock
zsmalloc: replace per zpage lock with pool->migrate_lock The zsmalloc has used a bit for spin_lock in zpage handle to keep zpage object alive during several operations. However, it causes the problem for PREEMPT_RT as well as introducing too complicated. This patch replaces the bit spin_lock with pool->migrate_lock rwlock. It could make the code simple as well as zsmalloc work under PREEMPT_RT. The drawback is the pool->migrate_lock is bigger granuarity than per zpage lock so the contention would be higher than old when both IO-related operations(i.e., zsmalloc, zsfree, zs_[map|unmap]) and compaction(page/zpage migration) are going in parallel(*, the migrate_lock is rwlock and IO related functions are all read side lock so there is no contention). However, the write-side is fast enough(dominant overhead is just page copy) so it wouldn't affect much. If the lock granurity becomes more problem later, we could introduce table locks based on handle as a hash value. Link: https://lkml.kernel.org/r/20211115185909.3949505-9-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:14:13 +00:00
* zspage->lock
*/
#include <linux/module.h>
#include <linux/kernel.h>
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
#include <linux/sched.h>
#include <linux/bitops.h>
#include <linux/errno.h>
#include <linux/highmem.h>
#include <linux/string.h>
#include <linux/slab.h>
mm: introduce include/linux/pgtable.h The include/linux/pgtable.h is going to be the home of generic page table manipulation functions. Start with moving asm-generic/pgtable.h to include/linux/pgtable.h and make the latter include asm/pgtable.h. Signed-off-by: Mike Rapoport <rppt@linux.ibm.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Borislav Petkov <bp@alien8.de> Cc: Brian Cain <bcain@codeaurora.org> Cc: Catalin Marinas <catalin.marinas@arm.com> Cc: Chris Zankel <chris@zankel.net> Cc: "David S. Miller" <davem@davemloft.net> Cc: Geert Uytterhoeven <geert@linux-m68k.org> Cc: Greentime Hu <green.hu@gmail.com> Cc: Greg Ungerer <gerg@linux-m68k.org> Cc: Guan Xuetao <gxt@pku.edu.cn> Cc: Guo Ren <guoren@kernel.org> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Helge Deller <deller@gmx.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: Ley Foon Tan <ley.foon.tan@intel.com> Cc: Mark Salter <msalter@redhat.com> Cc: Matthew Wilcox <willy@infradead.org> Cc: Matt Turner <mattst88@gmail.com> Cc: Max Filippov <jcmvbkbc@gmail.com> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: Michal Simek <monstr@monstr.eu> Cc: Nick Hu <nickhu@andestech.com> Cc: Paul Walmsley <paul.walmsley@sifive.com> Cc: Richard Weinberger <richard@nod.at> Cc: Rich Felker <dalias@libc.org> Cc: Russell King <linux@armlinux.org.uk> Cc: Stafford Horne <shorne@gmail.com> Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Tony Luck <tony.luck@intel.com> Cc: Vincent Chen <deanbo422@gmail.com> Cc: Vineet Gupta <vgupta@synopsys.com> Cc: Will Deacon <will@kernel.org> Cc: Yoshinori Sato <ysato@users.sourceforge.jp> Link: http://lkml.kernel.org/r/20200514170327.31389-3-rppt@kernel.org Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-06-09 04:32:38 +00:00
#include <linux/pgtable.h>
mm: reorder includes after introduction of linux/pgtable.h The replacement of <asm/pgrable.h> with <linux/pgtable.h> made the include of the latter in the middle of asm includes. Fix this up with the aid of the below script and manual adjustments here and there. import sys import re if len(sys.argv) is not 3: print "USAGE: %s <file> <header>" % (sys.argv[0]) sys.exit(1) hdr_to_move="#include <linux/%s>" % sys.argv[2] moved = False in_hdrs = False with open(sys.argv[1], "r") as f: lines = f.readlines() for _line in lines: line = _line.rstrip(' ') if line == hdr_to_move: continue if line.startswith("#include <linux/"): in_hdrs = True elif not moved and in_hdrs: moved = True print hdr_to_move print line Signed-off-by: Mike Rapoport <rppt@linux.ibm.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Borislav Petkov <bp@alien8.de> Cc: Brian Cain <bcain@codeaurora.org> Cc: Catalin Marinas <catalin.marinas@arm.com> Cc: Chris Zankel <chris@zankel.net> Cc: "David S. Miller" <davem@davemloft.net> Cc: Geert Uytterhoeven <geert@linux-m68k.org> Cc: Greentime Hu <green.hu@gmail.com> Cc: Greg Ungerer <gerg@linux-m68k.org> Cc: Guan Xuetao <gxt@pku.edu.cn> Cc: Guo Ren <guoren@kernel.org> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Helge Deller <deller@gmx.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: Ley Foon Tan <ley.foon.tan@intel.com> Cc: Mark Salter <msalter@redhat.com> Cc: Matthew Wilcox <willy@infradead.org> Cc: Matt Turner <mattst88@gmail.com> Cc: Max Filippov <jcmvbkbc@gmail.com> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: Michal Simek <monstr@monstr.eu> Cc: Nick Hu <nickhu@andestech.com> Cc: Paul Walmsley <paul.walmsley@sifive.com> Cc: Richard Weinberger <richard@nod.at> Cc: Rich Felker <dalias@libc.org> Cc: Russell King <linux@armlinux.org.uk> Cc: Stafford Horne <shorne@gmail.com> Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Tony Luck <tony.luck@intel.com> Cc: Vincent Chen <deanbo422@gmail.com> Cc: Vineet Gupta <vgupta@synopsys.com> Cc: Will Deacon <will@kernel.org> Cc: Yoshinori Sato <ysato@users.sourceforge.jp> Link: http://lkml.kernel.org/r/20200514170327.31389-4-rppt@kernel.org Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-06-09 04:32:42 +00:00
#include <asm/tlbflush.h>
#include <linux/cpumask.h>
#include <linux/cpu.h>
#include <linux/vmalloc.h>
#include <linux/preempt.h>
#include <linux/spinlock.h>
#include <linux/shrinker.h>
#include <linux/types.h>
mm/zsmalloc: add statistics support Keeping fragmentation of zsmalloc in a low level is our target. But now we still need to add the debug code in zsmalloc to get the quantitative data. This patch adds a new configuration CONFIG_ZSMALLOC_STAT to enable the statistics collection for developers. Currently only the objects statatitics in each class are collected. User can get the information via debugfs. cat /sys/kernel/debug/zsmalloc/zram0/... For example: After I copied "jdk-8u25-linux-x64.tar.gz" to zram with ext4 filesystem: class size obj_allocated obj_used pages_used 0 32 0 0 0 1 48 256 12 3 2 64 64 14 1 3 80 51 7 1 4 96 128 5 3 5 112 73 5 2 6 128 32 4 1 7 144 0 0 0 8 160 0 0 0 9 176 0 0 0 10 192 0 0 0 11 208 0 0 0 12 224 0 0 0 13 240 0 0 0 14 256 16 1 1 15 272 15 9 1 16 288 0 0 0 17 304 0 0 0 18 320 0 0 0 19 336 0 0 0 20 352 0 0 0 21 368 0 0 0 22 384 0 0 0 23 400 0 0 0 24 416 0 0 0 25 432 0 0 0 26 448 0 0 0 27 464 0 0 0 28 480 0 0 0 29 496 33 1 4 30 512 0 0 0 31 528 0 0 0 32 544 0 0 0 33 560 0 0 0 34 576 0 0 0 35 592 0 0 0 36 608 0 0 0 37 624 0 0 0 38 640 0 0 0 40 672 0 0 0 42 704 0 0 0 43 720 17 1 3 44 736 0 0 0 46 768 0 0 0 49 816 0 0 0 51 848 0 0 0 52 864 14 1 3 54 896 0 0 0 57 944 13 1 3 58 960 0 0 0 62 1024 4 1 1 66 1088 15 2 4 67 1104 0 0 0 71 1168 0 0 0 74 1216 0 0 0 76 1248 0 0 0 83 1360 3 1 1 91 1488 11 1 4 94 1536 0 0 0 100 1632 5 1 2 107 1744 0 0 0 111 1808 9 1 4 126 2048 4 4 2 144 2336 7 3 4 151 2448 0 0 0 168 2720 15 15 10 190 3072 28 27 21 202 3264 0 0 0 254 4096 36209 36209 36209 Total 37022 36326 36288 We can calculate the overall fragentation by the last line: Total 37022 36326 36288 (37022 - 36326) / 37022 = 1.87% Also by analysing objects alocated in every class we know why we got so low fragmentation: Most of the allocated objects is in <class 254>. And there is only 1 page in class 254 zspage. So, No fragmentation will be introduced by allocating objs in class 254. And in future, we can collect other zsmalloc statistics as we need and analyse them. Signed-off-by: Ganesh Mahendran <opensource.ganesh@gmail.com> Suggested-by: Minchan Kim <minchan@kernel.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Dan Streetman <ddstreet@ieee.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-12 23:00:54 +00:00
#include <linux/debugfs.h>
zsmalloc: move it under mm This patch moves zsmalloc under mm directory. Before that, description will explain why we have needed custom allocator. Zsmalloc is a new slab-based memory allocator for storing compressed pages. It is designed for low fragmentation and high allocation success rate on large object, but <= PAGE_SIZE allocations. zsmalloc differs from the kernel slab allocator in two primary ways to achieve these design goals. zsmalloc never requires high order page allocations to back slabs, or "size classes" in zsmalloc terms. Instead it allows multiple single-order pages to be stitched together into a "zspage" which backs the slab. This allows for higher allocation success rate under memory pressure. Also, zsmalloc allows objects to span page boundaries within the zspage. This allows for lower fragmentation than could be had with the kernel slab allocator for objects between PAGE_SIZE/2 and PAGE_SIZE. With the kernel slab allocator, if a page compresses to 60% of it original size, the memory savings gained through compression is lost in fragmentation because another object of the same size can't be stored in the leftover space. This ability to span pages results in zsmalloc allocations not being directly addressable by the user. The user is given an non-dereferencable handle in response to an allocation request. That handle must be mapped, using zs_map_object(), which returns a pointer to the mapped region that can be used. The mapping is necessary since the object data may reside in two different noncontigious pages. The zsmalloc fulfills the allocation needs for zram perfectly [sjenning@linux.vnet.ibm.com: borrow Seth's quote] Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Nitin Gupta <ngupta@vflare.org> Reviewed-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Cc: Bob Liu <bob.liu@oracle.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: Hugh Dickins <hughd@google.com> Cc: Jens Axboe <axboe@kernel.dk> Cc: Luigi Semenzato <semenzato@google.com> Cc: Mel Gorman <mgorman@suse.de> Cc: Pekka Enberg <penberg@kernel.org> Cc: Rik van Riel <riel@redhat.com> Cc: Seth Jennings <sjenning@linux.vnet.ibm.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-01-30 23:45:50 +00:00
#include <linux/zsmalloc.h>
#include <linux/zpool.h>
mm: fix build warnings in <linux/compaction.h> Randy reported below build error. > In file included from ../include/linux/balloon_compaction.h:48:0, > from ../mm/balloon_compaction.c:11: > ../include/linux/compaction.h:237:51: warning: 'struct node' declared inside parameter list [enabled by default] > static inline int compaction_register_node(struct node *node) > ../include/linux/compaction.h:237:51: warning: its scope is only this definition or declaration, which is probably not what you want [enabled by default] > ../include/linux/compaction.h:242:54: warning: 'struct node' declared inside parameter list [enabled by default] > static inline void compaction_unregister_node(struct node *node) > It was caused by non-lru page migration which needs compaction.h but compaction.h doesn't include any header to be standalone. I think proper header for non-lru page migration is migrate.h rather than compaction.h because migrate.h has already headers needed to work non-lru page migration indirectly like isolate_mode_t, migrate_mode MIGRATEPAGE_SUCCESS. [akpm@linux-foundation.org: revert mm-balloon-use-general-non-lru-movable-page-feature-fix.patch temp fix] Link: http://lkml.kernel.org/r/20160610003304.GE29779@bbox Signed-off-by: Minchan Kim <minchan@kernel.org> Reported-by: Randy Dunlap <rdunlap@infradead.org> Cc: Konstantin Khlebnikov <koct9i@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Cc: Gioh Kim <gi-oh.kim@profitbricks.com> Cc: Rafael Aquini <aquini@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:26:50 +00:00
#include <linux/migrate.h>
mm/zsmalloc.c: fix race condition in zs_destroy_pool In zs_destroy_pool() we call flush_work(&pool->free_work). However, we have no guarantee that migration isn't happening in the background at that time. Since migration can't directly free pages, it relies on free_work being scheduled to free the pages. But there's nothing preventing an in-progress migrate from queuing the work *after* zs_unregister_migration() has called flush_work(). Which would mean pages still pointing at the inode when we free it. Since we know at destroy time all objects should be free, no new migrations can come in (since zs_page_isolate() fails for fully-free zspages). This means it is sufficient to track a "# isolated zspages" count by class, and have the destroy logic ensure all such pages have drained before proceeding. Keeping that state under the class spinlock keeps the logic straightforward. In this case a memory leak could lead to an eventual crash if compaction hits the leaked page. This crash would only occur if people are changing their zswap backend at runtime (which eventually starts destruction). Link: http://lkml.kernel.org/r/20190809181751.219326-2-henryburns@google.com Fixes: 48b4800a1c6a ("zsmalloc: page migration support") Signed-off-by: Henry Burns <henryburns@google.com> Reviewed-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Henry Burns <henrywolfeburns@gmail.com> Cc: Minchan Kim <minchan@kernel.org> Cc: Shakeel Butt <shakeelb@google.com> Cc: Jonathan Adams <jwadams@google.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2019-08-25 00:55:06 +00:00
#include <linux/wait.h>
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
#include <linux/pagemap.h>
#include <linux/fs.h>
#include <linux/local_lock.h>
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
#define ZSPAGE_MAGIC 0x58
/*
* This must be power of 2 and greater than or equal to sizeof(link_free).
* These two conditions ensure that any 'struct link_free' itself doesn't
* span more than 1 page which avoids complex case of mapping 2 pages simply
* to restore link_free pointer values.
*/
#define ZS_ALIGN 8
zsmalloc: decouple handle and object Recently, we started to use zram heavily and some of issues popped. 1) external fragmentation I got a report from Juneho Choi that fork failed although there are plenty of free pages in the system. His investigation revealed zram is one of the culprit to make heavy fragmentation so there was no more contiguous 16K page for pgd to fork in the ARM. 2) non-movable pages Other problem of zram now is that inherently, user want to use zram as swap in small memory system so they use zRAM with CMA to use memory efficiently. However, unfortunately, it doesn't work well because zRAM cannot use CMA's movable pages unless it doesn't support compaction. I got several reports about that OOM happened with zram although there are lots of swap space and free space in CMA area. 3) internal fragmentation zRAM has started support memory limitation feature to limit memory usage and I sent a patchset(https://lkml.org/lkml/2014/9/21/148) for VM to be harmonized with zram-swap to stop anonymous page reclaim if zram consumed memory up to the limit although there are free space on the swap. One problem for that direction is zram has no way to know any hole in memory space zsmalloc allocated by internal fragmentation so zram would regard swap is full although there are free space in zsmalloc. For solving the issue, zram want to trigger compaction of zsmalloc before it decides full or not. This patchset is first step to support above issues. For that, it adds indirect layer between handle and object location and supports manual compaction to solve 3th problem first of all. After this patchset got merged, next step is to make VM aware of zsmalloc compaction so that generic compaction will move zsmalloced-pages automatically in runtime. In my imaginary experiment(ie, high compress ratio data with heavy swap in/out on 8G zram-swap), data is as follows, Before = zram allocated object : 60212066 bytes zram total used: 140103680 bytes ratio: 42.98 percent MemFree: 840192 kB Compaction After = frag ratio after compaction zram allocated object : 60212066 bytes zram total used: 76185600 bytes ratio: 79.03 percent MemFree: 901932 kB Juneho reported below in his real platform with small aging. So, I think the benefit would be bigger in real aging system for a long time. - frag_ratio increased 3% (ie, higher is better) - memfree increased about 6MB - In buddy info, Normal 2^3: 4, 2^2: 1: 2^1 increased, Highmem: 2^1 21 increased frag ratio after swap fragment used : 156677 kbytes total: 166092 kbytes frag_ratio : 94 meminfo before compaction MemFree: 83724 kB Node 0, zone Normal 13642 1364 57 10 61 17 9 5 4 0 0 Node 0, zone HighMem 425 29 1 0 0 0 0 0 0 0 0 num_migrated : 23630 compaction done frag ratio after compaction used : 156673 kbytes total: 160564 kbytes frag_ratio : 97 meminfo after compaction MemFree: 89060 kB Node 0, zone Normal 14076 1544 67 14 61 17 9 5 4 0 0 Node 0, zone HighMem 863 50 1 0 0 0 0 0 0 0 0 This patchset adds more logics(about 480 lines) in zsmalloc but when I tested heavy swapin/out program, the regression for swapin/out speed is marginal because most of overheads were caused by compress/decompress and other MM reclaim stuff. This patch (of 7): Currently, handle of zsmalloc encodes object's location directly so it makes support of migration hard. This patch decouples handle and object via adding indirect layer. For that, it allocates handle dynamically and returns it to user. The handle is the address allocated by slab allocation so it's unique and we could keep object's location in the memory space allocated for handle. With it, we can change object's position without changing handle itself. Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:23 +00:00
#define ZS_HANDLE_SIZE (sizeof(unsigned long))
/*
* Object location (<PFN>, <obj_idx>) is encoded as
* a single (unsigned long) handle value.
*
* Note that object index <obj_idx> starts from 0.
*
* This is made more complicated by various memory models and PAE.
*/
#ifndef MAX_POSSIBLE_PHYSMEM_BITS
#ifdef MAX_PHYSMEM_BITS
#define MAX_POSSIBLE_PHYSMEM_BITS MAX_PHYSMEM_BITS
#else
/*
* If this definition of MAX_PHYSMEM_BITS is used, OBJ_INDEX_BITS will just
* be PAGE_SHIFT
*/
#define MAX_POSSIBLE_PHYSMEM_BITS BITS_PER_LONG
#endif
#endif
#define _PFN_BITS (MAX_POSSIBLE_PHYSMEM_BITS - PAGE_SHIFT)
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
/*
* Head in allocated object should have OBJ_ALLOCATED_TAG
* to identify the object was allocated or not.
* It's okay to add the status bit in the least bit because
* header keeps handle which is 4byte-aligned address so we
* have room for two bit at least.
*/
#define OBJ_ALLOCATED_TAG 1
zsmalloc: fix a race with deferred_handles storing Currently, there is a race between zs_free() and zs_reclaim_page(): zs_reclaim_page() finds a handle to an allocated object, but before the eviction happens, an independent zs_free() call to the same handle could come in and overwrite the object value stored at the handle with the last deferred handle. When zs_reclaim_page() finally gets to call the eviction handler, it will see an invalid object value (i.e the previous deferred handle instead of the original object value). This race happens quite infrequently. We only managed to produce it with out-of-tree developmental code that triggers zsmalloc writeback with a much higher frequency than usual. This patch fixes this race by storing the deferred handle in the object header instead. We differentiate the deferred handle from the other two cases (handle for allocated object, and linkage for free object) with a new tag. If zspage reclamation succeeds, we will free these deferred handles by walking through the zspage objects. On the other hand, if zspage reclamation fails, we reconstruct the zspage freelist (with the deferred handle tag and allocated tag) before trying again with the reclamation. [arnd@arndb.de: avoid unused-function warning] Link: https://lkml.kernel.org/r/20230117170507.2651972-1-arnd@kernel.org Link: https://lkml.kernel.org/r/20230110231701.326724-1-nphamcs@gmail.com Fixes: 9997bc017549 ("zsmalloc: implement writeback mechanism for zsmalloc") Signed-off-by: Nhat Pham <nphamcs@gmail.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-01-10 23:17:01 +00:00
#ifdef CONFIG_ZPOOL
/*
* The second least-significant bit in the object's header identifies if the
* value stored at the header is a deferred handle from the last reclaim
* attempt.
*
* As noted above, this is valid because we have room for two bits.
*/
#define OBJ_DEFERRED_HANDLE_TAG 2
#define OBJ_TAG_BITS 2
#define OBJ_TAG_MASK (OBJ_ALLOCATED_TAG | OBJ_DEFERRED_HANDLE_TAG)
#else
#define OBJ_TAG_BITS 1
#define OBJ_TAG_MASK OBJ_ALLOCATED_TAG
#endif /* CONFIG_ZPOOL */
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
#define OBJ_INDEX_BITS (BITS_PER_LONG - _PFN_BITS - OBJ_TAG_BITS)
#define OBJ_INDEX_MASK ((_AC(1, UL) << OBJ_INDEX_BITS) - 1)
#define HUGE_BITS 1
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
#define FULLNESS_BITS 4
#define CLASS_BITS 8
#define ISOLATED_BITS 5
#define MAGIC_VAL_BITS 8
#define MAX(a, b) ((a) >= (b) ? (a) : (b))
#define ZS_MAX_PAGES_PER_ZSPAGE (_AC(CONFIG_ZSMALLOC_CHAIN_SIZE, UL))
/* ZS_MIN_ALLOC_SIZE must be multiple of ZS_ALIGN */
#define ZS_MIN_ALLOC_SIZE \
MAX(32, (ZS_MAX_PAGES_PER_ZSPAGE << PAGE_SHIFT >> OBJ_INDEX_BITS))
zsmalloc: decouple handle and object Recently, we started to use zram heavily and some of issues popped. 1) external fragmentation I got a report from Juneho Choi that fork failed although there are plenty of free pages in the system. His investigation revealed zram is one of the culprit to make heavy fragmentation so there was no more contiguous 16K page for pgd to fork in the ARM. 2) non-movable pages Other problem of zram now is that inherently, user want to use zram as swap in small memory system so they use zRAM with CMA to use memory efficiently. However, unfortunately, it doesn't work well because zRAM cannot use CMA's movable pages unless it doesn't support compaction. I got several reports about that OOM happened with zram although there are lots of swap space and free space in CMA area. 3) internal fragmentation zRAM has started support memory limitation feature to limit memory usage and I sent a patchset(https://lkml.org/lkml/2014/9/21/148) for VM to be harmonized with zram-swap to stop anonymous page reclaim if zram consumed memory up to the limit although there are free space on the swap. One problem for that direction is zram has no way to know any hole in memory space zsmalloc allocated by internal fragmentation so zram would regard swap is full although there are free space in zsmalloc. For solving the issue, zram want to trigger compaction of zsmalloc before it decides full or not. This patchset is first step to support above issues. For that, it adds indirect layer between handle and object location and supports manual compaction to solve 3th problem first of all. After this patchset got merged, next step is to make VM aware of zsmalloc compaction so that generic compaction will move zsmalloced-pages automatically in runtime. In my imaginary experiment(ie, high compress ratio data with heavy swap in/out on 8G zram-swap), data is as follows, Before = zram allocated object : 60212066 bytes zram total used: 140103680 bytes ratio: 42.98 percent MemFree: 840192 kB Compaction After = frag ratio after compaction zram allocated object : 60212066 bytes zram total used: 76185600 bytes ratio: 79.03 percent MemFree: 901932 kB Juneho reported below in his real platform with small aging. So, I think the benefit would be bigger in real aging system for a long time. - frag_ratio increased 3% (ie, higher is better) - memfree increased about 6MB - In buddy info, Normal 2^3: 4, 2^2: 1: 2^1 increased, Highmem: 2^1 21 increased frag ratio after swap fragment used : 156677 kbytes total: 166092 kbytes frag_ratio : 94 meminfo before compaction MemFree: 83724 kB Node 0, zone Normal 13642 1364 57 10 61 17 9 5 4 0 0 Node 0, zone HighMem 425 29 1 0 0 0 0 0 0 0 0 num_migrated : 23630 compaction done frag ratio after compaction used : 156673 kbytes total: 160564 kbytes frag_ratio : 97 meminfo after compaction MemFree: 89060 kB Node 0, zone Normal 14076 1544 67 14 61 17 9 5 4 0 0 Node 0, zone HighMem 863 50 1 0 0 0 0 0 0 0 0 This patchset adds more logics(about 480 lines) in zsmalloc but when I tested heavy swapin/out program, the regression for swapin/out speed is marginal because most of overheads were caused by compress/decompress and other MM reclaim stuff. This patch (of 7): Currently, handle of zsmalloc encodes object's location directly so it makes support of migration hard. This patch decouples handle and object via adding indirect layer. For that, it allocates handle dynamically and returns it to user. The handle is the address allocated by slab allocation so it's unique and we could keep object's location in the memory space allocated for handle. With it, we can change object's position without changing handle itself. Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:23 +00:00
/* each chunk includes extra space to keep handle */
#define ZS_MAX_ALLOC_SIZE PAGE_SIZE
/*
* On systems with 4K page size, this gives 255 size classes! There is a
* trader-off here:
* - Large number of size classes is potentially wasteful as free page are
* spread across these classes
* - Small number of size classes causes large internal fragmentation
* - Probably its better to use specific size classes (empirically
* determined). NOTE: all those class sizes must be set as multiple of
* ZS_ALIGN to make sure link_free itself never has to span 2 pages.
*
* ZS_MIN_ALLOC_SIZE and ZS_SIZE_CLASS_DELTA must be multiple of ZS_ALIGN
* (reason above)
*/
#define ZS_SIZE_CLASS_DELTA (PAGE_SIZE >> CLASS_BITS)
#define ZS_SIZE_CLASSES (DIV_ROUND_UP(ZS_MAX_ALLOC_SIZE - ZS_MIN_ALLOC_SIZE, \
ZS_SIZE_CLASS_DELTA) + 1)
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
/*
* Pages are distinguished by the ratio of used memory (that is the ratio
* of ->inuse objects to all objects that page can store). For example,
* INUSE_RATIO_10 means that the ratio of used objects is > 0% and <= 10%.
*
* The number of fullness groups is not random. It allows us to keep
* difference between the least busy page in the group (minimum permitted
* number of ->inuse objects) and the most busy page (maximum permitted
* number of ->inuse objects) at a reasonable value.
*/
enum fullness_group {
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
ZS_INUSE_RATIO_0,
ZS_INUSE_RATIO_10,
/* NOTE: 8 more fullness groups here */
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
ZS_INUSE_RATIO_99 = 10,
ZS_INUSE_RATIO_100,
NR_FULLNESS_GROUPS,
};
enum class_stat_type {
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
/* NOTE: stats for 12 fullness groups here: from inuse 0 to 100 */
ZS_OBJS_ALLOCATED = NR_FULLNESS_GROUPS,
ZS_OBJS_INUSE,
NR_CLASS_STAT_TYPES,
mm/zsmalloc: add statistics support Keeping fragmentation of zsmalloc in a low level is our target. But now we still need to add the debug code in zsmalloc to get the quantitative data. This patch adds a new configuration CONFIG_ZSMALLOC_STAT to enable the statistics collection for developers. Currently only the objects statatitics in each class are collected. User can get the information via debugfs. cat /sys/kernel/debug/zsmalloc/zram0/... For example: After I copied "jdk-8u25-linux-x64.tar.gz" to zram with ext4 filesystem: class size obj_allocated obj_used pages_used 0 32 0 0 0 1 48 256 12 3 2 64 64 14 1 3 80 51 7 1 4 96 128 5 3 5 112 73 5 2 6 128 32 4 1 7 144 0 0 0 8 160 0 0 0 9 176 0 0 0 10 192 0 0 0 11 208 0 0 0 12 224 0 0 0 13 240 0 0 0 14 256 16 1 1 15 272 15 9 1 16 288 0 0 0 17 304 0 0 0 18 320 0 0 0 19 336 0 0 0 20 352 0 0 0 21 368 0 0 0 22 384 0 0 0 23 400 0 0 0 24 416 0 0 0 25 432 0 0 0 26 448 0 0 0 27 464 0 0 0 28 480 0 0 0 29 496 33 1 4 30 512 0 0 0 31 528 0 0 0 32 544 0 0 0 33 560 0 0 0 34 576 0 0 0 35 592 0 0 0 36 608 0 0 0 37 624 0 0 0 38 640 0 0 0 40 672 0 0 0 42 704 0 0 0 43 720 17 1 3 44 736 0 0 0 46 768 0 0 0 49 816 0 0 0 51 848 0 0 0 52 864 14 1 3 54 896 0 0 0 57 944 13 1 3 58 960 0 0 0 62 1024 4 1 1 66 1088 15 2 4 67 1104 0 0 0 71 1168 0 0 0 74 1216 0 0 0 76 1248 0 0 0 83 1360 3 1 1 91 1488 11 1 4 94 1536 0 0 0 100 1632 5 1 2 107 1744 0 0 0 111 1808 9 1 4 126 2048 4 4 2 144 2336 7 3 4 151 2448 0 0 0 168 2720 15 15 10 190 3072 28 27 21 202 3264 0 0 0 254 4096 36209 36209 36209 Total 37022 36326 36288 We can calculate the overall fragentation by the last line: Total 37022 36326 36288 (37022 - 36326) / 37022 = 1.87% Also by analysing objects alocated in every class we know why we got so low fragmentation: Most of the allocated objects is in <class 254>. And there is only 1 page in class 254 zspage. So, No fragmentation will be introduced by allocating objs in class 254. And in future, we can collect other zsmalloc statistics as we need and analyse them. Signed-off-by: Ganesh Mahendran <opensource.ganesh@gmail.com> Suggested-by: Minchan Kim <minchan@kernel.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Dan Streetman <ddstreet@ieee.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-12 23:00:54 +00:00
};
struct zs_size_stat {
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
unsigned long objs[NR_CLASS_STAT_TYPES];
mm/zsmalloc: add statistics support Keeping fragmentation of zsmalloc in a low level is our target. But now we still need to add the debug code in zsmalloc to get the quantitative data. This patch adds a new configuration CONFIG_ZSMALLOC_STAT to enable the statistics collection for developers. Currently only the objects statatitics in each class are collected. User can get the information via debugfs. cat /sys/kernel/debug/zsmalloc/zram0/... For example: After I copied "jdk-8u25-linux-x64.tar.gz" to zram with ext4 filesystem: class size obj_allocated obj_used pages_used 0 32 0 0 0 1 48 256 12 3 2 64 64 14 1 3 80 51 7 1 4 96 128 5 3 5 112 73 5 2 6 128 32 4 1 7 144 0 0 0 8 160 0 0 0 9 176 0 0 0 10 192 0 0 0 11 208 0 0 0 12 224 0 0 0 13 240 0 0 0 14 256 16 1 1 15 272 15 9 1 16 288 0 0 0 17 304 0 0 0 18 320 0 0 0 19 336 0 0 0 20 352 0 0 0 21 368 0 0 0 22 384 0 0 0 23 400 0 0 0 24 416 0 0 0 25 432 0 0 0 26 448 0 0 0 27 464 0 0 0 28 480 0 0 0 29 496 33 1 4 30 512 0 0 0 31 528 0 0 0 32 544 0 0 0 33 560 0 0 0 34 576 0 0 0 35 592 0 0 0 36 608 0 0 0 37 624 0 0 0 38 640 0 0 0 40 672 0 0 0 42 704 0 0 0 43 720 17 1 3 44 736 0 0 0 46 768 0 0 0 49 816 0 0 0 51 848 0 0 0 52 864 14 1 3 54 896 0 0 0 57 944 13 1 3 58 960 0 0 0 62 1024 4 1 1 66 1088 15 2 4 67 1104 0 0 0 71 1168 0 0 0 74 1216 0 0 0 76 1248 0 0 0 83 1360 3 1 1 91 1488 11 1 4 94 1536 0 0 0 100 1632 5 1 2 107 1744 0 0 0 111 1808 9 1 4 126 2048 4 4 2 144 2336 7 3 4 151 2448 0 0 0 168 2720 15 15 10 190 3072 28 27 21 202 3264 0 0 0 254 4096 36209 36209 36209 Total 37022 36326 36288 We can calculate the overall fragentation by the last line: Total 37022 36326 36288 (37022 - 36326) / 37022 = 1.87% Also by analysing objects alocated in every class we know why we got so low fragmentation: Most of the allocated objects is in <class 254>. And there is only 1 page in class 254 zspage. So, No fragmentation will be introduced by allocating objs in class 254. And in future, we can collect other zsmalloc statistics as we need and analyse them. Signed-off-by: Ganesh Mahendran <opensource.ganesh@gmail.com> Suggested-by: Minchan Kim <minchan@kernel.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Dan Streetman <ddstreet@ieee.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-12 23:00:54 +00:00
};
#ifdef CONFIG_ZSMALLOC_STAT
static struct dentry *zs_stat_root;
mm/zsmalloc: add statistics support Keeping fragmentation of zsmalloc in a low level is our target. But now we still need to add the debug code in zsmalloc to get the quantitative data. This patch adds a new configuration CONFIG_ZSMALLOC_STAT to enable the statistics collection for developers. Currently only the objects statatitics in each class are collected. User can get the information via debugfs. cat /sys/kernel/debug/zsmalloc/zram0/... For example: After I copied "jdk-8u25-linux-x64.tar.gz" to zram with ext4 filesystem: class size obj_allocated obj_used pages_used 0 32 0 0 0 1 48 256 12 3 2 64 64 14 1 3 80 51 7 1 4 96 128 5 3 5 112 73 5 2 6 128 32 4 1 7 144 0 0 0 8 160 0 0 0 9 176 0 0 0 10 192 0 0 0 11 208 0 0 0 12 224 0 0 0 13 240 0 0 0 14 256 16 1 1 15 272 15 9 1 16 288 0 0 0 17 304 0 0 0 18 320 0 0 0 19 336 0 0 0 20 352 0 0 0 21 368 0 0 0 22 384 0 0 0 23 400 0 0 0 24 416 0 0 0 25 432 0 0 0 26 448 0 0 0 27 464 0 0 0 28 480 0 0 0 29 496 33 1 4 30 512 0 0 0 31 528 0 0 0 32 544 0 0 0 33 560 0 0 0 34 576 0 0 0 35 592 0 0 0 36 608 0 0 0 37 624 0 0 0 38 640 0 0 0 40 672 0 0 0 42 704 0 0 0 43 720 17 1 3 44 736 0 0 0 46 768 0 0 0 49 816 0 0 0 51 848 0 0 0 52 864 14 1 3 54 896 0 0 0 57 944 13 1 3 58 960 0 0 0 62 1024 4 1 1 66 1088 15 2 4 67 1104 0 0 0 71 1168 0 0 0 74 1216 0 0 0 76 1248 0 0 0 83 1360 3 1 1 91 1488 11 1 4 94 1536 0 0 0 100 1632 5 1 2 107 1744 0 0 0 111 1808 9 1 4 126 2048 4 4 2 144 2336 7 3 4 151 2448 0 0 0 168 2720 15 15 10 190 3072 28 27 21 202 3264 0 0 0 254 4096 36209 36209 36209 Total 37022 36326 36288 We can calculate the overall fragentation by the last line: Total 37022 36326 36288 (37022 - 36326) / 37022 = 1.87% Also by analysing objects alocated in every class we know why we got so low fragmentation: Most of the allocated objects is in <class 254>. And there is only 1 page in class 254 zspage. So, No fragmentation will be introduced by allocating objs in class 254. And in future, we can collect other zsmalloc statistics as we need and analyse them. Signed-off-by: Ganesh Mahendran <opensource.ganesh@gmail.com> Suggested-by: Minchan Kim <minchan@kernel.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Dan Streetman <ddstreet@ieee.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-12 23:00:54 +00:00
#endif
zsmalloc: introduce zs_huge_class_size() Patch series "zsmalloc/zram: drop zram's max_zpage_size", v3. ZRAM's max_zpage_size is a bad thing. It forces zsmalloc to store normal objects as huge ones, which results in bigger zsmalloc memory usage. Drop it and use actual zsmalloc huge-class value when decide if the object is huge or not. This patch (of 2): Not every object can be share its zspage with other objects, e.g. when the object is as big as zspage or nearly as big a zspage. For such objects zsmalloc has a so called huge class - every object which belongs to huge class consumes the entire zspage (which consists of a physical page). On x86_64, PAGE_SHIFT 12 box, the first non-huge class size is 3264, so starting down from size 3264, objects can share page(-s) and thus minimize memory wastage. ZRAM, however, has its own statically defined watermark for huge objects, namely "3 * PAGE_SIZE / 4 = 3072", and forcibly stores every object larger than this watermark (3072) as a PAGE_SIZE object, in other words, to a huge class, while zsmalloc can keep some of those objects in non-huge classes. This results in increased memory consumption. zsmalloc knows better if the object is huge or not. Introduce zs_huge_class_size() function which tells if the given object can be stored in one of non-huge classes or not. This will let us to drop ZRAM's huge object watermark and fully rely on zsmalloc when we decide if the object is huge. [sergey.senozhatsky.work@gmail.com: add pool param to zs_huge_class_size()] Link: http://lkml.kernel.org/r/20180314081833.1096-2-sergey.senozhatsky@gmail.com Link: http://lkml.kernel.org/r/20180306070639.7389-2-sergey.senozhatsky@gmail.com Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Mike Rapoport <rppt@linux.vnet.ibm.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-04-05 23:24:43 +00:00
static size_t huge_class_size;
struct size_class {
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
struct list_head fullness_list[NR_FULLNESS_GROUPS];
/*
* Size of objects stored in this class. Must be multiple
* of ZS_ALIGN.
*/
int size;
int objs_per_zspage;
/* Number of PAGE_SIZE sized pages to combine to form a 'zspage' */
int pages_per_zspage;
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
unsigned int index;
struct zs_size_stat stats;
};
/*
* Placed within free objects to form a singly linked list.
* For every zspage, zspage->freeobj gives head of this list.
*
* This must be power of 2 and less than or equal to ZS_ALIGN
*/
struct link_free {
zsmalloc: decouple handle and object Recently, we started to use zram heavily and some of issues popped. 1) external fragmentation I got a report from Juneho Choi that fork failed although there are plenty of free pages in the system. His investigation revealed zram is one of the culprit to make heavy fragmentation so there was no more contiguous 16K page for pgd to fork in the ARM. 2) non-movable pages Other problem of zram now is that inherently, user want to use zram as swap in small memory system so they use zRAM with CMA to use memory efficiently. However, unfortunately, it doesn't work well because zRAM cannot use CMA's movable pages unless it doesn't support compaction. I got several reports about that OOM happened with zram although there are lots of swap space and free space in CMA area. 3) internal fragmentation zRAM has started support memory limitation feature to limit memory usage and I sent a patchset(https://lkml.org/lkml/2014/9/21/148) for VM to be harmonized with zram-swap to stop anonymous page reclaim if zram consumed memory up to the limit although there are free space on the swap. One problem for that direction is zram has no way to know any hole in memory space zsmalloc allocated by internal fragmentation so zram would regard swap is full although there are free space in zsmalloc. For solving the issue, zram want to trigger compaction of zsmalloc before it decides full or not. This patchset is first step to support above issues. For that, it adds indirect layer between handle and object location and supports manual compaction to solve 3th problem first of all. After this patchset got merged, next step is to make VM aware of zsmalloc compaction so that generic compaction will move zsmalloced-pages automatically in runtime. In my imaginary experiment(ie, high compress ratio data with heavy swap in/out on 8G zram-swap), data is as follows, Before = zram allocated object : 60212066 bytes zram total used: 140103680 bytes ratio: 42.98 percent MemFree: 840192 kB Compaction After = frag ratio after compaction zram allocated object : 60212066 bytes zram total used: 76185600 bytes ratio: 79.03 percent MemFree: 901932 kB Juneho reported below in his real platform with small aging. So, I think the benefit would be bigger in real aging system for a long time. - frag_ratio increased 3% (ie, higher is better) - memfree increased about 6MB - In buddy info, Normal 2^3: 4, 2^2: 1: 2^1 increased, Highmem: 2^1 21 increased frag ratio after swap fragment used : 156677 kbytes total: 166092 kbytes frag_ratio : 94 meminfo before compaction MemFree: 83724 kB Node 0, zone Normal 13642 1364 57 10 61 17 9 5 4 0 0 Node 0, zone HighMem 425 29 1 0 0 0 0 0 0 0 0 num_migrated : 23630 compaction done frag ratio after compaction used : 156673 kbytes total: 160564 kbytes frag_ratio : 97 meminfo after compaction MemFree: 89060 kB Node 0, zone Normal 14076 1544 67 14 61 17 9 5 4 0 0 Node 0, zone HighMem 863 50 1 0 0 0 0 0 0 0 0 This patchset adds more logics(about 480 lines) in zsmalloc but when I tested heavy swapin/out program, the regression for swapin/out speed is marginal because most of overheads were caused by compress/decompress and other MM reclaim stuff. This patch (of 7): Currently, handle of zsmalloc encodes object's location directly so it makes support of migration hard. This patch decouples handle and object via adding indirect layer. For that, it allocates handle dynamically and returns it to user. The handle is the address allocated by slab allocation so it's unique and we could keep object's location in the memory space allocated for handle. With it, we can change object's position without changing handle itself. Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:23 +00:00
union {
/*
* Free object index;
zsmalloc: decouple handle and object Recently, we started to use zram heavily and some of issues popped. 1) external fragmentation I got a report from Juneho Choi that fork failed although there are plenty of free pages in the system. His investigation revealed zram is one of the culprit to make heavy fragmentation so there was no more contiguous 16K page for pgd to fork in the ARM. 2) non-movable pages Other problem of zram now is that inherently, user want to use zram as swap in small memory system so they use zRAM with CMA to use memory efficiently. However, unfortunately, it doesn't work well because zRAM cannot use CMA's movable pages unless it doesn't support compaction. I got several reports about that OOM happened with zram although there are lots of swap space and free space in CMA area. 3) internal fragmentation zRAM has started support memory limitation feature to limit memory usage and I sent a patchset(https://lkml.org/lkml/2014/9/21/148) for VM to be harmonized with zram-swap to stop anonymous page reclaim if zram consumed memory up to the limit although there are free space on the swap. One problem for that direction is zram has no way to know any hole in memory space zsmalloc allocated by internal fragmentation so zram would regard swap is full although there are free space in zsmalloc. For solving the issue, zram want to trigger compaction of zsmalloc before it decides full or not. This patchset is first step to support above issues. For that, it adds indirect layer between handle and object location and supports manual compaction to solve 3th problem first of all. After this patchset got merged, next step is to make VM aware of zsmalloc compaction so that generic compaction will move zsmalloced-pages automatically in runtime. In my imaginary experiment(ie, high compress ratio data with heavy swap in/out on 8G zram-swap), data is as follows, Before = zram allocated object : 60212066 bytes zram total used: 140103680 bytes ratio: 42.98 percent MemFree: 840192 kB Compaction After = frag ratio after compaction zram allocated object : 60212066 bytes zram total used: 76185600 bytes ratio: 79.03 percent MemFree: 901932 kB Juneho reported below in his real platform with small aging. So, I think the benefit would be bigger in real aging system for a long time. - frag_ratio increased 3% (ie, higher is better) - memfree increased about 6MB - In buddy info, Normal 2^3: 4, 2^2: 1: 2^1 increased, Highmem: 2^1 21 increased frag ratio after swap fragment used : 156677 kbytes total: 166092 kbytes frag_ratio : 94 meminfo before compaction MemFree: 83724 kB Node 0, zone Normal 13642 1364 57 10 61 17 9 5 4 0 0 Node 0, zone HighMem 425 29 1 0 0 0 0 0 0 0 0 num_migrated : 23630 compaction done frag ratio after compaction used : 156673 kbytes total: 160564 kbytes frag_ratio : 97 meminfo after compaction MemFree: 89060 kB Node 0, zone Normal 14076 1544 67 14 61 17 9 5 4 0 0 Node 0, zone HighMem 863 50 1 0 0 0 0 0 0 0 0 This patchset adds more logics(about 480 lines) in zsmalloc but when I tested heavy swapin/out program, the regression for swapin/out speed is marginal because most of overheads were caused by compress/decompress and other MM reclaim stuff. This patch (of 7): Currently, handle of zsmalloc encodes object's location directly so it makes support of migration hard. This patch decouples handle and object via adding indirect layer. For that, it allocates handle dynamically and returns it to user. The handle is the address allocated by slab allocation so it's unique and we could keep object's location in the memory space allocated for handle. With it, we can change object's position without changing handle itself. Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:23 +00:00
* It's valid for non-allocated object
*/
unsigned long next;
zsmalloc: decouple handle and object Recently, we started to use zram heavily and some of issues popped. 1) external fragmentation I got a report from Juneho Choi that fork failed although there are plenty of free pages in the system. His investigation revealed zram is one of the culprit to make heavy fragmentation so there was no more contiguous 16K page for pgd to fork in the ARM. 2) non-movable pages Other problem of zram now is that inherently, user want to use zram as swap in small memory system so they use zRAM with CMA to use memory efficiently. However, unfortunately, it doesn't work well because zRAM cannot use CMA's movable pages unless it doesn't support compaction. I got several reports about that OOM happened with zram although there are lots of swap space and free space in CMA area. 3) internal fragmentation zRAM has started support memory limitation feature to limit memory usage and I sent a patchset(https://lkml.org/lkml/2014/9/21/148) for VM to be harmonized with zram-swap to stop anonymous page reclaim if zram consumed memory up to the limit although there are free space on the swap. One problem for that direction is zram has no way to know any hole in memory space zsmalloc allocated by internal fragmentation so zram would regard swap is full although there are free space in zsmalloc. For solving the issue, zram want to trigger compaction of zsmalloc before it decides full or not. This patchset is first step to support above issues. For that, it adds indirect layer between handle and object location and supports manual compaction to solve 3th problem first of all. After this patchset got merged, next step is to make VM aware of zsmalloc compaction so that generic compaction will move zsmalloced-pages automatically in runtime. In my imaginary experiment(ie, high compress ratio data with heavy swap in/out on 8G zram-swap), data is as follows, Before = zram allocated object : 60212066 bytes zram total used: 140103680 bytes ratio: 42.98 percent MemFree: 840192 kB Compaction After = frag ratio after compaction zram allocated object : 60212066 bytes zram total used: 76185600 bytes ratio: 79.03 percent MemFree: 901932 kB Juneho reported below in his real platform with small aging. So, I think the benefit would be bigger in real aging system for a long time. - frag_ratio increased 3% (ie, higher is better) - memfree increased about 6MB - In buddy info, Normal 2^3: 4, 2^2: 1: 2^1 increased, Highmem: 2^1 21 increased frag ratio after swap fragment used : 156677 kbytes total: 166092 kbytes frag_ratio : 94 meminfo before compaction MemFree: 83724 kB Node 0, zone Normal 13642 1364 57 10 61 17 9 5 4 0 0 Node 0, zone HighMem 425 29 1 0 0 0 0 0 0 0 0 num_migrated : 23630 compaction done frag ratio after compaction used : 156673 kbytes total: 160564 kbytes frag_ratio : 97 meminfo after compaction MemFree: 89060 kB Node 0, zone Normal 14076 1544 67 14 61 17 9 5 4 0 0 Node 0, zone HighMem 863 50 1 0 0 0 0 0 0 0 0 This patchset adds more logics(about 480 lines) in zsmalloc but when I tested heavy swapin/out program, the regression for swapin/out speed is marginal because most of overheads were caused by compress/decompress and other MM reclaim stuff. This patch (of 7): Currently, handle of zsmalloc encodes object's location directly so it makes support of migration hard. This patch decouples handle and object via adding indirect layer. For that, it allocates handle dynamically and returns it to user. The handle is the address allocated by slab allocation so it's unique and we could keep object's location in the memory space allocated for handle. With it, we can change object's position without changing handle itself. Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:23 +00:00
/*
* Handle of allocated object.
*/
unsigned long handle;
zsmalloc: fix a race with deferred_handles storing Currently, there is a race between zs_free() and zs_reclaim_page(): zs_reclaim_page() finds a handle to an allocated object, but before the eviction happens, an independent zs_free() call to the same handle could come in and overwrite the object value stored at the handle with the last deferred handle. When zs_reclaim_page() finally gets to call the eviction handler, it will see an invalid object value (i.e the previous deferred handle instead of the original object value). This race happens quite infrequently. We only managed to produce it with out-of-tree developmental code that triggers zsmalloc writeback with a much higher frequency than usual. This patch fixes this race by storing the deferred handle in the object header instead. We differentiate the deferred handle from the other two cases (handle for allocated object, and linkage for free object) with a new tag. If zspage reclamation succeeds, we will free these deferred handles by walking through the zspage objects. On the other hand, if zspage reclamation fails, we reconstruct the zspage freelist (with the deferred handle tag and allocated tag) before trying again with the reclamation. [arnd@arndb.de: avoid unused-function warning] Link: https://lkml.kernel.org/r/20230117170507.2651972-1-arnd@kernel.org Link: https://lkml.kernel.org/r/20230110231701.326724-1-nphamcs@gmail.com Fixes: 9997bc017549 ("zsmalloc: implement writeback mechanism for zsmalloc") Signed-off-by: Nhat Pham <nphamcs@gmail.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-01-10 23:17:01 +00:00
#ifdef CONFIG_ZPOOL
/*
* Deferred handle of a reclaimed object.
