blob: 4023c9846860865a6461eaa3cc9ee27b620e5b8a [file] [log] [blame]
Jens Axboe771b53d02019-10-22 10:25:58 -06001// SPDX-License-Identifier: GPL-2.0
2/*
3 * Basic worker thread pool for io_uring
4 *
5 * Copyright (C) 2019 Jens Axboe
6 *
7 */
8#include <linux/kernel.h>
9#include <linux/init.h>
10#include <linux/errno.h>
11#include <linux/sched/signal.h>
12#include <linux/mm.h>
13#include <linux/mmu_context.h>
14#include <linux/sched/mm.h>
15#include <linux/percpu.h>
16#include <linux/slab.h>
17#include <linux/kthread.h>
18#include <linux/rculist_nulls.h>
Jens Axboe9392a272020-02-06 21:42:51 -070019#include <linux/fs_struct.h>
Jens Axboeaa96bf82020-04-03 11:26:26 -060020#include <linux/task_work.h>
Jens Axboe771b53d02019-10-22 10:25:58 -060021
22#include "io-wq.h"
23
24#define WORKER_IDLE_TIMEOUT (5 * HZ)
25
26enum {
27 IO_WORKER_F_UP = 1, /* up and active */
28 IO_WORKER_F_RUNNING = 2, /* account as running */
29 IO_WORKER_F_FREE = 4, /* worker on free list */
30 IO_WORKER_F_EXITING = 8, /* worker exiting */
31 IO_WORKER_F_FIXED = 16, /* static idle worker */
Jens Axboec5def4a2019-11-07 11:41:16 -070032 IO_WORKER_F_BOUND = 32, /* is doing bounded work */
Jens Axboe771b53d02019-10-22 10:25:58 -060033};
34
35enum {
36 IO_WQ_BIT_EXIT = 0, /* wq exiting */
37 IO_WQ_BIT_CANCEL = 1, /* cancel work on list */
Jens Axboeb60fda62019-11-19 08:37:07 -070038 IO_WQ_BIT_ERROR = 2, /* error on setup */
Jens Axboe771b53d02019-10-22 10:25:58 -060039};
40
41enum {
42 IO_WQE_FLAG_STALLED = 1, /* stalled on hash */
43};
44
45/*
46 * One for each thread in a wqe pool
47 */
48struct io_worker {
49 refcount_t ref;
50 unsigned flags;
51 struct hlist_nulls_node nulls_node;
Jens Axboee61df662019-11-13 13:54:49 -070052 struct list_head all_list;
Jens Axboe771b53d02019-10-22 10:25:58 -060053 struct task_struct *task;
Jens Axboe771b53d02019-10-22 10:25:58 -060054 struct io_wqe *wqe;
Jens Axboe36c2f922019-11-13 09:43:34 -070055
Jens Axboe771b53d02019-10-22 10:25:58 -060056 struct io_wq_work *cur_work;
Jens Axboe36c2f922019-11-13 09:43:34 -070057 spinlock_t lock;
Jens Axboe771b53d02019-10-22 10:25:58 -060058
59 struct rcu_head rcu;
60 struct mm_struct *mm;
Jens Axboecccf0ee2020-01-27 16:34:48 -070061 const struct cred *cur_creds;
62 const struct cred *saved_creds;
Jens Axboefcb323c2019-10-24 12:39:47 -060063 struct files_struct *restore_files;
Jens Axboe9392a272020-02-06 21:42:51 -070064 struct fs_struct *restore_fs;
Jens Axboe771b53d02019-10-22 10:25:58 -060065};
66
Jens Axboe771b53d02019-10-22 10:25:58 -060067#if BITS_PER_LONG == 64
68#define IO_WQ_HASH_ORDER 6
69#else
70#define IO_WQ_HASH_ORDER 5
71#endif
72
Pavel Begunkov86f3cd12020-03-23 22:57:22 +030073#define IO_WQ_NR_HASH_BUCKETS (1u << IO_WQ_HASH_ORDER)
74
Jens Axboec5def4a2019-11-07 11:41:16 -070075struct io_wqe_acct {
76 unsigned nr_workers;
77 unsigned max_workers;
78 atomic_t nr_running;
79};
80
