Do not try fetching a stack trace from the stack depot if the
stack_depot_disabled flag is enabled.
Link: https://lkml.kernel.org/r/c3bfa3b7ab00b2e48ab75a3fbb9c67555777cb08.1700502145.git.andreyknvl@google.com
Signed-off-by: Andrey Konovalov <andreyknvl@google.com>
Reviewed-by: Alexander Potapenko <glider@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Evgenii Stepanov <eugenis@google.com>
Cc: Marco Elver <elver@google.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Patch series "stackdepot: allow evicting stack traces", v4.
Currently, the stack depot grows indefinitely until it reaches its
capacity. Once that happens, the stack depot stops saving new stack
traces.
This creates a problem for using the stack depot for in-field testing and
in production.
For such uses, an ideal stack trace storage should:
1. Allow saving fresh stack traces on systems with a large uptime while
limiting the amount of memory used to store the traces;
2. Have a low performance impact.
Implementing #1 in the stack depot is impossible with the current
keep-forever approach. This series targets to address that. Issue #2 is
left to be addressed in a future series.
This series changes the stack depot implementation to allow evicting
unneeded stack traces from the stack depot. The users of the stack depot
can do that via new stack_depot_save_flags(STACK_DEPOT_FLAG_GET) and
stack_depot_put APIs.
Internal changes to the stack depot code include:
1. Storing stack traces in fixed-frame-sized slots (vs precisely-sized
slots in the current implementation); the slot size is controlled via
CONFIG_STACKDEPOT_MAX_FRAMES (default: 64 frames);
2. Keeping available slots in a freelist (vs keeping an offset to the next
free slot);
3. Using a read/write lock for synchronization (vs a lock-free approach
combined with a spinlock).
This series also integrates the eviction functionality into KASAN: the
tag-based modes evict stack traces when the corresponding entry leaves the
stack ring, and Generic KASAN evicts stack traces for objects once those
leave the quarantine.
With KASAN, despite wasting some space on rounding up the size of each
stack record, the total memory consumed by stack depot gets saturated due
to the eviction of irrelevant stack traces from the stack depot.
With the tag-based KASAN modes, the average total amount of memory used
for stack traces becomes ~0.5 MB (with the current default stack ring size
of 32k entries and the default CONFIG_STACKDEPOT_MAX_FRAMES of 64). With
Generic KASAN, the stack traces take up ~1 MB per 1 GB of RAM (as the
quarantine's size depends on the amount of RAM).
However, with KMSAN, the stack depot ends up using ~4x more memory per a
stack trace than before. Thus, for KMSAN, the stack depot capacity is
increased accordingly. KMSAN uses a lot of RAM for shadow memory anyway,
so the increased stack depot memory usage will not make a significant
difference.
Other users of the stack depot do not save stack traces as often as KASAN
and KMSAN. Thus, the increased memory usage is taken as an acceptable
trade-off. In the future, these other users can take advantage of the
eviction API to limit the memory waste.
There is no measurable boot time performance impact of these changes for
KASAN on x86-64. I haven't done any tests for arm64 modes (the stack
depot without performance optimizations is not suitable for intended use
of those anyway), but I expect a similar result. Obtaining and copying
stack trace frames when saving them into stack depot is what takes the
most time.
This series does not yet provide a way to configure the maximum size of
the stack depot externally (e.g. via a command-line parameter). This
will be added in a separate series, possibly together with the performance
improvement changes.
This patch (of 22):
Currently, if stack_depot_disable=off is passed to the kernel command-line
after stack_depot_disable=on, stack depot prints a message that it is
disabled, while it is actually enabled.
Fix this by moving printing the disabled message to
stack_depot_early_init. Place it before the
__stack_depot_early_init_requested check, so that the message is printed
even if early stack depot init has not been requested.
Also drop the stack_table = NULL assignment from disable_stack_depot, as
stack_table is NULL by default.
Link: https://lkml.kernel.org/r/cover.1700502145.git.andreyknvl@google.com
Link: https://lkml.kernel.org/r/73a25c5fff29f3357cd7a9330e85e09bc8da2cbe.1700502145.git.andreyknvl@google.com
Fixes: e1fdc40334 ("lib: stackdepot: add support to disable stack depot")
Signed-off-by: Andrey Konovalov <andreyknvl@google.com>
Reviewed-by: Marco Elver <elver@google.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Evgenii Stepanov <eugenis@google.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Change /sys/kernel/debug/kmemleak report format slightly, adding
"(extra info)" to the backtrace header:
from: " backtrace:"
to: " backtrace (crc <cksum>):"
The <cksum> allows a user to see recurring backtraces without
detailed/careful reading of multiline stacks. So after cycling
kmemleak-test a few times, I know some leaks are repeating.
bash-5.2# grep backtrace /sys/kernel/debug/kmemleak | wc
62 186 1792
bash-5.2# grep backtrace /sys/kernel/debug/kmemleak | sort -u | wc
37 111 1067
syzkaller parses kmemleak for "unreferenced object" only, so is
unaffected by this change. Other github repos are moribund.
Link: https://lkml.kernel.org/r/20231116224318.124209-3-jim.cromie@gmail.com
Signed-off-by: Jim Cromie <jim.cromie@gmail.com>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Patch series "tweak kmemleak report format".
These 2 patches make minor changes to the report:
1st strips "age <increasing>" from output. This makes the output
idempotent; unchanging until a new leak is reported.