*/
unsigned long deferred_handle;
#endif
zsmalloc: decouple handle and object Recently, we started to use zram heavily and some of issues popped. 1) external fragmentation I got a report from Juneho Choi that fork failed although there are plenty of free pages in the system. His investigation revealed zram is one of the culprit to make heavy fragmentation so there was no more contiguous 16K page for pgd to fork in the ARM. 2) non-movable pages Other problem of zram now is that inherently, user want to use zram as swap in small memory system so they use zRAM with CMA to use memory efficiently. However, unfortunately, it doesn't work well because zRAM cannot use CMA's movable pages unless it doesn't support compaction. I got several reports about that OOM happened with zram although there are lots of swap space and free space in CMA area. 3) internal fragmentation zRAM has started support memory limitation feature to limit memory usage and I sent a patchset(https://lkml.org/lkml/2014/9/21/148) for VM to be harmonized with zram-swap to stop anonymous page reclaim if zram consumed memory up to the limit although there are free space on the swap. One problem for that direction is zram has no way to know any hole in memory space zsmalloc allocated by internal fragmentation so zram would regard swap is full although there are free space in zsmalloc. For solving the issue, zram want to trigger compaction of zsmalloc before it decides full or not. This patchset is first step to support above issues. For that, it adds indirect layer between handle and object location and supports manual compaction to solve 3th problem first of all. After this patchset got merged, next step is to make VM aware of zsmalloc compaction so that generic compaction will move zsmalloced-pages automatically in runtime. In my imaginary experiment(ie, high compress ratio data with heavy swap in/out on 8G zram-swap), data is as follows, Before = zram allocated object : 60212066 bytes zram total used: 140103680 bytes ratio: 42.98 percent MemFree: 840192 kB Compaction After = frag ratio after compaction zram allocated object : 60212066 bytes zram total used: 76185600 bytes ratio: 79.03 percent MemFree: 901932 kB Juneho reported below in his real platform with small aging. So, I think the benefit would be bigger in real aging system for a long time. - frag_ratio increased 3% (ie, higher is better) - memfree increased about 6MB - In buddy info, Normal 2^3: 4, 2^2: 1: 2^1 increased, Highmem: 2^1 21 increased frag ratio after swap fragment used : 156677 kbytes total: 166092 kbytes frag_ratio : 94 meminfo before compaction MemFree: 83724 kB Node 0, zone Normal 13642 1364 57 10 61 17 9 5 4 0 0 Node 0, zone HighMem 425 29 1 0 0 0 0 0 0 0 0 num_migrated : 23630 compaction done frag ratio after compaction used : 156673 kbytes total: 160564 kbytes frag_ratio : 97 meminfo after compaction MemFree: 89060 kB Node 0, zone Normal 14076 1544 67 14 61 17 9 5 4 0 0 Node 0, zone HighMem 863 50 1 0 0 0 0 0 0 0 0 This patchset adds more logics(about 480 lines) in zsmalloc but when I tested heavy swapin/out program, the regression for swapin/out speed is marginal because most of overheads were caused by compress/decompress and other MM reclaim stuff. This patch (of 7): Currently, handle of zsmalloc encodes object's location directly so it makes support of migration hard. This patch decouples handle and object via adding indirect layer. For that, it allocates handle dynamically and returns it to user. The handle is the address allocated by slab allocation so it's unique and we could keep object's location in the memory space allocated for handle. With it, we can change object's position without changing handle itself. Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:23 +00:00
};
};
struct zs_pool {
const char *name;
mm/zsmalloc: add statistics support Keeping fragmentation of zsmalloc in a low level is our target. But now we still need to add the debug code in zsmalloc to get the quantitative data. This patch adds a new configuration CONFIG_ZSMALLOC_STAT to enable the statistics collection for developers. Currently only the objects statatitics in each class are collected. User can get the information via debugfs. cat /sys/kernel/debug/zsmalloc/zram0/... For example: After I copied "jdk-8u25-linux-x64.tar.gz" to zram with ext4 filesystem: class size obj_allocated obj_used pages_used 0 32 0 0 0 1 48 256 12 3 2 64 64 14 1 3 80 51 7 1 4 96 128 5 3 5 112 73 5 2 6 128 32 4 1 7 144 0 0 0 8 160 0 0 0 9 176 0 0 0 10 192 0 0 0 11 208 0 0 0 12 224 0 0 0 13 240 0 0 0 14 256 16 1 1 15 272 15 9 1 16 288 0 0 0 17 304 0 0 0 18 320 0 0 0 19 336 0 0 0 20 352 0 0 0 21 368 0 0 0 22 384 0 0 0 23 400 0 0 0 24 416 0 0 0 25 432 0 0 0 26 448 0 0 0 27 464 0 0 0 28 480 0 0 0 29 496 33 1 4 30 512 0 0 0 31 528 0 0 0 32 544 0 0 0 33 560 0 0 0 34 576 0 0 0 35 592 0 0 0 36 608 0 0 0 37 624 0 0 0 38 640 0 0 0 40 672 0 0 0 42 704 0 0 0 43 720 17 1 3 44 736 0 0 0 46 768 0 0 0 49 816 0 0 0 51 848 0 0 0 52 864 14 1 3 54 896 0 0 0 57 944 13 1 3 58 960 0 0 0 62 1024 4 1 1 66 1088 15 2 4 67 1104 0 0 0 71 1168 0 0 0 74 1216 0 0 0 76 1248 0 0 0 83 1360 3 1 1 91 1488 11 1 4 94 1536 0 0 0 100 1632 5 1 2 107 1744 0 0 0 111 1808 9 1 4 126 2048 4 4 2 144 2336 7 3 4 151 2448 0 0 0 168 2720 15 15 10 190 3072 28 27 21 202 3264 0 0 0 254 4096 36209 36209 36209 Total 37022 36326 36288 We can calculate the overall fragentation by the last line: Total 37022 36326 36288 (37022 - 36326) / 37022 = 1.87% Also by analysing objects alocated in every class we know why we got so low fragmentation: Most of the allocated objects is in <class 254>. And there is only 1 page in class 254 zspage. So, No fragmentation will be introduced by allocating objs in class 254. And in future, we can collect other zsmalloc statistics as we need and analyse them. Signed-off-by: Ganesh Mahendran <opensource.ganesh@gmail.com> Suggested-by: Minchan Kim <minchan@kernel.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Dan Streetman <ddstreet@ieee.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-12 23:00:54 +00:00
struct size_class *size_class[ZS_SIZE_CLASSES];
zsmalloc: decouple handle and object Recently, we started to use zram heavily and some of issues popped. 1) external fragmentation I got a report from Juneho Choi that fork failed although there are plenty of free pages in the system. His investigation revealed zram is one of the culprit to make heavy fragmentation so there was no more contiguous 16K page for pgd to fork in the ARM. 2) non-movable pages Other problem of zram now is that inherently, user want to use zram as swap in small memory system so they use zRAM with CMA to use memory efficiently. However, unfortunately, it doesn't work well because zRAM cannot use CMA's movable pages unless it doesn't support compaction. I got several reports about that OOM happened with zram although there are lots of swap space and free space in CMA area. 3) internal fragmentation zRAM has started support memory limitation feature to limit memory usage and I sent a patchset(https://lkml.org/lkml/2014/9/21/148) for VM to be harmonized with zram-swap to stop anonymous page reclaim if zram consumed memory up to the limit although there are free space on the swap. One problem for that direction is zram has no way to know any hole in memory space zsmalloc allocated by internal fragmentation so zram would regard swap is full although there are free space in zsmalloc. For solving the issue, zram want to trigger compaction of zsmalloc before it decides full or not. This patchset is first step to support above issues. For that, it adds indirect layer between handle and object location and supports manual compaction to solve 3th problem first of all. After this patchset got merged, next step is to make VM aware of zsmalloc compaction so that generic compaction will move zsmalloced-pages automatically in runtime. In my imaginary experiment(ie, high compress ratio data with heavy swap in/out on 8G zram-swap), data is as follows, Before = zram allocated object : 60212066 bytes zram total used: 140103680 bytes ratio: 42.98 percent MemFree: 840192 kB Compaction After = frag ratio after compaction zram allocated object : 60212066 bytes zram total used: 76185600 bytes ratio: 79.03 percent MemFree: 901932 kB Juneho reported below in his real platform with small aging. So, I think the benefit would be bigger in real aging system for a long time. - frag_ratio increased 3% (ie, higher is better) - memfree increased about 6MB - In buddy info, Normal 2^3: 4, 2^2: 1: 2^1 increased, Highmem: 2^1 21 increased frag ratio after swap fragment used : 156677 kbytes total: 166092 kbytes frag_ratio : 94 meminfo before compaction MemFree: 83724 kB Node 0, zone Normal 13642 1364 57 10 61 17 9 5 4 0 0 Node 0, zone HighMem 425 29 1 0 0 0 0 0 0 0 0 num_migrated : 23630 compaction done frag ratio after compaction used : 156673 kbytes total: 160564 kbytes frag_ratio : 97 meminfo after compaction MemFree: 89060 kB Node 0, zone Normal 14076 1544 67 14 61 17 9 5 4 0 0 Node 0, zone HighMem 863 50 1 0 0 0 0 0 0 0 0 This patchset adds more logics(about 480 lines) in zsmalloc but when I tested heavy swapin/out program, the regression for swapin/out speed is marginal because most of overheads were caused by compress/decompress and other MM reclaim stuff. This patch (of 7): Currently, handle of zsmalloc encodes object's location directly so it makes support of migration hard. This patch decouples handle and object via adding indirect layer. For that, it allocates handle dynamically and returns it to user. The handle is the address allocated by slab allocation so it's unique and we could keep object's location in the memory space allocated for handle. With it, we can change object's position without changing handle itself. Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:23 +00:00
struct kmem_cache *handle_cachep;
struct kmem_cache *zspage_cachep;
zsmalloc: move pages_allocated to zs_pool Currently, zram has no feature to limit memory so theoretically zram can deplete system memory. Users have asked for a limit several times as even without exhaustion zram makes it hard to control memory usage of the platform. This patchset adds the feature. Patch 1 makes zs_get_total_size_bytes faster because it would be used frequently in later patches for the new feature. Patch 2 changes zs_get_total_size_bytes's return unit from bytes to page so that zsmalloc doesn't need unnecessary operation(ie, << PAGE_SHIFT). Patch 3 adds new feature. I added the feature into zram layer, not zsmalloc because limiation is zram's requirement, not zsmalloc so any other user using zsmalloc(ie, zpool) shouldn't affected by unnecessary branch of zsmalloc. In future, if every users of zsmalloc want the feature, then, we could move the feature from client side to zsmalloc easily but vice versa would be painful. Patch 4 adds news facility to report maximum memory usage of zram so that this avoids user polling frequently via /sys/block/zram0/ mem_used_total and ensures transient max are not missed. This patch (of 4): pages_allocated has counted in size_class structure and when user of zsmalloc want to see total_size_bytes, it should gather all of count from each size_class to report the sum. It's not bad if user don't see the value often but if user start to see the value frequently, it would be not a good deal for performance pov. This patch moves the count from size_class to zs_pool so it could reduce memory footprint (from [255 * 8byte] to [sizeof(atomic_long_t)]). Signed-off-by: Minchan Kim <minchan@kernel.org> Reviewed-by: Dan Streetman <ddstreet@ieee.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: <juno.choi@lge.com> Cc: <seungho1.park@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjennings@variantweb.net> Reviewed-by: David Horner <ds2horner@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-10-09 22:29:48 +00:00
atomic_long_t pages_allocated;
mm/zsmalloc: add statistics support Keeping fragmentation of zsmalloc in a low level is our target. But now we still need to add the debug code in zsmalloc to get the quantitative data. This patch adds a new configuration CONFIG_ZSMALLOC_STAT to enable the statistics collection for developers. Currently only the objects statatitics in each class are collected. User can get the information via debugfs. cat /sys/kernel/debug/zsmalloc/zram0/... For example: After I copied "jdk-8u25-linux-x64.tar.gz" to zram with ext4 filesystem: class size obj_allocated obj_used pages_used 0 32 0 0 0 1 48 256 12 3 2 64 64 14 1 3 80 51 7 1 4 96 128 5 3 5 112 73 5 2 6 128 32 4 1 7 144 0 0 0 8 160 0 0 0 9 176 0 0 0 10 192 0 0 0 11 208 0 0 0 12 224 0 0 0 13 240 0 0 0 14 256 16 1 1 15 272 15 9 1 16 288 0 0 0 17 304 0 0 0 18 320 0 0 0 19 336 0 0 0 20 352 0 0 0 21 368 0 0 0 22 384 0 0 0 23 400 0 0 0 24 416 0 0 0 25 432 0 0 0 26 448 0 0 0 27 464 0 0 0 28 480 0 0 0 29 496 33 1 4 30 512 0 0 0 31 528 0 0 0 32 544 0 0 0 33 560 0 0 0 34 576 0 0 0 35 592 0 0 0 36 608 0 0 0 37 624 0 0 0 38 640 0 0 0 40 672 0 0 0 42 704 0 0 0 43 720 17 1 3 44 736 0 0 0 46 768 0 0 0 49 816 0 0 0 51 848 0 0 0 52 864 14 1 3 54 896 0 0 0 57 944 13 1 3 58 960 0 0 0 62 1024 4 1 1 66 1088 15 2 4 67 1104 0 0 0 71 1168 0 0 0 74 1216 0 0 0 76 1248 0 0 0 83 1360 3 1 1 91 1488 11 1 4 94 1536 0 0 0 100 1632 5 1 2 107 1744 0 0 0 111 1808 9 1 4 126 2048 4 4 2 144 2336 7 3 4 151 2448 0 0 0 168 2720 15 15 10 190 3072 28 27 21 202 3264 0 0 0 254 4096 36209 36209 36209 Total 37022 36326 36288 We can calculate the overall fragentation by the last line: Total 37022 36326 36288 (37022 - 36326) / 37022 = 1.87% Also by analysing objects alocated in every class we know why we got so low fragmentation: Most of the allocated objects is in <class 254>. And there is only 1 page in class 254 zspage. So, No fragmentation will be introduced by allocating objs in class 254. And in future, we can collect other zsmalloc statistics as we need and analyse them. Signed-off-by: Ganesh Mahendran <opensource.ganesh@gmail.com> Suggested-by: Minchan Kim <minchan@kernel.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Dan Streetman <ddstreet@ieee.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-12 23:00:54 +00:00
struct zs_pool_stats stats;
/* Compact classes */
struct shrinker shrinker;
#ifdef CONFIG_ZPOOL
/* List tracking the zspages in LRU order by most recently added object */
struct list_head lru;
struct zpool *zpool;
const struct zpool_ops *zpool_ops;
#endif
mm/zsmalloc: add statistics support Keeping fragmentation of zsmalloc in a low level is our target. But now we still need to add the debug code in zsmalloc to get the quantitative data. This patch adds a new configuration CONFIG_ZSMALLOC_STAT to enable the statistics collection for developers. Currently only the objects statatitics in each class are collected. User can get the information via debugfs. cat /sys/kernel/debug/zsmalloc/zram0/... For example: After I copied "jdk-8u25-linux-x64.tar.gz" to zram with ext4 filesystem: class size obj_allocated obj_used pages_used 0 32 0 0 0 1 48 256 12 3 2 64 64 14 1 3 80 51 7 1 4 96 128 5 3 5 112 73 5 2 6 128 32 4 1 7 144 0 0 0 8 160 0 0 0 9 176 0 0 0 10 192 0 0 0 11 208 0 0 0 12 224 0 0 0 13 240 0 0 0 14 256 16 1 1 15 272 15 9 1 16 288 0 0 0 17 304 0 0 0 18 320 0 0 0 19 336 0 0 0 20 352 0 0 0 21 368 0 0 0 22 384 0 0 0 23 400 0 0 0 24 416 0 0 0 25 432 0 0 0 26 448 0 0 0 27 464 0 0 0 28 480 0 0 0 29 496 33 1 4 30 512 0 0 0 31 528 0 0 0 32 544 0 0 0 33 560 0 0 0 34 576 0 0 0 35 592 0 0 0 36 608 0 0 0 37 624 0 0 0 38 640 0 0 0 40 672 0 0 0 42 704 0 0 0 43 720 17 1 3 44 736 0 0 0 46 768 0 0 0 49 816 0 0 0 51 848 0 0 0 52 864 14 1 3 54 896 0 0 0 57 944 13 1 3 58 960 0 0 0 62 1024 4 1 1 66 1088 15 2 4 67 1104 0 0 0 71 1168 0 0 0 74 1216 0 0 0 76 1248 0 0 0 83 1360 3 1 1 91 1488 11 1 4 94 1536 0 0 0 100 1632 5 1 2 107 1744 0 0 0 111 1808 9 1 4 126 2048 4 4 2 144 2336 7 3 4 151 2448 0 0 0 168 2720 15 15 10 190 3072 28 27 21 202 3264 0 0 0 254 4096 36209 36209 36209 Total 37022 36326 36288 We can calculate the overall fragentation by the last line: Total 37022 36326 36288 (37022 - 36326) / 37022 = 1.87% Also by analysing objects alocated in every class we know why we got so low fragmentation: Most of the allocated objects is in <class 254>. And there is only 1 page in class 254 zspage. So, No fragmentation will be introduced by allocating objs in class 254. And in future, we can collect other zsmalloc statistics as we need and analyse them. Signed-off-by: Ganesh Mahendran <opensource.ganesh@gmail.com> Suggested-by: Minchan Kim <minchan@kernel.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Dan Streetman <ddstreet@ieee.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-12 23:00:54 +00:00
#ifdef CONFIG_ZSMALLOC_STAT
struct dentry *stat_dentry;
#endif
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
#ifdef CONFIG_COMPACTION
struct work_struct free_work;
#endif
zsmalloc: consolidate zs_pool's migrate_lock and size_class's locks Currently, zsmalloc has a hierarchy of locks, which includes a pool-level migrate_lock, and a lock for each size class. We have to obtain both locks in the hotpath in most cases anyway, except for zs_malloc. This exception will no longer exist when we introduce a LRU into the zs_pool for the new writeback functionality - we will need to obtain a pool-level lock to synchronize LRU handling even in zs_malloc. In preparation for zsmalloc writeback, consolidate these locks into a single pool-level lock, which drastically reduces the complexity of synchronization in zsmalloc. We have also benchmarked the lock consolidation to see the performance effect of this change on zram. First, we ran a synthetic FS workload on a server machine with 36 cores (same machine for all runs), using fs_mark -d ../zram1mnt -s 100000 -n 2500 -t 32 -k before and after for btrfs and ext4 on zram (FS usage is 80%). Here is the result (unit is file/second): With lock consolidation (btrfs): Average: 13520.2, Median: 13531.0, Stddev: 137.5961482019028 Without lock consolidation (btrfs): Average: 13487.2, Median: 13575.0, Stddev: 309.08283679298665 With lock consolidation (ext4): Average: 16824.4, Median: 16839.0, Stddev: 89.97388510006668 Without lock consolidation (ext4) Average: 16958.0, Median: 16986.0, Stddev: 194.7370021336469 As you can see, we observe a 0.3% regression for btrfs, and a 0.9% regression for ext4. This is a small, barely measurable difference in my opinion. For a more realistic scenario, we also tries building the kernel on zram. Here is the time it takes (in seconds): With lock consolidation (btrfs): real Average: 319.6, Median: 320.0, Stddev: 0.8944271909999159 user Average: 6894.2, Median: 6895.0, Stddev: 25.528415540334656 sys Average: 521.4, Median: 522.0, Stddev: 1.51657508881031 Without lock consolidation (btrfs): real Average: 319.8, Median: 320.0, Stddev: 0.8366600265340756 user Average: 6896.6, Median: 6899.0, Stddev: 16.04057355583023 sys Average: 520.6, Median: 521.0, Stddev: 1.140175425099138 With lock consolidation (ext4): real Average: 320.0, Median: 319.0, Stddev: 1.4142135623730951 user Average: 6896.8, Median: 6878.0, Stddev: 28.621670111997307 sys Average: 521.2, Median: 521.0, Stddev: 1.7888543819998317 Without lock consolidation (ext4) real Average: 319.6, Median: 319.0, Stddev: 0.8944271909999159 user Average: 6886.2, Median: 6887.0, Stddev: 16.93221781102523 sys Average: 520.4, Median: 520.0, Stddev: 1.140175425099138 The difference is entirely within the noise of a typical run on zram. This hardly justifies the complexity of maintaining both the pool lock and the class lock. In fact, for writeback, we would need to introduce yet another lock to prevent data races on the pool's LRU, further complicating the lock handling logic. IMHO, it is just better to collapse all of these into a single pool-level lock. Link: https://lkml.kernel.org/r/20221128191616.1261026-4-nphamcs@gmail.com Signed-off-by: Nhat Pham <nphamcs@gmail.com> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Acked-by: Minchan Kim <minchan@kernel.org> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-11-28 19:16:12 +00:00
spinlock_t lock;
atomic_t compaction_in_progress;
};
struct zspage {
struct {
unsigned int huge:HUGE_BITS;
unsigned int fullness:FULLNESS_BITS;
unsigned int class:CLASS_BITS + 1;
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
unsigned int isolated:ISOLATED_BITS;
unsigned int magic:MAGIC_VAL_BITS;
};
unsigned int inuse;
unsigned int freeobj;
struct page *first_page;
struct list_head list; /* fullness list */
#ifdef CONFIG_ZPOOL
/* links the zspage to the lru list in the pool */
struct list_head lru;
bool under_reclaim;
#endif
struct zs_pool *pool;
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
rwlock_t lock;
};
struct mapping_area {
local_lock_t lock;
char *vm_buf; /* copy buffer for objects that span pages */
char *vm_addr; /* address of kmap_atomic()'ed pages */
enum zs_mapmode vm_mm; /* mapping mode */
};
/* huge object: pages_per_zspage == 1 && maxobj_per_zspage == 1 */
static void SetZsHugePage(struct zspage *zspage)
{
zspage->huge = 1;
}
static bool ZsHugePage(struct zspage *zspage)
{
return zspage->huge;
}
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
static void migrate_lock_init(struct zspage *zspage);
static void migrate_read_lock(struct zspage *zspage);
static void migrate_read_unlock(struct zspage *zspage);
#ifdef CONFIG_COMPACTION
zsmalloc: replace per zpage lock with pool->migrate_lock The zsmalloc has used a bit for spin_lock in zpage handle to keep zpage object alive during several operations. However, it causes the problem for PREEMPT_RT as well as introducing too complicated. This patch replaces the bit spin_lock with pool->migrate_lock rwlock. It could make the code simple as well as zsmalloc work under PREEMPT_RT. The drawback is the pool->migrate_lock is bigger granuarity than per zpage lock so the contention would be higher than old when both IO-related operations(i.e., zsmalloc, zsfree, zs_[map|unmap]) and compaction(page/zpage migration) are going in parallel(*, the migrate_lock is rwlock and IO related functions are all read side lock so there is no contention). However, the write-side is fast enough(dominant overhead is just page copy) so it wouldn't affect much. If the lock granurity becomes more problem later, we could introduce table locks based on handle as a hash value. Link: https://lkml.kernel.org/r/20211115185909.3949505-9-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:14:13 +00:00
static void migrate_write_lock(struct zspage *zspage);
static void migrate_write_lock_nested(struct zspage *zspage);
static void migrate_write_unlock(struct zspage *zspage);
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
static void kick_deferred_free(struct zs_pool *pool);
static void init_deferred_free(struct zs_pool *pool);
static void SetZsPageMovable(struct zs_pool *pool, struct zspage *zspage);
#else
zsmalloc: replace per zpage lock with pool->migrate_lock The zsmalloc has used a bit for spin_lock in zpage handle to keep zpage object alive during several operations. However, it causes the problem for PREEMPT_RT as well as introducing too complicated. This patch replaces the bit spin_lock with pool->migrate_lock rwlock. It could make the code simple as well as zsmalloc work under PREEMPT_RT. The drawback is the pool->migrate_lock is bigger granuarity than per zpage lock so the contention would be higher than old when both IO-related operations(i.e., zsmalloc, zsfree, zs_[map|unmap]) and compaction(page/zpage migration) are going in parallel(*, the migrate_lock is rwlock and IO related functions are all read side lock so there is no contention). However, the write-side is fast enough(dominant overhead is just page copy) so it wouldn't affect much. If the lock granurity becomes more problem later, we could introduce table locks based on handle as a hash value. Link: https://lkml.kernel.org/r/20211115185909.3949505-9-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:14:13 +00:00
static void migrate_write_lock(struct zspage *zspage) {}
static void migrate_write_lock_nested(struct zspage *zspage) {}
static void migrate_write_unlock(struct zspage *zspage) {}
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
static void kick_deferred_free(struct zs_pool *pool) {}
static void init_deferred_free(struct zs_pool *pool) {}
static void SetZsPageMovable(struct zs_pool *pool, struct zspage *zspage) {}
#endif
static int create_cache(struct zs_pool *pool)
zsmalloc: decouple handle and object Recently, we started to use zram heavily and some of issues popped. 1) external fragmentation I got a report from Juneho Choi that fork failed although there are plenty of free pages in the system. His investigation revealed zram is one of the culprit to make heavy fragmentation so there was no more contiguous 16K page for pgd to fork in the ARM. 2) non-movable pages Other problem of zram now is that inherently, user want to use zram as swap in small memory system so they use zRAM with CMA to use memory efficiently. However, unfortunately, it doesn't work well because zRAM cannot use CMA's movable pages unless it doesn't support compaction. I got several reports about that OOM happened with zram although there are lots of swap space and free space in CMA area. 3) internal fragmentation zRAM has started support memory limitation feature to limit memory usage and I sent a patchset(https://lkml.org/lkml/2014/9/21/148) for VM to be harmonized with zram-swap to stop anonymous page reclaim if zram consumed memory up to the limit although there are free space on the swap. One problem for that direction is zram has no way to know any hole in memory space zsmalloc allocated by internal fragmentation so zram would regard swap is full although there are free space in zsmalloc. For solving the issue, zram want to trigger compaction of zsmalloc before it decides full or not. This patchset is first step to support above issues. For that, it adds indirect layer between handle and object location and supports manual compaction to solve 3th problem first of all. After this patchset got merged, next step is to make VM aware of zsmalloc compaction so that generic compaction will move zsmalloced-pages automatically in runtime. In my imaginary experiment(ie, high compress ratio data with heavy swap in/out on 8G zram-swap), data is as follows, Before = zram allocated object : 60212066 bytes zram total used: 140103680 bytes ratio: 42.98 percent MemFree: 840192 kB Compaction After = frag ratio after compaction zram allocated object : 60212066 bytes zram total used: 76185600 bytes ratio: 79.03 percent MemFree: 901932 kB Juneho reported below in his real platform with small aging. So, I think the benefit would be bigger in real aging system for a long time. - frag_ratio increased 3% (ie, higher is better) - memfree increased about 6MB - In buddy info, Normal 2^3: 4, 2^2: 1: 2^1 increased, Highmem: 2^1 21 increased frag ratio after swap fragment used : 156677 kbytes total: 166092 kbytes frag_ratio : 94 meminfo before compaction MemFree: 83724 kB Node 0, zone Normal 13642 1364 57 10 61 17 9 5 4 0 0 Node 0, zone HighMem 425 29 1 0 0 0 0 0 0 0 0 num_migrated : 23630 compaction done frag ratio after compaction used : 156673 kbytes total: 160564 kbytes frag_ratio : 97 meminfo after compaction MemFree: 89060 kB Node 0, zone Normal 14076 1544 67 14 61 17 9 5 4 0 0 Node 0, zone HighMem 863 50 1 0 0 0 0 0 0 0 0 This patchset adds more logics(about 480 lines) in zsmalloc but when I tested heavy swapin/out program, the regression for swapin/out speed is marginal because most of overheads were caused by compress/decompress and other MM reclaim stuff. This patch (of 7): Currently, handle of zsmalloc encodes object's location directly so it makes support of migration hard. This patch decouples handle and object via adding indirect layer. For that, it allocates handle dynamically and returns it to user. The handle is the address allocated by slab allocation so it's unique and we could keep object's location in the memory space allocated for handle. With it, we can change object's position without changing handle itself. Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:23 +00:00
{
pool->handle_cachep = kmem_cache_create("zs_handle", ZS_HANDLE_SIZE,
0, 0, NULL);
if (!pool->handle_cachep)
return 1;
pool->zspage_cachep = kmem_cache_create("zspage", sizeof(struct zspage),
0, 0, NULL);
if (!pool->zspage_cachep) {
kmem_cache_destroy(pool->handle_cachep);
pool->handle_cachep = NULL;
return 1;
}
return 0;
zsmalloc: decouple handle and object Recently, we started to use zram heavily and some of issues popped. 1) external fragmentation I got a report from Juneho Choi that fork failed although there are plenty of free pages in the system. His investigation revealed zram is one of the culprit to make heavy fragmentation so there was no more contiguous 16K page for pgd to fork in the ARM. 2) non-movable pages Other problem of zram now is that inherently, user want to use zram as swap in small memory system so they use zRAM with CMA to use memory efficiently. However, unfortunately, it doesn't work well because zRAM cannot use CMA's movable pages unless it doesn't support compaction. I got several reports about that OOM happened with zram although there are lots of swap space and free space in CMA area. 3) internal fragmentation zRAM has started support memory limitation feature to limit memory usage and I sent a patchset(https://lkml.org/lkml/2014/9/21/148) for VM to be harmonized with zram-swap to stop anonymous page reclaim if zram consumed memory up to the limit although there are free space on the swap. One problem for that direction is zram has no way to know any hole in memory space zsmalloc allocated by internal fragmentation so zram would regard swap is full although there are free space in zsmalloc. For solving the issue, zram want to trigger compaction of zsmalloc before it decides full or not. This patchset is first step to support above issues. For that, it adds indirect layer between handle and object location and supports manual compaction to solve 3th problem first of all. After this patchset got merged, next step is to make VM aware of zsmalloc compaction so that generic compaction will move zsmalloced-pages automatically in runtime. In my imaginary experiment(ie, high compress ratio data with heavy swap in/out on 8G zram-swap), data is as follows, Before = zram allocated object : 60212066 bytes zram total used: 140103680 bytes ratio: 42.98 percent MemFree: 840192 kB Compaction After = frag ratio after compaction zram allocated object : 60212066 bytes zram total used: 76185600 bytes ratio: 79.03 percent MemFree: 901932 kB Juneho reported below in his real platform with small aging. So, I think the benefit would be bigger in real aging system for a long time. - frag_ratio increased 3% (ie, higher is better) - memfree increased about 6MB - In buddy info, Normal 2^3: 4, 2^2: 1: 2^1 increased, Highmem: 2^1 21 increased frag ratio after swap fragment used : 156677 kbytes total: 166092 kbytes frag_ratio : 94 meminfo before compaction MemFree: 83724 kB Node 0, zone Normal 13642 1364 57 10 61 17 9 5 4 0 0 Node 0, zone HighMem 425 29 1 0 0 0 0 0 0 0 0 num_migrated : 23630 compaction done frag ratio after compaction used : 156673 kbytes total: 160564 kbytes frag_ratio : 97 meminfo after compaction MemFree: 89060 kB Node 0, zone Normal 14076 1544 67 14 61 17 9 5 4 0 0 Node 0, zone HighMem 863 50 1 0 0 0 0 0 0 0 0 This patchset adds more logics(about 480 lines) in zsmalloc but when I tested heavy swapin/out program, the regression for swapin/out speed is marginal because most of overheads were caused by compress/decompress and other MM reclaim stuff. This patch (of 7): Currently, handle of zsmalloc encodes object's location directly so it makes support of migration hard. This patch decouples handle and object via adding indirect layer. For that, it allocates handle dynamically and returns it to user. The handle is the address allocated by slab allocation so it's unique and we could keep object's location in the memory space allocated for handle. With it, we can change object's position without changing handle itself. Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:23 +00:00
}
static void destroy_cache(struct zs_pool *pool)
zsmalloc: decouple handle and object Recently, we started to use zram heavily and some of issues popped. 1) external fragmentation I got a report from Juneho Choi that fork failed although there are plenty of free pages in the system. His investigation revealed zram is one of the culprit to make heavy fragmentation so there was no more contiguous 16K page for pgd to fork in the ARM. 2) non-movable pages Other problem of zram now is that inherently, user want to use zram as swap in small memory system so they use zRAM with CMA to use memory efficiently. However, unfortunately, it doesn't work well because zRAM cannot use CMA's movable pages unless it doesn't support compaction. I got several reports about that OOM happened with zram although there are lots of swap space and free space in CMA area. 3) internal fragmentation zRAM has started support memory limitation feature to limit memory usage and I sent a patchset(https://lkml.org/lkml/2014/9/21/148) for VM to be harmonized with zram-swap to stop anonymous page reclaim if zram consumed memory up to the limit although there are free space on the swap. One problem for that direction is zram has no way to know any hole in memory space zsmalloc allocated by internal fragmentation so zram would regard swap is full although there are free space in zsmalloc. For solving the issue, zram want to trigger compaction of zsmalloc before it decides full or not. This patchset is first step to support above issues. For that, it adds indirect layer between handle and object location and supports manual compaction to solve 3th problem first of all. After this patchset got merged, next step is to make VM aware of zsmalloc compaction so that generic compaction will move zsmalloced-pages automatically in runtime. In my imaginary experiment(ie, high compress ratio data with heavy swap in/out on 8G zram-swap), data is as follows, Before = zram allocated object : 60212066 bytes zram total used: 140103680 bytes ratio: 42.98 percent MemFree: 840192 kB Compaction After = frag ratio after compaction zram allocated object : 60212066 bytes zram total used: 76185600 bytes ratio: 79.03 percent MemFree: 901932 kB Juneho reported below in his real platform with small aging. So, I think the benefit would be bigger in real aging system for a long time. - frag_ratio increased 3% (ie, higher is better) - memfree increased about 6MB - In buddy info, Normal 2^3: 4, 2^2: 1: 2^1 increased, Highmem: 2^1 21 increased frag ratio after swap fragment used : 156677 kbytes total: 166092 kbytes frag_ratio : 94 meminfo before compaction MemFree: 83724 kB Node 0, zone Normal 13642 1364 57 10 61 17 9 5 4 0 0 Node 0, zone HighMem 425 29 1 0 0 0 0 0 0 0 0 num_migrated : 23630 compaction done frag ratio after compaction used : 156673 kbytes total: 160564 kbytes frag_ratio : 97 meminfo after compaction MemFree: 89060 kB Node 0, zone Normal 14076 1544 67 14 61 17 9 5 4 0 0 Node 0, zone HighMem 863 50 1 0 0 0 0 0 0 0 0 This patchset adds more logics(about 480 lines) in zsmalloc but when I tested heavy swapin/out program, the regression for swapin/out speed is marginal because most of overheads were caused by compress/decompress and other MM reclaim stuff. This patch (of 7): Currently, handle of zsmalloc encodes object's location directly so it makes support of migration hard. This patch decouples handle and object via adding indirect layer. For that, it allocates handle dynamically and returns it to user. The handle is the address allocated by slab allocation so it's unique and we could keep object's location in the memory space allocated for handle. With it, we can change object's position without changing handle itself. Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:23 +00:00
{
kmem_cache_destroy(pool->handle_cachep);
kmem_cache_destroy(pool->zspage_cachep);
zsmalloc: decouple handle and object Recently, we started to use zram heavily and some of issues popped. 1) external fragmentation I got a report from Juneho Choi that fork failed although there are plenty of free pages in the system. His investigation revealed zram is one of the culprit to make heavy fragmentation so there was no more contiguous 16K page for pgd to fork in the ARM. 2) non-movable pages Other problem of zram now is that inherently, user want to use zram as swap in small memory system so they use zRAM with CMA to use memory efficiently. However, unfortunately, it doesn't work well because zRAM cannot use CMA's movable pages unless it doesn't support compaction. I got several reports about that OOM happened with zram although there are lots of swap space and free space in CMA area. 3) internal fragmentation zRAM has started support memory limitation feature to limit memory usage and I sent a patchset(https://lkml.org/lkml/2014/9/21/148) for VM to be harmonized with zram-swap to stop anonymous page reclaim if zram consumed memory up to the limit although there are free space on the swap. One problem for that direction is zram has no way to know any hole in memory space zsmalloc allocated by internal fragmentation so zram would regard swap is full although there are free space in zsmalloc. For solving the issue, zram want to trigger compaction of zsmalloc before it decides full or not. This patchset is first step to support above issues. For that, it adds indirect layer between handle and object location and supports manual compaction to solve 3th problem first of all. After this patchset got merged, next step is to make VM aware of zsmalloc compaction so that generic compaction will move zsmalloced-pages automatically in runtime. In my imaginary experiment(ie, high compress ratio data with heavy swap in/out on 8G zram-swap), data is as follows, Before = zram allocated object : 60212066 bytes zram total used: 140103680 bytes ratio: 42.98 percent MemFree: 840192 kB Compaction After = frag ratio after compaction zram allocated object : 60212066 bytes zram total used: 76185600 bytes ratio: 79.03 percent MemFree: 901932 kB Juneho reported below in his real platform with small aging. So, I think the benefit would be bigger in real aging system for a long time. - frag_ratio increased 3% (ie, higher is better) - memfree increased about 6MB - In buddy info, Normal 2^3: 4, 2^2: 1: 2^1 increased, Highmem: 2^1 21 increased frag ratio after swap fragment used : 156677 kbytes total: 166092 kbytes frag_ratio : 94 meminfo before compaction MemFree: 83724 kB Node 0, zone Normal 13642 1364 57 10 61 17 9 5 4 0 0 Node 0, zone HighMem 425 29 1 0 0 0 0 0 0 0 0 num_migrated : 23630 compaction done frag ratio after compaction used : 156673 kbytes total: 160564 kbytes frag_ratio : 97 meminfo after compaction MemFree: 89060 kB Node 0, zone Normal 14076 1544 67 14 61 17 9 5 4 0 0 Node 0, zone HighMem 863 50 1 0 0 0 0 0 0 0 0 This patchset adds more logics(about 480 lines) in zsmalloc but when I tested heavy swapin/out program, the regression for swapin/out speed is marginal because most of overheads were caused by compress/decompress and other MM reclaim stuff. This patch (of 7): Currently, handle of zsmalloc encodes object's location directly so it makes support of migration hard. This patch decouples handle and object via adding indirect layer. For that, it allocates handle dynamically and returns it to user. The handle is the address allocated by slab allocation so it's unique and we could keep object's location in the memory space allocated for handle. With it, we can change object's position without changing handle itself. Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:23 +00:00
}
static unsigned long cache_alloc_handle(struct zs_pool *pool, gfp_t gfp)
zsmalloc: decouple handle and object Recently, we started to use zram heavily and some of issues popped. 1) external fragmentation I got a report from Juneho Choi that fork failed although there are plenty of free pages in the system. His investigation revealed zram is one of the culprit to make heavy fragmentation so there was no more contiguous 16K page for pgd to fork in the ARM. 2) non-movable pages Other problem of zram now is that inherently, user want to use zram as swap in small memory system so they use zRAM with CMA to use memory efficiently. However, unfortunately, it doesn't work well because zRAM cannot use CMA's movable pages unless it doesn't support compaction. I got several reports about that OOM happened with zram although there are lots of swap space and free space in CMA area. 3) internal fragmentation zRAM has started support memory limitation feature to limit memory usage and I sent a patchset(https://lkml.org/lkml/2014/9/21/148) for VM to be harmonized with zram-swap to stop anonymous page reclaim if zram consumed memory up to the limit although there are free space on the swap. One problem for that direction is zram has no way to know any hole in memory space zsmalloc allocated by internal fragmentation so zram would regard swap is full although there are free space in zsmalloc. For solving the issue, zram want to trigger compaction of zsmalloc before it decides full or not. This patchset is first step to support above issues. For that, it adds indirect layer between handle and object location and supports manual compaction to solve 3th problem first of all. After this patchset got merged, next step is to make VM aware of zsmalloc compaction so that generic compaction will move zsmalloced-pages automatically in runtime. In my imaginary experiment(ie, high compress ratio data with heavy swap in/out on 8G zram-swap), data is as follows, Before = zram allocated object : 60212066 bytes zram total used: 140103680 bytes ratio: 42.98 percent MemFree: 840192 kB Compaction After = frag ratio after compaction zram allocated object : 60212066 bytes zram total used: 76185600 bytes ratio: 79.03 percent MemFree: 901932 kB Juneho reported below in his real platform with small aging. So, I think the benefit would be bigger in real aging system for a long time. - frag_ratio increased 3% (ie, higher is better) - memfree increased about 6MB - In buddy info, Normal 2^3: 4, 2^2: 1: 2^1 increased, Highmem: 2^1 21 increased frag ratio after swap fragment used : 156677 kbytes total: 166092 kbytes frag_ratio : 94 meminfo before compaction MemFree: 83724 kB Node 0, zone Normal 13642 1364 57 10 61 17 9 5 4 0 0 Node 0, zone HighMem 425 29 1 0 0 0 0 0 0 0 0 num_migrated : 23630 compaction done frag ratio after compaction used : 156673 kbytes total: 160564 kbytes frag_ratio : 97 meminfo after compaction MemFree: 89060 kB Node 0, zone Normal 14076 1544 67 14 61 17 9 5 4 0 0 Node 0, zone HighMem 863 50 1 0 0 0 0 0 0 0 0 This patchset adds more logics(about 480 lines) in zsmalloc but when I tested heavy swapin/out program, the regression for swapin/out speed is marginal because most of overheads were caused by compress/decompress and other MM reclaim stuff. This patch (of 7): Currently, handle of zsmalloc encodes object's location directly so it makes support of migration hard. This patch decouples handle and object via adding indirect layer. For that, it allocates handle dynamically and returns it to user. The handle is the address allocated by slab allocation so it's unique and we could keep object's location in the memory space allocated for handle. With it, we can change object's position without changing handle itself. Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:23 +00:00
{
return (unsigned long)kmem_cache_alloc(pool->handle_cachep,
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
gfp & ~(__GFP_HIGHMEM|__GFP_MOVABLE));
zsmalloc: decouple handle and object Recently, we started to use zram heavily and some of issues popped. 1) external fragmentation I got a report from Juneho Choi that fork failed although there are plenty of free pages in the system. His investigation revealed zram is one of the culprit to make heavy fragmentation so there was no more contiguous 16K page for pgd to fork in the ARM. 2) non-movable pages Other problem of zram now is that inherently, user want to use zram as swap in small memory system so they use zRAM with CMA to use memory efficiently. However, unfortunately, it doesn't work well because zRAM cannot use CMA's movable pages unless it doesn't support compaction. I got several reports about that OOM happened with zram although there are lots of swap space and free space in CMA area. 3) internal fragmentation zRAM has started support memory limitation feature to limit memory usage and I sent a patchset(https://lkml.org/lkml/2014/9/21/148) for VM to be harmonized with zram-swap to stop anonymous page reclaim if zram consumed memory up to the limit although there are free space on the swap. One problem for that direction is zram has no way to know any hole in memory space zsmalloc allocated by internal fragmentation so zram would regard swap is full although there are free space in zsmalloc. For solving the issue, zram want to trigger compaction of zsmalloc before it decides full or not. This patchset is first step to support above issues. For that, it adds indirect layer between handle and object location and supports manual compaction to solve 3th problem first of all. After this patchset got merged, next step is to make VM aware of zsmalloc compaction so that generic compaction will move zsmalloced-pages automatically in runtime. In my imaginary experiment(ie, high compress ratio data with heavy swap in/out on 8G zram-swap), data is as follows, Before = zram allocated object : 60212066 bytes zram total used: 140103680 bytes ratio: 42.98 percent MemFree: 840192 kB Compaction After = frag ratio after compaction zram allocated object : 60212066 bytes zram total used: 76185600 bytes ratio: 79.03 percent MemFree: 901932 kB Juneho reported below in his real platform with small aging. So, I think the benefit would be bigger in real aging system for a long time. - frag_ratio increased 3% (ie, higher is better) - memfree increased about 6MB - In buddy info, Normal 2^3: 4, 2^2: 1: 2^1 increased, Highmem: 2^1 21 increased frag ratio after swap fragment used : 156677 kbytes total: 166092 kbytes frag_ratio : 94 meminfo before compaction MemFree: 83724 kB Node 0, zone Normal 13642 1364 57 10 61 17 9 5 4 0 0 Node 0, zone HighMem 425 29 1 0 0 0 0 0 0 0 0 num_migrated : 23630 compaction done frag ratio after compaction used : 156673 kbytes total: 160564 kbytes frag_ratio : 97 meminfo after compaction MemFree: 89060 kB Node 0, zone Normal 14076 1544 67 14 61 17 9 5 4 0 0 Node 0, zone HighMem 863 50 1 0 0 0 0 0 0 0 0 This patchset adds more logics(about 480 lines) in zsmalloc but when I tested heavy swapin/out program, the regression for swapin/out speed is marginal because most of overheads were caused by compress/decompress and other MM reclaim stuff. This patch (of 7): Currently, handle of zsmalloc encodes object's location directly so it makes support of migration hard. This patch decouples handle and object via adding indirect layer. For that, it allocates handle dynamically and returns it to user. The handle is the address allocated by slab allocation so it's unique and we could keep object's location in the memory space allocated for handle. With it, we can change object's position without changing handle itself. Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:23 +00:00
}
static void cache_free_handle(struct zs_pool *pool, unsigned long handle)
zsmalloc: decouple handle and object Recently, we started to use zram heavily and some of issues popped. 1) external fragmentation I got a report from Juneho Choi that fork failed although there are plenty of free pages in the system. His investigation revealed zram is one of the culprit to make heavy fragmentation so there was no more contiguous 16K page for pgd to fork in the ARM. 2) non-movable pages Other problem of zram now is that inherently, user want to use zram as swap in small memory system so they use zRAM with CMA to use memory efficiently. However, unfortunately, it doesn't work well because zRAM cannot use CMA's movable pages unless it doesn't support compaction. I got several reports about that OOM happened with zram although there are lots of swap space and free space in CMA area. 3) internal fragmentation zRAM has started support memory limitation feature to limit memory usage and I sent a patchset(https://lkml.org/lkml/2014/9/21/148) for VM to be harmonized with zram-swap to stop anonymous page reclaim if zram consumed memory up to the limit although there are free space on the swap. One problem for that direction is zram has no way to know any hole in memory space zsmalloc allocated by internal fragmentation so zram would regard swap is full although there are free space in zsmalloc. For solving the issue, zram want to trigger compaction of zsmalloc before it decides full or not. This patchset is first step to support above issues. For that, it adds indirect layer between handle and object location and supports manual compaction to solve 3th problem first of all. After this patchset got merged, next step is to make VM aware of zsmalloc compaction so that generic compaction will move zsmalloced-pages automatically in runtime. In my imaginary experiment(ie, high compress ratio data with heavy swap in/out on 8G zram-swap), data is as follows, Before = zram allocated object : 60212066 bytes zram total used: 140103680 bytes ratio: 42.98 percent MemFree: 840192 kB Compaction After = frag ratio after compaction zram allocated object : 60212066 bytes zram total used: 76185600 bytes ratio: 79.03 percent MemFree: 901932 kB Juneho reported below in his real platform with small aging. So, I think the benefit would be bigger in real aging system for a long time. - frag_ratio increased 3% (ie, higher is better) - memfree increased about 6MB - In buddy info, Normal 2^3: 4, 2^2: 1: 2^1 increased, Highmem: 2^1 21 increased frag ratio after swap fragment used : 156677 kbytes total: 166092 kbytes frag_ratio : 94 meminfo before compaction MemFree: 83724 kB Node 0, zone Normal 13642 1364 57 10 61 17 9 5 4 0 0 Node 0, zone HighMem 425 29 1 0 0 0 0 0 0 0 0 num_migrated : 23630 compaction done frag ratio after compaction used : 156673 kbytes total: 160564 kbytes frag_ratio : 97 meminfo after compaction MemFree: 89060 kB Node 0, zone Normal 14076 1544 67 14 61 17 9 5 4 0 0 Node 0, zone HighMem 863 50 1 0 0 0 0 0 0 0 0 This patchset adds more logics(about 480 lines) in zsmalloc but when I tested heavy swapin/out program, the regression for swapin/out speed is marginal because most of overheads were caused by compress/decompress and other MM reclaim stuff. This patch (of 7): Currently, handle of zsmalloc encodes object's location directly so it makes support of migration hard. This patch decouples handle and object via adding indirect layer. For that, it allocates handle dynamically and returns it to user. The handle is the address allocated by slab allocation so it's unique and we could keep object's location in the memory space allocated for handle. With it, we can change object's position without changing handle itself. Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:23 +00:00
{
kmem_cache_free(pool->handle_cachep, (void *)handle);
}
static struct zspage *cache_alloc_zspage(struct zs_pool *pool, gfp_t flags)
{
return kmem_cache_zalloc(pool->zspage_cachep,
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
flags & ~(__GFP_HIGHMEM|__GFP_MOVABLE));
}
static void cache_free_zspage(struct zs_pool *pool, struct zspage *zspage)
{
kmem_cache_free(pool->zspage_cachep, zspage);
}
zsmalloc: consolidate zs_pool's migrate_lock and size_class's locks Currently, zsmalloc has a hierarchy of locks, which includes a pool-level migrate_lock, and a lock for each size class. We have to obtain both locks in the hotpath in most cases anyway, except for zs_malloc. This exception will no longer exist when we introduce a LRU into the zs_pool for the new writeback functionality - we will need to obtain a pool-level lock to synchronize LRU handling even in zs_malloc. In preparation for zsmalloc writeback, consolidate these locks into a single pool-level lock, which drastically reduces the complexity of synchronization in zsmalloc. We have also benchmarked the lock consolidation to see the performance effect of this change on zram. First, we ran a synthetic FS workload on a server machine with 36 cores (same machine for all runs), using fs_mark -d ../zram1mnt -s 100000 -n 2500 -t 32 -k before and after for btrfs and ext4 on zram (FS usage is 80%). Here is the result (unit is file/second): With lock consolidation (btrfs): Average: 13520.2, Median: 13531.0, Stddev: 137.5961482019028 Without lock consolidation (btrfs): Average: 13487.2, Median: 13575.0, Stddev: 309.08283679298665 With lock consolidation (ext4): Average: 16824.4, Median: 16839.0, Stddev: 89.97388510006668 Without lock consolidation (ext4) Average: 16958.0, Median: 16986.0, Stddev: 194.7370021336469 As you can see, we observe a 0.3% regression for btrfs, and a 0.9% regression for ext4. This is a small, barely measurable difference in my opinion. For a more realistic scenario, we also tries building the kernel on zram. Here is the time it takes (in seconds): With lock consolidation (btrfs): real Average: 319.6, Median: 320.0, Stddev: 0.8944271909999159 user Average: 6894.2, Median: 6895.0, Stddev: 25.528415540334656 sys Average: 521.4, Median: 522.0, Stddev: 1.51657508881031 Without lock consolidation (btrfs): real Average: 319.8, Median: 320.0, Stddev: 0.8366600265340756 user Average: 6896.6, Median: 6899.0, Stddev: 16.04057355583023 sys Average: 520.6, Median: 521.0, Stddev: 1.140175425099138 With lock consolidation (ext4): real Average: 320.0, Median: 319.0, Stddev: 1.4142135623730951 user Average: 6896.8, Median: 6878.0, Stddev: 28.621670111997307 sys Average: 521.2, Median: 521.0, Stddev: 1.7888543819998317 Without lock consolidation (ext4) real Average: 319.6, Median: 319.0, Stddev: 0.8944271909999159 user Average: 6886.2, Median: 6887.0, Stddev: 16.93221781102523 sys Average: 520.4, Median: 520.0, Stddev: 1.140175425099138 The difference is entirely within the noise of a typical run on zram. This hardly justifies the complexity of maintaining both the pool lock and the class lock. In fact, for writeback, we would need to introduce yet another lock to prevent data races on the pool's LRU, further complicating the lock handling logic. IMHO, it is just better to collapse all of these into a single pool-level lock. Link: https://lkml.kernel.org/r/20221128191616.1261026-4-nphamcs@gmail.com Signed-off-by: Nhat Pham <nphamcs@gmail.com> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Acked-by: Minchan Kim <minchan@kernel.org> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-11-28 19:16:12 +00:00
/* pool->lock(which owns the handle) synchronizes races */
zsmalloc: decouple handle and object Recently, we started to use zram heavily and some of issues popped. 1) external fragmentation I got a report from Juneho Choi that fork failed although there are plenty of free pages in the system. His investigation revealed zram is one of the culprit to make heavy fragmentation so there was no more contiguous 16K page for pgd to fork in the ARM. 2) non-movable pages Other problem of zram now is that inherently, user want to use zram as swap in small memory system so they use zRAM with CMA to use memory efficiently. However, unfortunately, it doesn't work well because zRAM cannot use CMA's movable pages unless it doesn't support compaction. I got several reports about that OOM happened with zram although there are lots of swap space and free space in CMA area. 3) internal fragmentation zRAM has started support memory limitation feature to limit memory usage and I sent a patchset(https://lkml.org/lkml/2014/9/21/148) for VM to be harmonized with zram-swap to stop anonymous page reclaim if zram consumed memory up to the limit although there are free space on the swap. One problem for that direction is zram has no way to know any hole in memory space zsmalloc allocated by internal fragmentation so zram would regard swap is full although there are free space in zsmalloc. For solving the issue, zram want to trigger compaction of zsmalloc before it decides full or not. This patchset is first step to support above issues. For that, it adds indirect layer between handle and object location and supports manual compaction to solve 3th problem first of all. After this patchset got merged, next step is to make VM aware of zsmalloc compaction so that generic compaction will move zsmalloced-pages automatically in runtime. In my imaginary experiment(ie, high compress ratio data with heavy swap in/out on 8G zram-swap), data is as follows, Before = zram allocated object : 60212066 bytes zram total used: 140103680 bytes ratio: 42.98 percent MemFree: 840192 kB Compaction After = frag ratio after compaction zram allocated object : 60212066 bytes zram total used: 76185600 bytes ratio: 79.03 percent MemFree: 901932 kB Juneho reported below in his real platform with small aging. So, I think the benefit would be bigger in real aging system for a long time. - frag_ratio increased 3% (ie, higher is better) - memfree increased about 6MB - In buddy info, Normal 2^3: 4, 2^2: 1: 2^1 increased, Highmem: 2^1 21 increased frag ratio after swap fragment used : 156677 kbytes total: 166092 kbytes frag_ratio : 94 meminfo before compaction MemFree: 83724 kB Node 0, zone Normal 13642 1364 57 10 61 17 9 5 4 0 0 Node 0, zone HighMem 425 29 1 0 0 0 0 0 0 0 0 num_migrated : 23630 compaction done frag ratio after compaction used : 156673 kbytes total: 160564 kbytes frag_ratio : 97 meminfo after compaction MemFree: 89060 kB Node 0, zone Normal 14076 1544 67 14 61 17 9 5 4 0 0 Node 0, zone HighMem 863 50 1 0 0 0 0 0 0 0 0 This patchset adds more logics(about 480 lines) in zsmalloc but when I tested heavy swapin/out program, the regression for swapin/out speed is marginal because most of overheads were caused by compress/decompress and other MM reclaim stuff. This patch (of 7): Currently, handle of zsmalloc encodes object's location directly so it makes support of migration hard. This patch decouples handle and object via adding indirect layer. For that, it allocates handle dynamically and returns it to user. The handle is the address allocated by slab allocation so it's unique and we could keep object's location in the memory space allocated for handle. With it, we can change object's position without changing handle itself. Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:23 +00:00
static void record_obj(unsigned long handle, unsigned long obj)
{
zsmalloc: replace per zpage lock with pool->migrate_lock The zsmalloc has used a bit for spin_lock in zpage handle to keep zpage object alive during several operations. However, it causes the problem for PREEMPT_RT as well as introducing too complicated. This patch replaces the bit spin_lock with pool->migrate_lock rwlock. It could make the code simple as well as zsmalloc work under PREEMPT_RT. The drawback is the pool->migrate_lock is bigger granuarity than per zpage lock so the contention would be higher than old when both IO-related operations(i.e., zsmalloc, zsfree, zs_[map|unmap]) and compaction(page/zpage migration) are going in parallel(*, the migrate_lock is rwlock and IO related functions are all read side lock so there is no contention). However, the write-side is fast enough(dominant overhead is just page copy) so it wouldn't affect much. If the lock granurity becomes more problem later, we could introduce table locks based on handle as a hash value. Link: https://lkml.kernel.org/r/20211115185909.3949505-9-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:14:13 +00:00
*(unsigned long *)handle = obj;
zsmalloc: decouple handle and object Recently, we started to use zram heavily and some of issues popped. 1) external fragmentation I got a report from Juneho Choi that fork failed although there are plenty of free pages in the system. His investigation revealed zram is one of the culprit to make heavy fragmentation so there was no more contiguous 16K page for pgd to fork in the ARM. 2) non-movable pages Other problem of zram now is that inherently, user want to use zram as swap in small memory system so they use zRAM with CMA to use memory efficiently. However, unfortunately, it doesn't work well because zRAM cannot use CMA's movable pages unless it doesn't support compaction. I got several reports about that OOM happened with zram although there are lots of swap space and free space in CMA area. 3) internal fragmentation zRAM has started support memory limitation feature to limit memory usage and I sent a patchset(https://lkml.org/lkml/2014/9/21/148) for VM to be harmonized with zram-swap to stop anonymous page reclaim if zram consumed memory up to the limit although there are free space on the swap. One problem for that direction is zram has no way to know any hole in memory space zsmalloc allocated by internal fragmentation so zram would regard swap is full although there are free space in zsmalloc. For solving the issue, zram want to trigger compaction of zsmalloc before it decides full or not. This patchset is first step to support above issues. For that, it adds indirect layer between handle and object location and supports manual compaction to solve 3th problem first of all. After this patchset got merged, next step is to make VM aware of zsmalloc compaction so that generic compaction will move zsmalloced-pages automatically in runtime. In my imaginary experiment(ie, high compress ratio data with heavy swap in/out on 8G zram-swap), data is as follows, Before = zram allocated object : 60212066 bytes zram total used: 140103680 bytes ratio: 42.98 percent MemFree: 840192 kB Compaction After = frag ratio after compaction zram allocated object : 60212066 bytes zram total used: 76185600 bytes ratio: 79.03 percent MemFree: 901932 kB Juneho reported below in his real platform with small aging. So, I think the benefit would be bigger in real aging system for a long time. - frag_ratio increased 3% (ie, higher is better) - memfree increased about 6MB - In buddy info, Normal 2^3: 4, 2^2: 1: 2^1 increased, Highmem: 2^1 21 increased frag ratio after swap fragment used : 156677 kbytes total: 166092 kbytes frag_ratio : 94 meminfo before compaction MemFree: 83724 kB Node 0, zone Normal 13642 1364 57 10 61 17 9 5 4 0 0 Node 0, zone HighMem 425 29 1 0 0 0 0 0 0 0 0 num_migrated : 23630 compaction done frag ratio after compaction used : 156673 kbytes total: 160564 kbytes frag_ratio : 97 meminfo after compaction MemFree: 89060 kB Node 0, zone Normal 14076 1544 67 14 61 17 9 5 4 0 0 Node 0, zone HighMem 863 50 1 0 0 0 0 0 0 0 0 This patchset adds more logics(about 480 lines) in zsmalloc but when I tested heavy swapin/out program, the regression for swapin/out speed is marginal because most of overheads were caused by compress/decompress and other MM reclaim stuff. This patch (of 7): Currently, handle of zsmalloc encodes object's location directly so it makes support of migration hard. This patch decouples handle and object via adding indirect layer. For that, it allocates handle dynamically and returns it to user. The handle is the address allocated by slab allocation so it's unique and we could keep object's location in the memory space allocated for handle. With it, we can change object's position without changing handle itself. Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:23 +00:00
}
/* zpool driver */
#ifdef CONFIG_ZPOOL
static void *zs_zpool_create(const char *name, gfp_t gfp,
const struct zpool_ops *zpool_ops,
struct zpool *zpool)
{
/*
* Ignore global gfp flags: zs_malloc() may be invoked from
* different contexts and its caller must provide a valid
* gfp mask.
*/
struct zs_pool *pool = zs_create_pool(name);
if (pool) {
pool->zpool = zpool;
pool->zpool_ops = zpool_ops;
}
return pool;
}
static void zs_zpool_destroy(void *pool)
{
zs_destroy_pool(pool);
}
static int zs_zpool_malloc(void *pool, size_t size, gfp_t gfp,
unsigned long *handle)
{
*handle = zs_malloc(pool, size, gfp);
if (IS_ERR_VALUE(*handle))
return PTR_ERR((void *)*handle);
return 0;
}
static void zs_zpool_free(void *pool, unsigned long handle)
{
zs_free(pool, handle);
}
static int zs_reclaim_page(struct zs_pool *pool, unsigned int retries);
static int zs_zpool_shrink(void *pool, unsigned int pages,
unsigned int *reclaimed)
{
unsigned int total = 0;
int ret = -EINVAL;
while (total < pages) {
ret = zs_reclaim_page(pool, 8);
if (ret < 0)
break;
total++;
}
if (reclaimed)
*reclaimed = total;
return ret;
}
static void *zs_zpool_map(void *pool, unsigned long handle,
enum zpool_mapmode mm)
{
enum zs_mapmode zs_mm;
switch (mm) {
case ZPOOL_MM_RO:
zs_mm = ZS_MM_RO;
break;
case ZPOOL_MM_WO:
zs_mm = ZS_MM_WO;
break;
case ZPOOL_MM_RW:
default:
zs_mm = ZS_MM_RW;
break;
}
return zs_map_object(pool, handle, zs_mm);
}
static void zs_zpool_unmap(void *pool, unsigned long handle)
{
zs_unmap_object(pool, handle);
}
static u64 zs_zpool_total_size(void *pool)
{
return zs_get_total_pages(pool) << PAGE_SHIFT;
}
static struct zpool_driver zs_zpool_driver = {
.type = "zsmalloc",
.owner = THIS_MODULE,
.create = zs_zpool_create,
.destroy = zs_zpool_destroy,
.malloc_support_movable = true,
.malloc = zs_zpool_malloc,
.free = zs_zpool_free,
.shrink = zs_zpool_shrink,
.map = zs_zpool_map,
.unmap = zs_zpool_unmap,
.total_size = zs_zpool_total_size,
};
MODULE_ALIAS("zpool-zsmalloc");
#endif /* CONFIG_ZPOOL */
/* per-cpu VM mapping areas for zspage accesses that cross page boundaries */
static DEFINE_PER_CPU(struct mapping_area, zs_map_area) = {
.lock = INIT_LOCAL_LOCK(lock),
};
static __maybe_unused int is_first_page(struct page *page)
{
return PagePrivate(page);
}
zsmalloc: consolidate zs_pool's migrate_lock and size_class's locks Currently, zsmalloc has a hierarchy of locks, which includes a pool-level migrate_lock, and a lock for each size class. We have to obtain both locks in the hotpath in most cases anyway, except for zs_malloc. This exception will no longer exist when we introduce a LRU into the zs_pool for the new writeback functionality - we will need to obtain a pool-level lock to synchronize LRU handling even in zs_malloc. In preparation for zsmalloc writeback, consolidate these locks into a single pool-level lock, which drastically reduces the complexity of synchronization in zsmalloc. We have also benchmarked the lock consolidation to see the performance effect of this change on zram. First, we ran a synthetic FS workload on a server machine with 36 cores (same machine for all runs), using fs_mark -d ../zram1mnt -s 100000 -n 2500 -t 32 -k before and after for btrfs and ext4 on zram (FS usage is 80%). Here is the result (unit is file/second): With lock consolidation (btrfs): Average: 13520.2, Median: 13531.0, Stddev: 137.5961482019028 Without lock consolidation (btrfs): Average: 13487.2, Median: 13575.0, Stddev: 309.08283679298665 With lock consolidation (ext4): Average: 16824.4, Median: 16839.0, Stddev: 89.97388510006668 Without lock consolidation (ext4) Average: 16958.0, Median: 16986.0, Stddev: 194.7370021336469 As you can see, we observe a 0.3% regression for btrfs, and a 0.9% regression for ext4. This is a small, barely measurable difference in my opinion. For a more realistic scenario, we also tries building the kernel on zram. Here is the time it takes (in seconds): With lock consolidation (btrfs): real Average: 319.6, Median: 320.0, Stddev: 0.8944271909999159 user Average: 6894.2, Median: 6895.0, Stddev: 25.528415540334656 sys Average: 521.4, Median: 522.0, Stddev: 1.51657508881031 Without lock consolidation (btrfs): real Average: 319.8, Median: 320.0, Stddev: 0.8366600265340756 user Average: 6896.6, Median: 6899.0, Stddev: 16.04057355583023 sys Average: 520.6, Median: 521.0, Stddev: 1.140175425099138 With lock consolidation (ext4): real Average: 320.0, Median: 319.0, Stddev: 1.4142135623730951 user Average: 6896.8, Median: 6878.0, Stddev: 28.621670111997307 sys Average: 521.2, Median: 521.0, Stddev: 1.7888543819998317 Without lock consolidation (ext4) real Average: 319.6, Median: 319.0, Stddev: 0.8944271909999159 user Average: 6886.2, Median: 6887.0, Stddev: 16.93221781102523 sys Average: 520.4, Median: 520.0, Stddev: 1.140175425099138 The difference is entirely within the noise of a typical run on zram. This hardly justifies the complexity of maintaining both the pool lock and the class lock. In fact, for writeback, we would need to introduce yet another lock to prevent data races on the pool's LRU, further complicating the lock handling logic. IMHO, it is just better to collapse all of these into a single pool-level lock. Link: https://lkml.kernel.org/r/20221128191616.1261026-4-nphamcs@gmail.com Signed-off-by: Nhat Pham <nphamcs@gmail.com> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Acked-by: Minchan Kim <minchan@kernel.org> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-11-28 19:16:12 +00:00
/* Protected by pool->lock */
static inline int get_zspage_inuse(struct zspage *zspage)
{
return zspage->inuse;
}
static inline void mod_zspage_inuse(struct zspage *zspage, int val)
{
zspage->inuse += val;
}
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
static inline struct page *get_first_page(struct zspage *zspage)
{
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
struct page *first_page = zspage->first_page;
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
VM_BUG_ON_PAGE(!is_first_page(first_page), first_page);
return first_page;
}
static inline unsigned int get_first_obj_offset(struct page *page)
{
return page->page_type;
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
}
static inline void set_first_obj_offset(struct page *page, unsigned int offset)
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
{
page->page_type = offset;
}
static inline unsigned int get_freeobj(struct zspage *zspage)
{
return zspage->freeobj;
}
static inline void set_freeobj(struct zspage *zspage, unsigned int obj)
{
zspage->freeobj = obj;
}
static void get_zspage_mapping(struct zspage *zspage,
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
unsigned int *class_idx,
int *fullness)
{
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
BUG_ON(zspage->magic != ZSPAGE_MAGIC);
*fullness = zspage->fullness;
*class_idx = zspage->class;
}
zsmalloc: introduce some helper functions Patch series "zsmalloc: remove bit_spin_lock", v2. zsmalloc uses bit_spin_lock to minimize space overhead since it's zpage granularity lock. However, it causes zsmalloc non-working under PREEMPT_RT as well as adding too much complication. This patchset tries to replace the bit_spin_lock with per-pool rwlock. It also removes unnecessary zspage isolation logic from class, which was the other part too much complication added into zsmalloc. Last patch changes the get_cpu_var to local_lock to make it work in PREEMPT_RT. This patch (of 9): get_zspage_mapping returns fullness as well as class_idx. However, the fullness is usually not used since it could be stale in some contexts. It causes misleading as well as unnecessary instructions so this patch introduces zspage_class. obj_to_location also produces page and index but we don't need always the index, either so this patch introduces obj_to_page. Link: https://lkml.kernel.org/r/20211115185909.3949505-1-minchan@kernel.org Link: https://lkml.kernel.org/r/20211115185909.3949505-2-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:13:51 +00:00
static struct size_class *zspage_class(struct zs_pool *pool,
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
struct zspage *zspage)
zsmalloc: introduce some helper functions Patch series "zsmalloc: remove bit_spin_lock", v2. zsmalloc uses bit_spin_lock to minimize space overhead since it's zpage granularity lock. However, it causes zsmalloc non-working under PREEMPT_RT as well as adding too much complication. This patchset tries to replace the bit_spin_lock with per-pool rwlock. It also removes unnecessary zspage isolation logic from class, which was the other part too much complication added into zsmalloc. Last patch changes the get_cpu_var to local_lock to make it work in PREEMPT_RT. This patch (of 9): get_zspage_mapping returns fullness as well as class_idx. However, the fullness is usually not used since it could be stale in some contexts. It causes misleading as well as unnecessary instructions so this patch introduces zspage_class. obj_to_location also produces page and index but we don't need always the index, either so this patch introduces obj_to_page. Link: https://lkml.kernel.org/r/20211115185909.3949505-1-minchan@kernel.org Link: https://lkml.kernel.org/r/20211115185909.3949505-2-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:13:51 +00:00
{
return pool->size_class[zspage->class];
}
static void set_zspage_mapping(struct zspage *zspage,
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
unsigned int class_idx,
int fullness)
{
zspage->class = class_idx;
zspage->fullness = fullness;
}
/*
* zsmalloc divides the pool into various size classes where each
* class maintains a list of zspages where each zspage is divided
* into equal sized chunks. Each allocation falls into one of these
* classes depending on its size. This function returns index of the
* size class which has chunk size big enough to hold the given size.
*/
static int get_size_class_index(int size)
{
int idx = 0;
if (likely(size > ZS_MIN_ALLOC_SIZE))
idx = DIV_ROUND_UP(size - ZS_MIN_ALLOC_SIZE,
ZS_SIZE_CLASS_DELTA);
return min_t(int, ZS_SIZE_CLASSES - 1, idx);
}
static inline void class_stat_inc(struct size_class *class,
int type, unsigned long cnt)
{
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
class->stats.objs[type] += cnt;
}
static inline void class_stat_dec(struct size_class *class,
int type, unsigned long cnt)
{
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
class->stats.objs[type] -= cnt;
}
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
static inline unsigned long zs_stat_get(struct size_class *class, int type)
{
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
return class->stats.objs[type];
}
#ifdef CONFIG_ZSMALLOC_STAT
static void __init zs_stat_init(void)
{
if (!debugfs_initialized()) {
pr_warn("debugfs not available, stat dir not created\n");
return;
}
zs_stat_root = debugfs_create_dir("zsmalloc", NULL);
}
static void __exit zs_stat_exit(void)
{
debugfs_remove_recursive(zs_stat_root);
}
static unsigned long zs_can_compact(struct size_class *class);
static int zs_stats_size_show(struct seq_file *s, void *v)
{
int i, fg;
struct zs_pool *pool = s->private;
struct size_class *class;
int objs_per_zspage;
unsigned long obj_allocated, obj_used, pages_used, freeable;
unsigned long total_objs = 0, total_used_objs = 0, total_pages = 0;
unsigned long total_freeable = 0;
unsigned long inuse_totals[NR_FULLNESS_GROUPS] = {0, };
seq_printf(s, " %5s %5s %9s %9s %9s %9s %9s %9s %9s %9s %9s %9s %9s %13s %10s %10s %16s %8s\n",
"class", "size", "10%", "20%", "30%", "40%",
"50%", "60%", "70%", "80%", "90%", "99%", "100%",
"obj_allocated", "obj_used", "pages_used",
"pages_per_zspage", "freeable");
for (i = 0; i < ZS_SIZE_CLASSES; i++) {
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
class = pool->size_class[i];
if (class->index != i)
continue;
zsmalloc: consolidate zs_pool's migrate_lock and size_class's locks Currently, zsmalloc has a hierarchy of locks, which includes a pool-level migrate_lock, and a lock for each size class. We have to obtain both locks in the hotpath in most cases anyway, except for zs_malloc. This exception will no longer exist when we introduce a LRU into the zs_pool for the new writeback functionality - we will need to obtain a pool-level lock to synchronize LRU handling even in zs_malloc. In preparation for zsmalloc writeback, consolidate these locks into a single pool-level lock, which drastically reduces the complexity of synchronization in zsmalloc. We have also benchmarked the lock consolidation to see the performance effect of this change on zram. First, we ran a synthetic FS workload on a server machine with 36 cores (same machine for all runs), using fs_mark -d ../zram1mnt -s 100000 -n 2500 -t 32 -k before and after for btrfs and ext4 on zram (FS usage is 80%). Here is the result (unit is file/second): With lock consolidation (btrfs): Average: 13520.2, Median: 13531.0, Stddev: 137.5961482019028 Without lock consolidation (btrfs): Average: 13487.2, Median: 13575.0, Stddev: 309.08283679298665 With lock consolidation (ext4): Average: 16824.4, Median: 16839.0, Stddev: 89.97388510006668 Without lock consolidation (ext4) Average: 16958.0, Median: 16986.0, Stddev: 194.7370021336469 As you can see, we observe a 0.3% regression for btrfs, and a 0.9% regression for ext4. This is a small, barely measurable difference in my opinion. For a more realistic scenario, we also tries building the kernel on zram. Here is the time it takes (in seconds): With lock consolidation (btrfs): real Average: 319.6, Median: 320.0, Stddev: 0.8944271909999159 user Average: 6894.2, Median: 6895.0, Stddev: 25.528415540334656 sys Average: 521.4, Median: 522.0, Stddev: 1.51657508881031 Without lock consolidation (btrfs): real Average: 319.8, Median: 320.0, Stddev: 0.8366600265340756 user Average: 6896.6, Median: 6899.0, Stddev: 16.04057355583023 sys Average: 520.6, Median: 521.0, Stddev: 1.140175425099138 With lock consolidation (ext4): real Average: 320.0, Median: 319.0, Stddev: 1.4142135623730951 user Average: 6896.8, Median: 6878.0, Stddev: 28.621670111997307 sys Average: 521.2, Median: 521.0, Stddev: 1.7888543819998317 Without lock consolidation (ext4) real Average: 319.6, Median: 319.0, Stddev: 0.8944271909999159 user Average: 6886.2, Median: 6887.0, Stddev: 16.93221781102523 sys Average: 520.4, Median: 520.0, Stddev: 1.140175425099138 The difference is entirely within the noise of a typical run on zram. This hardly justifies the complexity of maintaining both the pool lock and the class lock. In fact, for writeback, we would need to introduce yet another lock to prevent data races on the pool's LRU, further complicating the lock handling logic. IMHO, it is just better to collapse all of these into a single pool-level lock. Link: https://lkml.kernel.org/r/20221128191616.1261026-4-nphamcs@gmail.com Signed-off-by: Nhat Pham <nphamcs@gmail.com> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Acked-by: Minchan Kim <minchan@kernel.org> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-11-28 19:16:12 +00:00
spin_lock(&pool->lock);
seq_printf(s, " %5u %5u ", i, class->size);
for (fg = ZS_INUSE_RATIO_10; fg < NR_FULLNESS_GROUPS; fg++) {
inuse_totals[fg] += zs_stat_get(class, fg);
seq_printf(s, "%9lu ", zs_stat_get(class, fg));
}
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
obj_allocated = zs_stat_get(class, ZS_OBJS_ALLOCATED);
obj_used = zs_stat_get(class, ZS_OBJS_INUSE);
freeable = zs_can_compact(class);
zsmalloc: consolidate zs_pool's migrate_lock and size_class's locks Currently, zsmalloc has a hierarchy of locks, which includes a pool-level migrate_lock, and a lock for each size class. We have to obtain both locks in the hotpath in most cases anyway, except for zs_malloc. This exception will no longer exist when we introduce a LRU into the zs_pool for the new writeback functionality - we will need to obtain a pool-level lock to synchronize LRU handling even in zs_malloc. In preparation for zsmalloc writeback, consolidate these locks into a single pool-level lock, which drastically reduces the complexity of synchronization in zsmalloc. We have also benchmarked the lock consolidation to see the performance effect of this change on zram. First, we ran a synthetic FS workload on a server machine with 36 cores (same machine for all runs), using fs_mark -d ../zram1mnt -s 100000 -n 2500 -t 32 -k before and after for btrfs and ext4 on zram (FS usage is 80%). Here is the result (unit is file/second): With lock consolidation (btrfs): Average: 13520.2, Median: 13531.0, Stddev: 137.5961482019028 Without lock consolidation (btrfs): Average: 13487.2, Median: 13575.0, Stddev: 309.08283679298665 With lock consolidation (ext4): Average: 16824.4, Median: 16839.0, Stddev: 89.97388510006668 Without lock consolidation (ext4) Average: 16958.0, Median: 16986.0, Stddev: 194.7370021336469 As you can see, we observe a 0.3% regression for btrfs, and a 0.9% regression for ext4. This is a small, barely measurable difference in my opinion. For a more realistic scenario, we also tries building the kernel on zram. Here is the time it takes (in seconds): With lock consolidation (btrfs): real Average: 319.6, Median: 320.0, Stddev: 0.8944271909999159 user Average: 6894.2, Median: 6895.0, Stddev: 25.528415540334656 sys Average: 521.4, Median: 522.0, Stddev: 1.51657508881031 Without lock consolidation (btrfs): real Average: 319.8, Median: 320.0, Stddev: 0.8366600265340756 user Average: 6896.6, Median: 6899.0, Stddev: 16.04057355583023 sys Average: 520.6, Median: 521.0, Stddev: 1.140175425099138 With lock consolidation (ext4): real Average: 320.0, Median: 319.0, Stddev: 1.4142135623730951 user Average: 6896.8, Median: 6878.0, Stddev: 28.621670111997307 sys Average: 521.2, Median: 521.0, Stddev: 1.7888543819998317 Without lock consolidation (ext4) real Average: 319.6, Median: 319.0, Stddev: 0.8944271909999159 user Average: 6886.2, Median: 6887.0, Stddev: 16.93221781102523 sys Average: 520.4, Median: 520.0, Stddev: 1.140175425099138 The difference is entirely within the noise of a typical run on zram. This hardly justifies the complexity of maintaining both the pool lock and the class lock. In fact, for writeback, we would need to introduce yet another lock to prevent data races on the pool's LRU, further complicating the lock handling logic. IMHO, it is just better to collapse all of these into a single pool-level lock. Link: https://lkml.kernel.org/r/20221128191616.1261026-4-nphamcs@gmail.com Signed-off-by: Nhat Pham <nphamcs@gmail.com> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Acked-by: Minchan Kim <minchan@kernel.org> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-11-28 19:16:12 +00:00
spin_unlock(&pool->lock);
objs_per_zspage = class->objs_per_zspage;
pages_used = obj_allocated / objs_per_zspage *
class->pages_per_zspage;
seq_printf(s, "%13lu %10lu %10lu %16d %8lu\n",
obj_allocated, obj_used, pages_used,
class->pages_per_zspage, freeable);
total_objs += obj_allocated;
total_used_objs += obj_used;
total_pages += pages_used;
total_freeable += freeable;
}
seq_puts(s, "\n");
seq_printf(s, " %5s %5s ", "Total", "");
for (fg = ZS_INUSE_RATIO_10; fg < NR_FULLNESS_GROUPS; fg++)
seq_printf(s, "%9lu ", inuse_totals[fg]);
seq_printf(s, "%13lu %10lu %10lu %16s %8lu\n",
total_objs, total_used_objs, total_pages, "",
total_freeable);
return 0;
}
DEFINE_SHOW_ATTRIBUTE(zs_stats_size);
static void zs_pool_stat_create(struct zs_pool *pool, const char *name)
{
if (!zs_stat_root) {
pr_warn("no root stat dir, not creating <%s> stat dir\n", name);
return;
}
pool->stat_dentry = debugfs_create_dir(name, zs_stat_root);
debugfs_create_file("classes", S_IFREG | 0444, pool->stat_dentry, pool,
&zs_stats_size_fops);
}
static void zs_pool_stat_destroy(struct zs_pool *pool)
{
debugfs_remove_recursive(pool->stat_dentry);
}
#else /* CONFIG_ZSMALLOC_STAT */
static void __init zs_stat_init(void)
{
}
static void __exit zs_stat_exit(void)
{
}
static inline void zs_pool_stat_create(struct zs_pool *pool, const char *name)
{
}
static inline void zs_pool_stat_destroy(struct zs_pool *pool)
{
}
#endif
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
/*
* For each size class, zspages are divided into different groups
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
* depending on their usage ratio. This function returns fullness
* status of the given page.
