system/linux
system/linux
1
0
Fork 0
mirror of https://github.com/torvalds/linux synced 2024-11-05 18:23:50 +00:00
Code Issues Projects Releases Packages Wiki Activity
linux/fs/io-wq.c

1212 lines
30 KiB
C
Raw Normal View History

io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
// SPDX-License-Identifier: GPL-2.0
/*
* Basic worker thread pool for io_uring
*
* Copyright (C) 2019 Jens Axboe
*
*/
#include <linux/kernel.h>
#include <linux/init.h>
#include <linux/errno.h>
#include <linux/sched/signal.h>
#include <linux/mm.h>
#include <linux/sched/mm.h>
#include <linux/percpu.h>
#include <linux/slab.h>
#include <linux/kthread.h>
#include <linux/rculist_nulls.h>
io-wq: add support for inheriting ->fs Some work items need this for relative path lookup, make it available like the other inherited credentials/mm/etc. Cc: stable@vger.kernel.org # 5.3+ Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-02-07 04:42:51 +00:00
#include <linux/fs_struct.h>
io_uring: use io-wq manager as backup task if task is exiting If the original task is (or has) exited, then the task work will not get queued properly. Allow for using the io-wq manager task to queue this work for execution, and ensure that the io-wq manager notices and runs this work if woken up (or exiting). Reported-by: Dan Melnic <dmm@fb.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-04-03 17:26:26 +00:00
#include <linux/task_work.h>
io_uring: add blkcg accounting to offloaded operations There are a few operations that are offloaded to the worker threads. In this case, we lose process context and end up in kthread context. This results in ios to be not accounted to the issuing cgroup and consequently end up as issued by root. Just like others, adopt the personality of the blkcg too when issuing via the workqueues. For the SQPOLL thread, it will live and attach in the inited cgroup's context. Signed-off-by: Dennis Zhou <dennis@kernel.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-16 20:41:05 +00:00
#include <linux/blk-cgroup.h>
io-wq: inherit audit loginuid and sessionid Make sure the async io-wq workers inherit the loginuid and sessionid from the original task, and restore them to unset once we're done with the async work item. While at it, disable the ability for kernel threads to write to their own loginuid. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-10-15 19:46:44 +00:00
#include <linux/audit.h>
io-wq: re-set NUMA node affinities if CPUs come online We correctly set io-wq NUMA node affinities when the io-wq context is setup, but if an entire node CPU set is offlined and then brought back online, the per node affinities are broken. Ensure that we set them again whenever a CPU comes online. This ensures that we always track the right node affinity. The usual cpuhp notifiers are used to drive it. Reported-by: Zhang Qiang <qiang.zhang@windriver.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-10-22 15:02:50 +00:00
#include <linux/cpu.h>
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
io-wq: re-set NUMA node affinities if CPUs come online We correctly set io-wq NUMA node affinities when the io-wq context is setup, but if an entire node CPU set is offlined and then brought back online, the per node affinities are broken. Ensure that we set them again whenever a CPU comes online. This ensures that we always track the right node affinity. The usual cpuhp notifiers are used to drive it. Reported-by: Zhang Qiang <qiang.zhang@windriver.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-10-22 15:02:50 +00:00
#include "../kernel/sched/sched.h"
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
#include "io-wq.h"
#define WORKER_IDLE_TIMEOUT (5 * HZ)
enum {
IO_WORKER_F_UP = 1, /* up and active */
IO_WORKER_F_RUNNING = 2, /* account as running */
IO_WORKER_F_FREE = 4, /* worker on free list */
io-wq: kill unused IO_WORKER_F_EXITING This flag is no longer used, remove it. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-26 18:37:46 +00:00
IO_WORKER_F_FIXED = 8, /* static idle worker */
IO_WORKER_F_BOUND = 16, /* is doing bounded work */
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
};
enum {
IO_WQ_BIT_EXIT = 0, /* wq exiting */
io-wq: kill now unused io_wq_cancel_all() io_uring no longer issues full cancelations on the io-wq, so remove any remnants of this code and the IO_WQ_BIT_CANCEL flag. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-12-20 17:47:42 +00:00
IO_WQ_BIT_ERROR = 1, /* error on setup */
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
};
enum {
IO_WQE_FLAG_STALLED = 1, /* stalled on hash */
};
/*
* One for each thread in a wqe pool
*/
struct io_worker {
refcount_t ref;
unsigned flags;
struct hlist_nulls_node nulls_node;
io-wq: ensure free/busy list browsing see all items We have two lists for workers in io-wq, a busy and a free list. For certain operations we want to browse all workers, and we currently do that by browsing the two separate lists. But since these lists are RCU protected, we can potentially miss workers if they move between the two lists while we're browsing them. Add a third list, all_list, that simply holds all workers. A worker is added to that list when it starts, and removed when it exits. This makes the worker iteration cleaner, too. Reported-by: Paul E. McKenney <paulmck@kernel.org> Reviewed-by: Paul E. McKenney <paulmck@kernel.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-13 20:54:49 +00:00
struct list_head all_list;
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
struct task_struct *task;
struct io_wqe *wqe;
io-wq: ensure we have a stable view of ->cur_work for cancellations worker->cur_work is currently protected by the lock of the wqe that the worker belongs to. When we send a signal to a worker, we need a stable view of ->cur_work, so we need to hold that lock. But this doesn't work so well, since we have the opposite order potentially on queueing work. If POLL_ADD is used with a signalfd, then io_poll_wake() is called with the signal lock, and that sometimes needs to insert work items. Add a specific worker lock that protects the current work item. Then we can guarantee that the task we're sending a signal is currently processing the exact work we think it is. Reported-by: Paul E. McKenney <paulmck@kernel.org> Reviewed-by: Paul E. McKenney <paulmck@kernel.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-13 16:43:34 +00:00
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
struct io_wq_work *cur_work;
io-wq: ensure we have a stable view of ->cur_work for cancellations worker->cur_work is currently protected by the lock of the wqe that the worker belongs to. When we send a signal to a worker, we need a stable view of ->cur_work, so we need to hold that lock. But this doesn't work so well, since we have the opposite order potentially on queueing work. If POLL_ADD is used with a signalfd, then io_poll_wake() is called with the signal lock, and that sometimes needs to insert work items. Add a specific worker lock that protects the current work item. Then we can guarantee that the task we're sending a signal is currently processing the exact work we think it is. Reported-by: Paul E. McKenney <paulmck@kernel.org> Reviewed-by: Paul E. McKenney <paulmck@kernel.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-13 16:43:34 +00:00
spinlock_t lock;
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
struct rcu_head rcu;
struct mm_struct *mm;
io_uring: add blkcg accounting to offloaded operations There are a few operations that are offloaded to the worker threads. In this case, we lose process context and end up in kthread context. This results in ios to be not accounted to the issuing cgroup and consequently end up as issued by root. Just like others, adopt the personality of the blkcg too when issuing via the workqueues. For the SQPOLL thread, it will live and attach in the inited cgroup's context. Signed-off-by: Dennis Zhou <dennis@kernel.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-16 20:41:05 +00:00
#ifdef CONFIG_BLK_CGROUP
struct cgroup_subsys_state *blkcg_css;
#endif
io_uring/io-wq: don't use static creds/mm assignments We currently setup the io_wq with a static set of mm and creds. Even for a single-use io-wq per io_uring, this is suboptimal as we have may have multiple enters of the ring. For sharing the io-wq backend, it doesn't work at all. Switch to passing in the creds and mm when the work item is setup. This means that async work is no longer deferred to the io_uring mm and creds, it is done with the current mm and creds. Flag this behavior with IORING_FEAT_CUR_PERSONALITY, so applications know they can rely on the current personality (mm and creds) being the same for direct issue and async issue. Reviewed-by: Stefan Metzmacher <metze@samba.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-01-27 23:34:48 +00:00
const struct cred *cur_creds;
const struct cred *saved_creds;
io_uring: io_uring: add support for async work inheriting files This is in preparation for adding opcodes that need to add new files in a process file table, system calls like open(2) or accept4(2). If an opcode needs this, it must set IO_WQ_WORK_NEEDS_FILES in the work item. If work that needs to get punted to async context have this set, the async worker will assume the original task file table before executing the work. Note that opcodes that need access to the current files of an application cannot be done through IORING_SETUP_SQPOLL. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-24 18:39:47 +00:00
struct files_struct *restore_files;
io_uring: reference ->nsproxy for file table commands If we don't get and assign the namespace for the async work, then certain paths just don't work properly (like /dev/stdin, /proc/mounts, etc). Anything that references the current namespace of the given task should be assigned for async work on behalf of that task. Cc: stable@vger.kernel.org # v5.5+ Reported-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-19 02:13:06 +00:00
struct nsproxy *restore_nsproxy;
io-wq: add support for inheriting ->fs Some work items need this for relative path lookup, make it available like the other inherited credentials/mm/etc. Cc: stable@vger.kernel.org # 5.3+ Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-02-07 04:42:51 +00:00
struct fs_struct *restore_fs;
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
};
#if BITS_PER_LONG == 64
#define IO_WQ_HASH_ORDER 6
#else
#define IO_WQ_HASH_ORDER 5
#endif
io-wq: handle hashed writes in chains We always punt async buffered writes to an io-wq helper, as the core kernel does not have IOCB_NOWAIT support for that. Most buffered async writes complete very quickly, as it's just a copy operation. This means that doing multiple locking roundtrips on the shared wqe lock for each buffered write is wasteful. Additionally, buffered writes are hashed work items, which means that any buffered write to a given file is serialized. Keep identicaly hashed work items contiguously in @wqe->work_list, and track a tail for each hash bucket. On dequeue of a hashed item, splice all of the same hash in one go using the tracked tail. Until the batch is done, the caller doesn't have to synchronize with the wqe or worker locks again. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-23 19:57:22 +00:00
#define IO_WQ_NR_HASH_BUCKETS (1u << IO_WQ_HASH_ORDER)
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
struct io_wqe_acct {
unsigned nr_workers;
unsigned max_workers;
atomic_t nr_running;
};
enum {
IO_WQ_ACCT_BOUND,
IO_WQ_ACCT_UNBOUND,
};
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
/*
* Per-node worker thread pool
*/
struct io_wqe {
struct {
io_wq: Make io_wqe::lock a raw_spinlock_t During a context switch the scheduler invokes wq_worker_sleeping() with disabled preemption. Disabling preemption is needed because it protects access to `worker->sleeping'. As an optimisation it avoids invoking schedule() within the schedule path as part of possible wake up (thus preempt_enable_no_resched() afterwards). The io-wq has been added to the mix in the same section with disabled preemption. This breaks on PREEMPT_RT because io_wq_worker_sleeping() acquires a spinlock_t. Also within the schedule() the spinlock_t must be acquired after tsk_is_pi_blocked() otherwise it will block on the sleeping lock again while scheduling out. While playing with `io_uring-bench' I didn't notice a significant latency spike after converting io_wqe::lock to a raw_spinlock_t. The latency was more or less the same. In order to keep the spinlock_t it would have to be moved after the tsk_is_pi_blocked() check which would introduce a branch instruction into the hot path. The lock is used to maintain the `work_list' and wakes one task up at most. Should io_wqe_cancel_pending_work() cause latency spikes, while searching for a specific item, then it would need to drop the lock during iterations. revert_creds() is also invoked under the lock. According to debug cred::non_rcu is 0. Otherwise it should be moved outside of the locked section because put_cred_rcu()->free_uid() acquires a sleeping lock. Convert io_wqe::lock to a raw_spinlock_t.c Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-01 08:41:46 +00:00
raw_spinlock_t lock;
io-wq: shrink io_wq_work a bit Currently we're using 40 bytes for the io_wq_work structure, and 16 of those is the doubly link list node. We don't need doubly linked lists, we always add to tail to keep things ordered, and any other use case is list traversal with deletion. For the deletion case, we can easily support any node deletion by keeping track of the previous entry. This shrinks io_wq_work to 32 bytes, and subsequently io_kiock from io_uring to 216 to 208 bytes. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-26 18:59:32 +00:00
struct io_wq_work_list work_list;
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
unsigned long hash_map;
unsigned flags;
} ____cacheline_aligned_in_smp;
int node;
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
struct io_wqe_acct acct[2];
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
io-wq: remove now redundant struct io_wq_nulls_list Since we don't iterate these lists anymore after commit: e61df66c69b1 ("io-wq: ensure free/busy list browsing see all items") we don't need to retain the nulls value we use for them. That means it's pretty pointless to wrap the hlist_nulls_head in a structure, so get rid of it. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-14 15:00:41 +00:00
struct hlist_nulls_head free_list;
io-wq: ensure free/busy list browsing see all items We have two lists for workers in io-wq, a busy and a free list. For certain operations we want to browse all workers, and we currently do that by browsing the two separate lists. But since these lists are RCU protected, we can potentially miss workers if they move between the two lists while we're browsing them. Add a third list, all_list, that simply holds all workers. A worker is added to that list when it starts, and removed when it exits. This makes the worker iteration cleaner, too. Reported-by: Paul E. McKenney <paulmck@kernel.org> Reviewed-by: Paul E. McKenney <paulmck@kernel.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-13 20:54:49 +00:00
struct list_head all_list;
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
struct io_wq *wq;
io-wq: handle hashed writes in chains We always punt async buffered writes to an io-wq helper, as the core kernel does not have IOCB_NOWAIT support for that. Most buffered async writes complete very quickly, as it's just a copy operation. This means that doing multiple locking roundtrips on the shared wqe lock for each buffered write is wasteful. Additionally, buffered writes are hashed work items, which means that any buffered write to a given file is serialized. Keep identicaly hashed work items contiguously in @wqe->work_list, and track a tail for each hash bucket. On dequeue of a hashed item, splice all of the same hash in one go using the tracked tail. Until the batch is done, the caller doesn't have to synchronize with the wqe or worker locks again. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-23 19:57:22 +00:00
struct io_wq_work *hash_tail[IO_WQ_NR_HASH_BUCKETS];
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
};
/*
* Per io_wq state
*/
struct io_wq {
struct io_wqe **wqes;
unsigned long state;
io_uring/io-wq: forward submission ref to async First it changes io-wq interfaces. It replaces {get,put}_work() with free_work(), which guaranteed to be called exactly once. It also enforces free_work() callback to be non-NULL. io_uring follows the changes and instead of putting a submission reference in io_put_req_async_completion(), it will be done in io_free_work(). As removes io_get_work() with corresponding refcount_inc(), the ref balance is maintained. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-04 13:14:12 +00:00
free_work_fn *free_work;
io_wq: add per-wq work handler instead of per work io_uring is the only user of io-wq, and now it uses only io-wq callback for all its requests, namely io_wq_submit_work(). Instead of storing work->runner callback in each instance of io_wq_work, keep it in io-wq itself. pros: - reduces io_wq_work size - more robust -- ->func won't be invalidated with mem{cpy,set}(req) - helps other work Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-06-08 18:08:20 +00:00
io_wq_work_fn *do_work;
io_wq: add get/put_work handlers to io_wq_create() For cancellation, we need to ensure that the work item stays valid for as long as ->cur_work is valid. Right now we can't safely dereference the work item even under the wqe->lock, because while the ->cur_work pointer will remain valid, the work could be completing and be freed in parallel. Only invoke ->get/put_work() on items we know that the caller queued themselves. Add IO_WQ_WORK_INTERNAL for io-wq to use, which is needed when we're queueing a flush item, for instance. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-13 05:31:31 +00:00
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
struct task_struct *manager;
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
struct user_struct *user;
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
refcount_t refs;
struct completion done;
io-wq: make the io_wq ref counted In preparation for sharing an io-wq across different users, add a reference count that manages destruction of it. Reviewed-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-01-23 22:33:32 +00:00
io-wq: re-set NUMA node affinities if CPUs come online We correctly set io-wq NUMA node affinities when the io-wq context is setup, but if an entire node CPU set is offlined and then brought back online, the per node affinities are broken. Ensure that we set them again whenever a CPU comes online. This ensures that we always track the right node affinity. The usual cpuhp notifiers are used to drive it. Reported-by: Zhang Qiang <qiang.zhang@windriver.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-10-22 15:02:50 +00:00
struct hlist_node cpuhp_node;
io-wq: make the io_wq ref counted In preparation for sharing an io-wq across different users, add a reference count that manages destruction of it. Reviewed-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-01-23 22:33:32 +00:00
refcount_t use_refs;
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
};
io-wq: re-set NUMA node affinities if CPUs come online We correctly set io-wq NUMA node affinities when the io-wq context is setup, but if an entire node CPU set is offlined and then brought back online, the per node affinities are broken. Ensure that we set them again whenever a CPU comes online. This ensures that we always track the right node affinity. The usual cpuhp notifiers are used to drive it. Reported-by: Zhang Qiang <qiang.zhang@windriver.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-10-22 15:02:50 +00:00
static enum cpuhp_state io_wq_online;
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
static bool io_worker_get(struct io_worker *worker)
{
return refcount_inc_not_zero(&worker->ref);
}
static void io_worker_release(struct io_worker *worker)
{
if (refcount_dec_and_test(&worker->ref))
wake_up_process(worker->task);
}
/*
* Note: drops the wqe->lock if returning true! The caller must re-acquire
* the lock in that case. Some callers need to restart handling if this
* happens, so we can't just re-acquire the lock on behalf of the caller.
*/
static bool __io_worker_unuse(struct io_wqe *wqe, struct io_worker *worker)
{
io_uring: io_uring: add support for async work inheriting files This is in preparation for adding opcodes that need to add new files in a process file table, system calls like open(2) or accept4(2). If an opcode needs this, it must set IO_WQ_WORK_NEEDS_FILES in the work item. If work that needs to get punted to async context have this set, the async worker will assume the original task file table before executing the work. Note that opcodes that need access to the current files of an application cannot be done through IORING_SETUP_SQPOLL. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-24 18:39:47 +00:00
bool dropped_lock = false;
io_uring/io-wq: don't use static creds/mm assignments We currently setup the io_wq with a static set of mm and creds. Even for a single-use io-wq per io_uring, this is suboptimal as we have may have multiple enters of the ring. For sharing the io-wq backend, it doesn't work at all. Switch to passing in the creds and mm when the work item is setup. This means that async work is no longer deferred to the io_uring mm and creds, it is done with the current mm and creds. Flag this behavior with IORING_FEAT_CUR_PERSONALITY, so applications know they can rely on the current personality (mm and creds) being the same for direct issue and async issue. Reviewed-by: Stefan Metzmacher <metze@samba.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-01-27 23:34:48 +00:00
if (worker->saved_creds) {
revert_creds(worker->saved_creds);
worker->cur_creds = worker->saved_creds = NULL;
io_uring: async workers should inherit the user creds If we don't inherit the original task creds, then we can confuse users like fuse that pass creds in the request header. See link below on identical aio issue. Link: https://lore.kernel.org/linux-fsdevel/26f0d78e-99ca-2f1b-78b9-433088053a61@scylladb.com/T/#u Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-25 15:52:30 +00:00
}
io_uring: io_uring: add support for async work inheriting files This is in preparation for adding opcodes that need to add new files in a process file table, system calls like open(2) or accept4(2). If an opcode needs this, it must set IO_WQ_WORK_NEEDS_FILES in the work item. If work that needs to get punted to async context have this set, the async worker will assume the original task file table before executing the work. Note that opcodes that need access to the current files of an application cannot be done through IORING_SETUP_SQPOLL. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-24 18:39:47 +00:00
if (current->files != worker->restore_files) {
__acquire(&wqe->lock);
io_wq: Make io_wqe::lock a raw_spinlock_t During a context switch the scheduler invokes wq_worker_sleeping() with disabled preemption. Disabling preemption is needed because it protects access to `worker->sleeping'. As an optimisation it avoids invoking schedule() within the schedule path as part of possible wake up (thus preempt_enable_no_resched() afterwards). The io-wq has been added to the mix in the same section with disabled preemption. This breaks on PREEMPT_RT because io_wq_worker_sleeping() acquires a spinlock_t. Also within the schedule() the spinlock_t must be acquired after tsk_is_pi_blocked() otherwise it will block on the sleeping lock again while scheduling out. While playing with `io_uring-bench' I didn't notice a significant latency spike after converting io_wqe::lock to a raw_spinlock_t. The latency was more or less the same. In order to keep the spinlock_t it would have to be moved after the tsk_is_pi_blocked() check which would introduce a branch instruction into the hot path. The lock is used to maintain the `work_list' and wakes one task up at most. Should io_wqe_cancel_pending_work() cause latency spikes, while searching for a specific item, then it would need to drop the lock during iterations. revert_creds() is also invoked under the lock. According to debug cred::non_rcu is 0. Otherwise it should be moved outside of the locked section because put_cred_rcu()->free_uid() acquires a sleeping lock. Convert io_wqe::lock to a raw_spinlock_t.c Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-01 08:41:46 +00:00
raw_spin_unlock_irq(&wqe->lock);
io_uring: io_uring: add support for async work inheriting files This is in preparation for adding opcodes that need to add new files in a process file table, system calls like open(2) or accept4(2). If an opcode needs this, it must set IO_WQ_WORK_NEEDS_FILES in the work item. If work that needs to get punted to async context have this set, the async worker will assume the original task file table before executing the work. Note that opcodes that need access to the current files of an application cannot be done through IORING_SETUP_SQPOLL. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-24 18:39:47 +00:00
dropped_lock = true;
task_lock(current);
current->files = worker->restore_files;
io_uring: reference ->nsproxy for file table commands If we don't get and assign the namespace for the async work, then certain paths just don't work properly (like /dev/stdin, /proc/mounts, etc). Anything that references the current namespace of the given task should be assigned for async work on behalf of that task. Cc: stable@vger.kernel.org # v5.5+ Reported-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-19 02:13:06 +00:00
current->nsproxy = worker->restore_nsproxy;
io_uring: io_uring: add support for async work inheriting files This is in preparation for adding opcodes that need to add new files in a process file table, system calls like open(2) or accept4(2). If an opcode needs this, it must set IO_WQ_WORK_NEEDS_FILES in the work item. If work that needs to get punted to async context have this set, the async worker will assume the original task file table before executing the work. Note that opcodes that need access to the current files of an application cannot be done through IORING_SETUP_SQPOLL. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-24 18:39:47 +00:00
task_unlock(current);
}
io-wq: add support for inheriting ->fs Some work items need this for relative path lookup, make it available like the other inherited credentials/mm/etc. Cc: stable@vger.kernel.org # 5.3+ Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-02-07 04:42:51 +00:00
if (current->fs != worker->restore_fs)
current->fs = worker->restore_fs;
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
/*
* If we have an active mm, we need to drop the wq lock before unusing
* it. If we do, return true and let the caller retry the idle loop.
*/
if (worker->mm) {
io_uring: io_uring: add support for async work inheriting files This is in preparation for adding opcodes that need to add new files in a process file table, system calls like open(2) or accept4(2). If an opcode needs this, it must set IO_WQ_WORK_NEEDS_FILES in the work item. If work that needs to get punted to async context have this set, the async worker will assume the original task file table before executing the work. Note that opcodes that need access to the current files of an application cannot be done through IORING_SETUP_SQPOLL. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-24 18:39:47 +00:00
if (!dropped_lock) {
__acquire(&wqe->lock);
io_wq: Make io_wqe::lock a raw_spinlock_t During a context switch the scheduler invokes wq_worker_sleeping() with disabled preemption. Disabling preemption is needed because it protects access to `worker->sleeping'. As an optimisation it avoids invoking schedule() within the schedule path as part of possible wake up (thus preempt_enable_no_resched() afterwards). The io-wq has been added to the mix in the same section with disabled preemption. This breaks on PREEMPT_RT because io_wq_worker_sleeping() acquires a spinlock_t. Also within the schedule() the spinlock_t must be acquired after tsk_is_pi_blocked() otherwise it will block on the sleeping lock again while scheduling out. While playing with `io_uring-bench' I didn't notice a significant latency spike after converting io_wqe::lock to a raw_spinlock_t. The latency was more or less the same. In order to keep the spinlock_t it would have to be moved after the tsk_is_pi_blocked() check which would introduce a branch instruction into the hot path. The lock is used to maintain the `work_list' and wakes one task up at most. Should io_wqe_cancel_pending_work() cause latency spikes, while searching for a specific item, then it would need to drop the lock during iterations. revert_creds() is also invoked under the lock. According to debug cred::non_rcu is 0. Otherwise it should be moved outside of the locked section because put_cred_rcu()->free_uid() acquires a sleeping lock. Convert io_wqe::lock to a raw_spinlock_t.c Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-01 08:41:46 +00:00
raw_spin_unlock_irq(&wqe->lock);
io_uring: io_uring: add support for async work inheriting files This is in preparation for adding opcodes that need to add new files in a process file table, system calls like open(2) or accept4(2). If an opcode needs this, it must set IO_WQ_WORK_NEEDS_FILES in the work item. If work that needs to get punted to async context have this set, the async worker will assume the original task file table before executing the work. Note that opcodes that need access to the current files of an application cannot be done through IORING_SETUP_SQPOLL. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-24 18:39:47 +00:00
dropped_lock = true;
}
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
__set_current_state(TASK_RUNNING);
kernel: better document the use_mm/unuse_mm API contract Switch the function documentation to kerneldoc comments, and add WARN_ON_ONCE asserts that the calling thread is a kernel thread and does not have ->mm set (or has ->mm set in the case of unuse_mm). Also give the functions a kthread_ prefix to better document the use case. [hch@lst.de: fix a comment typo, cover the newly merged use_mm/unuse_mm caller in vfio] Link: http://lkml.kernel.org/r/20200416053158.586887-3-hch@lst.de [sfr@canb.auug.org.au: powerpc/vas: fix up for {un}use_mm() rename] Link: http://lkml.kernel.org/r/20200422163935.5aa93ba5@canb.auug.org.au Signed-off-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Tested-by: Jens Axboe <axboe@kernel.dk> Reviewed-by: Jens Axboe <axboe@kernel.dk> Acked-by: Felix Kuehling <Felix.Kuehling@amd.com> Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> [usb] Acked-by: Haren Myneni <haren@linux.ibm.com> Cc: Alex Deucher <alexander.deucher@amd.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Felipe Balbi <balbi@kernel.org> Cc: Jason Wang <jasowang@redhat.com> Cc: "Michael S. Tsirkin" <mst@redhat.com> Cc: Zhenyu Wang <zhenyuw@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Link: http://lkml.kernel.org/r/20200404094101.672954-6-hch@lst.de Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-06-11 01:42:06 +00:00
kthread_unuse_mm(worker->mm);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
mmput(worker->mm);
worker->mm = NULL;
}
io_uring: add blkcg accounting to offloaded operations There are a few operations that are offloaded to the worker threads. In this case, we lose process context and end up in kthread context. This results in ios to be not accounted to the issuing cgroup and consequently end up as issued by root. Just like others, adopt the personality of the blkcg too when issuing via the workqueues. For the SQPOLL thread, it will live and attach in the inited cgroup's context. Signed-off-by: Dennis Zhou <dennis@kernel.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-16 20:41:05 +00:00
#ifdef CONFIG_BLK_CGROUP
if (worker->blkcg_css) {
kthread_associate_blkcg(NULL);
worker->blkcg_css = NULL;
}
#endif
io_uring: unify fsize with def->work_flags This one was missed in the earlier conversion, should be included like any of the other IO identity flags. Make sure we restore to RLIM_INIFITY when dropping the personality again. Fixes: 98447d65b4a7 ("io_uring: move io identity items into separate struct") Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-10-20 20:28:41 +00:00
if (current->signal->rlim[RLIMIT_FSIZE].rlim_cur != RLIM_INFINITY)
current->signal->rlim[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY;
io_uring: io_uring: add support for async work inheriting files This is in preparation for adding opcodes that need to add new files in a process file table, system calls like open(2) or accept4(2). If an opcode needs this, it must set IO_WQ_WORK_NEEDS_FILES in the work item. If work that needs to get punted to async context have this set, the async worker will assume the original task file table before executing the work. Note that opcodes that need access to the current files of an application cannot be done through IORING_SETUP_SQPOLL. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-24 18:39:47 +00:00
return dropped_lock;
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
}
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
static inline struct io_wqe_acct *io_work_get_acct(struct io_wqe *wqe,
struct io_wq_work *work)
{
if (work->flags & IO_WQ_WORK_UNBOUND)
return &wqe->acct[IO_WQ_ACCT_UNBOUND];
return &wqe->acct[IO_WQ_ACCT_BOUND];
}
static inline struct io_wqe_acct *io_wqe_get_acct(struct io_wqe *wqe,
struct io_worker *worker)
{
if (worker->flags & IO_WORKER_F_BOUND)
return &wqe->acct[IO_WQ_ACCT_BOUND];
return &wqe->acct[IO_WQ_ACCT_UNBOUND];
}
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
static void io_worker_exit(struct io_worker *worker)
{
struct io_wqe *wqe = worker->wqe;
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
struct io_wqe_acct *acct = io_wqe_get_acct(wqe, worker);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
/*
* If we're not at zero, someone else is holding a brief reference
* to the worker. Wait for that to go away.