2nd adds the backtrace.checksum to the "backtrace:" line. This lets a
user see repeats without actually reading the whole backtrace. So now
the backtrace line looks like this:
backtrace (crc 603070071):
I surveyed for un-wanted effects upon users:
Syzkaller parses kmemleak in executor/common_linux.h:
static void check_leaks(char** frames, int nframes)
It just counts occurrences of "unreferenced object", specifically it
does not look for "age", nor would it choke on "crc" being added.
github has 3 repos with "kmemleak" mentioned, all are moribund.
gitlab has 0 hits on "kmemleak".
This patch (of 2):
Displaying age is pretty, but counter-productive; it changes with
current-time, so it surrenders idempotency of the output, which breaks
simple hash-based cataloging of the records by the user.
The trouble: sequential reads, wo new leaks, get new results:
:#> sum /sys/kernel/debug/kmemleak
53439 74 /sys/kernel/debug/kmemleak
:#> sum /sys/kernel/debug/kmemleak
59066 74 /sys/kernel/debug/kmemleak
and age is why (nothing else changes):
:#> grep -v age /sys/kernel/debug/kmemleak | sum
58894 67
:#> grep -v age /sys/kernel/debug/kmemleak | sum
58894 67
Since jiffies is already printed in the "comm" line, age adds nothing.
Notably, syzkaller reads kmemleak only for "unreferenced object", and
won't care about this reform of age-ism. A few moribund github repos
mention it, but don't compile.
Link: https://lkml.kernel.org/r/20231116224318.124209-1-jim.cromie@gmail.com
Link: https://lkml.kernel.org/r/20231116224318.124209-2-jim.cromie@gmail.com
Signed-off-by: Jim Cromie <jim.cromie@gmail.com>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
There were already assertions that we were not passing a tail page to
error_remove_page(), so make the compiler enforce that by converting
everything to pass and use a folio.
Link: https://lkml.kernel.org/r/20231117161447.2461643-7-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Both callers now have a folio, so pass it in. Nothing downstream was
expecting a tail page; that's asserted in generic_error_remove_page(), for
example.
Link: https://lkml.kernel.org/r/20231117161447.2461643-6-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
This function was already explicitly calling compound_head();
unfortunately the compiler can't know that and elide the redundant calls
to compound_head() buried in page_mapping(), unlock_page(), etc. Switch
to using a folio, which does let us elide these calls.
Link: https://lkml.kernel.org/r/20231117161447.2461643-5-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
All three callers now have a folio; pass it in instead of the page.
Saves five calls to compound_head().
Link: https://lkml.kernel.org/r/20231117161447.2461643-4-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Replaces three hidden calls to compound_head() with one visible one.
Link: https://lkml.kernel.org/r/20231117161447.2461643-3-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Patch series "Convert aops->error_remove_page to ->error_remove_folio".
This is a memory-failure patch series which converts a lot of uses of page
APIs into folio APIs with the usual benefits.
This patch (of 6):
Replaces three hidden calls to compound_head() with one visible one.
Fix up a few comments while I'm modifying this function.
Link: https://lkml.kernel.org/r/20231117161447.2461643-1-willy@infradead.org
Link: https://lkml.kernel.org/r/20231117161447.2461643-2-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
ZRAM_MEMORY_TRACKING enables two features:
- per-entry ac-time tracking
- debugfs interface
The latter one is the reason why memory-tracking depends on DEBUG_FS,
while the former one is used far beyond debugging these days. Namely
ac-time is used for fine grained writeback of idle entries (pages).
Move ac-time tracking under its own config option so that it can be
enabled (along with writeback) on systems without DEBUG_FS.
[senozhatsky@chromium.org: ifdef fixup, per Dmytro]
Link: https://lkml.kernel.org/r/20231117013543.540280-1-senozhatsky@chromium.org
Link: https://lkml.kernel.org/r/20231115024223.4133148-1-senozhatsky@chromium.org
Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Dmytro Maluka <dmaluka@chromium.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
While investigating some complex memory allocation and free bugs
especially in multi-processes and multi-threads cases, from time to time,
I feel the free stack isn't sufficient as a page can be freed by processes
or threads other than the one allocating it. And other processes and
threads which free the page often have the exactly same free stack with
the one allocating the page. We can't know who free the page only through
the free stack though the current page_owner does tell us the pid and tgid
of the one allocating the page. This makes the bug investigation often
hard.
So this patch adds free pid and tgid in page_owner, so that we can easily
figure out if the freeing is crossing processes or threads.
Link: https://lkml.kernel.org/r/20231114034202.73098-1-v-songbaohua@oppo.com
Signed-off-by: Barry Song <v-songbaohua@oppo.com>
Cc: Audra Mitchell <audra@redhat.com>
Cc: Hyeonggon Yoo <42.hyeyoo@gmail.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Kassey Li <quic_yingangl@quicinc.com>
Cc: Kemeng Shi <shikemeng@huaweicloud.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
pte_bad() never existed unlike similar helpers at PMU, PUD, and PGD level.
This was added erroneously and hence should be dropped instead.
Link: https://lkml.kernel.org/r/20231114063456.339652-1-anshuman.khandual@arm.com
Signed-off-by: Anshuman Khandual <anshuman.khandual@arm.com>
Reviewed-by: Rick Edgecombe <rick.p.edgecombe@intel.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
KASan inline instrumentation can yield up to a 2x performance gain at the
cost of a larger binary.
Make inline instrumentation the default, as suggested in the bug report
below.