*/
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
static int get_fullness_group(struct size_class *class, struct zspage *zspage)
{
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
int inuse, objs_per_zspage, ratio;
inuse = get_zspage_inuse(zspage);
objs_per_zspage = class->objs_per_zspage;
if (inuse == 0)
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
return ZS_INUSE_RATIO_0;
if (inuse == objs_per_zspage)
return ZS_INUSE_RATIO_100;
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
ratio = 100 * inuse / objs_per_zspage;
/*
* Take integer division into consideration: a page with one inuse
* object out of 127 possible, will end up having 0 usage ratio,
* which is wrong as it belongs in ZS_INUSE_RATIO_10 fullness group.
*/
return ratio / 10 + 1;
}
/*
* Each size class maintains various freelists and zspages are assigned
* to one of these freelists based on the number of live objects they
* have. This functions inserts the given zspage into the freelist
* identified by <class, fullness_group>.
*/
static void insert_zspage(struct size_class *class,
struct zspage *zspage,
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
int fullness)
{
class_stat_inc(class, fullness, 1);
zsmalloc: remove insert_zspage() ->inuse optimization Patch series "zsmalloc: fine-grained fullness and new compaction algorithm", v4. Existing zsmalloc page fullness grouping leads to suboptimal page selection for both zs_malloc() and zs_compact(). This patchset reworks zsmalloc fullness grouping/classification. Additinally it also implements new compaction algorithm that is expected to use less CPU-cycles (as it potentially does fewer memcpy-s in zs_object_copy()). Test (synthetic) results can be seen in patch 0003. This patch (of 4): This optimization has no effect. It only ensures that when a zspage was added to its corresponding fullness list, its "inuse" counter was higher or lower than the "inuse" counter of the zspage at the head of the list. The intention was to keep busy zspages at the head, so they could be filled up and moved to the ZS_FULL fullness group more quickly. However, this doesn't work as the "inuse" counter of a zspage can be modified by obj_free() but the zspage may still belong to the same fullness list. So, fix_fullness_group() won't change the zspage's position in relation to the head's "inuse" counter, leading to a largely random order of zspages within the fullness list. For instance, consider a printout of the "inuse" counters of the first 10 zspages in a class that holds 93 objects per zspage: ZS_ALMOST_EMPTY: 36 67 68 64 35 54 63 52 As we can see the zspage with the lowest "inuse" counter is actually the head of the fullness list. Remove this pointless "optimisation". Link: https://lkml.kernel.org/r/20230304034835.2082479-1-senozhatsky@chromium.org Link: https://lkml.kernel.org/r/20230304034835.2082479-2-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:32 +00:00
list_add(&zspage->list, &class->fullness_list[fullness]);
}
/*
* This function removes the given zspage from the freelist identified
* by <class, fullness_group>.
*/
static void remove_zspage(struct size_class *class,
struct zspage *zspage,
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
int fullness)
{
VM_BUG_ON(list_empty(&class->fullness_list[fullness]));
list_del_init(&zspage->list);
class_stat_dec(class, fullness, 1);
}
/*
* Each size class maintains zspages in different fullness groups depending
* on the number of live objects they contain. When allocating or freeing
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
* objects, the fullness status of the page can change, for instance, from
* INUSE_RATIO_80 to INUSE_RATIO_70 when freeing an object. This function
* checks if such a status change has occurred for the given page and
* accordingly moves the page from the list of the old fullness group to that
* of the new fullness group.
*/
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
static int fix_fullness_group(struct size_class *class, struct zspage *zspage)
{
int class_idx;
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
int currfg, newfg;
get_zspage_mapping(zspage, &class_idx, &currfg);
newfg = get_fullness_group(class, zspage);
if (newfg == currfg)
goto out;
remove_zspage(class, zspage, currfg);
insert_zspage(class, zspage, newfg);
set_zspage_mapping(zspage, class_idx, newfg);
out:
return newfg;
}
static struct zspage *get_zspage(struct page *page)
{
struct zspage *zspage = (struct zspage *)page_private(page);
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
BUG_ON(zspage->magic != ZSPAGE_MAGIC);
return zspage;
}
static struct page *get_next_page(struct page *page)
{
struct zspage *zspage = get_zspage(page);
if (unlikely(ZsHugePage(zspage)))
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
return NULL;
return (struct page *)page->index;
}
/**
* obj_to_location - get (<page>, <obj_idx>) from encoded object value
* @obj: the encoded object value
* @page: page object resides in zspage
* @obj_idx: object index
*/
static void obj_to_location(unsigned long obj, struct page **page,
unsigned int *obj_idx)
{
obj >>= OBJ_TAG_BITS;
*page = pfn_to_page(obj >> OBJ_INDEX_BITS);
*obj_idx = (obj & OBJ_INDEX_MASK);
}
zsmalloc: introduce some helper functions Patch series "zsmalloc: remove bit_spin_lock", v2. zsmalloc uses bit_spin_lock to minimize space overhead since it's zpage granularity lock. However, it causes zsmalloc non-working under PREEMPT_RT as well as adding too much complication. This patchset tries to replace the bit_spin_lock with per-pool rwlock. It also removes unnecessary zspage isolation logic from class, which was the other part too much complication added into zsmalloc. Last patch changes the get_cpu_var to local_lock to make it work in PREEMPT_RT. This patch (of 9): get_zspage_mapping returns fullness as well as class_idx. However, the fullness is usually not used since it could be stale in some contexts. It causes misleading as well as unnecessary instructions so this patch introduces zspage_class. obj_to_location also produces page and index but we don't need always the index, either so this patch introduces obj_to_page. Link: https://lkml.kernel.org/r/20211115185909.3949505-1-minchan@kernel.org Link: https://lkml.kernel.org/r/20211115185909.3949505-2-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:13:51 +00:00
static void obj_to_page(unsigned long obj, struct page **page)
{
obj >>= OBJ_TAG_BITS;
*page = pfn_to_page(obj >> OBJ_INDEX_BITS);
}
/**
* location_to_obj - get obj value encoded from (<page>, <obj_idx>)
* @page: page object resides in zspage
* @obj_idx: object index
*/
static unsigned long location_to_obj(struct page *page, unsigned int obj_idx)
{
unsigned long obj;
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
obj = page_to_pfn(page) << OBJ_INDEX_BITS;
obj |= obj_idx & OBJ_INDEX_MASK;
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
obj <<= OBJ_TAG_BITS;
return obj;
}
zsmalloc: decouple handle and object Recently, we started to use zram heavily and some of issues popped. 1) external fragmentation I got a report from Juneho Choi that fork failed although there are plenty of free pages in the system. His investigation revealed zram is one of the culprit to make heavy fragmentation so there was no more contiguous 16K page for pgd to fork in the ARM. 2) non-movable pages Other problem of zram now is that inherently, user want to use zram as swap in small memory system so they use zRAM with CMA to use memory efficiently. However, unfortunately, it doesn't work well because zRAM cannot use CMA's movable pages unless it doesn't support compaction. I got several reports about that OOM happened with zram although there are lots of swap space and free space in CMA area. 3) internal fragmentation zRAM has started support memory limitation feature to limit memory usage and I sent a patchset(https://lkml.org/lkml/2014/9/21/148) for VM to be harmonized with zram-swap to stop anonymous page reclaim if zram consumed memory up to the limit although there are free space on the swap. One problem for that direction is zram has no way to know any hole in memory space zsmalloc allocated by internal fragmentation so zram would regard swap is full although there are free space in zsmalloc. For solving the issue, zram want to trigger compaction of zsmalloc before it decides full or not. This patchset is first step to support above issues. For that, it adds indirect layer between handle and object location and supports manual compaction to solve 3th problem first of all. After this patchset got merged, next step is to make VM aware of zsmalloc compaction so that generic compaction will move zsmalloced-pages automatically in runtime. In my imaginary experiment(ie, high compress ratio data with heavy swap in/out on 8G zram-swap), data is as follows, Before = zram allocated object : 60212066 bytes zram total used: 140103680 bytes ratio: 42.98 percent MemFree: 840192 kB Compaction After = frag ratio after compaction zram allocated object : 60212066 bytes zram total used: 76185600 bytes ratio: 79.03 percent MemFree: 901932 kB Juneho reported below in his real platform with small aging. So, I think the benefit would be bigger in real aging system for a long time. - frag_ratio increased 3% (ie, higher is better) - memfree increased about 6MB - In buddy info, Normal 2^3: 4, 2^2: 1: 2^1 increased, Highmem: 2^1 21 increased frag ratio after swap fragment used : 156677 kbytes total: 166092 kbytes frag_ratio : 94 meminfo before compaction MemFree: 83724 kB Node 0, zone Normal 13642 1364 57 10 61 17 9 5 4 0 0 Node 0, zone HighMem 425 29 1 0 0 0 0 0 0 0 0 num_migrated : 23630 compaction done frag ratio after compaction used : 156673 kbytes total: 160564 kbytes frag_ratio : 97 meminfo after compaction MemFree: 89060 kB Node 0, zone Normal 14076 1544 67 14 61 17 9 5 4 0 0 Node 0, zone HighMem 863 50 1 0 0 0 0 0 0 0 0 This patchset adds more logics(about 480 lines) in zsmalloc but when I tested heavy swapin/out program, the regression for swapin/out speed is marginal because most of overheads were caused by compress/decompress and other MM reclaim stuff. This patch (of 7): Currently, handle of zsmalloc encodes object's location directly so it makes support of migration hard. This patch decouples handle and object via adding indirect layer. For that, it allocates handle dynamically and returns it to user. The handle is the address allocated by slab allocation so it's unique and we could keep object's location in the memory space allocated for handle. With it, we can change object's position without changing handle itself. Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:23 +00:00
static unsigned long handle_to_obj(unsigned long handle)
{
return *(unsigned long *)handle;
}
zsmalloc: fix a race with deferred_handles storing Currently, there is a race between zs_free() and zs_reclaim_page(): zs_reclaim_page() finds a handle to an allocated object, but before the eviction happens, an independent zs_free() call to the same handle could come in and overwrite the object value stored at the handle with the last deferred handle. When zs_reclaim_page() finally gets to call the eviction handler, it will see an invalid object value (i.e the previous deferred handle instead of the original object value). This race happens quite infrequently. We only managed to produce it with out-of-tree developmental code that triggers zsmalloc writeback with a much higher frequency than usual. This patch fixes this race by storing the deferred handle in the object header instead. We differentiate the deferred handle from the other two cases (handle for allocated object, and linkage for free object) with a new tag. If zspage reclamation succeeds, we will free these deferred handles by walking through the zspage objects. On the other hand, if zspage reclamation fails, we reconstruct the zspage freelist (with the deferred handle tag and allocated tag) before trying again with the reclamation. [arnd@arndb.de: avoid unused-function warning] Link: https://lkml.kernel.org/r/20230117170507.2651972-1-arnd@kernel.org Link: https://lkml.kernel.org/r/20230110231701.326724-1-nphamcs@gmail.com Fixes: 9997bc017549 ("zsmalloc: implement writeback mechanism for zsmalloc") Signed-off-by: Nhat Pham <nphamcs@gmail.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-01-10 23:17:01 +00:00
static bool obj_tagged(struct page *page, void *obj, unsigned long *phandle,
int tag)
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
{
unsigned long handle;
struct zspage *zspage = get_zspage(page);
if (unlikely(ZsHugePage(zspage))) {
VM_BUG_ON_PAGE(!is_first_page(page), page);
handle = page->index;
} else
handle = *(unsigned long *)obj;
zsmalloc: fix a race with deferred_handles storing Currently, there is a race between zs_free() and zs_reclaim_page(): zs_reclaim_page() finds a handle to an allocated object, but before the eviction happens, an independent zs_free() call to the same handle could come in and overwrite the object value stored at the handle with the last deferred handle. When zs_reclaim_page() finally gets to call the eviction handler, it will see an invalid object value (i.e the previous deferred handle instead of the original object value). This race happens quite infrequently. We only managed to produce it with out-of-tree developmental code that triggers zsmalloc writeback with a much higher frequency than usual. This patch fixes this race by storing the deferred handle in the object header instead. We differentiate the deferred handle from the other two cases (handle for allocated object, and linkage for free object) with a new tag. If zspage reclamation succeeds, we will free these deferred handles by walking through the zspage objects. On the other hand, if zspage reclamation fails, we reconstruct the zspage freelist (with the deferred handle tag and allocated tag) before trying again with the reclamation. [arnd@arndb.de: avoid unused-function warning] Link: https://lkml.kernel.org/r/20230117170507.2651972-1-arnd@kernel.org Link: https://lkml.kernel.org/r/20230110231701.326724-1-nphamcs@gmail.com Fixes: 9997bc017549 ("zsmalloc: implement writeback mechanism for zsmalloc") Signed-off-by: Nhat Pham <nphamcs@gmail.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-01-10 23:17:01 +00:00
if (!(handle & tag))
return false;
zsmalloc: fix a race with deferred_handles storing Currently, there is a race between zs_free() and zs_reclaim_page(): zs_reclaim_page() finds a handle to an allocated object, but before the eviction happens, an independent zs_free() call to the same handle could come in and overwrite the object value stored at the handle with the last deferred handle. When zs_reclaim_page() finally gets to call the eviction handler, it will see an invalid object value (i.e the previous deferred handle instead of the original object value). This race happens quite infrequently. We only managed to produce it with out-of-tree developmental code that triggers zsmalloc writeback with a much higher frequency than usual. This patch fixes this race by storing the deferred handle in the object header instead. We differentiate the deferred handle from the other two cases (handle for allocated object, and linkage for free object) with a new tag. If zspage reclamation succeeds, we will free these deferred handles by walking through the zspage objects. On the other hand, if zspage reclamation fails, we reconstruct the zspage freelist (with the deferred handle tag and allocated tag) before trying again with the reclamation. [arnd@arndb.de: avoid unused-function warning] Link: https://lkml.kernel.org/r/20230117170507.2651972-1-arnd@kernel.org Link: https://lkml.kernel.org/r/20230110231701.326724-1-nphamcs@gmail.com Fixes: 9997bc017549 ("zsmalloc: implement writeback mechanism for zsmalloc") Signed-off-by: Nhat Pham <nphamcs@gmail.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-01-10 23:17:01 +00:00
/* Clear all tags before returning the handle */
*phandle = handle & ~OBJ_TAG_MASK;
return true;
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
}
zsmalloc: fix a race with deferred_handles storing Currently, there is a race between zs_free() and zs_reclaim_page(): zs_reclaim_page() finds a handle to an allocated object, but before the eviction happens, an independent zs_free() call to the same handle could come in and overwrite the object value stored at the handle with the last deferred handle. When zs_reclaim_page() finally gets to call the eviction handler, it will see an invalid object value (i.e the previous deferred handle instead of the original object value). This race happens quite infrequently. We only managed to produce it with out-of-tree developmental code that triggers zsmalloc writeback with a much higher frequency than usual. This patch fixes this race by storing the deferred handle in the object header instead. We differentiate the deferred handle from the other two cases (handle for allocated object, and linkage for free object) with a new tag. If zspage reclamation succeeds, we will free these deferred handles by walking through the zspage objects. On the other hand, if zspage reclamation fails, we reconstruct the zspage freelist (with the deferred handle tag and allocated tag) before trying again with the reclamation. [arnd@arndb.de: avoid unused-function warning] Link: https://lkml.kernel.org/r/20230117170507.2651972-1-arnd@kernel.org Link: https://lkml.kernel.org/r/20230110231701.326724-1-nphamcs@gmail.com Fixes: 9997bc017549 ("zsmalloc: implement writeback mechanism for zsmalloc") Signed-off-by: Nhat Pham <nphamcs@gmail.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-01-10 23:17:01 +00:00
static inline bool obj_allocated(struct page *page, void *obj, unsigned long *phandle)
{
return obj_tagged(page, obj, phandle, OBJ_ALLOCATED_TAG);
}
#ifdef CONFIG_ZPOOL
static bool obj_stores_deferred_handle(struct page *page, void *obj,
unsigned long *phandle)
{
return obj_tagged(page, obj, phandle, OBJ_DEFERRED_HANDLE_TAG);
}
#endif
static void reset_page(struct page *page)
{
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
__ClearPageMovable(page);
ClearPagePrivate(page);
set_page_private(page, 0);
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
page_mapcount_reset(page);
page->index = 0;
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
}
static int trylock_zspage(struct zspage *zspage)
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
{
struct page *cursor, *fail;
for (cursor = get_first_page(zspage); cursor != NULL; cursor =
get_next_page(cursor)) {
if (!trylock_page(cursor)) {
fail = cursor;
goto unlock;
}
}
return 1;
unlock:
for (cursor = get_first_page(zspage); cursor != fail; cursor =
get_next_page(cursor))
unlock_page(cursor);
return 0;
}
#ifdef CONFIG_ZPOOL
zsmalloc: fix a race with deferred_handles storing Currently, there is a race between zs_free() and zs_reclaim_page(): zs_reclaim_page() finds a handle to an allocated object, but before the eviction happens, an independent zs_free() call to the same handle could come in and overwrite the object value stored at the handle with the last deferred handle. When zs_reclaim_page() finally gets to call the eviction handler, it will see an invalid object value (i.e the previous deferred handle instead of the original object value). This race happens quite infrequently. We only managed to produce it with out-of-tree developmental code that triggers zsmalloc writeback with a much higher frequency than usual. This patch fixes this race by storing the deferred handle in the object header instead. We differentiate the deferred handle from the other two cases (handle for allocated object, and linkage for free object) with a new tag. If zspage reclamation succeeds, we will free these deferred handles by walking through the zspage objects. On the other hand, if zspage reclamation fails, we reconstruct the zspage freelist (with the deferred handle tag and allocated tag) before trying again with the reclamation. [arnd@arndb.de: avoid unused-function warning] Link: https://lkml.kernel.org/r/20230117170507.2651972-1-arnd@kernel.org Link: https://lkml.kernel.org/r/20230110231701.326724-1-nphamcs@gmail.com Fixes: 9997bc017549 ("zsmalloc: implement writeback mechanism for zsmalloc") Signed-off-by: Nhat Pham <nphamcs@gmail.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-01-10 23:17:01 +00:00
static unsigned long find_deferred_handle_obj(struct size_class *class,
struct page *page, int *obj_idx);
/*
* Free all the deferred handles whose objects are freed in zs_free.
*/
zsmalloc: fix a race with deferred_handles storing Currently, there is a race between zs_free() and zs_reclaim_page(): zs_reclaim_page() finds a handle to an allocated object, but before the eviction happens, an independent zs_free() call to the same handle could come in and overwrite the object value stored at the handle with the last deferred handle. When zs_reclaim_page() finally gets to call the eviction handler, it will see an invalid object value (i.e the previous deferred handle instead of the original object value). This race happens quite infrequently. We only managed to produce it with out-of-tree developmental code that triggers zsmalloc writeback with a much higher frequency than usual. This patch fixes this race by storing the deferred handle in the object header instead. We differentiate the deferred handle from the other two cases (handle for allocated object, and linkage for free object) with a new tag. If zspage reclamation succeeds, we will free these deferred handles by walking through the zspage objects. On the other hand, if zspage reclamation fails, we reconstruct the zspage freelist (with the deferred handle tag and allocated tag) before trying again with the reclamation. [arnd@arndb.de: avoid unused-function warning] Link: https://lkml.kernel.org/r/20230117170507.2651972-1-arnd@kernel.org Link: https://lkml.kernel.org/r/20230110231701.326724-1-nphamcs@gmail.com Fixes: 9997bc017549 ("zsmalloc: implement writeback mechanism for zsmalloc") Signed-off-by: Nhat Pham <nphamcs@gmail.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-01-10 23:17:01 +00:00
static void free_handles(struct zs_pool *pool, struct size_class *class,
struct zspage *zspage)
{
zsmalloc: fix a race with deferred_handles storing Currently, there is a race between zs_free() and zs_reclaim_page(): zs_reclaim_page() finds a handle to an allocated object, but before the eviction happens, an independent zs_free() call to the same handle could come in and overwrite the object value stored at the handle with the last deferred handle. When zs_reclaim_page() finally gets to call the eviction handler, it will see an invalid object value (i.e the previous deferred handle instead of the original object value). This race happens quite infrequently. We only managed to produce it with out-of-tree developmental code that triggers zsmalloc writeback with a much higher frequency than usual. This patch fixes this race by storing the deferred handle in the object header instead. We differentiate the deferred handle from the other two cases (handle for allocated object, and linkage for free object) with a new tag. If zspage reclamation succeeds, we will free these deferred handles by walking through the zspage objects. On the other hand, if zspage reclamation fails, we reconstruct the zspage freelist (with the deferred handle tag and allocated tag) before trying again with the reclamation. [arnd@arndb.de: avoid unused-function warning] Link: https://lkml.kernel.org/r/20230117170507.2651972-1-arnd@kernel.org Link: https://lkml.kernel.org/r/20230110231701.326724-1-nphamcs@gmail.com Fixes: 9997bc017549 ("zsmalloc: implement writeback mechanism for zsmalloc") Signed-off-by: Nhat Pham <nphamcs@gmail.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-01-10 23:17:01 +00:00
int obj_idx = 0;
struct page *page = get_first_page(zspage);
unsigned long handle;
zsmalloc: fix a race with deferred_handles storing Currently, there is a race between zs_free() and zs_reclaim_page(): zs_reclaim_page() finds a handle to an allocated object, but before the eviction happens, an independent zs_free() call to the same handle could come in and overwrite the object value stored at the handle with the last deferred handle. When zs_reclaim_page() finally gets to call the eviction handler, it will see an invalid object value (i.e the previous deferred handle instead of the original object value). This race happens quite infrequently. We only managed to produce it with out-of-tree developmental code that triggers zsmalloc writeback with a much higher frequency than usual. This patch fixes this race by storing the deferred handle in the object header instead. We differentiate the deferred handle from the other two cases (handle for allocated object, and linkage for free object) with a new tag. If zspage reclamation succeeds, we will free these deferred handles by walking through the zspage objects. On the other hand, if zspage reclamation fails, we reconstruct the zspage freelist (with the deferred handle tag and allocated tag) before trying again with the reclamation. [arnd@arndb.de: avoid unused-function warning] Link: https://lkml.kernel.org/r/20230117170507.2651972-1-arnd@kernel.org Link: https://lkml.kernel.org/r/20230110231701.326724-1-nphamcs@gmail.com Fixes: 9997bc017549 ("zsmalloc: implement writeback mechanism for zsmalloc") Signed-off-by: Nhat Pham <nphamcs@gmail.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-01-10 23:17:01 +00:00
while (1) {
handle = find_deferred_handle_obj(class, page, &obj_idx);
if (!handle) {
page = get_next_page(page);
if (!page)
break;
obj_idx = 0;
continue;
}
cache_free_handle(pool, handle);
zsmalloc: fix a race with deferred_handles storing Currently, there is a race between zs_free() and zs_reclaim_page(): zs_reclaim_page() finds a handle to an allocated object, but before the eviction happens, an independent zs_free() call to the same handle could come in and overwrite the object value stored at the handle with the last deferred handle. When zs_reclaim_page() finally gets to call the eviction handler, it will see an invalid object value (i.e the previous deferred handle instead of the original object value). This race happens quite infrequently. We only managed to produce it with out-of-tree developmental code that triggers zsmalloc writeback with a much higher frequency than usual. This patch fixes this race by storing the deferred handle in the object header instead. We differentiate the deferred handle from the other two cases (handle for allocated object, and linkage for free object) with a new tag. If zspage reclamation succeeds, we will free these deferred handles by walking through the zspage objects. On the other hand, if zspage reclamation fails, we reconstruct the zspage freelist (with the deferred handle tag and allocated tag) before trying again with the reclamation. [arnd@arndb.de: avoid unused-function warning] Link: https://lkml.kernel.org/r/20230117170507.2651972-1-arnd@kernel.org Link: https://lkml.kernel.org/r/20230110231701.326724-1-nphamcs@gmail.com Fixes: 9997bc017549 ("zsmalloc: implement writeback mechanism for zsmalloc") Signed-off-by: Nhat Pham <nphamcs@gmail.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-01-10 23:17:01 +00:00
obj_idx++;
}
}
#else
zsmalloc: fix a race with deferred_handles storing Currently, there is a race between zs_free() and zs_reclaim_page(): zs_reclaim_page() finds a handle to an allocated object, but before the eviction happens, an independent zs_free() call to the same handle could come in and overwrite the object value stored at the handle with the last deferred handle. When zs_reclaim_page() finally gets to call the eviction handler, it will see an invalid object value (i.e the previous deferred handle instead of the original object value). This race happens quite infrequently. We only managed to produce it with out-of-tree developmental code that triggers zsmalloc writeback with a much higher frequency than usual. This patch fixes this race by storing the deferred handle in the object header instead. We differentiate the deferred handle from the other two cases (handle for allocated object, and linkage for free object) with a new tag. If zspage reclamation succeeds, we will free these deferred handles by walking through the zspage objects. On the other hand, if zspage reclamation fails, we reconstruct the zspage freelist (with the deferred handle tag and allocated tag) before trying again with the reclamation. [arnd@arndb.de: avoid unused-function warning] Link: https://lkml.kernel.org/r/20230117170507.2651972-1-arnd@kernel.org Link: https://lkml.kernel.org/r/20230110231701.326724-1-nphamcs@gmail.com Fixes: 9997bc017549 ("zsmalloc: implement writeback mechanism for zsmalloc") Signed-off-by: Nhat Pham <nphamcs@gmail.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-01-10 23:17:01 +00:00
static inline void free_handles(struct zs_pool *pool, struct size_class *class,
struct zspage *zspage) {}
#endif
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
static void __free_zspage(struct zs_pool *pool, struct size_class *class,
struct zspage *zspage)
{
struct page *page, *next;
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
int fg;
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
unsigned int class_idx;
get_zspage_mapping(zspage, &class_idx, &fg);
zsmalloc: consolidate zs_pool's migrate_lock and size_class's locks Currently, zsmalloc has a hierarchy of locks, which includes a pool-level migrate_lock, and a lock for each size class. We have to obtain both locks in the hotpath in most cases anyway, except for zs_malloc. This exception will no longer exist when we introduce a LRU into the zs_pool for the new writeback functionality - we will need to obtain a pool-level lock to synchronize LRU handling even in zs_malloc. In preparation for zsmalloc writeback, consolidate these locks into a single pool-level lock, which drastically reduces the complexity of synchronization in zsmalloc. We have also benchmarked the lock consolidation to see the performance effect of this change on zram. First, we ran a synthetic FS workload on a server machine with 36 cores (same machine for all runs), using fs_mark -d ../zram1mnt -s 100000 -n 2500 -t 32 -k before and after for btrfs and ext4 on zram (FS usage is 80%). Here is the result (unit is file/second): With lock consolidation (btrfs): Average: 13520.2, Median: 13531.0, Stddev: 137.5961482019028 Without lock consolidation (btrfs): Average: 13487.2, Median: 13575.0, Stddev: 309.08283679298665 With lock consolidation (ext4): Average: 16824.4, Median: 16839.0, Stddev: 89.97388510006668 Without lock consolidation (ext4) Average: 16958.0, Median: 16986.0, Stddev: 194.7370021336469 As you can see, we observe a 0.3% regression for btrfs, and a 0.9% regression for ext4. This is a small, barely measurable difference in my opinion. For a more realistic scenario, we also tries building the kernel on zram. Here is the time it takes (in seconds): With lock consolidation (btrfs): real Average: 319.6, Median: 320.0, Stddev: 0.8944271909999159 user Average: 6894.2, Median: 6895.0, Stddev: 25.528415540334656 sys Average: 521.4, Median: 522.0, Stddev: 1.51657508881031 Without lock consolidation (btrfs): real Average: 319.8, Median: 320.0, Stddev: 0.8366600265340756 user Average: 6896.6, Median: 6899.0, Stddev: 16.04057355583023 sys Average: 520.6, Median: 521.0, Stddev: 1.140175425099138 With lock consolidation (ext4): real Average: 320.0, Median: 319.0, Stddev: 1.4142135623730951 user Average: 6896.8, Median: 6878.0, Stddev: 28.621670111997307 sys Average: 521.2, Median: 521.0, Stddev: 1.7888543819998317 Without lock consolidation (ext4) real Average: 319.6, Median: 319.0, Stddev: 0.8944271909999159 user Average: 6886.2, Median: 6887.0, Stddev: 16.93221781102523 sys Average: 520.4, Median: 520.0, Stddev: 1.140175425099138 The difference is entirely within the noise of a typical run on zram. This hardly justifies the complexity of maintaining both the pool lock and the class lock. In fact, for writeback, we would need to introduce yet another lock to prevent data races on the pool's LRU, further complicating the lock handling logic. IMHO, it is just better to collapse all of these into a single pool-level lock. Link: https://lkml.kernel.org/r/20221128191616.1261026-4-nphamcs@gmail.com Signed-off-by: Nhat Pham <nphamcs@gmail.com> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Acked-by: Minchan Kim <minchan@kernel.org> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-11-28 19:16:12 +00:00
assert_spin_locked(&pool->lock);
VM_BUG_ON(get_zspage_inuse(zspage));
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
VM_BUG_ON(fg != ZS_INUSE_RATIO_0);
/* Free all deferred handles from zs_free */
zsmalloc: fix a race with deferred_handles storing Currently, there is a race between zs_free() and zs_reclaim_page(): zs_reclaim_page() finds a handle to an allocated object, but before the eviction happens, an independent zs_free() call to the same handle could come in and overwrite the object value stored at the handle with the last deferred handle. When zs_reclaim_page() finally gets to call the eviction handler, it will see an invalid object value (i.e the previous deferred handle instead of the original object value). This race happens quite infrequently. We only managed to produce it with out-of-tree developmental code that triggers zsmalloc writeback with a much higher frequency than usual. This patch fixes this race by storing the deferred handle in the object header instead. We differentiate the deferred handle from the other two cases (handle for allocated object, and linkage for free object) with a new tag. If zspage reclamation succeeds, we will free these deferred handles by walking through the zspage objects. On the other hand, if zspage reclamation fails, we reconstruct the zspage freelist (with the deferred handle tag and allocated tag) before trying again with the reclamation. [arnd@arndb.de: avoid unused-function warning] Link: https://lkml.kernel.org/r/20230117170507.2651972-1-arnd@kernel.org Link: https://lkml.kernel.org/r/20230110231701.326724-1-nphamcs@gmail.com Fixes: 9997bc017549 ("zsmalloc: implement writeback mechanism for zsmalloc") Signed-off-by: Nhat Pham <nphamcs@gmail.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-01-10 23:17:01 +00:00
free_handles(pool, class, zspage);
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
next = page = get_first_page(zspage);
do {
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
VM_BUG_ON_PAGE(!PageLocked(page), page);
next = get_next_page(page);
reset_page(page);
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
unlock_page(page);
dec_zone_page_state(page, NR_ZSPAGES);
put_page(page);
page = next;
} while (page != NULL);
cache_free_zspage(pool, zspage);
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
class_stat_dec(class, ZS_OBJS_ALLOCATED, class->objs_per_zspage);
atomic_long_sub(class->pages_per_zspage, &pool->pages_allocated);
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
}
static void free_zspage(struct zs_pool *pool, struct size_class *class,
struct zspage *zspage)
{
VM_BUG_ON(get_zspage_inuse(zspage));
VM_BUG_ON(list_empty(&zspage->list));
zsmalloc: replace per zpage lock with pool->migrate_lock The zsmalloc has used a bit for spin_lock in zpage handle to keep zpage object alive during several operations. However, it causes the problem for PREEMPT_RT as well as introducing too complicated. This patch replaces the bit spin_lock with pool->migrate_lock rwlock. It could make the code simple as well as zsmalloc work under PREEMPT_RT. The drawback is the pool->migrate_lock is bigger granuarity than per zpage lock so the contention would be higher than old when both IO-related operations(i.e., zsmalloc, zsfree, zs_[map|unmap]) and compaction(page/zpage migration) are going in parallel(*, the migrate_lock is rwlock and IO related functions are all read side lock so there is no contention). However, the write-side is fast enough(dominant overhead is just page copy) so it wouldn't affect much. If the lock granurity becomes more problem later, we could introduce table locks based on handle as a hash value. Link: https://lkml.kernel.org/r/20211115185909.3949505-9-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:14:13 +00:00
/*
* Since zs_free couldn't be sleepable, this function cannot call
* lock_page. The page locks trylock_zspage got will be released
* by __free_zspage.
*/
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
if (!trylock_zspage(zspage)) {
kick_deferred_free(pool);
return;
}
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
remove_zspage(class, zspage, ZS_INUSE_RATIO_0);
#ifdef CONFIG_ZPOOL
list_del(&zspage->lru);
#endif
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
__free_zspage(pool, class, zspage);
}
/* Initialize a newly allocated zspage */
static void init_zspage(struct size_class *class, struct zspage *zspage)
{
unsigned int freeobj = 1;
unsigned long off = 0;
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
struct page *page = get_first_page(zspage);
while (page) {
struct page *next_page;
struct link_free *link;
void *vaddr;
set_first_obj_offset(page, off);
vaddr = kmap_atomic(page);
link = (struct link_free *)vaddr + off / sizeof(*link);
while ((off += class->size) < PAGE_SIZE) {
link->next = freeobj++ << OBJ_TAG_BITS;
link += class->size / sizeof(*link);
}
/*
* We now come to the last (full or partial) object on this
* page, which must point to the first object on the next
* page (if present)
*/
next_page = get_next_page(page);
if (next_page) {
link->next = freeobj++ << OBJ_TAG_BITS;
} else {
/*
* Reset OBJ_TAG_BITS bit to last link to tell
* whether it's allocated object or not.
*/
link->next = -1UL << OBJ_TAG_BITS;
}
kunmap_atomic(vaddr);
page = next_page;
off %= PAGE_SIZE;
}
#ifdef CONFIG_ZPOOL
INIT_LIST_HEAD(&zspage->lru);
zspage->under_reclaim = false;
#endif
set_freeobj(zspage, 0);
}
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
static void create_page_chain(struct size_class *class, struct zspage *zspage,
struct page *pages[])
{
int i;
struct page *page;
struct page *prev_page = NULL;
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
int nr_pages = class->pages_per_zspage;
/*
* Allocate individual pages and link them together as:
* 1. all pages are linked together using page->index
* 2. each sub-page point to zspage using page->private
*
* we set PG_private to identify the first page (i.e. no other sub-page
* has this flag set).
*/
for (i = 0; i < nr_pages; i++) {
page = pages[i];
set_page_private(page, (unsigned long)zspage);
page->index = 0;
if (i == 0) {
zspage->first_page = page;
SetPagePrivate(page);
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
if (unlikely(class->objs_per_zspage == 1 &&
class->pages_per_zspage == 1))
SetZsHugePage(zspage);
} else {
prev_page->index = (unsigned long)page;
}
prev_page = page;
}
}
/*
* Allocate a zspage for the given size class
*/
static struct zspage *alloc_zspage(struct zs_pool *pool,
struct size_class *class,
gfp_t gfp)
{
int i;
struct page *pages[ZS_MAX_PAGES_PER_ZSPAGE];
struct zspage *zspage = cache_alloc_zspage(pool, gfp);
if (!zspage)
return NULL;
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
zspage->magic = ZSPAGE_MAGIC;
migrate_lock_init(zspage);
for (i = 0; i < class->pages_per_zspage; i++) {
struct page *page;
page = alloc_page(gfp);
if (!page) {
while (--i >= 0) {
dec_zone_page_state(pages[i], NR_ZSPAGES);
__free_page(pages[i]);
}
cache_free_zspage(pool, zspage);
return NULL;
}
inc_zone_page_state(page, NR_ZSPAGES);
pages[i] = page;
}
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
create_page_chain(class, zspage, pages);
init_zspage(class, zspage);
zspage->pool = pool;
return zspage;
}
static struct zspage *find_get_zspage(struct size_class *class)
{
int i;
struct zspage *zspage;
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
for (i = ZS_INUSE_RATIO_99; i >= ZS_INUSE_RATIO_0; i--) {
zspage = list_first_entry_or_null(&class->fullness_list[i],
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
struct zspage, list);
if (zspage)
break;
}
return zspage;
}
static inline int __zs_cpu_up(struct mapping_area *area)
{
/*
* Make sure we don't leak memory if a cpu UP notification
* and zs_init() race and both call zs_cpu_up() on the same cpu
*/
if (area->vm_buf)
return 0;
mm/zsmalloc: support allocating obj with size of ZS_MAX_ALLOC_SIZE I sent a patch [1] for unnecessary check in zsmalloc. And Minchan Kim found zsmalloc even does not support allocating an obj with the size of ZS_MAX_ALLOC_SIZE in some situations. For example: In system with 64KB PAGE_SIZE and 32 bit of physical addr. Then: ZS_MIN_ALLOC_SIZE is 32 bytes which is calculated by: MAX(32, (ZS_MAX_PAGES_PER_ZSPAGE << PAGE_SHIFT >> OBJ_INDEX_BITS)) ZS_MAX_ALLOC_SIZE is 64KB(in current code, is PAGE_SIZE) ZS_SIZE_CLASS_DELTA is 256 bytes So, ZS_SIZE_CLASSES = (ZS_MAX_ALLOC_SIZE - ZS_MIN_ALLOC_SIZE) / ZS_SIZE_CLASS_DELTA + 1 = 256 In zs_create_pool(), the max size obj which can be allocated will be: ZS_MIN_ALLOC_SIZE + i * ZS_SIZE_CLASS_DELTA = 32 + 255*256 = 65312 We can see that 65312 < 65536 (ZS_MAX_ALLOC_SIZE). So we can NOT allocate objs with size ZS_MAX_ALLOC_SIZE(65536) which we promise upper users we can do. [1] http://lkml.iu.edu/hypermail/linux/kernel/1411.2/03835.html [2] http://lkml.iu.edu/hypermail/linux/kernel/1411.2/04534.html This patch fixes this issue by dynamiclly calculating zs_size_classes when module is loaded, allocates buffer with size ZS_MAX_ALLOC_SIZE. Then the max obj(size is ZS_MAX_ALLOC_SIZE) can be stored in it. [akpm@linux-foundation.org: restore ZS_SIZE_CLASSES to fix bisectability] Signed-off-by: Mahendran Ganesh <opensource.ganesh@gmail.com> Suggested-by: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-12-13 00:57:01 +00:00
area->vm_buf = kmalloc(ZS_MAX_ALLOC_SIZE, GFP_KERNEL);
if (!area->vm_buf)
return -ENOMEM;
return 0;
}
static inline void __zs_cpu_down(struct mapping_area *area)
{
mm/zsmalloc: support allocating obj with size of ZS_MAX_ALLOC_SIZE I sent a patch [1] for unnecessary check in zsmalloc. And Minchan Kim found zsmalloc even does not support allocating an obj with the size of ZS_MAX_ALLOC_SIZE in some situations. For example: In system with 64KB PAGE_SIZE and 32 bit of physical addr. Then: ZS_MIN_ALLOC_SIZE is 32 bytes which is calculated by: MAX(32, (ZS_MAX_PAGES_PER_ZSPAGE << PAGE_SHIFT >> OBJ_INDEX_BITS)) ZS_MAX_ALLOC_SIZE is 64KB(in current code, is PAGE_SIZE) ZS_SIZE_CLASS_DELTA is 256 bytes So, ZS_SIZE_CLASSES = (ZS_MAX_ALLOC_SIZE - ZS_MIN_ALLOC_SIZE) / ZS_SIZE_CLASS_DELTA + 1 = 256 In zs_create_pool(), the max size obj which can be allocated will be: ZS_MIN_ALLOC_SIZE + i * ZS_SIZE_CLASS_DELTA = 32 + 255*256 = 65312 We can see that 65312 < 65536 (ZS_MAX_ALLOC_SIZE). So we can NOT allocate objs with size ZS_MAX_ALLOC_SIZE(65536) which we promise upper users we can do. [1] http://lkml.iu.edu/hypermail/linux/kernel/1411.2/03835.html [2] http://lkml.iu.edu/hypermail/linux/kernel/1411.2/04534.html This patch fixes this issue by dynamiclly calculating zs_size_classes when module is loaded, allocates buffer with size ZS_MAX_ALLOC_SIZE. Then the max obj(size is ZS_MAX_ALLOC_SIZE) can be stored in it. [akpm@linux-foundation.org: restore ZS_SIZE_CLASSES to fix bisectability] Signed-off-by: Mahendran Ganesh <opensource.ganesh@gmail.com> Suggested-by: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-12-13 00:57:01 +00:00
kfree(area->vm_buf);
area->vm_buf = NULL;
}
static void *__zs_map_object(struct mapping_area *area,
struct page *pages[2], int off, int size)
{
int sizes[2];
void *addr;
char *buf = area->vm_buf;
/* disable page faults to match kmap_atomic() return conditions */
pagefault_disable();
/* no read fastpath */
if (area->vm_mm == ZS_MM_WO)
goto out;
sizes[0] = PAGE_SIZE - off;
sizes[1] = size - sizes[0];
/* copy object to per-cpu buffer */
addr = kmap_atomic(pages[0]);
memcpy(buf, addr + off, sizes[0]);
kunmap_atomic(addr);
addr = kmap_atomic(pages[1]);
memcpy(buf + sizes[0], addr, sizes[1]);
kunmap_atomic(addr);
out:
return area->vm_buf;
}
static void __zs_unmap_object(struct mapping_area *area,
struct page *pages[2], int off, int size)
{
int sizes[2];
void *addr;
zsmalloc: decouple handle and object Recently, we started to use zram heavily and some of issues popped. 1) external fragmentation I got a report from Juneho Choi that fork failed although there are plenty of free pages in the system. His investigation revealed zram is one of the culprit to make heavy fragmentation so there was no more contiguous 16K page for pgd to fork in the ARM. 2) non-movable pages Other problem of zram now is that inherently, user want to use zram as swap in small memory system so they use zRAM with CMA to use memory efficiently. However, unfortunately, it doesn't work well because zRAM cannot use CMA's movable pages unless it doesn't support compaction. I got several reports about that OOM happened with zram although there are lots of swap space and free space in CMA area. 3) internal fragmentation zRAM has started support memory limitation feature to limit memory usage and I sent a patchset(https://lkml.org/lkml/2014/9/21/148) for VM to be harmonized with zram-swap to stop anonymous page reclaim if zram consumed memory up to the limit although there are free space on the swap. One problem for that direction is zram has no way to know any hole in memory space zsmalloc allocated by internal fragmentation so zram would regard swap is full although there are free space in zsmalloc. For solving the issue, zram want to trigger compaction of zsmalloc before it decides full or not. This patchset is first step to support above issues. For that, it adds indirect layer between handle and object location and supports manual compaction to solve 3th problem first of all. After this patchset got merged, next step is to make VM aware of zsmalloc compaction so that generic compaction will move zsmalloced-pages automatically in runtime. In my imaginary experiment(ie, high compress ratio data with heavy swap in/out on 8G zram-swap), data is as follows, Before = zram allocated object : 60212066 bytes zram total used: 140103680 bytes ratio: 42.98 percent MemFree: 840192 kB Compaction After = frag ratio after compaction zram allocated object : 60212066 bytes zram total used: 76185600 bytes ratio: 79.03 percent MemFree: 901932 kB Juneho reported below in his real platform with small aging. So, I think the benefit would be bigger in real aging system for a long time. - frag_ratio increased 3% (ie, higher is better) - memfree increased about 6MB - In buddy info, Normal 2^3: 4, 2^2: 1: 2^1 increased, Highmem: 2^1 21 increased frag ratio after swap fragment used : 156677 kbytes total: 166092 kbytes frag_ratio : 94 meminfo before compaction MemFree: 83724 kB Node 0, zone Normal 13642 1364 57 10 61 17 9 5 4 0 0 Node 0, zone HighMem 425 29 1 0 0 0 0 0 0 0 0 num_migrated : 23630 compaction done frag ratio after compaction used : 156673 kbytes total: 160564 kbytes frag_ratio : 97 meminfo after compaction MemFree: 89060 kB Node 0, zone Normal 14076 1544 67 14 61 17 9 5 4 0 0 Node 0, zone HighMem 863 50 1 0 0 0 0 0 0 0 0 This patchset adds more logics(about 480 lines) in zsmalloc but when I tested heavy swapin/out program, the regression for swapin/out speed is marginal because most of overheads were caused by compress/decompress and other MM reclaim stuff. This patch (of 7): Currently, handle of zsmalloc encodes object's location directly so it makes support of migration hard. This patch decouples handle and object via adding indirect layer. For that, it allocates handle dynamically and returns it to user. The handle is the address allocated by slab allocation so it's unique and we could keep object's location in the memory space allocated for handle. With it, we can change object's position without changing handle itself. Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:23 +00:00
char *buf;
/* no write fastpath */
if (area->vm_mm == ZS_MM_RO)
goto out;
buf = area->vm_buf;
buf = buf + ZS_HANDLE_SIZE;
size -= ZS_HANDLE_SIZE;
off += ZS_HANDLE_SIZE;
zsmalloc: decouple handle and object Recently, we started to use zram heavily and some of issues popped. 1) external fragmentation I got a report from Juneho Choi that fork failed although there are plenty of free pages in the system. His investigation revealed zram is one of the culprit to make heavy fragmentation so there was no more contiguous 16K page for pgd to fork in the ARM. 2) non-movable pages Other problem of zram now is that inherently, user want to use zram as swap in small memory system so they use zRAM with CMA to use memory efficiently. However, unfortunately, it doesn't work well because zRAM cannot use CMA's movable pages unless it doesn't support compaction. I got several reports about that OOM happened with zram although there are lots of swap space and free space in CMA area. 3) internal fragmentation zRAM has started support memory limitation feature to limit memory usage and I sent a patchset(https://lkml.org/lkml/2014/9/21/148) for VM to be harmonized with zram-swap to stop anonymous page reclaim if zram consumed memory up to the limit although there are free space on the swap. One problem for that direction is zram has no way to know any hole in memory space zsmalloc allocated by internal fragmentation so zram would regard swap is full although there are free space in zsmalloc. For solving the issue, zram want to trigger compaction of zsmalloc before it decides full or not. This patchset is first step to support above issues. For that, it adds indirect layer between handle and object location and supports manual compaction to solve 3th problem first of all. After this patchset got merged, next step is to make VM aware of zsmalloc compaction so that generic compaction will move zsmalloced-pages automatically in runtime. In my imaginary experiment(ie, high compress ratio data with heavy swap in/out on 8G zram-swap), data is as follows, Before = zram allocated object : 60212066 bytes zram total used: 140103680 bytes ratio: 42.98 percent MemFree: 840192 kB Compaction After = frag ratio after compaction zram allocated object : 60212066 bytes zram total used: 76185600 bytes ratio: 79.03 percent MemFree: 901932 kB Juneho reported below in his real platform with small aging. So, I think the benefit would be bigger in real aging system for a long time. - frag_ratio increased 3% (ie, higher is better) - memfree increased about 6MB - In buddy info, Normal 2^3: 4, 2^2: 1: 2^1 increased, Highmem: 2^1 21 increased frag ratio after swap fragment used : 156677 kbytes total: 166092 kbytes frag_ratio : 94 meminfo before compaction MemFree: 83724 kB Node 0, zone Normal 13642 1364 57 10 61 17 9 5 4 0 0 Node 0, zone HighMem 425 29 1 0 0 0 0 0 0 0 0 num_migrated : 23630 compaction done frag ratio after compaction used : 156673 kbytes total: 160564 kbytes frag_ratio : 97 meminfo after compaction MemFree: 89060 kB Node 0, zone Normal 14076 1544 67 14 61 17 9 5 4 0 0 Node 0, zone HighMem 863 50 1 0 0 0 0 0 0 0 0 This patchset adds more logics(about 480 lines) in zsmalloc but when I tested heavy swapin/out program, the regression for swapin/out speed is marginal because most of overheads were caused by compress/decompress and other MM reclaim stuff. This patch (of 7): Currently, handle of zsmalloc encodes object's location directly so it makes support of migration hard. This patch decouples handle and object via adding indirect layer. For that, it allocates handle dynamically and returns it to user. The handle is the address allocated by slab allocation so it's unique and we could keep object's location in the memory space allocated for handle. With it, we can change object's position without changing handle itself. Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:23 +00:00
sizes[0] = PAGE_SIZE - off;
sizes[1] = size - sizes[0];
/* copy per-cpu buffer to object */
addr = kmap_atomic(pages[0]);
memcpy(addr + off, buf, sizes[0]);
kunmap_atomic(addr);
addr = kmap_atomic(pages[1]);
memcpy(addr, buf + sizes[0], sizes[1]);
kunmap_atomic(addr);
out:
/* enable page faults to match kunmap_atomic() return conditions */
pagefault_enable();
}
static int zs_cpu_prepare(unsigned int cpu)
{
struct mapping_area *area;
area = &per_cpu(zs_map_area, cpu);
return __zs_cpu_up(area);
}
static int zs_cpu_dead(unsigned int cpu)
{
struct mapping_area *area;
mm/zsmalloc: support allocating obj with size of ZS_MAX_ALLOC_SIZE I sent a patch [1] for unnecessary check in zsmalloc. And Minchan Kim found zsmalloc even does not support allocating an obj with the size of ZS_MAX_ALLOC_SIZE in some situations. For example: In system with 64KB PAGE_SIZE and 32 bit of physical addr. Then: ZS_MIN_ALLOC_SIZE is 32 bytes which is calculated by: MAX(32, (ZS_MAX_PAGES_PER_ZSPAGE << PAGE_SHIFT >> OBJ_INDEX_BITS)) ZS_MAX_ALLOC_SIZE is 64KB(in current code, is PAGE_SIZE) ZS_SIZE_CLASS_DELTA is 256 bytes So, ZS_SIZE_CLASSES = (ZS_MAX_ALLOC_SIZE - ZS_MIN_ALLOC_SIZE) / ZS_SIZE_CLASS_DELTA + 1 = 256 In zs_create_pool(), the max size obj which can be allocated will be: ZS_MIN_ALLOC_SIZE + i * ZS_SIZE_CLASS_DELTA = 32 + 255*256 = 65312 We can see that 65312 < 65536 (ZS_MAX_ALLOC_SIZE). So we can NOT allocate objs with size ZS_MAX_ALLOC_SIZE(65536) which we promise upper users we can do. [1] http://lkml.iu.edu/hypermail/linux/kernel/1411.2/03835.html [2] http://lkml.iu.edu/hypermail/linux/kernel/1411.2/04534.html This patch fixes this issue by dynamiclly calculating zs_size_classes when module is loaded, allocates buffer with size ZS_MAX_ALLOC_SIZE. Then the max obj(size is ZS_MAX_ALLOC_SIZE) can be stored in it. [akpm@linux-foundation.org: restore ZS_SIZE_CLASSES to fix bisectability] Signed-off-by: Mahendran Ganesh <opensource.ganesh@gmail.com> Suggested-by: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-12-13 00:57:01 +00:00
area = &per_cpu(zs_map_area, cpu);
__zs_cpu_down(area);
return 0;
}
static bool can_merge(struct size_class *prev, int pages_per_zspage,
int objs_per_zspage)
zsmalloc: merge size_class to reduce fragmentation zsmalloc has many size_classes to reduce fragmentation and they are in 16 bytes unit, for example, 16, 32, 48, etc., if PAGE_SIZE is 4096. And, zsmalloc has constraint that each zspage has 4 pages at maximum. In this situation, we can see interesting aspect. Let's think about size_class for 1488, 1472, ..., 1376. To prevent external fragmentation, they uses 4 pages per zspage and so all they can contain 11 objects at maximum. 16384 (4096 * 4) = 1488 * 11 + remains 16384 (4096 * 4) = 1472 * 11 + remains 16384 (4096 * 4) = ... 16384 (4096 * 4) = 1376 * 11 + remains It means that they have same characteristics and classification between them isn't needed. If we use one size_class for them, we can reduce fragementation and save some memory since both the 1488 and 1472 sized classes can only fit 11 objects into 4 pages, and an object that's 1472 bytes can fit into an object that's 1488 bytes, merging these classes to always use objects that are 1488 bytes will reduce the total number of size classes. And reducing the total number of size classes reduces overall fragmentation, because a wider range of compressed pages can fit into a single size class, leaving less unused objects in each size class. For this purpose, this patch implement size_class merging. If there is size_class that have same pages_per_zspage and same number of objects per zspage with previous size_class, we don't create new size_class. Instead, we use previous, same characteristic size_class. With this way, above example sizes (1488, 1472, ..., 1376) use just one size_class so we can get much more memory utilization. Below is result of my simple test. TEST ENV: EXT4 on zram, mount with discard option WORKLOAD: untar kernel source code, remove directory in descending order in size. (drivers arch fs sound include net Documentation firmware kernel tools) Each line represents orig_data_size, compr_data_size, mem_used_total, fragmentation overhead (mem_used - compr_data_size) and overhead ratio (overhead to compr_data_size), respectively, after untar and remove operation is executed. * untar-nomerge.out orig_size compr_size used_size overhead overhead_ratio 525.88MB 199.16MB 210.23MB 11.08MB 5.56% 288.32MB 97.43MB 105.63MB 8.20MB 8.41% 177.32MB 61.12MB 69.40MB 8.28MB 13.55% 146.47MB 47.32MB 56.10MB 8.78MB 18.55% 124.16MB 38.85MB 48.41MB 9.55MB 24.58% 103.93MB 31.68MB 40.93MB 9.25MB 29.21% 84.34MB 22.86MB 32.72MB 9.86MB 43.13% 66.87MB 14.83MB 23.83MB 9.00MB 60.70% 60.67MB 11.11MB 18.60MB 7.49MB 67.48% 55.86MB 8.83MB 16.61MB 7.77MB 88.03% 53.32MB 8.01MB 15.32MB 7.31MB 91.24% * untar-merge.out orig_size compr_size used_size overhead overhead_ratio 526.23MB 199.18MB 209.81MB 10.64MB 5.34% 288.68MB 97.45MB 104.08MB 6.63MB 6.80% 177.68MB 61.14MB 66.93MB 5.79MB 9.47% 146.83MB 47.34MB 52.79MB 5.45MB 11.51% 124.52MB 38.87MB 44.30MB 5.43MB 13.96% 104.29MB 31.70MB 36.83MB 5.13MB 16.19% 84.70MB 22.88MB 27.92MB 5.04MB 22.04% 67.11MB 14.83MB 19.26MB 4.43MB 29.86% 60.82MB 11.10MB 14.90MB 3.79MB 34.17% 55.90MB 8.82MB 12.61MB 3.79MB 42.97% 53.32MB 8.01MB 11.73MB 3.73MB 46.53% As you can see above result, merged one has better utilization (overhead ratio, 5th column) and uses less memory (mem_used_total, 3rd column). Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Reviewed-by: Dan Streetman <ddstreet@ieee.org> Cc: Luigi Semenzato <semenzato@google.com> Cc: <juno.choi@lge.com> Cc: "seungho1.park" <seungho1.park@lge.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-12-13 00:56:44 +00:00
{
if (prev->pages_per_zspage == pages_per_zspage &&
prev->objs_per_zspage == objs_per_zspage)
return true;
zsmalloc: merge size_class to reduce fragmentation zsmalloc has many size_classes to reduce fragmentation and they are in 16 bytes unit, for example, 16, 32, 48, etc., if PAGE_SIZE is 4096. And, zsmalloc has constraint that each zspage has 4 pages at maximum. In this situation, we can see interesting aspect. Let's think about size_class for 1488, 1472, ..., 1376. To prevent external fragmentation, they uses 4 pages per zspage and so all they can contain 11 objects at maximum. 16384 (4096 * 4) = 1488 * 11 + remains 16384 (4096 * 4) = 1472 * 11 + remains 16384 (4096 * 4) = ... 16384 (4096 * 4) = 1376 * 11 + remains It means that they have same characteristics and classification between them isn't needed. If we use one size_class for them, we can reduce fragementation and save some memory since both the 1488 and 1472 sized classes can only fit 11 objects into 4 pages, and an object that's 1472 bytes can fit into an object that's 1488 bytes, merging these classes to always use objects that are 1488 bytes will reduce the total number of size classes. And reducing the total number of size classes reduces overall fragmentation, because a wider range of compressed pages can fit into a single size class, leaving less unused objects in each size class. For this purpose, this patch implement size_class merging. If there is size_class that have same pages_per_zspage and same number of objects per zspage with previous size_class, we don't create new size_class. Instead, we use previous, same characteristic size_class. With this way, above example sizes (1488, 1472, ..., 1376) use just one size_class so we can get much more memory utilization. Below is result of my simple test. TEST ENV: EXT4 on zram, mount with discard option WORKLOAD: untar kernel source code, remove directory in descending order in size. (drivers arch fs sound include net Documentation firmware kernel tools) Each line represents orig_data_size, compr_data_size, mem_used_total, fragmentation overhead (mem_used - compr_data_size) and overhead ratio (overhead to compr_data_size), respectively, after untar and remove operation is executed. * untar-nomerge.out orig_size compr_size used_size overhead overhead_ratio 525.88MB 199.16MB 210.23MB 11.08MB 5.56% 288.32MB 97.43MB 105.63MB 8.20MB 8.41% 177.32MB 61.12MB 69.40MB 8.28MB 13.55% 146.47MB 47.32MB 56.10MB 8.78MB 18.55% 124.16MB 38.85MB 48.41MB 9.55MB 24.58% 103.93MB 31.68MB 40.93MB 9.25MB 29.21% 84.34MB 22.86MB 32.72MB 9.86MB 43.13% 66.87MB 14.83MB 23.83MB 9.00MB 60.70% 60.67MB 11.11MB 18.60MB 7.49MB 67.48% 55.86MB 8.83MB 16.61MB 7.77MB 88.03% 53.32MB 8.01MB 15.32MB 7.31MB 91.24% * untar-merge.out orig_size compr_size used_size overhead overhead_ratio 526.23MB 199.18MB 209.81MB 10.64MB 5.34% 288.68MB 97.45MB 104.08MB 6.63MB 6.80% 177.68MB 61.14MB 66.93MB 5.79MB 9.47% 146.83MB 47.34MB 52.79MB 5.45MB 11.51% 124.52MB 38.87MB 44.30MB 5.43MB 13.96% 104.29MB 31.70MB 36.83MB 5.13MB 16.19% 84.70MB 22.88MB 27.92MB 5.04MB 22.04% 67.11MB 14.83MB 19.26MB 4.43MB 29.86% 60.82MB 11.10MB 14.90MB 3.79MB 34.17% 55.90MB 8.82MB 12.61MB 3.79MB 42.97% 53.32MB 8.01MB 11.73MB 3.73MB 46.53% As you can see above result, merged one has better utilization (overhead ratio, 5th column) and uses less memory (mem_used_total, 3rd column). Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Reviewed-by: Dan Streetman <ddstreet@ieee.org> Cc: Luigi Semenzato <semenzato@google.com> Cc: <juno.choi@lge.com> Cc: "seungho1.park" <seungho1.park@lge.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-12-13 00:56:44 +00:00
return false;
zsmalloc: merge size_class to reduce fragmentation zsmalloc has many size_classes to reduce fragmentation and they are in 16 bytes unit, for example, 16, 32, 48, etc., if PAGE_SIZE is 4096. And, zsmalloc has constraint that each zspage has 4 pages at maximum. In this situation, we can see interesting aspect. Let's think about size_class for 1488, 1472, ..., 1376. To prevent external fragmentation, they uses 4 pages per zspage and so all they can contain 11 objects at maximum. 16384 (4096 * 4) = 1488 * 11 + remains 16384 (4096 * 4) = 1472 * 11 + remains 16384 (4096 * 4) = ... 16384 (4096 * 4) = 1376 * 11 + remains It means that they have same characteristics and classification between them isn't needed. If we use one size_class for them, we can reduce fragementation and save some memory since both the 1488 and 1472 sized classes can only fit 11 objects into 4 pages, and an object that's 1472 bytes can fit into an object that's 1488 bytes, merging these classes to always use objects that are 1488 bytes will reduce the total number of size classes. And reducing the total number of size classes reduces overall fragmentation, because a wider range of compressed pages can fit into a single size class, leaving less unused objects in each size class. For this purpose, this patch implement size_class merging. If there is size_class that have same pages_per_zspage and same number of objects per zspage with previous size_class, we don't create new size_class. Instead, we use previous, same characteristic size_class. With this way, above example sizes (1488, 1472, ..., 1376) use just one size_class so we can get much more memory utilization. Below is result of my simple test. TEST ENV: EXT4 on zram, mount with discard option WORKLOAD: untar kernel source code, remove directory in descending order in size. (drivers arch fs sound include net Documentation firmware kernel tools) Each line represents orig_data_size, compr_data_size, mem_used_total, fragmentation overhead (mem_used - compr_data_size) and overhead ratio (overhead to compr_data_size), respectively, after untar and remove operation is executed. * untar-nomerge.out orig_size compr_size used_size overhead overhead_ratio 525.88MB 199.16MB 210.23MB 11.08MB 5.56% 288.32MB 97.43MB 105.63MB 8.20MB 8.41% 177.32MB 61.12MB 69.40MB 8.28MB 13.55% 146.47MB 47.32MB 56.10MB 8.78MB 18.55% 124.16MB 38.85MB 48.41MB 9.55MB 24.58% 103.93MB 31.68MB 40.93MB 9.25MB 29.21% 84.34MB 22.86MB 32.72MB 9.86MB 43.13% 66.87MB 14.83MB 23.83MB 9.00MB 60.70% 60.67MB 11.11MB 18.60MB 7.49MB 67.48% 55.86MB 8.83MB 16.61MB 7.77MB 88.03% 53.32MB 8.01MB 15.32MB 7.31MB 91.24% * untar-merge.out orig_size compr_size used_size overhead overhead_ratio 526.23MB 199.18MB 209.81MB 10.64MB 5.34% 288.68MB 97.45MB 104.08MB 6.63MB 6.80% 177.68MB 61.14MB 66.93MB 5.79MB 9.47% 146.83MB 47.34MB 52.79MB 5.45MB 11.51% 124.52MB 38.87MB 44.30MB 5.43MB 13.96% 104.29MB 31.70MB 36.83MB 5.13MB 16.19% 84.70MB 22.88MB 27.92MB 5.04MB 22.04% 67.11MB 14.83MB 19.26MB 4.43MB 29.86% 60.82MB 11.10MB 14.90MB 3.79MB 34.17% 55.90MB 8.82MB 12.61MB 3.79MB 42.97% 53.32MB 8.01MB 11.73MB 3.73MB 46.53% As you can see above result, merged one has better utilization (overhead ratio, 5th column) and uses less memory (mem_used_total, 3rd column). Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Reviewed-by: Dan Streetman <ddstreet@ieee.org> Cc: Luigi Semenzato <semenzato@google.com> Cc: <juno.choi@lge.com> Cc: "seungho1.park" <seungho1.park@lge.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-12-13 00:56:44 +00:00
}
static bool zspage_full(struct size_class *class, struct zspage *zspage)
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
{
return get_zspage_inuse(zspage) == class->objs_per_zspage;
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
}
/**
* zs_lookup_class_index() - Returns index of the zsmalloc &size_class
* that hold objects of the provided size.
* @pool: zsmalloc pool to use
* @size: object size
*
* Context: Any context.
*
* Return: the index of the zsmalloc &size_class that hold objects of the
* provided size.
*/
unsigned int zs_lookup_class_index(struct zs_pool *pool, unsigned int size)
{
struct size_class *class;
class = pool->size_class[get_size_class_index(size)];
return class->index;
}
EXPORT_SYMBOL_GPL(zs_lookup_class_index);
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
unsigned long zs_get_total_pages(struct zs_pool *pool)
{
return atomic_long_read(&pool->pages_allocated);
}
EXPORT_SYMBOL_GPL(zs_get_total_pages);
/**
* zs_map_object - get address of allocated object from handle.