*/
set_current_state(TASK_INTERRUPTIBLE);
if (!refcount_dec_and_test(&worker->ref))
schedule();
__set_current_state(TASK_RUNNING);
preempt_disable();
current->flags &= ~PF_IO_WORKER;
if (worker->flags & IO_WORKER_F_RUNNING)
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
atomic_dec(&acct->nr_running);
if (!(worker->flags & IO_WORKER_F_BOUND))
atomic_dec(&wqe->wq->user->processes);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
worker->flags = 0;
preempt_enable();
io_wq: Make io_wqe::lock a raw_spinlock_t During a context switch the scheduler invokes wq_worker_sleeping() with disabled preemption. Disabling preemption is needed because it protects access to `worker->sleeping'. As an optimisation it avoids invoking schedule() within the schedule path as part of possible wake up (thus preempt_enable_no_resched() afterwards). The io-wq has been added to the mix in the same section with disabled preemption. This breaks on PREEMPT_RT because io_wq_worker_sleeping() acquires a spinlock_t. Also within the schedule() the spinlock_t must be acquired after tsk_is_pi_blocked() otherwise it will block on the sleeping lock again while scheduling out. While playing with `io_uring-bench' I didn't notice a significant latency spike after converting io_wqe::lock to a raw_spinlock_t. The latency was more or less the same. In order to keep the spinlock_t it would have to be moved after the tsk_is_pi_blocked() check which would introduce a branch instruction into the hot path. The lock is used to maintain the `work_list' and wakes one task up at most. Should io_wqe_cancel_pending_work() cause latency spikes, while searching for a specific item, then it would need to drop the lock during iterations. revert_creds() is also invoked under the lock. According to debug cred::non_rcu is 0. Otherwise it should be moved outside of the locked section because put_cred_rcu()->free_uid() acquires a sleeping lock. Convert io_wqe::lock to a raw_spinlock_t.c Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-01 08:41:46 +00:00
raw_spin_lock_irq(&wqe->lock);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
hlist_nulls_del_rcu(&worker->nulls_node);
io-wq: ensure free/busy list browsing see all items We have two lists for workers in io-wq, a busy and a free list. For certain operations we want to browse all workers, and we currently do that by browsing the two separate lists. But since these lists are RCU protected, we can potentially miss workers if they move between the two lists while we're browsing them. Add a third list, all_list, that simply holds all workers. A worker is added to that list when it starts, and removed when it exits. This makes the worker iteration cleaner, too. Reported-by: Paul E. McKenney <paulmck@kernel.org> Reviewed-by: Paul E. McKenney <paulmck@kernel.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-13 20:54:49 +00:00
list_del_rcu(&worker->all_list);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
if (__io_worker_unuse(wqe, worker)) {
__release(&wqe->lock);
io_wq: Make io_wqe::lock a raw_spinlock_t During a context switch the scheduler invokes wq_worker_sleeping() with disabled preemption. Disabling preemption is needed because it protects access to `worker->sleeping'. As an optimisation it avoids invoking schedule() within the schedule path as part of possible wake up (thus preempt_enable_no_resched() afterwards). The io-wq has been added to the mix in the same section with disabled preemption. This breaks on PREEMPT_RT because io_wq_worker_sleeping() acquires a spinlock_t. Also within the schedule() the spinlock_t must be acquired after tsk_is_pi_blocked() otherwise it will block on the sleeping lock again while scheduling out. While playing with `io_uring-bench' I didn't notice a significant latency spike after converting io_wqe::lock to a raw_spinlock_t. The latency was more or less the same. In order to keep the spinlock_t it would have to be moved after the tsk_is_pi_blocked() check which would introduce a branch instruction into the hot path. The lock is used to maintain the `work_list' and wakes one task up at most. Should io_wqe_cancel_pending_work() cause latency spikes, while searching for a specific item, then it would need to drop the lock during iterations. revert_creds() is also invoked under the lock. According to debug cred::non_rcu is 0. Otherwise it should be moved outside of the locked section because put_cred_rcu()->free_uid() acquires a sleeping lock. Convert io_wqe::lock to a raw_spinlock_t.c Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-01 08:41:46 +00:00
raw_spin_lock_irq(&wqe->lock);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
}
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
acct->nr_workers--;
io_wq: Make io_wqe::lock a raw_spinlock_t During a context switch the scheduler invokes wq_worker_sleeping() with disabled preemption. Disabling preemption is needed because it protects access to `worker->sleeping'. As an optimisation it avoids invoking schedule() within the schedule path as part of possible wake up (thus preempt_enable_no_resched() afterwards). The io-wq has been added to the mix in the same section with disabled preemption. This breaks on PREEMPT_RT because io_wq_worker_sleeping() acquires a spinlock_t. Also within the schedule() the spinlock_t must be acquired after tsk_is_pi_blocked() otherwise it will block on the sleeping lock again while scheduling out. While playing with `io_uring-bench' I didn't notice a significant latency spike after converting io_wqe::lock to a raw_spinlock_t. The latency was more or less the same. In order to keep the spinlock_t it would have to be moved after the tsk_is_pi_blocked() check which would introduce a branch instruction into the hot path. The lock is used to maintain the `work_list' and wakes one task up at most. Should io_wqe_cancel_pending_work() cause latency spikes, while searching for a specific item, then it would need to drop the lock during iterations. revert_creds() is also invoked under the lock. According to debug cred::non_rcu is 0. Otherwise it should be moved outside of the locked section because put_cred_rcu()->free_uid() acquires a sleeping lock. Convert io_wqe::lock to a raw_spinlock_t.c Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-01 08:41:46 +00:00
raw_spin_unlock_irq(&wqe->lock);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
io-wq: use kfree_rcu() to simplify the code The callback function of call_rcu() just calls kfree(), so we can use kfree_rcu() instead of call_rcu() + callback function. Signed-off-by: YueHaibing <yuehaibing@huawei.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-02 07:55:01 +00:00
kfree_rcu(worker, rcu);
io-wq: fix use-after-free in io_wq_worker_running The smart syzbot has found a reproducer for the following issue: ================================================================== BUG: KASAN: use-after-free in instrument_atomic_write include/linux/instrumented.h:71 [inline] BUG: KASAN: use-after-free in atomic_inc include/asm-generic/atomic-instrumented.h:240 [inline] BUG: KASAN: use-after-free in io_wqe_inc_running fs/io-wq.c:301 [inline] BUG: KASAN: use-after-free in io_wq_worker_running+0xde/0x110 fs/io-wq.c:613 Write of size 4 at addr ffff8882183db08c by task io_wqe_worker-0/7771 CPU: 0 PID: 7771 Comm: io_wqe_worker-0 Not tainted 5.9.0-rc4-syzkaller #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011 Call Trace: __dump_stack lib/dump_stack.c:77 [inline] dump_stack+0x198/0x1fd lib/dump_stack.c:118 print_address_description.constprop.0.cold+0xae/0x497 mm/kasan/report.c:383 __kasan_report mm/kasan/report.c:513 [inline] kasan_report.cold+0x1f/0x37 mm/kasan/report.c:530 check_memory_region_inline mm/kasan/generic.c:186 [inline] check_memory_region+0x13d/0x180 mm/kasan/generic.c:192 instrument_atomic_write include/linux/instrumented.h:71 [inline] atomic_inc include/asm-generic/atomic-instrumented.h:240 [inline] io_wqe_inc_running fs/io-wq.c:301 [inline] io_wq_worker_running+0xde/0x110 fs/io-wq.c:613 schedule_timeout+0x148/0x250 kernel/time/timer.c:1879 io_wqe_worker+0x517/0x10e0 fs/io-wq.c:580 kthread+0x3b5/0x4a0 kernel/kthread.c:292 ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:294 Allocated by task 7768: kasan_save_stack+0x1b/0x40 mm/kasan/common.c:48 kasan_set_track mm/kasan/common.c:56 [inline] __kasan_kmalloc.constprop.0+0xbf/0xd0 mm/kasan/common.c:461 kmem_cache_alloc_node_trace+0x17b/0x3f0 mm/slab.c:3594 kmalloc_node include/linux/slab.h:572 [inline] kzalloc_node include/linux/slab.h:677 [inline] io_wq_create+0x57b/0xa10 fs/io-wq.c:1064 io_init_wq_offload fs/io_uring.c:7432 [inline] io_sq_offload_start fs/io_uring.c:7504 [inline] io_uring_create fs/io_uring.c:8625 [inline] io_uring_setup+0x1836/0x28e0 fs/io_uring.c:8694 do_syscall_64+0x2d/0x70 arch/x86/entry/common.c:46 entry_SYSCALL_64_after_hwframe+0x44/0xa9 Freed by task 21: kasan_save_stack+0x1b/0x40 mm/kasan/common.c:48 kasan_set_track+0x1c/0x30 mm/kasan/common.c:56 kasan_set_free_info+0x1b/0x30 mm/kasan/generic.c:355 __kasan_slab_free+0xd8/0x120 mm/kasan/common.c:422 __cache_free mm/slab.c:3418 [inline] kfree+0x10e/0x2b0 mm/slab.c:3756 __io_wq_destroy fs/io-wq.c:1138 [inline] io_wq_destroy+0x2af/0x460 fs/io-wq.c:1146 io_finish_async fs/io_uring.c:6836 [inline] io_ring_ctx_free fs/io_uring.c:7870 [inline] io_ring_exit_work+0x1e4/0x6d0 fs/io_uring.c:7954 process_one_work+0x94c/0x1670 kernel/workqueue.c:2269 worker_thread+0x64c/0x1120 kernel/workqueue.c:2415 kthread+0x3b5/0x4a0 kernel/kthread.c:292 ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:294 The buggy address belongs to the object at ffff8882183db000 which belongs to the cache kmalloc-1k of size 1024 The buggy address is located 140 bytes inside of 1024-byte region [ffff8882183db000, ffff8882183db400) The buggy address belongs to the page: page:000000009bada22b refcount:1 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0x2183db flags: 0x57ffe0000000200(slab) raw: 057ffe0000000200 ffffea0008604c48 ffffea00086a8648 ffff8880aa040700 raw: 0000000000000000 ffff8882183db000 0000000100000002 0000000000000000 page dumped because: kasan: bad access detected Memory state around the buggy address: ffff8882183daf80: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ffff8882183db000: fa fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb >ffff8882183db080: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ^ ffff8882183db100: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ffff8882183db180: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ================================================================== which is down to the comment below, /* all workers gone, wq exit can proceed */ if (!nr_workers && refcount_dec_and_test(&wqe->wq->refs)) complete(&wqe->wq->done); because there might be multiple cases of wqe in a wq and we would wait for every worker in every wqe to go home before releasing wq's resources on destroying. To that end, rework wq's refcount by making it independent of the tracking of workers because after all they are two different things, and keeping it balanced when workers come and go. Note the manager kthread, like other workers, now holds a grab to wq during its lifetime. Finally to help destroy wq, check IO_WQ_BIT_EXIT upon creating worker and do nothing for exiting wq. Cc: stable@vger.kernel.org # v5.5+ Reported-by: syzbot+45fa0a195b941764e0f0@syzkaller.appspotmail.com Reported-by: syzbot+9af99580130003da82b1@syzkaller.appspotmail.com Cc: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Hillf Danton <hdanton@sina.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-26 13:26:55 +00:00
if (refcount_dec_and_test(&wqe->wq->refs))
complete(&wqe->wq->done);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
}
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
static inline bool io_wqe_run_queue(struct io_wqe *wqe)
__must_hold(wqe->lock)
{
io-wq: shrink io_wq_work a bit Currently we're using 40 bytes for the io_wq_work structure, and 16 of those is the doubly link list node. We don't need doubly linked lists, we always add to tail to keep things ordered, and any other use case is list traversal with deletion. For the deletion case, we can easily support any node deletion by keeping track of the previous entry. This shrinks io_wq_work to 32 bytes, and subsequently io_kiock from io_uring to 216 to 208 bytes. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-26 18:59:32 +00:00
if (!wq_list_empty(&wqe->work_list) &&
!(wqe->flags & IO_WQE_FLAG_STALLED))
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
return true;
return false;
}
/*
* Check head of free list for an available worker. If one isn't available,
* caller must wake up the wq manager to create one.
*/
static bool io_wqe_activate_free_worker(struct io_wqe *wqe)
__must_hold(RCU)
{
struct hlist_nulls_node *n;
struct io_worker *worker;
io-wq: remove now redundant struct io_wq_nulls_list Since we don't iterate these lists anymore after commit: e61df66c69b1 ("io-wq: ensure free/busy list browsing see all items") we don't need to retain the nulls value we use for them. That means it's pretty pointless to wrap the hlist_nulls_head in a structure, so get rid of it. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-14 15:00:41 +00:00
n = rcu_dereference(hlist_nulls_first_rcu(&wqe->free_list));
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
if (is_a_nulls(n))
return false;
worker = hlist_nulls_entry(n, struct io_worker, nulls_node);
if (io_worker_get(worker)) {
io-wq: remove worker->wait waitqueue We only have one cases of using the waitqueue to wake the worker, the rest are using wake_up_process(). Since we can save some cycles not fiddling with the waitqueue io_wqe_worker(), switch the work activation to task wakeup and get rid of the now unused wait_queue_head_t in struct io_worker. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-12-08 04:03:59 +00:00
wake_up_process(worker->task);
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
io_worker_release(worker);
return true;
}
return false;
}
/*
* We need a worker. If we find a free one, we're good. If not, and we're
* below the max number of workers, wake up the manager to create one.
*/
static void io_wqe_wake_worker(struct io_wqe *wqe, struct io_wqe_acct *acct)
{
bool ret;
/*
* Most likely an attempt to queue unbounded work on an io_wq that
* wasn't setup with any unbounded workers.
*/
WARN_ON_ONCE(!acct->max_workers);
rcu_read_lock();
ret = io_wqe_activate_free_worker(wqe);
rcu_read_unlock();
if (!ret && acct->nr_workers < acct->max_workers)
wake_up_process(wqe->wq->manager);
}
static void io_wqe_inc_running(struct io_wqe *wqe, struct io_worker *worker)
{
struct io_wqe_acct *acct = io_wqe_get_acct(wqe, worker);
atomic_inc(&acct->nr_running);
}
static void io_wqe_dec_running(struct io_wqe *wqe, struct io_worker *worker)
__must_hold(wqe->lock)
{
struct io_wqe_acct *acct = io_wqe_get_acct(wqe, worker);
if (atomic_dec_and_test(&acct->nr_running) && io_wqe_run_queue(wqe))
io_wqe_wake_worker(wqe, acct);
}
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
static void io_worker_start(struct io_wqe *wqe, struct io_worker *worker)
{
allow_kernel_signal(SIGINT);
current->flags |= PF_IO_WORKER;
worker->flags |= (IO_WORKER_F_UP | IO_WORKER_F_RUNNING);
io_uring: io_uring: add support for async work inheriting files This is in preparation for adding opcodes that need to add new files in a process file table, system calls like open(2) or accept4(2). If an opcode needs this, it must set IO_WQ_WORK_NEEDS_FILES in the work item. If work that needs to get punted to async context have this set, the async worker will assume the original task file table before executing the work. Note that opcodes that need access to the current files of an application cannot be done through IORING_SETUP_SQPOLL. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-24 18:39:47 +00:00
worker->restore_files = current->files;
io_uring: reference ->nsproxy for file table commands If we don't get and assign the namespace for the async work, then certain paths just don't work properly (like /dev/stdin, /proc/mounts, etc). Anything that references the current namespace of the given task should be assigned for async work on behalf of that task. Cc: stable@vger.kernel.org # v5.5+ Reported-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-19 02:13:06 +00:00
worker->restore_nsproxy = current->nsproxy;
io-wq: add support for inheriting ->fs Some work items need this for relative path lookup, make it available like the other inherited credentials/mm/etc. Cc: stable@vger.kernel.org # 5.3+ Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-02-07 04:42:51 +00:00
worker->restore_fs = current->fs;
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
io_wqe_inc_running(wqe, worker);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
}
/*
* Worker will start processing some work. Move it to the busy list, if
* it's currently on the freelist
*/
static void __io_worker_busy(struct io_wqe *wqe, struct io_worker *worker,
struct io_wq_work *work)
__must_hold(wqe->lock)
{
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
bool worker_bound, work_bound;
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
if (worker->flags & IO_WORKER_F_FREE) {
worker->flags &= ~IO_WORKER_F_FREE;
hlist_nulls_del_init_rcu(&worker->nulls_node);
}
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
/*
* If worker is moving from bound to unbound (or vice versa), then
* ensure we update the running accounting.
*/
io-wq: remove extra space characters These lines are indented an extra space character. Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-19 06:22:16 +00:00
worker_bound = (worker->flags & IO_WORKER_F_BOUND) != 0;
work_bound = (work->flags & IO_WQ_WORK_UNBOUND) == 0;
if (worker_bound != work_bound) {
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
io_wqe_dec_running(wqe, worker);
if (work_bound) {
worker->flags |= IO_WORKER_F_BOUND;
wqe->acct[IO_WQ_ACCT_UNBOUND].nr_workers--;
wqe->acct[IO_WQ_ACCT_BOUND].nr_workers++;
atomic_dec(&wqe->wq->user->processes);
} else {
worker->flags &= ~IO_WORKER_F_BOUND;
wqe->acct[IO_WQ_ACCT_UNBOUND].nr_workers++;
wqe->acct[IO_WQ_ACCT_BOUND].nr_workers--;
atomic_inc(&wqe->wq->user->processes);
}
io_wqe_inc_running(wqe, worker);
}
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
}
/*
* No work, worker going to sleep. Move to freelist, and unuse mm if we
* have one attached. Dropping the mm may potentially sleep, so we drop
* the lock in that case and return success. Since the caller has to
* retry the loop in that case (we changed task state), we don't regrab
* the lock if we return success.
*/
static bool __io_worker_idle(struct io_wqe *wqe, struct io_worker *worker)
__must_hold(wqe->lock)
{
if (!(worker->flags & IO_WORKER_F_FREE)) {
worker->flags |= IO_WORKER_F_FREE;
io-wq: remove now redundant struct io_wq_nulls_list Since we don't iterate these lists anymore after commit: e61df66c69b1 ("io-wq: ensure free/busy list browsing see all items") we don't need to retain the nulls value we use for them. That means it's pretty pointless to wrap the hlist_nulls_head in a structure, so get rid of it. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-14 15:00:41 +00:00
hlist_nulls_add_head_rcu(&worker->nulls_node, &wqe->free_list);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
}
return __io_worker_unuse(wqe, worker);
}
io-wq: hash dependent work Enable io-wq hashing stuff for dependent works simply by re-enqueueing such requests. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-13 21:31:05 +00:00
static inline unsigned int io_get_work_hash(struct io_wq_work *work)
{
return work->flags >> IO_WQ_HASH_SHIFT;
}
static struct io_wq_work *io_get_next_work(struct io_wqe *wqe)
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
__must_hold(wqe->lock)
{
io-wq: shrink io_wq_work a bit Currently we're using 40 bytes for the io_wq_work structure, and 16 of those is the doubly link list node. We don't need doubly linked lists, we always add to tail to keep things ordered, and any other use case is list traversal with deletion. For the deletion case, we can easily support any node deletion by keeping track of the previous entry. This shrinks io_wq_work to 32 bytes, and subsequently io_kiock from io_uring to 216 to 208 bytes. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-26 18:59:32 +00:00
struct io_wq_work_node *node, *prev;
io-wq: handle hashed writes in chains We always punt async buffered writes to an io-wq helper, as the core kernel does not have IOCB_NOWAIT support for that. Most buffered async writes complete very quickly, as it's just a copy operation. This means that doing multiple locking roundtrips on the shared wqe lock for each buffered write is wasteful. Additionally, buffered writes are hashed work items, which means that any buffered write to a given file is serialized. Keep identicaly hashed work items contiguously in @wqe->work_list, and track a tail for each hash bucket. On dequeue of a hashed item, splice all of the same hash in one go using the tracked tail. Until the batch is done, the caller doesn't have to synchronize with the wqe or worker locks again. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-23 19:57:22 +00:00
struct io_wq_work *work, *tail;
io-wq: hash dependent work Enable io-wq hashing stuff for dependent works simply by re-enqueueing such requests. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-13 21:31:05 +00:00
unsigned int hash;
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
io-wq: shrink io_wq_work a bit Currently we're using 40 bytes for the io_wq_work structure, and 16 of those is the doubly link list node. We don't need doubly linked lists, we always add to tail to keep things ordered, and any other use case is list traversal with deletion. For the deletion case, we can easily support any node deletion by keeping track of the previous entry. This shrinks io_wq_work to 32 bytes, and subsequently io_kiock from io_uring to 216 to 208 bytes. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-26 18:59:32 +00:00
wq_list_for_each(node, prev, &wqe->work_list) {
work = container_of(node, struct io_wq_work, list);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
/* not hashed, can run anytime */
io-wq: split hashing and enqueueing It's a preparation patch removing io_wq_enqueue_hashed(), which now should be done by io_wq_hash_work() + io_wq_enqueue(). Also, set hash value for dependant works, and do it as late as possible, because req->file can be unavailable before. This hash will be ignored by io-wq. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-13 21:31:04 +00:00
if (!io_wq_is_hashed(work)) {
io-wq: handle hashed writes in chains We always punt async buffered writes to an io-wq helper, as the core kernel does not have IOCB_NOWAIT support for that. Most buffered async writes complete very quickly, as it's just a copy operation. This means that doing multiple locking roundtrips on the shared wqe lock for each buffered write is wasteful. Additionally, buffered writes are hashed work items, which means that any buffered write to a given file is serialized. Keep identicaly hashed work items contiguously in @wqe->work_list, and track a tail for each hash bucket. On dequeue of a hashed item, splice all of the same hash in one go using the tracked tail. Until the batch is done, the caller doesn't have to synchronize with the wqe or worker locks again. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-23 19:57:22 +00:00
wq_list_del(&wqe->work_list, node, prev);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
return work;
}
/* hashed, can run if not already running */
io-wq: hash dependent work Enable io-wq hashing stuff for dependent works simply by re-enqueueing such requests. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-13 21:31:05 +00:00
hash = io_get_work_hash(work);
if (!(wqe->hash_map & BIT(hash))) {
wqe->hash_map |= BIT(hash);
io-wq: handle hashed writes in chains We always punt async buffered writes to an io-wq helper, as the core kernel does not have IOCB_NOWAIT support for that. Most buffered async writes complete very quickly, as it's just a copy operation. This means that doing multiple locking roundtrips on the shared wqe lock for each buffered write is wasteful. Additionally, buffered writes are hashed work items, which means that any buffered write to a given file is serialized. Keep identicaly hashed work items contiguously in @wqe->work_list, and track a tail for each hash bucket. On dequeue of a hashed item, splice all of the same hash in one go using the tracked tail. Until the batch is done, the caller doesn't have to synchronize with the wqe or worker locks again. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-23 19:57:22 +00:00
/* all items with this hash lie in [work, tail] */
tail = wqe->hash_tail[hash];
wqe->hash_tail[hash] = NULL;
wq_list_cut(&wqe->work_list, &tail->list, prev);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
return work;
}
}
return NULL;
}
io_uring/io-wq: don't use static creds/mm assignments We currently setup the io_wq with a static set of mm and creds. Even for a single-use io-wq per io_uring, this is suboptimal as we have may have multiple enters of the ring. For sharing the io-wq backend, it doesn't work at all. Switch to passing in the creds and mm when the work item is setup. This means that async work is no longer deferred to the io_uring mm and creds, it is done with the current mm and creds. Flag this behavior with IORING_FEAT_CUR_PERSONALITY, so applications know they can rely on the current personality (mm and creds) being the same for direct issue and async issue. Reviewed-by: Stefan Metzmacher <metze@samba.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-01-27 23:34:48 +00:00
static void io_wq_switch_mm(struct io_worker *worker, struct io_wq_work *work)
{
if (worker->mm) {
kernel: better document the use_mm/unuse_mm API contract Switch the function documentation to kerneldoc comments, and add WARN_ON_ONCE asserts that the calling thread is a kernel thread and does not have ->mm set (or has ->mm set in the case of unuse_mm). Also give the functions a kthread_ prefix to better document the use case. [hch@lst.de: fix a comment typo, cover the newly merged use_mm/unuse_mm caller in vfio] Link: http://lkml.kernel.org/r/20200416053158.586887-3-hch@lst.de [sfr@canb.auug.org.au: powerpc/vas: fix up for {un}use_mm() rename] Link: http://lkml.kernel.org/r/20200422163935.5aa93ba5@canb.auug.org.au Signed-off-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Tested-by: Jens Axboe <axboe@kernel.dk> Reviewed-by: Jens Axboe <axboe@kernel.dk> Acked-by: Felix Kuehling <Felix.Kuehling@amd.com> Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> [usb] Acked-by: Haren Myneni <haren@linux.ibm.com> Cc: Alex Deucher <alexander.deucher@amd.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Felipe Balbi <balbi@kernel.org> Cc: Jason Wang <jasowang@redhat.com> Cc: "Michael S. Tsirkin" <mst@redhat.com> Cc: Zhenyu Wang <zhenyuw@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Link: http://lkml.kernel.org/r/20200404094101.672954-6-hch@lst.de Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-06-11 01:42:06 +00:00
kthread_unuse_mm(worker->mm);
io_uring/io-wq: don't use static creds/mm assignments We currently setup the io_wq with a static set of mm and creds. Even for a single-use io-wq per io_uring, this is suboptimal as we have may have multiple enters of the ring. For sharing the io-wq backend, it doesn't work at all. Switch to passing in the creds and mm when the work item is setup. This means that async work is no longer deferred to the io_uring mm and creds, it is done with the current mm and creds. Flag this behavior with IORING_FEAT_CUR_PERSONALITY, so applications know they can rely on the current personality (mm and creds) being the same for direct issue and async issue. Reviewed-by: Stefan Metzmacher <metze@samba.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-01-27 23:34:48 +00:00
mmput(worker->mm);
worker->mm = NULL;
}
kernel: set USER_DS in kthread_use_mm Some architectures like arm64 and s390 require USER_DS to be set for kernel threads to access user address space, which is the whole purpose of kthread_use_mm, but other like x86 don't. That has lead to a huge mess where some callers are fixed up once they are tested on said architectures, while others linger around and yet other like io_uring try to do "clever" optimizations for what usually is just a trivial asignment to a member in the thread_struct for most architectures. Make kthread_use_mm set USER_DS, and kthread_unuse_mm restore to the previous value instead. Signed-off-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Tested-by: Jens Axboe <axboe@kernel.dk> Reviewed-by: Jens Axboe <axboe@kernel.dk> Acked-by: Michael S. Tsirkin <mst@redhat.com> Cc: Alex Deucher <alexander.deucher@amd.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Felipe Balbi <balbi@kernel.org> Cc: Felix Kuehling <Felix.Kuehling@amd.com> Cc: Jason Wang <jasowang@redhat.com> Cc: Zhenyu Wang <zhenyuw@linux.intel.com> Cc: Zhi Wang <zhi.a.wang@intel.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Link: http://lkml.kernel.org/r/20200404094101.672954-7-hch@lst.de Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2020-06-11 01:42:10 +00:00
io_uring: move io identity items into separate struct io-wq contains a pointer to the identity, which we just hold in io_kiocb for now. This is in preparation for putting this outside io_kiocb. The only exception is struct files_struct, which we'll need different rules for to avoid a circular dependency. No functional changes in this patch. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-10-14 16:48:51 +00:00
if (mmget_not_zero(work->identity->mm)) {
kthread_use_mm(work->identity->mm);
worker->mm = work->identity->mm;
io_uring/io-wq: don't use static creds/mm assignments We currently setup the io_wq with a static set of mm and creds. Even for a single-use io-wq per io_uring, this is suboptimal as we have may have multiple enters of the ring. For sharing the io-wq backend, it doesn't work at all. Switch to passing in the creds and mm when the work item is setup. This means that async work is no longer deferred to the io_uring mm and creds, it is done with the current mm and creds. Flag this behavior with IORING_FEAT_CUR_PERSONALITY, so applications know they can rely on the current personality (mm and creds) being the same for direct issue and async issue. Reviewed-by: Stefan Metzmacher <metze@samba.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-01-27 23:34:48 +00:00
return;
}
/* failed grabbing mm, ensure work gets cancelled */
work->flags |= IO_WQ_WORK_CANCEL;
}
io_uring: add blkcg accounting to offloaded operations There are a few operations that are offloaded to the worker threads. In this case, we lose process context and end up in kthread context. This results in ios to be not accounted to the issuing cgroup and consequently end up as issued by root. Just like others, adopt the personality of the blkcg too when issuing via the workqueues. For the SQPOLL thread, it will live and attach in the inited cgroup's context. Signed-off-by: Dennis Zhou <dennis@kernel.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-16 20:41:05 +00:00
static inline void io_wq_switch_blkcg(struct io_worker *worker,
struct io_wq_work *work)
{
#ifdef CONFIG_BLK_CGROUP
io_uring: pass required context in as flags We have a number of bits that decide what context to inherit. Set up io-wq flags for these instead. This is in preparation for always having the various members set, but not always needing them for all requests. No intended functional changes in this patch. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-10-14 15:23:55 +00:00
if (!(work->flags & IO_WQ_WORK_BLKCG))
return;
io_uring: move io identity items into separate struct io-wq contains a pointer to the identity, which we just hold in io_kiocb for now. This is in preparation for putting this outside io_kiocb. The only exception is struct files_struct, which we'll need different rules for to avoid a circular dependency. No functional changes in this patch. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-10-14 16:48:51 +00:00
if (work->identity->blkcg_css != worker->blkcg_css) {
kthread_associate_blkcg(work->identity->blkcg_css);
worker->blkcg_css = work->identity->blkcg_css;
io_uring: add blkcg accounting to offloaded operations There are a few operations that are offloaded to the worker threads. In this case, we lose process context and end up in kthread context. This results in ios to be not accounted to the issuing cgroup and consequently end up as issued by root. Just like others, adopt the personality of the blkcg too when issuing via the workqueues. For the SQPOLL thread, it will live and attach in the inited cgroup's context. Signed-off-by: Dennis Zhou <dennis@kernel.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-16 20:41:05 +00:00
}
#endif
}
io_uring/io-wq: don't use static creds/mm assignments We currently setup the io_wq with a static set of mm and creds. Even for a single-use io-wq per io_uring, this is suboptimal as we have may have multiple enters of the ring. For sharing the io-wq backend, it doesn't work at all. Switch to passing in the creds and mm when the work item is setup. This means that async work is no longer deferred to the io_uring mm and creds, it is done with the current mm and creds. Flag this behavior with IORING_FEAT_CUR_PERSONALITY, so applications know they can rely on the current personality (mm and creds) being the same for direct issue and async issue. Reviewed-by: Stefan Metzmacher <metze@samba.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-01-27 23:34:48 +00:00
static void io_wq_switch_creds(struct io_worker *worker,
struct io_wq_work *work)
{
io_uring: move io identity items into separate struct io-wq contains a pointer to the identity, which we just hold in io_kiocb for now. This is in preparation for putting this outside io_kiocb. The only exception is struct files_struct, which we'll need different rules for to avoid a circular dependency. No functional changes in this patch. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-10-14 16:48:51 +00:00
const struct cred *old_creds = override_creds(work->identity->creds);
io_uring/io-wq: don't use static creds/mm assignments We currently setup the io_wq with a static set of mm and creds. Even for a single-use io-wq per io_uring, this is suboptimal as we have may have multiple enters of the ring. For sharing the io-wq backend, it doesn't work at all. Switch to passing in the creds and mm when the work item is setup. This means that async work is no longer deferred to the io_uring mm and creds, it is done with the current mm and creds. Flag this behavior with IORING_FEAT_CUR_PERSONALITY, so applications know they can rely on the current personality (mm and creds) being the same for direct issue and async issue. Reviewed-by: Stefan Metzmacher <metze@samba.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-01-27 23:34:48 +00:00
io_uring: move io identity items into separate struct io-wq contains a pointer to the identity, which we just hold in io_kiocb for now. This is in preparation for putting this outside io_kiocb. The only exception is struct files_struct, which we'll need different rules for to avoid a circular dependency. No functional changes in this patch. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-10-14 16:48:51 +00:00
worker->cur_creds = work->identity->creds;
io_uring/io-wq: don't use static creds/mm assignments We currently setup the io_wq with a static set of mm and creds. Even for a single-use io-wq per io_uring, this is suboptimal as we have may have multiple enters of the ring. For sharing the io-wq backend, it doesn't work at all. Switch to passing in the creds and mm when the work item is setup. This means that async work is no longer deferred to the io_uring mm and creds, it is done with the current mm and creds. Flag this behavior with IORING_FEAT_CUR_PERSONALITY, so applications know they can rely on the current personality (mm and creds) being the same for direct issue and async issue. Reviewed-by: Stefan Metzmacher <metze@samba.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-01-27 23:34:48 +00:00
if (worker->saved_creds)
put_cred(old_creds); /* creds set by previous switch */
else
worker->saved_creds = old_creds;
}
io-wq: shuffle io_worker_handle_work() code This is a preparation patch, it adds some helpers and makes the next patches cleaner. - extract io_impersonate_work() and io_assign_current_work() - replace @next label with nested do-while - move put_work() right after NULL'ing cur_work. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-04 13:14:09 +00:00
static void io_impersonate_work(struct io_worker *worker,
struct io_wq_work *work)
{
io_uring: move io identity items into separate struct io-wq contains a pointer to the identity, which we just hold in io_kiocb for now. This is in preparation for putting this outside io_kiocb. The only exception is struct files_struct, which we'll need different rules for to avoid a circular dependency. No functional changes in this patch. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-10-14 16:48:51 +00:00
if ((work->flags & IO_WQ_WORK_FILES) &&
current->files != work->identity->files) {