When an architecture does not support inline instrumentation, it should
set ARCH_DISABLE_KASAN_INLINE, as done by PowerPC, for instance.
Link: https://lkml.kernel.org/r/20231109155101.186028-1-paul.heidekrueger@tum.de
Signed-off-by: Paul Heidekrüger <paul.heidekrueger@tum.de>
Reported-by: Andrey Konovalov <andreyknvl@gmail.com>
Reviewed-by: Marco Elver <elver@google.com>
Closes: https://bugzilla.kernel.org/show_bug.cgi?id=203495
Acked-by: Andrey Konovalov <andreyknvl@gmail.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Ryabinin <ryabinin.a.a@gmail.com>
Cc: Vincenzo Frascino <vincenzo.frascino@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
1. There is a "-1" missing in the page number calculation in
process_vm_rw_core. While this can't break anything, it can cause
unnecessary allocations in certain cases:
Consider handling an iovec ranging over PVM_MAX_PP_ARRAY_COUNT pages
that is also aligned to a page boundary. While pp_stack could hold
references to such an amount of pinned pages, nr_pages yields
(PVM_MAX_PP_ARRAY + 1) in process_vm_rw_core. Consequently, a larger
buffer is allocated with kmalloc for no reason.
For any page boundary aligned iovec that is a multiple of PAGE_SIZE
and larger than PVM_MAX_PP_ARRAY_COUNT pages, nr_pages will be too big
by 1 and thus kmalloc allocates excess space for one more pointer.
2. max_pages_per_loop is constant and there is no reason to have it as
a variable. A macro does the job just fine and saves memory.
3. Replaced "sizeof(struct pages *)" with "sizeof(struct page *)" to
have matching types for allocation and prevent confusion.
Link: https://lkml.kernel.org/r/20231111184859.44264-1-yjnworkstation@gmail.com
Signed-off-by: York Jasper Niebuhr <yjnworkstation@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
With commit cf8e865810 ("arch: Remove Itanium (IA-64) architecture"),
there is no need to keep the IA64-specific vma expansion.
Clean up the IA64-specific vma expansion implementation.
Link: https://lkml.kernel.org/r/20231113124728.3974-1-lukas.bulwahn@gmail.com
Signed-off-by: Lukas Bulwahn <lukas.bulwahn@gmail.com>
Acked-by: David Hildenbrand <david@redhat.com>
Cc: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
GFP_NOWAIT callers are always prepared for their allocations to fail
because they fail so frequently. Forcing the callers to remember to add
__GFP_NOWARN is just annoying and leads to an endless stream of patches
for the places where we forgot to add it.
We can now remove __GFP_NOWARN from all the callers which specify
GFP_NOWAIT, but I'd rather wait a cycle and send patches to each
maintainer instead of creating a big pile of merge conflicts.
Link: https://lkml.kernel.org/r/20231109211507.2262419-1-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
The duplication makes it seem like some work is required before uncharging
in the !PageHWPoison case. But it isn't, so we can simplify the code a
little.
Note the PageMemcgKmem check is redundant, but I've left it in as it
avoids an unnecessary function call.
Link: https://lkml.kernel.org/r/20231108164920.3401565-1-jackmanb@google.com
Signed-off-by: Brendan Jackman <jackmanb@google.com>
Reviewed-by: Yosry Ahmed <yosryahmed@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Muchun Song <muchun.song@linux.dev>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Shakeel Butt <shakeelb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
All callers are now converted to call mapping_evict_folio().
Link: https://lkml.kernel.org/r/20231108182809.602073-7-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
The only caller now has a folio, so pass it in and operate on it. Saves
many page->folio conversions and introduces only one folio->page
conversion when calling isolate_movable_page().
Link: https://lkml.kernel.org/r/20231108182809.602073-6-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
We already have the folio and the mapping, so replace the call to
invalidate_inode_page() with mapping_evict_folio().
Link: https://lkml.kernel.org/r/20231108182809.602073-4-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Convert vmf->page to a folio as soon as we're going to use it. This fixes
a bug if the fault handler returns a tail page with hardware poison; tail
pages have an invalid page->index, so we would fail to unmap the page from
the page tables. We actually have to unmap the entire folio (or
mapping_evict_folio() will fail), so use unmap_mapping_folio() instead.
This also saves various calls to compound_head() hidden in lock_page(),
put_page(), etc.
Link: https://lkml.kernel.org/r/20231108182809.602073-3-willy@infradead.org
Fixes: 793917d997 ("mm/readahead: Add large folio readahead")
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Patch series "Fix fault handler's handling of poisoned tail pages".
Since introducing the ability to have large folios in the page cache, it's
been possible to have a hwpoisoned tail page returned from the fault
handler. We handle this situation poorly; failing to remove the affected
page from use.
This isn't a minimal patch to fix it, it's a full conversion of all the
code surrounding it.
This patch (of 6):
invalidate_inode_page() does very little beyond calling
mapping_evict_folio(). Move the check for mapping being NULL into
mapping_evict_folio() and make it available to the rest of the MM for use
in the next few patches.
Link: https://lkml.kernel.org/r/20231108182809.602073-1-willy@infradead.org
Link: https://lkml.kernel.org/r/20231108182809.602073-2-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Nobody now checks the return value from any of these functions, so
add an assertion at the beginning of the function and return void.
Link: https://lkml.kernel.org/r/20231108204605.745109-5-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Steve French <sfrench@samba.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
In preparation for removing the return value entirely, stop testing it
in smb.