* @pool: pool from which the object was allocated
* @handle: handle returned from zs_malloc
* @mm: mapping mode to use
*
* Before using an object allocated from zs_malloc, it must be mapped using
* this function. When done with the object, it must be unmapped using
* zs_unmap_object.
*
* Only one object can be mapped per cpu at a time. There is no protection
* against nested mappings.
*
* This function returns with preemption and page faults disabled.
*/
void *zs_map_object(struct zs_pool *pool, unsigned long handle,
enum zs_mapmode mm)
{
struct zspage *zspage;
struct page *page;
unsigned long obj, off;
unsigned int obj_idx;
struct size_class *class;
struct mapping_area *area;
struct page *pages[2];
zsmalloc: decouple handle and object Recently, we started to use zram heavily and some of issues popped. 1) external fragmentation I got a report from Juneho Choi that fork failed although there are plenty of free pages in the system. His investigation revealed zram is one of the culprit to make heavy fragmentation so there was no more contiguous 16K page for pgd to fork in the ARM. 2) non-movable pages Other problem of zram now is that inherently, user want to use zram as swap in small memory system so they use zRAM with CMA to use memory efficiently. However, unfortunately, it doesn't work well because zRAM cannot use CMA's movable pages unless it doesn't support compaction. I got several reports about that OOM happened with zram although there are lots of swap space and free space in CMA area. 3) internal fragmentation zRAM has started support memory limitation feature to limit memory usage and I sent a patchset(https://lkml.org/lkml/2014/9/21/148) for VM to be harmonized with zram-swap to stop anonymous page reclaim if zram consumed memory up to the limit although there are free space on the swap. One problem for that direction is zram has no way to know any hole in memory space zsmalloc allocated by internal fragmentation so zram would regard swap is full although there are free space in zsmalloc. For solving the issue, zram want to trigger compaction of zsmalloc before it decides full or not. This patchset is first step to support above issues. For that, it adds indirect layer between handle and object location and supports manual compaction to solve 3th problem first of all. After this patchset got merged, next step is to make VM aware of zsmalloc compaction so that generic compaction will move zsmalloced-pages automatically in runtime. In my imaginary experiment(ie, high compress ratio data with heavy swap in/out on 8G zram-swap), data is as follows, Before = zram allocated object : 60212066 bytes zram total used: 140103680 bytes ratio: 42.98 percent MemFree: 840192 kB Compaction After = frag ratio after compaction zram allocated object : 60212066 bytes zram total used: 76185600 bytes ratio: 79.03 percent MemFree: 901932 kB Juneho reported below in his real platform with small aging. So, I think the benefit would be bigger in real aging system for a long time. - frag_ratio increased 3% (ie, higher is better) - memfree increased about 6MB - In buddy info, Normal 2^3: 4, 2^2: 1: 2^1 increased, Highmem: 2^1 21 increased frag ratio after swap fragment used : 156677 kbytes total: 166092 kbytes frag_ratio : 94 meminfo before compaction MemFree: 83724 kB Node 0, zone Normal 13642 1364 57 10 61 17 9 5 4 0 0 Node 0, zone HighMem 425 29 1 0 0 0 0 0 0 0 0 num_migrated : 23630 compaction done frag ratio after compaction used : 156673 kbytes total: 160564 kbytes frag_ratio : 97 meminfo after compaction MemFree: 89060 kB Node 0, zone Normal 14076 1544 67 14 61 17 9 5 4 0 0 Node 0, zone HighMem 863 50 1 0 0 0 0 0 0 0 0 This patchset adds more logics(about 480 lines) in zsmalloc but when I tested heavy swapin/out program, the regression for swapin/out speed is marginal because most of overheads were caused by compress/decompress and other MM reclaim stuff. This patch (of 7): Currently, handle of zsmalloc encodes object's location directly so it makes support of migration hard. This patch decouples handle and object via adding indirect layer. For that, it allocates handle dynamically and returns it to user. The handle is the address allocated by slab allocation so it's unique and we could keep object's location in the memory space allocated for handle. With it, we can change object's position without changing handle itself. Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:23 +00:00
void *ret;
zsmalloc: merge size_class to reduce fragmentation zsmalloc has many size_classes to reduce fragmentation and they are in 16 bytes unit, for example, 16, 32, 48, etc., if PAGE_SIZE is 4096. And, zsmalloc has constraint that each zspage has 4 pages at maximum. In this situation, we can see interesting aspect. Let's think about size_class for 1488, 1472, ..., 1376. To prevent external fragmentation, they uses 4 pages per zspage and so all they can contain 11 objects at maximum. 16384 (4096 * 4) = 1488 * 11 + remains 16384 (4096 * 4) = 1472 * 11 + remains 16384 (4096 * 4) = ... 16384 (4096 * 4) = 1376 * 11 + remains It means that they have same characteristics and classification between them isn't needed. If we use one size_class for them, we can reduce fragementation and save some memory since both the 1488 and 1472 sized classes can only fit 11 objects into 4 pages, and an object that's 1472 bytes can fit into an object that's 1488 bytes, merging these classes to always use objects that are 1488 bytes will reduce the total number of size classes. And reducing the total number of size classes reduces overall fragmentation, because a wider range of compressed pages can fit into a single size class, leaving less unused objects in each size class. For this purpose, this patch implement size_class merging. If there is size_class that have same pages_per_zspage and same number of objects per zspage with previous size_class, we don't create new size_class. Instead, we use previous, same characteristic size_class. With this way, above example sizes (1488, 1472, ..., 1376) use just one size_class so we can get much more memory utilization. Below is result of my simple test. TEST ENV: EXT4 on zram, mount with discard option WORKLOAD: untar kernel source code, remove directory in descending order in size. (drivers arch fs sound include net Documentation firmware kernel tools) Each line represents orig_data_size, compr_data_size, mem_used_total, fragmentation overhead (mem_used - compr_data_size) and overhead ratio (overhead to compr_data_size), respectively, after untar and remove operation is executed. * untar-nomerge.out orig_size compr_size used_size overhead overhead_ratio 525.88MB 199.16MB 210.23MB 11.08MB 5.56% 288.32MB 97.43MB 105.63MB 8.20MB 8.41% 177.32MB 61.12MB 69.40MB 8.28MB 13.55% 146.47MB 47.32MB 56.10MB 8.78MB 18.55% 124.16MB 38.85MB 48.41MB 9.55MB 24.58% 103.93MB 31.68MB 40.93MB 9.25MB 29.21% 84.34MB 22.86MB 32.72MB 9.86MB 43.13% 66.87MB 14.83MB 23.83MB 9.00MB 60.70% 60.67MB 11.11MB 18.60MB 7.49MB 67.48% 55.86MB 8.83MB 16.61MB 7.77MB 88.03% 53.32MB 8.01MB 15.32MB 7.31MB 91.24% * untar-merge.out orig_size compr_size used_size overhead overhead_ratio 526.23MB 199.18MB 209.81MB 10.64MB 5.34% 288.68MB 97.45MB 104.08MB 6.63MB 6.80% 177.68MB 61.14MB 66.93MB 5.79MB 9.47% 146.83MB 47.34MB 52.79MB 5.45MB 11.51% 124.52MB 38.87MB 44.30MB 5.43MB 13.96% 104.29MB 31.70MB 36.83MB 5.13MB 16.19% 84.70MB 22.88MB 27.92MB 5.04MB 22.04% 67.11MB 14.83MB 19.26MB 4.43MB 29.86% 60.82MB 11.10MB 14.90MB 3.79MB 34.17% 55.90MB 8.82MB 12.61MB 3.79MB 42.97% 53.32MB 8.01MB 11.73MB 3.73MB 46.53% As you can see above result, merged one has better utilization (overhead ratio, 5th column) and uses less memory (mem_used_total, 3rd column). Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Reviewed-by: Dan Streetman <ddstreet@ieee.org> Cc: Luigi Semenzato <semenzato@google.com> Cc: <juno.choi@lge.com> Cc: "seungho1.park" <seungho1.park@lge.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-12-13 00:56:44 +00:00
/*
* Because we use per-cpu mapping areas shared among the
* pools/users, we can't allow mapping in interrupt context
* because it can corrupt another users mappings.
zsmalloc: merge size_class to reduce fragmentation zsmalloc has many size_classes to reduce fragmentation and they are in 16 bytes unit, for example, 16, 32, 48, etc., if PAGE_SIZE is 4096. And, zsmalloc has constraint that each zspage has 4 pages at maximum. In this situation, we can see interesting aspect. Let's think about size_class for 1488, 1472, ..., 1376. To prevent external fragmentation, they uses 4 pages per zspage and so all they can contain 11 objects at maximum. 16384 (4096 * 4) = 1488 * 11 + remains 16384 (4096 * 4) = 1472 * 11 + remains 16384 (4096 * 4) = ... 16384 (4096 * 4) = 1376 * 11 + remains It means that they have same characteristics and classification between them isn't needed. If we use one size_class for them, we can reduce fragementation and save some memory since both the 1488 and 1472 sized classes can only fit 11 objects into 4 pages, and an object that's 1472 bytes can fit into an object that's 1488 bytes, merging these classes to always use objects that are 1488 bytes will reduce the total number of size classes. And reducing the total number of size classes reduces overall fragmentation, because a wider range of compressed pages can fit into a single size class, leaving less unused objects in each size class. For this purpose, this patch implement size_class merging. If there is size_class that have same pages_per_zspage and same number of objects per zspage with previous size_class, we don't create new size_class. Instead, we use previous, same characteristic size_class. With this way, above example sizes (1488, 1472, ..., 1376) use just one size_class so we can get much more memory utilization. Below is result of my simple test. TEST ENV: EXT4 on zram, mount with discard option WORKLOAD: untar kernel source code, remove directory in descending order in size. (drivers arch fs sound include net Documentation firmware kernel tools) Each line represents orig_data_size, compr_data_size, mem_used_total, fragmentation overhead (mem_used - compr_data_size) and overhead ratio (overhead to compr_data_size), respectively, after untar and remove operation is executed. * untar-nomerge.out orig_size compr_size used_size overhead overhead_ratio 525.88MB 199.16MB 210.23MB 11.08MB 5.56% 288.32MB 97.43MB 105.63MB 8.20MB 8.41% 177.32MB 61.12MB 69.40MB 8.28MB 13.55% 146.47MB 47.32MB 56.10MB 8.78MB 18.55% 124.16MB 38.85MB 48.41MB 9.55MB 24.58% 103.93MB 31.68MB 40.93MB 9.25MB 29.21% 84.34MB 22.86MB 32.72MB 9.86MB 43.13% 66.87MB 14.83MB 23.83MB 9.00MB 60.70% 60.67MB 11.11MB 18.60MB 7.49MB 67.48% 55.86MB 8.83MB 16.61MB 7.77MB 88.03% 53.32MB 8.01MB 15.32MB 7.31MB 91.24% * untar-merge.out orig_size compr_size used_size overhead overhead_ratio 526.23MB 199.18MB 209.81MB 10.64MB 5.34% 288.68MB 97.45MB 104.08MB 6.63MB 6.80% 177.68MB 61.14MB 66.93MB 5.79MB 9.47% 146.83MB 47.34MB 52.79MB 5.45MB 11.51% 124.52MB 38.87MB 44.30MB 5.43MB 13.96% 104.29MB 31.70MB 36.83MB 5.13MB 16.19% 84.70MB 22.88MB 27.92MB 5.04MB 22.04% 67.11MB 14.83MB 19.26MB 4.43MB 29.86% 60.82MB 11.10MB 14.90MB 3.79MB 34.17% 55.90MB 8.82MB 12.61MB 3.79MB 42.97% 53.32MB 8.01MB 11.73MB 3.73MB 46.53% As you can see above result, merged one has better utilization (overhead ratio, 5th column) and uses less memory (mem_used_total, 3rd column). Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Reviewed-by: Dan Streetman <ddstreet@ieee.org> Cc: Luigi Semenzato <semenzato@google.com> Cc: <juno.choi@lge.com> Cc: "seungho1.park" <seungho1.park@lge.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-12-13 00:56:44 +00:00
*/
BUG_ON(in_interrupt());
zsmalloc: replace per zpage lock with pool->migrate_lock The zsmalloc has used a bit for spin_lock in zpage handle to keep zpage object alive during several operations. However, it causes the problem for PREEMPT_RT as well as introducing too complicated. This patch replaces the bit spin_lock with pool->migrate_lock rwlock. It could make the code simple as well as zsmalloc work under PREEMPT_RT. The drawback is the pool->migrate_lock is bigger granuarity than per zpage lock so the contention would be higher than old when both IO-related operations(i.e., zsmalloc, zsfree, zs_[map|unmap]) and compaction(page/zpage migration) are going in parallel(*, the migrate_lock is rwlock and IO related functions are all read side lock so there is no contention). However, the write-side is fast enough(dominant overhead is just page copy) so it wouldn't affect much. If the lock granurity becomes more problem later, we could introduce table locks based on handle as a hash value. Link: https://lkml.kernel.org/r/20211115185909.3949505-9-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:14:13 +00:00
/* It guarantees it can get zspage from handle safely */
zsmalloc: consolidate zs_pool's migrate_lock and size_class's locks Currently, zsmalloc has a hierarchy of locks, which includes a pool-level migrate_lock, and a lock for each size class. We have to obtain both locks in the hotpath in most cases anyway, except for zs_malloc. This exception will no longer exist when we introduce a LRU into the zs_pool for the new writeback functionality - we will need to obtain a pool-level lock to synchronize LRU handling even in zs_malloc. In preparation for zsmalloc writeback, consolidate these locks into a single pool-level lock, which drastically reduces the complexity of synchronization in zsmalloc. We have also benchmarked the lock consolidation to see the performance effect of this change on zram. First, we ran a synthetic FS workload on a server machine with 36 cores (same machine for all runs), using fs_mark -d ../zram1mnt -s 100000 -n 2500 -t 32 -k before and after for btrfs and ext4 on zram (FS usage is 80%). Here is the result (unit is file/second): With lock consolidation (btrfs): Average: 13520.2, Median: 13531.0, Stddev: 137.5961482019028 Without lock consolidation (btrfs): Average: 13487.2, Median: 13575.0, Stddev: 309.08283679298665 With lock consolidation (ext4): Average: 16824.4, Median: 16839.0, Stddev: 89.97388510006668 Without lock consolidation (ext4) Average: 16958.0, Median: 16986.0, Stddev: 194.7370021336469 As you can see, we observe a 0.3% regression for btrfs, and a 0.9% regression for ext4. This is a small, barely measurable difference in my opinion. For a more realistic scenario, we also tries building the kernel on zram. Here is the time it takes (in seconds): With lock consolidation (btrfs): real Average: 319.6, Median: 320.0, Stddev: 0.8944271909999159 user Average: 6894.2, Median: 6895.0, Stddev: 25.528415540334656 sys Average: 521.4, Median: 522.0, Stddev: 1.51657508881031 Without lock consolidation (btrfs): real Average: 319.8, Median: 320.0, Stddev: 0.8366600265340756 user Average: 6896.6, Median: 6899.0, Stddev: 16.04057355583023 sys Average: 520.6, Median: 521.0, Stddev: 1.140175425099138 With lock consolidation (ext4): real Average: 320.0, Median: 319.0, Stddev: 1.4142135623730951 user Average: 6896.8, Median: 6878.0, Stddev: 28.621670111997307 sys Average: 521.2, Median: 521.0, Stddev: 1.7888543819998317 Without lock consolidation (ext4) real Average: 319.6, Median: 319.0, Stddev: 0.8944271909999159 user Average: 6886.2, Median: 6887.0, Stddev: 16.93221781102523 sys Average: 520.4, Median: 520.0, Stddev: 1.140175425099138 The difference is entirely within the noise of a typical run on zram. This hardly justifies the complexity of maintaining both the pool lock and the class lock. In fact, for writeback, we would need to introduce yet another lock to prevent data races on the pool's LRU, further complicating the lock handling logic. IMHO, it is just better to collapse all of these into a single pool-level lock. Link: https://lkml.kernel.org/r/20221128191616.1261026-4-nphamcs@gmail.com Signed-off-by: Nhat Pham <nphamcs@gmail.com> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Acked-by: Minchan Kim <minchan@kernel.org> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-11-28 19:16:12 +00:00
spin_lock(&pool->lock);
zsmalloc: decouple handle and object Recently, we started to use zram heavily and some of issues popped. 1) external fragmentation I got a report from Juneho Choi that fork failed although there are plenty of free pages in the system. His investigation revealed zram is one of the culprit to make heavy fragmentation so there was no more contiguous 16K page for pgd to fork in the ARM. 2) non-movable pages Other problem of zram now is that inherently, user want to use zram as swap in small memory system so they use zRAM with CMA to use memory efficiently. However, unfortunately, it doesn't work well because zRAM cannot use CMA's movable pages unless it doesn't support compaction. I got several reports about that OOM happened with zram although there are lots of swap space and free space in CMA area. 3) internal fragmentation zRAM has started support memory limitation feature to limit memory usage and I sent a patchset(https://lkml.org/lkml/2014/9/21/148) for VM to be harmonized with zram-swap to stop anonymous page reclaim if zram consumed memory up to the limit although there are free space on the swap. One problem for that direction is zram has no way to know any hole in memory space zsmalloc allocated by internal fragmentation so zram would regard swap is full although there are free space in zsmalloc. For solving the issue, zram want to trigger compaction of zsmalloc before it decides full or not. This patchset is first step to support above issues. For that, it adds indirect layer between handle and object location and supports manual compaction to solve 3th problem first of all. After this patchset got merged, next step is to make VM aware of zsmalloc compaction so that generic compaction will move zsmalloced-pages automatically in runtime. In my imaginary experiment(ie, high compress ratio data with heavy swap in/out on 8G zram-swap), data is as follows, Before = zram allocated object : 60212066 bytes zram total used: 140103680 bytes ratio: 42.98 percent MemFree: 840192 kB Compaction After = frag ratio after compaction zram allocated object : 60212066 bytes zram total used: 76185600 bytes ratio: 79.03 percent MemFree: 901932 kB Juneho reported below in his real platform with small aging. So, I think the benefit would be bigger in real aging system for a long time. - frag_ratio increased 3% (ie, higher is better) - memfree increased about 6MB - In buddy info, Normal 2^3: 4, 2^2: 1: 2^1 increased, Highmem: 2^1 21 increased frag ratio after swap fragment used : 156677 kbytes total: 166092 kbytes frag_ratio : 94 meminfo before compaction MemFree: 83724 kB Node 0, zone Normal 13642 1364 57 10 61 17 9 5 4 0 0 Node 0, zone HighMem 425 29 1 0 0 0 0 0 0 0 0 num_migrated : 23630 compaction done frag ratio after compaction used : 156673 kbytes total: 160564 kbytes frag_ratio : 97 meminfo after compaction MemFree: 89060 kB Node 0, zone Normal 14076 1544 67 14 61 17 9 5 4 0 0 Node 0, zone HighMem 863 50 1 0 0 0 0 0 0 0 0 This patchset adds more logics(about 480 lines) in zsmalloc but when I tested heavy swapin/out program, the regression for swapin/out speed is marginal because most of overheads were caused by compress/decompress and other MM reclaim stuff. This patch (of 7): Currently, handle of zsmalloc encodes object's location directly so it makes support of migration hard. This patch decouples handle and object via adding indirect layer. For that, it allocates handle dynamically and returns it to user. The handle is the address allocated by slab allocation so it's unique and we could keep object's location in the memory space allocated for handle. With it, we can change object's position without changing handle itself. Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:23 +00:00
obj = handle_to_obj(handle);
obj_to_location(obj, &page, &obj_idx);
zspage = get_zspage(page);
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
zsmalloc: replace per zpage lock with pool->migrate_lock The zsmalloc has used a bit for spin_lock in zpage handle to keep zpage object alive during several operations. However, it causes the problem for PREEMPT_RT as well as introducing too complicated. This patch replaces the bit spin_lock with pool->migrate_lock rwlock. It could make the code simple as well as zsmalloc work under PREEMPT_RT. The drawback is the pool->migrate_lock is bigger granuarity than per zpage lock so the contention would be higher than old when both IO-related operations(i.e., zsmalloc, zsfree, zs_[map|unmap]) and compaction(page/zpage migration) are going in parallel(*, the migrate_lock is rwlock and IO related functions are all read side lock so there is no contention). However, the write-side is fast enough(dominant overhead is just page copy) so it wouldn't affect much. If the lock granurity becomes more problem later, we could introduce table locks based on handle as a hash value. Link: https://lkml.kernel.org/r/20211115185909.3949505-9-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:14:13 +00:00
/*
zsmalloc: consolidate zs_pool's migrate_lock and size_class's locks Currently, zsmalloc has a hierarchy of locks, which includes a pool-level migrate_lock, and a lock for each size class. We have to obtain both locks in the hotpath in most cases anyway, except for zs_malloc. This exception will no longer exist when we introduce a LRU into the zs_pool for the new writeback functionality - we will need to obtain a pool-level lock to synchronize LRU handling even in zs_malloc. In preparation for zsmalloc writeback, consolidate these locks into a single pool-level lock, which drastically reduces the complexity of synchronization in zsmalloc. We have also benchmarked the lock consolidation to see the performance effect of this change on zram. First, we ran a synthetic FS workload on a server machine with 36 cores (same machine for all runs), using fs_mark -d ../zram1mnt -s 100000 -n 2500 -t 32 -k before and after for btrfs and ext4 on zram (FS usage is 80%). Here is the result (unit is file/second): With lock consolidation (btrfs): Average: 13520.2, Median: 13531.0, Stddev: 137.5961482019028 Without lock consolidation (btrfs): Average: 13487.2, Median: 13575.0, Stddev: 309.08283679298665 With lock consolidation (ext4): Average: 16824.4, Median: 16839.0, Stddev: 89.97388510006668 Without lock consolidation (ext4) Average: 16958.0, Median: 16986.0, Stddev: 194.7370021336469 As you can see, we observe a 0.3% regression for btrfs, and a 0.9% regression for ext4. This is a small, barely measurable difference in my opinion. For a more realistic scenario, we also tries building the kernel on zram. Here is the time it takes (in seconds): With lock consolidation (btrfs): real Average: 319.6, Median: 320.0, Stddev: 0.8944271909999159 user Average: 6894.2, Median: 6895.0, Stddev: 25.528415540334656 sys Average: 521.4, Median: 522.0, Stddev: 1.51657508881031 Without lock consolidation (btrfs): real Average: 319.8, Median: 320.0, Stddev: 0.8366600265340756 user Average: 6896.6, Median: 6899.0, Stddev: 16.04057355583023 sys Average: 520.6, Median: 521.0, Stddev: 1.140175425099138 With lock consolidation (ext4): real Average: 320.0, Median: 319.0, Stddev: 1.4142135623730951 user Average: 6896.8, Median: 6878.0, Stddev: 28.621670111997307 sys Average: 521.2, Median: 521.0, Stddev: 1.7888543819998317 Without lock consolidation (ext4) real Average: 319.6, Median: 319.0, Stddev: 0.8944271909999159 user Average: 6886.2, Median: 6887.0, Stddev: 16.93221781102523 sys Average: 520.4, Median: 520.0, Stddev: 1.140175425099138 The difference is entirely within the noise of a typical run on zram. This hardly justifies the complexity of maintaining both the pool lock and the class lock. In fact, for writeback, we would need to introduce yet another lock to prevent data races on the pool's LRU, further complicating the lock handling logic. IMHO, it is just better to collapse all of these into a single pool-level lock. Link: https://lkml.kernel.org/r/20221128191616.1261026-4-nphamcs@gmail.com Signed-off-by: Nhat Pham <nphamcs@gmail.com> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Acked-by: Minchan Kim <minchan@kernel.org> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-11-28 19:16:12 +00:00
* migration cannot move any zpages in this zspage. Here, pool->lock
zsmalloc: replace per zpage lock with pool->migrate_lock The zsmalloc has used a bit for spin_lock in zpage handle to keep zpage object alive during several operations. However, it causes the problem for PREEMPT_RT as well as introducing too complicated. This patch replaces the bit spin_lock with pool->migrate_lock rwlock. It could make the code simple as well as zsmalloc work under PREEMPT_RT. The drawback is the pool->migrate_lock is bigger granuarity than per zpage lock so the contention would be higher than old when both IO-related operations(i.e., zsmalloc, zsfree, zs_[map|unmap]) and compaction(page/zpage migration) are going in parallel(*, the migrate_lock is rwlock and IO related functions are all read side lock so there is no contention). However, the write-side is fast enough(dominant overhead is just page copy) so it wouldn't affect much. If the lock granurity becomes more problem later, we could introduce table locks based on handle as a hash value. Link: https://lkml.kernel.org/r/20211115185909.3949505-9-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:14:13 +00:00
* is too heavy since callers would take some time until they calls
* zs_unmap_object API so delegate the locking from class to zspage
* which is smaller granularity.
*/
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
migrate_read_lock(zspage);
zsmalloc: consolidate zs_pool's migrate_lock and size_class's locks Currently, zsmalloc has a hierarchy of locks, which includes a pool-level migrate_lock, and a lock for each size class. We have to obtain both locks in the hotpath in most cases anyway, except for zs_malloc. This exception will no longer exist when we introduce a LRU into the zs_pool for the new writeback functionality - we will need to obtain a pool-level lock to synchronize LRU handling even in zs_malloc. In preparation for zsmalloc writeback, consolidate these locks into a single pool-level lock, which drastically reduces the complexity of synchronization in zsmalloc. We have also benchmarked the lock consolidation to see the performance effect of this change on zram. First, we ran a synthetic FS workload on a server machine with 36 cores (same machine for all runs), using fs_mark -d ../zram1mnt -s 100000 -n 2500 -t 32 -k before and after for btrfs and ext4 on zram (FS usage is 80%). Here is the result (unit is file/second): With lock consolidation (btrfs): Average: 13520.2, Median: 13531.0, Stddev: 137.5961482019028 Without lock consolidation (btrfs): Average: 13487.2, Median: 13575.0, Stddev: 309.08283679298665 With lock consolidation (ext4): Average: 16824.4, Median: 16839.0, Stddev: 89.97388510006668 Without lock consolidation (ext4) Average: 16958.0, Median: 16986.0, Stddev: 194.7370021336469 As you can see, we observe a 0.3% regression for btrfs, and a 0.9% regression for ext4. This is a small, barely measurable difference in my opinion. For a more realistic scenario, we also tries building the kernel on zram. Here is the time it takes (in seconds): With lock consolidation (btrfs): real Average: 319.6, Median: 320.0, Stddev: 0.8944271909999159 user Average: 6894.2, Median: 6895.0, Stddev: 25.528415540334656 sys Average: 521.4, Median: 522.0, Stddev: 1.51657508881031 Without lock consolidation (btrfs): real Average: 319.8, Median: 320.0, Stddev: 0.8366600265340756 user Average: 6896.6, Median: 6899.0, Stddev: 16.04057355583023 sys Average: 520.6, Median: 521.0, Stddev: 1.140175425099138 With lock consolidation (ext4): real Average: 320.0, Median: 319.0, Stddev: 1.4142135623730951 user Average: 6896.8, Median: 6878.0, Stddev: 28.621670111997307 sys Average: 521.2, Median: 521.0, Stddev: 1.7888543819998317 Without lock consolidation (ext4) real Average: 319.6, Median: 319.0, Stddev: 0.8944271909999159 user Average: 6886.2, Median: 6887.0, Stddev: 16.93221781102523 sys Average: 520.4, Median: 520.0, Stddev: 1.140175425099138 The difference is entirely within the noise of a typical run on zram. This hardly justifies the complexity of maintaining both the pool lock and the class lock. In fact, for writeback, we would need to introduce yet another lock to prevent data races on the pool's LRU, further complicating the lock handling logic. IMHO, it is just better to collapse all of these into a single pool-level lock. Link: https://lkml.kernel.org/r/20221128191616.1261026-4-nphamcs@gmail.com Signed-off-by: Nhat Pham <nphamcs@gmail.com> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Acked-by: Minchan Kim <minchan@kernel.org> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-11-28 19:16:12 +00:00
spin_unlock(&pool->lock);
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
zsmalloc: introduce some helper functions Patch series "zsmalloc: remove bit_spin_lock", v2. zsmalloc uses bit_spin_lock to minimize space overhead since it's zpage granularity lock. However, it causes zsmalloc non-working under PREEMPT_RT as well as adding too much complication. This patchset tries to replace the bit_spin_lock with per-pool rwlock. It also removes unnecessary zspage isolation logic from class, which was the other part too much complication added into zsmalloc. Last patch changes the get_cpu_var to local_lock to make it work in PREEMPT_RT. This patch (of 9): get_zspage_mapping returns fullness as well as class_idx. However, the fullness is usually not used since it could be stale in some contexts. It causes misleading as well as unnecessary instructions so this patch introduces zspage_class. obj_to_location also produces page and index but we don't need always the index, either so this patch introduces obj_to_page. Link: https://lkml.kernel.org/r/20211115185909.3949505-1-minchan@kernel.org Link: https://lkml.kernel.org/r/20211115185909.3949505-2-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:13:51 +00:00
class = zspage_class(pool, zspage);
off = (class->size * obj_idx) & ~PAGE_MASK;
local_lock(&zs_map_area.lock);
area = this_cpu_ptr(&zs_map_area);
area->vm_mm = mm;
if (off + class->size <= PAGE_SIZE) {
/* this object is contained entirely within a page */
area->vm_addr = kmap_atomic(page);
zsmalloc: decouple handle and object Recently, we started to use zram heavily and some of issues popped. 1) external fragmentation I got a report from Juneho Choi that fork failed although there are plenty of free pages in the system. His investigation revealed zram is one of the culprit to make heavy fragmentation so there was no more contiguous 16K page for pgd to fork in the ARM. 2) non-movable pages Other problem of zram now is that inherently, user want to use zram as swap in small memory system so they use zRAM with CMA to use memory efficiently. However, unfortunately, it doesn't work well because zRAM cannot use CMA's movable pages unless it doesn't support compaction. I got several reports about that OOM happened with zram although there are lots of swap space and free space in CMA area. 3) internal fragmentation zRAM has started support memory limitation feature to limit memory usage and I sent a patchset(https://lkml.org/lkml/2014/9/21/148) for VM to be harmonized with zram-swap to stop anonymous page reclaim if zram consumed memory up to the limit although there are free space on the swap. One problem for that direction is zram has no way to know any hole in memory space zsmalloc allocated by internal fragmentation so zram would regard swap is full although there are free space in zsmalloc. For solving the issue, zram want to trigger compaction of zsmalloc before it decides full or not. This patchset is first step to support above issues. For that, it adds indirect layer between handle and object location and supports manual compaction to solve 3th problem first of all. After this patchset got merged, next step is to make VM aware of zsmalloc compaction so that generic compaction will move zsmalloced-pages automatically in runtime. In my imaginary experiment(ie, high compress ratio data with heavy swap in/out on 8G zram-swap), data is as follows, Before = zram allocated object : 60212066 bytes zram total used: 140103680 bytes ratio: 42.98 percent MemFree: 840192 kB Compaction After = frag ratio after compaction zram allocated object : 60212066 bytes zram total used: 76185600 bytes ratio: 79.03 percent MemFree: 901932 kB Juneho reported below in his real platform with small aging. So, I think the benefit would be bigger in real aging system for a long time. - frag_ratio increased 3% (ie, higher is better) - memfree increased about 6MB - In buddy info, Normal 2^3: 4, 2^2: 1: 2^1 increased, Highmem: 2^1 21 increased frag ratio after swap fragment used : 156677 kbytes total: 166092 kbytes frag_ratio : 94 meminfo before compaction MemFree: 83724 kB Node 0, zone Normal 13642 1364 57 10 61 17 9 5 4 0 0 Node 0, zone HighMem 425 29 1 0 0 0 0 0 0 0 0 num_migrated : 23630 compaction done frag ratio after compaction used : 156673 kbytes total: 160564 kbytes frag_ratio : 97 meminfo after compaction MemFree: 89060 kB Node 0, zone Normal 14076 1544 67 14 61 17 9 5 4 0 0 Node 0, zone HighMem 863 50 1 0 0 0 0 0 0 0 0 This patchset adds more logics(about 480 lines) in zsmalloc but when I tested heavy swapin/out program, the regression for swapin/out speed is marginal because most of overheads were caused by compress/decompress and other MM reclaim stuff. This patch (of 7): Currently, handle of zsmalloc encodes object's location directly so it makes support of migration hard. This patch decouples handle and object via adding indirect layer. For that, it allocates handle dynamically and returns it to user. The handle is the address allocated by slab allocation so it's unique and we could keep object's location in the memory space allocated for handle. With it, we can change object's position without changing handle itself. Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:23 +00:00
ret = area->vm_addr + off;
goto out;
}
/* this object spans two pages */
pages[0] = page;
pages[1] = get_next_page(page);
BUG_ON(!pages[1]);
zsmalloc: merge size_class to reduce fragmentation zsmalloc has many size_classes to reduce fragmentation and they are in 16 bytes unit, for example, 16, 32, 48, etc., if PAGE_SIZE is 4096. And, zsmalloc has constraint that each zspage has 4 pages at maximum. In this situation, we can see interesting aspect. Let's think about size_class for 1488, 1472, ..., 1376. To prevent external fragmentation, they uses 4 pages per zspage and so all they can contain 11 objects at maximum. 16384 (4096 * 4) = 1488 * 11 + remains 16384 (4096 * 4) = 1472 * 11 + remains 16384 (4096 * 4) = ... 16384 (4096 * 4) = 1376 * 11 + remains It means that they have same characteristics and classification between them isn't needed. If we use one size_class for them, we can reduce fragementation and save some memory since both the 1488 and 1472 sized classes can only fit 11 objects into 4 pages, and an object that's 1472 bytes can fit into an object that's 1488 bytes, merging these classes to always use objects that are 1488 bytes will reduce the total number of size classes. And reducing the total number of size classes reduces overall fragmentation, because a wider range of compressed pages can fit into a single size class, leaving less unused objects in each size class. For this purpose, this patch implement size_class merging. If there is size_class that have same pages_per_zspage and same number of objects per zspage with previous size_class, we don't create new size_class. Instead, we use previous, same characteristic size_class. With this way, above example sizes (1488, 1472, ..., 1376) use just one size_class so we can get much more memory utilization. Below is result of my simple test. TEST ENV: EXT4 on zram, mount with discard option WORKLOAD: untar kernel source code, remove directory in descending order in size. (drivers arch fs sound include net Documentation firmware kernel tools) Each line represents orig_data_size, compr_data_size, mem_used_total, fragmentation overhead (mem_used - compr_data_size) and overhead ratio (overhead to compr_data_size), respectively, after untar and remove operation is executed. * untar-nomerge.out orig_size compr_size used_size overhead overhead_ratio 525.88MB 199.16MB 210.23MB 11.08MB 5.56% 288.32MB 97.43MB 105.63MB 8.20MB 8.41% 177.32MB 61.12MB 69.40MB 8.28MB 13.55% 146.47MB 47.32MB 56.10MB 8.78MB 18.55% 124.16MB 38.85MB 48.41MB 9.55MB 24.58% 103.93MB 31.68MB 40.93MB 9.25MB 29.21% 84.34MB 22.86MB 32.72MB 9.86MB 43.13% 66.87MB 14.83MB 23.83MB 9.00MB 60.70% 60.67MB 11.11MB 18.60MB 7.49MB 67.48% 55.86MB 8.83MB 16.61MB 7.77MB 88.03% 53.32MB 8.01MB 15.32MB 7.31MB 91.24% * untar-merge.out orig_size compr_size used_size overhead overhead_ratio 526.23MB 199.18MB 209.81MB 10.64MB 5.34% 288.68MB 97.45MB 104.08MB 6.63MB 6.80% 177.68MB 61.14MB 66.93MB 5.79MB 9.47% 146.83MB 47.34MB 52.79MB 5.45MB 11.51% 124.52MB 38.87MB 44.30MB 5.43MB 13.96% 104.29MB 31.70MB 36.83MB 5.13MB 16.19% 84.70MB 22.88MB 27.92MB 5.04MB 22.04% 67.11MB 14.83MB 19.26MB 4.43MB 29.86% 60.82MB 11.10MB 14.90MB 3.79MB 34.17% 55.90MB 8.82MB 12.61MB 3.79MB 42.97% 53.32MB 8.01MB 11.73MB 3.73MB 46.53% As you can see above result, merged one has better utilization (overhead ratio, 5th column) and uses less memory (mem_used_total, 3rd column). Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Reviewed-by: Dan Streetman <ddstreet@ieee.org> Cc: Luigi Semenzato <semenzato@google.com> Cc: <juno.choi@lge.com> Cc: "seungho1.park" <seungho1.park@lge.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-12-13 00:56:44 +00:00
zsmalloc: decouple handle and object Recently, we started to use zram heavily and some of issues popped. 1) external fragmentation I got a report from Juneho Choi that fork failed although there are plenty of free pages in the system. His investigation revealed zram is one of the culprit to make heavy fragmentation so there was no more contiguous 16K page for pgd to fork in the ARM. 2) non-movable pages Other problem of zram now is that inherently, user want to use zram as swap in small memory system so they use zRAM with CMA to use memory efficiently. However, unfortunately, it doesn't work well because zRAM cannot use CMA's movable pages unless it doesn't support compaction. I got several reports about that OOM happened with zram although there are lots of swap space and free space in CMA area. 3) internal fragmentation zRAM has started support memory limitation feature to limit memory usage and I sent a patchset(https://lkml.org/lkml/2014/9/21/148) for VM to be harmonized with zram-swap to stop anonymous page reclaim if zram consumed memory up to the limit although there are free space on the swap. One problem for that direction is zram has no way to know any hole in memory space zsmalloc allocated by internal fragmentation so zram would regard swap is full although there are free space in zsmalloc. For solving the issue, zram want to trigger compaction of zsmalloc before it decides full or not. This patchset is first step to support above issues. For that, it adds indirect layer between handle and object location and supports manual compaction to solve 3th problem first of all. After this patchset got merged, next step is to make VM aware of zsmalloc compaction so that generic compaction will move zsmalloced-pages automatically in runtime. In my imaginary experiment(ie, high compress ratio data with heavy swap in/out on 8G zram-swap), data is as follows, Before = zram allocated object : 60212066 bytes zram total used: 140103680 bytes ratio: 42.98 percent MemFree: 840192 kB Compaction After = frag ratio after compaction zram allocated object : 60212066 bytes zram total used: 76185600 bytes ratio: 79.03 percent MemFree: 901932 kB Juneho reported below in his real platform with small aging. So, I think the benefit would be bigger in real aging system for a long time. - frag_ratio increased 3% (ie, higher is better) - memfree increased about 6MB - In buddy info, Normal 2^3: 4, 2^2: 1: 2^1 increased, Highmem: 2^1 21 increased frag ratio after swap fragment used : 156677 kbytes total: 166092 kbytes frag_ratio : 94 meminfo before compaction MemFree: 83724 kB Node 0, zone Normal 13642 1364 57 10 61 17 9 5 4 0 0 Node 0, zone HighMem 425 29 1 0 0 0 0 0 0 0 0 num_migrated : 23630 compaction done frag ratio after compaction used : 156673 kbytes total: 160564 kbytes frag_ratio : 97 meminfo after compaction MemFree: 89060 kB Node 0, zone Normal 14076 1544 67 14 61 17 9 5 4 0 0 Node 0, zone HighMem 863 50 1 0 0 0 0 0 0 0 0 This patchset adds more logics(about 480 lines) in zsmalloc but when I tested heavy swapin/out program, the regression for swapin/out speed is marginal because most of overheads were caused by compress/decompress and other MM reclaim stuff. This patch (of 7): Currently, handle of zsmalloc encodes object's location directly so it makes support of migration hard. This patch decouples handle and object via adding indirect layer. For that, it allocates handle dynamically and returns it to user. The handle is the address allocated by slab allocation so it's unique and we could keep object's location in the memory space allocated for handle. With it, we can change object's position without changing handle itself. Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:23 +00:00
ret = __zs_map_object(area, pages, off, class->size);
out:
if (likely(!ZsHugePage(zspage)))
ret += ZS_HANDLE_SIZE;
return ret;
}
EXPORT_SYMBOL_GPL(zs_map_object);
void zs_unmap_object(struct zs_pool *pool, unsigned long handle)
{
struct zspage *zspage;
struct page *page;
unsigned long obj, off;
unsigned int obj_idx;
struct size_class *class;
struct mapping_area *area;
zsmalloc: merge size_class to reduce fragmentation zsmalloc has many size_classes to reduce fragmentation and they are in 16 bytes unit, for example, 16, 32, 48, etc., if PAGE_SIZE is 4096. And, zsmalloc has constraint that each zspage has 4 pages at maximum. In this situation, we can see interesting aspect. Let's think about size_class for 1488, 1472, ..., 1376. To prevent external fragmentation, they uses 4 pages per zspage and so all they can contain 11 objects at maximum. 16384 (4096 * 4) = 1488 * 11 + remains 16384 (4096 * 4) = 1472 * 11 + remains 16384 (4096 * 4) = ... 16384 (4096 * 4) = 1376 * 11 + remains It means that they have same characteristics and classification between them isn't needed. If we use one size_class for them, we can reduce fragementation and save some memory since both the 1488 and 1472 sized classes can only fit 11 objects into 4 pages, and an object that's 1472 bytes can fit into an object that's 1488 bytes, merging these classes to always use objects that are 1488 bytes will reduce the total number of size classes. And reducing the total number of size classes reduces overall fragmentation, because a wider range of compressed pages can fit into a single size class, leaving less unused objects in each size class. For this purpose, this patch implement size_class merging. If there is size_class that have same pages_per_zspage and same number of objects per zspage with previous size_class, we don't create new size_class. Instead, we use previous, same characteristic size_class. With this way, above example sizes (1488, 1472, ..., 1376) use just one size_class so we can get much more memory utilization. Below is result of my simple test. TEST ENV: EXT4 on zram, mount with discard option WORKLOAD: untar kernel source code, remove directory in descending order in size. (drivers arch fs sound include net Documentation firmware kernel tools) Each line represents orig_data_size, compr_data_size, mem_used_total, fragmentation overhead (mem_used - compr_data_size) and overhead ratio (overhead to compr_data_size), respectively, after untar and remove operation is executed. * untar-nomerge.out orig_size compr_size used_size overhead overhead_ratio 525.88MB 199.16MB 210.23MB 11.08MB 5.56% 288.32MB 97.43MB 105.63MB 8.20MB 8.41% 177.32MB 61.12MB 69.40MB 8.28MB 13.55% 146.47MB 47.32MB 56.10MB 8.78MB 18.55% 124.16MB 38.85MB 48.41MB 9.55MB 24.58% 103.93MB 31.68MB 40.93MB 9.25MB 29.21% 84.34MB 22.86MB 32.72MB 9.86MB 43.13% 66.87MB 14.83MB 23.83MB 9.00MB 60.70% 60.67MB 11.11MB 18.60MB 7.49MB 67.48% 55.86MB 8.83MB 16.61MB 7.77MB 88.03% 53.32MB 8.01MB 15.32MB 7.31MB 91.24% * untar-merge.out orig_size compr_size used_size overhead overhead_ratio 526.23MB 199.18MB 209.81MB 10.64MB 5.34% 288.68MB 97.45MB 104.08MB 6.63MB 6.80% 177.68MB 61.14MB 66.93MB 5.79MB 9.47% 146.83MB 47.34MB 52.79MB 5.45MB 11.51% 124.52MB 38.87MB 44.30MB 5.43MB 13.96% 104.29MB 31.70MB 36.83MB 5.13MB 16.19% 84.70MB 22.88MB 27.92MB 5.04MB 22.04% 67.11MB 14.83MB 19.26MB 4.43MB 29.86% 60.82MB 11.10MB 14.90MB 3.79MB 34.17% 55.90MB 8.82MB 12.61MB 3.79MB 42.97% 53.32MB 8.01MB 11.73MB 3.73MB 46.53% As you can see above result, merged one has better utilization (overhead ratio, 5th column) and uses less memory (mem_used_total, 3rd column). Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Reviewed-by: Dan Streetman <ddstreet@ieee.org> Cc: Luigi Semenzato <semenzato@google.com> Cc: <juno.choi@lge.com> Cc: "seungho1.park" <seungho1.park@lge.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-12-13 00:56:44 +00:00
zsmalloc: decouple handle and object Recently, we started to use zram heavily and some of issues popped. 1) external fragmentation I got a report from Juneho Choi that fork failed although there are plenty of free pages in the system. His investigation revealed zram is one of the culprit to make heavy fragmentation so there was no more contiguous 16K page for pgd to fork in the ARM. 2) non-movable pages Other problem of zram now is that inherently, user want to use zram as swap in small memory system so they use zRAM with CMA to use memory efficiently. However, unfortunately, it doesn't work well because zRAM cannot use CMA's movable pages unless it doesn't support compaction. I got several reports about that OOM happened with zram although there are lots of swap space and free space in CMA area. 3) internal fragmentation zRAM has started support memory limitation feature to limit memory usage and I sent a patchset(https://lkml.org/lkml/2014/9/21/148) for VM to be harmonized with zram-swap to stop anonymous page reclaim if zram consumed memory up to the limit although there are free space on the swap. One problem for that direction is zram has no way to know any hole in memory space zsmalloc allocated by internal fragmentation so zram would regard swap is full although there are free space in zsmalloc. For solving the issue, zram want to trigger compaction of zsmalloc before it decides full or not. This patchset is first step to support above issues. For that, it adds indirect layer between handle and object location and supports manual compaction to solve 3th problem first of all. After this patchset got merged, next step is to make VM aware of zsmalloc compaction so that generic compaction will move zsmalloced-pages automatically in runtime. In my imaginary experiment(ie, high compress ratio data with heavy swap in/out on 8G zram-swap), data is as follows, Before = zram allocated object : 60212066 bytes zram total used: 140103680 bytes ratio: 42.98 percent MemFree: 840192 kB Compaction After = frag ratio after compaction zram allocated object : 60212066 bytes zram total used: 76185600 bytes ratio: 79.03 percent MemFree: 901932 kB Juneho reported below in his real platform with small aging. So, I think the benefit would be bigger in real aging system for a long time. - frag_ratio increased 3% (ie, higher is better) - memfree increased about 6MB - In buddy info, Normal 2^3: 4, 2^2: 1: 2^1 increased, Highmem: 2^1 21 increased frag ratio after swap fragment used : 156677 kbytes total: 166092 kbytes frag_ratio : 94 meminfo before compaction MemFree: 83724 kB Node 0, zone Normal 13642 1364 57 10 61 17 9 5 4 0 0 Node 0, zone HighMem 425 29 1 0 0 0 0 0 0 0 0 num_migrated : 23630 compaction done frag ratio after compaction used : 156673 kbytes total: 160564 kbytes frag_ratio : 97 meminfo after compaction MemFree: 89060 kB Node 0, zone Normal 14076 1544 67 14 61 17 9 5 4 0 0 Node 0, zone HighMem 863 50 1 0 0 0 0 0 0 0 0 This patchset adds more logics(about 480 lines) in zsmalloc but when I tested heavy swapin/out program, the regression for swapin/out speed is marginal because most of overheads were caused by compress/decompress and other MM reclaim stuff. This patch (of 7): Currently, handle of zsmalloc encodes object's location directly so it makes support of migration hard. This patch decouples handle and object via adding indirect layer. For that, it allocates handle dynamically and returns it to user. The handle is the address allocated by slab allocation so it's unique and we could keep object's location in the memory space allocated for handle. With it, we can change object's position without changing handle itself. Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:23 +00:00
obj = handle_to_obj(handle);
obj_to_location(obj, &page, &obj_idx);
zspage = get_zspage(page);
zsmalloc: introduce some helper functions Patch series "zsmalloc: remove bit_spin_lock", v2. zsmalloc uses bit_spin_lock to minimize space overhead since it's zpage granularity lock. However, it causes zsmalloc non-working under PREEMPT_RT as well as adding too much complication. This patchset tries to replace the bit_spin_lock with per-pool rwlock. It also removes unnecessary zspage isolation logic from class, which was the other part too much complication added into zsmalloc. Last patch changes the get_cpu_var to local_lock to make it work in PREEMPT_RT. This patch (of 9): get_zspage_mapping returns fullness as well as class_idx. However, the fullness is usually not used since it could be stale in some contexts. It causes misleading as well as unnecessary instructions so this patch introduces zspage_class. obj_to_location also produces page and index but we don't need always the index, either so this patch introduces obj_to_page. Link: https://lkml.kernel.org/r/20211115185909.3949505-1-minchan@kernel.org Link: https://lkml.kernel.org/r/20211115185909.3949505-2-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:13:51 +00:00
class = zspage_class(pool, zspage);
off = (class->size * obj_idx) & ~PAGE_MASK;
area = this_cpu_ptr(&zs_map_area);
if (off + class->size <= PAGE_SIZE)
kunmap_atomic(area->vm_addr);
else {
struct page *pages[2];
mm/zsmalloc: support allocating obj with size of ZS_MAX_ALLOC_SIZE I sent a patch [1] for unnecessary check in zsmalloc. And Minchan Kim found zsmalloc even does not support allocating an obj with the size of ZS_MAX_ALLOC_SIZE in some situations. For example: In system with 64KB PAGE_SIZE and 32 bit of physical addr. Then: ZS_MIN_ALLOC_SIZE is 32 bytes which is calculated by: MAX(32, (ZS_MAX_PAGES_PER_ZSPAGE << PAGE_SHIFT >> OBJ_INDEX_BITS)) ZS_MAX_ALLOC_SIZE is 64KB(in current code, is PAGE_SIZE) ZS_SIZE_CLASS_DELTA is 256 bytes So, ZS_SIZE_CLASSES = (ZS_MAX_ALLOC_SIZE - ZS_MIN_ALLOC_SIZE) / ZS_SIZE_CLASS_DELTA + 1 = 256 In zs_create_pool(), the max size obj which can be allocated will be: ZS_MIN_ALLOC_SIZE + i * ZS_SIZE_CLASS_DELTA = 32 + 255*256 = 65312 We can see that 65312 < 65536 (ZS_MAX_ALLOC_SIZE). So we can NOT allocate objs with size ZS_MAX_ALLOC_SIZE(65536) which we promise upper users we can do. [1] http://lkml.iu.edu/hypermail/linux/kernel/1411.2/03835.html [2] http://lkml.iu.edu/hypermail/linux/kernel/1411.2/04534.html This patch fixes this issue by dynamiclly calculating zs_size_classes when module is loaded, allocates buffer with size ZS_MAX_ALLOC_SIZE. Then the max obj(size is ZS_MAX_ALLOC_SIZE) can be stored in it. [akpm@linux-foundation.org: restore ZS_SIZE_CLASSES to fix bisectability] Signed-off-by: Mahendran Ganesh <opensource.ganesh@gmail.com> Suggested-by: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-12-13 00:57:01 +00:00
pages[0] = page;
pages[1] = get_next_page(page);
BUG_ON(!pages[1]);
__zs_unmap_object(area, pages, off, class->size);
}
local_unlock(&zs_map_area.lock);
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
migrate_read_unlock(zspage);
}
EXPORT_SYMBOL_GPL(zs_unmap_object);
zsmalloc: introduce zs_huge_class_size() Patch series "zsmalloc/zram: drop zram's max_zpage_size", v3. ZRAM's max_zpage_size is a bad thing. It forces zsmalloc to store normal objects as huge ones, which results in bigger zsmalloc memory usage. Drop it and use actual zsmalloc huge-class value when decide if the object is huge or not. This patch (of 2): Not every object can be share its zspage with other objects, e.g. when the object is as big as zspage or nearly as big a zspage. For such objects zsmalloc has a so called huge class - every object which belongs to huge class consumes the entire zspage (which consists of a physical page). On x86_64, PAGE_SHIFT 12 box, the first non-huge class size is 3264, so starting down from size 3264, objects can share page(-s) and thus minimize memory wastage. ZRAM, however, has its own statically defined watermark for huge objects, namely "3 * PAGE_SIZE / 4 = 3072", and forcibly stores every object larger than this watermark (3072) as a PAGE_SIZE object, in other words, to a huge class, while zsmalloc can keep some of those objects in non-huge classes. This results in increased memory consumption. zsmalloc knows better if the object is huge or not. Introduce zs_huge_class_size() function which tells if the given object can be stored in one of non-huge classes or not. This will let us to drop ZRAM's huge object watermark and fully rely on zsmalloc when we decide if the object is huge. [sergey.senozhatsky.work@gmail.com: add pool param to zs_huge_class_size()] Link: http://lkml.kernel.org/r/20180314081833.1096-2-sergey.senozhatsky@gmail.com Link: http://lkml.kernel.org/r/20180306070639.7389-2-sergey.senozhatsky@gmail.com Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Mike Rapoport <rppt@linux.vnet.ibm.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-04-05 23:24:43 +00:00
/**
* zs_huge_class_size() - Returns the size (in bytes) of the first huge
* zsmalloc &size_class.
* @pool: zsmalloc pool to use
*
* The function returns the size of the first huge class - any object of equal
* or bigger size will be stored in zspage consisting of a single physical
* page.
*
* Context: Any context.
*
* Return: the size (in bytes) of the first huge zsmalloc &size_class.
*/
size_t zs_huge_class_size(struct zs_pool *pool)
{
return huge_class_size;
}
EXPORT_SYMBOL_GPL(zs_huge_class_size);
static unsigned long obj_malloc(struct zs_pool *pool,
struct zspage *zspage, unsigned long handle)
{
int i, nr_page, offset;
unsigned long obj;
struct link_free *link;
struct size_class *class;
struct page *m_page;
unsigned long m_offset;
void *vaddr;
class = pool->size_class[zspage->class];
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
handle |= OBJ_ALLOCATED_TAG;
obj = get_freeobj(zspage);
offset = obj * class->size;
nr_page = offset >> PAGE_SHIFT;
m_offset = offset & ~PAGE_MASK;
m_page = get_first_page(zspage);
for (i = 0; i < nr_page; i++)
m_page = get_next_page(m_page);
vaddr = kmap_atomic(m_page);
link = (struct link_free *)vaddr + m_offset / sizeof(*link);
set_freeobj(zspage, link->next >> OBJ_TAG_BITS);
if (likely(!ZsHugePage(zspage)))
/* record handle in the header of allocated chunk */
link->handle = handle;
else
/* record handle to page->index */
zspage->first_page->index = handle;
kunmap_atomic(vaddr);
mod_zspage_inuse(zspage, 1);
obj = location_to_obj(m_page, obj);
return obj;
}
/**
* zs_malloc - Allocate block of given size from pool.
* @pool: pool to allocate from
* @size: size of block to allocate
* @gfp: gfp flags when allocating object
*
* On success, handle to the allocated object is returned,
* otherwise an ERR_PTR().
* Allocation requests with size > ZS_MAX_ALLOC_SIZE will fail.