io-wq: shuffle io_worker_handle_work() code This is a preparation patch, it adds some helpers and makes the next patches cleaner. - extract io_impersonate_work() and io_assign_current_work() - replace @next label with nested do-while - move put_work() right after NULL'ing cur_work. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-04 13:14:09 +00:00
task_lock(current);
io_uring: move io identity items into separate struct io-wq contains a pointer to the identity, which we just hold in io_kiocb for now. This is in preparation for putting this outside io_kiocb. The only exception is struct files_struct, which we'll need different rules for to avoid a circular dependency. No functional changes in this patch. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-10-14 16:48:51 +00:00
current->files = work->identity->files;
current->nsproxy = work->identity->nsproxy;
io-wq: shuffle io_worker_handle_work() code This is a preparation patch, it adds some helpers and makes the next patches cleaner. - extract io_impersonate_work() and io_assign_current_work() - replace @next label with nested do-while - move put_work() right after NULL'ing cur_work. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-04 13:14:09 +00:00
task_unlock(current);
io-wq: cancel request if it's asking for files and we don't have them This can't currently happen, but will be possible shortly. Handle missing files just like we do not being able to grab a needed mm, and mark the request as needing cancelation. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-10-30 15:36:41 +00:00
if (!work->identity->files) {
/* failed grabbing files, ensure work gets cancelled */
work->flags |= IO_WQ_WORK_CANCEL;
}
io-wq: shuffle io_worker_handle_work() code This is a preparation patch, it adds some helpers and makes the next patches cleaner. - extract io_impersonate_work() and io_assign_current_work() - replace @next label with nested do-while - move put_work() right after NULL'ing cur_work. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-04 13:14:09 +00:00
}
io_uring: move io identity items into separate struct io-wq contains a pointer to the identity, which we just hold in io_kiocb for now. This is in preparation for putting this outside io_kiocb. The only exception is struct files_struct, which we'll need different rules for to avoid a circular dependency. No functional changes in this patch. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-10-14 16:48:51 +00:00
if ((work->flags & IO_WQ_WORK_FS) && current->fs != work->identity->fs)
current->fs = work->identity->fs;
if ((work->flags & IO_WQ_WORK_MM) && work->identity->mm != worker->mm)
io-wq: shuffle io_worker_handle_work() code This is a preparation patch, it adds some helpers and makes the next patches cleaner. - extract io_impersonate_work() and io_assign_current_work() - replace @next label with nested do-while - move put_work() right after NULL'ing cur_work. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-04 13:14:09 +00:00
io_wq_switch_mm(worker, work);
io_uring: move io identity items into separate struct io-wq contains a pointer to the identity, which we just hold in io_kiocb for now. This is in preparation for putting this outside io_kiocb. The only exception is struct files_struct, which we'll need different rules for to avoid a circular dependency. No functional changes in this patch. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-10-14 16:48:51 +00:00
if ((work->flags & IO_WQ_WORK_CREDS) &&
worker->cur_creds != work->identity->creds)
io-wq: shuffle io_worker_handle_work() code This is a preparation patch, it adds some helpers and makes the next patches cleaner. - extract io_impersonate_work() and io_assign_current_work() - replace @next label with nested do-while - move put_work() right after NULL'ing cur_work. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-04 13:14:09 +00:00
io_wq_switch_creds(worker, work);
io_uring: unify fsize with def->work_flags This one was missed in the earlier conversion, should be included like any of the other IO identity flags. Make sure we restore to RLIM_INIFITY when dropping the personality again. Fixes: 98447d65b4a7 ("io_uring: move io identity items into separate struct") Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-10-20 20:28:41 +00:00
if (work->flags & IO_WQ_WORK_FSIZE)
current->signal->rlim[RLIMIT_FSIZE].rlim_cur = work->identity->fsize;
else if (current->signal->rlim[RLIMIT_FSIZE].rlim_cur != RLIM_INFINITY)
current->signal->rlim[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY;
io_uring: add blkcg accounting to offloaded operations There are a few operations that are offloaded to the worker threads. In this case, we lose process context and end up in kthread context. This results in ios to be not accounted to the issuing cgroup and consequently end up as issued by root. Just like others, adopt the personality of the blkcg too when issuing via the workqueues. For the SQPOLL thread, it will live and attach in the inited cgroup's context. Signed-off-by: Dennis Zhou <dennis@kernel.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-16 20:41:05 +00:00
io_wq_switch_blkcg(worker, work);
io-wq: inherit audit loginuid and sessionid Make sure the async io-wq workers inherit the loginuid and sessionid from the original task, and restore them to unset once we're done with the async work item. While at it, disable the ability for kernel threads to write to their own loginuid. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-10-15 19:46:44 +00:00
#ifdef CONFIG_AUDIT
current->loginuid = work->identity->loginuid;
current->sessionid = work->identity->sessionid;
#endif
io-wq: shuffle io_worker_handle_work() code This is a preparation patch, it adds some helpers and makes the next patches cleaner. - extract io_impersonate_work() and io_assign_current_work() - replace @next label with nested do-while - move put_work() right after NULL'ing cur_work. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-04 13:14:09 +00:00
}
static void io_assign_current_work(struct io_worker *worker,
struct io_wq_work *work)
{
io-wq: don't resched if there is no work This little tweak restores the behaviour that was before the recent io_worker_handle_work() optimisation patches. It makes the function do cond_resched() and flush_signals() only if there is an actual work to execute. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-13 21:31:03 +00:00
if (work) {
/* flush pending signals before assigning new work */
if (signal_pending(current))
flush_signals(current);
cond_resched();
}
io-wq: shuffle io_worker_handle_work() code This is a preparation patch, it adds some helpers and makes the next patches cleaner. - extract io_impersonate_work() and io_assign_current_work() - replace @next label with nested do-while - move put_work() right after NULL'ing cur_work. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-04 13:14:09 +00:00
io-wq: inherit audit loginuid and sessionid Make sure the async io-wq workers inherit the loginuid and sessionid from the original task, and restore them to unset once we're done with the async work item. While at it, disable the ability for kernel threads to write to their own loginuid. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-10-15 19:46:44 +00:00
#ifdef CONFIG_AUDIT
current->loginuid = KUIDT_INIT(AUDIT_UID_UNSET);
current->sessionid = AUDIT_SID_UNSET;
#endif
io-wq: shuffle io_worker_handle_work() code This is a preparation patch, it adds some helpers and makes the next patches cleaner. - extract io_impersonate_work() and io_assign_current_work() - replace @next label with nested do-while - move put_work() right after NULL'ing cur_work. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-04 13:14:09 +00:00
spin_lock_irq(&worker->lock);
worker->cur_work = work;
spin_unlock_irq(&worker->lock);
}
io-wq: hash dependent work Enable io-wq hashing stuff for dependent works simply by re-enqueueing such requests. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-13 21:31:05 +00:00
static void io_wqe_enqueue(struct io_wqe *wqe, struct io_wq_work *work);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
static void io_worker_handle_work(struct io_worker *worker)
__releases(wqe->lock)
{
struct io_wqe *wqe = worker->wqe;
struct io_wq *wq = wqe->wq;
do {
io-wq: handle hashed writes in chains We always punt async buffered writes to an io-wq helper, as the core kernel does not have IOCB_NOWAIT support for that. Most buffered async writes complete very quickly, as it's just a copy operation. This means that doing multiple locking roundtrips on the shared wqe lock for each buffered write is wasteful. Additionally, buffered writes are hashed work items, which means that any buffered write to a given file is serialized. Keep identicaly hashed work items contiguously in @wqe->work_list, and track a tail for each hash bucket. On dequeue of a hashed item, splice all of the same hash in one go using the tracked tail. Until the batch is done, the caller doesn't have to synchronize with the wqe or worker locks again. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-23 19:57:22 +00:00
struct io_wq_work *work;
io-wq: optimise out *next_work() double lock When executing non-linked hashed work, io_worker_handle_work() will lock-unlock wqe->lock to update hash, and then immediately lock-unlock to get next work. Optimise this case and do lock/unlock only once. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-04 13:14:11 +00:00
get_next:
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
/*
* If we got some work, mark us as busy. If we didn't, but
* the list isn't empty, it means we stalled on hashed work.
* Mark us stalled so we don't keep looking for work when we
* can't make progress, any work completion or insertion will
* clear the stalled flag.
*/
io-wq: hash dependent work Enable io-wq hashing stuff for dependent works simply by re-enqueueing such requests. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-13 21:31:05 +00:00
work = io_get_next_work(wqe);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
if (work)
__io_worker_busy(wqe, worker, work);
io-wq: shrink io_wq_work a bit Currently we're using 40 bytes for the io_wq_work structure, and 16 of those is the doubly link list node. We don't need doubly linked lists, we always add to tail to keep things ordered, and any other use case is list traversal with deletion. For the deletion case, we can easily support any node deletion by keeping track of the previous entry. This shrinks io_wq_work to 32 bytes, and subsequently io_kiock from io_uring to 216 to 208 bytes. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-26 18:59:32 +00:00
else if (!wq_list_empty(&wqe->work_list))
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
wqe->flags |= IO_WQE_FLAG_STALLED;
io_wq: Make io_wqe::lock a raw_spinlock_t During a context switch the scheduler invokes wq_worker_sleeping() with disabled preemption. Disabling preemption is needed because it protects access to `worker->sleeping'. As an optimisation it avoids invoking schedule() within the schedule path as part of possible wake up (thus preempt_enable_no_resched() afterwards). The io-wq has been added to the mix in the same section with disabled preemption. This breaks on PREEMPT_RT because io_wq_worker_sleeping() acquires a spinlock_t. Also within the schedule() the spinlock_t must be acquired after tsk_is_pi_blocked() otherwise it will block on the sleeping lock again while scheduling out. While playing with `io_uring-bench' I didn't notice a significant latency spike after converting io_wqe::lock to a raw_spinlock_t. The latency was more or less the same. In order to keep the spinlock_t it would have to be moved after the tsk_is_pi_blocked() check which would introduce a branch instruction into the hot path. The lock is used to maintain the `work_list' and wakes one task up at most. Should io_wqe_cancel_pending_work() cause latency spikes, while searching for a specific item, then it would need to drop the lock during iterations. revert_creds() is also invoked under the lock. According to debug cred::non_rcu is 0. Otherwise it should be moved outside of the locked section because put_cred_rcu()->free_uid() acquires a sleeping lock. Convert io_wqe::lock to a raw_spinlock_t.c Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-01 08:41:46 +00:00
raw_spin_unlock_irq(&wqe->lock);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
if (!work)
break;
io-wq: optimise locking in io_worker_handle_work() There are 2 optimisations: - Now, io_worker_handler_work() do io_assign_current_work() twice per request, and each one adds lock/unlock(worker->lock) pair. The first is to reset worker->cur_work to NULL, and the second to set a real work shortly after. If there is a dependant work, set it immediately, that effectively removes the extra NULL'ing. - And there is no use in taking wqe->lock for linked works, as they are not hashed now. Optimise it out. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-04 13:14:10 +00:00
io_assign_current_work(worker, work);
io-wq: ensure we have a stable view of ->cur_work for cancellations worker->cur_work is currently protected by the lock of the wqe that the worker belongs to. When we send a signal to a worker, we need a stable view of ->cur_work, so we need to hold that lock. But this doesn't work so well, since we have the opposite order potentially on queueing work. If POLL_ADD is used with a signalfd, then io_poll_wake() is called with the signal lock, and that sometimes needs to insert work items. Add a specific worker lock that protects the current work item. Then we can guarantee that the task we're sending a signal is currently processing the exact work we think it is. Reported-by: Paul E. McKenney <paulmck@kernel.org> Reviewed-by: Paul E. McKenney <paulmck@kernel.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-13 16:43:34 +00:00
io-wq: shuffle io_worker_handle_work() code This is a preparation patch, it adds some helpers and makes the next patches cleaner. - extract io_impersonate_work() and io_assign_current_work() - replace @next label with nested do-while - move put_work() right after NULL'ing cur_work. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-04 13:14:09 +00:00
/* handle a whole dependent link */
do {
io-wq: handle hashed writes in chains We always punt async buffered writes to an io-wq helper, as the core kernel does not have IOCB_NOWAIT support for that. Most buffered async writes complete very quickly, as it's just a copy operation. This means that doing multiple locking roundtrips on the shared wqe lock for each buffered write is wasteful. Additionally, buffered writes are hashed work items, which means that any buffered write to a given file is serialized. Keep identicaly hashed work items contiguously in @wqe->work_list, and track a tail for each hash bucket. On dequeue of a hashed item, splice all of the same hash in one go using the tracked tail. Until the batch is done, the caller doesn't have to synchronize with the wqe or worker locks again. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-23 19:57:22 +00:00
struct io_wq_work *old_work, *next_hashed, *linked;
io-wq: update hash bits Linked requests are hashed, remove a comment stating otherwise. Also move hash bits to emphasise that we don't carry it through loop iteration and set it every time. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-07-25 11:42:00 +00:00
unsigned int hash = io_get_work_hash(work);
io-wq: shuffle io_worker_handle_work() code This is a preparation patch, it adds some helpers and makes the next patches cleaner. - extract io_impersonate_work() and io_assign_current_work() - replace @next label with nested do-while - move put_work() right after NULL'ing cur_work. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-04 13:14:09 +00:00
io-wq: handle hashed writes in chains We always punt async buffered writes to an io-wq helper, as the core kernel does not have IOCB_NOWAIT support for that. Most buffered async writes complete very quickly, as it's just a copy operation. This means that doing multiple locking roundtrips on the shared wqe lock for each buffered write is wasteful. Additionally, buffered writes are hashed work items, which means that any buffered write to a given file is serialized. Keep identicaly hashed work items contiguously in @wqe->work_list, and track a tail for each hash bucket. On dequeue of a hashed item, splice all of the same hash in one go using the tracked tail. Until the batch is done, the caller doesn't have to synchronize with the wqe or worker locks again. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-23 19:57:22 +00:00
next_hashed = wq_next_work(work);
io-wq: optimise locking in io_worker_handle_work() There are 2 optimisations: - Now, io_worker_handler_work() do io_assign_current_work() twice per request, and each one adds lock/unlock(worker->lock) pair. The first is to reset worker->cur_work to NULL, and the second to set a real work shortly after. If there is a dependant work, set it immediately, that effectively removes the extra NULL'ing. - And there is no use in taking wqe->lock for linked works, as they are not hashed now. Optimise it out. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-04 13:14:10 +00:00
io_impersonate_work(worker, work);
io-wq: shuffle io_worker_handle_work() code This is a preparation patch, it adds some helpers and makes the next patches cleaner. - extract io_impersonate_work() and io_assign_current_work() - replace @next label with nested do-while - move put_work() right after NULL'ing cur_work. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-04 13:14:09 +00:00
io-wq: return next work from ->do_work() directly It's easier to return next work from ->do_work() than having an in-out argument. Looks nicer and easier to compile. Also, merge io_wq_assign_next() into its only user. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-06-25 15:20:54 +00:00
old_work = work;
linked = wq->do_work(work);
io-wq: handle hashed writes in chains We always punt async buffered writes to an io-wq helper, as the core kernel does not have IOCB_NOWAIT support for that. Most buffered async writes complete very quickly, as it's just a copy operation. This means that doing multiple locking roundtrips on the shared wqe lock for each buffered write is wasteful. Additionally, buffered writes are hashed work items, which means that any buffered write to a given file is serialized. Keep identicaly hashed work items contiguously in @wqe->work_list, and track a tail for each hash bucket. On dequeue of a hashed item, splice all of the same hash in one go using the tracked tail. Until the batch is done, the caller doesn't have to synchronize with the wqe or worker locks again. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-23 19:57:22 +00:00
work = next_hashed;
if (!work && linked && !io_wq_is_hashed(linked)) {
work = linked;
linked = NULL;
}
io_assign_current_work(worker, work);
io_uring/io-wq: forward submission ref to async First it changes io-wq interfaces. It replaces {get,put}_work() with free_work(), which guaranteed to be called exactly once. It also enforces free_work() callback to be non-NULL. io_uring follows the changes and instead of putting a submission reference in io_put_req_async_completion(), it will be done in io_free_work(). As removes io_get_work() with corresponding refcount_inc(), the ref balance is maintained. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-04 13:14:12 +00:00
wq->free_work(old_work);
io-wq: shuffle io_worker_handle_work() code This is a preparation patch, it adds some helpers and makes the next patches cleaner. - extract io_impersonate_work() and io_assign_current_work() - replace @next label with nested do-while - move put_work() right after NULL'ing cur_work. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-04 13:14:09 +00:00
io-wq: handle hashed writes in chains We always punt async buffered writes to an io-wq helper, as the core kernel does not have IOCB_NOWAIT support for that. Most buffered async writes complete very quickly, as it's just a copy operation. This means that doing multiple locking roundtrips on the shared wqe lock for each buffered write is wasteful. Additionally, buffered writes are hashed work items, which means that any buffered write to a given file is serialized. Keep identicaly hashed work items contiguously in @wqe->work_list, and track a tail for each hash bucket. On dequeue of a hashed item, splice all of the same hash in one go using the tracked tail. Until the batch is done, the caller doesn't have to synchronize with the wqe or worker locks again. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-23 19:57:22 +00:00
if (linked)
io_wqe_enqueue(wqe, linked);
if (hash != -1U && !next_hashed) {
io_wq: Make io_wqe::lock a raw_spinlock_t During a context switch the scheduler invokes wq_worker_sleeping() with disabled preemption. Disabling preemption is needed because it protects access to `worker->sleeping'. As an optimisation it avoids invoking schedule() within the schedule path as part of possible wake up (thus preempt_enable_no_resched() afterwards). The io-wq has been added to the mix in the same section with disabled preemption. This breaks on PREEMPT_RT because io_wq_worker_sleeping() acquires a spinlock_t. Also within the schedule() the spinlock_t must be acquired after tsk_is_pi_blocked() otherwise it will block on the sleeping lock again while scheduling out. While playing with `io_uring-bench' I didn't notice a significant latency spike after converting io_wqe::lock to a raw_spinlock_t. The latency was more or less the same. In order to keep the spinlock_t it would have to be moved after the tsk_is_pi_blocked() check which would introduce a branch instruction into the hot path. The lock is used to maintain the `work_list' and wakes one task up at most. Should io_wqe_cancel_pending_work() cause latency spikes, while searching for a specific item, then it would need to drop the lock during iterations. revert_creds() is also invoked under the lock. According to debug cred::non_rcu is 0. Otherwise it should be moved outside of the locked section because put_cred_rcu()->free_uid() acquires a sleeping lock. Convert io_wqe::lock to a raw_spinlock_t.c Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-01 08:41:46 +00:00
raw_spin_lock_irq(&wqe->lock);
io-wq: shuffle io_worker_handle_work() code This is a preparation patch, it adds some helpers and makes the next patches cleaner. - extract io_impersonate_work() and io_assign_current_work() - replace @next label with nested do-while - move put_work() right after NULL'ing cur_work. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-04 13:14:09 +00:00
wqe->hash_map &= ~BIT_ULL(hash);
wqe->flags &= ~IO_WQE_FLAG_STALLED;
io-wq: optimise out *next_work() double lock When executing non-linked hashed work, io_worker_handle_work() will lock-unlock wqe->lock to update hash, and then immediately lock-unlock to get next work. Optimise this case and do lock/unlock only once. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-04 13:14:11 +00:00
/* skip unnecessary unlock-lock wqe->lock */
if (!work)
goto get_next;
io_wq: Make io_wqe::lock a raw_spinlock_t During a context switch the scheduler invokes wq_worker_sleeping() with disabled preemption. Disabling preemption is needed because it protects access to `worker->sleeping'. As an optimisation it avoids invoking schedule() within the schedule path as part of possible wake up (thus preempt_enable_no_resched() afterwards). The io-wq has been added to the mix in the same section with disabled preemption. This breaks on PREEMPT_RT because io_wq_worker_sleeping() acquires a spinlock_t. Also within the schedule() the spinlock_t must be acquired after tsk_is_pi_blocked() otherwise it will block on the sleeping lock again while scheduling out. While playing with `io_uring-bench' I didn't notice a significant latency spike after converting io_wqe::lock to a raw_spinlock_t. The latency was more or less the same. In order to keep the spinlock_t it would have to be moved after the tsk_is_pi_blocked() check which would introduce a branch instruction into the hot path. The lock is used to maintain the `work_list' and wakes one task up at most. Should io_wqe_cancel_pending_work() cause latency spikes, while searching for a specific item, then it would need to drop the lock during iterations. revert_creds() is also invoked under the lock. According to debug cred::non_rcu is 0. Otherwise it should be moved outside of the locked section because put_cred_rcu()->free_uid() acquires a sleeping lock. Convert io_wqe::lock to a raw_spinlock_t.c Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-01 08:41:46 +00:00
raw_spin_unlock_irq(&wqe->lock);
io_wq: add get/put_work handlers to io_wq_create() For cancellation, we need to ensure that the work item stays valid for as long as ->cur_work is valid. Right now we can't safely dereference the work item even under the wqe->lock, because while the ->cur_work pointer will remain valid, the work could be completing and be freed in parallel. Only invoke ->get/put_work() on items we know that the caller queued themselves. Add IO_WQ_WORK_INTERNAL for io-wq to use, which is needed when we're queueing a flush item, for instance. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-13 05:31:31 +00:00
}
io-wq: optimise locking in io_worker_handle_work() There are 2 optimisations: - Now, io_worker_handler_work() do io_assign_current_work() twice per request, and each one adds lock/unlock(worker->lock) pair. The first is to reset worker->cur_work to NULL, and the second to set a real work shortly after. If there is a dependant work, set it immediately, that effectively removes the extra NULL'ing. - And there is no use in taking wqe->lock for linked works, as they are not hashed now. Optimise it out. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-04 13:14:10 +00:00
} while (work);
io_wq: add get/put_work handlers to io_wq_create() For cancellation, we need to ensure that the work item stays valid for as long as ->cur_work is valid. Right now we can't safely dereference the work item even under the wqe->lock, because while the ->cur_work pointer will remain valid, the work could be completing and be freed in parallel. Only invoke ->get/put_work() on items we know that the caller queued themselves. Add IO_WQ_WORK_INTERNAL for io-wq to use, which is needed when we're queueing a flush item, for instance. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-13 05:31:31 +00:00
io_wq: Make io_wqe::lock a raw_spinlock_t During a context switch the scheduler invokes wq_worker_sleeping() with disabled preemption. Disabling preemption is needed because it protects access to `worker->sleeping'. As an optimisation it avoids invoking schedule() within the schedule path as part of possible wake up (thus preempt_enable_no_resched() afterwards). The io-wq has been added to the mix in the same section with disabled preemption. This breaks on PREEMPT_RT because io_wq_worker_sleeping() acquires a spinlock_t. Also within the schedule() the spinlock_t must be acquired after tsk_is_pi_blocked() otherwise it will block on the sleeping lock again while scheduling out. While playing with `io_uring-bench' I didn't notice a significant latency spike after converting io_wqe::lock to a raw_spinlock_t. The latency was more or less the same. In order to keep the spinlock_t it would have to be moved after the tsk_is_pi_blocked() check which would introduce a branch instruction into the hot path. The lock is used to maintain the `work_list' and wakes one task up at most. Should io_wqe_cancel_pending_work() cause latency spikes, while searching for a specific item, then it would need to drop the lock during iterations. revert_creds() is also invoked under the lock. According to debug cred::non_rcu is 0. Otherwise it should be moved outside of the locked section because put_cred_rcu()->free_uid() acquires a sleeping lock. Convert io_wqe::lock to a raw_spinlock_t.c Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-01 08:41:46 +00:00
raw_spin_lock_irq(&wqe->lock);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
} while (1);
}
static int io_wqe_worker(void *data)
{
struct io_worker *worker = data;
struct io_wqe *wqe = worker->wqe;
struct io_wq *wq = wqe->wq;
io_worker_start(wqe, worker);
while (!test_bit(IO_WQ_BIT_EXIT, &wq->state)) {
io-wq: remove worker->wait waitqueue We only have one cases of using the waitqueue to wake the worker, the rest are using wake_up_process(). Since we can save some cycles not fiddling with the waitqueue io_wqe_worker(), switch the work activation to task wakeup and get rid of the now unused wait_queue_head_t in struct io_worker. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-12-08 04:03:59 +00:00
set_current_state(TASK_INTERRUPTIBLE);
io-wq: briefly spin for new work after finishing work To avoid going to sleep only to get woken shortly thereafter, spin briefly for new work upon completion of work. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-12-08 04:06:46 +00:00
loop:
io_wq: Make io_wqe::lock a raw_spinlock_t During a context switch the scheduler invokes wq_worker_sleeping() with disabled preemption. Disabling preemption is needed because it protects access to `worker->sleeping'. As an optimisation it avoids invoking schedule() within the schedule path as part of possible wake up (thus preempt_enable_no_resched() afterwards). The io-wq has been added to the mix in the same section with disabled preemption. This breaks on PREEMPT_RT because io_wq_worker_sleeping() acquires a spinlock_t. Also within the schedule() the spinlock_t must be acquired after tsk_is_pi_blocked() otherwise it will block on the sleeping lock again while scheduling out. While playing with `io_uring-bench' I didn't notice a significant latency spike after converting io_wqe::lock to a raw_spinlock_t. The latency was more or less the same. In order to keep the spinlock_t it would have to be moved after the tsk_is_pi_blocked() check which would introduce a branch instruction into the hot path. The lock is used to maintain the `work_list' and wakes one task up at most. Should io_wqe_cancel_pending_work() cause latency spikes, while searching for a specific item, then it would need to drop the lock during iterations. revert_creds() is also invoked under the lock. According to debug cred::non_rcu is 0. Otherwise it should be moved outside of the locked section because put_cred_rcu()->free_uid() acquires a sleeping lock. Convert io_wqe::lock to a raw_spinlock_t.c Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-01 08:41:46 +00:00
raw_spin_lock_irq(&wqe->lock);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
if (io_wqe_run_queue(wqe)) {
__set_current_state(TASK_RUNNING);
io_worker_handle_work(worker);
io-wq: briefly spin for new work after finishing work To avoid going to sleep only to get woken shortly thereafter, spin briefly for new work upon completion of work. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-12-08 04:06:46 +00:00
goto loop;
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
}
/* drops the lock on success, retry */
if (__io_worker_idle(wqe, worker)) {
__release(&wqe->lock);
io-wq: briefly spin for new work after finishing work To avoid going to sleep only to get woken shortly thereafter, spin briefly for new work upon completion of work. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-12-08 04:06:46 +00:00
goto loop;
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
}
io_wq: Make io_wqe::lock a raw_spinlock_t During a context switch the scheduler invokes wq_worker_sleeping() with disabled preemption. Disabling preemption is needed because it protects access to `worker->sleeping'. As an optimisation it avoids invoking schedule() within the schedule path as part of possible wake up (thus preempt_enable_no_resched() afterwards). The io-wq has been added to the mix in the same section with disabled preemption. This breaks on PREEMPT_RT because io_wq_worker_sleeping() acquires a spinlock_t. Also within the schedule() the spinlock_t must be acquired after tsk_is_pi_blocked() otherwise it will block on the sleeping lock again while scheduling out. While playing with `io_uring-bench' I didn't notice a significant latency spike after converting io_wqe::lock to a raw_spinlock_t. The latency was more or less the same. In order to keep the spinlock_t it would have to be moved after the tsk_is_pi_blocked() check which would introduce a branch instruction into the hot path. The lock is used to maintain the `work_list' and wakes one task up at most. Should io_wqe_cancel_pending_work() cause latency spikes, while searching for a specific item, then it would need to drop the lock during iterations. revert_creds() is also invoked under the lock. According to debug cred::non_rcu is 0. Otherwise it should be moved outside of the locked section because put_cred_rcu()->free_uid() acquires a sleeping lock. Convert io_wqe::lock to a raw_spinlock_t.c Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-01 08:41:46 +00:00
raw_spin_unlock_irq(&wqe->lock);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
if (signal_pending(current))
flush_signals(current);
if (schedule_timeout(WORKER_IDLE_TIMEOUT))
continue;
/* timed out, exit unless we're the fixed worker */
if (test_bit(IO_WQ_BIT_EXIT, &wq->state) ||
!(worker->flags & IO_WORKER_F_FIXED))
break;
}
if (test_bit(IO_WQ_BIT_EXIT, &wq->state)) {
io_wq: Make io_wqe::lock a raw_spinlock_t During a context switch the scheduler invokes wq_worker_sleeping() with disabled preemption. Disabling preemption is needed because it protects access to `worker->sleeping'. As an optimisation it avoids invoking schedule() within the schedule path as part of possible wake up (thus preempt_enable_no_resched() afterwards). The io-wq has been added to the mix in the same section with disabled preemption. This breaks on PREEMPT_RT because io_wq_worker_sleeping() acquires a spinlock_t. Also within the schedule() the spinlock_t must be acquired after tsk_is_pi_blocked() otherwise it will block on the sleeping lock again while scheduling out. While playing with `io_uring-bench' I didn't notice a significant latency spike after converting io_wqe::lock to a raw_spinlock_t. The latency was more or less the same. In order to keep the spinlock_t it would have to be moved after the tsk_is_pi_blocked() check which would introduce a branch instruction into the hot path. The lock is used to maintain the `work_list' and wakes one task up at most. Should io_wqe_cancel_pending_work() cause latency spikes, while searching for a specific item, then it would need to drop the lock during iterations. revert_creds() is also invoked under the lock. According to debug cred::non_rcu is 0. Otherwise it should be moved outside of the locked section because put_cred_rcu()->free_uid() acquires a sleeping lock. Convert io_wqe::lock to a raw_spinlock_t.c Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-01 08:41:46 +00:00
raw_spin_lock_irq(&wqe->lock);
io-wq: shrink io_wq_work a bit Currently we're using 40 bytes for the io_wq_work structure, and 16 of those is the doubly link list node. We don't need doubly linked lists, we always add to tail to keep things ordered, and any other use case is list traversal with deletion. For the deletion case, we can easily support any node deletion by keeping track of the previous entry. This shrinks io_wq_work to 32 bytes, and subsequently io_kiock from io_uring to 216 to 208 bytes. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-26 18:59:32 +00:00
if (!wq_list_empty(&wqe->work_list))
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
io_worker_handle_work(worker);
else
io_wq: Make io_wqe::lock a raw_spinlock_t During a context switch the scheduler invokes wq_worker_sleeping() with disabled preemption. Disabling preemption is needed because it protects access to `worker->sleeping'. As an optimisation it avoids invoking schedule() within the schedule path as part of possible wake up (thus preempt_enable_no_resched() afterwards). The io-wq has been added to the mix in the same section with disabled preemption. This breaks on PREEMPT_RT because io_wq_worker_sleeping() acquires a spinlock_t. Also within the schedule() the spinlock_t must be acquired after tsk_is_pi_blocked() otherwise it will block on the sleeping lock again while scheduling out. While playing with `io_uring-bench' I didn't notice a significant latency spike after converting io_wqe::lock to a raw_spinlock_t. The latency was more or less the same. In order to keep the spinlock_t it would have to be moved after the tsk_is_pi_blocked() check which would introduce a branch instruction into the hot path. The lock is used to maintain the `work_list' and wakes one task up at most. Should io_wqe_cancel_pending_work() cause latency spikes, while searching for a specific item, then it would need to drop the lock during iterations. revert_creds() is also invoked under the lock. According to debug cred::non_rcu is 0. Otherwise it should be moved outside of the locked section because put_cred_rcu()->free_uid() acquires a sleeping lock. Convert io_wqe::lock to a raw_spinlock_t.c Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-01 08:41:46 +00:00
raw_spin_unlock_irq(&wqe->lock);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
}
io_worker_exit(worker);
return 0;
}
/*
* Called when a worker is scheduled in. Mark us as currently running.