Link: https://lkml.kernel.org/r/20231108204605.745109-4-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: David Howells <dhowells@redhat.com>
Cc: Steve French <sfrench@samba.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
In preparation for removing the return value entirely, stop testing it
in afs.
Link: https://lkml.kernel.org/r/20231108204605.745109-3-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: David Howells <dhowells@redhat.com>
Cc: Steve French <sfrench@samba.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Patch series "Make folio_start_writeback return void".
Most of the folio flag-setting functions return void.
folio_start_writeback is gratuitously different; the only two filesystems
that do anything with the return value emit debug messages if it's already
set, and we can (and should) do that internally without bothering the
filesystem to do it.
This patch (of 4):
There are no more callers of this wrapper.
Link: https://lkml.kernel.org/r/20231108204605.745109-1-willy@infradead.org
Link: https://lkml.kernel.org/r/20231108204605.745109-2-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: David Howells <dhowells@redhat.com>
Cc: Steve French <sfrench@samba.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Use folio_fill_tail() to implement the unstuffing and folio_end_read() to
simultaneously mark the folio uptodate and unlock it. Unifies a couple of
code paths.
Link: https://lkml.kernel.org/r/20231107212643.3490372-4-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Andreas Gruenbacher <agruenba@redhat.com>
Cc: Andreas Dilger <adilger.kernel@dilger.ca>
Cc: Darrick J. Wong <djwong@kernel.org>
Cc: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
The iomap code was limited to PAGE_SIZE bytes; generalise it to cover
an arbitrary-sized folio, and move it to be a common helper.
[akpm@linux-foundation.org: fix folio_fill_tail(), per Andreas Gruenbacher]
Link: https://lkml.kernel.org/r/20231107212643.3490372-3-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Andreas Gruenbacher <agruenba@redhat.com>
Cc: Andreas Dilger <adilger.kernel@dilger.ca>
Cc: Darrick J. Wong <djwong@kernel.org>
Cc: Theodore Ts'o <tytso@mit.edu>
Cc: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Patch series "Add folio_zero_tail() and folio_fill_tail()".
I'm trying to make it easier for filesystems with tailpacking / stuffing /
inline data to use folios. The primary function here is
folio_fill_tail(). You give it a pointer to memory where the data
currently is, and it takes care of copying it into the folio at that
offset. That works for gfs2 & iomap. Then There's Ext4. Rather than gin
up some kind of specialist "Here's a two pointers to two blocks of memory"
routine, just let it do its current thing, and let it call
folio_zero_tail(), which is also called by folio_fill_tail().
Other filesystems can be converted later; these ones seemed like good
examples as they're already partly or completely converted to folios.
This patch (of 3):
Instead of unmapping the folio after copying the data to it, then mapping
it again to zero the tail, provide folio_zero_tail() to zero the tail of
an already-mapped folio.
[akpm@linux-foundation.org: fix kerneldoc argument ordering]
Link: https://lkml.kernel.org/r/20231107212643.3490372-1-willy@infradead.org
Link: https://lkml.kernel.org/r/20231107212643.3490372-2-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Andreas Gruenbacher <agruenba@redhat.com>
Cc: Darrick J. Wong <djwong@kernel.org>
Cc: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Right now, tests read page flags from /proc/pid/pagemap files. With this
change, tests will check that PAGEMAP_SCAN return correct information too.
[colin.i.king@gmail.com: fix spelling mistake "succedded" -> "succeeded"]
Link: https://lkml.kernel.org/r/20231121093104.1728332-1-colin.i.king@gmail.com
Link: https://lkml.kernel.org/r/20231106220959.296568-2-avagin@google.com
Signed-off-by: Andrei Vagin <avagin@google.com>
Signed-off-by: Colin Ian King <colin.i.king@gmail.com>
Reviewed-by: Muhammad Usama Anjum <usama.anjum@collabora.com>
Tested-by: Muhammad Usama Anjum <usama.anjum@collabora.com>
Cc: Michał Mirosław <mirq-linux@rere.qmqm.pl>
[avagin@google.com: allow running tests on old kernels]
Link: https://lkml.kernel.org/r/20231117181127.2574897-1-avagin@google.com
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
The PAGEMAP_SCAN ioctl returns information regarding page table entries.
It is more efficient compared to reading pagemap files. CRIU can start to
utilize this ioctl, but it needs info about soft-dirty bits to track
memory changes.
We are aware of a new method for tracking memory changes implemented in
the PAGEMAP_SCAN ioctl. For CRIU, the primary advantage of this method is
its usability by unprivileged users. However, it is not feasible to
transparently replace the soft-dirty tracker with the new one. The main
problem here is userfault descriptors that have to be preserved between
pre-dump iterations. It means criu continues supporting the soft-dirty
method to avoid breakage for current users. The new method will be
implemented as a separate feature.
[avagin@google.com: update tools/include/uapi/linux/fs.h]
Link: https://lkml.kernel.org/r/20231107164139.576046-1-avagin@google.com
Link: https://lkml.kernel.org/r/20231106220959.296568-1-avagin@google.com
Signed-off-by: Andrei Vagin <avagin@google.com>
Reviewed-by: Muhammad Usama Anjum <usama.anjum@collabora.com>
Cc: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Simplify code pattern of 'folio->index + folio_nr_pages(folio)' by using
the existing helper folio_next_index() in filemap_get_folios_contig().