*/
unsigned long zs_malloc(struct zs_pool *pool, size_t size, gfp_t gfp)
{
zsmalloc: decouple handle and object Recently, we started to use zram heavily and some of issues popped. 1) external fragmentation I got a report from Juneho Choi that fork failed although there are plenty of free pages in the system. His investigation revealed zram is one of the culprit to make heavy fragmentation so there was no more contiguous 16K page for pgd to fork in the ARM. 2) non-movable pages Other problem of zram now is that inherently, user want to use zram as swap in small memory system so they use zRAM with CMA to use memory efficiently. However, unfortunately, it doesn't work well because zRAM cannot use CMA's movable pages unless it doesn't support compaction. I got several reports about that OOM happened with zram although there are lots of swap space and free space in CMA area. 3) internal fragmentation zRAM has started support memory limitation feature to limit memory usage and I sent a patchset(https://lkml.org/lkml/2014/9/21/148) for VM to be harmonized with zram-swap to stop anonymous page reclaim if zram consumed memory up to the limit although there are free space on the swap. One problem for that direction is zram has no way to know any hole in memory space zsmalloc allocated by internal fragmentation so zram would regard swap is full although there are free space in zsmalloc. For solving the issue, zram want to trigger compaction of zsmalloc before it decides full or not. This patchset is first step to support above issues. For that, it adds indirect layer between handle and object location and supports manual compaction to solve 3th problem first of all. After this patchset got merged, next step is to make VM aware of zsmalloc compaction so that generic compaction will move zsmalloced-pages automatically in runtime. In my imaginary experiment(ie, high compress ratio data with heavy swap in/out on 8G zram-swap), data is as follows, Before = zram allocated object : 60212066 bytes zram total used: 140103680 bytes ratio: 42.98 percent MemFree: 840192 kB Compaction After = frag ratio after compaction zram allocated object : 60212066 bytes zram total used: 76185600 bytes ratio: 79.03 percent MemFree: 901932 kB Juneho reported below in his real platform with small aging. So, I think the benefit would be bigger in real aging system for a long time. - frag_ratio increased 3% (ie, higher is better) - memfree increased about 6MB - In buddy info, Normal 2^3: 4, 2^2: 1: 2^1 increased, Highmem: 2^1 21 increased frag ratio after swap fragment used : 156677 kbytes total: 166092 kbytes frag_ratio : 94 meminfo before compaction MemFree: 83724 kB Node 0, zone Normal 13642 1364 57 10 61 17 9 5 4 0 0 Node 0, zone HighMem 425 29 1 0 0 0 0 0 0 0 0 num_migrated : 23630 compaction done frag ratio after compaction used : 156673 kbytes total: 160564 kbytes frag_ratio : 97 meminfo after compaction MemFree: 89060 kB Node 0, zone Normal 14076 1544 67 14 61 17 9 5 4 0 0 Node 0, zone HighMem 863 50 1 0 0 0 0 0 0 0 0 This patchset adds more logics(about 480 lines) in zsmalloc but when I tested heavy swapin/out program, the regression for swapin/out speed is marginal because most of overheads were caused by compress/decompress and other MM reclaim stuff. This patch (of 7): Currently, handle of zsmalloc encodes object's location directly so it makes support of migration hard. This patch decouples handle and object via adding indirect layer. For that, it allocates handle dynamically and returns it to user. The handle is the address allocated by slab allocation so it's unique and we could keep object's location in the memory space allocated for handle. With it, we can change object's position without changing handle itself. Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:23 +00:00
unsigned long handle, obj;
struct size_class *class;
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
int newfg;
struct zspage *zspage;
if (unlikely(!size || size > ZS_MAX_ALLOC_SIZE))
return (unsigned long)ERR_PTR(-EINVAL);
zsmalloc: decouple handle and object Recently, we started to use zram heavily and some of issues popped. 1) external fragmentation I got a report from Juneho Choi that fork failed although there are plenty of free pages in the system. His investigation revealed zram is one of the culprit to make heavy fragmentation so there was no more contiguous 16K page for pgd to fork in the ARM. 2) non-movable pages Other problem of zram now is that inherently, user want to use zram as swap in small memory system so they use zRAM with CMA to use memory efficiently. However, unfortunately, it doesn't work well because zRAM cannot use CMA's movable pages unless it doesn't support compaction. I got several reports about that OOM happened with zram although there are lots of swap space and free space in CMA area. 3) internal fragmentation zRAM has started support memory limitation feature to limit memory usage and I sent a patchset(https://lkml.org/lkml/2014/9/21/148) for VM to be harmonized with zram-swap to stop anonymous page reclaim if zram consumed memory up to the limit although there are free space on the swap. One problem for that direction is zram has no way to know any hole in memory space zsmalloc allocated by internal fragmentation so zram would regard swap is full although there are free space in zsmalloc. For solving the issue, zram want to trigger compaction of zsmalloc before it decides full or not. This patchset is first step to support above issues. For that, it adds indirect layer between handle and object location and supports manual compaction to solve 3th problem first of all. After this patchset got merged, next step is to make VM aware of zsmalloc compaction so that generic compaction will move zsmalloced-pages automatically in runtime. In my imaginary experiment(ie, high compress ratio data with heavy swap in/out on 8G zram-swap), data is as follows, Before = zram allocated object : 60212066 bytes zram total used: 140103680 bytes ratio: 42.98 percent MemFree: 840192 kB Compaction After = frag ratio after compaction zram allocated object : 60212066 bytes zram total used: 76185600 bytes ratio: 79.03 percent MemFree: 901932 kB Juneho reported below in his real platform with small aging. So, I think the benefit would be bigger in real aging system for a long time. - frag_ratio increased 3% (ie, higher is better) - memfree increased about 6MB - In buddy info, Normal 2^3: 4, 2^2: 1: 2^1 increased, Highmem: 2^1 21 increased frag ratio after swap fragment used : 156677 kbytes total: 166092 kbytes frag_ratio : 94 meminfo before compaction MemFree: 83724 kB Node 0, zone Normal 13642 1364 57 10 61 17 9 5 4 0 0 Node 0, zone HighMem 425 29 1 0 0 0 0 0 0 0 0 num_migrated : 23630 compaction done frag ratio after compaction used : 156673 kbytes total: 160564 kbytes frag_ratio : 97 meminfo after compaction MemFree: 89060 kB Node 0, zone Normal 14076 1544 67 14 61 17 9 5 4 0 0 Node 0, zone HighMem 863 50 1 0 0 0 0 0 0 0 0 This patchset adds more logics(about 480 lines) in zsmalloc but when I tested heavy swapin/out program, the regression for swapin/out speed is marginal because most of overheads were caused by compress/decompress and other MM reclaim stuff. This patch (of 7): Currently, handle of zsmalloc encodes object's location directly so it makes support of migration hard. This patch decouples handle and object via adding indirect layer. For that, it allocates handle dynamically and returns it to user. The handle is the address allocated by slab allocation so it's unique and we could keep object's location in the memory space allocated for handle. With it, we can change object's position without changing handle itself. Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:23 +00:00
handle = cache_alloc_handle(pool, gfp);
zsmalloc: decouple handle and object Recently, we started to use zram heavily and some of issues popped. 1) external fragmentation I got a report from Juneho Choi that fork failed although there are plenty of free pages in the system. His investigation revealed zram is one of the culprit to make heavy fragmentation so there was no more contiguous 16K page for pgd to fork in the ARM. 2) non-movable pages Other problem of zram now is that inherently, user want to use zram as swap in small memory system so they use zRAM with CMA to use memory efficiently. However, unfortunately, it doesn't work well because zRAM cannot use CMA's movable pages unless it doesn't support compaction. I got several reports about that OOM happened with zram although there are lots of swap space and free space in CMA area. 3) internal fragmentation zRAM has started support memory limitation feature to limit memory usage and I sent a patchset(https://lkml.org/lkml/2014/9/21/148) for VM to be harmonized with zram-swap to stop anonymous page reclaim if zram consumed memory up to the limit although there are free space on the swap. One problem for that direction is zram has no way to know any hole in memory space zsmalloc allocated by internal fragmentation so zram would regard swap is full although there are free space in zsmalloc. For solving the issue, zram want to trigger compaction of zsmalloc before it decides full or not. This patchset is first step to support above issues. For that, it adds indirect layer between handle and object location and supports manual compaction to solve 3th problem first of all. After this patchset got merged, next step is to make VM aware of zsmalloc compaction so that generic compaction will move zsmalloced-pages automatically in runtime. In my imaginary experiment(ie, high compress ratio data with heavy swap in/out on 8G zram-swap), data is as follows, Before = zram allocated object : 60212066 bytes zram total used: 140103680 bytes ratio: 42.98 percent MemFree: 840192 kB Compaction After = frag ratio after compaction zram allocated object : 60212066 bytes zram total used: 76185600 bytes ratio: 79.03 percent MemFree: 901932 kB Juneho reported below in his real platform with small aging. So, I think the benefit would be bigger in real aging system for a long time. - frag_ratio increased 3% (ie, higher is better) - memfree increased about 6MB - In buddy info, Normal 2^3: 4, 2^2: 1: 2^1 increased, Highmem: 2^1 21 increased frag ratio after swap fragment used : 156677 kbytes total: 166092 kbytes frag_ratio : 94 meminfo before compaction MemFree: 83724 kB Node 0, zone Normal 13642 1364 57 10 61 17 9 5 4 0 0 Node 0, zone HighMem 425 29 1 0 0 0 0 0 0 0 0 num_migrated : 23630 compaction done frag ratio after compaction used : 156673 kbytes total: 160564 kbytes frag_ratio : 97 meminfo after compaction MemFree: 89060 kB Node 0, zone Normal 14076 1544 67 14 61 17 9 5 4 0 0 Node 0, zone HighMem 863 50 1 0 0 0 0 0 0 0 0 This patchset adds more logics(about 480 lines) in zsmalloc but when I tested heavy swapin/out program, the regression for swapin/out speed is marginal because most of overheads were caused by compress/decompress and other MM reclaim stuff. This patch (of 7): Currently, handle of zsmalloc encodes object's location directly so it makes support of migration hard. This patch decouples handle and object via adding indirect layer. For that, it allocates handle dynamically and returns it to user. The handle is the address allocated by slab allocation so it's unique and we could keep object's location in the memory space allocated for handle. With it, we can change object's position without changing handle itself. Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:23 +00:00
if (!handle)
return (unsigned long)ERR_PTR(-ENOMEM);
zsmalloc: decouple handle and object Recently, we started to use zram heavily and some of issues popped. 1) external fragmentation I got a report from Juneho Choi that fork failed although there are plenty of free pages in the system. His investigation revealed zram is one of the culprit to make heavy fragmentation so there was no more contiguous 16K page for pgd to fork in the ARM. 2) non-movable pages Other problem of zram now is that inherently, user want to use zram as swap in small memory system so they use zRAM with CMA to use memory efficiently. However, unfortunately, it doesn't work well because zRAM cannot use CMA's movable pages unless it doesn't support compaction. I got several reports about that OOM happened with zram although there are lots of swap space and free space in CMA area. 3) internal fragmentation zRAM has started support memory limitation feature to limit memory usage and I sent a patchset(https://lkml.org/lkml/2014/9/21/148) for VM to be harmonized with zram-swap to stop anonymous page reclaim if zram consumed memory up to the limit although there are free space on the swap. One problem for that direction is zram has no way to know any hole in memory space zsmalloc allocated by internal fragmentation so zram would regard swap is full although there are free space in zsmalloc. For solving the issue, zram want to trigger compaction of zsmalloc before it decides full or not. This patchset is first step to support above issues. For that, it adds indirect layer between handle and object location and supports manual compaction to solve 3th problem first of all. After this patchset got merged, next step is to make VM aware of zsmalloc compaction so that generic compaction will move zsmalloced-pages automatically in runtime. In my imaginary experiment(ie, high compress ratio data with heavy swap in/out on 8G zram-swap), data is as follows, Before = zram allocated object : 60212066 bytes zram total used: 140103680 bytes ratio: 42.98 percent MemFree: 840192 kB Compaction After = frag ratio after compaction zram allocated object : 60212066 bytes zram total used: 76185600 bytes ratio: 79.03 percent MemFree: 901932 kB Juneho reported below in his real platform with small aging. So, I think the benefit would be bigger in real aging system for a long time. - frag_ratio increased 3% (ie, higher is better) - memfree increased about 6MB - In buddy info, Normal 2^3: 4, 2^2: 1: 2^1 increased, Highmem: 2^1 21 increased frag ratio after swap fragment used : 156677 kbytes total: 166092 kbytes frag_ratio : 94 meminfo before compaction MemFree: 83724 kB Node 0, zone Normal 13642 1364 57 10 61 17 9 5 4 0 0 Node 0, zone HighMem 425 29 1 0 0 0 0 0 0 0 0 num_migrated : 23630 compaction done frag ratio after compaction used : 156673 kbytes total: 160564 kbytes frag_ratio : 97 meminfo after compaction MemFree: 89060 kB Node 0, zone Normal 14076 1544 67 14 61 17 9 5 4 0 0 Node 0, zone HighMem 863 50 1 0 0 0 0 0 0 0 0 This patchset adds more logics(about 480 lines) in zsmalloc but when I tested heavy swapin/out program, the regression for swapin/out speed is marginal because most of overheads were caused by compress/decompress and other MM reclaim stuff. This patch (of 7): Currently, handle of zsmalloc encodes object's location directly so it makes support of migration hard. This patch decouples handle and object via adding indirect layer. For that, it allocates handle dynamically and returns it to user. The handle is the address allocated by slab allocation so it's unique and we could keep object's location in the memory space allocated for handle. With it, we can change object's position without changing handle itself. Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:23 +00:00
/* extra space in chunk to keep the handle */
size += ZS_HANDLE_SIZE;
zsmalloc: merge size_class to reduce fragmentation zsmalloc has many size_classes to reduce fragmentation and they are in 16 bytes unit, for example, 16, 32, 48, etc., if PAGE_SIZE is 4096. And, zsmalloc has constraint that each zspage has 4 pages at maximum. In this situation, we can see interesting aspect. Let's think about size_class for 1488, 1472, ..., 1376. To prevent external fragmentation, they uses 4 pages per zspage and so all they can contain 11 objects at maximum. 16384 (4096 * 4) = 1488 * 11 + remains 16384 (4096 * 4) = 1472 * 11 + remains 16384 (4096 * 4) = ... 16384 (4096 * 4) = 1376 * 11 + remains It means that they have same characteristics and classification between them isn't needed. If we use one size_class for them, we can reduce fragementation and save some memory since both the 1488 and 1472 sized classes can only fit 11 objects into 4 pages, and an object that's 1472 bytes can fit into an object that's 1488 bytes, merging these classes to always use objects that are 1488 bytes will reduce the total number of size classes. And reducing the total number of size classes reduces overall fragmentation, because a wider range of compressed pages can fit into a single size class, leaving less unused objects in each size class. For this purpose, this patch implement size_class merging. If there is size_class that have same pages_per_zspage and same number of objects per zspage with previous size_class, we don't create new size_class. Instead, we use previous, same characteristic size_class. With this way, above example sizes (1488, 1472, ..., 1376) use just one size_class so we can get much more memory utilization. Below is result of my simple test. TEST ENV: EXT4 on zram, mount with discard option WORKLOAD: untar kernel source code, remove directory in descending order in size. (drivers arch fs sound include net Documentation firmware kernel tools) Each line represents orig_data_size, compr_data_size, mem_used_total, fragmentation overhead (mem_used - compr_data_size) and overhead ratio (overhead to compr_data_size), respectively, after untar and remove operation is executed. * untar-nomerge.out orig_size compr_size used_size overhead overhead_ratio 525.88MB 199.16MB 210.23MB 11.08MB 5.56% 288.32MB 97.43MB 105.63MB 8.20MB 8.41% 177.32MB 61.12MB 69.40MB 8.28MB 13.55% 146.47MB 47.32MB 56.10MB 8.78MB 18.55% 124.16MB 38.85MB 48.41MB 9.55MB 24.58% 103.93MB 31.68MB 40.93MB 9.25MB 29.21% 84.34MB 22.86MB 32.72MB 9.86MB 43.13% 66.87MB 14.83MB 23.83MB 9.00MB 60.70% 60.67MB 11.11MB 18.60MB 7.49MB 67.48% 55.86MB 8.83MB 16.61MB 7.77MB 88.03% 53.32MB 8.01MB 15.32MB 7.31MB 91.24% * untar-merge.out orig_size compr_size used_size overhead overhead_ratio 526.23MB 199.18MB 209.81MB 10.64MB 5.34% 288.68MB 97.45MB 104.08MB 6.63MB 6.80% 177.68MB 61.14MB 66.93MB 5.79MB 9.47% 146.83MB 47.34MB 52.79MB 5.45MB 11.51% 124.52MB 38.87MB 44.30MB 5.43MB 13.96% 104.29MB 31.70MB 36.83MB 5.13MB 16.19% 84.70MB 22.88MB 27.92MB 5.04MB 22.04% 67.11MB 14.83MB 19.26MB 4.43MB 29.86% 60.82MB 11.10MB 14.90MB 3.79MB 34.17% 55.90MB 8.82MB 12.61MB 3.79MB 42.97% 53.32MB 8.01MB 11.73MB 3.73MB 46.53% As you can see above result, merged one has better utilization (overhead ratio, 5th column) and uses less memory (mem_used_total, 3rd column). Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Reviewed-by: Dan Streetman <ddstreet@ieee.org> Cc: Luigi Semenzato <semenzato@google.com> Cc: <juno.choi@lge.com> Cc: "seungho1.park" <seungho1.park@lge.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-12-13 00:56:44 +00:00
class = pool->size_class[get_size_class_index(size)];
zsmalloc: consolidate zs_pool's migrate_lock and size_class's locks Currently, zsmalloc has a hierarchy of locks, which includes a pool-level migrate_lock, and a lock for each size class. We have to obtain both locks in the hotpath in most cases anyway, except for zs_malloc. This exception will no longer exist when we introduce a LRU into the zs_pool for the new writeback functionality - we will need to obtain a pool-level lock to synchronize LRU handling even in zs_malloc. In preparation for zsmalloc writeback, consolidate these locks into a single pool-level lock, which drastically reduces the complexity of synchronization in zsmalloc. We have also benchmarked the lock consolidation to see the performance effect of this change on zram. First, we ran a synthetic FS workload on a server machine with 36 cores (same machine for all runs), using fs_mark -d ../zram1mnt -s 100000 -n 2500 -t 32 -k before and after for btrfs and ext4 on zram (FS usage is 80%). Here is the result (unit is file/second): With lock consolidation (btrfs): Average: 13520.2, Median: 13531.0, Stddev: 137.5961482019028 Without lock consolidation (btrfs): Average: 13487.2, Median: 13575.0, Stddev: 309.08283679298665 With lock consolidation (ext4): Average: 16824.4, Median: 16839.0, Stddev: 89.97388510006668 Without lock consolidation (ext4) Average: 16958.0, Median: 16986.0, Stddev: 194.7370021336469 As you can see, we observe a 0.3% regression for btrfs, and a 0.9% regression for ext4. This is a small, barely measurable difference in my opinion. For a more realistic scenario, we also tries building the kernel on zram. Here is the time it takes (in seconds): With lock consolidation (btrfs): real Average: 319.6, Median: 320.0, Stddev: 0.8944271909999159 user Average: 6894.2, Median: 6895.0, Stddev: 25.528415540334656 sys Average: 521.4, Median: 522.0, Stddev: 1.51657508881031 Without lock consolidation (btrfs): real Average: 319.8, Median: 320.0, Stddev: 0.8366600265340756 user Average: 6896.6, Median: 6899.0, Stddev: 16.04057355583023 sys Average: 520.6, Median: 521.0, Stddev: 1.140175425099138 With lock consolidation (ext4): real Average: 320.0, Median: 319.0, Stddev: 1.4142135623730951 user Average: 6896.8, Median: 6878.0, Stddev: 28.621670111997307 sys Average: 521.2, Median: 521.0, Stddev: 1.7888543819998317 Without lock consolidation (ext4) real Average: 319.6, Median: 319.0, Stddev: 0.8944271909999159 user Average: 6886.2, Median: 6887.0, Stddev: 16.93221781102523 sys Average: 520.4, Median: 520.0, Stddev: 1.140175425099138 The difference is entirely within the noise of a typical run on zram. This hardly justifies the complexity of maintaining both the pool lock and the class lock. In fact, for writeback, we would need to introduce yet another lock to prevent data races on the pool's LRU, further complicating the lock handling logic. IMHO, it is just better to collapse all of these into a single pool-level lock. Link: https://lkml.kernel.org/r/20221128191616.1261026-4-nphamcs@gmail.com Signed-off-by: Nhat Pham <nphamcs@gmail.com> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Acked-by: Minchan Kim <minchan@kernel.org> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-11-28 19:16:12 +00:00
/* pool->lock effectively protects the zpage migration */
spin_lock(&pool->lock);
zspage = find_get_zspage(class);
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
if (likely(zspage)) {
obj = obj_malloc(pool, zspage, handle);
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
/* Now move the zspage to another fullness group, if required */
fix_fullness_group(class, zspage);
record_obj(handle, obj);
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
class_stat_inc(class, ZS_OBJS_INUSE, 1);
zsmalloc: move LRU update from zs_map_object() to zs_malloc() Under memory pressure, we sometimes observe the following crash: [ 5694.832838] ------------[ cut here ]------------ [ 5694.842093] list_del corruption, ffff888014b6a448->next is LIST_POISON1 (dead000000000100) [ 5694.858677] WARNING: CPU: 33 PID: 418824 at lib/list_debug.c:47 __list_del_entry_valid+0x42/0x80 [ 5694.961820] CPU: 33 PID: 418824 Comm: fuse_counters.s Kdump: loaded Tainted: G S 5.19.0-0_fbk3_rc3_hoangnhatpzsdynshrv41_10870_g85a9558a25de #1 [ 5694.990194] Hardware name: Wiwynn Twin Lakes MP/Twin Lakes Passive MP, BIOS YMM16 05/24/2021 [ 5695.007072] RIP: 0010:__list_del_entry_valid+0x42/0x80 [ 5695.017351] Code: 08 48 83 c2 22 48 39 d0 74 24 48 8b 10 48 39 f2 75 2c 48 8b 51 08 b0 01 48 39 f2 75 34 c3 48 c7 c7 55 d7 78 82 e8 4e 45 3b 00 <0f> 0b eb 31 48 c7 c7 27 a8 70 82 e8 3e 45 3b 00 0f 0b eb 21 48 c7 [ 5695.054919] RSP: 0018:ffffc90027aef4f0 EFLAGS: 00010246 [ 5695.065366] RAX: 41fe484987275300 RBX: ffff888008988180 RCX: 0000000000000000 [ 5695.079636] RDX: ffff88886006c280 RSI: ffff888860060480 RDI: ffff888860060480 [ 5695.093904] RBP: 0000000000000002 R08: 0000000000000000 R09: ffffc90027aef370 [ 5695.108175] R10: 0000000000000000 R11: ffffffff82fdf1c0 R12: 0000000010000002 [ 5695.122447] R13: ffff888014b6a448 R14: ffff888014b6a420 R15: 00000000138dc240 [ 5695.136717] FS: 00007f23a7d3f740(0000) GS:ffff888860040000(0000) knlGS:0000000000000000 [ 5695.152899] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 5695.164388] CR2: 0000560ceaab6ac0 CR3: 000000001c06c001 CR4: 00000000007706e0 [ 5695.178659] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 [ 5695.192927] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 [ 5695.207197] PKRU: 55555554 [ 5695.212602] Call Trace: [ 5695.217486] <TASK> [ 5695.221674] zs_map_object+0x91/0x270 [ 5695.229000] zswap_frontswap_store+0x33d/0x870 [ 5695.237885] ? do_raw_spin_lock+0x5d/0xa0 [ 5695.245899] __frontswap_store+0x51/0xb0 [ 5695.253742] swap_writepage+0x3c/0x60 [ 5695.261063] shrink_page_list+0x738/0x1230 [ 5695.269255] shrink_lruvec+0x5ec/0xcd0 [ 5695.276749] ? shrink_slab+0x187/0x5f0 [ 5695.284240] ? mem_cgroup_iter+0x6e/0x120 [ 5695.292255] shrink_node+0x293/0x7b0 [ 5695.299402] do_try_to_free_pages+0xea/0x550 [ 5695.307940] try_to_free_pages+0x19a/0x490 [ 5695.316126] __folio_alloc+0x19ff/0x3e40 [ 5695.323971] ? __filemap_get_folio+0x8a/0x4e0 [ 5695.332681] ? walk_component+0x2a8/0xb50 [ 5695.340697] ? generic_permission+0xda/0x2a0 [ 5695.349231] ? __filemap_get_folio+0x8a/0x4e0 [ 5695.357940] ? walk_component+0x2a8/0xb50 [ 5695.365955] vma_alloc_folio+0x10e/0x570 [ 5695.373796] ? walk_component+0x52/0xb50 [ 5695.381634] wp_page_copy+0x38c/0xc10 [ 5695.388953] ? filename_lookup+0x378/0xbc0 [ 5695.397140] handle_mm_fault+0x87f/0x1800 [ 5695.405157] do_user_addr_fault+0x1bd/0x570 [ 5695.413520] exc_page_fault+0x5d/0x110 [ 5695.421017] asm_exc_page_fault+0x22/0x30 After some investigation, I have found the following issue: unlike other zswap backends, zsmalloc performs the LRU list update at the object mapping time, rather than when the slot for the object is allocated. This deviation was discussed and agreed upon during the review process of the zsmalloc writeback patch series: https://lore.kernel.org/lkml/Y3flcAXNxxrvy3ZH@cmpxchg.org/ Unfortunately, this introduces a subtle bug that occurs when there is a concurrent store and reclaim, which interleave as follows: zswap_frontswap_store() shrink_worker() zs_malloc() zs_zpool_shrink() spin_lock(&pool->lock) zs_reclaim_page() zspage = find_get_zspage() spin_unlock(&pool->lock) spin_lock(&pool->lock) zspage = list_first_entry(&pool->lru) list_del(&zspage->lru) zspage->lru.next = LIST_POISON1 zspage->lru.prev = LIST_POISON2 spin_unlock(&pool->lock) zs_map_object() spin_lock(&pool->lock) if (!list_empty(&zspage->lru)) list_del(&zspage->lru) CHECK_DATA_CORRUPTION(next == LIST_POISON1) /* BOOM */ With the current upstream code, this issue rarely happens. zswap only triggers writeback when the pool is already full, at which point all further store attempts are short-circuited. This creates an implicit pseudo-serialization between reclaim and store. I am working on a new zswap shrinking mechanism, which makes interleaving reclaim and store more likely, exposing this bug. zbud and z3fold do not have this problem, because they perform the LRU list update in the alloc function, while still holding the pool's lock. This patch fixes the aforementioned bug by moving the LRU update back to zs_malloc(), analogous to zbud and z3fold. Link: https://lkml.kernel.org/r/20230505185054.2417128-1-nphamcs@gmail.com Fixes: 64f768c6b32e ("zsmalloc: add a LRU to zs_pool to keep track of zspages in LRU order") Signed-off-by: Nhat Pham <nphamcs@gmail.com> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-05-05 18:50:54 +00:00
goto out;
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
}
mm/zsmalloc: add statistics support Keeping fragmentation of zsmalloc in a low level is our target. But now we still need to add the debug code in zsmalloc to get the quantitative data. This patch adds a new configuration CONFIG_ZSMALLOC_STAT to enable the statistics collection for developers. Currently only the objects statatitics in each class are collected. User can get the information via debugfs. cat /sys/kernel/debug/zsmalloc/zram0/... For example: After I copied "jdk-8u25-linux-x64.tar.gz" to zram with ext4 filesystem: class size obj_allocated obj_used pages_used 0 32 0 0 0 1 48 256 12 3 2 64 64 14 1 3 80 51 7 1 4 96 128 5 3 5 112 73 5 2 6 128 32 4 1 7 144 0 0 0 8 160 0 0 0 9 176 0 0 0 10 192 0 0 0 11 208 0 0 0 12 224 0 0 0 13 240 0 0 0 14 256 16 1 1 15 272 15 9 1 16 288 0 0 0 17 304 0 0 0 18 320 0 0 0 19 336 0 0 0 20 352 0 0 0 21 368 0 0 0 22 384 0 0 0 23 400 0 0 0 24 416 0 0 0 25 432 0 0 0 26 448 0 0 0 27 464 0 0 0 28 480 0 0 0 29 496 33 1 4 30 512 0 0 0 31 528 0 0 0 32 544 0 0 0 33 560 0 0 0 34 576 0 0 0 35 592 0 0 0 36 608 0 0 0 37 624 0 0 0 38 640 0 0 0 40 672 0 0 0 42 704 0 0 0 43 720 17 1 3 44 736 0 0 0 46 768 0 0 0 49 816 0 0 0 51 848 0 0 0 52 864 14 1 3 54 896 0 0 0 57 944 13 1 3 58 960 0 0 0 62 1024 4 1 1 66 1088 15 2 4 67 1104 0 0 0 71 1168 0 0 0 74 1216 0 0 0 76 1248 0 0 0 83 1360 3 1 1 91 1488 11 1 4 94 1536 0 0 0 100 1632 5 1 2 107 1744 0 0 0 111 1808 9 1 4 126 2048 4 4 2 144 2336 7 3 4 151 2448 0 0 0 168 2720 15 15 10 190 3072 28 27 21 202 3264 0 0 0 254 4096 36209 36209 36209 Total 37022 36326 36288 We can calculate the overall fragentation by the last line: Total 37022 36326 36288 (37022 - 36326) / 37022 = 1.87% Also by analysing objects alocated in every class we know why we got so low fragmentation: Most of the allocated objects is in <class 254>. And there is only 1 page in class 254 zspage. So, No fragmentation will be introduced by allocating objs in class 254. And in future, we can collect other zsmalloc statistics as we need and analyse them. Signed-off-by: Ganesh Mahendran <opensource.ganesh@gmail.com> Suggested-by: Minchan Kim <minchan@kernel.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Dan Streetman <ddstreet@ieee.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-12 23:00:54 +00:00
zsmalloc: consolidate zs_pool's migrate_lock and size_class's locks Currently, zsmalloc has a hierarchy of locks, which includes a pool-level migrate_lock, and a lock for each size class. We have to obtain both locks in the hotpath in most cases anyway, except for zs_malloc. This exception will no longer exist when we introduce a LRU into the zs_pool for the new writeback functionality - we will need to obtain a pool-level lock to synchronize LRU handling even in zs_malloc. In preparation for zsmalloc writeback, consolidate these locks into a single pool-level lock, which drastically reduces the complexity of synchronization in zsmalloc. We have also benchmarked the lock consolidation to see the performance effect of this change on zram. First, we ran a synthetic FS workload on a server machine with 36 cores (same machine for all runs), using fs_mark -d ../zram1mnt -s 100000 -n 2500 -t 32 -k before and after for btrfs and ext4 on zram (FS usage is 80%). Here is the result (unit is file/second): With lock consolidation (btrfs): Average: 13520.2, Median: 13531.0, Stddev: 137.5961482019028 Without lock consolidation (btrfs): Average: 13487.2, Median: 13575.0, Stddev: 309.08283679298665 With lock consolidation (ext4): Average: 16824.4, Median: 16839.0, Stddev: 89.97388510006668 Without lock consolidation (ext4) Average: 16958.0, Median: 16986.0, Stddev: 194.7370021336469 As you can see, we observe a 0.3% regression for btrfs, and a 0.9% regression for ext4. This is a small, barely measurable difference in my opinion. For a more realistic scenario, we also tries building the kernel on zram. Here is the time it takes (in seconds): With lock consolidation (btrfs): real Average: 319.6, Median: 320.0, Stddev: 0.8944271909999159 user Average: 6894.2, Median: 6895.0, Stddev: 25.528415540334656 sys Average: 521.4, Median: 522.0, Stddev: 1.51657508881031 Without lock consolidation (btrfs): real Average: 319.8, Median: 320.0, Stddev: 0.8366600265340756 user Average: 6896.6, Median: 6899.0, Stddev: 16.04057355583023 sys Average: 520.6, Median: 521.0, Stddev: 1.140175425099138 With lock consolidation (ext4): real Average: 320.0, Median: 319.0, Stddev: 1.4142135623730951 user Average: 6896.8, Median: 6878.0, Stddev: 28.621670111997307 sys Average: 521.2, Median: 521.0, Stddev: 1.7888543819998317 Without lock consolidation (ext4) real Average: 319.6, Median: 319.0, Stddev: 0.8944271909999159 user Average: 6886.2, Median: 6887.0, Stddev: 16.93221781102523 sys Average: 520.4, Median: 520.0, Stddev: 1.140175425099138 The difference is entirely within the noise of a typical run on zram. This hardly justifies the complexity of maintaining both the pool lock and the class lock. In fact, for writeback, we would need to introduce yet another lock to prevent data races on the pool's LRU, further complicating the lock handling logic. IMHO, it is just better to collapse all of these into a single pool-level lock. Link: https://lkml.kernel.org/r/20221128191616.1261026-4-nphamcs@gmail.com Signed-off-by: Nhat Pham <nphamcs@gmail.com> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Acked-by: Minchan Kim <minchan@kernel.org> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-11-28 19:16:12 +00:00
spin_unlock(&pool->lock);
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
zspage = alloc_zspage(pool, class, gfp);
if (!zspage) {
cache_free_handle(pool, handle);
return (unsigned long)ERR_PTR(-ENOMEM);
}
zsmalloc: consolidate zs_pool's migrate_lock and size_class's locks Currently, zsmalloc has a hierarchy of locks, which includes a pool-level migrate_lock, and a lock for each size class. We have to obtain both locks in the hotpath in most cases anyway, except for zs_malloc. This exception will no longer exist when we introduce a LRU into the zs_pool for the new writeback functionality - we will need to obtain a pool-level lock to synchronize LRU handling even in zs_malloc. In preparation for zsmalloc writeback, consolidate these locks into a single pool-level lock, which drastically reduces the complexity of synchronization in zsmalloc. We have also benchmarked the lock consolidation to see the performance effect of this change on zram. First, we ran a synthetic FS workload on a server machine with 36 cores (same machine for all runs), using fs_mark -d ../zram1mnt -s 100000 -n 2500 -t 32 -k before and after for btrfs and ext4 on zram (FS usage is 80%). Here is the result (unit is file/second): With lock consolidation (btrfs): Average: 13520.2, Median: 13531.0, Stddev: 137.5961482019028 Without lock consolidation (btrfs): Average: 13487.2, Median: 13575.0, Stddev: 309.08283679298665 With lock consolidation (ext4): Average: 16824.4, Median: 16839.0, Stddev: 89.97388510006668 Without lock consolidation (ext4) Average: 16958.0, Median: 16986.0, Stddev: 194.7370021336469 As you can see, we observe a 0.3% regression for btrfs, and a 0.9% regression for ext4. This is a small, barely measurable difference in my opinion. For a more realistic scenario, we also tries building the kernel on zram. Here is the time it takes (in seconds): With lock consolidation (btrfs): real Average: 319.6, Median: 320.0, Stddev: 0.8944271909999159 user Average: 6894.2, Median: 6895.0, Stddev: 25.528415540334656 sys Average: 521.4, Median: 522.0, Stddev: 1.51657508881031 Without lock consolidation (btrfs): real Average: 319.8, Median: 320.0, Stddev: 0.8366600265340756 user Average: 6896.6, Median: 6899.0, Stddev: 16.04057355583023 sys Average: 520.6, Median: 521.0, Stddev: 1.140175425099138 With lock consolidation (ext4): real Average: 320.0, Median: 319.0, Stddev: 1.4142135623730951 user Average: 6896.8, Median: 6878.0, Stddev: 28.621670111997307 sys Average: 521.2, Median: 521.0, Stddev: 1.7888543819998317 Without lock consolidation (ext4) real Average: 319.6, Median: 319.0, Stddev: 0.8944271909999159 user Average: 6886.2, Median: 6887.0, Stddev: 16.93221781102523 sys Average: 520.4, Median: 520.0, Stddev: 1.140175425099138 The difference is entirely within the noise of a typical run on zram. This hardly justifies the complexity of maintaining both the pool lock and the class lock. In fact, for writeback, we would need to introduce yet another lock to prevent data races on the pool's LRU, further complicating the lock handling logic. IMHO, it is just better to collapse all of these into a single pool-level lock. Link: https://lkml.kernel.org/r/20221128191616.1261026-4-nphamcs@gmail.com Signed-off-by: Nhat Pham <nphamcs@gmail.com> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Acked-by: Minchan Kim <minchan@kernel.org> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-11-28 19:16:12 +00:00
spin_lock(&pool->lock);
obj = obj_malloc(pool, zspage, handle);
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
newfg = get_fullness_group(class, zspage);
insert_zspage(class, zspage, newfg);
set_zspage_mapping(zspage, class->index, newfg);
zsmalloc: decouple handle and object Recently, we started to use zram heavily and some of issues popped. 1) external fragmentation I got a report from Juneho Choi that fork failed although there are plenty of free pages in the system. His investigation revealed zram is one of the culprit to make heavy fragmentation so there was no more contiguous 16K page for pgd to fork in the ARM. 2) non-movable pages Other problem of zram now is that inherently, user want to use zram as swap in small memory system so they use zRAM with CMA to use memory efficiently. However, unfortunately, it doesn't work well because zRAM cannot use CMA's movable pages unless it doesn't support compaction. I got several reports about that OOM happened with zram although there are lots of swap space and free space in CMA area. 3) internal fragmentation zRAM has started support memory limitation feature to limit memory usage and I sent a patchset(https://lkml.org/lkml/2014/9/21/148) for VM to be harmonized with zram-swap to stop anonymous page reclaim if zram consumed memory up to the limit although there are free space on the swap. One problem for that direction is zram has no way to know any hole in memory space zsmalloc allocated by internal fragmentation so zram would regard swap is full although there are free space in zsmalloc. For solving the issue, zram want to trigger compaction of zsmalloc before it decides full or not. This patchset is first step to support above issues. For that, it adds indirect layer between handle and object location and supports manual compaction to solve 3th problem first of all. After this patchset got merged, next step is to make VM aware of zsmalloc compaction so that generic compaction will move zsmalloced-pages automatically in runtime. In my imaginary experiment(ie, high compress ratio data with heavy swap in/out on 8G zram-swap), data is as follows, Before = zram allocated object : 60212066 bytes zram total used: 140103680 bytes ratio: 42.98 percent MemFree: 840192 kB Compaction After = frag ratio after compaction zram allocated object : 60212066 bytes zram total used: 76185600 bytes ratio: 79.03 percent MemFree: 901932 kB Juneho reported below in his real platform with small aging. So, I think the benefit would be bigger in real aging system for a long time. - frag_ratio increased 3% (ie, higher is better) - memfree increased about 6MB - In buddy info, Normal 2^3: 4, 2^2: 1: 2^1 increased, Highmem: 2^1 21 increased frag ratio after swap fragment used : 156677 kbytes total: 166092 kbytes frag_ratio : 94 meminfo before compaction MemFree: 83724 kB Node 0, zone Normal 13642 1364 57 10 61 17 9 5 4 0 0 Node 0, zone HighMem 425 29 1 0 0 0 0 0 0 0 0 num_migrated : 23630 compaction done frag ratio after compaction used : 156673 kbytes total: 160564 kbytes frag_ratio : 97 meminfo after compaction MemFree: 89060 kB Node 0, zone Normal 14076 1544 67 14 61 17 9 5 4 0 0 Node 0, zone HighMem 863 50 1 0 0 0 0 0 0 0 0 This patchset adds more logics(about 480 lines) in zsmalloc but when I tested heavy swapin/out program, the regression for swapin/out speed is marginal because most of overheads were caused by compress/decompress and other MM reclaim stuff. This patch (of 7): Currently, handle of zsmalloc encodes object's location directly so it makes support of migration hard. This patch decouples handle and object via adding indirect layer. For that, it allocates handle dynamically and returns it to user. The handle is the address allocated by slab allocation so it's unique and we could keep object's location in the memory space allocated for handle. With it, we can change object's position without changing handle itself. Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:23 +00:00
record_obj(handle, obj);
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
atomic_long_add(class->pages_per_zspage, &pool->pages_allocated);
class_stat_inc(class, ZS_OBJS_ALLOCATED, class->objs_per_zspage);
class_stat_inc(class, ZS_OBJS_INUSE, 1);
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
/* We completely set up zspage so mark them as movable */
SetZsPageMovable(pool, zspage);
zsmalloc: move LRU update from zs_map_object() to zs_malloc() Under memory pressure, we sometimes observe the following crash: [ 5694.832838] ------------[ cut here ]------------ [ 5694.842093] list_del corruption, ffff888014b6a448->next is LIST_POISON1 (dead000000000100) [ 5694.858677] WARNING: CPU: 33 PID: 418824 at lib/list_debug.c:47 __list_del_entry_valid+0x42/0x80 [ 5694.961820] CPU: 33 PID: 418824 Comm: fuse_counters.s Kdump: loaded Tainted: G S 5.19.0-0_fbk3_rc3_hoangnhatpzsdynshrv41_10870_g85a9558a25de #1 [ 5694.990194] Hardware name: Wiwynn Twin Lakes MP/Twin Lakes Passive MP, BIOS YMM16 05/24/2021 [ 5695.007072] RIP: 0010:__list_del_entry_valid+0x42/0x80 [ 5695.017351] Code: 08 48 83 c2 22 48 39 d0 74 24 48 8b 10 48 39 f2 75 2c 48 8b 51 08 b0 01 48 39 f2 75 34 c3 48 c7 c7 55 d7 78 82 e8 4e 45 3b 00 <0f> 0b eb 31 48 c7 c7 27 a8 70 82 e8 3e 45 3b 00 0f 0b eb 21 48 c7 [ 5695.054919] RSP: 0018:ffffc90027aef4f0 EFLAGS: 00010246 [ 5695.065366] RAX: 41fe484987275300 RBX: ffff888008988180 RCX: 0000000000000000 [ 5695.079636] RDX: ffff88886006c280 RSI: ffff888860060480 RDI: ffff888860060480 [ 5695.093904] RBP: 0000000000000002 R08: 0000000000000000 R09: ffffc90027aef370 [ 5695.108175] R10: 0000000000000000 R11: ffffffff82fdf1c0 R12: 0000000010000002 [ 5695.122447] R13: ffff888014b6a448 R14: ffff888014b6a420 R15: 00000000138dc240 [ 5695.136717] FS: 00007f23a7d3f740(0000) GS:ffff888860040000(0000) knlGS:0000000000000000 [ 5695.152899] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 5695.164388] CR2: 0000560ceaab6ac0 CR3: 000000001c06c001 CR4: 00000000007706e0 [ 5695.178659] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 [ 5695.192927] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 [ 5695.207197] PKRU: 55555554 [ 5695.212602] Call Trace: [ 5695.217486] <TASK> [ 5695.221674] zs_map_object+0x91/0x270 [ 5695.229000] zswap_frontswap_store+0x33d/0x870 [ 5695.237885] ? do_raw_spin_lock+0x5d/0xa0 [ 5695.245899] __frontswap_store+0x51/0xb0 [ 5695.253742] swap_writepage+0x3c/0x60 [ 5695.261063] shrink_page_list+0x738/0x1230 [ 5695.269255] shrink_lruvec+0x5ec/0xcd0 [ 5695.276749] ? shrink_slab+0x187/0x5f0 [ 5695.284240] ? mem_cgroup_iter+0x6e/0x120 [ 5695.292255] shrink_node+0x293/0x7b0 [ 5695.299402] do_try_to_free_pages+0xea/0x550 [ 5695.307940] try_to_free_pages+0x19a/0x490 [ 5695.316126] __folio_alloc+0x19ff/0x3e40 [ 5695.323971] ? __filemap_get_folio+0x8a/0x4e0 [ 5695.332681] ? walk_component+0x2a8/0xb50 [ 5695.340697] ? generic_permission+0xda/0x2a0 [ 5695.349231] ? __filemap_get_folio+0x8a/0x4e0 [ 5695.357940] ? walk_component+0x2a8/0xb50 [ 5695.365955] vma_alloc_folio+0x10e/0x570 [ 5695.373796] ? walk_component+0x52/0xb50 [ 5695.381634] wp_page_copy+0x38c/0xc10 [ 5695.388953] ? filename_lookup+0x378/0xbc0 [ 5695.397140] handle_mm_fault+0x87f/0x1800 [ 5695.405157] do_user_addr_fault+0x1bd/0x570 [ 5695.413520] exc_page_fault+0x5d/0x110 [ 5695.421017] asm_exc_page_fault+0x22/0x30 After some investigation, I have found the following issue: unlike other zswap backends, zsmalloc performs the LRU list update at the object mapping time, rather than when the slot for the object is allocated. This deviation was discussed and agreed upon during the review process of the zsmalloc writeback patch series: https://lore.kernel.org/lkml/Y3flcAXNxxrvy3ZH@cmpxchg.org/ Unfortunately, this introduces a subtle bug that occurs when there is a concurrent store and reclaim, which interleave as follows: zswap_frontswap_store() shrink_worker() zs_malloc() zs_zpool_shrink() spin_lock(&pool->lock) zs_reclaim_page() zspage = find_get_zspage() spin_unlock(&pool->lock) spin_lock(&pool->lock) zspage = list_first_entry(&pool->lru) list_del(&zspage->lru) zspage->lru.next = LIST_POISON1 zspage->lru.prev = LIST_POISON2 spin_unlock(&pool->lock) zs_map_object() spin_lock(&pool->lock) if (!list_empty(&zspage->lru)) list_del(&zspage->lru) CHECK_DATA_CORRUPTION(next == LIST_POISON1) /* BOOM */ With the current upstream code, this issue rarely happens. zswap only triggers writeback when the pool is already full, at which point all further store attempts are short-circuited. This creates an implicit pseudo-serialization between reclaim and store. I am working on a new zswap shrinking mechanism, which makes interleaving reclaim and store more likely, exposing this bug. zbud and z3fold do not have this problem, because they perform the LRU list update in the alloc function, while still holding the pool's lock. This patch fixes the aforementioned bug by moving the LRU update back to zs_malloc(), analogous to zbud and z3fold. Link: https://lkml.kernel.org/r/20230505185054.2417128-1-nphamcs@gmail.com Fixes: 64f768c6b32e ("zsmalloc: add a LRU to zs_pool to keep track of zspages in LRU order") Signed-off-by: Nhat Pham <nphamcs@gmail.com> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-05-05 18:50:54 +00:00
out:
#ifdef CONFIG_ZPOOL
/* Add/move zspage to beginning of LRU */
if (!list_empty(&zspage->lru))
list_del(&zspage->lru);
list_add(&zspage->lru, &pool->lru);
#endif
zsmalloc: consolidate zs_pool's migrate_lock and size_class's locks Currently, zsmalloc has a hierarchy of locks, which includes a pool-level migrate_lock, and a lock for each size class. We have to obtain both locks in the hotpath in most cases anyway, except for zs_malloc. This exception will no longer exist when we introduce a LRU into the zs_pool for the new writeback functionality - we will need to obtain a pool-level lock to synchronize LRU handling even in zs_malloc. In preparation for zsmalloc writeback, consolidate these locks into a single pool-level lock, which drastically reduces the complexity of synchronization in zsmalloc. We have also benchmarked the lock consolidation to see the performance effect of this change on zram. First, we ran a synthetic FS workload on a server machine with 36 cores (same machine for all runs), using fs_mark -d ../zram1mnt -s 100000 -n 2500 -t 32 -k before and after for btrfs and ext4 on zram (FS usage is 80%). Here is the result (unit is file/second): With lock consolidation (btrfs): Average: 13520.2, Median: 13531.0, Stddev: 137.5961482019028 Without lock consolidation (btrfs): Average: 13487.2, Median: 13575.0, Stddev: 309.08283679298665 With lock consolidation (ext4): Average: 16824.4, Median: 16839.0, Stddev: 89.97388510006668 Without lock consolidation (ext4) Average: 16958.0, Median: 16986.0, Stddev: 194.7370021336469 As you can see, we observe a 0.3% regression for btrfs, and a 0.9% regression for ext4. This is a small, barely measurable difference in my opinion. For a more realistic scenario, we also tries building the kernel on zram. Here is the time it takes (in seconds): With lock consolidation (btrfs): real Average: 319.6, Median: 320.0, Stddev: 0.8944271909999159 user Average: 6894.2, Median: 6895.0, Stddev: 25.528415540334656 sys Average: 521.4, Median: 522.0, Stddev: 1.51657508881031 Without lock consolidation (btrfs): real Average: 319.8, Median: 320.0, Stddev: 0.8366600265340756 user Average: 6896.6, Median: 6899.0, Stddev: 16.04057355583023 sys Average: 520.6, Median: 521.0, Stddev: 1.140175425099138 With lock consolidation (ext4): real Average: 320.0, Median: 319.0, Stddev: 1.4142135623730951 user Average: 6896.8, Median: 6878.0, Stddev: 28.621670111997307 sys Average: 521.2, Median: 521.0, Stddev: 1.7888543819998317 Without lock consolidation (ext4) real Average: 319.6, Median: 319.0, Stddev: 0.8944271909999159 user Average: 6886.2, Median: 6887.0, Stddev: 16.93221781102523 sys Average: 520.4, Median: 520.0, Stddev: 1.140175425099138 The difference is entirely within the noise of a typical run on zram. This hardly justifies the complexity of maintaining both the pool lock and the class lock. In fact, for writeback, we would need to introduce yet another lock to prevent data races on the pool's LRU, further complicating the lock handling logic. IMHO, it is just better to collapse all of these into a single pool-level lock. Link: https://lkml.kernel.org/r/20221128191616.1261026-4-nphamcs@gmail.com Signed-off-by: Nhat Pham <nphamcs@gmail.com> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Acked-by: Minchan Kim <minchan@kernel.org> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-11-28 19:16:12 +00:00
spin_unlock(&pool->lock);
zsmalloc: decouple handle and object Recently, we started to use zram heavily and some of issues popped. 1) external fragmentation I got a report from Juneho Choi that fork failed although there are plenty of free pages in the system. His investigation revealed zram is one of the culprit to make heavy fragmentation so there was no more contiguous 16K page for pgd to fork in the ARM. 2) non-movable pages Other problem of zram now is that inherently, user want to use zram as swap in small memory system so they use zRAM with CMA to use memory efficiently. However, unfortunately, it doesn't work well because zRAM cannot use CMA's movable pages unless it doesn't support compaction. I got several reports about that OOM happened with zram although there are lots of swap space and free space in CMA area. 3) internal fragmentation zRAM has started support memory limitation feature to limit memory usage and I sent a patchset(https://lkml.org/lkml/2014/9/21/148) for VM to be harmonized with zram-swap to stop anonymous page reclaim if zram consumed memory up to the limit although there are free space on the swap. One problem for that direction is zram has no way to know any hole in memory space zsmalloc allocated by internal fragmentation so zram would regard swap is full although there are free space in zsmalloc. For solving the issue, zram want to trigger compaction of zsmalloc before it decides full or not. This patchset is first step to support above issues. For that, it adds indirect layer between handle and object location and supports manual compaction to solve 3th problem first of all. After this patchset got merged, next step is to make VM aware of zsmalloc compaction so that generic compaction will move zsmalloced-pages automatically in runtime. In my imaginary experiment(ie, high compress ratio data with heavy swap in/out on 8G zram-swap), data is as follows, Before = zram allocated object : 60212066 bytes zram total used: 140103680 bytes ratio: 42.98 percent MemFree: 840192 kB Compaction After = frag ratio after compaction zram allocated object : 60212066 bytes zram total used: 76185600 bytes ratio: 79.03 percent MemFree: 901932 kB Juneho reported below in his real platform with small aging. So, I think the benefit would be bigger in real aging system for a long time. - frag_ratio increased 3% (ie, higher is better) - memfree increased about 6MB - In buddy info, Normal 2^3: 4, 2^2: 1: 2^1 increased, Highmem: 2^1 21 increased frag ratio after swap fragment used : 156677 kbytes total: 166092 kbytes frag_ratio : 94 meminfo before compaction MemFree: 83724 kB Node 0, zone Normal 13642 1364 57 10 61 17 9 5 4 0 0 Node 0, zone HighMem 425 29 1 0 0 0 0 0 0 0 0 num_migrated : 23630 compaction done frag ratio after compaction used : 156673 kbytes total: 160564 kbytes frag_ratio : 97 meminfo after compaction MemFree: 89060 kB Node 0, zone Normal 14076 1544 67 14 61 17 9 5 4 0 0 Node 0, zone HighMem 863 50 1 0 0 0 0 0 0 0 0 This patchset adds more logics(about 480 lines) in zsmalloc but when I tested heavy swapin/out program, the regression for swapin/out speed is marginal because most of overheads were caused by compress/decompress and other MM reclaim stuff. This patch (of 7): Currently, handle of zsmalloc encodes object's location directly so it makes support of migration hard. This patch decouples handle and object via adding indirect layer. For that, it allocates handle dynamically and returns it to user. The handle is the address allocated by slab allocation so it's unique and we could keep object's location in the memory space allocated for handle. With it, we can change object's position without changing handle itself. Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:23 +00:00
return handle;
}
EXPORT_SYMBOL_GPL(zs_malloc);
zsmalloc: fix a race with deferred_handles storing Currently, there is a race between zs_free() and zs_reclaim_page(): zs_reclaim_page() finds a handle to an allocated object, but before the eviction happens, an independent zs_free() call to the same handle could come in and overwrite the object value stored at the handle with the last deferred handle. When zs_reclaim_page() finally gets to call the eviction handler, it will see an invalid object value (i.e the previous deferred handle instead of the original object value). This race happens quite infrequently. We only managed to produce it with out-of-tree developmental code that triggers zsmalloc writeback with a much higher frequency than usual. This patch fixes this race by storing the deferred handle in the object header instead. We differentiate the deferred handle from the other two cases (handle for allocated object, and linkage for free object) with a new tag. If zspage reclamation succeeds, we will free these deferred handles by walking through the zspage objects. On the other hand, if zspage reclamation fails, we reconstruct the zspage freelist (with the deferred handle tag and allocated tag) before trying again with the reclamation. [arnd@arndb.de: avoid unused-function warning] Link: https://lkml.kernel.org/r/20230117170507.2651972-1-arnd@kernel.org Link: https://lkml.kernel.org/r/20230110231701.326724-1-nphamcs@gmail.com Fixes: 9997bc017549 ("zsmalloc: implement writeback mechanism for zsmalloc") Signed-off-by: Nhat Pham <nphamcs@gmail.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-01-10 23:17:01 +00:00
static void obj_free(int class_size, unsigned long obj, unsigned long *handle)
{
struct link_free *link;
struct zspage *zspage;
struct page *f_page;
unsigned long f_offset;
unsigned int f_objidx;
void *vaddr;
zsmalloc: decouple handle and object Recently, we started to use zram heavily and some of issues popped. 1) external fragmentation I got a report from Juneho Choi that fork failed although there are plenty of free pages in the system. His investigation revealed zram is one of the culprit to make heavy fragmentation so there was no more contiguous 16K page for pgd to fork in the ARM. 2) non-movable pages Other problem of zram now is that inherently, user want to use zram as swap in small memory system so they use zRAM with CMA to use memory efficiently. However, unfortunately, it doesn't work well because zRAM cannot use CMA's movable pages unless it doesn't support compaction. I got several reports about that OOM happened with zram although there are lots of swap space and free space in CMA area. 3) internal fragmentation zRAM has started support memory limitation feature to limit memory usage and I sent a patchset(https://lkml.org/lkml/2014/9/21/148) for VM to be harmonized with zram-swap to stop anonymous page reclaim if zram consumed memory up to the limit although there are free space on the swap. One problem for that direction is zram has no way to know any hole in memory space zsmalloc allocated by internal fragmentation so zram would regard swap is full although there are free space in zsmalloc. For solving the issue, zram want to trigger compaction of zsmalloc before it decides full or not. This patchset is first step to support above issues. For that, it adds indirect layer between handle and object location and supports manual compaction to solve 3th problem first of all. After this patchset got merged, next step is to make VM aware of zsmalloc compaction so that generic compaction will move zsmalloced-pages automatically in runtime. In my imaginary experiment(ie, high compress ratio data with heavy swap in/out on 8G zram-swap), data is as follows, Before = zram allocated object : 60212066 bytes zram total used: 140103680 bytes ratio: 42.98 percent MemFree: 840192 kB Compaction After = frag ratio after compaction zram allocated object : 60212066 bytes zram total used: 76185600 bytes ratio: 79.03 percent MemFree: 901932 kB Juneho reported below in his real platform with small aging. So, I think the benefit would be bigger in real aging system for a long time. - frag_ratio increased 3% (ie, higher is better) - memfree increased about 6MB - In buddy info, Normal 2^3: 4, 2^2: 1: 2^1 increased, Highmem: 2^1 21 increased frag ratio after swap fragment used : 156677 kbytes total: 166092 kbytes frag_ratio : 94 meminfo before compaction MemFree: 83724 kB Node 0, zone Normal 13642 1364 57 10 61 17 9 5 4 0 0 Node 0, zone HighMem 425 29 1 0 0 0 0 0 0 0 0 num_migrated : 23630 compaction done frag ratio after compaction used : 156673 kbytes total: 160564 kbytes frag_ratio : 97 meminfo after compaction MemFree: 89060 kB Node 0, zone Normal 14076 1544 67 14 61 17 9 5 4 0 0 Node 0, zone HighMem 863 50 1 0 0 0 0 0 0 0 0 This patchset adds more logics(about 480 lines) in zsmalloc but when I tested heavy swapin/out program, the regression for swapin/out speed is marginal because most of overheads were caused by compress/decompress and other MM reclaim stuff. This patch (of 7): Currently, handle of zsmalloc encodes object's location directly so it makes support of migration hard. This patch decouples handle and object via adding indirect layer. For that, it allocates handle dynamically and returns it to user. The handle is the address allocated by slab allocation so it's unique and we could keep object's location in the memory space allocated for handle. With it, we can change object's position without changing handle itself. Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:23 +00:00
obj_to_location(obj, &f_page, &f_objidx);
f_offset = (class_size * f_objidx) & ~PAGE_MASK;
zspage = get_zspage(f_page);
vaddr = kmap_atomic(f_page);
link = (struct link_free *)(vaddr + f_offset);
zsmalloc: fix a race with deferred_handles storing Currently, there is a race between zs_free() and zs_reclaim_page(): zs_reclaim_page() finds a handle to an allocated object, but before the eviction happens, an independent zs_free() call to the same handle could come in and overwrite the object value stored at the handle with the last deferred handle. When zs_reclaim_page() finally gets to call the eviction handler, it will see an invalid object value (i.e the previous deferred handle instead of the original object value). This race happens quite infrequently. We only managed to produce it with out-of-tree developmental code that triggers zsmalloc writeback with a much higher frequency than usual. This patch fixes this race by storing the deferred handle in the object header instead. We differentiate the deferred handle from the other two cases (handle for allocated object, and linkage for free object) with a new tag. If zspage reclamation succeeds, we will free these deferred handles by walking through the zspage objects. On the other hand, if zspage reclamation fails, we reconstruct the zspage freelist (with the deferred handle tag and allocated tag) before trying again with the reclamation. [arnd@arndb.de: avoid unused-function warning] Link: https://lkml.kernel.org/r/20230117170507.2651972-1-arnd@kernel.org Link: https://lkml.kernel.org/r/20230110231701.326724-1-nphamcs@gmail.com Fixes: 9997bc017549 ("zsmalloc: implement writeback mechanism for zsmalloc") Signed-off-by: Nhat Pham <nphamcs@gmail.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-01-10 23:17:01 +00:00
if (handle) {
#ifdef CONFIG_ZPOOL
/* Stores the (deferred) handle in the object's header */
*handle |= OBJ_DEFERRED_HANDLE_TAG;
*handle &= ~OBJ_ALLOCATED_TAG;
if (likely(!ZsHugePage(zspage)))
link->deferred_handle = *handle;
else
f_page->index = *handle;
#endif
} else {
/* Insert this object in containing zspage's freelist */
if (likely(!ZsHugePage(zspage)))
link->next = get_freeobj(zspage) << OBJ_TAG_BITS;
else
f_page->index = 0;
set_freeobj(zspage, f_objidx);
}
kunmap_atomic(vaddr);
mod_zspage_inuse(zspage, -1);
}
void zs_free(struct zs_pool *pool, unsigned long handle)
{
struct zspage *zspage;
struct page *f_page;
unsigned long obj;
struct size_class *class;
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
int fullness;
if (IS_ERR_OR_NULL((void *)handle))
return;
zsmalloc: replace per zpage lock with pool->migrate_lock The zsmalloc has used a bit for spin_lock in zpage handle to keep zpage object alive during several operations. However, it causes the problem for PREEMPT_RT as well as introducing too complicated. This patch replaces the bit spin_lock with pool->migrate_lock rwlock. It could make the code simple as well as zsmalloc work under PREEMPT_RT. The drawback is the pool->migrate_lock is bigger granuarity than per zpage lock so the contention would be higher than old when both IO-related operations(i.e., zsmalloc, zsfree, zs_[map|unmap]) and compaction(page/zpage migration) are going in parallel(*, the migrate_lock is rwlock and IO related functions are all read side lock so there is no contention). However, the write-side is fast enough(dominant overhead is just page copy) so it wouldn't affect much. If the lock granurity becomes more problem later, we could introduce table locks based on handle as a hash value. Link: https://lkml.kernel.org/r/20211115185909.3949505-9-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:14:13 +00:00
/*
zsmalloc: consolidate zs_pool's migrate_lock and size_class's locks Currently, zsmalloc has a hierarchy of locks, which includes a pool-level migrate_lock, and a lock for each size class. We have to obtain both locks in the hotpath in most cases anyway, except for zs_malloc. This exception will no longer exist when we introduce a LRU into the zs_pool for the new writeback functionality - we will need to obtain a pool-level lock to synchronize LRU handling even in zs_malloc. In preparation for zsmalloc writeback, consolidate these locks into a single pool-level lock, which drastically reduces the complexity of synchronization in zsmalloc. We have also benchmarked the lock consolidation to see the performance effect of this change on zram. First, we ran a synthetic FS workload on a server machine with 36 cores (same machine for all runs), using fs_mark -d ../zram1mnt -s 100000 -n 2500 -t 32 -k before and after for btrfs and ext4 on zram (FS usage is 80%). Here is the result (unit is file/second): With lock consolidation (btrfs): Average: 13520.2, Median: 13531.0, Stddev: 137.5961482019028 Without lock consolidation (btrfs): Average: 13487.2, Median: 13575.0, Stddev: 309.08283679298665 With lock consolidation (ext4): Average: 16824.4, Median: 16839.0, Stddev: 89.97388510006668 Without lock consolidation (ext4) Average: 16958.0, Median: 16986.0, Stddev: 194.7370021336469 As you can see, we observe a 0.3% regression for btrfs, and a 0.9% regression for ext4. This is a small, barely measurable difference in my opinion. For a more realistic scenario, we also tries building the kernel on zram. Here is the time it takes (in seconds): With lock consolidation (btrfs): real Average: 319.6, Median: 320.0, Stddev: 0.8944271909999159 user Average: 6894.2, Median: 6895.0, Stddev: 25.528415540334656 sys Average: 521.4, Median: 522.0, Stddev: 1.51657508881031 Without lock consolidation (btrfs): real Average: 319.8, Median: 320.0, Stddev: 0.8366600265340756 user Average: 6896.6, Median: 6899.0, Stddev: 16.04057355583023 sys Average: 520.6, Median: 521.0, Stddev: 1.140175425099138 With lock consolidation (ext4): real Average: 320.0, Median: 319.0, Stddev: 1.4142135623730951 user Average: 6896.8, Median: 6878.0, Stddev: 28.621670111997307 sys Average: 521.2, Median: 521.0, Stddev: 1.7888543819998317 Without lock consolidation (ext4) real Average: 319.6, Median: 319.0, Stddev: 0.8944271909999159 user Average: 6886.2, Median: 6887.0, Stddev: 16.93221781102523 sys Average: 520.4, Median: 520.0, Stddev: 1.140175425099138 The difference is entirely within the noise of a typical run on zram. This hardly justifies the complexity of maintaining both the pool lock and the class lock. In fact, for writeback, we would need to introduce yet another lock to prevent data races on the pool's LRU, further complicating the lock handling logic. IMHO, it is just better to collapse all of these into a single pool-level lock. Link: https://lkml.kernel.org/r/20221128191616.1261026-4-nphamcs@gmail.com Signed-off-by: Nhat Pham <nphamcs@gmail.com> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Acked-by: Minchan Kim <minchan@kernel.org> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-11-28 19:16:12 +00:00
* The pool->lock protects the race with zpage's migration
zsmalloc: replace per zpage lock with pool->migrate_lock The zsmalloc has used a bit for spin_lock in zpage handle to keep zpage object alive during several operations. However, it causes the problem for PREEMPT_RT as well as introducing too complicated. This patch replaces the bit spin_lock with pool->migrate_lock rwlock. It could make the code simple as well as zsmalloc work under PREEMPT_RT. The drawback is the pool->migrate_lock is bigger granuarity than per zpage lock so the contention would be higher than old when both IO-related operations(i.e., zsmalloc, zsfree, zs_[map|unmap]) and compaction(page/zpage migration) are going in parallel(*, the migrate_lock is rwlock and IO related functions are all read side lock so there is no contention). However, the write-side is fast enough(dominant overhead is just page copy) so it wouldn't affect much. If the lock granurity becomes more problem later, we could introduce table locks based on handle as a hash value. Link: https://lkml.kernel.org/r/20211115185909.3949505-9-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:14:13 +00:00
* so it's safe to get the page from handle.
*/
zsmalloc: consolidate zs_pool's migrate_lock and size_class's locks Currently, zsmalloc has a hierarchy of locks, which includes a pool-level migrate_lock, and a lock for each size class. We have to obtain both locks in the hotpath in most cases anyway, except for zs_malloc. This exception will no longer exist when we introduce a LRU into the zs_pool for the new writeback functionality - we will need to obtain a pool-level lock to synchronize LRU handling even in zs_malloc. In preparation for zsmalloc writeback, consolidate these locks into a single pool-level lock, which drastically reduces the complexity of synchronization in zsmalloc. We have also benchmarked the lock consolidation to see the performance effect of this change on zram. First, we ran a synthetic FS workload on a server machine with 36 cores (same machine for all runs), using fs_mark -d ../zram1mnt -s 100000 -n 2500 -t 32 -k before and after for btrfs and ext4 on zram (FS usage is 80%). Here is the result (unit is file/second): With lock consolidation (btrfs): Average: 13520.2, Median: 13531.0, Stddev: 137.5961482019028 Without lock consolidation (btrfs): Average: 13487.2, Median: 13575.0, Stddev: 309.08283679298665 With lock consolidation (ext4): Average: 16824.4, Median: 16839.0, Stddev: 89.97388510006668 Without lock consolidation (ext4) Average: 16958.0, Median: 16986.0, Stddev: 194.7370021336469 As you can see, we observe a 0.3% regression for btrfs, and a 0.9% regression for ext4. This is a small, barely measurable difference in my opinion. For a more realistic scenario, we also tries building the kernel on zram. Here is the time it takes (in seconds): With lock consolidation (btrfs): real Average: 319.6, Median: 320.0, Stddev: 0.8944271909999159 user Average: 6894.2, Median: 6895.0, Stddev: 25.528415540334656 sys Average: 521.4, Median: 522.0, Stddev: 1.51657508881031 Without lock consolidation (btrfs): real Average: 319.8, Median: 320.0, Stddev: 0.8366600265340756 user Average: 6896.6, Median: 6899.0, Stddev: 16.04057355583023 sys Average: 520.6, Median: 521.0, Stddev: 1.140175425099138 With lock consolidation (ext4): real Average: 320.0, Median: 319.0, Stddev: 1.4142135623730951 user Average: 6896.8, Median: 6878.0, Stddev: 28.621670111997307 sys Average: 521.2, Median: 521.0, Stddev: 1.7888543819998317 Without lock consolidation (ext4) real Average: 319.6, Median: 319.0, Stddev: 0.8944271909999159 user Average: 6886.2, Median: 6887.0, Stddev: 16.93221781102523 sys Average: 520.4, Median: 520.0, Stddev: 1.140175425099138 The difference is entirely within the noise of a typical run on zram. This hardly justifies the complexity of maintaining both the pool lock and the class lock. In fact, for writeback, we would need to introduce yet another lock to prevent data races on the pool's LRU, further complicating the lock handling logic. IMHO, it is just better to collapse all of these into a single pool-level lock. Link: https://lkml.kernel.org/r/20221128191616.1261026-4-nphamcs@gmail.com Signed-off-by: Nhat Pham <nphamcs@gmail.com> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Acked-by: Minchan Kim <minchan@kernel.org> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-11-28 19:16:12 +00:00
spin_lock(&pool->lock);
obj = handle_to_obj(handle);
zsmalloc: introduce some helper functions Patch series "zsmalloc: remove bit_spin_lock", v2. zsmalloc uses bit_spin_lock to minimize space overhead since it's zpage granularity lock. However, it causes zsmalloc non-working under PREEMPT_RT as well as adding too much complication. This patchset tries to replace the bit_spin_lock with per-pool rwlock. It also removes unnecessary zspage isolation logic from class, which was the other part too much complication added into zsmalloc. Last patch changes the get_cpu_var to local_lock to make it work in PREEMPT_RT. This patch (of 9): get_zspage_mapping returns fullness as well as class_idx. However, the fullness is usually not used since it could be stale in some contexts. It causes misleading as well as unnecessary instructions so this patch introduces zspage_class. obj_to_location also produces page and index but we don't need always the index, either so this patch introduces obj_to_page. Link: https://lkml.kernel.org/r/20211115185909.3949505-1-minchan@kernel.org Link: https://lkml.kernel.org/r/20211115185909.3949505-2-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:13:51 +00:00
obj_to_page(obj, &f_page);
zspage = get_zspage(f_page);
zsmalloc: introduce some helper functions Patch series "zsmalloc: remove bit_spin_lock", v2. zsmalloc uses bit_spin_lock to minimize space overhead since it's zpage granularity lock. However, it causes zsmalloc non-working under PREEMPT_RT as well as adding too much complication. This patchset tries to replace the bit_spin_lock with per-pool rwlock. It also removes unnecessary zspage isolation logic from class, which was the other part too much complication added into zsmalloc. Last patch changes the get_cpu_var to local_lock to make it work in PREEMPT_RT. This patch (of 9): get_zspage_mapping returns fullness as well as class_idx. However, the fullness is usually not used since it could be stale in some contexts. It causes misleading as well as unnecessary instructions so this patch introduces zspage_class. obj_to_location also produces page and index but we don't need always the index, either so this patch introduces obj_to_page. Link: https://lkml.kernel.org/r/20211115185909.3949505-1-minchan@kernel.org Link: https://lkml.kernel.org/r/20211115185909.3949505-2-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:13:51 +00:00
class = zspage_class(pool, zspage);
zsmalloc: replace per zpage lock with pool->migrate_lock The zsmalloc has used a bit for spin_lock in zpage handle to keep zpage object alive during several operations. However, it causes the problem for PREEMPT_RT as well as introducing too complicated. This patch replaces the bit spin_lock with pool->migrate_lock rwlock. It could make the code simple as well as zsmalloc work under PREEMPT_RT. The drawback is the pool->migrate_lock is bigger granuarity than per zpage lock so the contention would be higher than old when both IO-related operations(i.e., zsmalloc, zsfree, zs_[map|unmap]) and compaction(page/zpage migration) are going in parallel(*, the migrate_lock is rwlock and IO related functions are all read side lock so there is no contention). However, the write-side is fast enough(dominant overhead is just page copy) so it wouldn't affect much. If the lock granurity becomes more problem later, we could introduce table locks based on handle as a hash value. Link: https://lkml.kernel.org/r/20211115185909.3949505-9-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:14:13 +00:00
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
class_stat_dec(class, ZS_OBJS_INUSE, 1);
#ifdef CONFIG_ZPOOL
if (zspage->under_reclaim) {
/*
* Reclaim needs the handles during writeback. It'll free
* them along with the zspage when it's done with them.
*
zsmalloc: fix a race with deferred_handles storing Currently, there is a race between zs_free() and zs_reclaim_page(): zs_reclaim_page() finds a handle to an allocated object, but before the eviction happens, an independent zs_free() call to the same handle could come in and overwrite the object value stored at the handle with the last deferred handle. When zs_reclaim_page() finally gets to call the eviction handler, it will see an invalid object value (i.e the previous deferred handle instead of the original object value). This race happens quite infrequently. We only managed to produce it with out-of-tree developmental code that triggers zsmalloc writeback with a much higher frequency than usual. This patch fixes this race by storing the deferred handle in the object header instead. We differentiate the deferred handle from the other two cases (handle for allocated object, and linkage for free object) with a new tag. If zspage reclamation succeeds, we will free these deferred handles by walking through the zspage objects. On the other hand, if zspage reclamation fails, we reconstruct the zspage freelist (with the deferred handle tag and allocated tag) before trying again with the reclamation. [arnd@arndb.de: avoid unused-function warning] Link: https://lkml.kernel.org/r/20230117170507.2651972-1-arnd@kernel.org Link: https://lkml.kernel.org/r/20230110231701.326724-1-nphamcs@gmail.com Fixes: 9997bc017549 ("zsmalloc: implement writeback mechanism for zsmalloc") Signed-off-by: Nhat Pham <nphamcs@gmail.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-01-10 23:17:01 +00:00
* Record current deferred handle in the object's header.