*/
void io_wq_worker_running(struct task_struct *tsk)
{
struct io_worker *worker = kthread_data(tsk);
struct io_wqe *wqe = worker->wqe;
if (!(worker->flags & IO_WORKER_F_UP))
return;
if (worker->flags & IO_WORKER_F_RUNNING)
return;
worker->flags |= IO_WORKER_F_RUNNING;
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
io_wqe_inc_running(wqe, worker);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
}
/*
* Called when worker is going to sleep. If there are no workers currently
* running and we have work pending, wake up a free one or have the manager
* set one up.
*/
void io_wq_worker_sleeping(struct task_struct *tsk)
{
struct io_worker *worker = kthread_data(tsk);
struct io_wqe *wqe = worker->wqe;
if (!(worker->flags & IO_WORKER_F_UP))
return;
if (!(worker->flags & IO_WORKER_F_RUNNING))
return;
worker->flags &= ~IO_WORKER_F_RUNNING;
io_wq: Make io_wqe::lock a raw_spinlock_t During a context switch the scheduler invokes wq_worker_sleeping() with disabled preemption. Disabling preemption is needed because it protects access to `worker->sleeping'. As an optimisation it avoids invoking schedule() within the schedule path as part of possible wake up (thus preempt_enable_no_resched() afterwards). The io-wq has been added to the mix in the same section with disabled preemption. This breaks on PREEMPT_RT because io_wq_worker_sleeping() acquires a spinlock_t. Also within the schedule() the spinlock_t must be acquired after tsk_is_pi_blocked() otherwise it will block on the sleeping lock again while scheduling out. While playing with `io_uring-bench' I didn't notice a significant latency spike after converting io_wqe::lock to a raw_spinlock_t. The latency was more or less the same. In order to keep the spinlock_t it would have to be moved after the tsk_is_pi_blocked() check which would introduce a branch instruction into the hot path. The lock is used to maintain the `work_list' and wakes one task up at most. Should io_wqe_cancel_pending_work() cause latency spikes, while searching for a specific item, then it would need to drop the lock during iterations. revert_creds() is also invoked under the lock. According to debug cred::non_rcu is 0. Otherwise it should be moved outside of the locked section because put_cred_rcu()->free_uid() acquires a sleeping lock. Convert io_wqe::lock to a raw_spinlock_t.c Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-01 08:41:46 +00:00
raw_spin_lock_irq(&wqe->lock);
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
io_wqe_dec_running(wqe, worker);
io_wq: Make io_wqe::lock a raw_spinlock_t During a context switch the scheduler invokes wq_worker_sleeping() with disabled preemption. Disabling preemption is needed because it protects access to `worker->sleeping'. As an optimisation it avoids invoking schedule() within the schedule path as part of possible wake up (thus preempt_enable_no_resched() afterwards). The io-wq has been added to the mix in the same section with disabled preemption. This breaks on PREEMPT_RT because io_wq_worker_sleeping() acquires a spinlock_t. Also within the schedule() the spinlock_t must be acquired after tsk_is_pi_blocked() otherwise it will block on the sleeping lock again while scheduling out. While playing with `io_uring-bench' I didn't notice a significant latency spike after converting io_wqe::lock to a raw_spinlock_t. The latency was more or less the same. In order to keep the spinlock_t it would have to be moved after the tsk_is_pi_blocked() check which would introduce a branch instruction into the hot path. The lock is used to maintain the `work_list' and wakes one task up at most. Should io_wqe_cancel_pending_work() cause latency spikes, while searching for a specific item, then it would need to drop the lock during iterations. revert_creds() is also invoked under the lock. According to debug cred::non_rcu is 0. Otherwise it should be moved outside of the locked section because put_cred_rcu()->free_uid() acquires a sleeping lock. Convert io_wqe::lock to a raw_spinlock_t.c Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-01 08:41:46 +00:00
raw_spin_unlock_irq(&wqe->lock);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
}
io-wq: wait for io_wq_create() to setup necessary workers We currently have a race where if setup is really slow, we can be calling io_wq_destroy() before we're done setting up. This will cause the caller to get stuck waiting for the manager to set things up, but the manager already exited. Fix this by doing a sync setup of the manager. This also fixes the case where if we failed creating workers, we'd also get stuck. In practice this race window was really small, as we already wait for the manager to start. Hence someone would have to call io_wq_destroy() after the task has started, but before it started the first loop. The reported test case forked tons of these, which is why it became an issue. Reported-by: syzbot+0f1cc17f85154f400465@syzkaller.appspotmail.com Fixes: 771b53d033e8 ("io-wq: small threadpool implementation for io_uring") Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-19 15:37:07 +00:00
static bool create_io_worker(struct io_wq *wq, struct io_wqe *wqe, int index)
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
{
io-wq: fix use-after-free in io_wq_worker_running The smart syzbot has found a reproducer for the following issue: ================================================================== BUG: KASAN: use-after-free in instrument_atomic_write include/linux/instrumented.h:71 [inline] BUG: KASAN: use-after-free in atomic_inc include/asm-generic/atomic-instrumented.h:240 [inline] BUG: KASAN: use-after-free in io_wqe_inc_running fs/io-wq.c:301 [inline] BUG: KASAN: use-after-free in io_wq_worker_running+0xde/0x110 fs/io-wq.c:613 Write of size 4 at addr ffff8882183db08c by task io_wqe_worker-0/7771 CPU: 0 PID: 7771 Comm: io_wqe_worker-0 Not tainted 5.9.0-rc4-syzkaller #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011 Call Trace: __dump_stack lib/dump_stack.c:77 [inline] dump_stack+0x198/0x1fd lib/dump_stack.c:118 print_address_description.constprop.0.cold+0xae/0x497 mm/kasan/report.c:383 __kasan_report mm/kasan/report.c:513 [inline] kasan_report.cold+0x1f/0x37 mm/kasan/report.c:530 check_memory_region_inline mm/kasan/generic.c:186 [inline] check_memory_region+0x13d/0x180 mm/kasan/generic.c:192 instrument_atomic_write include/linux/instrumented.h:71 [inline] atomic_inc include/asm-generic/atomic-instrumented.h:240 [inline] io_wqe_inc_running fs/io-wq.c:301 [inline] io_wq_worker_running+0xde/0x110 fs/io-wq.c:613 schedule_timeout+0x148/0x250 kernel/time/timer.c:1879 io_wqe_worker+0x517/0x10e0 fs/io-wq.c:580 kthread+0x3b5/0x4a0 kernel/kthread.c:292 ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:294 Allocated by task 7768: kasan_save_stack+0x1b/0x40 mm/kasan/common.c:48 kasan_set_track mm/kasan/common.c:56 [inline] __kasan_kmalloc.constprop.0+0xbf/0xd0 mm/kasan/common.c:461 kmem_cache_alloc_node_trace+0x17b/0x3f0 mm/slab.c:3594 kmalloc_node include/linux/slab.h:572 [inline] kzalloc_node include/linux/slab.h:677 [inline] io_wq_create+0x57b/0xa10 fs/io-wq.c:1064 io_init_wq_offload fs/io_uring.c:7432 [inline] io_sq_offload_start fs/io_uring.c:7504 [inline] io_uring_create fs/io_uring.c:8625 [inline] io_uring_setup+0x1836/0x28e0 fs/io_uring.c:8694 do_syscall_64+0x2d/0x70 arch/x86/entry/common.c:46 entry_SYSCALL_64_after_hwframe+0x44/0xa9 Freed by task 21: kasan_save_stack+0x1b/0x40 mm/kasan/common.c:48 kasan_set_track+0x1c/0x30 mm/kasan/common.c:56 kasan_set_free_info+0x1b/0x30 mm/kasan/generic.c:355 __kasan_slab_free+0xd8/0x120 mm/kasan/common.c:422 __cache_free mm/slab.c:3418 [inline] kfree+0x10e/0x2b0 mm/slab.c:3756 __io_wq_destroy fs/io-wq.c:1138 [inline] io_wq_destroy+0x2af/0x460 fs/io-wq.c:1146 io_finish_async fs/io_uring.c:6836 [inline] io_ring_ctx_free fs/io_uring.c:7870 [inline] io_ring_exit_work+0x1e4/0x6d0 fs/io_uring.c:7954 process_one_work+0x94c/0x1670 kernel/workqueue.c:2269 worker_thread+0x64c/0x1120 kernel/workqueue.c:2415 kthread+0x3b5/0x4a0 kernel/kthread.c:292 ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:294 The buggy address belongs to the object at ffff8882183db000 which belongs to the cache kmalloc-1k of size 1024 The buggy address is located 140 bytes inside of 1024-byte region [ffff8882183db000, ffff8882183db400) The buggy address belongs to the page: page:000000009bada22b refcount:1 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0x2183db flags: 0x57ffe0000000200(slab) raw: 057ffe0000000200 ffffea0008604c48 ffffea00086a8648 ffff8880aa040700 raw: 0000000000000000 ffff8882183db000 0000000100000002 0000000000000000 page dumped because: kasan: bad access detected Memory state around the buggy address: ffff8882183daf80: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ffff8882183db000: fa fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb >ffff8882183db080: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ^ ffff8882183db100: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ffff8882183db180: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ================================================================== which is down to the comment below, /* all workers gone, wq exit can proceed */ if (!nr_workers && refcount_dec_and_test(&wqe->wq->refs)) complete(&wqe->wq->done); because there might be multiple cases of wqe in a wq and we would wait for every worker in every wqe to go home before releasing wq's resources on destroying. To that end, rework wq's refcount by making it independent of the tracking of workers because after all they are two different things, and keeping it balanced when workers come and go. Note the manager kthread, like other workers, now holds a grab to wq during its lifetime. Finally to help destroy wq, check IO_WQ_BIT_EXIT upon creating worker and do nothing for exiting wq. Cc: stable@vger.kernel.org # v5.5+ Reported-by: syzbot+45fa0a195b941764e0f0@syzkaller.appspotmail.com Reported-by: syzbot+9af99580130003da82b1@syzkaller.appspotmail.com Cc: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Hillf Danton <hdanton@sina.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-26 13:26:55 +00:00
struct io_wqe_acct *acct = &wqe->acct[index];
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
struct io_worker *worker;
io_uring: use kzalloc instead of kcalloc for single-element allocations These allocations are single-element allocations, so don't use the array allocation wrapper for them. Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-26 16:39:45 +00:00
worker = kzalloc_node(sizeof(*worker), GFP_KERNEL, wqe->node);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
if (!worker)
io-wq: wait for io_wq_create() to setup necessary workers We currently have a race where if setup is really slow, we can be calling io_wq_destroy() before we're done setting up. This will cause the caller to get stuck waiting for the manager to set things up, but the manager already exited. Fix this by doing a sync setup of the manager. This also fixes the case where if we failed creating workers, we'd also get stuck. In practice this race window was really small, as we already wait for the manager to start. Hence someone would have to call io_wq_destroy() after the task has started, but before it started the first loop. The reported test case forked tons of these, which is why it became an issue. Reported-by: syzbot+0f1cc17f85154f400465@syzkaller.appspotmail.com Fixes: 771b53d033e8 ("io-wq: small threadpool implementation for io_uring") Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-19 15:37:07 +00:00
return false;
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
refcount_set(&worker->ref, 1);
worker->nulls_node.pprev = NULL;
worker->wqe = wqe;
io-wq: ensure we have a stable view of ->cur_work for cancellations worker->cur_work is currently protected by the lock of the wqe that the worker belongs to. When we send a signal to a worker, we need a stable view of ->cur_work, so we need to hold that lock. But this doesn't work so well, since we have the opposite order potentially on queueing work. If POLL_ADD is used with a signalfd, then io_poll_wake() is called with the signal lock, and that sometimes needs to insert work items. Add a specific worker lock that protects the current work item. Then we can guarantee that the task we're sending a signal is currently processing the exact work we think it is. Reported-by: Paul E. McKenney <paulmck@kernel.org> Reviewed-by: Paul E. McKenney <paulmck@kernel.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-13 16:43:34 +00:00
spin_lock_init(&worker->lock);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
worker->task = kthread_create_on_node(io_wqe_worker, worker, wqe->node,
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
"io_wqe_worker-%d/%d", index, wqe->node);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
if (IS_ERR(worker->task)) {
kfree(worker);
io-wq: wait for io_wq_create() to setup necessary workers We currently have a race where if setup is really slow, we can be calling io_wq_destroy() before we're done setting up. This will cause the caller to get stuck waiting for the manager to set things up, but the manager already exited. Fix this by doing a sync setup of the manager. This also fixes the case where if we failed creating workers, we'd also get stuck. In practice this race window was really small, as we already wait for the manager to start. Hence someone would have to call io_wq_destroy() after the task has started, but before it started the first loop. The reported test case forked tons of these, which is why it became an issue. Reported-by: syzbot+0f1cc17f85154f400465@syzkaller.appspotmail.com Fixes: 771b53d033e8 ("io-wq: small threadpool implementation for io_uring") Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-19 15:37:07 +00:00
return false;
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
}
io-wq: assign NUMA node locality if appropriate There was an assumption that kthread_create_on_node() would properly set NUMA affinities in terms of CPUs allowed, but it doesn't. Make sure we do this when creating an io-wq context on NUMA. Cc: stable@vger.kernel.org Stefan Metzmacher <metze@samba.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-10-15 16:13:07 +00:00
kthread_bind_mask(worker->task, cpumask_of_node(wqe->node));
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
io_wq: Make io_wqe::lock a raw_spinlock_t During a context switch the scheduler invokes wq_worker_sleeping() with disabled preemption. Disabling preemption is needed because it protects access to `worker->sleeping'. As an optimisation it avoids invoking schedule() within the schedule path as part of possible wake up (thus preempt_enable_no_resched() afterwards). The io-wq has been added to the mix in the same section with disabled preemption. This breaks on PREEMPT_RT because io_wq_worker_sleeping() acquires a spinlock_t. Also within the schedule() the spinlock_t must be acquired after tsk_is_pi_blocked() otherwise it will block on the sleeping lock again while scheduling out. While playing with `io_uring-bench' I didn't notice a significant latency spike after converting io_wqe::lock to a raw_spinlock_t. The latency was more or less the same. In order to keep the spinlock_t it would have to be moved after the tsk_is_pi_blocked() check which would introduce a branch instruction into the hot path. The lock is used to maintain the `work_list' and wakes one task up at most. Should io_wqe_cancel_pending_work() cause latency spikes, while searching for a specific item, then it would need to drop the lock during iterations. revert_creds() is also invoked under the lock. According to debug cred::non_rcu is 0. Otherwise it should be moved outside of the locked section because put_cred_rcu()->free_uid() acquires a sleeping lock. Convert io_wqe::lock to a raw_spinlock_t.c Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-01 08:41:46 +00:00
raw_spin_lock_irq(&wqe->lock);
io-wq: remove now redundant struct io_wq_nulls_list Since we don't iterate these lists anymore after commit: e61df66c69b1 ("io-wq: ensure free/busy list browsing see all items") we don't need to retain the nulls value we use for them. That means it's pretty pointless to wrap the hlist_nulls_head in a structure, so get rid of it. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-14 15:00:41 +00:00
hlist_nulls_add_head_rcu(&worker->nulls_node, &wqe->free_list);
io-wq: ensure free/busy list browsing see all items We have two lists for workers in io-wq, a busy and a free list. For certain operations we want to browse all workers, and we currently do that by browsing the two separate lists. But since these lists are RCU protected, we can potentially miss workers if they move between the two lists while we're browsing them. Add a third list, all_list, that simply holds all workers. A worker is added to that list when it starts, and removed when it exits. This makes the worker iteration cleaner, too. Reported-by: Paul E. McKenney <paulmck@kernel.org> Reviewed-by: Paul E. McKenney <paulmck@kernel.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-13 20:54:49 +00:00
list_add_tail_rcu(&worker->all_list, &wqe->all_list);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
worker->flags |= IO_WORKER_F_FREE;
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
if (index == IO_WQ_ACCT_BOUND)
worker->flags |= IO_WORKER_F_BOUND;
if (!acct->nr_workers && (worker->flags & IO_WORKER_F_BOUND))
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
worker->flags |= IO_WORKER_F_FIXED;
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
acct->nr_workers++;
io_wq: Make io_wqe::lock a raw_spinlock_t During a context switch the scheduler invokes wq_worker_sleeping() with disabled preemption. Disabling preemption is needed because it protects access to `worker->sleeping'. As an optimisation it avoids invoking schedule() within the schedule path as part of possible wake up (thus preempt_enable_no_resched() afterwards). The io-wq has been added to the mix in the same section with disabled preemption. This breaks on PREEMPT_RT because io_wq_worker_sleeping() acquires a spinlock_t. Also within the schedule() the spinlock_t must be acquired after tsk_is_pi_blocked() otherwise it will block on the sleeping lock again while scheduling out. While playing with `io_uring-bench' I didn't notice a significant latency spike after converting io_wqe::lock to a raw_spinlock_t. The latency was more or less the same. In order to keep the spinlock_t it would have to be moved after the tsk_is_pi_blocked() check which would introduce a branch instruction into the hot path. The lock is used to maintain the `work_list' and wakes one task up at most. Should io_wqe_cancel_pending_work() cause latency spikes, while searching for a specific item, then it would need to drop the lock during iterations. revert_creds() is also invoked under the lock. According to debug cred::non_rcu is 0. Otherwise it should be moved outside of the locked section because put_cred_rcu()->free_uid() acquires a sleeping lock. Convert io_wqe::lock to a raw_spinlock_t.c Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-01 08:41:46 +00:00
raw_spin_unlock_irq(&wqe->lock);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
if (index == IO_WQ_ACCT_UNBOUND)
atomic_inc(&wq->user->processes);
io-wq: fix use-after-free in io_wq_worker_running The smart syzbot has found a reproducer for the following issue: ================================================================== BUG: KASAN: use-after-free in instrument_atomic_write include/linux/instrumented.h:71 [inline] BUG: KASAN: use-after-free in atomic_inc include/asm-generic/atomic-instrumented.h:240 [inline] BUG: KASAN: use-after-free in io_wqe_inc_running fs/io-wq.c:301 [inline] BUG: KASAN: use-after-free in io_wq_worker_running+0xde/0x110 fs/io-wq.c:613 Write of size 4 at addr ffff8882183db08c by task io_wqe_worker-0/7771 CPU: 0 PID: 7771 Comm: io_wqe_worker-0 Not tainted 5.9.0-rc4-syzkaller #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011 Call Trace: __dump_stack lib/dump_stack.c:77 [inline] dump_stack+0x198/0x1fd lib/dump_stack.c:118 print_address_description.constprop.0.cold+0xae/0x497 mm/kasan/report.c:383 __kasan_report mm/kasan/report.c:513 [inline] kasan_report.cold+0x1f/0x37 mm/kasan/report.c:530 check_memory_region_inline mm/kasan/generic.c:186 [inline] check_memory_region+0x13d/0x180 mm/kasan/generic.c:192 instrument_atomic_write include/linux/instrumented.h:71 [inline] atomic_inc include/asm-generic/atomic-instrumented.h:240 [inline] io_wqe_inc_running fs/io-wq.c:301 [inline] io_wq_worker_running+0xde/0x110 fs/io-wq.c:613 schedule_timeout+0x148/0x250 kernel/time/timer.c:1879 io_wqe_worker+0x517/0x10e0 fs/io-wq.c:580 kthread+0x3b5/0x4a0 kernel/kthread.c:292 ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:294 Allocated by task 7768: kasan_save_stack+0x1b/0x40 mm/kasan/common.c:48 kasan_set_track mm/kasan/common.c:56 [inline] __kasan_kmalloc.constprop.0+0xbf/0xd0 mm/kasan/common.c:461 kmem_cache_alloc_node_trace+0x17b/0x3f0 mm/slab.c:3594 kmalloc_node include/linux/slab.h:572 [inline] kzalloc_node include/linux/slab.h:677 [inline] io_wq_create+0x57b/0xa10 fs/io-wq.c:1064 io_init_wq_offload fs/io_uring.c:7432 [inline] io_sq_offload_start fs/io_uring.c:7504 [inline] io_uring_create fs/io_uring.c:8625 [inline] io_uring_setup+0x1836/0x28e0 fs/io_uring.c:8694 do_syscall_64+0x2d/0x70 arch/x86/entry/common.c:46 entry_SYSCALL_64_after_hwframe+0x44/0xa9 Freed by task 21: kasan_save_stack+0x1b/0x40 mm/kasan/common.c:48 kasan_set_track+0x1c/0x30 mm/kasan/common.c:56 kasan_set_free_info+0x1b/0x30 mm/kasan/generic.c:355 __kasan_slab_free+0xd8/0x120 mm/kasan/common.c:422 __cache_free mm/slab.c:3418 [inline] kfree+0x10e/0x2b0 mm/slab.c:3756 __io_wq_destroy fs/io-wq.c:1138 [inline] io_wq_destroy+0x2af/0x460 fs/io-wq.c:1146 io_finish_async fs/io_uring.c:6836 [inline] io_ring_ctx_free fs/io_uring.c:7870 [inline] io_ring_exit_work+0x1e4/0x6d0 fs/io_uring.c:7954 process_one_work+0x94c/0x1670 kernel/workqueue.c:2269 worker_thread+0x64c/0x1120 kernel/workqueue.c:2415 kthread+0x3b5/0x4a0 kernel/kthread.c:292 ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:294 The buggy address belongs to the object at ffff8882183db000 which belongs to the cache kmalloc-1k of size 1024 The buggy address is located 140 bytes inside of 1024-byte region [ffff8882183db000, ffff8882183db400) The buggy address belongs to the page: page:000000009bada22b refcount:1 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0x2183db flags: 0x57ffe0000000200(slab) raw: 057ffe0000000200 ffffea0008604c48 ffffea00086a8648 ffff8880aa040700 raw: 0000000000000000 ffff8882183db000 0000000100000002 0000000000000000 page dumped because: kasan: bad access detected Memory state around the buggy address: ffff8882183daf80: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ffff8882183db000: fa fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb >ffff8882183db080: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ^ ffff8882183db100: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ffff8882183db180: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ================================================================== which is down to the comment below, /* all workers gone, wq exit can proceed */ if (!nr_workers && refcount_dec_and_test(&wqe->wq->refs)) complete(&wqe->wq->done); because there might be multiple cases of wqe in a wq and we would wait for every worker in every wqe to go home before releasing wq's resources on destroying. To that end, rework wq's refcount by making it independent of the tracking of workers because after all they are two different things, and keeping it balanced when workers come and go. Note the manager kthread, like other workers, now holds a grab to wq during its lifetime. Finally to help destroy wq, check IO_WQ_BIT_EXIT upon creating worker and do nothing for exiting wq. Cc: stable@vger.kernel.org # v5.5+ Reported-by: syzbot+45fa0a195b941764e0f0@syzkaller.appspotmail.com Reported-by: syzbot+9af99580130003da82b1@syzkaller.appspotmail.com Cc: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Hillf Danton <hdanton@sina.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-26 13:26:55 +00:00
refcount_inc(&wq->refs);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
wake_up_process(worker->task);
io-wq: wait for io_wq_create() to setup necessary workers We currently have a race where if setup is really slow, we can be calling io_wq_destroy() before we're done setting up. This will cause the caller to get stuck waiting for the manager to set things up, but the manager already exited. Fix this by doing a sync setup of the manager. This also fixes the case where if we failed creating workers, we'd also get stuck. In practice this race window was really small, as we already wait for the manager to start. Hence someone would have to call io_wq_destroy() after the task has started, but before it started the first loop. The reported test case forked tons of these, which is why it became an issue. Reported-by: syzbot+0f1cc17f85154f400465@syzkaller.appspotmail.com Fixes: 771b53d033e8 ("io-wq: small threadpool implementation for io_uring") Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-19 15:37:07 +00:00
return true;
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
}
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
static inline bool io_wqe_need_worker(struct io_wqe *wqe, int index)
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
__must_hold(wqe->lock)
{
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
struct io_wqe_acct *acct = &wqe->acct[index];
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
/* if we have available workers or no work, no need */
io-wq: remove now redundant struct io_wq_nulls_list Since we don't iterate these lists anymore after commit: e61df66c69b1 ("io-wq: ensure free/busy list browsing see all items") we don't need to retain the nulls value we use for them. That means it's pretty pointless to wrap the hlist_nulls_head in a structure, so get rid of it. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-14 15:00:41 +00:00
if (!hlist_nulls_empty(&wqe->free_list) || !io_wqe_run_queue(wqe))
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
return false;
return acct->nr_workers < acct->max_workers;
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
}
io-wq: fix use-after-free in io_wq_worker_running The smart syzbot has found a reproducer for the following issue: ================================================================== BUG: KASAN: use-after-free in instrument_atomic_write include/linux/instrumented.h:71 [inline] BUG: KASAN: use-after-free in atomic_inc include/asm-generic/atomic-instrumented.h:240 [inline] BUG: KASAN: use-after-free in io_wqe_inc_running fs/io-wq.c:301 [inline] BUG: KASAN: use-after-free in io_wq_worker_running+0xde/0x110 fs/io-wq.c:613 Write of size 4 at addr ffff8882183db08c by task io_wqe_worker-0/7771 CPU: 0 PID: 7771 Comm: io_wqe_worker-0 Not tainted 5.9.0-rc4-syzkaller #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011 Call Trace: __dump_stack lib/dump_stack.c:77 [inline] dump_stack+0x198/0x1fd lib/dump_stack.c:118 print_address_description.constprop.0.cold+0xae/0x497 mm/kasan/report.c:383 __kasan_report mm/kasan/report.c:513 [inline] kasan_report.cold+0x1f/0x37 mm/kasan/report.c:530 check_memory_region_inline mm/kasan/generic.c:186 [inline] check_memory_region+0x13d/0x180 mm/kasan/generic.c:192 instrument_atomic_write include/linux/instrumented.h:71 [inline] atomic_inc include/asm-generic/atomic-instrumented.h:240 [inline] io_wqe_inc_running fs/io-wq.c:301 [inline] io_wq_worker_running+0xde/0x110 fs/io-wq.c:613 schedule_timeout+0x148/0x250 kernel/time/timer.c:1879 io_wqe_worker+0x517/0x10e0 fs/io-wq.c:580 kthread+0x3b5/0x4a0 kernel/kthread.c:292 ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:294 Allocated by task 7768: kasan_save_stack+0x1b/0x40 mm/kasan/common.c:48 kasan_set_track mm/kasan/common.c:56 [inline] __kasan_kmalloc.constprop.0+0xbf/0xd0 mm/kasan/common.c:461 kmem_cache_alloc_node_trace+0x17b/0x3f0 mm/slab.c:3594 kmalloc_node include/linux/slab.h:572 [inline] kzalloc_node include/linux/slab.h:677 [inline] io_wq_create+0x57b/0xa10 fs/io-wq.c:1064 io_init_wq_offload fs/io_uring.c:7432 [inline] io_sq_offload_start fs/io_uring.c:7504 [inline] io_uring_create fs/io_uring.c:8625 [inline] io_uring_setup+0x1836/0x28e0 fs/io_uring.c:8694 do_syscall_64+0x2d/0x70 arch/x86/entry/common.c:46 entry_SYSCALL_64_after_hwframe+0x44/0xa9 Freed by task 21: kasan_save_stack+0x1b/0x40 mm/kasan/common.c:48 kasan_set_track+0x1c/0x30 mm/kasan/common.c:56 kasan_set_free_info+0x1b/0x30 mm/kasan/generic.c:355 __kasan_slab_free+0xd8/0x120 mm/kasan/common.c:422 __cache_free mm/slab.c:3418 [inline] kfree+0x10e/0x2b0 mm/slab.c:3756 __io_wq_destroy fs/io-wq.c:1138 [inline] io_wq_destroy+0x2af/0x460 fs/io-wq.c:1146 io_finish_async fs/io_uring.c:6836 [inline] io_ring_ctx_free fs/io_uring.c:7870 [inline] io_ring_exit_work+0x1e4/0x6d0 fs/io_uring.c:7954 process_one_work+0x94c/0x1670 kernel/workqueue.c:2269 worker_thread+0x64c/0x1120 kernel/workqueue.c:2415 kthread+0x3b5/0x4a0 kernel/kthread.c:292 ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:294 The buggy address belongs to the object at ffff8882183db000 which belongs to the cache kmalloc-1k of size 1024 The buggy address is located 140 bytes inside of 1024-byte region [ffff8882183db000, ffff8882183db400) The buggy address belongs to the page: page:000000009bada22b refcount:1 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0x2183db flags: 0x57ffe0000000200(slab) raw: 057ffe0000000200 ffffea0008604c48 ffffea00086a8648 ffff8880aa040700 raw: 0000000000000000 ffff8882183db000 0000000100000002 0000000000000000 page dumped because: kasan: bad access detected Memory state around the buggy address: ffff8882183daf80: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ffff8882183db000: fa fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb >ffff8882183db080: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ^ ffff8882183db100: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ffff8882183db180: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ================================================================== which is down to the comment below, /* all workers gone, wq exit can proceed */ if (!nr_workers && refcount_dec_and_test(&wqe->wq->refs)) complete(&wqe->wq->done); because there might be multiple cases of wqe in a wq and we would wait for every worker in every wqe to go home before releasing wq's resources on destroying. To that end, rework wq's refcount by making it independent of the tracking of workers because after all they are two different things, and keeping it balanced when workers come and go. Note the manager kthread, like other workers, now holds a grab to wq during its lifetime. Finally to help destroy wq, check IO_WQ_BIT_EXIT upon creating worker and do nothing for exiting wq. Cc: stable@vger.kernel.org # v5.5+ Reported-by: syzbot+45fa0a195b941764e0f0@syzkaller.appspotmail.com Reported-by: syzbot+9af99580130003da82b1@syzkaller.appspotmail.com Cc: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Hillf Danton <hdanton@sina.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-26 13:26:55 +00:00
/*
* Iterate the passed in list and call the specific function for each
* worker that isn't exiting
*/
static bool io_wq_for_each_worker(struct io_wqe *wqe,
bool (*func)(struct io_worker *, void *),
void *data)
{
struct io_worker *worker;
bool ret = false;
list_for_each_entry_rcu(worker, &wqe->all_list, all_list) {
if (io_worker_get(worker)) {
/* no task if node is/was offline */
if (worker->task)
ret = func(worker, data);
io_worker_release(worker);
if (ret)
break;
}
}
return ret;
}
static bool io_wq_worker_wake(struct io_worker *worker, void *data)
{
wake_up_process(worker->task);
return false;
}
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
/*
* Manager thread. Tasked with creating new workers, if we need them.