Link: https://lkml.kernel.org/r/20231107024635.4512-1-duminjie@vivo.com
Signed-off-by: Minjie Du <duminjie@vivo.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Large amounts of memory managed by the kmem driver may come in via CXL,
and it is often desirable to have the memmap for this memory on the new
memory itself.
Enroll kmem-managed memory for memmap_on_memory semantics if the dax
region originates via CXL. For non-CXL dax regions, retain the existing
default behavior of hot adding without memmap_on_memory semantics.
Link: https://lkml.kernel.org/r/20231107-vv-kmem_memmap-v10-3-1253ec050ed0@intel.com
Signed-off-by: Vishal Verma <vishal.l.verma@intel.com>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: "Huang, Ying" <ying.huang@intel.com>
Tested-by: Li Zhijian <lizhijian@fujitsu.com> [cxl.kmem and nvdimm.kmem]
Cc: Michal Hocko <mhocko@suse.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Dave Jiang <dave.jiang@intel.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Cc: Fan Ni <fan.ni@samsung.com>
Cc: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
The MHP_MEMMAP_ON_MEMORY flag for hotplugged memory is restricted to
'memblock_size' chunks of memory being added. Adding a larger span of
memory precludes memmap_on_memory semantics.
For users of hotplug such as kmem, large amounts of memory might get added
from the CXL subsystem. In some cases, this amount may exceed the
available 'main memory' to store the memmap for the memory being added.
In this case, it is useful to have a way to place the memmap on the memory
being added, even if it means splitting the addition into memblock-sized
chunks.
Change add_memory_resource() to loop over memblock-sized chunks of memory
if caller requested memmap_on_memory, and if other conditions for it are
met. Teach try_remove_memory() to also expect that a memory range being
removed might have been split up into memblock sized chunks, and to loop
through those as needed.
This does preclude being able to use PUD mappings in the direct map; a
proposal to how this could be optimized in the future is laid out here[1].
[1]: https://lore.kernel.org/linux-mm/b6753402-2de9-25b2-36e9-eacd49752b19@redhat.com/
Link: https://lkml.kernel.org/r/20231107-vv-kmem_memmap-v10-2-1253ec050ed0@intel.com
Signed-off-by: Vishal Verma <vishal.l.verma@intel.com>
Suggested-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: "Huang, Ying" <ying.huang@intel.com>
Acked-by: David Hildenbrand <david@redhat.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Dave Jiang <dave.jiang@intel.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Cc: Fan Ni <fan.ni@samsung.com>
Cc: Jeff Moyer <jmoyer@redhat.com>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Patch series "mm: use memmap_on_memory semantics for dax/kmem", v10.
The dax/kmem driver can potentially hot-add large amounts of memory
originating from CXL memory expanders, or NVDIMMs, or other 'device
memories'. There is a chance there isn't enough regular system memory
available to fit the memmap for this new memory. It's therefore
desirable, if all other conditions are met, for the kmem managed memory to
place its memmap on the newly added memory itself.
The main hurdle for accomplishing this for kmem is that memmap_on_memory
can only be done if the memory being added is equal to the size of one
memblock. To overcome this, allow the hotplug code to split an
add_memory() request into memblock-sized chunks, and try_remove_memory()
to also expect and handle such a scenario.
Patch 1 replaces an open-coded kmemdup()
Patch 2 teaches the memory_hotplug code to allow for splitting
add_memory() and remove_memory() requests over memblock sized chunks.
Patch 3 allows the dax region drivers to request memmap_on_memory
semantics. CXL dax regions default this to 'on', all others default to
off to keep existing behavior unchanged.
This patch (of 3):
A review of the memmap_on_memory modifications to add_memory_resource()
revealed an instance of an open-coded kmemdup(). Replace it with
kmemdup().
Link: https://lkml.kernel.org/r/20231107-vv-kmem_memmap-v10-0-1253ec050ed0@intel.com
Link: https://lkml.kernel.org/r/20231107-vv-kmem_memmap-v10-1-1253ec050ed0@intel.com
Signed-off-by: Vishal Verma <vishal.l.verma@intel.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Fan Ni <fan.ni@samsung.com>
Reported-by: Dan Williams <dan.j.williams@intel.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Dave Jiang <dave.jiang@intel.com>
Cc: "Huang, Ying" <ying.huang@intel.com>
Cc: Jeff Moyer <jmoyer@redhat.com>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Sanity check that makes sure the nodes cover all memory loops over
numa_meminfo to count the pages that have node id assigned by the
firmware, then loops again over memblock.memory to find the total amount
of memory and in the end checks that the difference between the total
memory and memory that covered by nodes is less than some threshold.
Worse, the loop over numa_meminfo calls __absent_pages_in_range() that
also partially traverses memblock.memory.
It's much simpler and more efficient to have a single traversal of
memblock.memory that verifies that amount of memory not covered by nodes
is less than a threshold.
Introduce memblock_validate_numa_coverage() that does exactly that and use
it instead of numa_meminfo_cover_memory().