*/
zsmalloc: fix a race with deferred_handles storing Currently, there is a race between zs_free() and zs_reclaim_page(): zs_reclaim_page() finds a handle to an allocated object, but before the eviction happens, an independent zs_free() call to the same handle could come in and overwrite the object value stored at the handle with the last deferred handle. When zs_reclaim_page() finally gets to call the eviction handler, it will see an invalid object value (i.e the previous deferred handle instead of the original object value). This race happens quite infrequently. We only managed to produce it with out-of-tree developmental code that triggers zsmalloc writeback with a much higher frequency than usual. This patch fixes this race by storing the deferred handle in the object header instead. We differentiate the deferred handle from the other two cases (handle for allocated object, and linkage for free object) with a new tag. If zspage reclamation succeeds, we will free these deferred handles by walking through the zspage objects. On the other hand, if zspage reclamation fails, we reconstruct the zspage freelist (with the deferred handle tag and allocated tag) before trying again with the reclamation. [arnd@arndb.de: avoid unused-function warning] Link: https://lkml.kernel.org/r/20230117170507.2651972-1-arnd@kernel.org Link: https://lkml.kernel.org/r/20230110231701.326724-1-nphamcs@gmail.com Fixes: 9997bc017549 ("zsmalloc: implement writeback mechanism for zsmalloc") Signed-off-by: Nhat Pham <nphamcs@gmail.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-01-10 23:17:01 +00:00
obj_free(class->size, obj, &handle);
spin_unlock(&pool->lock);
return;
}
#endif
zsmalloc: fix a race with deferred_handles storing Currently, there is a race between zs_free() and zs_reclaim_page(): zs_reclaim_page() finds a handle to an allocated object, but before the eviction happens, an independent zs_free() call to the same handle could come in and overwrite the object value stored at the handle with the last deferred handle. When zs_reclaim_page() finally gets to call the eviction handler, it will see an invalid object value (i.e the previous deferred handle instead of the original object value). This race happens quite infrequently. We only managed to produce it with out-of-tree developmental code that triggers zsmalloc writeback with a much higher frequency than usual. This patch fixes this race by storing the deferred handle in the object header instead. We differentiate the deferred handle from the other two cases (handle for allocated object, and linkage for free object) with a new tag. If zspage reclamation succeeds, we will free these deferred handles by walking through the zspage objects. On the other hand, if zspage reclamation fails, we reconstruct the zspage freelist (with the deferred handle tag and allocated tag) before trying again with the reclamation. [arnd@arndb.de: avoid unused-function warning] Link: https://lkml.kernel.org/r/20230117170507.2651972-1-arnd@kernel.org Link: https://lkml.kernel.org/r/20230110231701.326724-1-nphamcs@gmail.com Fixes: 9997bc017549 ("zsmalloc: implement writeback mechanism for zsmalloc") Signed-off-by: Nhat Pham <nphamcs@gmail.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-01-10 23:17:01 +00:00
obj_free(class->size, obj, NULL);
fullness = fix_fullness_group(class, zspage);
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
if (fullness == ZS_INUSE_RATIO_0)
free_zspage(pool, class, zspage);
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
zsmalloc: consolidate zs_pool's migrate_lock and size_class's locks Currently, zsmalloc has a hierarchy of locks, which includes a pool-level migrate_lock, and a lock for each size class. We have to obtain both locks in the hotpath in most cases anyway, except for zs_malloc. This exception will no longer exist when we introduce a LRU into the zs_pool for the new writeback functionality - we will need to obtain a pool-level lock to synchronize LRU handling even in zs_malloc. In preparation for zsmalloc writeback, consolidate these locks into a single pool-level lock, which drastically reduces the complexity of synchronization in zsmalloc. We have also benchmarked the lock consolidation to see the performance effect of this change on zram. First, we ran a synthetic FS workload on a server machine with 36 cores (same machine for all runs), using fs_mark -d ../zram1mnt -s 100000 -n 2500 -t 32 -k before and after for btrfs and ext4 on zram (FS usage is 80%). Here is the result (unit is file/second): With lock consolidation (btrfs): Average: 13520.2, Median: 13531.0, Stddev: 137.5961482019028 Without lock consolidation (btrfs): Average: 13487.2, Median: 13575.0, Stddev: 309.08283679298665 With lock consolidation (ext4): Average: 16824.4, Median: 16839.0, Stddev: 89.97388510006668 Without lock consolidation (ext4) Average: 16958.0, Median: 16986.0, Stddev: 194.7370021336469 As you can see, we observe a 0.3% regression for btrfs, and a 0.9% regression for ext4. This is a small, barely measurable difference in my opinion. For a more realistic scenario, we also tries building the kernel on zram. Here is the time it takes (in seconds): With lock consolidation (btrfs): real Average: 319.6, Median: 320.0, Stddev: 0.8944271909999159 user Average: 6894.2, Median: 6895.0, Stddev: 25.528415540334656 sys Average: 521.4, Median: 522.0, Stddev: 1.51657508881031 Without lock consolidation (btrfs): real Average: 319.8, Median: 320.0, Stddev: 0.8366600265340756 user Average: 6896.6, Median: 6899.0, Stddev: 16.04057355583023 sys Average: 520.6, Median: 521.0, Stddev: 1.140175425099138 With lock consolidation (ext4): real Average: 320.0, Median: 319.0, Stddev: 1.4142135623730951 user Average: 6896.8, Median: 6878.0, Stddev: 28.621670111997307 sys Average: 521.2, Median: 521.0, Stddev: 1.7888543819998317 Without lock consolidation (ext4) real Average: 319.6, Median: 319.0, Stddev: 0.8944271909999159 user Average: 6886.2, Median: 6887.0, Stddev: 16.93221781102523 sys Average: 520.4, Median: 520.0, Stddev: 1.140175425099138 The difference is entirely within the noise of a typical run on zram. This hardly justifies the complexity of maintaining both the pool lock and the class lock. In fact, for writeback, we would need to introduce yet another lock to prevent data races on the pool's LRU, further complicating the lock handling logic. IMHO, it is just better to collapse all of these into a single pool-level lock. Link: https://lkml.kernel.org/r/20221128191616.1261026-4-nphamcs@gmail.com Signed-off-by: Nhat Pham <nphamcs@gmail.com> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Acked-by: Minchan Kim <minchan@kernel.org> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-11-28 19:16:12 +00:00
spin_unlock(&pool->lock);
cache_free_handle(pool, handle);
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
}
EXPORT_SYMBOL_GPL(zs_free);
static void zs_object_copy(struct size_class *class, unsigned long dst,
unsigned long src)
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
{
struct page *s_page, *d_page;
unsigned int s_objidx, d_objidx;
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
unsigned long s_off, d_off;
void *s_addr, *d_addr;
int s_size, d_size, size;
int written = 0;
s_size = d_size = class->size;
obj_to_location(src, &s_page, &s_objidx);
obj_to_location(dst, &d_page, &d_objidx);
s_off = (class->size * s_objidx) & ~PAGE_MASK;
d_off = (class->size * d_objidx) & ~PAGE_MASK;
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
if (s_off + class->size > PAGE_SIZE)
s_size = PAGE_SIZE - s_off;
if (d_off + class->size > PAGE_SIZE)
d_size = PAGE_SIZE - d_off;
s_addr = kmap_atomic(s_page);
d_addr = kmap_atomic(d_page);
while (1) {
size = min(s_size, d_size);
memcpy(d_addr + d_off, s_addr + s_off, size);
written += size;
if (written == class->size)
break;
s_off += size;
s_size -= size;
d_off += size;
d_size -= size;
/*
* Calling kunmap_atomic(d_addr) is necessary. kunmap_atomic()
* calls must occurs in reverse order of calls to kmap_atomic().
* So, to call kunmap_atomic(s_addr) we should first call
* kunmap_atomic(d_addr). For more details see
* Documentation/mm/highmem.rst.
*/
if (s_off >= PAGE_SIZE) {
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
kunmap_atomic(d_addr);
kunmap_atomic(s_addr);
s_page = get_next_page(s_page);
s_addr = kmap_atomic(s_page);
d_addr = kmap_atomic(d_page);
s_size = class->size - written;
s_off = 0;
}
if (d_off >= PAGE_SIZE) {
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
kunmap_atomic(d_addr);
d_page = get_next_page(d_page);
d_addr = kmap_atomic(d_page);
d_size = class->size - written;
d_off = 0;
}
}
kunmap_atomic(d_addr);
kunmap_atomic(s_addr);
}
/*
zsmalloc: fix a race with deferred_handles storing Currently, there is a race between zs_free() and zs_reclaim_page(): zs_reclaim_page() finds a handle to an allocated object, but before the eviction happens, an independent zs_free() call to the same handle could come in and overwrite the object value stored at the handle with the last deferred handle. When zs_reclaim_page() finally gets to call the eviction handler, it will see an invalid object value (i.e the previous deferred handle instead of the original object value). This race happens quite infrequently. We only managed to produce it with out-of-tree developmental code that triggers zsmalloc writeback with a much higher frequency than usual. This patch fixes this race by storing the deferred handle in the object header instead. We differentiate the deferred handle from the other two cases (handle for allocated object, and linkage for free object) with a new tag. If zspage reclamation succeeds, we will free these deferred handles by walking through the zspage objects. On the other hand, if zspage reclamation fails, we reconstruct the zspage freelist (with the deferred handle tag and allocated tag) before trying again with the reclamation. [arnd@arndb.de: avoid unused-function warning] Link: https://lkml.kernel.org/r/20230117170507.2651972-1-arnd@kernel.org Link: https://lkml.kernel.org/r/20230110231701.326724-1-nphamcs@gmail.com Fixes: 9997bc017549 ("zsmalloc: implement writeback mechanism for zsmalloc") Signed-off-by: Nhat Pham <nphamcs@gmail.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-01-10 23:17:01 +00:00
* Find object with a certain tag in zspage from index object and
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
* return handle.
*/
zsmalloc: fix a race with deferred_handles storing Currently, there is a race between zs_free() and zs_reclaim_page(): zs_reclaim_page() finds a handle to an allocated object, but before the eviction happens, an independent zs_free() call to the same handle could come in and overwrite the object value stored at the handle with the last deferred handle. When zs_reclaim_page() finally gets to call the eviction handler, it will see an invalid object value (i.e the previous deferred handle instead of the original object value). This race happens quite infrequently. We only managed to produce it with out-of-tree developmental code that triggers zsmalloc writeback with a much higher frequency than usual. This patch fixes this race by storing the deferred handle in the object header instead. We differentiate the deferred handle from the other two cases (handle for allocated object, and linkage for free object) with a new tag. If zspage reclamation succeeds, we will free these deferred handles by walking through the zspage objects. On the other hand, if zspage reclamation fails, we reconstruct the zspage freelist (with the deferred handle tag and allocated tag) before trying again with the reclamation. [arnd@arndb.de: avoid unused-function warning] Link: https://lkml.kernel.org/r/20230117170507.2651972-1-arnd@kernel.org Link: https://lkml.kernel.org/r/20230110231701.326724-1-nphamcs@gmail.com Fixes: 9997bc017549 ("zsmalloc: implement writeback mechanism for zsmalloc") Signed-off-by: Nhat Pham <nphamcs@gmail.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-01-10 23:17:01 +00:00
static unsigned long find_tagged_obj(struct size_class *class,
struct page *page, int *obj_idx, int tag)
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
{
unsigned int offset;
int index = *obj_idx;
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
unsigned long handle = 0;
void *addr = kmap_atomic(page);
offset = get_first_obj_offset(page);
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
offset += class->size * index;
while (offset < PAGE_SIZE) {
zsmalloc: fix a race with deferred_handles storing Currently, there is a race between zs_free() and zs_reclaim_page(): zs_reclaim_page() finds a handle to an allocated object, but before the eviction happens, an independent zs_free() call to the same handle could come in and overwrite the object value stored at the handle with the last deferred handle. When zs_reclaim_page() finally gets to call the eviction handler, it will see an invalid object value (i.e the previous deferred handle instead of the original object value). This race happens quite infrequently. We only managed to produce it with out-of-tree developmental code that triggers zsmalloc writeback with a much higher frequency than usual. This patch fixes this race by storing the deferred handle in the object header instead. We differentiate the deferred handle from the other two cases (handle for allocated object, and linkage for free object) with a new tag. If zspage reclamation succeeds, we will free these deferred handles by walking through the zspage objects. On the other hand, if zspage reclamation fails, we reconstruct the zspage freelist (with the deferred handle tag and allocated tag) before trying again with the reclamation. [arnd@arndb.de: avoid unused-function warning] Link: https://lkml.kernel.org/r/20230117170507.2651972-1-arnd@kernel.org Link: https://lkml.kernel.org/r/20230110231701.326724-1-nphamcs@gmail.com Fixes: 9997bc017549 ("zsmalloc: implement writeback mechanism for zsmalloc") Signed-off-by: Nhat Pham <nphamcs@gmail.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-01-10 23:17:01 +00:00
if (obj_tagged(page, addr + offset, &handle, tag))
zsmalloc: replace per zpage lock with pool->migrate_lock The zsmalloc has used a bit for spin_lock in zpage handle to keep zpage object alive during several operations. However, it causes the problem for PREEMPT_RT as well as introducing too complicated. This patch replaces the bit spin_lock with pool->migrate_lock rwlock. It could make the code simple as well as zsmalloc work under PREEMPT_RT. The drawback is the pool->migrate_lock is bigger granuarity than per zpage lock so the contention would be higher than old when both IO-related operations(i.e., zsmalloc, zsfree, zs_[map|unmap]) and compaction(page/zpage migration) are going in parallel(*, the migrate_lock is rwlock and IO related functions are all read side lock so there is no contention). However, the write-side is fast enough(dominant overhead is just page copy) so it wouldn't affect much. If the lock granurity becomes more problem later, we could introduce table locks based on handle as a hash value. Link: https://lkml.kernel.org/r/20211115185909.3949505-9-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:14:13 +00:00
break;
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
offset += class->size;
index++;
}
kunmap_atomic(addr);
*obj_idx = index;
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
return handle;
}
zsmalloc: fix a race with deferred_handles storing Currently, there is a race between zs_free() and zs_reclaim_page(): zs_reclaim_page() finds a handle to an allocated object, but before the eviction happens, an independent zs_free() call to the same handle could come in and overwrite the object value stored at the handle with the last deferred handle. When zs_reclaim_page() finally gets to call the eviction handler, it will see an invalid object value (i.e the previous deferred handle instead of the original object value). This race happens quite infrequently. We only managed to produce it with out-of-tree developmental code that triggers zsmalloc writeback with a much higher frequency than usual. This patch fixes this race by storing the deferred handle in the object header instead. We differentiate the deferred handle from the other two cases (handle for allocated object, and linkage for free object) with a new tag. If zspage reclamation succeeds, we will free these deferred handles by walking through the zspage objects. On the other hand, if zspage reclamation fails, we reconstruct the zspage freelist (with the deferred handle tag and allocated tag) before trying again with the reclamation. [arnd@arndb.de: avoid unused-function warning] Link: https://lkml.kernel.org/r/20230117170507.2651972-1-arnd@kernel.org Link: https://lkml.kernel.org/r/20230110231701.326724-1-nphamcs@gmail.com Fixes: 9997bc017549 ("zsmalloc: implement writeback mechanism for zsmalloc") Signed-off-by: Nhat Pham <nphamcs@gmail.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-01-10 23:17:01 +00:00
/*
* Find alloced object in zspage from index object and
* return handle.
*/
static unsigned long find_alloced_obj(struct size_class *class,
struct page *page, int *obj_idx)
{
return find_tagged_obj(class, page, obj_idx, OBJ_ALLOCATED_TAG);
}
#ifdef CONFIG_ZPOOL
/*
* Find object storing a deferred handle in header in zspage from index object
* and return handle.
*/
static unsigned long find_deferred_handle_obj(struct size_class *class,
struct page *page, int *obj_idx)
{
return find_tagged_obj(class, page, obj_idx, OBJ_DEFERRED_HANDLE_TAG);
}
#endif
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
struct zs_compact_control {
/* Source spage for migration which could be a subpage of zspage */
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
struct page *s_page;
/* Destination page for migration which should be a first page
* of zspage. */
struct page *d_page;
/* Starting object index within @s_page which used for live object
* in the subpage. */
int obj_idx;
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
};
zsmalloc: rework compaction algorithm The zsmalloc compaction algorithm has the potential to waste some CPU cycles, particularly when compacting pages within the same fullness group. This is due to the way it selects the head page of the fullness list for source and destination pages, and how it reinserts those pages during each iteration. The algorithm may first use a page as a migration destination and then as a migration source, leading to an unnecessary back-and-forth movement of objects. Consider the following fullness list: PageA PageB PageC PageD PageE During the first iteration, the compaction algorithm will select PageA as the source and PageB as the destination. All of PageA's objects will be moved to PageB, and then PageA will be released while PageB is reinserted into the fullness list. PageB PageC PageD PageE During the next iteration, the compaction algorithm will again select the head of the list as the source and destination, meaning that PageB will now serve as the source and PageC as the destination. This will result in the objects being moved away from PageB, the same objects that were just moved to PageB in the previous iteration. To prevent this avalanche effect, the compaction algorithm should not reinsert the destination page between iterations. By doing so, the most optimal page will continue to be used and its usage ratio will increase, reducing internal fragmentation. The destination page should only be reinserted into the fullness list if: - It becomes full - No source page is available. TEST ==== It's very challenging to reliably test this series. I ended up developing my own synthetic test that has 100% reproducibility. The test generates significan fragmentation (for each size class) and then performs compaction for each class individually and tracks the number of memcpy() in zs_object_copy(), so that we can compare the amount work compaction does on per-class basis. Total amount of work (zram mm_stat objs_moved) ---------------------------------------------- Old fullness grouping, old compaction algorithm: 323977 memcpy() in zs_object_copy(). Old fullness grouping, new compaction algorithm: 262944 memcpy() in zs_object_copy(). New fullness grouping, new compaction algorithm: 213978 memcpy() in zs_object_copy(). Per-class compaction memcpy() comparison (T-test) ------------------------------------------------- x Old fullness grouping, old compaction algorithm + Old fullness grouping, new compaction algorithm N Min Max Median Avg Stddev x 140 349 3513 2461 2314.1214 806.03271 + 140 289 2778 2006 1878.1714 641.02073 Difference at 95.0% confidence -435.95 +/- 170.595 -18.8387% +/- 7.37193% (Student's t, pooled s = 728.216) x Old fullness grouping, old compaction algorithm + New fullness grouping, new compaction algorithm N Min Max Median Avg Stddev x 140 349 3513 2461 2314.1214 806.03271 + 140 226 2279 1644 1528.4143 524.85268 Difference at 95.0% confidence -785.707 +/- 159.331 -33.9527% +/- 6.88516% (Student's t, pooled s = 680.132) Link: https://lkml.kernel.org/r/20230304034835.2082479-4-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:34 +00:00
static void migrate_zspage(struct zs_pool *pool, struct size_class *class,
struct zs_compact_control *cc)
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
{
unsigned long used_obj, free_obj;
unsigned long handle;
struct page *s_page = cc->s_page;
struct page *d_page = cc->d_page;
int obj_idx = cc->obj_idx;
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
while (1) {
handle = find_alloced_obj(class, s_page, &obj_idx);
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
if (!handle) {
s_page = get_next_page(s_page);
if (!s_page)
break;
obj_idx = 0;
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
continue;
}
/* Stop if there is no more space */
zsmalloc: rework compaction algorithm The zsmalloc compaction algorithm has the potential to waste some CPU cycles, particularly when compacting pages within the same fullness group. This is due to the way it selects the head page of the fullness list for source and destination pages, and how it reinserts those pages during each iteration. The algorithm may first use a page as a migration destination and then as a migration source, leading to an unnecessary back-and-forth movement of objects. Consider the following fullness list: PageA PageB PageC PageD PageE During the first iteration, the compaction algorithm will select PageA as the source and PageB as the destination. All of PageA's objects will be moved to PageB, and then PageA will be released while PageB is reinserted into the fullness list. PageB PageC PageD PageE During the next iteration, the compaction algorithm will again select the head of the list as the source and destination, meaning that PageB will now serve as the source and PageC as the destination. This will result in the objects being moved away from PageB, the same objects that were just moved to PageB in the previous iteration. To prevent this avalanche effect, the compaction algorithm should not reinsert the destination page between iterations. By doing so, the most optimal page will continue to be used and its usage ratio will increase, reducing internal fragmentation. The destination page should only be reinserted into the fullness list if: - It becomes full - No source page is available. TEST ==== It's very challenging to reliably test this series. I ended up developing my own synthetic test that has 100% reproducibility. The test generates significan fragmentation (for each size class) and then performs compaction for each class individually and tracks the number of memcpy() in zs_object_copy(), so that we can compare the amount work compaction does on per-class basis. Total amount of work (zram mm_stat objs_moved) ---------------------------------------------- Old fullness grouping, old compaction algorithm: 323977 memcpy() in zs_object_copy(). Old fullness grouping, new compaction algorithm: 262944 memcpy() in zs_object_copy(). New fullness grouping, new compaction algorithm: 213978 memcpy() in zs_object_copy(). Per-class compaction memcpy() comparison (T-test) ------------------------------------------------- x Old fullness grouping, old compaction algorithm + Old fullness grouping, new compaction algorithm N Min Max Median Avg Stddev x 140 349 3513 2461 2314.1214 806.03271 + 140 289 2778 2006 1878.1714 641.02073 Difference at 95.0% confidence -435.95 +/- 170.595 -18.8387% +/- 7.37193% (Student's t, pooled s = 728.216) x Old fullness grouping, old compaction algorithm + New fullness grouping, new compaction algorithm N Min Max Median Avg Stddev x 140 349 3513 2461 2314.1214 806.03271 + 140 226 2279 1644 1528.4143 524.85268 Difference at 95.0% confidence -785.707 +/- 159.331 -33.9527% +/- 6.88516% (Student's t, pooled s = 680.132) Link: https://lkml.kernel.org/r/20230304034835.2082479-4-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:34 +00:00
if (zspage_full(class, get_zspage(d_page)))
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
break;
used_obj = handle_to_obj(handle);
free_obj = obj_malloc(pool, get_zspage(d_page), handle);
zs_object_copy(class, free_obj, used_obj);
obj_idx++;
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
record_obj(handle, free_obj);
zsmalloc: fix a race with deferred_handles storing Currently, there is a race between zs_free() and zs_reclaim_page(): zs_reclaim_page() finds a handle to an allocated object, but before the eviction happens, an independent zs_free() call to the same handle could come in and overwrite the object value stored at the handle with the last deferred handle. When zs_reclaim_page() finally gets to call the eviction handler, it will see an invalid object value (i.e the previous deferred handle instead of the original object value). This race happens quite infrequently. We only managed to produce it with out-of-tree developmental code that triggers zsmalloc writeback with a much higher frequency than usual. This patch fixes this race by storing the deferred handle in the object header instead. We differentiate the deferred handle from the other two cases (handle for allocated object, and linkage for free object) with a new tag. If zspage reclamation succeeds, we will free these deferred handles by walking through the zspage objects. On the other hand, if zspage reclamation fails, we reconstruct the zspage freelist (with the deferred handle tag and allocated tag) before trying again with the reclamation. [arnd@arndb.de: avoid unused-function warning] Link: https://lkml.kernel.org/r/20230117170507.2651972-1-arnd@kernel.org Link: https://lkml.kernel.org/r/20230110231701.326724-1-nphamcs@gmail.com Fixes: 9997bc017549 ("zsmalloc: implement writeback mechanism for zsmalloc") Signed-off-by: Nhat Pham <nphamcs@gmail.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-01-10 23:17:01 +00:00
obj_free(class->size, used_obj, NULL);
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
}
/* Remember last position in this iteration */
cc->s_page = s_page;
cc->obj_idx = obj_idx;
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
}
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
static struct zspage *isolate_src_zspage(struct size_class *class)
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
{
struct zspage *zspage;
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
int fg;
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
for (fg = ZS_INUSE_RATIO_10; fg <= ZS_INUSE_RATIO_99; fg++) {
zspage = list_first_entry_or_null(&class->fullness_list[fg],
struct zspage, list);
if (zspage) {
remove_zspage(class, zspage, fg);
return zspage;
}
}
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
return zspage;
}
static struct zspage *isolate_dst_zspage(struct size_class *class)
{
struct zspage *zspage;
int fg;
for (fg = ZS_INUSE_RATIO_99; fg >= ZS_INUSE_RATIO_10; fg--) {
zspage = list_first_entry_or_null(&class->fullness_list[fg],
struct zspage, list);
if (zspage) {
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
remove_zspage(class, zspage, fg);
return zspage;
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
}
}
return zspage;
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
}
/*
* putback_zspage - add @zspage into right class's fullness list
* @class: destination class
* @zspage: target page
*
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
* Return @zspage's fullness status
*/
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
static int putback_zspage(struct size_class *class, struct zspage *zspage)
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
{
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
int fullness;
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
fullness = get_fullness_group(class, zspage);
insert_zspage(class, zspage, fullness);
set_zspage_mapping(zspage, class->index, fullness);
return fullness;
}
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
#if defined(CONFIG_ZPOOL) || defined(CONFIG_COMPACTION)
/*
* To prevent zspage destroy during migration, zspage freeing should
* hold locks of all pages in the zspage.
*/
static void lock_zspage(struct zspage *zspage)
{
struct page *curr_page, *page;
/*
* Pages we haven't locked yet can be migrated off the list while we're
* trying to lock them, so we need to be careful and only attempt to
* lock each page under migrate_read_lock(). Otherwise, the page we lock
* may no longer belong to the zspage. This means that we may wait for
* the wrong page to unlock, so we must take a reference to the page
* prior to waiting for it to unlock outside migrate_read_lock().
*/
while (1) {
migrate_read_lock(zspage);
page = get_first_page(zspage);
if (trylock_page(page))
break;
get_page(page);
migrate_read_unlock(zspage);
wait_on_page_locked(page);
put_page(page);
}
curr_page = page;
while ((page = get_next_page(curr_page))) {
if (trylock_page(page)) {
curr_page = page;
} else {
get_page(page);
migrate_read_unlock(zspage);
wait_on_page_locked(page);
put_page(page);
migrate_read_lock(zspage);
}
}
migrate_read_unlock(zspage);
}
#endif /* defined(CONFIG_ZPOOL) || defined(CONFIG_COMPACTION) */
#ifdef CONFIG_ZPOOL
/*
* Unlocks all the pages of the zspage.
*
* pool->lock must be held before this function is called
* to prevent the underlying pages from migrating.
*/
static void unlock_zspage(struct zspage *zspage)
{
struct page *page = get_first_page(zspage);
do {
unlock_page(page);
} while ((page = get_next_page(page)) != NULL);
}
#endif /* CONFIG_ZPOOL */
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
static void migrate_lock_init(struct zspage *zspage)
{
rwlock_init(&zspage->lock);
}
static void migrate_read_lock(struct zspage *zspage) __acquires(&zspage->lock)
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
{
read_lock(&zspage->lock);
}
static void migrate_read_unlock(struct zspage *zspage) __releases(&zspage->lock)
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
{
read_unlock(&zspage->lock);
}
#ifdef CONFIG_COMPACTION
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
static void migrate_write_lock(struct zspage *zspage)
{
write_lock(&zspage->lock);
}
zsmalloc: replace per zpage lock with pool->migrate_lock The zsmalloc has used a bit for spin_lock in zpage handle to keep zpage object alive during several operations. However, it causes the problem for PREEMPT_RT as well as introducing too complicated. This patch replaces the bit spin_lock with pool->migrate_lock rwlock. It could make the code simple as well as zsmalloc work under PREEMPT_RT. The drawback is the pool->migrate_lock is bigger granuarity than per zpage lock so the contention would be higher than old when both IO-related operations(i.e., zsmalloc, zsfree, zs_[map|unmap]) and compaction(page/zpage migration) are going in parallel(*, the migrate_lock is rwlock and IO related functions are all read side lock so there is no contention). However, the write-side is fast enough(dominant overhead is just page copy) so it wouldn't affect much. If the lock granurity becomes more problem later, we could introduce table locks based on handle as a hash value. Link: https://lkml.kernel.org/r/20211115185909.3949505-9-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:14:13 +00:00
static void migrate_write_lock_nested(struct zspage *zspage)
{
write_lock_nested(&zspage->lock, SINGLE_DEPTH_NESTING);
}
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
static void migrate_write_unlock(struct zspage *zspage)
{
write_unlock(&zspage->lock);
}
/* Number of isolated subpage for *page migration* in this zspage */
static void inc_zspage_isolation(struct zspage *zspage)
{
zspage->isolated++;
}
static void dec_zspage_isolation(struct zspage *zspage)
{
VM_BUG_ON(zspage->isolated == 0);
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
zspage->isolated--;
}
static const struct movable_operations zsmalloc_mops;
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
static void replace_sub_page(struct size_class *class, struct zspage *zspage,
struct page *newpage, struct page *oldpage)
{
struct page *page;
struct page *pages[ZS_MAX_PAGES_PER_ZSPAGE] = {NULL, };
int idx = 0;
page = get_first_page(zspage);
do {
if (page == oldpage)
pages[idx] = newpage;
else
pages[idx] = page;
idx++;
} while ((page = get_next_page(page)) != NULL);
create_page_chain(class, zspage, pages);
set_first_obj_offset(newpage, get_first_obj_offset(oldpage));
if (unlikely(ZsHugePage(zspage)))
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
newpage->index = oldpage->index;
__SetPageMovable(newpage, &zsmalloc_mops);
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
}
static bool zs_page_isolate(struct page *page, isolate_mode_t mode)
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
{
struct zspage *zspage;
/*
* Page is locked so zspage couldn't be destroyed. For detail, look at
* lock_zspage in free_zspage.
*/
VM_BUG_ON_PAGE(PageIsolated(page), page);
zspage = get_zspage(page);
migrate_write_lock(zspage);
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
inc_zspage_isolation(zspage);
migrate_write_unlock(zspage);
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
return true;
}
static int zs_page_migrate(struct page *newpage, struct page *page,
enum migrate_mode mode)
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
{
struct zs_pool *pool;
struct size_class *class;
struct zspage *zspage;
struct page *dummy;
void *s_addr, *d_addr, *addr;
unsigned int offset;
unsigned long handle;
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
unsigned long old_obj, new_obj;
unsigned int obj_idx;
mm/migrate: new migrate mode MIGRATE_SYNC_NO_COPY Introduce a new migration mode that allow to offload the copy to a device DMA engine. This changes the workflow of migration and not all address_space migratepage callback can support this. This is intended to be use by migrate_vma() which itself is use for thing like HMM (see include/linux/hmm.h). No additional per-filesystem migratepage testing is needed. I disables MIGRATE_SYNC_NO_COPY in all problematic migratepage() callback and i added comment in those to explain why (part of this patch). The commit message is unclear it should say that any callback that wish to support this new mode need to be aware of the difference in the migration flow from other mode. Some of these callbacks do extra locking while copying (aio, zsmalloc, balloon, ...) and for DMA to be effective you want to copy multiple pages in one DMA operations. But in the problematic case you can not easily hold the extra lock accross multiple call to this callback. Usual flow is: For each page { 1 - lock page 2 - call migratepage() callback 3 - (extra locking in some migratepage() callback) 4 - migrate page state (freeze refcount, update page cache, buffer head, ...) 5 - copy page 6 - (unlock any extra lock of migratepage() callback) 7 - return from migratepage() callback 8 - unlock page } The new mode MIGRATE_SYNC_NO_COPY: 1 - lock multiple pages For each page { 2 - call migratepage() callback 3 - abort in all problematic migratepage() callback 4 - migrate page state (freeze refcount, update page cache, buffer head, ...) } // finished all calls to migratepage() callback 5 - DMA copy multiple pages 6 - unlock all the pages To support MIGRATE_SYNC_NO_COPY in the problematic case we would need a new callback migratepages() (for instance) that deals with multiple pages in one transaction. Because the problematic cases are not important for current usage I did not wanted to complexify this patchset even more for no good reason. Link: http://lkml.kernel.org/r/20170817000548.32038-14-jglisse@redhat.com Signed-off-by: Jérôme Glisse <jglisse@redhat.com> Cc: Aneesh Kumar <aneesh.kumar@linux.vnet.ibm.com> Cc: Balbir Singh <bsingharora@gmail.com> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Dan Williams <dan.j.williams@intel.com> Cc: David Nellans <dnellans@nvidia.com> Cc: Evgeny Baskakov <ebaskakov@nvidia.com> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: John Hubbard <jhubbard@nvidia.com> Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Mark Hairgrove <mhairgrove@nvidia.com> Cc: Michal Hocko <mhocko@kernel.org> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Cc: Ross Zwisler <ross.zwisler@linux.intel.com> Cc: Sherry Cheung <SCheung@nvidia.com> Cc: Subhash Gutti <sgutti@nvidia.com> Cc: Vladimir Davydov <vdavydov.dev@gmail.com> Cc: Bob Liu <liubo95@huawei.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-08 23:12:06 +00:00
/*
* We cannot support the _NO_COPY case here, because copy needs to
* happen under the zs lock, which does not work with
* MIGRATE_SYNC_NO_COPY workflow.
*/
if (mode == MIGRATE_SYNC_NO_COPY)
return -EINVAL;
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
VM_BUG_ON_PAGE(!PageIsolated(page), page);
/* The page is locked, so this pointer must remain valid */
zspage = get_zspage(page);
pool = zspage->pool;
zsmalloc: replace per zpage lock with pool->migrate_lock The zsmalloc has used a bit for spin_lock in zpage handle to keep zpage object alive during several operations. However, it causes the problem for PREEMPT_RT as well as introducing too complicated. This patch replaces the bit spin_lock with pool->migrate_lock rwlock. It could make the code simple as well as zsmalloc work under PREEMPT_RT. The drawback is the pool->migrate_lock is bigger granuarity than per zpage lock so the contention would be higher than old when both IO-related operations(i.e., zsmalloc, zsfree, zs_[map|unmap]) and compaction(page/zpage migration) are going in parallel(*, the migrate_lock is rwlock and IO related functions are all read side lock so there is no contention). However, the write-side is fast enough(dominant overhead is just page copy) so it wouldn't affect much. If the lock granurity becomes more problem later, we could introduce table locks based on handle as a hash value. Link: https://lkml.kernel.org/r/20211115185909.3949505-9-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:14:13 +00:00
/*
zsmalloc: consolidate zs_pool's migrate_lock and size_class's locks Currently, zsmalloc has a hierarchy of locks, which includes a pool-level migrate_lock, and a lock for each size class. We have to obtain both locks in the hotpath in most cases anyway, except for zs_malloc. This exception will no longer exist when we introduce a LRU into the zs_pool for the new writeback functionality - we will need to obtain a pool-level lock to synchronize LRU handling even in zs_malloc. In preparation for zsmalloc writeback, consolidate these locks into a single pool-level lock, which drastically reduces the complexity of synchronization in zsmalloc. We have also benchmarked the lock consolidation to see the performance effect of this change on zram. First, we ran a synthetic FS workload on a server machine with 36 cores (same machine for all runs), using fs_mark -d ../zram1mnt -s 100000 -n 2500 -t 32 -k before and after for btrfs and ext4 on zram (FS usage is 80%). Here is the result (unit is file/second): With lock consolidation (btrfs): Average: 13520.2, Median: 13531.0, Stddev: 137.5961482019028 Without lock consolidation (btrfs): Average: 13487.2, Median: 13575.0, Stddev: 309.08283679298665 With lock consolidation (ext4): Average: 16824.4, Median: 16839.0, Stddev: 89.97388510006668 Without lock consolidation (ext4) Average: 16958.0, Median: 16986.0, Stddev: 194.7370021336469 As you can see, we observe a 0.3% regression for btrfs, and a 0.9% regression for ext4. This is a small, barely measurable difference in my opinion. For a more realistic scenario, we also tries building the kernel on zram. Here is the time it takes (in seconds): With lock consolidation (btrfs): real Average: 319.6, Median: 320.0, Stddev: 0.8944271909999159 user Average: 6894.2, Median: 6895.0, Stddev: 25.528415540334656 sys Average: 521.4, Median: 522.0, Stddev: 1.51657508881031 Without lock consolidation (btrfs): real Average: 319.8, Median: 320.0, Stddev: 0.8366600265340756 user Average: 6896.6, Median: 6899.0, Stddev: 16.04057355583023 sys Average: 520.6, Median: 521.0, Stddev: 1.140175425099138 With lock consolidation (ext4): real Average: 320.0, Median: 319.0, Stddev: 1.4142135623730951 user Average: 6896.8, Median: 6878.0, Stddev: 28.621670111997307 sys Average: 521.2, Median: 521.0, Stddev: 1.7888543819998317 Without lock consolidation (ext4) real Average: 319.6, Median: 319.0, Stddev: 0.8944271909999159 user Average: 6886.2, Median: 6887.0, Stddev: 16.93221781102523 sys Average: 520.4, Median: 520.0, Stddev: 1.140175425099138 The difference is entirely within the noise of a typical run on zram. This hardly justifies the complexity of maintaining both the pool lock and the class lock. In fact, for writeback, we would need to introduce yet another lock to prevent data races on the pool's LRU, further complicating the lock handling logic. IMHO, it is just better to collapse all of these into a single pool-level lock. Link: https://lkml.kernel.org/r/20221128191616.1261026-4-nphamcs@gmail.com Signed-off-by: Nhat Pham <nphamcs@gmail.com> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Acked-by: Minchan Kim <minchan@kernel.org> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-11-28 19:16:12 +00:00
* The pool's lock protects the race between zpage migration
zsmalloc: replace per zpage lock with pool->migrate_lock The zsmalloc has used a bit for spin_lock in zpage handle to keep zpage object alive during several operations. However, it causes the problem for PREEMPT_RT as well as introducing too complicated. This patch replaces the bit spin_lock with pool->migrate_lock rwlock. It could make the code simple as well as zsmalloc work under PREEMPT_RT. The drawback is the pool->migrate_lock is bigger granuarity than per zpage lock so the contention would be higher than old when both IO-related operations(i.e., zsmalloc, zsfree, zs_[map|unmap]) and compaction(page/zpage migration) are going in parallel(*, the migrate_lock is rwlock and IO related functions are all read side lock so there is no contention). However, the write-side is fast enough(dominant overhead is just page copy) so it wouldn't affect much. If the lock granurity becomes more problem later, we could introduce table locks based on handle as a hash value. Link: https://lkml.kernel.org/r/20211115185909.3949505-9-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:14:13 +00:00
* and zs_free.
*/
zsmalloc: consolidate zs_pool's migrate_lock and size_class's locks Currently, zsmalloc has a hierarchy of locks, which includes a pool-level migrate_lock, and a lock for each size class. We have to obtain both locks in the hotpath in most cases anyway, except for zs_malloc. This exception will no longer exist when we introduce a LRU into the zs_pool for the new writeback functionality - we will need to obtain a pool-level lock to synchronize LRU handling even in zs_malloc. In preparation for zsmalloc writeback, consolidate these locks into a single pool-level lock, which drastically reduces the complexity of synchronization in zsmalloc. We have also benchmarked the lock consolidation to see the performance effect of this change on zram. First, we ran a synthetic FS workload on a server machine with 36 cores (same machine for all runs), using fs_mark -d ../zram1mnt -s 100000 -n 2500 -t 32 -k before and after for btrfs and ext4 on zram (FS usage is 80%). Here is the result (unit is file/second): With lock consolidation (btrfs): Average: 13520.2, Median: 13531.0, Stddev: 137.5961482019028 Without lock consolidation (btrfs): Average: 13487.2, Median: 13575.0, Stddev: 309.08283679298665 With lock consolidation (ext4): Average: 16824.4, Median: 16839.0, Stddev: 89.97388510006668 Without lock consolidation (ext4) Average: 16958.0, Median: 16986.0, Stddev: 194.7370021336469 As you can see, we observe a 0.3% regression for btrfs, and a 0.9% regression for ext4. This is a small, barely measurable difference in my opinion. For a more realistic scenario, we also tries building the kernel on zram. Here is the time it takes (in seconds): With lock consolidation (btrfs): real Average: 319.6, Median: 320.0, Stddev: 0.8944271909999159 user Average: 6894.2, Median: 6895.0, Stddev: 25.528415540334656 sys Average: 521.4, Median: 522.0, Stddev: 1.51657508881031 Without lock consolidation (btrfs): real Average: 319.8, Median: 320.0, Stddev: 0.8366600265340756 user Average: 6896.6, Median: 6899.0, Stddev: 16.04057355583023 sys Average: 520.6, Median: 521.0, Stddev: 1.140175425099138 With lock consolidation (ext4): real Average: 320.0, Median: 319.0, Stddev: 1.4142135623730951 user Average: 6896.8, Median: 6878.0, Stddev: 28.621670111997307 sys Average: 521.2, Median: 521.0, Stddev: 1.7888543819998317 Without lock consolidation (ext4) real Average: 319.6, Median: 319.0, Stddev: 0.8944271909999159 user Average: 6886.2, Median: 6887.0, Stddev: 16.93221781102523 sys Average: 520.4, Median: 520.0, Stddev: 1.140175425099138 The difference is entirely within the noise of a typical run on zram. This hardly justifies the complexity of maintaining both the pool lock and the class lock. In fact, for writeback, we would need to introduce yet another lock to prevent data races on the pool's LRU, further complicating the lock handling logic. IMHO, it is just better to collapse all of these into a single pool-level lock. Link: https://lkml.kernel.org/r/20221128191616.1261026-4-nphamcs@gmail.com Signed-off-by: Nhat Pham <nphamcs@gmail.com> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Acked-by: Minchan Kim <minchan@kernel.org> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-11-28 19:16:12 +00:00
spin_lock(&pool->lock);
zsmalloc: introduce some helper functions Patch series "zsmalloc: remove bit_spin_lock", v2. zsmalloc uses bit_spin_lock to minimize space overhead since it's zpage granularity lock. However, it causes zsmalloc non-working under PREEMPT_RT as well as adding too much complication. This patchset tries to replace the bit_spin_lock with per-pool rwlock. It also removes unnecessary zspage isolation logic from class, which was the other part too much complication added into zsmalloc. Last patch changes the get_cpu_var to local_lock to make it work in PREEMPT_RT. This patch (of 9): get_zspage_mapping returns fullness as well as class_idx. However, the fullness is usually not used since it could be stale in some contexts. It causes misleading as well as unnecessary instructions so this patch introduces zspage_class. obj_to_location also produces page and index but we don't need always the index, either so this patch introduces obj_to_page. Link: https://lkml.kernel.org/r/20211115185909.3949505-1-minchan@kernel.org Link: https://lkml.kernel.org/r/20211115185909.3949505-2-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:13:51 +00:00
class = zspage_class(pool, zspage);
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
zsmalloc: replace per zpage lock with pool->migrate_lock The zsmalloc has used a bit for spin_lock in zpage handle to keep zpage object alive during several operations. However, it causes the problem for PREEMPT_RT as well as introducing too complicated. This patch replaces the bit spin_lock with pool->migrate_lock rwlock. It could make the code simple as well as zsmalloc work under PREEMPT_RT. The drawback is the pool->migrate_lock is bigger granuarity than per zpage lock so the contention would be higher than old when both IO-related operations(i.e., zsmalloc, zsfree, zs_[map|unmap]) and compaction(page/zpage migration) are going in parallel(*, the migrate_lock is rwlock and IO related functions are all read side lock so there is no contention). However, the write-side is fast enough(dominant overhead is just page copy) so it wouldn't affect much. If the lock granurity becomes more problem later, we could introduce table locks based on handle as a hash value. Link: https://lkml.kernel.org/r/20211115185909.3949505-9-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:14:13 +00:00
/* the migrate_write_lock protects zpage access via zs_map_object */
migrate_write_lock(zspage);
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
zsmalloc: replace per zpage lock with pool->migrate_lock The zsmalloc has used a bit for spin_lock in zpage handle to keep zpage object alive during several operations. However, it causes the problem for PREEMPT_RT as well as introducing too complicated. This patch replaces the bit spin_lock with pool->migrate_lock rwlock. It could make the code simple as well as zsmalloc work under PREEMPT_RT. The drawback is the pool->migrate_lock is bigger granuarity than per zpage lock so the contention would be higher than old when both IO-related operations(i.e., zsmalloc, zsfree, zs_[map|unmap]) and compaction(page/zpage migration) are going in parallel(*, the migrate_lock is rwlock and IO related functions are all read side lock so there is no contention). However, the write-side is fast enough(dominant overhead is just page copy) so it wouldn't affect much. If the lock granurity becomes more problem later, we could introduce table locks based on handle as a hash value. Link: https://lkml.kernel.org/r/20211115185909.3949505-9-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:14:13 +00:00
offset = get_first_obj_offset(page);
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
s_addr = kmap_atomic(page);
/*
* Here, any user cannot access all objects in the zspage so let's move.
*/
d_addr = kmap_atomic(newpage);
memcpy(d_addr, s_addr, PAGE_SIZE);
kunmap_atomic(d_addr);
zsmalloc: replace per zpage lock with pool->migrate_lock The zsmalloc has used a bit for spin_lock in zpage handle to keep zpage object alive during several operations. However, it causes the problem for PREEMPT_RT as well as introducing too complicated. This patch replaces the bit spin_lock with pool->migrate_lock rwlock. It could make the code simple as well as zsmalloc work under PREEMPT_RT. The drawback is the pool->migrate_lock is bigger granuarity than per zpage lock so the contention would be higher than old when both IO-related operations(i.e., zsmalloc, zsfree, zs_[map|unmap]) and compaction(page/zpage migration) are going in parallel(*, the migrate_lock is rwlock and IO related functions are all read side lock so there is no contention). However, the write-side is fast enough(dominant overhead is just page copy) so it wouldn't affect much. If the lock granurity becomes more problem later, we could introduce table locks based on handle as a hash value. Link: https://lkml.kernel.org/r/20211115185909.3949505-9-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:14:13 +00:00
for (addr = s_addr + offset; addr < s_addr + PAGE_SIZE;
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
addr += class->size) {
if (obj_allocated(page, addr, &handle)) {
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
old_obj = handle_to_obj(handle);
obj_to_location(old_obj, &dummy, &obj_idx);
new_obj = (unsigned long)location_to_obj(newpage,
obj_idx);
record_obj(handle, new_obj);
}
}
zsmalloc: replace per zpage lock with pool->migrate_lock The zsmalloc has used a bit for spin_lock in zpage handle to keep zpage object alive during several operations. However, it causes the problem for PREEMPT_RT as well as introducing too complicated. This patch replaces the bit spin_lock with pool->migrate_lock rwlock. It could make the code simple as well as zsmalloc work under PREEMPT_RT. The drawback is the pool->migrate_lock is bigger granuarity than per zpage lock so the contention would be higher than old when both IO-related operations(i.e., zsmalloc, zsfree, zs_[map|unmap]) and compaction(page/zpage migration) are going in parallel(*, the migrate_lock is rwlock and IO related functions are all read side lock so there is no contention). However, the write-side is fast enough(dominant overhead is just page copy) so it wouldn't affect much. If the lock granurity becomes more problem later, we could introduce table locks based on handle as a hash value. Link: https://lkml.kernel.org/r/20211115185909.3949505-9-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:14:13 +00:00
kunmap_atomic(s_addr);
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
replace_sub_page(class, zspage, newpage, page);
zsmalloc: replace per zpage lock with pool->migrate_lock The zsmalloc has used a bit for spin_lock in zpage handle to keep zpage object alive during several operations. However, it causes the problem for PREEMPT_RT as well as introducing too complicated. This patch replaces the bit spin_lock with pool->migrate_lock rwlock. It could make the code simple as well as zsmalloc work under PREEMPT_RT. The drawback is the pool->migrate_lock is bigger granuarity than per zpage lock so the contention would be higher than old when both IO-related operations(i.e., zsmalloc, zsfree, zs_[map|unmap]) and compaction(page/zpage migration) are going in parallel(*, the migrate_lock is rwlock and IO related functions are all read side lock so there is no contention). However, the write-side is fast enough(dominant overhead is just page copy) so it wouldn't affect much. If the lock granurity becomes more problem later, we could introduce table locks based on handle as a hash value. Link: https://lkml.kernel.org/r/20211115185909.3949505-9-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:14:13 +00:00
/*
* Since we complete the data copy and set up new zspage structure,
zsmalloc: consolidate zs_pool's migrate_lock and size_class's locks Currently, zsmalloc has a hierarchy of locks, which includes a pool-level migrate_lock, and a lock for each size class. We have to obtain both locks in the hotpath in most cases anyway, except for zs_malloc. This exception will no longer exist when we introduce a LRU into the zs_pool for the new writeback functionality - we will need to obtain a pool-level lock to synchronize LRU handling even in zs_malloc. In preparation for zsmalloc writeback, consolidate these locks into a single pool-level lock, which drastically reduces the complexity of synchronization in zsmalloc. We have also benchmarked the lock consolidation to see the performance effect of this change on zram. First, we ran a synthetic FS workload on a server machine with 36 cores (same machine for all runs), using fs_mark -d ../zram1mnt -s 100000 -n 2500 -t 32 -k before and after for btrfs and ext4 on zram (FS usage is 80%). Here is the result (unit is file/second): With lock consolidation (btrfs): Average: 13520.2, Median: 13531.0, Stddev: 137.5961482019028 Without lock consolidation (btrfs): Average: 13487.2, Median: 13575.0, Stddev: 309.08283679298665 With lock consolidation (ext4): Average: 16824.4, Median: 16839.0, Stddev: 89.97388510006668 Without lock consolidation (ext4) Average: 16958.0, Median: 16986.0, Stddev: 194.7370021336469 As you can see, we observe a 0.3% regression for btrfs, and a 0.9% regression for ext4. This is a small, barely measurable difference in my opinion. For a more realistic scenario, we also tries building the kernel on zram. Here is the time it takes (in seconds): With lock consolidation (btrfs): real Average: 319.6, Median: 320.0, Stddev: 0.8944271909999159 user Average: 6894.2, Median: 6895.0, Stddev: 25.528415540334656 sys Average: 521.4, Median: 522.0, Stddev: 1.51657508881031 Without lock consolidation (btrfs): real Average: 319.8, Median: 320.0, Stddev: 0.8366600265340756 user Average: 6896.6, Median: 6899.0, Stddev: 16.04057355583023 sys Average: 520.6, Median: 521.0, Stddev: 1.140175425099138 With lock consolidation (ext4): real Average: 320.0, Median: 319.0, Stddev: 1.4142135623730951 user Average: 6896.8, Median: 6878.0, Stddev: 28.621670111997307 sys Average: 521.2, Median: 521.0, Stddev: 1.7888543819998317 Without lock consolidation (ext4) real Average: 319.6, Median: 319.0, Stddev: 0.8944271909999159 user Average: 6886.2, Median: 6887.0, Stddev: 16.93221781102523 sys Average: 520.4, Median: 520.0, Stddev: 1.140175425099138 The difference is entirely within the noise of a typical run on zram. This hardly justifies the complexity of maintaining both the pool lock and the class lock. In fact, for writeback, we would need to introduce yet another lock to prevent data races on the pool's LRU, further complicating the lock handling logic. IMHO, it is just better to collapse all of these into a single pool-level lock. Link: https://lkml.kernel.org/r/20221128191616.1261026-4-nphamcs@gmail.com Signed-off-by: Nhat Pham <nphamcs@gmail.com> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Acked-by: Minchan Kim <minchan@kernel.org> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-11-28 19:16:12 +00:00
* it's okay to release the pool's lock.
zsmalloc: replace per zpage lock with pool->migrate_lock The zsmalloc has used a bit for spin_lock in zpage handle to keep zpage object alive during several operations. However, it causes the problem for PREEMPT_RT as well as introducing too complicated. This patch replaces the bit spin_lock with pool->migrate_lock rwlock. It could make the code simple as well as zsmalloc work under PREEMPT_RT. The drawback is the pool->migrate_lock is bigger granuarity than per zpage lock so the contention would be higher than old when both IO-related operations(i.e., zsmalloc, zsfree, zs_[map|unmap]) and compaction(page/zpage migration) are going in parallel(*, the migrate_lock is rwlock and IO related functions are all read side lock so there is no contention). However, the write-side is fast enough(dominant overhead is just page copy) so it wouldn't affect much. If the lock granurity becomes more problem later, we could introduce table locks based on handle as a hash value. Link: https://lkml.kernel.org/r/20211115185909.3949505-9-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:14:13 +00:00
*/
zsmalloc: consolidate zs_pool's migrate_lock and size_class's locks Currently, zsmalloc has a hierarchy of locks, which includes a pool-level migrate_lock, and a lock for each size class. We have to obtain both locks in the hotpath in most cases anyway, except for zs_malloc. This exception will no longer exist when we introduce a LRU into the zs_pool for the new writeback functionality - we will need to obtain a pool-level lock to synchronize LRU handling even in zs_malloc. In preparation for zsmalloc writeback, consolidate these locks into a single pool-level lock, which drastically reduces the complexity of synchronization in zsmalloc. We have also benchmarked the lock consolidation to see the performance effect of this change on zram. First, we ran a synthetic FS workload on a server machine with 36 cores (same machine for all runs), using fs_mark -d ../zram1mnt -s 100000 -n 2500 -t 32 -k before and after for btrfs and ext4 on zram (FS usage is 80%). Here is the result (unit is file/second): With lock consolidation (btrfs): Average: 13520.2, Median: 13531.0, Stddev: 137.5961482019028 Without lock consolidation (btrfs): Average: 13487.2, Median: 13575.0, Stddev: 309.08283679298665 With lock consolidation (ext4): Average: 16824.4, Median: 16839.0, Stddev: 89.97388510006668 Without lock consolidation (ext4) Average: 16958.0, Median: 16986.0, Stddev: 194.7370021336469 As you can see, we observe a 0.3% regression for btrfs, and a 0.9% regression for ext4. This is a small, barely measurable difference in my opinion. For a more realistic scenario, we also tries building the kernel on zram. Here is the time it takes (in seconds): With lock consolidation (btrfs): real Average: 319.6, Median: 320.0, Stddev: 0.8944271909999159 user Average: 6894.2, Median: 6895.0, Stddev: 25.528415540334656 sys Average: 521.4, Median: 522.0, Stddev: 1.51657508881031 Without lock consolidation (btrfs): real Average: 319.8, Median: 320.0, Stddev: 0.8366600265340756 user Average: 6896.6, Median: 6899.0, Stddev: 16.04057355583023 sys Average: 520.6, Median: 521.0, Stddev: 1.140175425099138 With lock consolidation (ext4): real Average: 320.0, Median: 319.0, Stddev: 1.4142135623730951 user Average: 6896.8, Median: 6878.0, Stddev: 28.621670111997307 sys Average: 521.2, Median: 521.0, Stddev: 1.7888543819998317 Without lock consolidation (ext4) real Average: 319.6, Median: 319.0, Stddev: 0.8944271909999159 user Average: 6886.2, Median: 6887.0, Stddev: 16.93221781102523 sys Average: 520.4, Median: 520.0, Stddev: 1.140175425099138 The difference is entirely within the noise of a typical run on zram. This hardly justifies the complexity of maintaining both the pool lock and the class lock. In fact, for writeback, we would need to introduce yet another lock to prevent data races on the pool's LRU, further complicating the lock handling logic. IMHO, it is just better to collapse all of these into a single pool-level lock. Link: https://lkml.kernel.org/r/20221128191616.1261026-4-nphamcs@gmail.com Signed-off-by: Nhat Pham <nphamcs@gmail.com> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Acked-by: Minchan Kim <minchan@kernel.org> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-11-28 19:16:12 +00:00
spin_unlock(&pool->lock);
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
dec_zspage_isolation(zspage);
zsmalloc: replace per zpage lock with pool->migrate_lock The zsmalloc has used a bit for spin_lock in zpage handle to keep zpage object alive during several operations. However, it causes the problem for PREEMPT_RT as well as introducing too complicated. This patch replaces the bit spin_lock with pool->migrate_lock rwlock. It could make the code simple as well as zsmalloc work under PREEMPT_RT. The drawback is the pool->migrate_lock is bigger granuarity than per zpage lock so the contention would be higher than old when both IO-related operations(i.e., zsmalloc, zsfree, zs_[map|unmap]) and compaction(page/zpage migration) are going in parallel(*, the migrate_lock is rwlock and IO related functions are all read side lock so there is no contention). However, the write-side is fast enough(dominant overhead is just page copy) so it wouldn't affect much. If the lock granurity becomes more problem later, we could introduce table locks based on handle as a hash value. Link: https://lkml.kernel.org/r/20211115185909.3949505-9-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:14:13 +00:00
migrate_write_unlock(zspage);
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
zsmalloc: replace per zpage lock with pool->migrate_lock The zsmalloc has used a bit for spin_lock in zpage handle to keep zpage object alive during several operations. However, it causes the problem for PREEMPT_RT as well as introducing too complicated. This patch replaces the bit spin_lock with pool->migrate_lock rwlock. It could make the code simple as well as zsmalloc work under PREEMPT_RT. The drawback is the pool->migrate_lock is bigger granuarity than per zpage lock so the contention would be higher than old when both IO-related operations(i.e., zsmalloc, zsfree, zs_[map|unmap]) and compaction(page/zpage migration) are going in parallel(*, the migrate_lock is rwlock and IO related functions are all read side lock so there is no contention). However, the write-side is fast enough(dominant overhead is just page copy) so it wouldn't affect much. If the lock granurity becomes more problem later, we could introduce table locks based on handle as a hash value. Link: https://lkml.kernel.org/r/20211115185909.3949505-9-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:14:13 +00:00
get_page(newpage);
if (page_zone(newpage) != page_zone(page)) {
dec_zone_page_state(page, NR_ZSPAGES);
inc_zone_page_state(newpage, NR_ZSPAGES);
}
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
reset_page(page);
put_page(page);
zsmalloc: replace per zpage lock with pool->migrate_lock The zsmalloc has used a bit for spin_lock in zpage handle to keep zpage object alive during several operations. However, it causes the problem for PREEMPT_RT as well as introducing too complicated. This patch replaces the bit spin_lock with pool->migrate_lock rwlock. It could make the code simple as well as zsmalloc work under PREEMPT_RT. The drawback is the pool->migrate_lock is bigger granuarity than per zpage lock so the contention would be higher than old when both IO-related operations(i.e., zsmalloc, zsfree, zs_[map|unmap]) and compaction(page/zpage migration) are going in parallel(*, the migrate_lock is rwlock and IO related functions are all read side lock so there is no contention). However, the write-side is fast enough(dominant overhead is just page copy) so it wouldn't affect much. If the lock granurity becomes more problem later, we could introduce table locks based on handle as a hash value. Link: https://lkml.kernel.org/r/20211115185909.3949505-9-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:14:13 +00:00
return MIGRATEPAGE_SUCCESS;
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
}
static void zs_page_putback(struct page *page)
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
{
struct zspage *zspage;
VM_BUG_ON_PAGE(!PageIsolated(page), page);
zspage = get_zspage(page);
migrate_write_lock(zspage);
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
dec_zspage_isolation(zspage);
migrate_write_unlock(zspage);
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
}
static const struct movable_operations zsmalloc_mops = {
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
.isolate_page = zs_page_isolate,
.migrate_page = zs_page_migrate,
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
.putback_page = zs_page_putback,
};
/*
* Caller should hold page_lock of all pages in the zspage
* In here, we cannot use zspage meta data.