*/
static int io_wq_manager(void *data)
{
struct io_wq *wq = data;
io-wq: fix handling of NUMA node IDs There are several things that can go wrong in the current code on NUMA systems, especially if not all nodes are online all the time: - If the identifiers of the online nodes do not form a single contiguous block starting at zero, wq->wqes will be too small, and OOB memory accesses will occur e.g. in the loop in io_wq_create(). - If a node comes online between the call to num_online_nodes() and the for_each_node() loop in io_wq_create(), an OOB write will occur. - If a node comes online between io_wq_create() and io_wq_enqueue(), a lookup is performed for an element that doesn't exist, and an OOB read will probably occur. Fix it by: - using nr_node_ids instead of num_online_nodes() for the allocation size; nr_node_ids is calculated by setup_nr_node_ids() to be bigger than the highest node ID that could possibly come online at some point, even if those nodes' identifiers are not a contiguous block - creating workers for all possible CPUs, not just all online ones This is basically what the normal workqueue code also does, as far as I can tell. Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-26 18:10:20 +00:00
int node;
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
io-wq: wait for io_wq_create() to setup necessary workers We currently have a race where if setup is really slow, we can be calling io_wq_destroy() before we're done setting up. This will cause the caller to get stuck waiting for the manager to set things up, but the manager already exited. Fix this by doing a sync setup of the manager. This also fixes the case where if we failed creating workers, we'd also get stuck. In practice this race window was really small, as we already wait for the manager to start. Hence someone would have to call io_wq_destroy() after the task has started, but before it started the first loop. The reported test case forked tons of these, which is why it became an issue. Reported-by: syzbot+0f1cc17f85154f400465@syzkaller.appspotmail.com Fixes: 771b53d033e8 ("io-wq: small threadpool implementation for io_uring") Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-19 15:37:07 +00:00
/* create fixed workers */
io-wq: fix use-after-free in io_wq_worker_running The smart syzbot has found a reproducer for the following issue: ================================================================== BUG: KASAN: use-after-free in instrument_atomic_write include/linux/instrumented.h:71 [inline] BUG: KASAN: use-after-free in atomic_inc include/asm-generic/atomic-instrumented.h:240 [inline] BUG: KASAN: use-after-free in io_wqe_inc_running fs/io-wq.c:301 [inline] BUG: KASAN: use-after-free in io_wq_worker_running+0xde/0x110 fs/io-wq.c:613 Write of size 4 at addr ffff8882183db08c by task io_wqe_worker-0/7771 CPU: 0 PID: 7771 Comm: io_wqe_worker-0 Not tainted 5.9.0-rc4-syzkaller #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011 Call Trace: __dump_stack lib/dump_stack.c:77 [inline] dump_stack+0x198/0x1fd lib/dump_stack.c:118 print_address_description.constprop.0.cold+0xae/0x497 mm/kasan/report.c:383 __kasan_report mm/kasan/report.c:513 [inline] kasan_report.cold+0x1f/0x37 mm/kasan/report.c:530 check_memory_region_inline mm/kasan/generic.c:186 [inline] check_memory_region+0x13d/0x180 mm/kasan/generic.c:192 instrument_atomic_write include/linux/instrumented.h:71 [inline] atomic_inc include/asm-generic/atomic-instrumented.h:240 [inline] io_wqe_inc_running fs/io-wq.c:301 [inline] io_wq_worker_running+0xde/0x110 fs/io-wq.c:613 schedule_timeout+0x148/0x250 kernel/time/timer.c:1879 io_wqe_worker+0x517/0x10e0 fs/io-wq.c:580 kthread+0x3b5/0x4a0 kernel/kthread.c:292 ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:294 Allocated by task 7768: kasan_save_stack+0x1b/0x40 mm/kasan/common.c:48 kasan_set_track mm/kasan/common.c:56 [inline] __kasan_kmalloc.constprop.0+0xbf/0xd0 mm/kasan/common.c:461 kmem_cache_alloc_node_trace+0x17b/0x3f0 mm/slab.c:3594 kmalloc_node include/linux/slab.h:572 [inline] kzalloc_node include/linux/slab.h:677 [inline] io_wq_create+0x57b/0xa10 fs/io-wq.c:1064 io_init_wq_offload fs/io_uring.c:7432 [inline] io_sq_offload_start fs/io_uring.c:7504 [inline] io_uring_create fs/io_uring.c:8625 [inline] io_uring_setup+0x1836/0x28e0 fs/io_uring.c:8694 do_syscall_64+0x2d/0x70 arch/x86/entry/common.c:46 entry_SYSCALL_64_after_hwframe+0x44/0xa9 Freed by task 21: kasan_save_stack+0x1b/0x40 mm/kasan/common.c:48 kasan_set_track+0x1c/0x30 mm/kasan/common.c:56 kasan_set_free_info+0x1b/0x30 mm/kasan/generic.c:355 __kasan_slab_free+0xd8/0x120 mm/kasan/common.c:422 __cache_free mm/slab.c:3418 [inline] kfree+0x10e/0x2b0 mm/slab.c:3756 __io_wq_destroy fs/io-wq.c:1138 [inline] io_wq_destroy+0x2af/0x460 fs/io-wq.c:1146 io_finish_async fs/io_uring.c:6836 [inline] io_ring_ctx_free fs/io_uring.c:7870 [inline] io_ring_exit_work+0x1e4/0x6d0 fs/io_uring.c:7954 process_one_work+0x94c/0x1670 kernel/workqueue.c:2269 worker_thread+0x64c/0x1120 kernel/workqueue.c:2415 kthread+0x3b5/0x4a0 kernel/kthread.c:292 ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:294 The buggy address belongs to the object at ffff8882183db000 which belongs to the cache kmalloc-1k of size 1024 The buggy address is located 140 bytes inside of 1024-byte region [ffff8882183db000, ffff8882183db400) The buggy address belongs to the page: page:000000009bada22b refcount:1 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0x2183db flags: 0x57ffe0000000200(slab) raw: 057ffe0000000200 ffffea0008604c48 ffffea00086a8648 ffff8880aa040700 raw: 0000000000000000 ffff8882183db000 0000000100000002 0000000000000000 page dumped because: kasan: bad access detected Memory state around the buggy address: ffff8882183daf80: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ffff8882183db000: fa fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb >ffff8882183db080: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ^ ffff8882183db100: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ffff8882183db180: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ================================================================== which is down to the comment below, /* all workers gone, wq exit can proceed */ if (!nr_workers && refcount_dec_and_test(&wqe->wq->refs)) complete(&wqe->wq->done); because there might be multiple cases of wqe in a wq and we would wait for every worker in every wqe to go home before releasing wq's resources on destroying. To that end, rework wq's refcount by making it independent of the tracking of workers because after all they are two different things, and keeping it balanced when workers come and go. Note the manager kthread, like other workers, now holds a grab to wq during its lifetime. Finally to help destroy wq, check IO_WQ_BIT_EXIT upon creating worker and do nothing for exiting wq. Cc: stable@vger.kernel.org # v5.5+ Reported-by: syzbot+45fa0a195b941764e0f0@syzkaller.appspotmail.com Reported-by: syzbot+9af99580130003da82b1@syzkaller.appspotmail.com Cc: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Hillf Danton <hdanton@sina.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-26 13:26:55 +00:00
refcount_set(&wq->refs, 1);
io-wq: fix handling of NUMA node IDs There are several things that can go wrong in the current code on NUMA systems, especially if not all nodes are online all the time: - If the identifiers of the online nodes do not form a single contiguous block starting at zero, wq->wqes will be too small, and OOB memory accesses will occur e.g. in the loop in io_wq_create(). - If a node comes online between the call to num_online_nodes() and the for_each_node() loop in io_wq_create(), an OOB write will occur. - If a node comes online between io_wq_create() and io_wq_enqueue(), a lookup is performed for an element that doesn't exist, and an OOB read will probably occur. Fix it by: - using nr_node_ids instead of num_online_nodes() for the allocation size; nr_node_ids is calculated by setup_nr_node_ids() to be bigger than the highest node ID that could possibly come online at some point, even if those nodes' identifiers are not a contiguous block - creating workers for all possible CPUs, not just all online ones This is basically what the normal workqueue code also does, as far as I can tell. Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-26 18:10:20 +00:00
for_each_node(node) {
io-wq: don't call kXalloc_node() with non-online node Glauber reports a crash on init on a box he has: RIP: 0010:__alloc_pages_nodemask+0x132/0x340 Code: 18 01 75 04 41 80 ce 80 89 e8 48 8b 54 24 08 8b 74 24 1c c1 e8 0c 48 8b 3c 24 83 e0 01 88 44 24 20 48 85 d2 0f 85 74 01 00 00 <3b> 77 08 0f 82 6b 01 00 00 48 89 7c 24 10 89 ea 48 8b 07 b9 00 02 RSP: 0018:ffffb8be4d0b7c28 EFLAGS: 00010246 RAX: 0000000000000000 RBX: 0000000000000000 RCX: 000000000000e8e8 RDX: 0000000000000000 RSI: 0000000000000002 RDI: 0000000000002080 RBP: 0000000000012cc0 R08: 0000000000000000 R09: 0000000000000002 R10: 0000000000000dc0 R11: ffff995c60400100 R12: 0000000000000000 R13: 0000000000012cc0 R14: 0000000000000001 R15: ffff995c60db00f0 FS: 00007f4d115ca900(0000) GS:ffff995c60d80000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000000002088 CR3: 00000017cca66002 CR4: 00000000007606e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 PKRU: 55555554 Call Trace: alloc_slab_page+0x46/0x320 new_slab+0x9d/0x4e0 ___slab_alloc+0x507/0x6a0 ? io_wq_create+0xb4/0x2a0 __slab_alloc+0x1c/0x30 kmem_cache_alloc_node_trace+0xa6/0x260 io_wq_create+0xb4/0x2a0 io_uring_setup+0x97f/0xaa0 ? io_remove_personalities+0x30/0x30 ? io_poll_trigger_evfd+0x30/0x30 do_syscall_64+0x5b/0x1c0 entry_SYSCALL_64_after_hwframe+0x44/0xa9 RIP: 0033:0x7f4d116cb1ed which is due to the 'wqe' and 'worker' allocation being node affine. But it isn't valid to call the node affine allocation if the node isn't online. Setup structures for even offline nodes, as usual, but skip them in terms of thread setup to not waste resources. If the node isn't online, just alloc memory with NUMA_NO_NODE. Reported-by: Glauber Costa <glauber@scylladb.com> Tested-by: Glauber Costa <glauber@scylladb.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-02-11 13:30:06 +00:00
if (!node_online(node))
continue;
io-wq: fix use-after-free in io_wq_worker_running The smart syzbot has found a reproducer for the following issue: ================================================================== BUG: KASAN: use-after-free in instrument_atomic_write include/linux/instrumented.h:71 [inline] BUG: KASAN: use-after-free in atomic_inc include/asm-generic/atomic-instrumented.h:240 [inline] BUG: KASAN: use-after-free in io_wqe_inc_running fs/io-wq.c:301 [inline] BUG: KASAN: use-after-free in io_wq_worker_running+0xde/0x110 fs/io-wq.c:613 Write of size 4 at addr ffff8882183db08c by task io_wqe_worker-0/7771 CPU: 0 PID: 7771 Comm: io_wqe_worker-0 Not tainted 5.9.0-rc4-syzkaller #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011 Call Trace: __dump_stack lib/dump_stack.c:77 [inline] dump_stack+0x198/0x1fd lib/dump_stack.c:118 print_address_description.constprop.0.cold+0xae/0x497 mm/kasan/report.c:383 __kasan_report mm/kasan/report.c:513 [inline] kasan_report.cold+0x1f/0x37 mm/kasan/report.c:530 check_memory_region_inline mm/kasan/generic.c:186 [inline] check_memory_region+0x13d/0x180 mm/kasan/generic.c:192 instrument_atomic_write include/linux/instrumented.h:71 [inline] atomic_inc include/asm-generic/atomic-instrumented.h:240 [inline] io_wqe_inc_running fs/io-wq.c:301 [inline] io_wq_worker_running+0xde/0x110 fs/io-wq.c:613 schedule_timeout+0x148/0x250 kernel/time/timer.c:1879 io_wqe_worker+0x517/0x10e0 fs/io-wq.c:580 kthread+0x3b5/0x4a0 kernel/kthread.c:292 ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:294 Allocated by task 7768: kasan_save_stack+0x1b/0x40 mm/kasan/common.c:48 kasan_set_track mm/kasan/common.c:56 [inline] __kasan_kmalloc.constprop.0+0xbf/0xd0 mm/kasan/common.c:461 kmem_cache_alloc_node_trace+0x17b/0x3f0 mm/slab.c:3594 kmalloc_node include/linux/slab.h:572 [inline] kzalloc_node include/linux/slab.h:677 [inline] io_wq_create+0x57b/0xa10 fs/io-wq.c:1064 io_init_wq_offload fs/io_uring.c:7432 [inline] io_sq_offload_start fs/io_uring.c:7504 [inline] io_uring_create fs/io_uring.c:8625 [inline] io_uring_setup+0x1836/0x28e0 fs/io_uring.c:8694 do_syscall_64+0x2d/0x70 arch/x86/entry/common.c:46 entry_SYSCALL_64_after_hwframe+0x44/0xa9 Freed by task 21: kasan_save_stack+0x1b/0x40 mm/kasan/common.c:48 kasan_set_track+0x1c/0x30 mm/kasan/common.c:56 kasan_set_free_info+0x1b/0x30 mm/kasan/generic.c:355 __kasan_slab_free+0xd8/0x120 mm/kasan/common.c:422 __cache_free mm/slab.c:3418 [inline] kfree+0x10e/0x2b0 mm/slab.c:3756 __io_wq_destroy fs/io-wq.c:1138 [inline] io_wq_destroy+0x2af/0x460 fs/io-wq.c:1146 io_finish_async fs/io_uring.c:6836 [inline] io_ring_ctx_free fs/io_uring.c:7870 [inline] io_ring_exit_work+0x1e4/0x6d0 fs/io_uring.c:7954 process_one_work+0x94c/0x1670 kernel/workqueue.c:2269 worker_thread+0x64c/0x1120 kernel/workqueue.c:2415 kthread+0x3b5/0x4a0 kernel/kthread.c:292 ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:294 The buggy address belongs to the object at ffff8882183db000 which belongs to the cache kmalloc-1k of size 1024 The buggy address is located 140 bytes inside of 1024-byte region [ffff8882183db000, ffff8882183db400) The buggy address belongs to the page: page:000000009bada22b refcount:1 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0x2183db flags: 0x57ffe0000000200(slab) raw: 057ffe0000000200 ffffea0008604c48 ffffea00086a8648 ffff8880aa040700 raw: 0000000000000000 ffff8882183db000 0000000100000002 0000000000000000 page dumped because: kasan: bad access detected Memory state around the buggy address: ffff8882183daf80: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ffff8882183db000: fa fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb >ffff8882183db080: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ^ ffff8882183db100: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ffff8882183db180: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ================================================================== which is down to the comment below, /* all workers gone, wq exit can proceed */ if (!nr_workers && refcount_dec_and_test(&wqe->wq->refs)) complete(&wqe->wq->done); because there might be multiple cases of wqe in a wq and we would wait for every worker in every wqe to go home before releasing wq's resources on destroying. To that end, rework wq's refcount by making it independent of the tracking of workers because after all they are two different things, and keeping it balanced when workers come and go. Note the manager kthread, like other workers, now holds a grab to wq during its lifetime. Finally to help destroy wq, check IO_WQ_BIT_EXIT upon creating worker and do nothing for exiting wq. Cc: stable@vger.kernel.org # v5.5+ Reported-by: syzbot+45fa0a195b941764e0f0@syzkaller.appspotmail.com Reported-by: syzbot+9af99580130003da82b1@syzkaller.appspotmail.com Cc: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Hillf Danton <hdanton@sina.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-26 13:26:55 +00:00
if (create_io_worker(wq, wq->wqes[node], IO_WQ_ACCT_BOUND))
continue;
set_bit(IO_WQ_BIT_ERROR, &wq->state);
set_bit(IO_WQ_BIT_EXIT, &wq->state);
goto out;
io-wq: wait for io_wq_create() to setup necessary workers We currently have a race where if setup is really slow, we can be calling io_wq_destroy() before we're done setting up. This will cause the caller to get stuck waiting for the manager to set things up, but the manager already exited. Fix this by doing a sync setup of the manager. This also fixes the case where if we failed creating workers, we'd also get stuck. In practice this race window was really small, as we already wait for the manager to start. Hence someone would have to call io_wq_destroy() after the task has started, but before it started the first loop. The reported test case forked tons of these, which is why it became an issue. Reported-by: syzbot+0f1cc17f85154f400465@syzkaller.appspotmail.com Fixes: 771b53d033e8 ("io-wq: small threadpool implementation for io_uring") Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-19 15:37:07 +00:00
}
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
io-wq: wait for io_wq_create() to setup necessary workers We currently have a race where if setup is really slow, we can be calling io_wq_destroy() before we're done setting up. This will cause the caller to get stuck waiting for the manager to set things up, but the manager already exited. Fix this by doing a sync setup of the manager. This also fixes the case where if we failed creating workers, we'd also get stuck. In practice this race window was really small, as we already wait for the manager to start. Hence someone would have to call io_wq_destroy() after the task has started, but before it started the first loop. The reported test case forked tons of these, which is why it became an issue. Reported-by: syzbot+0f1cc17f85154f400465@syzkaller.appspotmail.com Fixes: 771b53d033e8 ("io-wq: small threadpool implementation for io_uring") Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-19 15:37:07 +00:00
complete(&wq->done);
while (!kthread_should_stop()) {
io_uring: use io-wq manager as backup task if task is exiting If the original task is (or has) exited, then the task work will not get queued properly. Allow for using the io-wq manager task to queue this work for execution, and ensure that the io-wq manager notices and runs this work if woken up (or exiting). Reported-by: Dan Melnic <dmm@fb.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-04-03 17:26:26 +00:00
if (current->task_works)
task_work_run();
io-wq: fix handling of NUMA node IDs There are several things that can go wrong in the current code on NUMA systems, especially if not all nodes are online all the time: - If the identifiers of the online nodes do not form a single contiguous block starting at zero, wq->wqes will be too small, and OOB memory accesses will occur e.g. in the loop in io_wq_create(). - If a node comes online between the call to num_online_nodes() and the for_each_node() loop in io_wq_create(), an OOB write will occur. - If a node comes online between io_wq_create() and io_wq_enqueue(), a lookup is performed for an element that doesn't exist, and an OOB read will probably occur. Fix it by: - using nr_node_ids instead of num_online_nodes() for the allocation size; nr_node_ids is calculated by setup_nr_node_ids() to be bigger than the highest node ID that could possibly come online at some point, even if those nodes' identifiers are not a contiguous block - creating workers for all possible CPUs, not just all online ones This is basically what the normal workqueue code also does, as far as I can tell. Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-26 18:10:20 +00:00
for_each_node(node) {
struct io_wqe *wqe = wq->wqes[node];
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
bool fork_worker[2] = { false, false };
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
io-wq: don't call kXalloc_node() with non-online node Glauber reports a crash on init on a box he has: RIP: 0010:__alloc_pages_nodemask+0x132/0x340 Code: 18 01 75 04 41 80 ce 80 89 e8 48 8b 54 24 08 8b 74 24 1c c1 e8 0c 48 8b 3c 24 83 e0 01 88 44 24 20 48 85 d2 0f 85 74 01 00 00 <3b> 77 08 0f 82 6b 01 00 00 48 89 7c 24 10 89 ea 48 8b 07 b9 00 02 RSP: 0018:ffffb8be4d0b7c28 EFLAGS: 00010246 RAX: 0000000000000000 RBX: 0000000000000000 RCX: 000000000000e8e8 RDX: 0000000000000000 RSI: 0000000000000002 RDI: 0000000000002080 RBP: 0000000000012cc0 R08: 0000000000000000 R09: 0000000000000002 R10: 0000000000000dc0 R11: ffff995c60400100 R12: 0000000000000000 R13: 0000000000012cc0 R14: 0000000000000001 R15: ffff995c60db00f0 FS: 00007f4d115ca900(0000) GS:ffff995c60d80000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000000002088 CR3: 00000017cca66002 CR4: 00000000007606e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 PKRU: 55555554 Call Trace: alloc_slab_page+0x46/0x320 new_slab+0x9d/0x4e0 ___slab_alloc+0x507/0x6a0 ? io_wq_create+0xb4/0x2a0 __slab_alloc+0x1c/0x30 kmem_cache_alloc_node_trace+0xa6/0x260 io_wq_create+0xb4/0x2a0 io_uring_setup+0x97f/0xaa0 ? io_remove_personalities+0x30/0x30 ? io_poll_trigger_evfd+0x30/0x30 do_syscall_64+0x5b/0x1c0 entry_SYSCALL_64_after_hwframe+0x44/0xa9 RIP: 0033:0x7f4d116cb1ed which is due to the 'wqe' and 'worker' allocation being node affine. But it isn't valid to call the node affine allocation if the node isn't online. Setup structures for even offline nodes, as usual, but skip them in terms of thread setup to not waste resources. If the node isn't online, just alloc memory with NUMA_NO_NODE. Reported-by: Glauber Costa <glauber@scylladb.com> Tested-by: Glauber Costa <glauber@scylladb.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-02-11 13:30:06 +00:00
if (!node_online(node))
continue;
io_wq: Make io_wqe::lock a raw_spinlock_t During a context switch the scheduler invokes wq_worker_sleeping() with disabled preemption. Disabling preemption is needed because it protects access to `worker->sleeping'. As an optimisation it avoids invoking schedule() within the schedule path as part of possible wake up (thus preempt_enable_no_resched() afterwards). The io-wq has been added to the mix in the same section with disabled preemption. This breaks on PREEMPT_RT because io_wq_worker_sleeping() acquires a spinlock_t. Also within the schedule() the spinlock_t must be acquired after tsk_is_pi_blocked() otherwise it will block on the sleeping lock again while scheduling out. While playing with `io_uring-bench' I didn't notice a significant latency spike after converting io_wqe::lock to a raw_spinlock_t. The latency was more or less the same. In order to keep the spinlock_t it would have to be moved after the tsk_is_pi_blocked() check which would introduce a branch instruction into the hot path. The lock is used to maintain the `work_list' and wakes one task up at most. Should io_wqe_cancel_pending_work() cause latency spikes, while searching for a specific item, then it would need to drop the lock during iterations. revert_creds() is also invoked under the lock. According to debug cred::non_rcu is 0. Otherwise it should be moved outside of the locked section because put_cred_rcu()->free_uid() acquires a sleeping lock. Convert io_wqe::lock to a raw_spinlock_t.c Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-01 08:41:46 +00:00
raw_spin_lock_irq(&wqe->lock);
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
if (io_wqe_need_worker(wqe, IO_WQ_ACCT_BOUND))
fork_worker[IO_WQ_ACCT_BOUND] = true;
if (io_wqe_need_worker(wqe, IO_WQ_ACCT_UNBOUND))
fork_worker[IO_WQ_ACCT_UNBOUND] = true;
io_wq: Make io_wqe::lock a raw_spinlock_t During a context switch the scheduler invokes wq_worker_sleeping() with disabled preemption. Disabling preemption is needed because it protects access to `worker->sleeping'. As an optimisation it avoids invoking schedule() within the schedule path as part of possible wake up (thus preempt_enable_no_resched() afterwards). The io-wq has been added to the mix in the same section with disabled preemption. This breaks on PREEMPT_RT because io_wq_worker_sleeping() acquires a spinlock_t. Also within the schedule() the spinlock_t must be acquired after tsk_is_pi_blocked() otherwise it will block on the sleeping lock again while scheduling out. While playing with `io_uring-bench' I didn't notice a significant latency spike after converting io_wqe::lock to a raw_spinlock_t. The latency was more or less the same. In order to keep the spinlock_t it would have to be moved after the tsk_is_pi_blocked() check which would introduce a branch instruction into the hot path. The lock is used to maintain the `work_list' and wakes one task up at most. Should io_wqe_cancel_pending_work() cause latency spikes, while searching for a specific item, then it would need to drop the lock during iterations. revert_creds() is also invoked under the lock. According to debug cred::non_rcu is 0. Otherwise it should be moved outside of the locked section because put_cred_rcu()->free_uid() acquires a sleeping lock. Convert io_wqe::lock to a raw_spinlock_t.c Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-01 08:41:46 +00:00
raw_spin_unlock_irq(&wqe->lock);
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
if (fork_worker[IO_WQ_ACCT_BOUND])
create_io_worker(wq, wqe, IO_WQ_ACCT_BOUND);
if (fork_worker[IO_WQ_ACCT_UNBOUND])
create_io_worker(wq, wqe, IO_WQ_ACCT_UNBOUND);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
}
set_current_state(TASK_INTERRUPTIBLE);
schedule_timeout(HZ);
}
io_uring: use io-wq manager as backup task if task is exiting If the original task is (or has) exited, then the task work will not get queued properly. Allow for using the io-wq manager task to queue this work for execution, and ensure that the io-wq manager notices and runs this work if woken up (or exiting). Reported-by: Dan Melnic <dmm@fb.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-04-03 17:26:26 +00:00
if (current->task_works)
task_work_run();