Link: https://lkml.kernel.org/r/20231026020329.327329-1-zhiguangni01@gmail.com
Signed-off-by: Liam Ni <zhiguangni01@gmail.com>
Reviewed-by: Mike Rapoport (IBM) <rppt@kernel.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Bibo Mao <maobibo@loongson.cn>
Cc: Binbin Zhou <zhoubinbin@loongson.cn>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Feiyang Chen <chenfeiyang@loongson.cn>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Huacai Chen <chenhuacai@kernel.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: WANG Xuerui <kernel@xen0n.name>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
I can observe an obvious tlb flush hotspot when splitting a pte-mapped THP
on my ARM64 server, and the distribution of this hotspot is as follows:
- 16.85% split_huge_page_to_list
+ 7.80% down_write
- 7.49% try_to_migrate
- 7.48% rmap_walk_anon
7.23% ptep_clear_flush
+ 1.52% __split_huge_page
The reason is that the split_huge_page_to_list() will build migration
entries for each subpage of a pte-mapped Anon THP by try_to_migrate(), or
unmap for file THP, and it will clear and tlb flush for each subpage's
pte. Moreover, the split_huge_page_to_list() will set TTU_SPLIT_HUGE_PMD
flag to ensure the THP is already a pte-mapped THP before splitting it to
some normal pages.
Actually, there is no need to flush tlb for each subpage immediately,
instead we can batch tlb flush for the pte-mapped THP to improve the
performance.
After this patch, we can see the batch tlb flush can improve the latency
obviously when running thpscale.
k6.5-base patched
Amean fault-both-1 1071.17 ( 0.00%) 901.83 * 15.81%*
Amean fault-both-3 2386.08 ( 0.00%) 1865.32 * 21.82%*
Amean fault-both-5 2851.10 ( 0.00%) 2273.84 * 20.25%*
Amean fault-both-7 3679.91 ( 0.00%) 2881.66 * 21.69%*
Amean fault-both-12 5916.66 ( 0.00%) 4369.55 * 26.15%*
Amean fault-both-18 7981.36 ( 0.00%) 6303.57 * 21.02%*
Amean fault-both-24 10950.79 ( 0.00%) 8752.56 * 20.07%*
Amean fault-both-30 14077.35 ( 0.00%) 10170.01 * 27.76%*
Amean fault-both-32 13061.57 ( 0.00%) 11630.08 * 10.96%*
Link: https://lkml.kernel.org/r/431d9fb6823036369dcb1d3b2f63732f01df21a7.1698488264.git.baolin.wang@linux.alibaba.com
Signed-off-by: Baolin Wang <baolin.wang@linux.alibaba.com>
Reviewed-by: "Huang, Ying" <ying.huang@intel.com>
Reviewed-by: Yang Shi <shy828301@gmail.com>
Reviewed-by: Alistair Popple <apopple@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
In dup_mmap(), using __mt_dup() to duplicate the old maple tree and then
directly replacing the entries of VMAs in the new maple tree can result in
better performance. __mt_dup() uses DFS pre-order to duplicate the maple
tree, so it is efficient.
The average time complexity of __mt_dup() is O(n), where n is the number
of VMAs. The proof of the time complexity is provided in the commit log
that introduces __mt_dup(). After duplicating the maple tree, each
element is traversed and replaced (ignoring the cases of deletion, which
are rare). Since it is only a replacement operation for each element,
this process is also O(n).
Analyzing the exact time complexity of the previous algorithm is
challenging because each insertion can involve appending to a node,
pushing data to adjacent nodes, or even splitting nodes. The frequency of
each action is difficult to calculate. The worst-case scenario for a
single insertion is when the tree undergoes splitting at every level. If
we consider each insertion as the worst-case scenario, we can determine
that the upper bound of the time complexity is O(n*log(n)), although this
is a loose upper bound. However, based on the test data, it appears that
the actual time complexity is likely to be O(n).
As the entire maple tree is duplicated using __mt_dup(), if dup_mmap()
fails, there will be a portion of VMAs that have not been duplicated in
the maple tree. To handle this, we mark the failure point with
XA_ZERO_ENTRY. In exit_mmap(), if this marker is encountered, stop
releasing VMAs that have not been duplicated after this point.
There is a "spawn" in byte-unixbench[1], which can be used to test the
performance of fork(). I modified it slightly to make it work with
different number of VMAs.
Below are the test results. The first row shows the number of VMAs. The
second and third rows show the number of fork() calls per ten seconds,
corresponding to next-20231006 and the this patchset, respectively. The
test results were obtained with CPU binding to avoid scheduler load
balancing that could cause unstable results. There are still some
fluctuations in the test results, but at least they are better than the
original performance.
21 121 221 421 821 1621 3221 6421 12821 25621 51221
112100 76261 54227 34035 20195 11112 6017 3161 1606 802 393
114558 83067 65008 45824 28751 16072 8922 4747 2436 1233 599
2.19% 8.92% 19.88% 34.64% 42.37% 44.64% 48.28% 50.17% 51.68% 53.74% 52.42%
[1] https://github.com/kdlucas/byte-unixbench/tree/master
Link: https://lkml.kernel.org/r/20231027033845.90608-11-zhangpeng.00@bytedance.com
Signed-off-by: Peng Zhang <zhangpeng.00@bytedance.com>
Suggested-by: Liam R. Howlett <Liam.Howlett@oracle.com>
Reviewed-by: Liam R. Howlett <Liam.Howlett@oracle.com>
Cc: Christian Brauner <brauner@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Mateusz Guzik <mjguzik@gmail.com>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michael S. Tsirkin <mst@redhat.com>
Cc: Mike Christie <michael.christie@oracle.com>
Cc: Nicholas Piggin <npiggin@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
When destroying maple tree, preserve its attributes and then turn it into
an empty tree. This allows it to be reused without needing to be
reinitialized.