*/
static void async_free_zspage(struct work_struct *work)
{
int i;
struct size_class *class;
unsigned int class_idx;
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
int fullness;
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
struct zspage *zspage, *tmp;
LIST_HEAD(free_pages);
struct zs_pool *pool = container_of(work, struct zs_pool,
free_work);
for (i = 0; i < ZS_SIZE_CLASSES; i++) {
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
class = pool->size_class[i];
if (class->index != i)
continue;
zsmalloc: consolidate zs_pool's migrate_lock and size_class's locks Currently, zsmalloc has a hierarchy of locks, which includes a pool-level migrate_lock, and a lock for each size class. We have to obtain both locks in the hotpath in most cases anyway, except for zs_malloc. This exception will no longer exist when we introduce a LRU into the zs_pool for the new writeback functionality - we will need to obtain a pool-level lock to synchronize LRU handling even in zs_malloc. In preparation for zsmalloc writeback, consolidate these locks into a single pool-level lock, which drastically reduces the complexity of synchronization in zsmalloc. We have also benchmarked the lock consolidation to see the performance effect of this change on zram. First, we ran a synthetic FS workload on a server machine with 36 cores (same machine for all runs), using fs_mark -d ../zram1mnt -s 100000 -n 2500 -t 32 -k before and after for btrfs and ext4 on zram (FS usage is 80%). Here is the result (unit is file/second): With lock consolidation (btrfs): Average: 13520.2, Median: 13531.0, Stddev: 137.5961482019028 Without lock consolidation (btrfs): Average: 13487.2, Median: 13575.0, Stddev: 309.08283679298665 With lock consolidation (ext4): Average: 16824.4, Median: 16839.0, Stddev: 89.97388510006668 Without lock consolidation (ext4) Average: 16958.0, Median: 16986.0, Stddev: 194.7370021336469 As you can see, we observe a 0.3% regression for btrfs, and a 0.9% regression for ext4. This is a small, barely measurable difference in my opinion. For a more realistic scenario, we also tries building the kernel on zram. Here is the time it takes (in seconds): With lock consolidation (btrfs): real Average: 319.6, Median: 320.0, Stddev: 0.8944271909999159 user Average: 6894.2, Median: 6895.0, Stddev: 25.528415540334656 sys Average: 521.4, Median: 522.0, Stddev: 1.51657508881031 Without lock consolidation (btrfs): real Average: 319.8, Median: 320.0, Stddev: 0.8366600265340756 user Average: 6896.6, Median: 6899.0, Stddev: 16.04057355583023 sys Average: 520.6, Median: 521.0, Stddev: 1.140175425099138 With lock consolidation (ext4): real Average: 320.0, Median: 319.0, Stddev: 1.4142135623730951 user Average: 6896.8, Median: 6878.0, Stddev: 28.621670111997307 sys Average: 521.2, Median: 521.0, Stddev: 1.7888543819998317 Without lock consolidation (ext4) real Average: 319.6, Median: 319.0, Stddev: 0.8944271909999159 user Average: 6886.2, Median: 6887.0, Stddev: 16.93221781102523 sys Average: 520.4, Median: 520.0, Stddev: 1.140175425099138 The difference is entirely within the noise of a typical run on zram. This hardly justifies the complexity of maintaining both the pool lock and the class lock. In fact, for writeback, we would need to introduce yet another lock to prevent data races on the pool's LRU, further complicating the lock handling logic. IMHO, it is just better to collapse all of these into a single pool-level lock. Link: https://lkml.kernel.org/r/20221128191616.1261026-4-nphamcs@gmail.com Signed-off-by: Nhat Pham <nphamcs@gmail.com> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Acked-by: Minchan Kim <minchan@kernel.org> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-11-28 19:16:12 +00:00
spin_lock(&pool->lock);
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
list_splice_init(&class->fullness_list[ZS_INUSE_RATIO_0],
&free_pages);
zsmalloc: consolidate zs_pool's migrate_lock and size_class's locks Currently, zsmalloc has a hierarchy of locks, which includes a pool-level migrate_lock, and a lock for each size class. We have to obtain both locks in the hotpath in most cases anyway, except for zs_malloc. This exception will no longer exist when we introduce a LRU into the zs_pool for the new writeback functionality - we will need to obtain a pool-level lock to synchronize LRU handling even in zs_malloc. In preparation for zsmalloc writeback, consolidate these locks into a single pool-level lock, which drastically reduces the complexity of synchronization in zsmalloc. We have also benchmarked the lock consolidation to see the performance effect of this change on zram. First, we ran a synthetic FS workload on a server machine with 36 cores (same machine for all runs), using fs_mark -d ../zram1mnt -s 100000 -n 2500 -t 32 -k before and after for btrfs and ext4 on zram (FS usage is 80%). Here is the result (unit is file/second): With lock consolidation (btrfs): Average: 13520.2, Median: 13531.0, Stddev: 137.5961482019028 Without lock consolidation (btrfs): Average: 13487.2, Median: 13575.0, Stddev: 309.08283679298665 With lock consolidation (ext4): Average: 16824.4, Median: 16839.0, Stddev: 89.97388510006668 Without lock consolidation (ext4) Average: 16958.0, Median: 16986.0, Stddev: 194.7370021336469 As you can see, we observe a 0.3% regression for btrfs, and a 0.9% regression for ext4. This is a small, barely measurable difference in my opinion. For a more realistic scenario, we also tries building the kernel on zram. Here is the time it takes (in seconds): With lock consolidation (btrfs): real Average: 319.6, Median: 320.0, Stddev: 0.8944271909999159 user Average: 6894.2, Median: 6895.0, Stddev: 25.528415540334656 sys Average: 521.4, Median: 522.0, Stddev: 1.51657508881031 Without lock consolidation (btrfs): real Average: 319.8, Median: 320.0, Stddev: 0.8366600265340756 user Average: 6896.6, Median: 6899.0, Stddev: 16.04057355583023 sys Average: 520.6, Median: 521.0, Stddev: 1.140175425099138 With lock consolidation (ext4): real Average: 320.0, Median: 319.0, Stddev: 1.4142135623730951 user Average: 6896.8, Median: 6878.0, Stddev: 28.621670111997307 sys Average: 521.2, Median: 521.0, Stddev: 1.7888543819998317 Without lock consolidation (ext4) real Average: 319.6, Median: 319.0, Stddev: 0.8944271909999159 user Average: 6886.2, Median: 6887.0, Stddev: 16.93221781102523 sys Average: 520.4, Median: 520.0, Stddev: 1.140175425099138 The difference is entirely within the noise of a typical run on zram. This hardly justifies the complexity of maintaining both the pool lock and the class lock. In fact, for writeback, we would need to introduce yet another lock to prevent data races on the pool's LRU, further complicating the lock handling logic. IMHO, it is just better to collapse all of these into a single pool-level lock. Link: https://lkml.kernel.org/r/20221128191616.1261026-4-nphamcs@gmail.com Signed-off-by: Nhat Pham <nphamcs@gmail.com> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Acked-by: Minchan Kim <minchan@kernel.org> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-11-28 19:16:12 +00:00
spin_unlock(&pool->lock);
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
}
list_for_each_entry_safe(zspage, tmp, &free_pages, list) {
list_del(&zspage->list);
lock_zspage(zspage);
get_zspage_mapping(zspage, &class_idx, &fullness);
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
VM_BUG_ON(fullness != ZS_INUSE_RATIO_0);
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
class = pool->size_class[class_idx];
zsmalloc: consolidate zs_pool's migrate_lock and size_class's locks Currently, zsmalloc has a hierarchy of locks, which includes a pool-level migrate_lock, and a lock for each size class. We have to obtain both locks in the hotpath in most cases anyway, except for zs_malloc. This exception will no longer exist when we introduce a LRU into the zs_pool for the new writeback functionality - we will need to obtain a pool-level lock to synchronize LRU handling even in zs_malloc. In preparation for zsmalloc writeback, consolidate these locks into a single pool-level lock, which drastically reduces the complexity of synchronization in zsmalloc. We have also benchmarked the lock consolidation to see the performance effect of this change on zram. First, we ran a synthetic FS workload on a server machine with 36 cores (same machine for all runs), using fs_mark -d ../zram1mnt -s 100000 -n 2500 -t 32 -k before and after for btrfs and ext4 on zram (FS usage is 80%). Here is the result (unit is file/second): With lock consolidation (btrfs): Average: 13520.2, Median: 13531.0, Stddev: 137.5961482019028 Without lock consolidation (btrfs): Average: 13487.2, Median: 13575.0, Stddev: 309.08283679298665 With lock consolidation (ext4): Average: 16824.4, Median: 16839.0, Stddev: 89.97388510006668 Without lock consolidation (ext4) Average: 16958.0, Median: 16986.0, Stddev: 194.7370021336469 As you can see, we observe a 0.3% regression for btrfs, and a 0.9% regression for ext4. This is a small, barely measurable difference in my opinion. For a more realistic scenario, we also tries building the kernel on zram. Here is the time it takes (in seconds): With lock consolidation (btrfs): real Average: 319.6, Median: 320.0, Stddev: 0.8944271909999159 user Average: 6894.2, Median: 6895.0, Stddev: 25.528415540334656 sys Average: 521.4, Median: 522.0, Stddev: 1.51657508881031 Without lock consolidation (btrfs): real Average: 319.8, Median: 320.0, Stddev: 0.8366600265340756 user Average: 6896.6, Median: 6899.0, Stddev: 16.04057355583023 sys Average: 520.6, Median: 521.0, Stddev: 1.140175425099138 With lock consolidation (ext4): real Average: 320.0, Median: 319.0, Stddev: 1.4142135623730951 user Average: 6896.8, Median: 6878.0, Stddev: 28.621670111997307 sys Average: 521.2, Median: 521.0, Stddev: 1.7888543819998317 Without lock consolidation (ext4) real Average: 319.6, Median: 319.0, Stddev: 0.8944271909999159 user Average: 6886.2, Median: 6887.0, Stddev: 16.93221781102523 sys Average: 520.4, Median: 520.0, Stddev: 1.140175425099138 The difference is entirely within the noise of a typical run on zram. This hardly justifies the complexity of maintaining both the pool lock and the class lock. In fact, for writeback, we would need to introduce yet another lock to prevent data races on the pool's LRU, further complicating the lock handling logic. IMHO, it is just better to collapse all of these into a single pool-level lock. Link: https://lkml.kernel.org/r/20221128191616.1261026-4-nphamcs@gmail.com Signed-off-by: Nhat Pham <nphamcs@gmail.com> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Acked-by: Minchan Kim <minchan@kernel.org> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-11-28 19:16:12 +00:00
spin_lock(&pool->lock);
#ifdef CONFIG_ZPOOL
list_del(&zspage->lru);
#endif
__free_zspage(pool, class, zspage);
zsmalloc: consolidate zs_pool's migrate_lock and size_class's locks Currently, zsmalloc has a hierarchy of locks, which includes a pool-level migrate_lock, and a lock for each size class. We have to obtain both locks in the hotpath in most cases anyway, except for zs_malloc. This exception will no longer exist when we introduce a LRU into the zs_pool for the new writeback functionality - we will need to obtain a pool-level lock to synchronize LRU handling even in zs_malloc. In preparation for zsmalloc writeback, consolidate these locks into a single pool-level lock, which drastically reduces the complexity of synchronization in zsmalloc. We have also benchmarked the lock consolidation to see the performance effect of this change on zram. First, we ran a synthetic FS workload on a server machine with 36 cores (same machine for all runs), using fs_mark -d ../zram1mnt -s 100000 -n 2500 -t 32 -k before and after for btrfs and ext4 on zram (FS usage is 80%). Here is the result (unit is file/second): With lock consolidation (btrfs): Average: 13520.2, Median: 13531.0, Stddev: 137.5961482019028 Without lock consolidation (btrfs): Average: 13487.2, Median: 13575.0, Stddev: 309.08283679298665 With lock consolidation (ext4): Average: 16824.4, Median: 16839.0, Stddev: 89.97388510006668 Without lock consolidation (ext4) Average: 16958.0, Median: 16986.0, Stddev: 194.7370021336469 As you can see, we observe a 0.3% regression for btrfs, and a 0.9% regression for ext4. This is a small, barely measurable difference in my opinion. For a more realistic scenario, we also tries building the kernel on zram. Here is the time it takes (in seconds): With lock consolidation (btrfs): real Average: 319.6, Median: 320.0, Stddev: 0.8944271909999159 user Average: 6894.2, Median: 6895.0, Stddev: 25.528415540334656 sys Average: 521.4, Median: 522.0, Stddev: 1.51657508881031 Without lock consolidation (btrfs): real Average: 319.8, Median: 320.0, Stddev: 0.8366600265340756 user Average: 6896.6, Median: 6899.0, Stddev: 16.04057355583023 sys Average: 520.6, Median: 521.0, Stddev: 1.140175425099138 With lock consolidation (ext4): real Average: 320.0, Median: 319.0, Stddev: 1.4142135623730951 user Average: 6896.8, Median: 6878.0, Stddev: 28.621670111997307 sys Average: 521.2, Median: 521.0, Stddev: 1.7888543819998317 Without lock consolidation (ext4) real Average: 319.6, Median: 319.0, Stddev: 0.8944271909999159 user Average: 6886.2, Median: 6887.0, Stddev: 16.93221781102523 sys Average: 520.4, Median: 520.0, Stddev: 1.140175425099138 The difference is entirely within the noise of a typical run on zram. This hardly justifies the complexity of maintaining both the pool lock and the class lock. In fact, for writeback, we would need to introduce yet another lock to prevent data races on the pool's LRU, further complicating the lock handling logic. IMHO, it is just better to collapse all of these into a single pool-level lock. Link: https://lkml.kernel.org/r/20221128191616.1261026-4-nphamcs@gmail.com Signed-off-by: Nhat Pham <nphamcs@gmail.com> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Acked-by: Minchan Kim <minchan@kernel.org> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-11-28 19:16:12 +00:00
spin_unlock(&pool->lock);
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
}
};
static void kick_deferred_free(struct zs_pool *pool)
{
schedule_work(&pool->free_work);
}
static void zs_flush_migration(struct zs_pool *pool)
{
flush_work(&pool->free_work);
}
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
static void init_deferred_free(struct zs_pool *pool)
{
INIT_WORK(&pool->free_work, async_free_zspage);
}
static void SetZsPageMovable(struct zs_pool *pool, struct zspage *zspage)
{
struct page *page = get_first_page(zspage);
do {
WARN_ON(!trylock_page(page));
__SetPageMovable(page, &zsmalloc_mops);
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
unlock_page(page);
} while ((page = get_next_page(page)) != NULL);
}
#else
static inline void zs_flush_migration(struct zs_pool *pool) { }
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
#endif
zsmalloc: introduce zs_can_compact() function This function checks if class compaction will free any pages. Rephrasing -- do we have enough unused objects to form at least one ZS_EMPTY page and free it. It aborts compaction if class compaction will not result in any (further) savings. EXAMPLE (this debug output is not part of this patch set): - class size - number of allocated objects - number of used objects - max objects per zspage - pages per zspage - estimated number of pages that will be freed [..] class-512 objs:544 inuse:540 maxobj-per-zspage:8 pages-per-zspage:1 zspages-to-free:0 ... class-512 compaction is useless. break class-496 objs:660 inuse:570 maxobj-per-zspage:33 pages-per-zspage:4 zspages-to-free:2 class-496 objs:627 inuse:570 maxobj-per-zspage:33 pages-per-zspage:4 zspages-to-free:1 class-496 objs:594 inuse:570 maxobj-per-zspage:33 pages-per-zspage:4 zspages-to-free:0 ... class-496 compaction is useless. break class-448 objs:657 inuse:617 maxobj-per-zspage:9 pages-per-zspage:1 zspages-to-free:4 class-448 objs:648 inuse:617 maxobj-per-zspage:9 pages-per-zspage:1 zspages-to-free:3 class-448 objs:639 inuse:617 maxobj-per-zspage:9 pages-per-zspage:1 zspages-to-free:2 class-448 objs:630 inuse:617 maxobj-per-zspage:9 pages-per-zspage:1 zspages-to-free:1 class-448 objs:621 inuse:617 maxobj-per-zspage:9 pages-per-zspage:1 zspages-to-free:0 ... class-448 compaction is useless. break class-432 objs:728 inuse:685 maxobj-per-zspage:28 pages-per-zspage:3 zspages-to-free:1 class-432 objs:700 inuse:685 maxobj-per-zspage:28 pages-per-zspage:3 zspages-to-free:0 ... class-432 compaction is useless. break class-416 objs:819 inuse:705 maxobj-per-zspage:39 pages-per-zspage:4 zspages-to-free:2 class-416 objs:780 inuse:705 maxobj-per-zspage:39 pages-per-zspage:4 zspages-to-free:1 class-416 objs:741 inuse:705 maxobj-per-zspage:39 pages-per-zspage:4 zspages-to-free:0 ... class-416 compaction is useless. break class-400 objs:690 inuse:674 maxobj-per-zspage:10 pages-per-zspage:1 zspages-to-free:1 class-400 objs:680 inuse:674 maxobj-per-zspage:10 pages-per-zspage:1 zspages-to-free:0 ... class-400 compaction is useless. break class-384 objs:736 inuse:709 maxobj-per-zspage:32 pages-per-zspage:3 zspages-to-free:0 ... class-384 compaction is useless. break [..] Every "compaction is useless" indicates that we saved CPU cycles. class-512 has 544 object allocated 540 objects used 8 objects per-page Even if we have a ALMOST_EMPTY zspage, we still don't have enough room to migrate all of its objects and free this zspage; so compaction will not make a lot of sense, it's better to just leave it as is. Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Acked-by: Minchan Kim <minchan@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-09-08 22:04:30 +00:00
/*
*
* Based on the number of unused allocated objects calculate
* and return the number of pages that we can free.
*/
static unsigned long zs_can_compact(struct size_class *class)
{
unsigned long obj_wasted;
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
unsigned long obj_allocated = zs_stat_get(class, ZS_OBJS_ALLOCATED);
unsigned long obj_used = zs_stat_get(class, ZS_OBJS_INUSE);
zsmalloc: introduce zs_can_compact() function This function checks if class compaction will free any pages. Rephrasing -- do we have enough unused objects to form at least one ZS_EMPTY page and free it. It aborts compaction if class compaction will not result in any (further) savings. EXAMPLE (this debug output is not part of this patch set): - class size - number of allocated objects - number of used objects - max objects per zspage - pages per zspage - estimated number of pages that will be freed [..] class-512 objs:544 inuse:540 maxobj-per-zspage:8 pages-per-zspage:1 zspages-to-free:0 ... class-512 compaction is useless. break class-496 objs:660 inuse:570 maxobj-per-zspage:33 pages-per-zspage:4 zspages-to-free:2 class-496 objs:627 inuse:570 maxobj-per-zspage:33 pages-per-zspage:4 zspages-to-free:1 class-496 objs:594 inuse:570 maxobj-per-zspage:33 pages-per-zspage:4 zspages-to-free:0 ... class-496 compaction is useless. break class-448 objs:657 inuse:617 maxobj-per-zspage:9 pages-per-zspage:1 zspages-to-free:4 class-448 objs:648 inuse:617 maxobj-per-zspage:9 pages-per-zspage:1 zspages-to-free:3 class-448 objs:639 inuse:617 maxobj-per-zspage:9 pages-per-zspage:1 zspages-to-free:2 class-448 objs:630 inuse:617 maxobj-per-zspage:9 pages-per-zspage:1 zspages-to-free:1 class-448 objs:621 inuse:617 maxobj-per-zspage:9 pages-per-zspage:1 zspages-to-free:0 ... class-448 compaction is useless. break class-432 objs:728 inuse:685 maxobj-per-zspage:28 pages-per-zspage:3 zspages-to-free:1 class-432 objs:700 inuse:685 maxobj-per-zspage:28 pages-per-zspage:3 zspages-to-free:0 ... class-432 compaction is useless. break class-416 objs:819 inuse:705 maxobj-per-zspage:39 pages-per-zspage:4 zspages-to-free:2 class-416 objs:780 inuse:705 maxobj-per-zspage:39 pages-per-zspage:4 zspages-to-free:1 class-416 objs:741 inuse:705 maxobj-per-zspage:39 pages-per-zspage:4 zspages-to-free:0 ... class-416 compaction is useless. break class-400 objs:690 inuse:674 maxobj-per-zspage:10 pages-per-zspage:1 zspages-to-free:1 class-400 objs:680 inuse:674 maxobj-per-zspage:10 pages-per-zspage:1 zspages-to-free:0 ... class-400 compaction is useless. break class-384 objs:736 inuse:709 maxobj-per-zspage:32 pages-per-zspage:3 zspages-to-free:0 ... class-384 compaction is useless. break [..] Every "compaction is useless" indicates that we saved CPU cycles. class-512 has 544 object allocated 540 objects used 8 objects per-page Even if we have a ALMOST_EMPTY zspage, we still don't have enough room to migrate all of its objects and free this zspage; so compaction will not make a lot of sense, it's better to just leave it as is. Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Acked-by: Minchan Kim <minchan@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-09-08 22:04:30 +00:00
zsmalloc: fix zs_can_compact() integer overflow zs_can_compact() has two race conditions in its core calculation: unsigned long obj_wasted = zs_stat_get(class, OBJ_ALLOCATED) - zs_stat_get(class, OBJ_USED); 1) classes are not locked, so the numbers of allocated and used objects can change by the concurrent ops happening on other CPUs 2) shrinker invokes it from preemptible context Depending on the circumstances, thus, OBJ_ALLOCATED can become less than OBJ_USED, which can result in either very high or negative `total_scan' value calculated later in do_shrink_slab(). do_shrink_slab() has some logic to prevent those cases: vmscan: shrink_slab: zs_shrinker_scan+0x0/0x28 [zsmalloc] negative objects to delete nr=-62 vmscan: shrink_slab: zs_shrinker_scan+0x0/0x28 [zsmalloc] negative objects to delete nr=-62 vmscan: shrink_slab: zs_shrinker_scan+0x0/0x28 [zsmalloc] negative objects to delete nr=-64 vmscan: shrink_slab: zs_shrinker_scan+0x0/0x28 [zsmalloc] negative objects to delete nr=-62 vmscan: shrink_slab: zs_shrinker_scan+0x0/0x28 [zsmalloc] negative objects to delete nr=-62 vmscan: shrink_slab: zs_shrinker_scan+0x0/0x28 [zsmalloc] negative objects to delete nr=-62 However, due to the way `total_scan' is calculated, not every shrinker->count_objects() overflow can be spotted and handled. To demonstrate the latter, I added some debugging code to do_shrink_slab() (x86_64) and the results were: vmscan: OVERFLOW: shrinker->count_objects() == -1 [18446744073709551615] vmscan: but total_scan > 0: 92679974445502 vmscan: resulting total_scan: 92679974445502 [..] vmscan: OVERFLOW: shrinker->count_objects() == -1 [18446744073709551615] vmscan: but total_scan > 0: 22634041808232578 vmscan: resulting total_scan: 22634041808232578 Even though shrinker->count_objects() has returned an overflowed value, the resulting `total_scan' is positive, and, what is more worrisome, it is insanely huge. This value is getting used later on in shrinker->scan_objects() loop: while (total_scan >= batch_size || total_scan >= freeable) { unsigned long ret; unsigned long nr_to_scan = min(batch_size, total_scan); shrinkctl->nr_to_scan = nr_to_scan; ret = shrinker->scan_objects(shrinker, shrinkctl); if (ret == SHRINK_STOP) break; freed += ret; count_vm_events(SLABS_SCANNED, nr_to_scan); total_scan -= nr_to_scan; cond_resched(); } `total_scan >= batch_size' is true for a very-very long time and 'total_scan >= freeable' is also true for quite some time, because `freeable < 0' and `total_scan' is large enough, for example, 22634041808232578. The only break condition, in the given scheme of things, is shrinker->scan_objects() == SHRINK_STOP test, which is a bit too weak to rely on, especially in heavy zsmalloc-usage scenarios. To fix the issue, take a pool stat snapshot and use it instead of racy zs_stat_get() calls. Link: http://lkml.kernel.org/r/20160509140052.3389-1-sergey.senozhatsky@gmail.com Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Minchan Kim <minchan@kernel.org> Cc: <stable@vger.kernel.org> [4.3+] Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-05-09 23:28:49 +00:00
if (obj_allocated <= obj_used)
return 0;
zsmalloc: introduce zs_can_compact() function This function checks if class compaction will free any pages. Rephrasing -- do we have enough unused objects to form at least one ZS_EMPTY page and free it. It aborts compaction if class compaction will not result in any (further) savings. EXAMPLE (this debug output is not part of this patch set): - class size - number of allocated objects - number of used objects - max objects per zspage - pages per zspage - estimated number of pages that will be freed [..] class-512 objs:544 inuse:540 maxobj-per-zspage:8 pages-per-zspage:1 zspages-to-free:0 ... class-512 compaction is useless. break class-496 objs:660 inuse:570 maxobj-per-zspage:33 pages-per-zspage:4 zspages-to-free:2 class-496 objs:627 inuse:570 maxobj-per-zspage:33 pages-per-zspage:4 zspages-to-free:1 class-496 objs:594 inuse:570 maxobj-per-zspage:33 pages-per-zspage:4 zspages-to-free:0 ... class-496 compaction is useless. break class-448 objs:657 inuse:617 maxobj-per-zspage:9 pages-per-zspage:1 zspages-to-free:4 class-448 objs:648 inuse:617 maxobj-per-zspage:9 pages-per-zspage:1 zspages-to-free:3 class-448 objs:639 inuse:617 maxobj-per-zspage:9 pages-per-zspage:1 zspages-to-free:2 class-448 objs:630 inuse:617 maxobj-per-zspage:9 pages-per-zspage:1 zspages-to-free:1 class-448 objs:621 inuse:617 maxobj-per-zspage:9 pages-per-zspage:1 zspages-to-free:0 ... class-448 compaction is useless. break class-432 objs:728 inuse:685 maxobj-per-zspage:28 pages-per-zspage:3 zspages-to-free:1 class-432 objs:700 inuse:685 maxobj-per-zspage:28 pages-per-zspage:3 zspages-to-free:0 ... class-432 compaction is useless. break class-416 objs:819 inuse:705 maxobj-per-zspage:39 pages-per-zspage:4 zspages-to-free:2 class-416 objs:780 inuse:705 maxobj-per-zspage:39 pages-per-zspage:4 zspages-to-free:1 class-416 objs:741 inuse:705 maxobj-per-zspage:39 pages-per-zspage:4 zspages-to-free:0 ... class-416 compaction is useless. break class-400 objs:690 inuse:674 maxobj-per-zspage:10 pages-per-zspage:1 zspages-to-free:1 class-400 objs:680 inuse:674 maxobj-per-zspage:10 pages-per-zspage:1 zspages-to-free:0 ... class-400 compaction is useless. break class-384 objs:736 inuse:709 maxobj-per-zspage:32 pages-per-zspage:3 zspages-to-free:0 ... class-384 compaction is useless. break [..] Every "compaction is useless" indicates that we saved CPU cycles. class-512 has 544 object allocated 540 objects used 8 objects per-page Even if we have a ALMOST_EMPTY zspage, we still don't have enough room to migrate all of its objects and free this zspage; so compaction will not make a lot of sense, it's better to just leave it as is. Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Acked-by: Minchan Kim <minchan@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-09-08 22:04:30 +00:00
zsmalloc: fix zs_can_compact() integer overflow zs_can_compact() has two race conditions in its core calculation: unsigned long obj_wasted = zs_stat_get(class, OBJ_ALLOCATED) - zs_stat_get(class, OBJ_USED); 1) classes are not locked, so the numbers of allocated and used objects can change by the concurrent ops happening on other CPUs 2) shrinker invokes it from preemptible context Depending on the circumstances, thus, OBJ_ALLOCATED can become less than OBJ_USED, which can result in either very high or negative `total_scan' value calculated later in do_shrink_slab(). do_shrink_slab() has some logic to prevent those cases: vmscan: shrink_slab: zs_shrinker_scan+0x0/0x28 [zsmalloc] negative objects to delete nr=-62 vmscan: shrink_slab: zs_shrinker_scan+0x0/0x28 [zsmalloc] negative objects to delete nr=-62 vmscan: shrink_slab: zs_shrinker_scan+0x0/0x28 [zsmalloc] negative objects to delete nr=-64 vmscan: shrink_slab: zs_shrinker_scan+0x0/0x28 [zsmalloc] negative objects to delete nr=-62 vmscan: shrink_slab: zs_shrinker_scan+0x0/0x28 [zsmalloc] negative objects to delete nr=-62 vmscan: shrink_slab: zs_shrinker_scan+0x0/0x28 [zsmalloc] negative objects to delete nr=-62 However, due to the way `total_scan' is calculated, not every shrinker->count_objects() overflow can be spotted and handled. To demonstrate the latter, I added some debugging code to do_shrink_slab() (x86_64) and the results were: vmscan: OVERFLOW: shrinker->count_objects() == -1 [18446744073709551615] vmscan: but total_scan > 0: 92679974445502 vmscan: resulting total_scan: 92679974445502 [..] vmscan: OVERFLOW: shrinker->count_objects() == -1 [18446744073709551615] vmscan: but total_scan > 0: 22634041808232578 vmscan: resulting total_scan: 22634041808232578 Even though shrinker->count_objects() has returned an overflowed value, the resulting `total_scan' is positive, and, what is more worrisome, it is insanely huge. This value is getting used later on in shrinker->scan_objects() loop: while (total_scan >= batch_size || total_scan >= freeable) { unsigned long ret; unsigned long nr_to_scan = min(batch_size, total_scan); shrinkctl->nr_to_scan = nr_to_scan; ret = shrinker->scan_objects(shrinker, shrinkctl); if (ret == SHRINK_STOP) break; freed += ret; count_vm_events(SLABS_SCANNED, nr_to_scan); total_scan -= nr_to_scan; cond_resched(); } `total_scan >= batch_size' is true for a very-very long time and 'total_scan >= freeable' is also true for quite some time, because `freeable < 0' and `total_scan' is large enough, for example, 22634041808232578. The only break condition, in the given scheme of things, is shrinker->scan_objects() == SHRINK_STOP test, which is a bit too weak to rely on, especially in heavy zsmalloc-usage scenarios. To fix the issue, take a pool stat snapshot and use it instead of racy zs_stat_get() calls. Link: http://lkml.kernel.org/r/20160509140052.3389-1-sergey.senozhatsky@gmail.com Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Minchan Kim <minchan@kernel.org> Cc: <stable@vger.kernel.org> [4.3+] Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-05-09 23:28:49 +00:00
obj_wasted = obj_allocated - obj_used;
obj_wasted /= class->objs_per_zspage;
zsmalloc: introduce zs_can_compact() function This function checks if class compaction will free any pages. Rephrasing -- do we have enough unused objects to form at least one ZS_EMPTY page and free it. It aborts compaction if class compaction will not result in any (further) savings. EXAMPLE (this debug output is not part of this patch set): - class size - number of allocated objects - number of used objects - max objects per zspage - pages per zspage - estimated number of pages that will be freed [..] class-512 objs:544 inuse:540 maxobj-per-zspage:8 pages-per-zspage:1 zspages-to-free:0 ... class-512 compaction is useless. break class-496 objs:660 inuse:570 maxobj-per-zspage:33 pages-per-zspage:4 zspages-to-free:2 class-496 objs:627 inuse:570 maxobj-per-zspage:33 pages-per-zspage:4 zspages-to-free:1 class-496 objs:594 inuse:570 maxobj-per-zspage:33 pages-per-zspage:4 zspages-to-free:0 ... class-496 compaction is useless. break class-448 objs:657 inuse:617 maxobj-per-zspage:9 pages-per-zspage:1 zspages-to-free:4 class-448 objs:648 inuse:617 maxobj-per-zspage:9 pages-per-zspage:1 zspages-to-free:3 class-448 objs:639 inuse:617 maxobj-per-zspage:9 pages-per-zspage:1 zspages-to-free:2 class-448 objs:630 inuse:617 maxobj-per-zspage:9 pages-per-zspage:1 zspages-to-free:1 class-448 objs:621 inuse:617 maxobj-per-zspage:9 pages-per-zspage:1 zspages-to-free:0 ... class-448 compaction is useless. break class-432 objs:728 inuse:685 maxobj-per-zspage:28 pages-per-zspage:3 zspages-to-free:1 class-432 objs:700 inuse:685 maxobj-per-zspage:28 pages-per-zspage:3 zspages-to-free:0 ... class-432 compaction is useless. break class-416 objs:819 inuse:705 maxobj-per-zspage:39 pages-per-zspage:4 zspages-to-free:2 class-416 objs:780 inuse:705 maxobj-per-zspage:39 pages-per-zspage:4 zspages-to-free:1 class-416 objs:741 inuse:705 maxobj-per-zspage:39 pages-per-zspage:4 zspages-to-free:0 ... class-416 compaction is useless. break class-400 objs:690 inuse:674 maxobj-per-zspage:10 pages-per-zspage:1 zspages-to-free:1 class-400 objs:680 inuse:674 maxobj-per-zspage:10 pages-per-zspage:1 zspages-to-free:0 ... class-400 compaction is useless. break class-384 objs:736 inuse:709 maxobj-per-zspage:32 pages-per-zspage:3 zspages-to-free:0 ... class-384 compaction is useless. break [..] Every "compaction is useless" indicates that we saved CPU cycles. class-512 has 544 object allocated 540 objects used 8 objects per-page Even if we have a ALMOST_EMPTY zspage, we still don't have enough room to migrate all of its objects and free this zspage; so compaction will not make a lot of sense, it's better to just leave it as is. Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Acked-by: Minchan Kim <minchan@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-09-08 22:04:30 +00:00
return obj_wasted * class->pages_per_zspage;
zsmalloc: introduce zs_can_compact() function This function checks if class compaction will free any pages. Rephrasing -- do we have enough unused objects to form at least one ZS_EMPTY page and free it. It aborts compaction if class compaction will not result in any (further) savings. EXAMPLE (this debug output is not part of this patch set): - class size - number of allocated objects - number of used objects - max objects per zspage - pages per zspage - estimated number of pages that will be freed [..] class-512 objs:544 inuse:540 maxobj-per-zspage:8 pages-per-zspage:1 zspages-to-free:0 ... class-512 compaction is useless. break class-496 objs:660 inuse:570 maxobj-per-zspage:33 pages-per-zspage:4 zspages-to-free:2 class-496 objs:627 inuse:570 maxobj-per-zspage:33 pages-per-zspage:4 zspages-to-free:1 class-496 objs:594 inuse:570 maxobj-per-zspage:33 pages-per-zspage:4 zspages-to-free:0 ... class-496 compaction is useless. break class-448 objs:657 inuse:617 maxobj-per-zspage:9 pages-per-zspage:1 zspages-to-free:4 class-448 objs:648 inuse:617 maxobj-per-zspage:9 pages-per-zspage:1 zspages-to-free:3 class-448 objs:639 inuse:617 maxobj-per-zspage:9 pages-per-zspage:1 zspages-to-free:2 class-448 objs:630 inuse:617 maxobj-per-zspage:9 pages-per-zspage:1 zspages-to-free:1 class-448 objs:621 inuse:617 maxobj-per-zspage:9 pages-per-zspage:1 zspages-to-free:0 ... class-448 compaction is useless. break class-432 objs:728 inuse:685 maxobj-per-zspage:28 pages-per-zspage:3 zspages-to-free:1 class-432 objs:700 inuse:685 maxobj-per-zspage:28 pages-per-zspage:3 zspages-to-free:0 ... class-432 compaction is useless. break class-416 objs:819 inuse:705 maxobj-per-zspage:39 pages-per-zspage:4 zspages-to-free:2 class-416 objs:780 inuse:705 maxobj-per-zspage:39 pages-per-zspage:4 zspages-to-free:1 class-416 objs:741 inuse:705 maxobj-per-zspage:39 pages-per-zspage:4 zspages-to-free:0 ... class-416 compaction is useless. break class-400 objs:690 inuse:674 maxobj-per-zspage:10 pages-per-zspage:1 zspages-to-free:1 class-400 objs:680 inuse:674 maxobj-per-zspage:10 pages-per-zspage:1 zspages-to-free:0 ... class-400 compaction is useless. break class-384 objs:736 inuse:709 maxobj-per-zspage:32 pages-per-zspage:3 zspages-to-free:0 ... class-384 compaction is useless. break [..] Every "compaction is useless" indicates that we saved CPU cycles. class-512 has 544 object allocated 540 objects used 8 objects per-page Even if we have a ALMOST_EMPTY zspage, we still don't have enough room to migrate all of its objects and free this zspage; so compaction will not make a lot of sense, it's better to just leave it as is. Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Acked-by: Minchan Kim <minchan@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-09-08 22:04:30 +00:00
}
static unsigned long __zs_compact(struct zs_pool *pool,
struct size_class *class)
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
{
struct zs_compact_control cc;
zsmalloc: rework compaction algorithm The zsmalloc compaction algorithm has the potential to waste some CPU cycles, particularly when compacting pages within the same fullness group. This is due to the way it selects the head page of the fullness list for source and destination pages, and how it reinserts those pages during each iteration. The algorithm may first use a page as a migration destination and then as a migration source, leading to an unnecessary back-and-forth movement of objects. Consider the following fullness list: PageA PageB PageC PageD PageE During the first iteration, the compaction algorithm will select PageA as the source and PageB as the destination. All of PageA's objects will be moved to PageB, and then PageA will be released while PageB is reinserted into the fullness list. PageB PageC PageD PageE During the next iteration, the compaction algorithm will again select the head of the list as the source and destination, meaning that PageB will now serve as the source and PageC as the destination. This will result in the objects being moved away from PageB, the same objects that were just moved to PageB in the previous iteration. To prevent this avalanche effect, the compaction algorithm should not reinsert the destination page between iterations. By doing so, the most optimal page will continue to be used and its usage ratio will increase, reducing internal fragmentation. The destination page should only be reinserted into the fullness list if: - It becomes full - No source page is available. TEST ==== It's very challenging to reliably test this series. I ended up developing my own synthetic test that has 100% reproducibility. The test generates significan fragmentation (for each size class) and then performs compaction for each class individually and tracks the number of memcpy() in zs_object_copy(), so that we can compare the amount work compaction does on per-class basis. Total amount of work (zram mm_stat objs_moved) ---------------------------------------------- Old fullness grouping, old compaction algorithm: 323977 memcpy() in zs_object_copy(). Old fullness grouping, new compaction algorithm: 262944 memcpy() in zs_object_copy(). New fullness grouping, new compaction algorithm: 213978 memcpy() in zs_object_copy(). Per-class compaction memcpy() comparison (T-test) ------------------------------------------------- x Old fullness grouping, old compaction algorithm + Old fullness grouping, new compaction algorithm N Min Max Median Avg Stddev x 140 349 3513 2461 2314.1214 806.03271 + 140 289 2778 2006 1878.1714 641.02073 Difference at 95.0% confidence -435.95 +/- 170.595 -18.8387% +/- 7.37193% (Student's t, pooled s = 728.216) x Old fullness grouping, old compaction algorithm + New fullness grouping, new compaction algorithm N Min Max Median Avg Stddev x 140 349 3513 2461 2314.1214 806.03271 + 140 226 2279 1644 1528.4143 524.85268 Difference at 95.0% confidence -785.707 +/- 159.331 -33.9527% +/- 6.88516% (Student's t, pooled s = 680.132) Link: https://lkml.kernel.org/r/20230304034835.2082479-4-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:34 +00:00
struct zspage *src_zspage = NULL;
struct zspage *dst_zspage = NULL;
unsigned long pages_freed = 0;
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
zsmalloc: consolidate zs_pool's migrate_lock and size_class's locks Currently, zsmalloc has a hierarchy of locks, which includes a pool-level migrate_lock, and a lock for each size class. We have to obtain both locks in the hotpath in most cases anyway, except for zs_malloc. This exception will no longer exist when we introduce a LRU into the zs_pool for the new writeback functionality - we will need to obtain a pool-level lock to synchronize LRU handling even in zs_malloc. In preparation for zsmalloc writeback, consolidate these locks into a single pool-level lock, which drastically reduces the complexity of synchronization in zsmalloc. We have also benchmarked the lock consolidation to see the performance effect of this change on zram. First, we ran a synthetic FS workload on a server machine with 36 cores (same machine for all runs), using fs_mark -d ../zram1mnt -s 100000 -n 2500 -t 32 -k before and after for btrfs and ext4 on zram (FS usage is 80%). Here is the result (unit is file/second): With lock consolidation (btrfs): Average: 13520.2, Median: 13531.0, Stddev: 137.5961482019028 Without lock consolidation (btrfs): Average: 13487.2, Median: 13575.0, Stddev: 309.08283679298665 With lock consolidation (ext4): Average: 16824.4, Median: 16839.0, Stddev: 89.97388510006668 Without lock consolidation (ext4) Average: 16958.0, Median: 16986.0, Stddev: 194.7370021336469 As you can see, we observe a 0.3% regression for btrfs, and a 0.9% regression for ext4. This is a small, barely measurable difference in my opinion. For a more realistic scenario, we also tries building the kernel on zram. Here is the time it takes (in seconds): With lock consolidation (btrfs): real Average: 319.6, Median: 320.0, Stddev: 0.8944271909999159 user Average: 6894.2, Median: 6895.0, Stddev: 25.528415540334656 sys Average: 521.4, Median: 522.0, Stddev: 1.51657508881031 Without lock consolidation (btrfs): real Average: 319.8, Median: 320.0, Stddev: 0.8366600265340756 user Average: 6896.6, Median: 6899.0, Stddev: 16.04057355583023 sys Average: 520.6, Median: 521.0, Stddev: 1.140175425099138 With lock consolidation (ext4): real Average: 320.0, Median: 319.0, Stddev: 1.4142135623730951 user Average: 6896.8, Median: 6878.0, Stddev: 28.621670111997307 sys Average: 521.2, Median: 521.0, Stddev: 1.7888543819998317 Without lock consolidation (ext4) real Average: 319.6, Median: 319.0, Stddev: 0.8944271909999159 user Average: 6886.2, Median: 6887.0, Stddev: 16.93221781102523 sys Average: 520.4, Median: 520.0, Stddev: 1.140175425099138 The difference is entirely within the noise of a typical run on zram. This hardly justifies the complexity of maintaining both the pool lock and the class lock. In fact, for writeback, we would need to introduce yet another lock to prevent data races on the pool's LRU, further complicating the lock handling logic. IMHO, it is just better to collapse all of these into a single pool-level lock. Link: https://lkml.kernel.org/r/20221128191616.1261026-4-nphamcs@gmail.com Signed-off-by: Nhat Pham <nphamcs@gmail.com> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Acked-by: Minchan Kim <minchan@kernel.org> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-11-28 19:16:12 +00:00
/*
* protect the race between zpage migration and zs_free
* as well as zpage allocation/free
*/
spin_lock(&pool->lock);
zsmalloc: rework compaction algorithm The zsmalloc compaction algorithm has the potential to waste some CPU cycles, particularly when compacting pages within the same fullness group. This is due to the way it selects the head page of the fullness list for source and destination pages, and how it reinserts those pages during each iteration. The algorithm may first use a page as a migration destination and then as a migration source, leading to an unnecessary back-and-forth movement of objects. Consider the following fullness list: PageA PageB PageC PageD PageE During the first iteration, the compaction algorithm will select PageA as the source and PageB as the destination. All of PageA's objects will be moved to PageB, and then PageA will be released while PageB is reinserted into the fullness list. PageB PageC PageD PageE During the next iteration, the compaction algorithm will again select the head of the list as the source and destination, meaning that PageB will now serve as the source and PageC as the destination. This will result in the objects being moved away from PageB, the same objects that were just moved to PageB in the previous iteration. To prevent this avalanche effect, the compaction algorithm should not reinsert the destination page between iterations. By doing so, the most optimal page will continue to be used and its usage ratio will increase, reducing internal fragmentation. The destination page should only be reinserted into the fullness list if: - It becomes full - No source page is available. TEST ==== It's very challenging to reliably test this series. I ended up developing my own synthetic test that has 100% reproducibility. The test generates significan fragmentation (for each size class) and then performs compaction for each class individually and tracks the number of memcpy() in zs_object_copy(), so that we can compare the amount work compaction does on per-class basis. Total amount of work (zram mm_stat objs_moved) ---------------------------------------------- Old fullness grouping, old compaction algorithm: 323977 memcpy() in zs_object_copy(). Old fullness grouping, new compaction algorithm: 262944 memcpy() in zs_object_copy(). New fullness grouping, new compaction algorithm: 213978 memcpy() in zs_object_copy(). Per-class compaction memcpy() comparison (T-test) ------------------------------------------------- x Old fullness grouping, old compaction algorithm + Old fullness grouping, new compaction algorithm N Min Max Median Avg Stddev x 140 349 3513 2461 2314.1214 806.03271 + 140 289 2778 2006 1878.1714 641.02073 Difference at 95.0% confidence -435.95 +/- 170.595 -18.8387% +/- 7.37193% (Student's t, pooled s = 728.216) x Old fullness grouping, old compaction algorithm + New fullness grouping, new compaction algorithm N Min Max Median Avg Stddev x 140 349 3513 2461 2314.1214 806.03271 + 140 226 2279 1644 1528.4143 524.85268 Difference at 95.0% confidence -785.707 +/- 159.331 -33.9527% +/- 6.88516% (Student's t, pooled s = 680.132) Link: https://lkml.kernel.org/r/20230304034835.2082479-4-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:34 +00:00
while (zs_can_compact(class)) {
int fg;
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
zsmalloc: rework compaction algorithm The zsmalloc compaction algorithm has the potential to waste some CPU cycles, particularly when compacting pages within the same fullness group. This is due to the way it selects the head page of the fullness list for source and destination pages, and how it reinserts those pages during each iteration. The algorithm may first use a page as a migration destination and then as a migration source, leading to an unnecessary back-and-forth movement of objects. Consider the following fullness list: PageA PageB PageC PageD PageE During the first iteration, the compaction algorithm will select PageA as the source and PageB as the destination. All of PageA's objects will be moved to PageB, and then PageA will be released while PageB is reinserted into the fullness list. PageB PageC PageD PageE During the next iteration, the compaction algorithm will again select the head of the list as the source and destination, meaning that PageB will now serve as the source and PageC as the destination. This will result in the objects being moved away from PageB, the same objects that were just moved to PageB in the previous iteration. To prevent this avalanche effect, the compaction algorithm should not reinsert the destination page between iterations. By doing so, the most optimal page will continue to be used and its usage ratio will increase, reducing internal fragmentation. The destination page should only be reinserted into the fullness list if: - It becomes full - No source page is available. TEST ==== It's very challenging to reliably test this series. I ended up developing my own synthetic test that has 100% reproducibility. The test generates significan fragmentation (for each size class) and then performs compaction for each class individually and tracks the number of memcpy() in zs_object_copy(), so that we can compare the amount work compaction does on per-class basis. Total amount of work (zram mm_stat objs_moved) ---------------------------------------------- Old fullness grouping, old compaction algorithm: 323977 memcpy() in zs_object_copy(). Old fullness grouping, new compaction algorithm: 262944 memcpy() in zs_object_copy(). New fullness grouping, new compaction algorithm: 213978 memcpy() in zs_object_copy(). Per-class compaction memcpy() comparison (T-test) ------------------------------------------------- x Old fullness grouping, old compaction algorithm + Old fullness grouping, new compaction algorithm N Min Max Median Avg Stddev x 140 349 3513 2461 2314.1214 806.03271 + 140 289 2778 2006 1878.1714 641.02073 Difference at 95.0% confidence -435.95 +/- 170.595 -18.8387% +/- 7.37193% (Student's t, pooled s = 728.216) x Old fullness grouping, old compaction algorithm + New fullness grouping, new compaction algorithm N Min Max Median Avg Stddev x 140 349 3513 2461 2314.1214 806.03271 + 140 226 2279 1644 1528.4143 524.85268 Difference at 95.0% confidence -785.707 +/- 159.331 -33.9527% +/- 6.88516% (Student's t, pooled s = 680.132) Link: https://lkml.kernel.org/r/20230304034835.2082479-4-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:34 +00:00
if (!dst_zspage) {
dst_zspage = isolate_dst_zspage(class);
if (!dst_zspage)
break;
migrate_write_lock(dst_zspage);
cc.d_page = get_first_page(dst_zspage);
}
src_zspage = isolate_src_zspage(class);
if (!src_zspage)
zsmalloc: introduce zs_can_compact() function This function checks if class compaction will free any pages. Rephrasing -- do we have enough unused objects to form at least one ZS_EMPTY page and free it. It aborts compaction if class compaction will not result in any (further) savings. EXAMPLE (this debug output is not part of this patch set): - class size - number of allocated objects - number of used objects - max objects per zspage - pages per zspage - estimated number of pages that will be freed [..] class-512 objs:544 inuse:540 maxobj-per-zspage:8 pages-per-zspage:1 zspages-to-free:0 ... class-512 compaction is useless. break class-496 objs:660 inuse:570 maxobj-per-zspage:33 pages-per-zspage:4 zspages-to-free:2 class-496 objs:627 inuse:570 maxobj-per-zspage:33 pages-per-zspage:4 zspages-to-free:1 class-496 objs:594 inuse:570 maxobj-per-zspage:33 pages-per-zspage:4 zspages-to-free:0 ... class-496 compaction is useless. break class-448 objs:657 inuse:617 maxobj-per-zspage:9 pages-per-zspage:1 zspages-to-free:4 class-448 objs:648 inuse:617 maxobj-per-zspage:9 pages-per-zspage:1 zspages-to-free:3 class-448 objs:639 inuse:617 maxobj-per-zspage:9 pages-per-zspage:1 zspages-to-free:2 class-448 objs:630 inuse:617 maxobj-per-zspage:9 pages-per-zspage:1 zspages-to-free:1 class-448 objs:621 inuse:617 maxobj-per-zspage:9 pages-per-zspage:1 zspages-to-free:0 ... class-448 compaction is useless. break class-432 objs:728 inuse:685 maxobj-per-zspage:28 pages-per-zspage:3 zspages-to-free:1 class-432 objs:700 inuse:685 maxobj-per-zspage:28 pages-per-zspage:3 zspages-to-free:0 ... class-432 compaction is useless. break class-416 objs:819 inuse:705 maxobj-per-zspage:39 pages-per-zspage:4 zspages-to-free:2 class-416 objs:780 inuse:705 maxobj-per-zspage:39 pages-per-zspage:4 zspages-to-free:1 class-416 objs:741 inuse:705 maxobj-per-zspage:39 pages-per-zspage:4 zspages-to-free:0 ... class-416 compaction is useless. break class-400 objs:690 inuse:674 maxobj-per-zspage:10 pages-per-zspage:1 zspages-to-free:1 class-400 objs:680 inuse:674 maxobj-per-zspage:10 pages-per-zspage:1 zspages-to-free:0 ... class-400 compaction is useless. break class-384 objs:736 inuse:709 maxobj-per-zspage:32 pages-per-zspage:3 zspages-to-free:0 ... class-384 compaction is useless. break [..] Every "compaction is useless" indicates that we saved CPU cycles. class-512 has 544 object allocated 540 objects used 8 objects per-page Even if we have a ALMOST_EMPTY zspage, we still don't have enough room to migrate all of its objects and free this zspage; so compaction will not make a lot of sense, it's better to just leave it as is. Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Acked-by: Minchan Kim <minchan@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-09-08 22:04:30 +00:00
break;
zsmalloc: rework compaction algorithm The zsmalloc compaction algorithm has the potential to waste some CPU cycles, particularly when compacting pages within the same fullness group. This is due to the way it selects the head page of the fullness list for source and destination pages, and how it reinserts those pages during each iteration. The algorithm may first use a page as a migration destination and then as a migration source, leading to an unnecessary back-and-forth movement of objects. Consider the following fullness list: PageA PageB PageC PageD PageE During the first iteration, the compaction algorithm will select PageA as the source and PageB as the destination. All of PageA's objects will be moved to PageB, and then PageA will be released while PageB is reinserted into the fullness list. PageB PageC PageD PageE During the next iteration, the compaction algorithm will again select the head of the list as the source and destination, meaning that PageB will now serve as the source and PageC as the destination. This will result in the objects being moved away from PageB, the same objects that were just moved to PageB in the previous iteration. To prevent this avalanche effect, the compaction algorithm should not reinsert the destination page between iterations. By doing so, the most optimal page will continue to be used and its usage ratio will increase, reducing internal fragmentation. The destination page should only be reinserted into the fullness list if: - It becomes full - No source page is available. TEST ==== It's very challenging to reliably test this series. I ended up developing my own synthetic test that has 100% reproducibility. The test generates significan fragmentation (for each size class) and then performs compaction for each class individually and tracks the number of memcpy() in zs_object_copy(), so that we can compare the amount work compaction does on per-class basis. Total amount of work (zram mm_stat objs_moved) ---------------------------------------------- Old fullness grouping, old compaction algorithm: 323977 memcpy() in zs_object_copy(). Old fullness grouping, new compaction algorithm: 262944 memcpy() in zs_object_copy(). New fullness grouping, new compaction algorithm: 213978 memcpy() in zs_object_copy(). Per-class compaction memcpy() comparison (T-test) ------------------------------------------------- x Old fullness grouping, old compaction algorithm + Old fullness grouping, new compaction algorithm N Min Max Median Avg Stddev x 140 349 3513 2461 2314.1214 806.03271 + 140 289 2778 2006 1878.1714 641.02073 Difference at 95.0% confidence -435.95 +/- 170.595 -18.8387% +/- 7.37193% (Student's t, pooled s = 728.216) x Old fullness grouping, old compaction algorithm + New fullness grouping, new compaction algorithm N Min Max Median Avg Stddev x 140 349 3513 2461 2314.1214 806.03271 + 140 226 2279 1644 1528.4143 524.85268 Difference at 95.0% confidence -785.707 +/- 159.331 -33.9527% +/- 6.88516% (Student's t, pooled s = 680.132) Link: https://lkml.kernel.org/r/20230304034835.2082479-4-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:34 +00:00
migrate_write_lock_nested(src_zspage);
cc.obj_idx = 0;
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
cc.s_page = get_first_page(src_zspage);
zsmalloc: rework compaction algorithm The zsmalloc compaction algorithm has the potential to waste some CPU cycles, particularly when compacting pages within the same fullness group. This is due to the way it selects the head page of the fullness list for source and destination pages, and how it reinserts those pages during each iteration. The algorithm may first use a page as a migration destination and then as a migration source, leading to an unnecessary back-and-forth movement of objects. Consider the following fullness list: PageA PageB PageC PageD PageE During the first iteration, the compaction algorithm will select PageA as the source and PageB as the destination. All of PageA's objects will be moved to PageB, and then PageA will be released while PageB is reinserted into the fullness list. PageB PageC PageD PageE During the next iteration, the compaction algorithm will again select the head of the list as the source and destination, meaning that PageB will now serve as the source and PageC as the destination. This will result in the objects being moved away from PageB, the same objects that were just moved to PageB in the previous iteration. To prevent this avalanche effect, the compaction algorithm should not reinsert the destination page between iterations. By doing so, the most optimal page will continue to be used and its usage ratio will increase, reducing internal fragmentation. The destination page should only be reinserted into the fullness list if: - It becomes full - No source page is available. TEST ==== It's very challenging to reliably test this series. I ended up developing my own synthetic test that has 100% reproducibility. The test generates significan fragmentation (for each size class) and then performs compaction for each class individually and tracks the number of memcpy() in zs_object_copy(), so that we can compare the amount work compaction does on per-class basis. Total amount of work (zram mm_stat objs_moved) ---------------------------------------------- Old fullness grouping, old compaction algorithm: 323977 memcpy() in zs_object_copy(). Old fullness grouping, new compaction algorithm: 262944 memcpy() in zs_object_copy(). New fullness grouping, new compaction algorithm: 213978 memcpy() in zs_object_copy(). Per-class compaction memcpy() comparison (T-test) ------------------------------------------------- x Old fullness grouping, old compaction algorithm + Old fullness grouping, new compaction algorithm N Min Max Median Avg Stddev x 140 349 3513 2461 2314.1214 806.03271 + 140 289 2778 2006 1878.1714 641.02073 Difference at 95.0% confidence -435.95 +/- 170.595 -18.8387% +/- 7.37193% (Student's t, pooled s = 728.216) x Old fullness grouping, old compaction algorithm + New fullness grouping, new compaction algorithm N Min Max Median Avg Stddev x 140 349 3513 2461 2314.1214 806.03271 + 140 226 2279 1644 1528.4143 524.85268 Difference at 95.0% confidence -785.707 +/- 159.331 -33.9527% +/- 6.88516% (Student's t, pooled s = 680.132) Link: https://lkml.kernel.org/r/20230304034835.2082479-4-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:34 +00:00
migrate_zspage(pool, class, &cc);
fg = putback_zspage(class, src_zspage);
migrate_write_unlock(src_zspage);
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
zsmalloc: rework compaction algorithm The zsmalloc compaction algorithm has the potential to waste some CPU cycles, particularly when compacting pages within the same fullness group. This is due to the way it selects the head page of the fullness list for source and destination pages, and how it reinserts those pages during each iteration. The algorithm may first use a page as a migration destination and then as a migration source, leading to an unnecessary back-and-forth movement of objects. Consider the following fullness list: PageA PageB PageC PageD PageE During the first iteration, the compaction algorithm will select PageA as the source and PageB as the destination. All of PageA's objects will be moved to PageB, and then PageA will be released while PageB is reinserted into the fullness list. PageB PageC PageD PageE During the next iteration, the compaction algorithm will again select the head of the list as the source and destination, meaning that PageB will now serve as the source and PageC as the destination. This will result in the objects being moved away from PageB, the same objects that were just moved to PageB in the previous iteration. To prevent this avalanche effect, the compaction algorithm should not reinsert the destination page between iterations. By doing so, the most optimal page will continue to be used and its usage ratio will increase, reducing internal fragmentation. The destination page should only be reinserted into the fullness list if: - It becomes full - No source page is available. TEST ==== It's very challenging to reliably test this series. I ended up developing my own synthetic test that has 100% reproducibility. The test generates significan fragmentation (for each size class) and then performs compaction for each class individually and tracks the number of memcpy() in zs_object_copy(), so that we can compare the amount work compaction does on per-class basis. Total amount of work (zram mm_stat objs_moved) ---------------------------------------------- Old fullness grouping, old compaction algorithm: 323977 memcpy() in zs_object_copy(). Old fullness grouping, new compaction algorithm: 262944 memcpy() in zs_object_copy(). New fullness grouping, new compaction algorithm: 213978 memcpy() in zs_object_copy(). Per-class compaction memcpy() comparison (T-test) ------------------------------------------------- x Old fullness grouping, old compaction algorithm + Old fullness grouping, new compaction algorithm N Min Max Median Avg Stddev x 140 349 3513 2461 2314.1214 806.03271 + 140 289 2778 2006 1878.1714 641.02073 Difference at 95.0% confidence -435.95 +/- 170.595 -18.8387% +/- 7.37193% (Student's t, pooled s = 728.216) x Old fullness grouping, old compaction algorithm + New fullness grouping, new compaction algorithm N Min Max Median Avg Stddev x 140 349 3513 2461 2314.1214 806.03271 + 140 226 2279 1644 1528.4143 524.85268 Difference at 95.0% confidence -785.707 +/- 159.331 -33.9527% +/- 6.88516% (Student's t, pooled s = 680.132) Link: https://lkml.kernel.org/r/20230304034835.2082479-4-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:34 +00:00
if (fg == ZS_INUSE_RATIO_0) {
free_zspage(pool, class, src_zspage);
pages_freed += class->pages_per_zspage;
}
src_zspage = NULL;
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
zsmalloc: rework compaction algorithm The zsmalloc compaction algorithm has the potential to waste some CPU cycles, particularly when compacting pages within the same fullness group. This is due to the way it selects the head page of the fullness list for source and destination pages, and how it reinserts those pages during each iteration. The algorithm may first use a page as a migration destination and then as a migration source, leading to an unnecessary back-and-forth movement of objects. Consider the following fullness list: PageA PageB PageC PageD PageE During the first iteration, the compaction algorithm will select PageA as the source and PageB as the destination. All of PageA's objects will be moved to PageB, and then PageA will be released while PageB is reinserted into the fullness list. PageB PageC PageD PageE During the next iteration, the compaction algorithm will again select the head of the list as the source and destination, meaning that PageB will now serve as the source and PageC as the destination. This will result in the objects being moved away from PageB, the same objects that were just moved to PageB in the previous iteration. To prevent this avalanche effect, the compaction algorithm should not reinsert the destination page between iterations. By doing so, the most optimal page will continue to be used and its usage ratio will increase, reducing internal fragmentation. The destination page should only be reinserted into the fullness list if: - It becomes full - No source page is available. TEST ==== It's very challenging to reliably test this series. I ended up developing my own synthetic test that has 100% reproducibility. The test generates significan fragmentation (for each size class) and then performs compaction for each class individually and tracks the number of memcpy() in zs_object_copy(), so that we can compare the amount work compaction does on per-class basis. Total amount of work (zram mm_stat objs_moved) ---------------------------------------------- Old fullness grouping, old compaction algorithm: 323977 memcpy() in zs_object_copy(). Old fullness grouping, new compaction algorithm: 262944 memcpy() in zs_object_copy(). New fullness grouping, new compaction algorithm: 213978 memcpy() in zs_object_copy(). Per-class compaction memcpy() comparison (T-test) ------------------------------------------------- x Old fullness grouping, old compaction algorithm + Old fullness grouping, new compaction algorithm N Min Max Median Avg Stddev x 140 349 3513 2461 2314.1214 806.03271 + 140 289 2778 2006 1878.1714 641.02073 Difference at 95.0% confidence -435.95 +/- 170.595 -18.8387% +/- 7.37193% (Student's t, pooled s = 728.216) x Old fullness grouping, old compaction algorithm + New fullness grouping, new compaction algorithm N Min Max Median Avg Stddev x 140 349 3513 2461 2314.1214 806.03271 + 140 226 2279 1644 1528.4143 524.85268 Difference at 95.0% confidence -785.707 +/- 159.331 -33.9527% +/- 6.88516% (Student's t, pooled s = 680.132) Link: https://lkml.kernel.org/r/20230304034835.2082479-4-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:34 +00:00
if (get_fullness_group(class, dst_zspage) == ZS_INUSE_RATIO_100
|| spin_is_contended(&pool->lock)) {
putback_zspage(class, dst_zspage);
zsmalloc: replace per zpage lock with pool->migrate_lock The zsmalloc has used a bit for spin_lock in zpage handle to keep zpage object alive during several operations. However, it causes the problem for PREEMPT_RT as well as introducing too complicated. This patch replaces the bit spin_lock with pool->migrate_lock rwlock. It could make the code simple as well as zsmalloc work under PREEMPT_RT. The drawback is the pool->migrate_lock is bigger granuarity than per zpage lock so the contention would be higher than old when both IO-related operations(i.e., zsmalloc, zsfree, zs_[map|unmap]) and compaction(page/zpage migration) are going in parallel(*, the migrate_lock is rwlock and IO related functions are all read side lock so there is no contention). However, the write-side is fast enough(dominant overhead is just page copy) so it wouldn't affect much. If the lock granurity becomes more problem later, we could introduce table locks based on handle as a hash value. Link: https://lkml.kernel.org/r/20211115185909.3949505-9-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:14:13 +00:00
migrate_write_unlock(dst_zspage);
dst_zspage = NULL;
zsmalloc: rework compaction algorithm The zsmalloc compaction algorithm has the potential to waste some CPU cycles, particularly when compacting pages within the same fullness group. This is due to the way it selects the head page of the fullness list for source and destination pages, and how it reinserts those pages during each iteration. The algorithm may first use a page as a migration destination and then as a migration source, leading to an unnecessary back-and-forth movement of objects. Consider the following fullness list: PageA PageB PageC PageD PageE During the first iteration, the compaction algorithm will select PageA as the source and PageB as the destination. All of PageA's objects will be moved to PageB, and then PageA will be released while PageB is reinserted into the fullness list. PageB PageC PageD PageE During the next iteration, the compaction algorithm will again select the head of the list as the source and destination, meaning that PageB will now serve as the source and PageC as the destination. This will result in the objects being moved away from PageB, the same objects that were just moved to PageB in the previous iteration. To prevent this avalanche effect, the compaction algorithm should not reinsert the destination page between iterations. By doing so, the most optimal page will continue to be used and its usage ratio will increase, reducing internal fragmentation. The destination page should only be reinserted into the fullness list if: - It becomes full - No source page is available. TEST ==== It's very challenging to reliably test this series. I ended up developing my own synthetic test that has 100% reproducibility. The test generates significan fragmentation (for each size class) and then performs compaction for each class individually and tracks the number of memcpy() in zs_object_copy(), so that we can compare the amount work compaction does on per-class basis. Total amount of work (zram mm_stat objs_moved) ---------------------------------------------- Old fullness grouping, old compaction algorithm: 323977 memcpy() in zs_object_copy(). Old fullness grouping, new compaction algorithm: 262944 memcpy() in zs_object_copy(). New fullness grouping, new compaction algorithm: 213978 memcpy() in zs_object_copy(). Per-class compaction memcpy() comparison (T-test) ------------------------------------------------- x Old fullness grouping, old compaction algorithm + Old fullness grouping, new compaction algorithm N Min Max Median Avg Stddev x 140 349 3513 2461 2314.1214 806.03271 + 140 289 2778 2006 1878.1714 641.02073 Difference at 95.0% confidence -435.95 +/- 170.595 -18.8387% +/- 7.37193% (Student's t, pooled s = 728.216) x Old fullness grouping, old compaction algorithm + New fullness grouping, new compaction algorithm N Min Max Median Avg Stddev x 140 349 3513 2461 2314.1214 806.03271 + 140 226 2279 1644 1528.4143 524.85268 Difference at 95.0% confidence -785.707 +/- 159.331 -33.9527% +/- 6.88516% (Student's t, pooled s = 680.132) Link: https://lkml.kernel.org/r/20230304034835.2082479-4-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:34 +00:00
spin_unlock(&pool->lock);
cond_resched();
spin_lock(&pool->lock);
}
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
}
zsmalloc: replace per zpage lock with pool->migrate_lock The zsmalloc has used a bit for spin_lock in zpage handle to keep zpage object alive during several operations. However, it causes the problem for PREEMPT_RT as well as introducing too complicated. This patch replaces the bit spin_lock with pool->migrate_lock rwlock. It could make the code simple as well as zsmalloc work under PREEMPT_RT. The drawback is the pool->migrate_lock is bigger granuarity than per zpage lock so the contention would be higher than old when both IO-related operations(i.e., zsmalloc, zsfree, zs_[map|unmap]) and compaction(page/zpage migration) are going in parallel(*, the migrate_lock is rwlock and IO related functions are all read side lock so there is no contention). However, the write-side is fast enough(dominant overhead is just page copy) so it wouldn't affect much. If the lock granurity becomes more problem later, we could introduce table locks based on handle as a hash value. Link: https://lkml.kernel.org/r/20211115185909.3949505-9-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:14:13 +00:00
if (src_zspage) {
putback_zspage(class, src_zspage);
zsmalloc: replace per zpage lock with pool->migrate_lock The zsmalloc has used a bit for spin_lock in zpage handle to keep zpage object alive during several operations. However, it causes the problem for PREEMPT_RT as well as introducing too complicated. This patch replaces the bit spin_lock with pool->migrate_lock rwlock. It could make the code simple as well as zsmalloc work under PREEMPT_RT. The drawback is the pool->migrate_lock is bigger granuarity than per zpage lock so the contention would be higher than old when both IO-related operations(i.e., zsmalloc, zsfree, zs_[map|unmap]) and compaction(page/zpage migration) are going in parallel(*, the migrate_lock is rwlock and IO related functions are all read side lock so there is no contention). However, the write-side is fast enough(dominant overhead is just page copy) so it wouldn't affect much. If the lock granurity becomes more problem later, we could introduce table locks based on handle as a hash value. Link: https://lkml.kernel.org/r/20211115185909.3949505-9-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Tested-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Cc: Mike Galbraith <umgwanakikbuti@gmail.com> Cc: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2022-01-22 06:14:13 +00:00
migrate_write_unlock(src_zspage);
}
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
zsmalloc: rework compaction algorithm The zsmalloc compaction algorithm has the potential to waste some CPU cycles, particularly when compacting pages within the same fullness group. This is due to the way it selects the head page of the fullness list for source and destination pages, and how it reinserts those pages during each iteration. The algorithm may first use a page as a migration destination and then as a migration source, leading to an unnecessary back-and-forth movement of objects. Consider the following fullness list: PageA PageB PageC PageD PageE During the first iteration, the compaction algorithm will select PageA as the source and PageB as the destination. All of PageA's objects will be moved to PageB, and then PageA will be released while PageB is reinserted into the fullness list. PageB PageC PageD PageE During the next iteration, the compaction algorithm will again select the head of the list as the source and destination, meaning that PageB will now serve as the source and PageC as the destination. This will result in the objects being moved away from PageB, the same objects that were just moved to PageB in the previous iteration. To prevent this avalanche effect, the compaction algorithm should not reinsert the destination page between iterations. By doing so, the most optimal page will continue to be used and its usage ratio will increase, reducing internal fragmentation. The destination page should only be reinserted into the fullness list if: - It becomes full - No source page is available. TEST ==== It's very challenging to reliably test this series. I ended up developing my own synthetic test that has 100% reproducibility. The test generates significan fragmentation (for each size class) and then performs compaction for each class individually and tracks the number of memcpy() in zs_object_copy(), so that we can compare the amount work compaction does on per-class basis. Total amount of work (zram mm_stat objs_moved) ---------------------------------------------- Old fullness grouping, old compaction algorithm: 323977 memcpy() in zs_object_copy(). Old fullness grouping, new compaction algorithm: 262944 memcpy() in zs_object_copy(). New fullness grouping, new compaction algorithm: 213978 memcpy() in zs_object_copy(). Per-class compaction memcpy() comparison (T-test) ------------------------------------------------- x Old fullness grouping, old compaction algorithm + Old fullness grouping, new compaction algorithm N Min Max Median Avg Stddev x 140 349 3513 2461 2314.1214 806.03271 + 140 289 2778 2006 1878.1714 641.02073 Difference at 95.0% confidence -435.95 +/- 170.595 -18.8387% +/- 7.37193% (Student's t, pooled s = 728.216) x Old fullness grouping, old compaction algorithm + New fullness grouping, new compaction algorithm N Min Max Median Avg Stddev x 140 349 3513 2461 2314.1214 806.03271 + 140 226 2279 1644 1528.4143 524.85268 Difference at 95.0% confidence -785.707 +/- 159.331 -33.9527% +/- 6.88516% (Student's t, pooled s = 680.132) Link: https://lkml.kernel.org/r/20230304034835.2082479-4-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:34 +00:00
if (dst_zspage) {
putback_zspage(class, dst_zspage);
migrate_write_unlock(dst_zspage);
}
zsmalloc: consolidate zs_pool's migrate_lock and size_class's locks Currently, zsmalloc has a hierarchy of locks, which includes a pool-level migrate_lock, and a lock for each size class. We have to obtain both locks in the hotpath in most cases anyway, except for zs_malloc. This exception will no longer exist when we introduce a LRU into the zs_pool for the new writeback functionality - we will need to obtain a pool-level lock to synchronize LRU handling even in zs_malloc. In preparation for zsmalloc writeback, consolidate these locks into a single pool-level lock, which drastically reduces the complexity of synchronization in zsmalloc. We have also benchmarked the lock consolidation to see the performance effect of this change on zram. First, we ran a synthetic FS workload on a server machine with 36 cores (same machine for all runs), using fs_mark -d ../zram1mnt -s 100000 -n 2500 -t 32 -k before and after for btrfs and ext4 on zram (FS usage is 80%). Here is the result (unit is file/second): With lock consolidation (btrfs): Average: 13520.2, Median: 13531.0, Stddev: 137.5961482019028 Without lock consolidation (btrfs): Average: 13487.2, Median: 13575.0, Stddev: 309.08283679298665 With lock consolidation (ext4): Average: 16824.4, Median: 16839.0, Stddev: 89.97388510006668 Without lock consolidation (ext4) Average: 16958.0, Median: 16986.0, Stddev: 194.7370021336469 As you can see, we observe a 0.3% regression for btrfs, and a 0.9% regression for ext4. This is a small, barely measurable difference in my opinion. For a more realistic scenario, we also tries building the kernel on zram. Here is the time it takes (in seconds): With lock consolidation (btrfs): real Average: 319.6, Median: 320.0, Stddev: 0.8944271909999159 user Average: 6894.2, Median: 6895.0, Stddev: 25.528415540334656 sys Average: 521.4, Median: 522.0, Stddev: 1.51657508881031 Without lock consolidation (btrfs): real Average: 319.8, Median: 320.0, Stddev: 0.8366600265340756 user Average: 6896.6, Median: 6899.0, Stddev: 16.04057355583023 sys Average: 520.6, Median: 521.0, Stddev: 1.140175425099138 With lock consolidation (ext4): real Average: 320.0, Median: 319.0, Stddev: 1.4142135623730951 user Average: 6896.8, Median: 6878.0, Stddev: 28.621670111997307 sys Average: 521.2, Median: 521.0, Stddev: 1.7888543819998317 Without lock consolidation (ext4) real Average: 319.6, Median: 319.0, Stddev: 0.8944271909999159 user Average: 6886.2, Median: 6887.0, Stddev: 16.93221781102523 sys Average: 520.4, Median: 520.0, Stddev: 1.140175425099138 The difference is entirely within the noise of a typical run on zram. This hardly justifies the complexity of maintaining both the pool lock and the class lock. In fact, for writeback, we would need to introduce yet another lock to prevent data races on the pool's LRU, further complicating the lock handling logic. IMHO, it is just better to collapse all of these into a single pool-level lock. Link: https://lkml.kernel.org/r/20221128191616.1261026-4-nphamcs@gmail.com Signed-off-by: Nhat Pham <nphamcs@gmail.com> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Acked-by: Minchan Kim <minchan@kernel.org> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-11-28 19:16:12 +00:00
spin_unlock(&pool->lock);
return pages_freed;
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
}
unsigned long zs_compact(struct zs_pool *pool)
{
int i;
struct size_class *class;
unsigned long pages_freed = 0;
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
/*
* Pool compaction is performed under pool->lock so it is basically
* single-threaded. Having more than one thread in __zs_compact()
* will increase pool->lock contention, which will impact other
* zsmalloc operations that need pool->lock.