io-wq: fix use-after-free in io_wq_worker_running The smart syzbot has found a reproducer for the following issue: ================================================================== BUG: KASAN: use-after-free in instrument_atomic_write include/linux/instrumented.h:71 [inline] BUG: KASAN: use-after-free in atomic_inc include/asm-generic/atomic-instrumented.h:240 [inline] BUG: KASAN: use-after-free in io_wqe_inc_running fs/io-wq.c:301 [inline] BUG: KASAN: use-after-free in io_wq_worker_running+0xde/0x110 fs/io-wq.c:613 Write of size 4 at addr ffff8882183db08c by task io_wqe_worker-0/7771 CPU: 0 PID: 7771 Comm: io_wqe_worker-0 Not tainted 5.9.0-rc4-syzkaller #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011 Call Trace: __dump_stack lib/dump_stack.c:77 [inline] dump_stack+0x198/0x1fd lib/dump_stack.c:118 print_address_description.constprop.0.cold+0xae/0x497 mm/kasan/report.c:383 __kasan_report mm/kasan/report.c:513 [inline] kasan_report.cold+0x1f/0x37 mm/kasan/report.c:530 check_memory_region_inline mm/kasan/generic.c:186 [inline] check_memory_region+0x13d/0x180 mm/kasan/generic.c:192 instrument_atomic_write include/linux/instrumented.h:71 [inline] atomic_inc include/asm-generic/atomic-instrumented.h:240 [inline] io_wqe_inc_running fs/io-wq.c:301 [inline] io_wq_worker_running+0xde/0x110 fs/io-wq.c:613 schedule_timeout+0x148/0x250 kernel/time/timer.c:1879 io_wqe_worker+0x517/0x10e0 fs/io-wq.c:580 kthread+0x3b5/0x4a0 kernel/kthread.c:292 ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:294 Allocated by task 7768: kasan_save_stack+0x1b/0x40 mm/kasan/common.c:48 kasan_set_track mm/kasan/common.c:56 [inline] __kasan_kmalloc.constprop.0+0xbf/0xd0 mm/kasan/common.c:461 kmem_cache_alloc_node_trace+0x17b/0x3f0 mm/slab.c:3594 kmalloc_node include/linux/slab.h:572 [inline] kzalloc_node include/linux/slab.h:677 [inline] io_wq_create+0x57b/0xa10 fs/io-wq.c:1064 io_init_wq_offload fs/io_uring.c:7432 [inline] io_sq_offload_start fs/io_uring.c:7504 [inline] io_uring_create fs/io_uring.c:8625 [inline] io_uring_setup+0x1836/0x28e0 fs/io_uring.c:8694 do_syscall_64+0x2d/0x70 arch/x86/entry/common.c:46 entry_SYSCALL_64_after_hwframe+0x44/0xa9 Freed by task 21: kasan_save_stack+0x1b/0x40 mm/kasan/common.c:48 kasan_set_track+0x1c/0x30 mm/kasan/common.c:56 kasan_set_free_info+0x1b/0x30 mm/kasan/generic.c:355 __kasan_slab_free+0xd8/0x120 mm/kasan/common.c:422 __cache_free mm/slab.c:3418 [inline] kfree+0x10e/0x2b0 mm/slab.c:3756 __io_wq_destroy fs/io-wq.c:1138 [inline] io_wq_destroy+0x2af/0x460 fs/io-wq.c:1146 io_finish_async fs/io_uring.c:6836 [inline] io_ring_ctx_free fs/io_uring.c:7870 [inline] io_ring_exit_work+0x1e4/0x6d0 fs/io_uring.c:7954 process_one_work+0x94c/0x1670 kernel/workqueue.c:2269 worker_thread+0x64c/0x1120 kernel/workqueue.c:2415 kthread+0x3b5/0x4a0 kernel/kthread.c:292 ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:294 The buggy address belongs to the object at ffff8882183db000 which belongs to the cache kmalloc-1k of size 1024 The buggy address is located 140 bytes inside of 1024-byte region [ffff8882183db000, ffff8882183db400) The buggy address belongs to the page: page:000000009bada22b refcount:1 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0x2183db flags: 0x57ffe0000000200(slab) raw: 057ffe0000000200 ffffea0008604c48 ffffea00086a8648 ffff8880aa040700 raw: 0000000000000000 ffff8882183db000 0000000100000002 0000000000000000 page dumped because: kasan: bad access detected Memory state around the buggy address: ffff8882183daf80: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ffff8882183db000: fa fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb >ffff8882183db080: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ^ ffff8882183db100: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ffff8882183db180: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ================================================================== which is down to the comment below, /* all workers gone, wq exit can proceed */ if (!nr_workers && refcount_dec_and_test(&wqe->wq->refs)) complete(&wqe->wq->done); because there might be multiple cases of wqe in a wq and we would wait for every worker in every wqe to go home before releasing wq's resources on destroying. To that end, rework wq's refcount by making it independent of the tracking of workers because after all they are two different things, and keeping it balanced when workers come and go. Note the manager kthread, like other workers, now holds a grab to wq during its lifetime. Finally to help destroy wq, check IO_WQ_BIT_EXIT upon creating worker and do nothing for exiting wq. Cc: stable@vger.kernel.org # v5.5+ Reported-by: syzbot+45fa0a195b941764e0f0@syzkaller.appspotmail.com Reported-by: syzbot+9af99580130003da82b1@syzkaller.appspotmail.com Cc: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Hillf Danton <hdanton@sina.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-26 13:26:55 +00:00
out:
if (refcount_dec_and_test(&wq->refs)) {
io-wq: wait for io_wq_create() to setup necessary workers We currently have a race where if setup is really slow, we can be calling io_wq_destroy() before we're done setting up. This will cause the caller to get stuck waiting for the manager to set things up, but the manager already exited. Fix this by doing a sync setup of the manager. This also fixes the case where if we failed creating workers, we'd also get stuck. In practice this race window was really small, as we already wait for the manager to start. Hence someone would have to call io_wq_destroy() after the task has started, but before it started the first loop. The reported test case forked tons of these, which is why it became an issue. Reported-by: syzbot+0f1cc17f85154f400465@syzkaller.appspotmail.com Fixes: 771b53d033e8 ("io-wq: small threadpool implementation for io_uring") Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-19 15:37:07 +00:00
complete(&wq->done);
io-wq: fix use-after-free in io_wq_worker_running The smart syzbot has found a reproducer for the following issue: ================================================================== BUG: KASAN: use-after-free in instrument_atomic_write include/linux/instrumented.h:71 [inline] BUG: KASAN: use-after-free in atomic_inc include/asm-generic/atomic-instrumented.h:240 [inline] BUG: KASAN: use-after-free in io_wqe_inc_running fs/io-wq.c:301 [inline] BUG: KASAN: use-after-free in io_wq_worker_running+0xde/0x110 fs/io-wq.c:613 Write of size 4 at addr ffff8882183db08c by task io_wqe_worker-0/7771 CPU: 0 PID: 7771 Comm: io_wqe_worker-0 Not tainted 5.9.0-rc4-syzkaller #0 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011 Call Trace: __dump_stack lib/dump_stack.c:77 [inline] dump_stack+0x198/0x1fd lib/dump_stack.c:118 print_address_description.constprop.0.cold+0xae/0x497 mm/kasan/report.c:383 __kasan_report mm/kasan/report.c:513 [inline] kasan_report.cold+0x1f/0x37 mm/kasan/report.c:530 check_memory_region_inline mm/kasan/generic.c:186 [inline] check_memory_region+0x13d/0x180 mm/kasan/generic.c:192 instrument_atomic_write include/linux/instrumented.h:71 [inline] atomic_inc include/asm-generic/atomic-instrumented.h:240 [inline] io_wqe_inc_running fs/io-wq.c:301 [inline] io_wq_worker_running+0xde/0x110 fs/io-wq.c:613 schedule_timeout+0x148/0x250 kernel/time/timer.c:1879 io_wqe_worker+0x517/0x10e0 fs/io-wq.c:580 kthread+0x3b5/0x4a0 kernel/kthread.c:292 ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:294 Allocated by task 7768: kasan_save_stack+0x1b/0x40 mm/kasan/common.c:48 kasan_set_track mm/kasan/common.c:56 [inline] __kasan_kmalloc.constprop.0+0xbf/0xd0 mm/kasan/common.c:461 kmem_cache_alloc_node_trace+0x17b/0x3f0 mm/slab.c:3594 kmalloc_node include/linux/slab.h:572 [inline] kzalloc_node include/linux/slab.h:677 [inline] io_wq_create+0x57b/0xa10 fs/io-wq.c:1064 io_init_wq_offload fs/io_uring.c:7432 [inline] io_sq_offload_start fs/io_uring.c:7504 [inline] io_uring_create fs/io_uring.c:8625 [inline] io_uring_setup+0x1836/0x28e0 fs/io_uring.c:8694 do_syscall_64+0x2d/0x70 arch/x86/entry/common.c:46 entry_SYSCALL_64_after_hwframe+0x44/0xa9 Freed by task 21: kasan_save_stack+0x1b/0x40 mm/kasan/common.c:48 kasan_set_track+0x1c/0x30 mm/kasan/common.c:56 kasan_set_free_info+0x1b/0x30 mm/kasan/generic.c:355 __kasan_slab_free+0xd8/0x120 mm/kasan/common.c:422 __cache_free mm/slab.c:3418 [inline] kfree+0x10e/0x2b0 mm/slab.c:3756 __io_wq_destroy fs/io-wq.c:1138 [inline] io_wq_destroy+0x2af/0x460 fs/io-wq.c:1146 io_finish_async fs/io_uring.c:6836 [inline] io_ring_ctx_free fs/io_uring.c:7870 [inline] io_ring_exit_work+0x1e4/0x6d0 fs/io_uring.c:7954 process_one_work+0x94c/0x1670 kernel/workqueue.c:2269 worker_thread+0x64c/0x1120 kernel/workqueue.c:2415 kthread+0x3b5/0x4a0 kernel/kthread.c:292 ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:294 The buggy address belongs to the object at ffff8882183db000 which belongs to the cache kmalloc-1k of size 1024 The buggy address is located 140 bytes inside of 1024-byte region [ffff8882183db000, ffff8882183db400) The buggy address belongs to the page: page:000000009bada22b refcount:1 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0x2183db flags: 0x57ffe0000000200(slab) raw: 057ffe0000000200 ffffea0008604c48 ffffea00086a8648 ffff8880aa040700 raw: 0000000000000000 ffff8882183db000 0000000100000002 0000000000000000 page dumped because: kasan: bad access detected Memory state around the buggy address: ffff8882183daf80: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ffff8882183db000: fa fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb >ffff8882183db080: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ^ ffff8882183db100: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ffff8882183db180: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb ================================================================== which is down to the comment below, /* all workers gone, wq exit can proceed */ if (!nr_workers && refcount_dec_and_test(&wqe->wq->refs)) complete(&wqe->wq->done); because there might be multiple cases of wqe in a wq and we would wait for every worker in every wqe to go home before releasing wq's resources on destroying. To that end, rework wq's refcount by making it independent of the tracking of workers because after all they are two different things, and keeping it balanced when workers come and go. Note the manager kthread, like other workers, now holds a grab to wq during its lifetime. Finally to help destroy wq, check IO_WQ_BIT_EXIT upon creating worker and do nothing for exiting wq. Cc: stable@vger.kernel.org # v5.5+ Reported-by: syzbot+45fa0a195b941764e0f0@syzkaller.appspotmail.com Reported-by: syzbot+9af99580130003da82b1@syzkaller.appspotmail.com Cc: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Hillf Danton <hdanton@sina.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-26 13:26:55 +00:00
return 0;
}
/* if ERROR is set and we get here, we have workers to wake */
if (test_bit(IO_WQ_BIT_ERROR, &wq->state)) {
rcu_read_lock();
for_each_node(node)
io_wq_for_each_worker(wq->wqes[node], io_wq_worker_wake, NULL);
rcu_read_unlock();
}
io-wq: wait for io_wq_create() to setup necessary workers We currently have a race where if setup is really slow, we can be calling io_wq_destroy() before we're done setting up. This will cause the caller to get stuck waiting for the manager to set things up, but the manager already exited. Fix this by doing a sync setup of the manager. This also fixes the case where if we failed creating workers, we'd also get stuck. In practice this race window was really small, as we already wait for the manager to start. Hence someone would have to call io_wq_destroy() after the task has started, but before it started the first loop. The reported test case forked tons of these, which is why it became an issue. Reported-by: syzbot+0f1cc17f85154f400465@syzkaller.appspotmail.com Fixes: 771b53d033e8 ("io-wq: small threadpool implementation for io_uring") Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-19 15:37:07 +00:00
return 0;
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
}
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
static bool io_wq_can_queue(struct io_wqe *wqe, struct io_wqe_acct *acct,
struct io_wq_work *work)
{
bool free_worker;
if (!(work->flags & IO_WQ_WORK_UNBOUND))
return true;
if (atomic_read(&acct->nr_running))
return true;
rcu_read_lock();
io-wq: remove now redundant struct io_wq_nulls_list Since we don't iterate these lists anymore after commit: e61df66c69b1 ("io-wq: ensure free/busy list browsing see all items") we don't need to retain the nulls value we use for them. That means it's pretty pointless to wrap the hlist_nulls_head in a structure, so get rid of it. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-14 15:00:41 +00:00
free_worker = !hlist_nulls_empty(&wqe->free_list);
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
rcu_read_unlock();
if (free_worker)
return true;
if (atomic_read(&wqe->wq->user->processes) >= acct->max_workers &&
!(capable(CAP_SYS_RESOURCE) || capable(CAP_SYS_ADMIN)))
return false;
return true;
}
io_uring/io-wq: forward submission ref to async First it changes io-wq interfaces. It replaces {get,put}_work() with free_work(), which guaranteed to be called exactly once. It also enforces free_work() callback to be non-NULL. io_uring follows the changes and instead of putting a submission reference in io_put_req_async_completion(), it will be done in io_free_work(). As removes io_get_work() with corresponding refcount_inc(), the ref balance is maintained. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-04 13:14:12 +00:00
static void io_run_cancel(struct io_wq_work *work, struct io_wqe *wqe)
io-wq: fix IO_WQ_WORK_NO_CANCEL cancellation To cancel a work, io-wq sets IO_WQ_WORK_CANCEL and executes the callback. However, IO_WQ_WORK_NO_CANCEL works will just execute and may return next work, which will be ignored and lost. Cancel the whole link. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-01 16:18:19 +00:00
{
io_uring/io-wq: forward submission ref to async First it changes io-wq interfaces. It replaces {get,put}_work() with free_work(), which guaranteed to be called exactly once. It also enforces free_work() callback to be non-NULL. io_uring follows the changes and instead of putting a submission reference in io_put_req_async_completion(), it will be done in io_free_work(). As removes io_get_work() with corresponding refcount_inc(), the ref balance is maintained. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-04 13:14:12 +00:00
struct io_wq *wq = wqe->wq;
io-wq: fix IO_WQ_WORK_NO_CANCEL cancellation To cancel a work, io-wq sets IO_WQ_WORK_CANCEL and executes the callback. However, IO_WQ_WORK_NO_CANCEL works will just execute and may return next work, which will be ignored and lost. Cancel the whole link. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-01 16:18:19 +00:00
do {
struct io_wq_work *old_work = work;
work->flags |= IO_WQ_WORK_CANCEL;
io-wq: return next work from ->do_work() directly It's easier to return next work from ->do_work() than having an in-out argument. Looks nicer and easier to compile. Also, merge io_wq_assign_next() into its only user. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-06-25 15:20:54 +00:00
work = wq->do_work(work);
io_uring/io-wq: forward submission ref to async First it changes io-wq interfaces. It replaces {get,put}_work() with free_work(), which guaranteed to be called exactly once. It also enforces free_work() callback to be non-NULL. io_uring follows the changes and instead of putting a submission reference in io_put_req_async_completion(), it will be done in io_free_work(). As removes io_get_work() with corresponding refcount_inc(), the ref balance is maintained. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-04 13:14:12 +00:00
wq->free_work(old_work);
io-wq: fix IO_WQ_WORK_NO_CANCEL cancellation To cancel a work, io-wq sets IO_WQ_WORK_CANCEL and executes the callback. However, IO_WQ_WORK_NO_CANCEL works will just execute and may return next work, which will be ignored and lost. Cancel the whole link. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-01 16:18:19 +00:00
} while (work);
}
io-wq: handle hashed writes in chains We always punt async buffered writes to an io-wq helper, as the core kernel does not have IOCB_NOWAIT support for that. Most buffered async writes complete very quickly, as it's just a copy operation. This means that doing multiple locking roundtrips on the shared wqe lock for each buffered write is wasteful. Additionally, buffered writes are hashed work items, which means that any buffered write to a given file is serialized. Keep identicaly hashed work items contiguously in @wqe->work_list, and track a tail for each hash bucket. On dequeue of a hashed item, splice all of the same hash in one go using the tracked tail. Until the batch is done, the caller doesn't have to synchronize with the wqe or worker locks again. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-23 19:57:22 +00:00
static void io_wqe_insert_work(struct io_wqe *wqe, struct io_wq_work *work)
{
unsigned int hash;
struct io_wq_work *tail;
if (!io_wq_is_hashed(work)) {
append:
wq_list_add_tail(&work->list, &wqe->work_list);
return;
}
hash = io_get_work_hash(work);
tail = wqe->hash_tail[hash];
wqe->hash_tail[hash] = work;
if (!tail)
goto append;
wq_list_add_after(&work->list, &tail->list, &wqe->work_list);
}
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
static void io_wqe_enqueue(struct io_wqe *wqe, struct io_wq_work *work)
{
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
struct io_wqe_acct *acct = io_work_get_acct(wqe, work);
io-wq: support concurrent non-blocking work io-wq assumes that work will complete fast (and not block), so it doesn't create a new worker when work is enqueued, if we already have at least one worker running. This is done on the assumption that if work is running, then it will complete fast. Add an option to force io-wq to fork a new worker for work queued. This is signaled by setting IO_WQ_WORK_CONCURRENT on the work item. For that case, io-wq will create a new worker, even though workers are already running. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-12-17 15:46:33 +00:00
int work_flags;
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
unsigned long flags;
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
/*
* Do early check to see if we need a new unbound worker, and if we do,
* if we're allowed to do so. This isn't 100% accurate as there's a
* gap between this check and incrementing the value, but that's OK.
* It's close enough to not be an issue, fork() has the same delay.
*/
if (unlikely(!io_wq_can_queue(wqe, acct, work))) {
io_uring/io-wq: forward submission ref to async First it changes io-wq interfaces. It replaces {get,put}_work() with free_work(), which guaranteed to be called exactly once. It also enforces free_work() callback to be non-NULL. io_uring follows the changes and instead of putting a submission reference in io_put_req_async_completion(), it will be done in io_free_work(). As removes io_get_work() with corresponding refcount_inc(), the ref balance is maintained. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-04 13:14:12 +00:00
io_run_cancel(work, wqe);
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
return;
}
io-wq: support concurrent non-blocking work io-wq assumes that work will complete fast (and not block), so it doesn't create a new worker when work is enqueued, if we already have at least one worker running. This is done on the assumption that if work is running, then it will complete fast. Add an option to force io-wq to fork a new worker for work queued. This is signaled by setting IO_WQ_WORK_CONCURRENT on the work item. For that case, io-wq will create a new worker, even though workers are already running. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-12-17 15:46:33 +00:00
work_flags = work->flags;
io_wq: Make io_wqe::lock a raw_spinlock_t During a context switch the scheduler invokes wq_worker_sleeping() with disabled preemption. Disabling preemption is needed because it protects access to `worker->sleeping'. As an optimisation it avoids invoking schedule() within the schedule path as part of possible wake up (thus preempt_enable_no_resched() afterwards). The io-wq has been added to the mix in the same section with disabled preemption. This breaks on PREEMPT_RT because io_wq_worker_sleeping() acquires a spinlock_t. Also within the schedule() the spinlock_t must be acquired after tsk_is_pi_blocked() otherwise it will block on the sleeping lock again while scheduling out. While playing with `io_uring-bench' I didn't notice a significant latency spike after converting io_wqe::lock to a raw_spinlock_t. The latency was more or less the same. In order to keep the spinlock_t it would have to be moved after the tsk_is_pi_blocked() check which would introduce a branch instruction into the hot path. The lock is used to maintain the `work_list' and wakes one task up at most. Should io_wqe_cancel_pending_work() cause latency spikes, while searching for a specific item, then it would need to drop the lock during iterations. revert_creds() is also invoked under the lock. According to debug cred::non_rcu is 0. Otherwise it should be moved outside of the locked section because put_cred_rcu()->free_uid() acquires a sleeping lock. Convert io_wqe::lock to a raw_spinlock_t.c Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-01 08:41:46 +00:00
raw_spin_lock_irqsave(&wqe->lock, flags);
io-wq: handle hashed writes in chains We always punt async buffered writes to an io-wq helper, as the core kernel does not have IOCB_NOWAIT support for that. Most buffered async writes complete very quickly, as it's just a copy operation. This means that doing multiple locking roundtrips on the shared wqe lock for each buffered write is wasteful. Additionally, buffered writes are hashed work items, which means that any buffered write to a given file is serialized. Keep identicaly hashed work items contiguously in @wqe->work_list, and track a tail for each hash bucket. On dequeue of a hashed item, splice all of the same hash in one go using the tracked tail. Until the batch is done, the caller doesn't have to synchronize with the wqe or worker locks again. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-23 19:57:22 +00:00
io_wqe_insert_work(wqe, work);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
wqe->flags &= ~IO_WQE_FLAG_STALLED;
io_wq: Make io_wqe::lock a raw_spinlock_t During a context switch the scheduler invokes wq_worker_sleeping() with disabled preemption. Disabling preemption is needed because it protects access to `worker->sleeping'. As an optimisation it avoids invoking schedule() within the schedule path as part of possible wake up (thus preempt_enable_no_resched() afterwards). The io-wq has been added to the mix in the same section with disabled preemption. This breaks on PREEMPT_RT because io_wq_worker_sleeping() acquires a spinlock_t. Also within the schedule() the spinlock_t must be acquired after tsk_is_pi_blocked() otherwise it will block on the sleeping lock again while scheduling out. While playing with `io_uring-bench' I didn't notice a significant latency spike after converting io_wqe::lock to a raw_spinlock_t. The latency was more or less the same. In order to keep the spinlock_t it would have to be moved after the tsk_is_pi_blocked() check which would introduce a branch instruction into the hot path. The lock is used to maintain the `work_list' and wakes one task up at most. Should io_wqe_cancel_pending_work() cause latency spikes, while searching for a specific item, then it would need to drop the lock during iterations. revert_creds() is also invoked under the lock. According to debug cred::non_rcu is 0. Otherwise it should be moved outside of the locked section because put_cred_rcu()->free_uid() acquires a sleeping lock. Convert io_wqe::lock to a raw_spinlock_t.c Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-01 08:41:46 +00:00
raw_spin_unlock_irqrestore(&wqe->lock, flags);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
io-wq: support concurrent non-blocking work io-wq assumes that work will complete fast (and not block), so it doesn't create a new worker when work is enqueued, if we already have at least one worker running. This is done on the assumption that if work is running, then it will complete fast. Add an option to force io-wq to fork a new worker for work queued. This is signaled by setting IO_WQ_WORK_CONCURRENT on the work item. For that case, io-wq will create a new worker, even though workers are already running. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-12-17 15:46:33 +00:00
if ((work_flags & IO_WQ_WORK_CONCURRENT) ||
!atomic_read(&acct->nr_running))
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
io_wqe_wake_worker(wqe, acct);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
}
void io_wq_enqueue(struct io_wq *wq, struct io_wq_work *work)
{
struct io_wqe *wqe = wq->wqes[numa_node_id()];
io_wqe_enqueue(wqe, work);
}
/*
io-wq: split hashing and enqueueing It's a preparation patch removing io_wq_enqueue_hashed(), which now should be done by io_wq_hash_work() + io_wq_enqueue(). Also, set hash value for dependant works, and do it as late as possible, because req->file can be unavailable before. This hash will be ignored by io-wq. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-13 21:31:04 +00:00
* Work items that hash to the same value will not be done in parallel.
* Used to limit concurrent writes, generally hashed by inode.