Link: https://lkml.kernel.org/r/20231027033845.90608-10-zhangpeng.00@bytedance.com
Signed-off-by: Peng Zhang <zhangpeng.00@bytedance.com>
Reviewed-by: Liam R. Howlett <Liam.Howlett@oracle.com>
Cc: Christian Brauner <brauner@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Mateusz Guzik <mjguzik@gmail.com>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michael S. Tsirkin <mst@redhat.com>
Cc: Mike Christie <michael.christie@oracle.com>
Cc: Nicholas Piggin <npiggin@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Updated check_forking() and bench_forking() to use __mt_dup() to duplicate
maple tree.
Link: https://lkml.kernel.org/r/20231027033845.90608-9-zhangpeng.00@bytedance.com
Signed-off-by: Peng Zhang <zhangpeng.00@bytedance.com>
Reviewed-by: Liam R. Howlett <Liam.Howlett@oracle.com>
Cc: Christian Brauner <brauner@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Mateusz Guzik <mjguzik@gmail.com>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michael S. Tsirkin <mst@redhat.com>
Cc: Mike Christie <michael.christie@oracle.com>
Cc: Nicholas Piggin <npiggin@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Skip other tests when BENCH is enabled so that performance can be measured
in user space.
Link: https://lkml.kernel.org/r/20231027033845.90608-8-zhangpeng.00@bytedance.com
Signed-off-by: Peng Zhang <zhangpeng.00@bytedance.com>
Reviewed-by: Liam R. Howlett <Liam.Howlett@oracle.com>
Cc: Christian Brauner <brauner@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Mateusz Guzik <mjguzik@gmail.com>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michael S. Tsirkin <mst@redhat.com>
Cc: Mike Christie <michael.christie@oracle.com>
Cc: Nicholas Piggin <npiggin@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Introduce the new interface mtree_dup() in the documentation.
Link: https://lkml.kernel.org/r/20231027033845.90608-7-zhangpeng.00@bytedance.com
Signed-off-by: Peng Zhang <zhangpeng.00@bytedance.com>
Reviewed-by: Liam R. Howlett <Liam.Howlett@oracle.com>
Cc: Christian Brauner <brauner@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Mateusz Guzik <mjguzik@gmail.com>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michael S. Tsirkin <mst@redhat.com>
Cc: Mike Christie <michael.christie@oracle.com>
Cc: Nicholas Piggin <npiggin@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Add test for mtree_dup().
Test by duplicating different maple trees and then comparing the two
trees. Includes tests for duplicating full trees and memory allocation
failures on different nodes.
Link: https://lkml.kernel.org/r/20231027033845.90608-6-zhangpeng.00@bytedance.com
Signed-off-by: Peng Zhang <zhangpeng.00@bytedance.com>
Reviewed-by: Liam R. Howlett <Liam.Howlett@oracle.com>
Cc: Christian Brauner <brauner@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Mateusz Guzik <mjguzik@gmail.com>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michael S. Tsirkin <mst@redhat.com>
Cc: Mike Christie <michael.christie@oracle.com>
Cc: Nicholas Piggin <npiggin@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
When kmem_cache_alloc_bulk() fails to allocate, leave the freed pointers
in the array. This enables a more accurate simulation of the kernel's
behavior and allows for testing potential double-free scenarios.
Link: https://lkml.kernel.org/r/20231027033845.90608-5-zhangpeng.00@bytedance.com
Signed-off-by: Peng Zhang <zhangpeng.00@bytedance.com>
Reviewed-by: Liam R. Howlett <Liam.Howlett@oracle.com>
Cc: Christian Brauner <brauner@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Mateusz Guzik <mjguzik@gmail.com>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michael S. Tsirkin <mst@redhat.com>
Cc: Mike Christie <michael.christie@oracle.com>
Cc: Nicholas Piggin <npiggin@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Introduce interfaces __mt_dup() and mtree_dup(), which are used to
duplicate a maple tree. They duplicate a maple tree in Depth-First Search
(DFS) pre-order traversal. It uses memcopy() to copy nodes in the source
tree and allocate new child nodes in non-leaf nodes. The new node is
exactly the same as the source node except for all the addresses stored in
it. It will be faster than traversing all elements in the source tree and
inserting them one by one into the new tree. The time complexity of these
two functions is O(n).
The difference between __mt_dup() and mtree_dup() is that mtree_dup()
handles locks internally.
Analysis of the average time complexity of this algorithm:
For simplicity, let's assume that the maximum branching factor of all
non-leaf nodes is 16 (in allocation mode, it is 10), and the tree is a
full tree.
Under the given conditions, if there is a maple tree with n elements, the
number of its leaves is n/16. From bottom to top, the number of nodes in
each level is 1/16 of the number of nodes in the level below. So the
total number of nodes in the entire tree is given by the sum of n/16 +
n/16^2 + n/16^3 + ... + 1. This is a geometric series, and it has log(n)
terms with base 16. According to the formula for the sum of a geometric
series, the sum of this series can be calculated as (n-1)/15. Each node
has only one parent node pointer, which can be considered as an edge. In
total, there are (n-1)/15-1 edges.
This algorithm consists of two operations:
1. Traversing all nodes in DFS order.
2. For each node, making a copy and performing necessary modifications
to create a new node.