*/
if (atomic_xchg(&pool->compaction_in_progress, 1))
return 0;
for (i = ZS_SIZE_CLASSES - 1; i >= 0; i--) {
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
class = pool->size_class[i];
if (class->index != i)
continue;
pages_freed += __zs_compact(pool, class);
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
}
atomic_long_add(pages_freed, &pool->stats.pages_compacted);
atomic_set(&pool->compaction_in_progress, 0);
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
return pages_freed;
zsmalloc: support compaction This patch provides core functions for migration of zsmalloc. Migraion policy is simple as follows. for each size class { while { src_page = get zs_page from ZS_ALMOST_EMPTY if (!src_page) break; dst_page = get zs_page from ZS_ALMOST_FULL if (!dst_page) dst_page = get zs_page from ZS_ALMOST_EMPTY if (!dst_page) break; migrate(from src_page, to dst_page); } } For migration, we need to identify which objects in zspage are allocated to migrate them out. We could know it by iterating of freed objects in a zspage because first_page of zspage keeps free objects singly-linked list but it's not efficient. Instead, this patch adds a tag(ie, OBJ_ALLOCATED_TAG) in header of each object(ie, handle) so we could check whether the object is allocated easily. This patch adds another status bit in handle to synchronize between user access through zs_map_object and migration. During migration, we cannot move objects user are using due to data coherency between old object and new object. [akpm@linux-foundation.org: zsmalloc.c needs sched.h for cond_resched()] Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:30 +00:00
}
EXPORT_SYMBOL_GPL(zs_compact);
void zs_pool_stats(struct zs_pool *pool, struct zs_pool_stats *stats)
{
memcpy(stats, &pool->stats, sizeof(struct zs_pool_stats));
}
EXPORT_SYMBOL_GPL(zs_pool_stats);
static unsigned long zs_shrinker_scan(struct shrinker *shrinker,
struct shrink_control *sc)
{
unsigned long pages_freed;
struct zs_pool *pool = container_of(shrinker, struct zs_pool,
shrinker);
/*
* Compact classes and calculate compaction delta.
* Can run concurrently with a manually triggered
* (by user) compaction.
*/
pages_freed = zs_compact(pool);
return pages_freed ? pages_freed : SHRINK_STOP;
}
static unsigned long zs_shrinker_count(struct shrinker *shrinker,
struct shrink_control *sc)
{
int i;
struct size_class *class;
unsigned long pages_to_free = 0;
struct zs_pool *pool = container_of(shrinker, struct zs_pool,
shrinker);
for (i = ZS_SIZE_CLASSES - 1; i >= 0; i--) {
class = pool->size_class[i];
if (class->index != i)
continue;
pages_to_free += zs_can_compact(class);
}
return pages_to_free;
}
static void zs_unregister_shrinker(struct zs_pool *pool)
{
unregister_shrinker(&pool->shrinker);
}
static int zs_register_shrinker(struct zs_pool *pool)
{
pool->shrinker.scan_objects = zs_shrinker_scan;
pool->shrinker.count_objects = zs_shrinker_count;
pool->shrinker.batch = 0;
pool->shrinker.seeks = DEFAULT_SEEKS;
mm: shrinkers: provide shrinkers with names Currently shrinkers are anonymous objects. For debugging purposes they can be identified by count/scan function names, but it's not always useful: e.g. for superblock's shrinkers it's nice to have at least an idea of to which superblock the shrinker belongs. This commit adds names to shrinkers. register_shrinker() and prealloc_shrinker() functions are extended to take a format and arguments to master a name. In some cases it's not possible to determine a good name at the time when a shrinker is allocated. For such cases shrinker_debugfs_rename() is provided. The expected format is: <subsystem>-<shrinker_type>[:<instance>]-<id> For some shrinkers an instance can be encoded as (MAJOR:MINOR) pair. After this change the shrinker debugfs directory looks like: $ cd /sys/kernel/debug/shrinker/ $ ls dquota-cache-16 sb-devpts-28 sb-proc-47 sb-tmpfs-42 mm-shadow-18 sb-devtmpfs-5 sb-proc-48 sb-tmpfs-43 mm-zspool:zram0-34 sb-hugetlbfs-17 sb-pstore-31 sb-tmpfs-44 rcu-kfree-0 sb-hugetlbfs-33 sb-rootfs-2 sb-tmpfs-49 sb-aio-20 sb-iomem-12 sb-securityfs-6 sb-tracefs-13 sb-anon_inodefs-15 sb-mqueue-21 sb-selinuxfs-22 sb-xfs:vda1-36 sb-bdev-3 sb-nsfs-4 sb-sockfs-8 sb-zsmalloc-19 sb-bpf-32 sb-pipefs-14 sb-sysfs-26 thp-deferred_split-10 sb-btrfs:vda2-24 sb-proc-25 sb-tmpfs-1 thp-zero-9 sb-cgroup2-30 sb-proc-39 sb-tmpfs-27 xfs-buf:vda1-37 sb-configfs-23 sb-proc-41 sb-tmpfs-29 xfs-inodegc:vda1-38 sb-dax-11 sb-proc-45 sb-tmpfs-35 sb-debugfs-7 sb-proc-46 sb-tmpfs-40 [roman.gushchin@linux.dev: fix build warnings] Link: https://lkml.kernel.org/r/Yr+ZTnLb9lJk6fJO@castle Reported-by: kernel test robot <lkp@intel.com> Link: https://lkml.kernel.org/r/20220601032227.4076670-4-roman.gushchin@linux.dev Signed-off-by: Roman Gushchin <roman.gushchin@linux.dev> Cc: Christophe JAILLET <christophe.jaillet@wanadoo.fr> Cc: Dave Chinner <dchinner@redhat.com> Cc: Hillf Danton <hdanton@sina.com> Cc: Kent Overstreet <kent.overstreet@gmail.com> Cc: Muchun Song <songmuchun@bytedance.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-06-01 03:22:24 +00:00
return register_shrinker(&pool->shrinker, "mm-zspool:%s",
pool->name);
}
static int calculate_zspage_chain_size(int class_size)
{
int i, min_waste = INT_MAX;
int chain_size = 1;
if (is_power_of_2(class_size))
return chain_size;
for (i = 1; i <= ZS_MAX_PAGES_PER_ZSPAGE; i++) {
int waste;
waste = (i * PAGE_SIZE) % class_size;
if (waste < min_waste) {
min_waste = waste;
chain_size = i;
}
}
return chain_size;
}
/**
* zs_create_pool - Creates an allocation pool to work from.
* @name: pool name to be created
*
* This function must be called before anything when using
* the zsmalloc allocator.
*
* On success, a pointer to the newly created pool is returned,
* otherwise NULL.
*/
struct zs_pool *zs_create_pool(const char *name)
{
int i;
struct zs_pool *pool;
struct size_class *prev_class = NULL;
pool = kzalloc(sizeof(*pool), GFP_KERNEL);
if (!pool)
return NULL;
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
init_deferred_free(pool);
zsmalloc: consolidate zs_pool's migrate_lock and size_class's locks Currently, zsmalloc has a hierarchy of locks, which includes a pool-level migrate_lock, and a lock for each size class. We have to obtain both locks in the hotpath in most cases anyway, except for zs_malloc. This exception will no longer exist when we introduce a LRU into the zs_pool for the new writeback functionality - we will need to obtain a pool-level lock to synchronize LRU handling even in zs_malloc. In preparation for zsmalloc writeback, consolidate these locks into a single pool-level lock, which drastically reduces the complexity of synchronization in zsmalloc. We have also benchmarked the lock consolidation to see the performance effect of this change on zram. First, we ran a synthetic FS workload on a server machine with 36 cores (same machine for all runs), using fs_mark -d ../zram1mnt -s 100000 -n 2500 -t 32 -k before and after for btrfs and ext4 on zram (FS usage is 80%). Here is the result (unit is file/second): With lock consolidation (btrfs): Average: 13520.2, Median: 13531.0, Stddev: 137.5961482019028 Without lock consolidation (btrfs): Average: 13487.2, Median: 13575.0, Stddev: 309.08283679298665 With lock consolidation (ext4): Average: 16824.4, Median: 16839.0, Stddev: 89.97388510006668 Without lock consolidation (ext4) Average: 16958.0, Median: 16986.0, Stddev: 194.7370021336469 As you can see, we observe a 0.3% regression for btrfs, and a 0.9% regression for ext4. This is a small, barely measurable difference in my opinion. For a more realistic scenario, we also tries building the kernel on zram. Here is the time it takes (in seconds): With lock consolidation (btrfs): real Average: 319.6, Median: 320.0, Stddev: 0.8944271909999159 user Average: 6894.2, Median: 6895.0, Stddev: 25.528415540334656 sys Average: 521.4, Median: 522.0, Stddev: 1.51657508881031 Without lock consolidation (btrfs): real Average: 319.8, Median: 320.0, Stddev: 0.8366600265340756 user Average: 6896.6, Median: 6899.0, Stddev: 16.04057355583023 sys Average: 520.6, Median: 521.0, Stddev: 1.140175425099138 With lock consolidation (ext4): real Average: 320.0, Median: 319.0, Stddev: 1.4142135623730951 user Average: 6896.8, Median: 6878.0, Stddev: 28.621670111997307 sys Average: 521.2, Median: 521.0, Stddev: 1.7888543819998317 Without lock consolidation (ext4) real Average: 319.6, Median: 319.0, Stddev: 0.8944271909999159 user Average: 6886.2, Median: 6887.0, Stddev: 16.93221781102523 sys Average: 520.4, Median: 520.0, Stddev: 1.140175425099138 The difference is entirely within the noise of a typical run on zram. This hardly justifies the complexity of maintaining both the pool lock and the class lock. In fact, for writeback, we would need to introduce yet another lock to prevent data races on the pool's LRU, further complicating the lock handling logic. IMHO, it is just better to collapse all of these into a single pool-level lock. Link: https://lkml.kernel.org/r/20221128191616.1261026-4-nphamcs@gmail.com Signed-off-by: Nhat Pham <nphamcs@gmail.com> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Acked-by: Minchan Kim <minchan@kernel.org> Acked-by: Johannes Weiner <hannes@cmpxchg.org> Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2022-11-28 19:16:12 +00:00
spin_lock_init(&pool->lock);
atomic_set(&pool->compaction_in_progress, 0);
zsmalloc: decouple handle and object Recently, we started to use zram heavily and some of issues popped. 1) external fragmentation I got a report from Juneho Choi that fork failed although there are plenty of free pages in the system. His investigation revealed zram is one of the culprit to make heavy fragmentation so there was no more contiguous 16K page for pgd to fork in the ARM. 2) non-movable pages Other problem of zram now is that inherently, user want to use zram as swap in small memory system so they use zRAM with CMA to use memory efficiently. However, unfortunately, it doesn't work well because zRAM cannot use CMA's movable pages unless it doesn't support compaction. I got several reports about that OOM happened with zram although there are lots of swap space and free space in CMA area. 3) internal fragmentation zRAM has started support memory limitation feature to limit memory usage and I sent a patchset(https://lkml.org/lkml/2014/9/21/148) for VM to be harmonized with zram-swap to stop anonymous page reclaim if zram consumed memory up to the limit although there are free space on the swap. One problem for that direction is zram has no way to know any hole in memory space zsmalloc allocated by internal fragmentation so zram would regard swap is full although there are free space in zsmalloc. For solving the issue, zram want to trigger compaction of zsmalloc before it decides full or not. This patchset is first step to support above issues. For that, it adds indirect layer between handle and object location and supports manual compaction to solve 3th problem first of all. After this patchset got merged, next step is to make VM aware of zsmalloc compaction so that generic compaction will move zsmalloced-pages automatically in runtime. In my imaginary experiment(ie, high compress ratio data with heavy swap in/out on 8G zram-swap), data is as follows, Before = zram allocated object : 60212066 bytes zram total used: 140103680 bytes ratio: 42.98 percent MemFree: 840192 kB Compaction After = frag ratio after compaction zram allocated object : 60212066 bytes zram total used: 76185600 bytes ratio: 79.03 percent MemFree: 901932 kB Juneho reported below in his real platform with small aging. So, I think the benefit would be bigger in real aging system for a long time. - frag_ratio increased 3% (ie, higher is better) - memfree increased about 6MB - In buddy info, Normal 2^3: 4, 2^2: 1: 2^1 increased, Highmem: 2^1 21 increased frag ratio after swap fragment used : 156677 kbytes total: 166092 kbytes frag_ratio : 94 meminfo before compaction MemFree: 83724 kB Node 0, zone Normal 13642 1364 57 10 61 17 9 5 4 0 0 Node 0, zone HighMem 425 29 1 0 0 0 0 0 0 0 0 num_migrated : 23630 compaction done frag ratio after compaction used : 156673 kbytes total: 160564 kbytes frag_ratio : 97 meminfo after compaction MemFree: 89060 kB Node 0, zone Normal 14076 1544 67 14 61 17 9 5 4 0 0 Node 0, zone HighMem 863 50 1 0 0 0 0 0 0 0 0 This patchset adds more logics(about 480 lines) in zsmalloc but when I tested heavy swapin/out program, the regression for swapin/out speed is marginal because most of overheads were caused by compress/decompress and other MM reclaim stuff. This patch (of 7): Currently, handle of zsmalloc encodes object's location directly so it makes support of migration hard. This patch decouples handle and object via adding indirect layer. For that, it allocates handle dynamically and returns it to user. The handle is the address allocated by slab allocation so it's unique and we could keep object's location in the memory space allocated for handle. With it, we can change object's position without changing handle itself. Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:23 +00:00
pool->name = kstrdup(name, GFP_KERNEL);
if (!pool->name)
goto err;
if (create_cache(pool))
zsmalloc: decouple handle and object Recently, we started to use zram heavily and some of issues popped. 1) external fragmentation I got a report from Juneho Choi that fork failed although there are plenty of free pages in the system. His investigation revealed zram is one of the culprit to make heavy fragmentation so there was no more contiguous 16K page for pgd to fork in the ARM. 2) non-movable pages Other problem of zram now is that inherently, user want to use zram as swap in small memory system so they use zRAM with CMA to use memory efficiently. However, unfortunately, it doesn't work well because zRAM cannot use CMA's movable pages unless it doesn't support compaction. I got several reports about that OOM happened with zram although there are lots of swap space and free space in CMA area. 3) internal fragmentation zRAM has started support memory limitation feature to limit memory usage and I sent a patchset(https://lkml.org/lkml/2014/9/21/148) for VM to be harmonized with zram-swap to stop anonymous page reclaim if zram consumed memory up to the limit although there are free space on the swap. One problem for that direction is zram has no way to know any hole in memory space zsmalloc allocated by internal fragmentation so zram would regard swap is full although there are free space in zsmalloc. For solving the issue, zram want to trigger compaction of zsmalloc before it decides full or not. This patchset is first step to support above issues. For that, it adds indirect layer between handle and object location and supports manual compaction to solve 3th problem first of all. After this patchset got merged, next step is to make VM aware of zsmalloc compaction so that generic compaction will move zsmalloced-pages automatically in runtime. In my imaginary experiment(ie, high compress ratio data with heavy swap in/out on 8G zram-swap), data is as follows, Before = zram allocated object : 60212066 bytes zram total used: 140103680 bytes ratio: 42.98 percent MemFree: 840192 kB Compaction After = frag ratio after compaction zram allocated object : 60212066 bytes zram total used: 76185600 bytes ratio: 79.03 percent MemFree: 901932 kB Juneho reported below in his real platform with small aging. So, I think the benefit would be bigger in real aging system for a long time. - frag_ratio increased 3% (ie, higher is better) - memfree increased about 6MB - In buddy info, Normal 2^3: 4, 2^2: 1: 2^1 increased, Highmem: 2^1 21 increased frag ratio after swap fragment used : 156677 kbytes total: 166092 kbytes frag_ratio : 94 meminfo before compaction MemFree: 83724 kB Node 0, zone Normal 13642 1364 57 10 61 17 9 5 4 0 0 Node 0, zone HighMem 425 29 1 0 0 0 0 0 0 0 0 num_migrated : 23630 compaction done frag ratio after compaction used : 156673 kbytes total: 160564 kbytes frag_ratio : 97 meminfo after compaction MemFree: 89060 kB Node 0, zone Normal 14076 1544 67 14 61 17 9 5 4 0 0 Node 0, zone HighMem 863 50 1 0 0 0 0 0 0 0 0 This patchset adds more logics(about 480 lines) in zsmalloc but when I tested heavy swapin/out program, the regression for swapin/out speed is marginal because most of overheads were caused by compress/decompress and other MM reclaim stuff. This patch (of 7): Currently, handle of zsmalloc encodes object's location directly so it makes support of migration hard. This patch decouples handle and object via adding indirect layer. For that, it allocates handle dynamically and returns it to user. The handle is the address allocated by slab allocation so it's unique and we could keep object's location in the memory space allocated for handle. With it, we can change object's position without changing handle itself. Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Juneho Choi <juno.choi@lge.com> Cc: Gunho Lee <gunho.lee@lge.com> Cc: Luigi Semenzato <semenzato@google.com> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Jerome Marchand <jmarchan@redhat.com> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-04-15 23:15:23 +00:00
goto err;
/*
* Iterate reversely, because, size of size_class that we want to use
* for merging should be larger or equal to current size.
*/
for (i = ZS_SIZE_CLASSES - 1; i >= 0; i--) {
int size;
int pages_per_zspage;
int objs_per_zspage;
struct size_class *class;
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
int fullness;
size = ZS_MIN_ALLOC_SIZE + i * ZS_SIZE_CLASS_DELTA;
if (size > ZS_MAX_ALLOC_SIZE)
size = ZS_MAX_ALLOC_SIZE;
pages_per_zspage = calculate_zspage_chain_size(size);
objs_per_zspage = pages_per_zspage * PAGE_SIZE / size;
zsmalloc: introduce zs_huge_class_size() Patch series "zsmalloc/zram: drop zram's max_zpage_size", v3. ZRAM's max_zpage_size is a bad thing. It forces zsmalloc to store normal objects as huge ones, which results in bigger zsmalloc memory usage. Drop it and use actual zsmalloc huge-class value when decide if the object is huge or not. This patch (of 2): Not every object can be share its zspage with other objects, e.g. when the object is as big as zspage or nearly as big a zspage. For such objects zsmalloc has a so called huge class - every object which belongs to huge class consumes the entire zspage (which consists of a physical page). On x86_64, PAGE_SHIFT 12 box, the first non-huge class size is 3264, so starting down from size 3264, objects can share page(-s) and thus minimize memory wastage. ZRAM, however, has its own statically defined watermark for huge objects, namely "3 * PAGE_SIZE / 4 = 3072", and forcibly stores every object larger than this watermark (3072) as a PAGE_SIZE object, in other words, to a huge class, while zsmalloc can keep some of those objects in non-huge classes. This results in increased memory consumption. zsmalloc knows better if the object is huge or not. Introduce zs_huge_class_size() function which tells if the given object can be stored in one of non-huge classes or not. This will let us to drop ZRAM's huge object watermark and fully rely on zsmalloc when we decide if the object is huge. [sergey.senozhatsky.work@gmail.com: add pool param to zs_huge_class_size()] Link: http://lkml.kernel.org/r/20180314081833.1096-2-sergey.senozhatsky@gmail.com Link: http://lkml.kernel.org/r/20180306070639.7389-2-sergey.senozhatsky@gmail.com Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Mike Rapoport <rppt@linux.vnet.ibm.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-04-05 23:24:43 +00:00
/*
* We iterate from biggest down to smallest classes,
* so huge_class_size holds the size of the first huge
* class. Any object bigger than or equal to that will
* endup in the huge class.
*/
if (pages_per_zspage != 1 && objs_per_zspage != 1 &&
!huge_class_size) {
huge_class_size = size;
/*
* The object uses ZS_HANDLE_SIZE bytes to store the
* handle. We need to subtract it, because zs_malloc()
* unconditionally adds handle size before it performs
* size class search - so object may be smaller than
* huge class size, yet it still can end up in the huge
* class because it grows by ZS_HANDLE_SIZE extra bytes
* right before class lookup.
*/
huge_class_size -= (ZS_HANDLE_SIZE - 1);
}
/*
* size_class is used for normal zsmalloc operation such
* as alloc/free for that size. Although it is natural that we
* have one size_class for each size, there is a chance that we
* can get more memory utilization if we use one size_class for
* many different sizes whose size_class have same
* characteristics. So, we makes size_class point to
* previous size_class if possible.
*/
if (prev_class) {
if (can_merge(prev_class, pages_per_zspage, objs_per_zspage)) {
pool->size_class[i] = prev_class;
continue;
}
}
class = kzalloc(sizeof(struct size_class), GFP_KERNEL);
if (!class)
goto err;
class->size = size;
class->index = i;
class->pages_per_zspage = pages_per_zspage;
class->objs_per_zspage = objs_per_zspage;
pool->size_class[i] = class;
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
fullness = ZS_INUSE_RATIO_0;
while (fullness < NR_FULLNESS_GROUPS) {
INIT_LIST_HEAD(&class->fullness_list[fullness]);
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
fullness++;
}
prev_class = class;
}
/* debug only, don't abort if it fails */
zs_pool_stat_create(pool, name);
mm/zsmalloc: add statistics support Keeping fragmentation of zsmalloc in a low level is our target. But now we still need to add the debug code in zsmalloc to get the quantitative data. This patch adds a new configuration CONFIG_ZSMALLOC_STAT to enable the statistics collection for developers. Currently only the objects statatitics in each class are collected. User can get the information via debugfs. cat /sys/kernel/debug/zsmalloc/zram0/... For example: After I copied "jdk-8u25-linux-x64.tar.gz" to zram with ext4 filesystem: class size obj_allocated obj_used pages_used 0 32 0 0 0 1 48 256 12 3 2 64 64 14 1 3 80 51 7 1 4 96 128 5 3 5 112 73 5 2 6 128 32 4 1 7 144 0 0 0 8 160 0 0 0 9 176 0 0 0 10 192 0 0 0 11 208 0 0 0 12 224 0 0 0 13 240 0 0 0 14 256 16 1 1 15 272 15 9 1 16 288 0 0 0 17 304 0 0 0 18 320 0 0 0 19 336 0 0 0 20 352 0 0 0 21 368 0 0 0 22 384 0 0 0 23 400 0 0 0 24 416 0 0 0 25 432 0 0 0 26 448 0 0 0 27 464 0 0 0 28 480 0 0 0 29 496 33 1 4 30 512 0 0 0 31 528 0 0 0 32 544 0 0 0 33 560 0 0 0 34 576 0 0 0 35 592 0 0 0 36 608 0 0 0 37 624 0 0 0 38 640 0 0 0 40 672 0 0 0 42 704 0 0 0 43 720 17 1 3 44 736 0 0 0 46 768 0 0 0 49 816 0 0 0 51 848 0 0 0 52 864 14 1 3 54 896 0 0 0 57 944 13 1 3 58 960 0 0 0 62 1024 4 1 1 66 1088 15 2 4 67 1104 0 0 0 71 1168 0 0 0 74 1216 0 0 0 76 1248 0 0 0 83 1360 3 1 1 91 1488 11 1 4 94 1536 0 0 0 100 1632 5 1 2 107 1744 0 0 0 111 1808 9 1 4 126 2048 4 4 2 144 2336 7 3 4 151 2448 0 0 0 168 2720 15 15 10 190 3072 28 27 21 202 3264 0 0 0 254 4096 36209 36209 36209 Total 37022 36326 36288 We can calculate the overall fragentation by the last line: Total 37022 36326 36288 (37022 - 36326) / 37022 = 1.87% Also by analysing objects alocated in every class we know why we got so low fragmentation: Most of the allocated objects is in <class 254>. And there is only 1 page in class 254 zspage. So, No fragmentation will be introduced by allocating objs in class 254. And in future, we can collect other zsmalloc statistics as we need and analyse them. Signed-off-by: Ganesh Mahendran <opensource.ganesh@gmail.com> Suggested-by: Minchan Kim <minchan@kernel.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Dan Streetman <ddstreet@ieee.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-12 23:00:54 +00:00
/*
* Not critical since shrinker is only used to trigger internal
* defragmentation of the pool which is pretty optional thing. If
* registration fails we still can use the pool normally and user can
* trigger compaction manually. Thus, ignore return code.
*/
zs_register_shrinker(pool);
#ifdef CONFIG_ZPOOL
INIT_LIST_HEAD(&pool->lru);
#endif
return pool;
err:
zs_destroy_pool(pool);
return NULL;
}
EXPORT_SYMBOL_GPL(zs_create_pool);
void zs_destroy_pool(struct zs_pool *pool)
{
int i;
zs_unregister_shrinker(pool);
zs_flush_migration(pool);
mm/zsmalloc: add statistics support Keeping fragmentation of zsmalloc in a low level is our target. But now we still need to add the debug code in zsmalloc to get the quantitative data. This patch adds a new configuration CONFIG_ZSMALLOC_STAT to enable the statistics collection for developers. Currently only the objects statatitics in each class are collected. User can get the information via debugfs. cat /sys/kernel/debug/zsmalloc/zram0/... For example: After I copied "jdk-8u25-linux-x64.tar.gz" to zram with ext4 filesystem: class size obj_allocated obj_used pages_used 0 32 0 0 0 1 48 256 12 3 2 64 64 14 1 3 80 51 7 1 4 96 128 5 3 5 112 73 5 2 6 128 32 4 1 7 144 0 0 0 8 160 0 0 0 9 176 0 0 0 10 192 0 0 0 11 208 0 0 0 12 224 0 0 0 13 240 0 0 0 14 256 16 1 1 15 272 15 9 1 16 288 0 0 0 17 304 0 0 0 18 320 0 0 0 19 336 0 0 0 20 352 0 0 0 21 368 0 0 0 22 384 0 0 0 23 400 0 0 0 24 416 0 0 0 25 432 0 0 0 26 448 0 0 0 27 464 0 0 0 28 480 0 0 0 29 496 33 1 4 30 512 0 0 0 31 528 0 0 0 32 544 0 0 0 33 560 0 0 0 34 576 0 0 0 35 592 0 0 0 36 608 0 0 0 37 624 0 0 0 38 640 0 0 0 40 672 0 0 0 42 704 0 0 0 43 720 17 1 3 44 736 0 0 0 46 768 0 0 0 49 816 0 0 0 51 848 0 0 0 52 864 14 1 3 54 896 0 0 0 57 944 13 1 3 58 960 0 0 0 62 1024 4 1 1 66 1088 15 2 4 67 1104 0 0 0 71 1168 0 0 0 74 1216 0 0 0 76 1248 0 0 0 83 1360 3 1 1 91 1488 11 1 4 94 1536 0 0 0 100 1632 5 1 2 107 1744 0 0 0 111 1808 9 1 4 126 2048 4 4 2 144 2336 7 3 4 151 2448 0 0 0 168 2720 15 15 10 190 3072 28 27 21 202 3264 0 0 0 254 4096 36209 36209 36209 Total 37022 36326 36288 We can calculate the overall fragentation by the last line: Total 37022 36326 36288 (37022 - 36326) / 37022 = 1.87% Also by analysing objects alocated in every class we know why we got so low fragmentation: Most of the allocated objects is in <class 254>. And there is only 1 page in class 254 zspage. So, No fragmentation will be introduced by allocating objs in class 254. And in future, we can collect other zsmalloc statistics as we need and analyse them. Signed-off-by: Ganesh Mahendran <opensource.ganesh@gmail.com> Suggested-by: Minchan Kim <minchan@kernel.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Dan Streetman <ddstreet@ieee.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-12 23:00:54 +00:00
zs_pool_stat_destroy(pool);
for (i = 0; i < ZS_SIZE_CLASSES; i++) {
int fg;
struct size_class *class = pool->size_class[i];
if (!class)
continue;
if (class->index != i)
continue;
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
for (fg = ZS_INUSE_RATIO_0; fg < NR_FULLNESS_GROUPS; fg++) {
if (list_empty(&class->fullness_list[fg]))
continue;
pr_err("Class-%d fullness group %d is not empty\n",
class->size, fg);
}
kfree(class);
}
destroy_cache(pool);
mm/zsmalloc: add statistics support Keeping fragmentation of zsmalloc in a low level is our target. But now we still need to add the debug code in zsmalloc to get the quantitative data. This patch adds a new configuration CONFIG_ZSMALLOC_STAT to enable the statistics collection for developers. Currently only the objects statatitics in each class are collected. User can get the information via debugfs. cat /sys/kernel/debug/zsmalloc/zram0/... For example: After I copied "jdk-8u25-linux-x64.tar.gz" to zram with ext4 filesystem: class size obj_allocated obj_used pages_used 0 32 0 0 0 1 48 256 12 3 2 64 64 14 1 3 80 51 7 1 4 96 128 5 3 5 112 73 5 2 6 128 32 4 1 7 144 0 0 0 8 160 0 0 0 9 176 0 0 0 10 192 0 0 0 11 208 0 0 0 12 224 0 0 0 13 240 0 0 0 14 256 16 1 1 15 272 15 9 1 16 288 0 0 0 17 304 0 0 0 18 320 0 0 0 19 336 0 0 0 20 352 0 0 0 21 368 0 0 0 22 384 0 0 0 23 400 0 0 0 24 416 0 0 0 25 432 0 0 0 26 448 0 0 0 27 464 0 0 0 28 480 0 0 0 29 496 33 1 4 30 512 0 0 0 31 528 0 0 0 32 544 0 0 0 33 560 0 0 0 34 576 0 0 0 35 592 0 0 0 36 608 0 0 0 37 624 0 0 0 38 640 0 0 0 40 672 0 0 0 42 704 0 0 0 43 720 17 1 3 44 736 0 0 0 46 768 0 0 0 49 816 0 0 0 51 848 0 0 0 52 864 14 1 3 54 896 0 0 0 57 944 13 1 3 58 960 0 0 0 62 1024 4 1 1 66 1088 15 2 4 67 1104 0 0 0 71 1168 0 0 0 74 1216 0 0 0 76 1248 0 0 0 83 1360 3 1 1 91 1488 11 1 4 94 1536 0 0 0 100 1632 5 1 2 107 1744 0 0 0 111 1808 9 1 4 126 2048 4 4 2 144 2336 7 3 4 151 2448 0 0 0 168 2720 15 15 10 190 3072 28 27 21 202 3264 0 0 0 254 4096 36209 36209 36209 Total 37022 36326 36288 We can calculate the overall fragentation by the last line: Total 37022 36326 36288 (37022 - 36326) / 37022 = 1.87% Also by analysing objects alocated in every class we know why we got so low fragmentation: Most of the allocated objects is in <class 254>. And there is only 1 page in class 254 zspage. So, No fragmentation will be introduced by allocating objs in class 254. And in future, we can collect other zsmalloc statistics as we need and analyse them. Signed-off-by: Ganesh Mahendran <opensource.ganesh@gmail.com> Suggested-by: Minchan Kim <minchan@kernel.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Dan Streetman <ddstreet@ieee.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-12 23:00:54 +00:00
kfree(pool->name);
kfree(pool);
}
EXPORT_SYMBOL_GPL(zs_destroy_pool);
#ifdef CONFIG_ZPOOL
zsmalloc: fix a race with deferred_handles storing Currently, there is a race between zs_free() and zs_reclaim_page(): zs_reclaim_page() finds a handle to an allocated object, but before the eviction happens, an independent zs_free() call to the same handle could come in and overwrite the object value stored at the handle with the last deferred handle. When zs_reclaim_page() finally gets to call the eviction handler, it will see an invalid object value (i.e the previous deferred handle instead of the original object value). This race happens quite infrequently. We only managed to produce it with out-of-tree developmental code that triggers zsmalloc writeback with a much higher frequency than usual. This patch fixes this race by storing the deferred handle in the object header instead. We differentiate the deferred handle from the other two cases (handle for allocated object, and linkage for free object) with a new tag. If zspage reclamation succeeds, we will free these deferred handles by walking through the zspage objects. On the other hand, if zspage reclamation fails, we reconstruct the zspage freelist (with the deferred handle tag and allocated tag) before trying again with the reclamation. [arnd@arndb.de: avoid unused-function warning] Link: https://lkml.kernel.org/r/20230117170507.2651972-1-arnd@kernel.org Link: https://lkml.kernel.org/r/20230110231701.326724-1-nphamcs@gmail.com Fixes: 9997bc017549 ("zsmalloc: implement writeback mechanism for zsmalloc") Signed-off-by: Nhat Pham <nphamcs@gmail.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-01-10 23:17:01 +00:00
static void restore_freelist(struct zs_pool *pool, struct size_class *class,
struct zspage *zspage)
{
unsigned int obj_idx = 0;
unsigned long handle, off = 0; /* off is within-page offset */
struct page *page = get_first_page(zspage);
struct link_free *prev_free = NULL;
void *prev_page_vaddr = NULL;
/* in case no free object found */
set_freeobj(zspage, (unsigned int)(-1UL));
while (page) {
void *vaddr = kmap_atomic(page);
struct page *next_page;
while (off < PAGE_SIZE) {
void *obj_addr = vaddr + off;
/* skip allocated object */
if (obj_allocated(page, obj_addr, &handle)) {
obj_idx++;
off += class->size;
continue;
}
/* free deferred handle from reclaim attempt */
if (obj_stores_deferred_handle(page, obj_addr, &handle))
cache_free_handle(pool, handle);
if (prev_free)
prev_free->next = obj_idx << OBJ_TAG_BITS;
else /* first free object found */
set_freeobj(zspage, obj_idx);
prev_free = (struct link_free *)vaddr + off / sizeof(*prev_free);
/* if last free object in a previous page, need to unmap */
if (prev_page_vaddr) {
kunmap_atomic(prev_page_vaddr);
prev_page_vaddr = NULL;
}
obj_idx++;
off += class->size;
}
/*
* Handle the last (full or partial) object on this page.
*/
next_page = get_next_page(page);
if (next_page) {
if (!prev_free || prev_page_vaddr) {
/*
* There is no free object in this page, so we can safely
* unmap it.
*/
kunmap_atomic(vaddr);
} else {
/* update prev_page_vaddr since prev_free is on this page */
prev_page_vaddr = vaddr;
}
} else { /* this is the last page */
if (prev_free) {
/*
* Reset OBJ_TAG_BITS bit to last link to tell
* whether it's allocated object or not.
*/
prev_free->next = -1UL << OBJ_TAG_BITS;
}
/* unmap previous page (if not done yet) */
if (prev_page_vaddr) {
kunmap_atomic(prev_page_vaddr);
prev_page_vaddr = NULL;
}
kunmap_atomic(vaddr);
}
page = next_page;
off %= PAGE_SIZE;
}
}
static int zs_reclaim_page(struct zs_pool *pool, unsigned int retries)
{
int i, obj_idx, ret = 0;
unsigned long handle;
struct zspage *zspage;
struct page *page;
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
int fullness;
/* Lock LRU and fullness list */
spin_lock(&pool->lock);
if (list_empty(&pool->lru)) {
spin_unlock(&pool->lock);
return -EINVAL;
}
for (i = 0; i < retries; i++) {
struct size_class *class;
zspage = list_last_entry(&pool->lru, struct zspage, lru);
list_del(&zspage->lru);
/* zs_free may free objects, but not the zspage and handles */
zspage->under_reclaim = true;
class = zspage_class(pool, zspage);
fullness = get_fullness_group(class, zspage);
/* Lock out object allocations and object compaction */
remove_zspage(class, zspage, fullness);
spin_unlock(&pool->lock);
cond_resched();
/* Lock backing pages into place */
lock_zspage(zspage);
obj_idx = 0;
page = get_first_page(zspage);
while (1) {
handle = find_alloced_obj(class, page, &obj_idx);
if (!handle) {
page = get_next_page(page);
if (!page)
break;
obj_idx = 0;
continue;
}
/*
* This will write the object and call zs_free.
*
* zs_free will free the object, but the
* under_reclaim flag prevents it from freeing
* the zspage altogether. This is necessary so
* that we can continue working with the
* zspage potentially after the last object
* has been freed.
*/
ret = pool->zpool_ops->evict(pool->zpool, handle);
if (ret)
goto next;
obj_idx++;
}
next:
/* For freeing the zspage, or putting it back in the pool and LRU list. */
spin_lock(&pool->lock);
zspage->under_reclaim = false;
if (!get_zspage_inuse(zspage)) {
/*
* Fullness went stale as zs_free() won't touch it
* while the page is removed from the pool. Fix it
* up for the check in __free_zspage().
*/
zsmalloc: fine-grained inuse ratio based fullness grouping Each zspage maintains ->inuse counter which keeps track of the number of objects stored in the zspage. The ->inuse counter also determines the zspage's "fullness group" which is calculated as the ratio of the "inuse" objects to the total number of objects the zspage can hold (objs_per_zspage). The closer the ->inuse counter is to objs_per_zspage, the better. Each size class maintains several fullness lists, that keep track of zspages of particular "fullness". Pages within each fullness list are stored in random order with regard to the ->inuse counter. This is because sorting the zspages by ->inuse counter each time obj_malloc() or obj_free() is called would be too expensive. However, the ->inuse counter is still a crucial factor in many situations. For the two major zsmalloc operations, zs_malloc() and zs_compact(), we typically select the head zspage from the corresponding fullness list as the best candidate zspage. However, this assumption is not always accurate. For the zs_malloc() operation, the optimal candidate zspage should have the highest ->inuse counter. This is because the goal is to maximize the number of ZS_FULL zspages and make full use of all allocated memory. For the zs_compact() operation, the optimal source zspage should have the lowest ->inuse counter. This is because compaction needs to move objects in use to another page before it can release the zspage and return its physical pages to the buddy allocator. The fewer objects in use, the quicker compaction can release the zspage. Additionally, compaction is measured by the number of pages it releases. This patch reworks the fullness grouping mechanism. Instead of having two groups - ZS_ALMOST_EMPTY (usage ratio below 3/4) and ZS_ALMOST_FULL (usage ration above 3/4) - that result in too many zspages being included in the ALMOST_EMPTY group for specific classes, size classes maintain a larger number of fullness lists that give strict guarantees on the minimum and maximum ->inuse values within each group. Each group represents a 10% change in the ->inuse ratio compared to neighboring groups. In essence, there are groups for zspages with 0%, 10%, 20% usage ratios, and so on, up to 100%. This enhances the selection of candidate zspages for both zs_malloc() and zs_compact(). A printout of the ->inuse counters of the first 7 zspages per (random) class fullness group: class-768 objs_per_zspage 16: fullness 100%: empty fullness 99%: empty fullness 90%: empty fullness 80%: empty fullness 70%: empty fullness 60%: 8 8 9 9 8 8 8 fullness 50%: empty fullness 40%: 5 5 6 5 5 5 5 fullness 30%: 4 4 4 4 4 4 4 fullness 20%: 2 3 2 3 3 2 2 fullness 10%: 1 1 1 1 1 1 1 fullness 0%: empty The zs_malloc() function searches through the groups of pages starting with the one having the highest usage ratio. This means that it always selects a zspage from the group with the least internal fragmentation (highest usage ratio) and makes it even less fragmented by increasing its usage ratio. The zs_compact() function, on the other hand, begins by scanning the group with the highest fragmentation (lowest usage ratio) to locate the source page. The first available zspage is selected, and then the function moves downward to find a destination zspage in the group with the lowest internal fragmentation (highest usage ratio). Link: https://lkml.kernel.org/r/20230304034835.2082479-3-senozhatsky@chromium.org Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Yosry Ahmed <yosryahmed@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-03-04 03:48:33 +00:00
zspage->fullness = ZS_INUSE_RATIO_0;
__free_zspage(pool, class, zspage);
spin_unlock(&pool->lock);
return 0;
}
zsmalloc: fix a race with deferred_handles storing Currently, there is a race between zs_free() and zs_reclaim_page(): zs_reclaim_page() finds a handle to an allocated object, but before the eviction happens, an independent zs_free() call to the same handle could come in and overwrite the object value stored at the handle with the last deferred handle. When zs_reclaim_page() finally gets to call the eviction handler, it will see an invalid object value (i.e the previous deferred handle instead of the original object value). This race happens quite infrequently. We only managed to produce it with out-of-tree developmental code that triggers zsmalloc writeback with a much higher frequency than usual. This patch fixes this race by storing the deferred handle in the object header instead. We differentiate the deferred handle from the other two cases (handle for allocated object, and linkage for free object) with a new tag. If zspage reclamation succeeds, we will free these deferred handles by walking through the zspage objects. On the other hand, if zspage reclamation fails, we reconstruct the zspage freelist (with the deferred handle tag and allocated tag) before trying again with the reclamation. [arnd@arndb.de: avoid unused-function warning] Link: https://lkml.kernel.org/r/20230117170507.2651972-1-arnd@kernel.org Link: https://lkml.kernel.org/r/20230110231701.326724-1-nphamcs@gmail.com Fixes: 9997bc017549 ("zsmalloc: implement writeback mechanism for zsmalloc") Signed-off-by: Nhat Pham <nphamcs@gmail.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de> Suggested-by: Johannes Weiner <hannes@cmpxchg.org> Cc: Dan Streetman <ddstreet@ieee.org> Cc: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Sergey Senozhatsky <senozhatsky@chromium.org> Cc: Seth Jennings <sjenning@redhat.com> Cc: Vitaly Wool <vitaly.wool@konsulko.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
2023-01-10 23:17:01 +00:00
/*
* Eviction fails on one of the handles, so we need to restore zspage.
* We need to rebuild its freelist (and free stored deferred handles),
* put it back to the correct size class, and add it to the LRU list.
*/
restore_freelist(pool, class, zspage);
putback_zspage(class, zspage);
list_add(&zspage->lru, &pool->lru);
unlock_zspage(zspage);
}
spin_unlock(&pool->lock);
return -EAGAIN;
}
#endif /* CONFIG_ZPOOL */
static int __init zs_init(void)
{
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
int ret;
ret = cpuhp_setup_state(CPUHP_MM_ZS_PREPARE, "mm/zsmalloc:prepare",
zs_cpu_prepare, zs_cpu_dead);
mm/zsmalloc: add statistics support Keeping fragmentation of zsmalloc in a low level is our target. But now we still need to add the debug code in zsmalloc to get the quantitative data. This patch adds a new configuration CONFIG_ZSMALLOC_STAT to enable the statistics collection for developers. Currently only the objects statatitics in each class are collected. User can get the information via debugfs. cat /sys/kernel/debug/zsmalloc/zram0/... For example: After I copied "jdk-8u25-linux-x64.tar.gz" to zram with ext4 filesystem: class size obj_allocated obj_used pages_used 0 32 0 0 0 1 48 256 12 3 2 64 64 14 1 3 80 51 7 1 4 96 128 5 3 5 112 73 5 2 6 128 32 4 1 7 144 0 0 0 8 160 0 0 0 9 176 0 0 0 10 192 0 0 0 11 208 0 0 0 12 224 0 0 0 13 240 0 0 0 14 256 16 1 1 15 272 15 9 1 16 288 0 0 0 17 304 0 0 0 18 320 0 0 0 19 336 0 0 0 20 352 0 0 0 21 368 0 0 0 22 384 0 0 0 23 400 0 0 0 24 416 0 0 0 25 432 0 0 0 26 448 0 0 0 27 464 0 0 0 28 480 0 0 0 29 496 33 1 4 30 512 0 0 0 31 528 0 0 0 32 544 0 0 0 33 560 0 0 0 34 576 0 0 0 35 592 0 0 0 36 608 0 0 0 37 624 0 0 0 38 640 0 0 0 40 672 0 0 0 42 704 0 0 0 43 720 17 1 3 44 736 0 0 0 46 768 0 0 0 49 816 0 0 0 51 848 0 0 0 52 864 14 1 3 54 896 0 0 0 57 944 13 1 3 58 960 0 0 0 62 1024 4 1 1 66 1088 15 2 4 67 1104 0 0 0 71 1168 0 0 0 74 1216 0 0 0 76 1248 0 0 0 83 1360 3 1 1 91 1488 11 1 4 94 1536 0 0 0 100 1632 5 1 2 107 1744 0 0 0 111 1808 9 1 4 126 2048 4 4 2 144 2336 7 3 4 151 2448 0 0 0 168 2720 15 15 10 190 3072 28 27 21 202 3264 0 0 0 254 4096 36209 36209 36209 Total 37022 36326 36288 We can calculate the overall fragentation by the last line: Total 37022 36326 36288 (37022 - 36326) / 37022 = 1.87% Also by analysing objects alocated in every class we know why we got so low fragmentation: Most of the allocated objects is in <class 254>. And there is only 1 page in class 254 zspage. So, No fragmentation will be introduced by allocating objs in class 254. And in future, we can collect other zsmalloc statistics as we need and analyse them. Signed-off-by: Ganesh Mahendran <opensource.ganesh@gmail.com> Suggested-by: Minchan Kim <minchan@kernel.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Dan Streetman <ddstreet@ieee.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-12 23:00:54 +00:00
if (ret)
goto out;
#ifdef CONFIG_ZPOOL
zpool_register_driver(&zs_zpool_driver);
#endif
mm/zsmalloc: add statistics support Keeping fragmentation of zsmalloc in a low level is our target. But now we still need to add the debug code in zsmalloc to get the quantitative data. This patch adds a new configuration CONFIG_ZSMALLOC_STAT to enable the statistics collection for developers. Currently only the objects statatitics in each class are collected. User can get the information via debugfs. cat /sys/kernel/debug/zsmalloc/zram0/... For example: After I copied "jdk-8u25-linux-x64.tar.gz" to zram with ext4 filesystem: class size obj_allocated obj_used pages_used 0 32 0 0 0 1 48 256 12 3 2 64 64 14 1 3 80 51 7 1 4 96 128 5 3 5 112 73 5 2 6 128 32 4 1 7 144 0 0 0 8 160 0 0 0 9 176 0 0 0 10 192 0 0 0 11 208 0 0 0 12 224 0 0 0 13 240 0 0 0 14 256 16 1 1 15 272 15 9 1 16 288 0 0 0 17 304 0 0 0 18 320 0 0 0 19 336 0 0 0 20 352 0 0 0 21 368 0 0 0 22 384 0 0 0 23 400 0 0 0 24 416 0 0 0 25 432 0 0 0 26 448 0 0 0 27 464 0 0 0 28 480 0 0 0 29 496 33 1 4 30 512 0 0 0 31 528 0 0 0 32 544 0 0 0 33 560 0 0 0 34 576 0 0 0 35 592 0 0 0 36 608 0 0 0 37 624 0 0 0 38 640 0 0 0 40 672 0 0 0 42 704 0 0 0 43 720 17 1 3 44 736 0 0 0 46 768 0 0 0 49 816 0 0 0 51 848 0 0 0 52 864 14 1 3 54 896 0 0 0 57 944 13 1 3 58 960 0 0 0 62 1024 4 1 1 66 1088 15 2 4 67 1104 0 0 0 71 1168 0 0 0 74 1216 0 0 0 76 1248 0 0 0 83 1360 3 1 1 91 1488 11 1 4 94 1536 0 0 0 100 1632 5 1 2 107 1744 0 0 0 111 1808 9 1 4 126 2048 4 4 2 144 2336 7 3 4 151 2448 0 0 0 168 2720 15 15 10 190 3072 28 27 21 202 3264 0 0 0 254 4096 36209 36209 36209 Total 37022 36326 36288 We can calculate the overall fragentation by the last line: Total 37022 36326 36288 (37022 - 36326) / 37022 = 1.87% Also by analysing objects alocated in every class we know why we got so low fragmentation: Most of the allocated objects is in <class 254>. And there is only 1 page in class 254 zspage. So, No fragmentation will be introduced by allocating objs in class 254. And in future, we can collect other zsmalloc statistics as we need and analyse them. Signed-off-by: Ganesh Mahendran <opensource.ganesh@gmail.com> Suggested-by: Minchan Kim <minchan@kernel.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Dan Streetman <ddstreet@ieee.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-12 23:00:54 +00:00
zs_stat_init();
return 0;
mm/zsmalloc: add statistics support Keeping fragmentation of zsmalloc in a low level is our target. But now we still need to add the debug code in zsmalloc to get the quantitative data. This patch adds a new configuration CONFIG_ZSMALLOC_STAT to enable the statistics collection for developers. Currently only the objects statatitics in each class are collected. User can get the information via debugfs. cat /sys/kernel/debug/zsmalloc/zram0/... For example: After I copied "jdk-8u25-linux-x64.tar.gz" to zram with ext4 filesystem: class size obj_allocated obj_used pages_used 0 32 0 0 0 1 48 256 12 3 2 64 64 14 1 3 80 51 7 1 4 96 128 5 3 5 112 73 5 2 6 128 32 4 1 7 144 0 0 0 8 160 0 0 0 9 176 0 0 0 10 192 0 0 0 11 208 0 0 0 12 224 0 0 0 13 240 0 0 0 14 256 16 1 1 15 272 15 9 1 16 288 0 0 0 17 304 0 0 0 18 320 0 0 0 19 336 0 0 0 20 352 0 0 0 21 368 0 0 0 22 384 0 0 0 23 400 0 0 0 24 416 0 0 0 25 432 0 0 0 26 448 0 0 0 27 464 0 0 0 28 480 0 0 0 29 496 33 1 4 30 512 0 0 0 31 528 0 0 0 32 544 0 0 0 33 560 0 0 0 34 576 0 0 0 35 592 0 0 0 36 608 0 0 0 37 624 0 0 0 38 640 0 0 0 40 672 0 0 0 42 704 0 0 0 43 720 17 1 3 44 736 0 0 0 46 768 0 0 0 49 816 0 0 0 51 848 0 0 0 52 864 14 1 3 54 896 0 0 0 57 944 13 1 3 58 960 0 0 0 62 1024 4 1 1 66 1088 15 2 4 67 1104 0 0 0 71 1168 0 0 0 74 1216 0 0 0 76 1248 0 0 0 83 1360 3 1 1 91 1488 11 1 4 94 1536 0 0 0 100 1632 5 1 2 107 1744 0 0 0 111 1808 9 1 4 126 2048 4 4 2 144 2336 7 3 4 151 2448 0 0 0 168 2720 15 15 10 190 3072 28 27 21 202 3264 0 0 0 254 4096 36209 36209 36209 Total 37022 36326 36288 We can calculate the overall fragentation by the last line: Total 37022 36326 36288 (37022 - 36326) / 37022 = 1.87% Also by analysing objects alocated in every class we know why we got so low fragmentation: Most of the allocated objects is in <class 254>. And there is only 1 page in class 254 zspage. So, No fragmentation will be introduced by allocating objs in class 254. And in future, we can collect other zsmalloc statistics as we need and analyse them. Signed-off-by: Ganesh Mahendran <opensource.ganesh@gmail.com> Suggested-by: Minchan Kim <minchan@kernel.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Dan Streetman <ddstreet@ieee.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-12 23:00:54 +00:00
zsmalloc: page migration support This patch introduces run-time migration feature for zspage. For migration, VM uses page.lru field so it would be better to not use page.next field which is unified with page.lru for own purpose. For that, firstly, we can get first object offset of the page via runtime calculation instead of using page.index so we can use page.index as link for page chaining instead of page.next. In case of huge object, it stores handle to page.index instead of next link of page chaining because huge object doesn't need to next link for page chaining. So get_next_page need to identify huge object to return NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag. For migration, it supports three functions * zs_page_isolate It isolates a zspage which includes a subpage VM want to migrate from class so anyone cannot allocate new object from the zspage. We could try to isolate a zspage by the number of subpage so subsequent isolation trial of other subpage of the zpsage shouldn't fail. For that, we introduce zspage.isolated count. With that, zs_page_isolate can know whether zspage is already isolated or not for migration so if it is isolated for migration, subsequent isolation trial can be successful without trying further isolation. * zs_page_migrate First of all, it holds write-side zspage->lock to prevent migrate other subpage in zspage. Then, lock all objects in the page VM want to migrate. The reason we should lock all objects in the page is due to race between zs_map_object and zs_page_migrate. zs_map_object zs_page_migrate pin_tag(handle) obj = handle_to_obj(handle) obj_to_location(obj, &page, &obj_idx); write_lock(&zspage->lock) if (!trypin_tag(handle)) goto unpin_object zspage = get_zspage(page); read_lock(&zspage->lock); If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be stale by migration so it goes crash. If it locks all of objects successfully, it copies content from old page to new one, finally, create new zspage chain with new page. And if it's last isolated subpage in the zspage, put the zspage back to class. * zs_page_putback It returns isolated zspage to right fullness_group list if it fails to migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage freeing to workqueue. See below about async zspage freeing. This patch introduces asynchronous zspage free. The reason to need it is we need page_lock to clear PG_movable but unfortunately, zs_free path should be atomic so the apporach is try to grab page_lock. If it got page_lock of all of pages successfully, it can free zspage immediately. Otherwise, it queues free request and free zspage via workqueue in process context. If zs_free finds the zspage is isolated when it try to free zspage, it delays the freeing until zs_page_putback finds it so it will free free the zspage finally. In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First of all, it will use ZS_EMPTY list for delay freeing. And with adding ZS_FULL list, it makes to identify whether zspage is isolated or not via list_empty(&zspage->list) test. [minchan@kernel.org: zsmalloc: keep first object offset in struct page] Link: http://lkml.kernel.org/r/1465788015-23195-1-git-send-email-minchan@kernel.org [minchan@kernel.org: zsmalloc: zspage sanity check] Link: http://lkml.kernel.org/r/20160603010129.GC3304@bbox Link: http://lkml.kernel.org/r/1464736881-24886-12-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-07-26 22:23:31 +00:00
out:
mm/zsmalloc: add statistics support Keeping fragmentation of zsmalloc in a low level is our target. But now we still need to add the debug code in zsmalloc to get the quantitative data. This patch adds a new configuration CONFIG_ZSMALLOC_STAT to enable the statistics collection for developers. Currently only the objects statatitics in each class are collected. User can get the information via debugfs. cat /sys/kernel/debug/zsmalloc/zram0/... For example: After I copied "jdk-8u25-linux-x64.tar.gz" to zram with ext4 filesystem: class size obj_allocated obj_used pages_used 0 32 0 0 0 1 48 256 12 3 2 64 64 14 1 3 80 51 7 1 4 96 128 5 3 5 112 73 5 2 6 128 32 4 1 7 144 0 0 0 8 160 0 0 0 9 176 0 0 0 10 192 0 0 0 11 208 0 0 0 12 224 0 0 0 13 240 0 0 0 14 256 16 1 1 15 272 15 9 1 16 288 0 0 0 17 304 0 0 0 18 320 0 0 0 19 336 0 0 0 20 352 0 0 0 21 368 0 0 0 22 384 0 0 0 23 400 0 0 0 24 416 0 0 0 25 432 0 0 0 26 448 0 0 0 27 464 0 0 0 28 480 0 0 0 29 496 33 1 4 30 512 0 0 0 31 528 0 0 0 32 544 0 0 0 33 560 0 0 0 34 576 0 0 0 35 592 0 0 0 36 608 0 0 0 37 624 0 0 0 38 640 0 0 0 40 672 0 0 0 42 704 0 0 0 43 720 17 1 3 44 736 0 0 0 46 768 0 0 0 49 816 0 0 0 51 848 0 0 0 52 864 14 1 3 54 896 0 0 0 57 944 13 1 3 58 960 0 0 0 62 1024 4 1 1 66 1088 15 2 4 67 1104 0 0 0 71 1168 0 0 0 74 1216 0 0 0 76 1248 0 0 0 83 1360 3 1 1 91 1488 11 1 4 94 1536 0 0 0 100 1632 5 1 2 107 1744 0 0 0 111 1808 9 1 4 126 2048 4 4 2 144 2336 7 3 4 151 2448 0 0 0 168 2720 15 15 10 190 3072 28 27 21 202 3264 0 0 0 254 4096 36209 36209 36209 Total 37022 36326 36288 We can calculate the overall fragentation by the last line: Total 37022 36326 36288 (37022 - 36326) / 37022 = 1.87% Also by analysing objects alocated in every class we know why we got so low fragmentation: Most of the allocated objects is in <class 254>. And there is only 1 page in class 254 zspage. So, No fragmentation will be introduced by allocating objs in class 254. And in future, we can collect other zsmalloc statistics as we need and analyse them. Signed-off-by: Ganesh Mahendran <opensource.ganesh@gmail.com> Suggested-by: Minchan Kim <minchan@kernel.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Dan Streetman <ddstreet@ieee.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-12 23:00:54 +00:00
return ret;
}
static void __exit zs_exit(void)
{
#ifdef CONFIG_ZPOOL
zpool_unregister_driver(&zs_zpool_driver);
#endif
cpuhp_remove_state(CPUHP_MM_ZS_PREPARE);
mm/zsmalloc: add statistics support Keeping fragmentation of zsmalloc in a low level is our target. But now we still need to add the debug code in zsmalloc to get the quantitative data. This patch adds a new configuration CONFIG_ZSMALLOC_STAT to enable the statistics collection for developers. Currently only the objects statatitics in each class are collected. User can get the information via debugfs. cat /sys/kernel/debug/zsmalloc/zram0/... For example: After I copied "jdk-8u25-linux-x64.tar.gz" to zram with ext4 filesystem: class size obj_allocated obj_used pages_used 0 32 0 0 0 1 48 256 12 3 2 64 64 14 1 3 80 51 7 1 4 96 128 5 3 5 112 73 5 2 6 128 32 4 1 7 144 0 0 0 8 160 0 0 0 9 176 0 0 0 10 192 0 0 0 11 208 0 0 0 12 224 0 0 0 13 240 0 0 0 14 256 16 1 1 15 272 15 9 1 16 288 0 0 0 17 304 0 0 0 18 320 0 0 0 19 336 0 0 0 20 352 0 0 0 21 368 0 0 0 22 384 0 0 0 23 400 0 0 0 24 416 0 0 0 25 432 0 0 0 26 448 0 0 0 27 464 0 0 0 28 480 0 0 0 29 496 33 1 4 30 512 0 0 0 31 528 0 0 0 32 544 0 0 0 33 560 0 0 0 34 576 0 0 0 35 592 0 0 0 36 608 0 0 0 37 624 0 0 0 38 640 0 0 0 40 672 0 0 0 42 704 0 0 0 43 720 17 1 3 44 736 0 0 0 46 768 0 0 0 49 816 0 0 0 51 848 0 0 0 52 864 14 1 3 54 896 0 0 0 57 944 13 1 3 58 960 0 0 0 62 1024 4 1 1 66 1088 15 2 4 67 1104 0 0 0 71 1168 0 0 0 74 1216 0 0 0 76 1248 0 0 0 83 1360 3 1 1 91 1488 11 1 4 94 1536 0 0 0 100 1632 5 1 2 107 1744 0 0 0 111 1808 9 1 4 126 2048 4 4 2 144 2336 7 3 4 151 2448 0 0 0 168 2720 15 15 10 190 3072 28 27 21 202 3264 0 0 0 254 4096 36209 36209 36209 Total 37022 36326 36288 We can calculate the overall fragentation by the last line: Total 37022 36326 36288 (37022 - 36326) / 37022 = 1.87% Also by analysing objects alocated in every class we know why we got so low fragmentation: Most of the allocated objects is in <class 254>. And there is only 1 page in class 254 zspage. So, No fragmentation will be introduced by allocating objs in class 254. And in future, we can collect other zsmalloc statistics as we need and analyse them. Signed-off-by: Ganesh Mahendran <opensource.ganesh@gmail.com> Suggested-by: Minchan Kim <minchan@kernel.org> Acked-by: Minchan Kim <minchan@kernel.org> Cc: Nitin Gupta <ngupta@vflare.org> Cc: Seth Jennings <sjennings@variantweb.net> Cc: Dan Streetman <ddstreet@ieee.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2015-02-12 23:00:54 +00:00
zs_stat_exit();
}
module_init(zs_init);
module_exit(zs_exit);
MODULE_LICENSE("Dual BSD/GPL");
MODULE_AUTHOR("Nitin Gupta <ngupta@vflare.org>");