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
*/
io-wq: split hashing and enqueueing It's a preparation patch removing io_wq_enqueue_hashed(), which now should be done by io_wq_hash_work() + io_wq_enqueue(). Also, set hash value for dependant works, and do it as late as possible, because req->file can be unavailable before. This hash will be ignored by io-wq. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-13 21:31:04 +00:00
void io_wq_hash_work(struct io_wq_work *work, void *val)
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
{
io-wq: split hashing and enqueueing It's a preparation patch removing io_wq_enqueue_hashed(), which now should be done by io_wq_hash_work() + io_wq_enqueue(). Also, set hash value for dependant works, and do it as late as possible, because req->file can be unavailable before. This hash will be ignored by io-wq. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-13 21:31:04 +00:00
unsigned int bit;
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
bit = hash_ptr(val, IO_WQ_HASH_ORDER);
work->flags |= (IO_WQ_WORK_HASHED | (bit << IO_WQ_HASH_SHIFT));
}
io_uring: support for generic async request cancel This adds support for IORING_OP_ASYNC_CANCEL, which will attempt to cancel requests that have been punted to async context and are now in-flight. This works for regular read/write requests to files, as long as they haven't been started yet. For socket based IO (or things like accept4(2)), we can cancel work that is already running as well. To cancel a request, the sqe must have ->addr set to the user_data of the request it wishes to cancel. If the request is cancelled successfully, the original request is completed with -ECANCELED and the cancel request is completed with a result of 0. If the request was already running, the original may or may not complete in error. The cancel request will complete with -EALREADY for that case. And finally, if the request to cancel wasn't found, the cancel request is completed with -ENOENT. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-29 03:49:21 +00:00
struct io_cb_cancel_data {
io-wq: remove duplicated cancel code Deduplicate cancellation parts, as many of them looks the same, as do e.g. - io_wqe_cancel_cb_work() and io_wqe_cancel_work() - io_wq_worker_cancel() and io_work_cancel() Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-06 22:15:39 +00:00
work_cancel_fn *fn;
void *data;
io-wq: add an option to cancel all matched reqs This adds support for cancelling all io-wq works matching a predicate. It isn't used yet, so no change in observable behaviour. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-06-15 07:24:03 +00:00
int nr_running;
int nr_pending;
bool cancel_all;
io_uring: support for generic async request cancel This adds support for IORING_OP_ASYNC_CANCEL, which will attempt to cancel requests that have been punted to async context and are now in-flight. This works for regular read/write requests to files, as long as they haven't been started yet. For socket based IO (or things like accept4(2)), we can cancel work that is already running as well. To cancel a request, the sqe must have ->addr set to the user_data of the request it wishes to cancel. If the request is cancelled successfully, the original request is completed with -ECANCELED and the cancel request is completed with a result of 0. If the request was already running, the original may or may not complete in error. The cancel request will complete with -EALREADY for that case. And finally, if the request to cancel wasn't found, the cancel request is completed with -ENOENT. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-29 03:49:21 +00:00
};
io-wq: remove duplicated cancel code Deduplicate cancellation parts, as many of them looks the same, as do e.g. - io_wqe_cancel_cb_work() and io_wqe_cancel_work() - io_wq_worker_cancel() and io_work_cancel() Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-06 22:15:39 +00:00
static bool io_wq_worker_cancel(struct io_worker *worker, void *data)
io_uring: support for generic async request cancel This adds support for IORING_OP_ASYNC_CANCEL, which will attempt to cancel requests that have been punted to async context and are now in-flight. This works for regular read/write requests to files, as long as they haven't been started yet. For socket based IO (or things like accept4(2)), we can cancel work that is already running as well. To cancel a request, the sqe must have ->addr set to the user_data of the request it wishes to cancel. If the request is cancelled successfully, the original request is completed with -ECANCELED and the cancel request is completed with a result of 0. If the request was already running, the original may or may not complete in error. The cancel request will complete with -EALREADY for that case. And finally, if the request to cancel wasn't found, the cancel request is completed with -ENOENT. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-29 03:49:21 +00:00
{
io-wq: remove duplicated cancel code Deduplicate cancellation parts, as many of them looks the same, as do e.g. - io_wqe_cancel_cb_work() and io_wqe_cancel_work() - io_wq_worker_cancel() and io_work_cancel() Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-06 22:15:39 +00:00
struct io_cb_cancel_data *match = data;
io-wq: use proper nesting IRQ disabling spinlocks for cancel We don't know what context we'll be called in for cancel, it could very well be with IRQs disabled already. Use the IRQ saving variants of the locking primitives. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-05 20:51:51 +00:00
unsigned long flags;
io_uring: support for generic async request cancel This adds support for IORING_OP_ASYNC_CANCEL, which will attempt to cancel requests that have been punted to async context and are now in-flight. This works for regular read/write requests to files, as long as they haven't been started yet. For socket based IO (or things like accept4(2)), we can cancel work that is already running as well. To cancel a request, the sqe must have ->addr set to the user_data of the request it wishes to cancel. If the request is cancelled successfully, the original request is completed with -ECANCELED and the cancel request is completed with a result of 0. If the request was already running, the original may or may not complete in error. The cancel request will complete with -EALREADY for that case. And finally, if the request to cancel wasn't found, the cancel request is completed with -ENOENT. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-29 03:49:21 +00:00
/*
* Hold the lock to avoid ->cur_work going out of scope, caller
io-wq: ensure we have a stable view of ->cur_work for cancellations worker->cur_work is currently protected by the lock of the wqe that the worker belongs to. When we send a signal to a worker, we need a stable view of ->cur_work, so we need to hold that lock. But this doesn't work so well, since we have the opposite order potentially on queueing work. If POLL_ADD is used with a signalfd, then io_poll_wake() is called with the signal lock, and that sometimes needs to insert work items. Add a specific worker lock that protects the current work item. Then we can guarantee that the task we're sending a signal is currently processing the exact work we think it is. Reported-by: Paul E. McKenney <paulmck@kernel.org> Reviewed-by: Paul E. McKenney <paulmck@kernel.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-13 16:43:34 +00:00
* may dereference the passed in work.
io_uring: support for generic async request cancel This adds support for IORING_OP_ASYNC_CANCEL, which will attempt to cancel requests that have been punted to async context and are now in-flight. This works for regular read/write requests to files, as long as they haven't been started yet. For socket based IO (or things like accept4(2)), we can cancel work that is already running as well. To cancel a request, the sqe must have ->addr set to the user_data of the request it wishes to cancel. If the request is cancelled successfully, the original request is completed with -ECANCELED and the cancel request is completed with a result of 0. If the request was already running, the original may or may not complete in error. The cancel request will complete with -EALREADY for that case. And finally, if the request to cancel wasn't found, the cancel request is completed with -ENOENT. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-29 03:49:21 +00:00
*/
io-wq: ensure we have a stable view of ->cur_work for cancellations worker->cur_work is currently protected by the lock of the wqe that the worker belongs to. When we send a signal to a worker, we need a stable view of ->cur_work, so we need to hold that lock. But this doesn't work so well, since we have the opposite order potentially on queueing work. If POLL_ADD is used with a signalfd, then io_poll_wake() is called with the signal lock, and that sometimes needs to insert work items. Add a specific worker lock that protects the current work item. Then we can guarantee that the task we're sending a signal is currently processing the exact work we think it is. Reported-by: Paul E. McKenney <paulmck@kernel.org> Reviewed-by: Paul E. McKenney <paulmck@kernel.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-13 16:43:34 +00:00
spin_lock_irqsave(&worker->lock, flags);
io_uring: support for generic async request cancel This adds support for IORING_OP_ASYNC_CANCEL, which will attempt to cancel requests that have been punted to async context and are now in-flight. This works for regular read/write requests to files, as long as they haven't been started yet. For socket based IO (or things like accept4(2)), we can cancel work that is already running as well. To cancel a request, the sqe must have ->addr set to the user_data of the request it wishes to cancel. If the request is cancelled successfully, the original request is completed with -ECANCELED and the cancel request is completed with a result of 0. If the request was already running, the original may or may not complete in error. The cancel request will complete with -EALREADY for that case. And finally, if the request to cancel wasn't found, the cancel request is completed with -ENOENT. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-29 03:49:21 +00:00
if (worker->cur_work &&
io-wq: add support for uncancellable work Not all work can be cancelled, some of it we may need to guarantee that it runs to completion. Allow the caller to set IO_WQ_WORK_NO_CANCEL on work that must not be cancelled. Note that the caller work function must also check for IO_WQ_WORK_NO_CANCEL on work that is marked IO_WQ_WORK_CANCEL. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-12-12 02:29:43 +00:00
!(worker->cur_work->flags & IO_WQ_WORK_NO_CANCEL) &&
io-wq: remove duplicated cancel code Deduplicate cancellation parts, as many of them looks the same, as do e.g. - io_wqe_cancel_cb_work() and io_wqe_cancel_work() - io_wq_worker_cancel() and io_work_cancel() Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-06 22:15:39 +00:00
match->fn(worker->cur_work, match->data)) {
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
send_sig(SIGINT, worker->task, 1);
io-wq: add an option to cancel all matched reqs This adds support for cancelling all io-wq works matching a predicate. It isn't used yet, so no change in observable behaviour. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-06-15 07:24:03 +00:00
match->nr_running++;
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
}
io-wq: ensure we have a stable view of ->cur_work for cancellations worker->cur_work is currently protected by the lock of the wqe that the worker belongs to. When we send a signal to a worker, we need a stable view of ->cur_work, so we need to hold that lock. But this doesn't work so well, since we have the opposite order potentially on queueing work. If POLL_ADD is used with a signalfd, then io_poll_wake() is called with the signal lock, and that sometimes needs to insert work items. Add a specific worker lock that protects the current work item. Then we can guarantee that the task we're sending a signal is currently processing the exact work we think it is. Reported-by: Paul E. McKenney <paulmck@kernel.org> Reviewed-by: Paul E. McKenney <paulmck@kernel.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-13 16:43:34 +00:00
spin_unlock_irqrestore(&worker->lock, flags);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
io-wq: add an option to cancel all matched reqs This adds support for cancelling all io-wq works matching a predicate. It isn't used yet, so no change in observable behaviour. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-06-15 07:24:03 +00:00
return match->nr_running && !match->cancel_all;
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
}
io-wq: fix hang after cancelling pending hashed work Don't forget to update wqe->hash_tail after cancelling a pending work item, if it was hashed. Cc: stable@vger.kernel.org # 5.7+ Reported-by: Dmitry Shulyak <yashulyak@gmail.com> Fixes: 86f3cd1b589a1 ("io-wq: handle hashed writes in chains") Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-08-23 17:33:10 +00:00
static inline void io_wqe_remove_pending(struct io_wqe *wqe,
struct io_wq_work *work,
struct io_wq_work_node *prev)
{
unsigned int hash = io_get_work_hash(work);
struct io_wq_work *prev_work = NULL;
if (io_wq_is_hashed(work) && work == wqe->hash_tail[hash]) {
if (prev)
prev_work = container_of(prev, struct io_wq_work, list);
if (prev_work && io_get_work_hash(prev_work) == hash)
wqe->hash_tail[hash] = prev_work;
else
wqe->hash_tail[hash] = NULL;
}
wq_list_del(&wqe->work_list, &work->list, prev);
}
io-wq: add an option to cancel all matched reqs This adds support for cancelling all io-wq works matching a predicate. It isn't used yet, so no change in observable behaviour. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-06-15 07:24:03 +00:00
static void io_wqe_cancel_pending_work(struct io_wqe *wqe,
io-wq: reorder cancellation pending -> running Go all over all pending lists and cancel works there, and only then try to match running requests. No functional changes here, just a preparation for bulk cancellation. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-06-15 07:24:02 +00:00
struct io_cb_cancel_data *match)
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
{
io-wq: shrink io_wq_work a bit Currently we're using 40 bytes for the io_wq_work structure, and 16 of those is the doubly link list node. We don't need doubly linked lists, we always add to tail to keep things ordered, and any other use case is list traversal with deletion. For the deletion case, we can easily support any node deletion by keeping track of the previous entry. This shrinks io_wq_work to 32 bytes, and subsequently io_kiock from io_uring to 216 to 208 bytes. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-26 18:59:32 +00:00
struct io_wq_work_node *node, *prev;
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
struct io_wq_work *work;
io-wq: use proper nesting IRQ disabling spinlocks for cancel We don't know what context we'll be called in for cancel, it could very well be with IRQs disabled already. Use the IRQ saving variants of the locking primitives. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-05 20:51:51 +00:00
unsigned long flags;
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
io-wq: add an option to cancel all matched reqs This adds support for cancelling all io-wq works matching a predicate. It isn't used yet, so no change in observable behaviour. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-06-15 07:24:03 +00:00
retry:
io_wq: Make io_wqe::lock a raw_spinlock_t During a context switch the scheduler invokes wq_worker_sleeping() with disabled preemption. Disabling preemption is needed because it protects access to `worker->sleeping'. As an optimisation it avoids invoking schedule() within the schedule path as part of possible wake up (thus preempt_enable_no_resched() afterwards). The io-wq has been added to the mix in the same section with disabled preemption. This breaks on PREEMPT_RT because io_wq_worker_sleeping() acquires a spinlock_t. Also within the schedule() the spinlock_t must be acquired after tsk_is_pi_blocked() otherwise it will block on the sleeping lock again while scheduling out. While playing with `io_uring-bench' I didn't notice a significant latency spike after converting io_wqe::lock to a raw_spinlock_t. The latency was more or less the same. In order to keep the spinlock_t it would have to be moved after the tsk_is_pi_blocked() check which would introduce a branch instruction into the hot path. The lock is used to maintain the `work_list' and wakes one task up at most. Should io_wqe_cancel_pending_work() cause latency spikes, while searching for a specific item, then it would need to drop the lock during iterations. revert_creds() is also invoked under the lock. According to debug cred::non_rcu is 0. Otherwise it should be moved outside of the locked section because put_cred_rcu()->free_uid() acquires a sleeping lock. Convert io_wqe::lock to a raw_spinlock_t.c Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-01 08:41:46 +00:00
raw_spin_lock_irqsave(&wqe->lock, flags);
io-wq: shrink io_wq_work a bit Currently we're using 40 bytes for the io_wq_work structure, and 16 of those is the doubly link list node. We don't need doubly linked lists, we always add to tail to keep things ordered, and any other use case is list traversal with deletion. For the deletion case, we can easily support any node deletion by keeping track of the previous entry. This shrinks io_wq_work to 32 bytes, and subsequently io_kiock from io_uring to 216 to 208 bytes. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-26 18:59:32 +00:00
wq_list_for_each(node, prev, &wqe->work_list) {
work = container_of(node, struct io_wq_work, list);
io-wq: add an option to cancel all matched reqs This adds support for cancelling all io-wq works matching a predicate. It isn't used yet, so no change in observable behaviour. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-06-15 07:24:03 +00:00
if (!match->fn(work, match->data))
continue;
io-wq: fix hang after cancelling pending hashed work Don't forget to update wqe->hash_tail after cancelling a pending work item, if it was hashed. Cc: stable@vger.kernel.org # 5.7+ Reported-by: Dmitry Shulyak <yashulyak@gmail.com> Fixes: 86f3cd1b589a1 ("io-wq: handle hashed writes in chains") Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-08-23 17:33:10 +00:00
io_wqe_remove_pending(wqe, work, prev);
io_wq: Make io_wqe::lock a raw_spinlock_t During a context switch the scheduler invokes wq_worker_sleeping() with disabled preemption. Disabling preemption is needed because it protects access to `worker->sleeping'. As an optimisation it avoids invoking schedule() within the schedule path as part of possible wake up (thus preempt_enable_no_resched() afterwards). The io-wq has been added to the mix in the same section with disabled preemption. This breaks on PREEMPT_RT because io_wq_worker_sleeping() acquires a spinlock_t. Also within the schedule() the spinlock_t must be acquired after tsk_is_pi_blocked() otherwise it will block on the sleeping lock again while scheduling out. While playing with `io_uring-bench' I didn't notice a significant latency spike after converting io_wqe::lock to a raw_spinlock_t. The latency was more or less the same. In order to keep the spinlock_t it would have to be moved after the tsk_is_pi_blocked() check which would introduce a branch instruction into the hot path. The lock is used to maintain the `work_list' and wakes one task up at most. Should io_wqe_cancel_pending_work() cause latency spikes, while searching for a specific item, then it would need to drop the lock during iterations. revert_creds() is also invoked under the lock. According to debug cred::non_rcu is 0. Otherwise it should be moved outside of the locked section because put_cred_rcu()->free_uid() acquires a sleeping lock. Convert io_wqe::lock to a raw_spinlock_t.c Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-01 08:41:46 +00:00
raw_spin_unlock_irqrestore(&wqe->lock, flags);
io-wq: add an option to cancel all matched reqs This adds support for cancelling all io-wq works matching a predicate. It isn't used yet, so no change in observable behaviour. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-06-15 07:24:03 +00:00
io_run_cancel(work, wqe);
match->nr_pending++;
if (!match->cancel_all)
return;
/* not safe to continue after unlock */
goto retry;
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
}
io_wq: Make io_wqe::lock a raw_spinlock_t During a context switch the scheduler invokes wq_worker_sleeping() with disabled preemption. Disabling preemption is needed because it protects access to `worker->sleeping'. As an optimisation it avoids invoking schedule() within the schedule path as part of possible wake up (thus preempt_enable_no_resched() afterwards). The io-wq has been added to the mix in the same section with disabled preemption. This breaks on PREEMPT_RT because io_wq_worker_sleeping() acquires a spinlock_t. Also within the schedule() the spinlock_t must be acquired after tsk_is_pi_blocked() otherwise it will block on the sleeping lock again while scheduling out. While playing with `io_uring-bench' I didn't notice a significant latency spike after converting io_wqe::lock to a raw_spinlock_t. The latency was more or less the same. In order to keep the spinlock_t it would have to be moved after the tsk_is_pi_blocked() check which would introduce a branch instruction into the hot path. The lock is used to maintain the `work_list' and wakes one task up at most. Should io_wqe_cancel_pending_work() cause latency spikes, while searching for a specific item, then it would need to drop the lock during iterations. revert_creds() is also invoked under the lock. According to debug cred::non_rcu is 0. Otherwise it should be moved outside of the locked section because put_cred_rcu()->free_uid() acquires a sleeping lock. Convert io_wqe::lock to a raw_spinlock_t.c Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-01 08:41:46 +00:00
raw_spin_unlock_irqrestore(&wqe->lock, flags);
io-wq: reorder cancellation pending -> running Go all over all pending lists and cancel works there, and only then try to match running requests. No functional changes here, just a preparation for bulk cancellation. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-06-15 07:24:02 +00:00
}
io-wq: add an option to cancel all matched reqs This adds support for cancelling all io-wq works matching a predicate. It isn't used yet, so no change in observable behaviour. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-06-15 07:24:03 +00:00
static void io_wqe_cancel_running_work(struct io_wqe *wqe,
io-wq: reorder cancellation pending -> running Go all over all pending lists and cancel works there, and only then try to match running requests. No functional changes here, just a preparation for bulk cancellation. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-06-15 07:24:02 +00:00
struct io_cb_cancel_data *match)
{
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
rcu_read_lock();
io-wq: add an option to cancel all matched reqs This adds support for cancelling all io-wq works matching a predicate. It isn't used yet, so no change in observable behaviour. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-06-15 07:24:03 +00:00
io_wq_for_each_worker(wqe, io_wq_worker_cancel, match);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
rcu_read_unlock();
}
io-wq: remove duplicated cancel code Deduplicate cancellation parts, as many of them looks the same, as do e.g. - io_wqe_cancel_cb_work() and io_wqe_cancel_work() - io_wq_worker_cancel() and io_work_cancel() Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-06 22:15:39 +00:00
enum io_wq_cancel io_wq_cancel_cb(struct io_wq *wq, work_cancel_fn *cancel,
io-wq: add an option to cancel all matched reqs This adds support for cancelling all io-wq works matching a predicate. It isn't used yet, so no change in observable behaviour. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-06-15 07:24:03 +00:00
void *data, bool cancel_all)
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
{
io-wq: remove duplicated cancel code Deduplicate cancellation parts, as many of them looks the same, as do e.g. - io_wqe_cancel_cb_work() and io_wqe_cancel_work() - io_wq_worker_cancel() and io_work_cancel() Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-06 22:15:39 +00:00
struct io_cb_cancel_data match = {
io-wq: add an option to cancel all matched reqs This adds support for cancelling all io-wq works matching a predicate. It isn't used yet, so no change in observable behaviour. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-06-15 07:24:03 +00:00
.fn = cancel,
.data = data,
.cancel_all = cancel_all,
io-wq: make io_wqe_cancel_work() take a match handler We want to use the cancel functionality for canceling based on not just the work itself. Instead of matching on the work address manually, allow a match handler to tell us if we found the right work item or not. No functional changes in this patch. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-02-09 02:13:32 +00:00
};
io-wq: fix handling of NUMA node IDs There are several things that can go wrong in the current code on NUMA systems, especially if not all nodes are online all the time: - If the identifiers of the online nodes do not form a single contiguous block starting at zero, wq->wqes will be too small, and OOB memory accesses will occur e.g. in the loop in io_wq_create(). - If a node comes online between the call to num_online_nodes() and the for_each_node() loop in io_wq_create(), an OOB write will occur. - If a node comes online between io_wq_create() and io_wq_enqueue(), a lookup is performed for an element that doesn't exist, and an OOB read will probably occur. Fix it by: - using nr_node_ids instead of num_online_nodes() for the allocation size; nr_node_ids is calculated by setup_nr_node_ids() to be bigger than the highest node ID that could possibly come online at some point, even if those nodes' identifiers are not a contiguous block - creating workers for all possible CPUs, not just all online ones This is basically what the normal workqueue code also does, as far as I can tell. Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-26 18:10:20 +00:00
int node;
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
io-wq: reorder cancellation pending -> running Go all over all pending lists and cancel works there, and only then try to match running requests. No functional changes here, just a preparation for bulk cancellation. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-06-15 07:24:02 +00:00
/*
* First check pending list, if we're lucky we can just remove it
* from there. CANCEL_OK means that the work is returned as-new,
* no completion will be posted for it.
*/
io-wq: fix handling of NUMA node IDs There are several things that can go wrong in the current code on NUMA systems, especially if not all nodes are online all the time: - If the identifiers of the online nodes do not form a single contiguous block starting at zero, wq->wqes will be too small, and OOB memory accesses will occur e.g. in the loop in io_wq_create(). - If a node comes online between the call to num_online_nodes() and the for_each_node() loop in io_wq_create(), an OOB write will occur. - If a node comes online between io_wq_create() and io_wq_enqueue(), a lookup is performed for an element that doesn't exist, and an OOB read will probably occur. Fix it by: - using nr_node_ids instead of num_online_nodes() for the allocation size; nr_node_ids is calculated by setup_nr_node_ids() to be bigger than the highest node ID that could possibly come online at some point, even if those nodes' identifiers are not a contiguous block - creating workers for all possible CPUs, not just all online ones This is basically what the normal workqueue code also does, as far as I can tell. Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-26 18:10:20 +00:00
for_each_node(node) {
struct io_wqe *wqe = wq->wqes[node];
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
io-wq: add an option to cancel all matched reqs This adds support for cancelling all io-wq works matching a predicate. It isn't used yet, so no change in observable behaviour. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-06-15 07:24:03 +00:00
io_wqe_cancel_pending_work(wqe, &match);
if (match.nr_pending && !match.cancel_all)
io-wq: reorder cancellation pending -> running Go all over all pending lists and cancel works there, and only then try to match running requests. No functional changes here, just a preparation for bulk cancellation. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-06-15 07:24:02 +00:00
return IO_WQ_CANCEL_OK;
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
}
io-wq: reorder cancellation pending -> running Go all over all pending lists and cancel works there, and only then try to match running requests. No functional changes here, just a preparation for bulk cancellation. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-06-15 07:24:02 +00:00
/*
* Now check if a free (going busy) or busy worker has the work
* currently running. If we find it there, we'll return CANCEL_RUNNING
* as an indication that we attempt to signal cancellation. The
* completion will run normally in this case.