For the first part, DFS traversal will visit each edge twice. Let
T(ascend) represent the cost of taking one step downwards, and T(descend)
represent the cost of taking one step upwards. And both of them are
constants (although mas_ascend() may not be, as it contains a loop, but
here we ignore it and treat it as a constant). So the time spent on the
first part can be represented as ((n-1)/15-1) * (T(ascend) + T(descend)).
For the second part, each node will be copied, and the cost of copying a
node is denoted as T(copy_node). For each non-leaf node, it is necessary
to reallocate all child nodes, and the cost of this operation is denoted
as T(dup_alloc). The behavior behind memory allocation is complex and not
specific to the maple tree operation. Here, we assume that the time
required for a single allocation is constant. Since the size of a node is
fixed, both of these symbols are also constants. We can calculate that
the time spent on the second part is ((n-1)/15) * T(copy_node) + ((n-1)/15
- n/16) * T(dup_alloc).
Adding both parts together, the total time spent by the algorithm can be
represented as:
((n-1)/15) * (T(ascend) + T(descend) + T(copy_node) + T(dup_alloc)) -
n/16 * T(dup_alloc) - (T(ascend) + T(descend))
Let C1 = T(ascend) + T(descend) + T(copy_node) + T(dup_alloc)
Let C2 = T(dup_alloc)
Let C3 = T(ascend) + T(descend)
Finally, the expression can be simplified as:
((16 * C1 - 15 * C2) / (15 * 16)) * n - (C1 / 15 + C3).
This is a linear function, so the average time complexity is O(n).
Link: https://lkml.kernel.org/r/20231027033845.90608-4-zhangpeng.00@bytedance.com
Signed-off-by: Peng Zhang <zhangpeng.00@bytedance.com>
Suggested-by: Liam R. Howlett <Liam.Howlett@oracle.com>
Cc: Christian Brauner <brauner@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Mateusz Guzik <mjguzik@gmail.com>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michael S. Tsirkin <mst@redhat.com>
Cc: Mike Christie <michael.christie@oracle.com>
Cc: Nicholas Piggin <npiggin@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
In some cases, nested locks may be needed, so {mtree,mas}_lock_nested is
introduced. For example, when duplicating maple tree, we need to hold the
locks of two trees, in which case nested locks are needed.
At the same time, add the definition of spin_lock_nested() in tools for
testing.
Link: https://lkml.kernel.org/r/20231027033845.90608-3-zhangpeng.00@bytedance.com
Signed-off-by: Peng Zhang <zhangpeng.00@bytedance.com>
Reviewed-by: Liam R. Howlett <Liam.Howlett@oracle.com>
Cc: Christian Brauner <brauner@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Mateusz Guzik <mjguzik@gmail.com>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michael S. Tsirkin <mst@redhat.com>
Cc: Mike Christie <michael.christie@oracle.com>
Cc: Nicholas Piggin <npiggin@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Patch series "Introduce __mt_dup() to improve the performance of fork()", v7.
This series introduces __mt_dup() to improve the performance of fork().
During the duplication process of mmap, all VMAs are traversed and
inserted one by one into the new maple tree, causing the maple tree to be
rebalanced multiple times. Balancing the maple tree is a costly
operation. To duplicate VMAs more efficiently, mtree_dup() and __mt_dup()
are introduced for the maple tree. They can efficiently duplicate a maple
tree.
Here are some algorithmic details about {mtree,__mt}_dup(). We perform a
DFS pre-order traversal of all nodes in the source maple tree. During
this process, we fully copy the nodes from the source tree to the new
tree. This involves memory allocation, and when encountering a new node,
if it is a non-leaf node, all its child nodes are allocated at once.
This idea was originally from Liam R. Howlett's Maple Tree Work email,
and I added some of my own ideas to implement it. Some previous
discussions can be found in [1]. For a more detailed analysis of the
algorithm, please refer to the logs for patch [3/10] and patch [10/10].
There is a "spawn" in byte-unixbench[2], which can be used to test the
performance of fork(). I modified it slightly to make it work with
different number of VMAs.
Below are the test results. The first row shows the number of VMAs. The
second and third rows show the number of fork() calls per ten seconds,
corresponding to next-20231006 and the this patchset, respectively. The
test results were obtained with CPU binding to avoid scheduler load
balancing that could cause unstable results. There are still some
fluctuations in the test results, but at least they are better than the
original performance.
21 121 221 421 821 1621 3221 6421 12821 25621 51221
112100 76261 54227 34035 20195 11112 6017 3161 1606 802 393
114558 83067 65008 45824 28751 16072 8922 4747 2436 1233 599
2.19% 8.92% 19.88% 34.64% 42.37% 44.64% 48.28% 50.17% 51.68% 53.74% 52.42%
Thanks to Liam and Matthew for the review.
This patch (of 10):
Add two helpers:
1. mt_free_one(), used to free a maple node.
2. mt_attr(), used to obtain the attributes of maple tree.
Link: https://lkml.kernel.org/r/20231027033845.90608-1-zhangpeng.00@bytedance.com
Link: https://lkml.kernel.org/r/20231027033845.90608-2-zhangpeng.00@bytedance.com
Signed-off-by: Peng Zhang <zhangpeng.00@bytedance.com>
Reviewed-by: Liam R. Howlett <Liam.Howlett@oracle.com>
Cc: Christian Brauner <brauner@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Mateusz Guzik <mjguzik@gmail.com>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michael S. Tsirkin <mst@redhat.com>
Cc: Mike Christie <michael.christie@oracle.com>
Cc: Nicholas Piggin <npiggin@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