*/
for_each_node(node) {
struct io_wqe *wqe = wq->wqes[node];
io-wq: add an option to cancel all matched reqs This adds support for cancelling all io-wq works matching a predicate. It isn't used yet, so no change in observable behaviour. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-06-15 07:24:03 +00:00
io_wqe_cancel_running_work(wqe, &match);
if (match.nr_running && !match.cancel_all)
io-wq: reorder cancellation pending -> running Go all over all pending lists and cancel works there, and only then try to match running requests. No functional changes here, just a preparation for bulk cancellation. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-06-15 07:24:02 +00:00
return IO_WQ_CANCEL_RUNNING;
}
io-wq: add an option to cancel all matched reqs This adds support for cancelling all io-wq works matching a predicate. It isn't used yet, so no change in observable behaviour. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-06-15 07:24:03 +00:00
if (match.nr_running)
return IO_WQ_CANCEL_RUNNING;
if (match.nr_pending)
return IO_WQ_CANCEL_OK;
io-wq: reorder cancellation pending -> running Go all over all pending lists and cancel works there, and only then try to match running requests. No functional changes here, just a preparation for bulk cancellation. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-06-15 07:24:02 +00:00
return IO_WQ_CANCEL_NOTFOUND;
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
}
io-wq: have io_wq_create() take a 'data' argument We currently pass in 4 arguments outside of the bounded size. In preparation for adding one more argument, let's bundle them up in a struct to make it more readable. No functional changes in this patch. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-25 15:49:20 +00:00
struct io_wq *io_wq_create(unsigned bounded, struct io_wq_data *data)
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
{
io-wq: fix handling of NUMA node IDs There are several things that can go wrong in the current code on NUMA systems, especially if not all nodes are online all the time: - If the identifiers of the online nodes do not form a single contiguous block starting at zero, wq->wqes will be too small, and OOB memory accesses will occur e.g. in the loop in io_wq_create(). - If a node comes online between the call to num_online_nodes() and the for_each_node() loop in io_wq_create(), an OOB write will occur. - If a node comes online between io_wq_create() and io_wq_enqueue(), a lookup is performed for an element that doesn't exist, and an OOB read will probably occur. Fix it by: - using nr_node_ids instead of num_online_nodes() for the allocation size; nr_node_ids is calculated by setup_nr_node_ids() to be bigger than the highest node ID that could possibly come online at some point, even if those nodes' identifiers are not a contiguous block - creating workers for all possible CPUs, not just all online ones This is basically what the normal workqueue code also does, as far as I can tell. Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-26 18:10:20 +00:00
int ret = -ENOMEM, node;
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
struct io_wq *wq;
io_wq: add per-wq work handler instead of per work io_uring is the only user of io-wq, and now it uses only io-wq callback for all its requests, namely io_wq_submit_work(). Instead of storing work->runner callback in each instance of io_wq_work, keep it in io-wq itself. pros: - reduces io_wq_work size - more robust -- ->func won't be invalidated with mem{cpy,set}(req) - helps other work Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-06-08 18:08:20 +00:00
if (WARN_ON_ONCE(!data->free_work || !data->do_work))
io_uring/io-wq: forward submission ref to async First it changes io-wq interfaces. It replaces {get,put}_work() with free_work(), which guaranteed to be called exactly once. It also enforces free_work() callback to be non-NULL. io_uring follows the changes and instead of putting a submission reference in io_put_req_async_completion(), it will be done in io_free_work(). As removes io_get_work() with corresponding refcount_inc(), the ref balance is maintained. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-04 13:14:12 +00:00
return ERR_PTR(-EINVAL);
io_uring: use kzalloc instead of kcalloc for single-element allocations These allocations are single-element allocations, so don't use the array allocation wrapper for them. Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-26 16:39:45 +00:00
wq = kzalloc(sizeof(*wq), GFP_KERNEL);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
if (!wq)
return ERR_PTR(-ENOMEM);
io-wq: fix handling of NUMA node IDs There are several things that can go wrong in the current code on NUMA systems, especially if not all nodes are online all the time: - If the identifiers of the online nodes do not form a single contiguous block starting at zero, wq->wqes will be too small, and OOB memory accesses will occur e.g. in the loop in io_wq_create(). - If a node comes online between the call to num_online_nodes() and the for_each_node() loop in io_wq_create(), an OOB write will occur. - If a node comes online between io_wq_create() and io_wq_enqueue(), a lookup is performed for an element that doesn't exist, and an OOB read will probably occur. Fix it by: - using nr_node_ids instead of num_online_nodes() for the allocation size; nr_node_ids is calculated by setup_nr_node_ids() to be bigger than the highest node ID that could possibly come online at some point, even if those nodes' identifiers are not a contiguous block - creating workers for all possible CPUs, not just all online ones This is basically what the normal workqueue code also does, as far as I can tell. Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-26 18:10:20 +00:00
wq->wqes = kcalloc(nr_node_ids, sizeof(struct io_wqe *), GFP_KERNEL);
io-wq: re-set NUMA node affinities if CPUs come online We correctly set io-wq NUMA node affinities when the io-wq context is setup, but if an entire node CPU set is offlined and then brought back online, the per node affinities are broken. Ensure that we set them again whenever a CPU comes online. This ensures that we always track the right node affinity. The usual cpuhp notifiers are used to drive it. Reported-by: Zhang Qiang <qiang.zhang@windriver.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-10-22 15:02:50 +00:00
if (!wq->wqes)
goto err_wq;
ret = cpuhp_state_add_instance_nocalls(io_wq_online, &wq->cpuhp_node);
if (ret)
goto err_wqes;
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
io_uring/io-wq: forward submission ref to async First it changes io-wq interfaces. It replaces {get,put}_work() with free_work(), which guaranteed to be called exactly once. It also enforces free_work() callback to be non-NULL. io_uring follows the changes and instead of putting a submission reference in io_put_req_async_completion(), it will be done in io_free_work(). As removes io_get_work() with corresponding refcount_inc(), the ref balance is maintained. Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-03-04 13:14:12 +00:00
wq->free_work = data->free_work;
io_wq: add per-wq work handler instead of per work io_uring is the only user of io-wq, and now it uses only io-wq callback for all its requests, namely io_wq_submit_work(). Instead of storing work->runner callback in each instance of io_wq_work, keep it in io-wq itself. pros: - reduces io_wq_work size - more robust -- ->func won't be invalidated with mem{cpy,set}(req) - helps other work Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-06-08 18:08:20 +00:00
wq->do_work = data->do_work;
io_wq: add get/put_work handlers to io_wq_create() For cancellation, we need to ensure that the work item stays valid for as long as ->cur_work is valid. Right now we can't safely dereference the work item even under the wqe->lock, because while the ->cur_work pointer will remain valid, the work could be completing and be freed in parallel. Only invoke ->get/put_work() on items we know that the caller queued themselves. Add IO_WQ_WORK_INTERNAL for io-wq to use, which is needed when we're queueing a flush item, for instance. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-13 05:31:31 +00:00
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
/* caller must already hold a reference to this */
io-wq: have io_wq_create() take a 'data' argument We currently pass in 4 arguments outside of the bounded size. In preparation for adding one more argument, let's bundle them up in a struct to make it more readable. No functional changes in this patch. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-25 15:49:20 +00:00
wq->user = data->user;
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
io-wq: re-set NUMA node affinities if CPUs come online We correctly set io-wq NUMA node affinities when the io-wq context is setup, but if an entire node CPU set is offlined and then brought back online, the per node affinities are broken. Ensure that we set them again whenever a CPU comes online. This ensures that we always track the right node affinity. The usual cpuhp notifiers are used to drive it. Reported-by: Zhang Qiang <qiang.zhang@windriver.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-10-22 15:02:50 +00:00
ret = -ENOMEM;
io-wq: fix handling of NUMA node IDs There are several things that can go wrong in the current code on NUMA systems, especially if not all nodes are online all the time: - If the identifiers of the online nodes do not form a single contiguous block starting at zero, wq->wqes will be too small, and OOB memory accesses will occur e.g. in the loop in io_wq_create(). - If a node comes online between the call to num_online_nodes() and the for_each_node() loop in io_wq_create(), an OOB write will occur. - If a node comes online between io_wq_create() and io_wq_enqueue(), a lookup is performed for an element that doesn't exist, and an OOB read will probably occur. Fix it by: - using nr_node_ids instead of num_online_nodes() for the allocation size; nr_node_ids is calculated by setup_nr_node_ids() to be bigger than the highest node ID that could possibly come online at some point, even if those nodes' identifiers are not a contiguous block - creating workers for all possible CPUs, not just all online ones This is basically what the normal workqueue code also does, as far as I can tell. Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-26 18:10:20 +00:00
for_each_node(node) {
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
struct io_wqe *wqe;
io-wq: don't call kXalloc_node() with non-online node Glauber reports a crash on init on a box he has: RIP: 0010:__alloc_pages_nodemask+0x132/0x340 Code: 18 01 75 04 41 80 ce 80 89 e8 48 8b 54 24 08 8b 74 24 1c c1 e8 0c 48 8b 3c 24 83 e0 01 88 44 24 20 48 85 d2 0f 85 74 01 00 00 <3b> 77 08 0f 82 6b 01 00 00 48 89 7c 24 10 89 ea 48 8b 07 b9 00 02 RSP: 0018:ffffb8be4d0b7c28 EFLAGS: 00010246 RAX: 0000000000000000 RBX: 0000000000000000 RCX: 000000000000e8e8 RDX: 0000000000000000 RSI: 0000000000000002 RDI: 0000000000002080 RBP: 0000000000012cc0 R08: 0000000000000000 R09: 0000000000000002 R10: 0000000000000dc0 R11: ffff995c60400100 R12: 0000000000000000 R13: 0000000000012cc0 R14: 0000000000000001 R15: ffff995c60db00f0 FS: 00007f4d115ca900(0000) GS:ffff995c60d80000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000000002088 CR3: 00000017cca66002 CR4: 00000000007606e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 PKRU: 55555554 Call Trace: alloc_slab_page+0x46/0x320 new_slab+0x9d/0x4e0 ___slab_alloc+0x507/0x6a0 ? io_wq_create+0xb4/0x2a0 __slab_alloc+0x1c/0x30 kmem_cache_alloc_node_trace+0xa6/0x260 io_wq_create+0xb4/0x2a0 io_uring_setup+0x97f/0xaa0 ? io_remove_personalities+0x30/0x30 ? io_poll_trigger_evfd+0x30/0x30 do_syscall_64+0x5b/0x1c0 entry_SYSCALL_64_after_hwframe+0x44/0xa9 RIP: 0033:0x7f4d116cb1ed which is due to the 'wqe' and 'worker' allocation being node affine. But it isn't valid to call the node affine allocation if the node isn't online. Setup structures for even offline nodes, as usual, but skip them in terms of thread setup to not waste resources. If the node isn't online, just alloc memory with NUMA_NO_NODE. Reported-by: Glauber Costa <glauber@scylladb.com> Tested-by: Glauber Costa <glauber@scylladb.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-02-11 13:30:06 +00:00
int alloc_node = node;
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
io-wq: don't call kXalloc_node() with non-online node Glauber reports a crash on init on a box he has: RIP: 0010:__alloc_pages_nodemask+0x132/0x340 Code: 18 01 75 04 41 80 ce 80 89 e8 48 8b 54 24 08 8b 74 24 1c c1 e8 0c 48 8b 3c 24 83 e0 01 88 44 24 20 48 85 d2 0f 85 74 01 00 00 <3b> 77 08 0f 82 6b 01 00 00 48 89 7c 24 10 89 ea 48 8b 07 b9 00 02 RSP: 0018:ffffb8be4d0b7c28 EFLAGS: 00010246 RAX: 0000000000000000 RBX: 0000000000000000 RCX: 000000000000e8e8 RDX: 0000000000000000 RSI: 0000000000000002 RDI: 0000000000002080 RBP: 0000000000012cc0 R08: 0000000000000000 R09: 0000000000000002 R10: 0000000000000dc0 R11: ffff995c60400100 R12: 0000000000000000 R13: 0000000000012cc0 R14: 0000000000000001 R15: ffff995c60db00f0 FS: 00007f4d115ca900(0000) GS:ffff995c60d80000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000000002088 CR3: 00000017cca66002 CR4: 00000000007606e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 PKRU: 55555554 Call Trace: alloc_slab_page+0x46/0x320 new_slab+0x9d/0x4e0 ___slab_alloc+0x507/0x6a0 ? io_wq_create+0xb4/0x2a0 __slab_alloc+0x1c/0x30 kmem_cache_alloc_node_trace+0xa6/0x260 io_wq_create+0xb4/0x2a0 io_uring_setup+0x97f/0xaa0 ? io_remove_personalities+0x30/0x30 ? io_poll_trigger_evfd+0x30/0x30 do_syscall_64+0x5b/0x1c0 entry_SYSCALL_64_after_hwframe+0x44/0xa9 RIP: 0033:0x7f4d116cb1ed which is due to the 'wqe' and 'worker' allocation being node affine. But it isn't valid to call the node affine allocation if the node isn't online. Setup structures for even offline nodes, as usual, but skip them in terms of thread setup to not waste resources. If the node isn't online, just alloc memory with NUMA_NO_NODE. Reported-by: Glauber Costa <glauber@scylladb.com> Tested-by: Glauber Costa <glauber@scylladb.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-02-11 13:30:06 +00:00
if (!node_online(alloc_node))
alloc_node = NUMA_NO_NODE;
wqe = kzalloc_node(sizeof(struct io_wqe), GFP_KERNEL, alloc_node);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
if (!wqe)
io-wq: fix handling of NUMA node IDs There are several things that can go wrong in the current code on NUMA systems, especially if not all nodes are online all the time: - If the identifiers of the online nodes do not form a single contiguous block starting at zero, wq->wqes will be too small, and OOB memory accesses will occur e.g. in the loop in io_wq_create(). - If a node comes online between the call to num_online_nodes() and the for_each_node() loop in io_wq_create(), an OOB write will occur. - If a node comes online between io_wq_create() and io_wq_enqueue(), a lookup is performed for an element that doesn't exist, and an OOB read will probably occur. Fix it by: - using nr_node_ids instead of num_online_nodes() for the allocation size; nr_node_ids is calculated by setup_nr_node_ids() to be bigger than the highest node ID that could possibly come online at some point, even if those nodes' identifiers are not a contiguous block - creating workers for all possible CPUs, not just all online ones This is basically what the normal workqueue code also does, as far as I can tell. Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-26 18:10:20 +00:00
goto err;
wq->wqes[node] = wqe;
io-wq: don't call kXalloc_node() with non-online node Glauber reports a crash on init on a box he has: RIP: 0010:__alloc_pages_nodemask+0x132/0x340 Code: 18 01 75 04 41 80 ce 80 89 e8 48 8b 54 24 08 8b 74 24 1c c1 e8 0c 48 8b 3c 24 83 e0 01 88 44 24 20 48 85 d2 0f 85 74 01 00 00 <3b> 77 08 0f 82 6b 01 00 00 48 89 7c 24 10 89 ea 48 8b 07 b9 00 02 RSP: 0018:ffffb8be4d0b7c28 EFLAGS: 00010246 RAX: 0000000000000000 RBX: 0000000000000000 RCX: 000000000000e8e8 RDX: 0000000000000000 RSI: 0000000000000002 RDI: 0000000000002080 RBP: 0000000000012cc0 R08: 0000000000000000 R09: 0000000000000002 R10: 0000000000000dc0 R11: ffff995c60400100 R12: 0000000000000000 R13: 0000000000012cc0 R14: 0000000000000001 R15: ffff995c60db00f0 FS: 00007f4d115ca900(0000) GS:ffff995c60d80000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000000002088 CR3: 00000017cca66002 CR4: 00000000007606e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 PKRU: 55555554 Call Trace: alloc_slab_page+0x46/0x320 new_slab+0x9d/0x4e0 ___slab_alloc+0x507/0x6a0 ? io_wq_create+0xb4/0x2a0 __slab_alloc+0x1c/0x30 kmem_cache_alloc_node_trace+0xa6/0x260 io_wq_create+0xb4/0x2a0 io_uring_setup+0x97f/0xaa0 ? io_remove_personalities+0x30/0x30 ? io_poll_trigger_evfd+0x30/0x30 do_syscall_64+0x5b/0x1c0 entry_SYSCALL_64_after_hwframe+0x44/0xa9 RIP: 0033:0x7f4d116cb1ed which is due to the 'wqe' and 'worker' allocation being node affine. But it isn't valid to call the node affine allocation if the node isn't online. Setup structures for even offline nodes, as usual, but skip them in terms of thread setup to not waste resources. If the node isn't online, just alloc memory with NUMA_NO_NODE. Reported-by: Glauber Costa <glauber@scylladb.com> Tested-by: Glauber Costa <glauber@scylladb.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-02-11 13:30:06 +00:00
wqe->node = alloc_node;
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
wqe->acct[IO_WQ_ACCT_BOUND].max_workers = bounded;
atomic_set(&wqe->acct[IO_WQ_ACCT_BOUND].nr_running, 0);
io-wq: have io_wq_create() take a 'data' argument We currently pass in 4 arguments outside of the bounded size. In preparation for adding one more argument, let's bundle them up in a struct to make it more readable. No functional changes in this patch. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-25 15:49:20 +00:00
if (wq->user) {
io-wq: add support for bounded vs unbunded work io_uring supports request types that basically have two different lifetimes: 1) Bounded completion time. These are requests like disk reads or writes, which we know will finish in a finite amount of time. 2) Unbounded completion time. These are generally networked IO, where we have no idea how long they will take to complete. Another example is POLL commands. This patch provides support for io-wq to handle these differently, so we don't starve bounded requests by tying up workers for too long. By default all work is bounded, unless otherwise specified in the work item. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-07 18:41:16 +00:00
wqe->acct[IO_WQ_ACCT_UNBOUND].max_workers =
task_rlimit(current, RLIMIT_NPROC);
}
atomic_set(&wqe->acct[IO_WQ_ACCT_UNBOUND].nr_running, 0);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
wqe->wq = wq;
io_wq: Make io_wqe::lock a raw_spinlock_t During a context switch the scheduler invokes wq_worker_sleeping() with disabled preemption. Disabling preemption is needed because it protects access to `worker->sleeping'. As an optimisation it avoids invoking schedule() within the schedule path as part of possible wake up (thus preempt_enable_no_resched() afterwards). The io-wq has been added to the mix in the same section with disabled preemption. This breaks on PREEMPT_RT because io_wq_worker_sleeping() acquires a spinlock_t. Also within the schedule() the spinlock_t must be acquired after tsk_is_pi_blocked() otherwise it will block on the sleeping lock again while scheduling out. While playing with `io_uring-bench' I didn't notice a significant latency spike after converting io_wqe::lock to a raw_spinlock_t. The latency was more or less the same. In order to keep the spinlock_t it would have to be moved after the tsk_is_pi_blocked() check which would introduce a branch instruction into the hot path. The lock is used to maintain the `work_list' and wakes one task up at most. Should io_wqe_cancel_pending_work() cause latency spikes, while searching for a specific item, then it would need to drop the lock during iterations. revert_creds() is also invoked under the lock. According to debug cred::non_rcu is 0. Otherwise it should be moved outside of the locked section because put_cred_rcu()->free_uid() acquires a sleeping lock. Convert io_wqe::lock to a raw_spinlock_t.c Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-09-01 08:41:46 +00:00
raw_spin_lock_init(&wqe->lock);
io-wq: shrink io_wq_work a bit Currently we're using 40 bytes for the io_wq_work structure, and 16 of those is the doubly link list node. We don't need doubly linked lists, we always add to tail to keep things ordered, and any other use case is list traversal with deletion. For the deletion case, we can easily support any node deletion by keeping track of the previous entry. This shrinks io_wq_work to 32 bytes, and subsequently io_kiock from io_uring to 216 to 208 bytes. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-26 18:59:32 +00:00
INIT_WQ_LIST(&wqe->work_list);
io-wq: remove now redundant struct io_wq_nulls_list Since we don't iterate these lists anymore after commit: e61df66c69b1 ("io-wq: ensure free/busy list browsing see all items") we don't need to retain the nulls value we use for them. That means it's pretty pointless to wrap the hlist_nulls_head in a structure, so get rid of it. Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-14 15:00:41 +00:00
INIT_HLIST_NULLS_HEAD(&wqe->free_list, 0);
io-wq: ensure free/busy list browsing see all items We have two lists for workers in io-wq, a busy and a free list. For certain operations we want to browse all workers, and we currently do that by browsing the two separate lists. But since these lists are RCU protected, we can potentially miss workers if they move between the two lists while we're browsing them. Add a third list, all_list, that simply holds all workers. A worker is added to that list when it starts, and removed when it exits. This makes the worker iteration cleaner, too. Reported-by: Paul E. McKenney <paulmck@kernel.org> Reviewed-by: Paul E. McKenney <paulmck@kernel.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-13 20:54:49 +00:00
INIT_LIST_HEAD(&wqe->all_list);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
}
init_completion(&wq->done);
wq->manager = kthread_create(io_wq_manager, wq, "io_wq_manager");
if (!IS_ERR(wq->manager)) {
wake_up_process(wq->manager);
io-wq: wait for io_wq_create() to setup necessary workers We currently have a race where if setup is really slow, we can be calling io_wq_destroy() before we're done setting up. This will cause the caller to get stuck waiting for the manager to set things up, but the manager already exited. Fix this by doing a sync setup of the manager. This also fixes the case where if we failed creating workers, we'd also get stuck. In practice this race window was really small, as we already wait for the manager to start. Hence someone would have to call io_wq_destroy() after the task has started, but before it started the first loop. The reported test case forked tons of these, which is why it became an issue. Reported-by: syzbot+0f1cc17f85154f400465@syzkaller.appspotmail.com Fixes: 771b53d033e8 ("io-wq: small threadpool implementation for io_uring") Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-19 15:37:07 +00:00
wait_for_completion(&wq->done);
if (test_bit(IO_WQ_BIT_ERROR, &wq->state)) {
ret = -ENOMEM;
goto err;
}
io-wq: make the io_wq ref counted In preparation for sharing an io-wq across different users, add a reference count that manages destruction of it. Reviewed-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-01-23 22:33:32 +00:00
refcount_set(&wq->use_refs, 1);
io-wq: wait for io_wq_create() to setup necessary workers We currently have a race where if setup is really slow, we can be calling io_wq_destroy() before we're done setting up. This will cause the caller to get stuck waiting for the manager to set things up, but the manager already exited. Fix this by doing a sync setup of the manager. This also fixes the case where if we failed creating workers, we'd also get stuck. In practice this race window was really small, as we already wait for the manager to start. Hence someone would have to call io_wq_destroy() after the task has started, but before it started the first loop. The reported test case forked tons of these, which is why it became an issue. Reported-by: syzbot+0f1cc17f85154f400465@syzkaller.appspotmail.com Fixes: 771b53d033e8 ("io-wq: small threadpool implementation for io_uring") Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-19 15:37:07 +00:00
reinit_completion(&wq->done);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
return wq;
}
ret = PTR_ERR(wq->manager);
complete(&wq->done);
io-wq: wait for io_wq_create() to setup necessary workers We currently have a race where if setup is really slow, we can be calling io_wq_destroy() before we're done setting up. This will cause the caller to get stuck waiting for the manager to set things up, but the manager already exited. Fix this by doing a sync setup of the manager. This also fixes the case where if we failed creating workers, we'd also get stuck. In practice this race window was really small, as we already wait for the manager to start. Hence someone would have to call io_wq_destroy() after the task has started, but before it started the first loop. The reported test case forked tons of these, which is why it became an issue. Reported-by: syzbot+0f1cc17f85154f400465@syzkaller.appspotmail.com Fixes: 771b53d033e8 ("io-wq: small threadpool implementation for io_uring") Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-19 15:37:07 +00:00
err:
io-wq: re-set NUMA node affinities if CPUs come online We correctly set io-wq NUMA node affinities when the io-wq context is setup, but if an entire node CPU set is offlined and then brought back online, the per node affinities are broken. Ensure that we set them again whenever a CPU comes online. This ensures that we always track the right node affinity. The usual cpuhp notifiers are used to drive it. Reported-by: Zhang Qiang <qiang.zhang@windriver.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-10-22 15:02:50 +00:00
cpuhp_state_remove_instance_nocalls(io_wq_online, &wq->cpuhp_node);
io-wq: fix handling of NUMA node IDs There are several things that can go wrong in the current code on NUMA systems, especially if not all nodes are online all the time: - If the identifiers of the online nodes do not form a single contiguous block starting at zero, wq->wqes will be too small, and OOB memory accesses will occur e.g. in the loop in io_wq_create(). - If a node comes online between the call to num_online_nodes() and the for_each_node() loop in io_wq_create(), an OOB write will occur. - If a node comes online between io_wq_create() and io_wq_enqueue(), a lookup is performed for an element that doesn't exist, and an OOB read will probably occur. Fix it by: - using nr_node_ids instead of num_online_nodes() for the allocation size; nr_node_ids is calculated by setup_nr_node_ids() to be bigger than the highest node ID that could possibly come online at some point, even if those nodes' identifiers are not a contiguous block - creating workers for all possible CPUs, not just all online ones This is basically what the normal workqueue code also does, as far as I can tell. Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-26 18:10:20 +00:00
for_each_node(node)
kfree(wq->wqes[node]);
io-wq: re-set NUMA node affinities if CPUs come online We correctly set io-wq NUMA node affinities when the io-wq context is setup, but if an entire node CPU set is offlined and then brought back online, the per node affinities are broken. Ensure that we set them again whenever a CPU comes online. This ensures that we always track the right node affinity. The usual cpuhp notifiers are used to drive it. Reported-by: Zhang Qiang <qiang.zhang@windriver.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-10-22 15:02:50 +00:00
err_wqes:
io-wq: wait for io_wq_create() to setup necessary workers We currently have a race where if setup is really slow, we can be calling io_wq_destroy() before we're done setting up. This will cause the caller to get stuck waiting for the manager to set things up, but the manager already exited. Fix this by doing a sync setup of the manager. This also fixes the case where if we failed creating workers, we'd also get stuck. In practice this race window was really small, as we already wait for the manager to start. Hence someone would have to call io_wq_destroy() after the task has started, but before it started the first loop. The reported test case forked tons of these, which is why it became an issue. Reported-by: syzbot+0f1cc17f85154f400465@syzkaller.appspotmail.com Fixes: 771b53d033e8 ("io-wq: small threadpool implementation for io_uring") Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-19 15:37:07 +00:00
kfree(wq->wqes);
io-wq: re-set NUMA node affinities if CPUs come online We correctly set io-wq NUMA node affinities when the io-wq context is setup, but if an entire node CPU set is offlined and then brought back online, the per node affinities are broken. Ensure that we set them again whenever a CPU comes online. This ensures that we always track the right node affinity. The usual cpuhp notifiers are used to drive it. Reported-by: Zhang Qiang <qiang.zhang@windriver.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-10-22 15:02:50 +00:00
err_wq:
io-wq: wait for io_wq_create() to setup necessary workers We currently have a race where if setup is really slow, we can be calling io_wq_destroy() before we're done setting up. This will cause the caller to get stuck waiting for the manager to set things up, but the manager already exited. Fix this by doing a sync setup of the manager. This also fixes the case where if we failed creating workers, we'd also get stuck. In practice this race window was really small, as we already wait for the manager to start. Hence someone would have to call io_wq_destroy() after the task has started, but before it started the first loop. The reported test case forked tons of these, which is why it became an issue. Reported-by: syzbot+0f1cc17f85154f400465@syzkaller.appspotmail.com Fixes: 771b53d033e8 ("io-wq: small threadpool implementation for io_uring") Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-19 15:37:07 +00:00
kfree(wq);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
return ERR_PTR(ret);
}
io-wq: allow grabbing existing io-wq Export a helper to attach to an existing io-wq, rather than setting up a new one. This is doable now that we have reference counted io_wq's. Reported-by: Jens Axboe <axboe@kernel.dk> Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-01-28 00:15:47 +00:00
bool io_wq_get(struct io_wq *wq, struct io_wq_data *data)
{
io_wq: add per-wq work handler instead of per work io_uring is the only user of io-wq, and now it uses only io-wq callback for all its requests, namely io_wq_submit_work(). Instead of storing work->runner callback in each instance of io_wq_work, keep it in io-wq itself. pros: - reduces io_wq_work size - more robust -- ->func won't be invalidated with mem{cpy,set}(req) - helps other work Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-06-08 18:08:20 +00:00
if (data->free_work != wq->free_work || data->do_work != wq->do_work)
io-wq: allow grabbing existing io-wq Export a helper to attach to an existing io-wq, rather than setting up a new one. This is doable now that we have reference counted io_wq's. Reported-by: Jens Axboe <axboe@kernel.dk> Signed-off-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-01-28 00:15:47 +00:00
return false;
return refcount_inc_not_zero(&wq->use_refs);
}
io-wq: make the io_wq ref counted In preparation for sharing an io-wq across different users, add a reference count that manages destruction of it. Reviewed-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-01-23 22:33:32 +00:00
static void __io_wq_destroy(struct io_wq *wq)
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
{
io-wq: fix handling of NUMA node IDs There are several things that can go wrong in the current code on NUMA systems, especially if not all nodes are online all the time: - If the identifiers of the online nodes do not form a single contiguous block starting at zero, wq->wqes will be too small, and OOB memory accesses will occur e.g. in the loop in io_wq_create(). - If a node comes online between the call to num_online_nodes() and the for_each_node() loop in io_wq_create(), an OOB write will occur. - If a node comes online between io_wq_create() and io_wq_enqueue(), a lookup is performed for an element that doesn't exist, and an OOB read will probably occur. Fix it by: - using nr_node_ids instead of num_online_nodes() for the allocation size; nr_node_ids is calculated by setup_nr_node_ids() to be bigger than the highest node ID that could possibly come online at some point, even if those nodes' identifiers are not a contiguous block - creating workers for all possible CPUs, not just all online ones This is basically what the normal workqueue code also does, as far as I can tell. Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-26 18:10:20 +00:00
int node;
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
io-wq: re-set NUMA node affinities if CPUs come online We correctly set io-wq NUMA node affinities when the io-wq context is setup, but if an entire node CPU set is offlined and then brought back online, the per node affinities are broken. Ensure that we set them again whenever a CPU comes online. This ensures that we always track the right node affinity. The usual cpuhp notifiers are used to drive it. Reported-by: Zhang Qiang <qiang.zhang@windriver.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-10-22 15:02:50 +00:00
cpuhp_state_remove_instance_nocalls(io_wq_online, &wq->cpuhp_node);
io-wq: wait for io_wq_create() to setup necessary workers We currently have a race where if setup is really slow, we can be calling io_wq_destroy() before we're done setting up. This will cause the caller to get stuck waiting for the manager to set things up, but the manager already exited. Fix this by doing a sync setup of the manager. This also fixes the case where if we failed creating workers, we'd also get stuck. In practice this race window was really small, as we already wait for the manager to start. Hence someone would have to call io_wq_destroy() after the task has started, but before it started the first loop. The reported test case forked tons of these, which is why it became an issue. Reported-by: syzbot+0f1cc17f85154f400465@syzkaller.appspotmail.com Fixes: 771b53d033e8 ("io-wq: small threadpool implementation for io_uring") Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-19 15:37:07 +00:00
set_bit(IO_WQ_BIT_EXIT, &wq->state);
if (wq->manager)
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
kthread_stop(wq->manager);
rcu_read_lock();
io-wq: fix handling of NUMA node IDs There are several things that can go wrong in the current code on NUMA systems, especially if not all nodes are online all the time: - If the identifiers of the online nodes do not form a single contiguous block starting at zero, wq->wqes will be too small, and OOB memory accesses will occur e.g. in the loop in io_wq_create(). - If a node comes online between the call to num_online_nodes() and the for_each_node() loop in io_wq_create(), an OOB write will occur. - If a node comes online between io_wq_create() and io_wq_enqueue(), a lookup is performed for an element that doesn't exist, and an OOB read will probably occur. Fix it by: - using nr_node_ids instead of num_online_nodes() for the allocation size; nr_node_ids is calculated by setup_nr_node_ids() to be bigger than the highest node ID that could possibly come online at some point, even if those nodes' identifiers are not a contiguous block - creating workers for all possible CPUs, not just all online ones This is basically what the normal workqueue code also does, as far as I can tell. Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-26 18:10:20 +00:00
for_each_node(node)
io_wq_for_each_worker(wq->wqes[node], io_wq_worker_wake, NULL);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
rcu_read_unlock();
wait_for_completion(&wq->done);
io-wq: fix handling of NUMA node IDs There are several things that can go wrong in the current code on NUMA systems, especially if not all nodes are online all the time: - If the identifiers of the online nodes do not form a single contiguous block starting at zero, wq->wqes will be too small, and OOB memory accesses will occur e.g. in the loop in io_wq_create(). - If a node comes online between the call to num_online_nodes() and the for_each_node() loop in io_wq_create(), an OOB write will occur. - If a node comes online between io_wq_create() and io_wq_enqueue(), a lookup is performed for an element that doesn't exist, and an OOB read will probably occur. Fix it by: - using nr_node_ids instead of num_online_nodes() for the allocation size; nr_node_ids is calculated by setup_nr_node_ids() to be bigger than the highest node ID that could possibly come online at some point, even if those nodes' identifiers are not a contiguous block - creating workers for all possible CPUs, not just all online ones This is basically what the normal workqueue code also does, as far as I can tell. Signed-off-by: Jann Horn <jannh@google.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-11-26 18:10:20 +00:00
for_each_node(node)
kfree(wq->wqes[node]);
io-wq: small threadpool implementation for io_uring This adds support for io-wq, a smaller and specialized thread pool implementation. This is meant to replace workqueues for io_uring. Among the reasons for this addition are: - We can assign memory context smarter and more persistently if we manage the life time of threads. - We can drop various work-arounds we have in io_uring, like the async_list. - We can implement hashed work insertion, to manage concurrency of buffered writes without needing a) an extra workqueue, or b) needlessly making the concurrency of said workqueue very low which hurts performance of multiple buffered file writers. - We can implement cancel through signals, for cancelling interruptible work like read/write (or send/recv) to/from sockets. - We need the above cancel for being able to assign and use file tables from a process. - We can implement a more thorough cancel operation in general. - We need it to move towards a syslet/threadlet model for even faster async execution. For that we need to take ownership of the used threads. This list is just off the top of my head. Performance should be the same, or better, at least that's what I've seen in my testing. io-wq supports basic NUMA functionality, setting up a pool per node. io-wq hooks up to the scheduler schedule in/out just like workqueue and uses that to drive the need for more/less workers. Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2019-10-22 16:25:58 +00:00
kfree(wq->wqes);
kfree(wq);
}
io-wq: make the io_wq ref counted In preparation for sharing an io-wq across different users, add a reference count that manages destruction of it. Reviewed-by: Pavel Begunkov <asml.silence@gmail.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-01-23 22:33:32 +00:00
void io_wq_destroy(struct io_wq *wq)
{
if (refcount_dec_and_test(&wq->use_refs))
__io_wq_destroy(wq);
}
io_uring: use io-wq manager as backup task if task is exiting If the original task is (or has) exited, then the task work will not get queued properly. Allow for using the io-wq manager task to queue this work for execution, and ensure that the io-wq manager notices and runs this work if woken up (or exiting). Reported-by: Dan Melnic <dmm@fb.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-04-03 17:26:26 +00:00
struct task_struct *io_wq_get_task(struct io_wq *wq)
{
return wq->manager;
}
io-wq: re-set NUMA node affinities if CPUs come online We correctly set io-wq NUMA node affinities when the io-wq context is setup, but if an entire node CPU set is offlined and then brought back online, the per node affinities are broken. Ensure that we set them again whenever a CPU comes online. This ensures that we always track the right node affinity. The usual cpuhp notifiers are used to drive it. Reported-by: Zhang Qiang <qiang.zhang@windriver.com> Signed-off-by: Jens Axboe <axboe@kernel.dk>
2020-10-22 15:02:50 +00:00
static bool io_wq_worker_affinity(struct io_worker *worker, void *data)
{
struct task_struct *task = worker->task;
struct rq_flags rf;
struct rq *rq;
rq = task_rq_lock(task, &rf);
do_set_cpus_allowed(task, cpumask_of_node(worker->wqe->node));
task->flags |= PF_NO_SETAFFINITY;
task_rq_unlock(rq, task, &rf);
return false;
}
static int io_wq_cpu_online(unsigned int cpu, struct hlist_node *node)
{
struct io_wq *wq = hlist_entry_safe(node, struct io_wq, cpuhp_node);
int i;
rcu_read_lock();
for_each_node(i)
io_wq_for_each_worker(wq->wqes[i], io_wq_worker_affinity, NULL);
rcu_read_unlock();
return 0;
}
static __init int io_wq_init(void)
{
int ret;
ret = cpuhp_setup_state_multi(CPUHP_AP_ONLINE_DYN, "io-wq/online",
io_wq_cpu_online, NULL);
if (ret < 0)
return ret;
io_wq_online = ret;
return 0;
}
subsys_initcall(io_wq_init);
Reference in a new issue Copy permalink