81enum {
82 IO_WQ_ACCT_BOUND,
83 IO_WQ_ACCT_UNBOUND,
84};
85
Jens Axboe771b53d02019-10-22 10:25:58 -060086/*
87 * Per-node worker thread pool
88 */
89struct io_wqe {
90 struct {
91 spinlock_t lock;
Jens Axboe6206f0e2019-11-26 11:59:32 -070092 struct io_wq_work_list work_list;
Jens Axboe771b53d02019-10-22 10:25:58 -060093 unsigned long hash_map;
94 unsigned flags;
95 } ____cacheline_aligned_in_smp;
96
97 int node;
Jens Axboec5def4a2019-11-07 11:41:16 -070098 struct io_wqe_acct acct[2];
Jens Axboe771b53d02019-10-22 10:25:58 -060099
Jens Axboe021d1cd2019-11-14 08:00:41 -0700100 struct hlist_nulls_head free_list;
Jens Axboee61df662019-11-13 13:54:49 -0700101 struct list_head all_list;
Jens Axboe771b53d02019-10-22 10:25:58 -0600102
103 struct io_wq *wq;
Pavel Begunkov86f3cd12020-03-23 22:57:22 +0300104 struct io_wq_work *hash_tail[IO_WQ_NR_HASH_BUCKETS];
Jens Axboe771b53d02019-10-22 10:25:58 -0600105};
106
107/*
108 * Per io_wq state
109 */
110struct io_wq {
111 struct io_wqe **wqes;
112 unsigned long state;
Jens Axboe771b53d02019-10-22 10:25:58 -0600113
Pavel Begunkove9fd9392020-03-04 16:14:12 +0300114 free_work_fn *free_work;
Jens Axboe7d723062019-11-12 22:31:31 -0700115
Jens Axboe771b53d02019-10-22 10:25:58 -0600116 struct task_struct *manager;
Jens Axboec5def4a2019-11-07 11:41:16 -0700117 struct user_struct *user;
Jens Axboe771b53d02019-10-22 10:25:58 -0600118 refcount_t refs;
119 struct completion done;
Jens Axboe848f7e12020-01-23 15:33:32 -0700120
121 refcount_t use_refs;
Jens Axboe771b53d02019-10-22 10:25:58 -0600122};
123
Jens Axboe771b53d02019-10-22 10:25:58 -0600124static bool io_worker_get(struct io_worker *worker)
125{
126 return refcount_inc_not_zero(&worker->ref);
127}
128
129static void io_worker_release(struct io_worker *worker)
130{
131 if (refcount_dec_and_test(&worker->ref))
132 wake_up_process(worker->task);
133}
134
135/*
136 * Note: drops the wqe->lock if returning true! The caller must re-acquire
137 * the lock in that case. Some callers need to restart handling if this
138 * happens, so we can't just re-acquire the lock on behalf of the caller.
139 */
140static bool __io_worker_unuse(struct io_wqe *wqe, struct io_worker *worker)
141{
Jens Axboefcb323c2019-10-24 12:39:47 -0600142 bool dropped_lock = false;
143
Jens Axboecccf0ee2020-01-27 16:34:48 -0700144 if (worker->saved_creds) {
145 revert_creds(worker->saved_creds);
146 worker->cur_creds = worker->saved_creds = NULL;
Jens Axboe181e4482019-11-25 08:52:30 -0700147 }
148
Jens Axboefcb323c2019-10-24 12:39:47 -0600149 if (current->files != worker->restore_files) {
150 __acquire(&wqe->lock);
151 spin_unlock_irq(&wqe->lock);
152 dropped_lock = true;
153
154 task_lock(current);
155 current->files = worker->restore_files;
156 task_unlock(current);
157 }
158
Jens Axboe9392a272020-02-06 21:42:51 -0700159 if (current->fs != worker->restore_fs)
160 current->fs = worker->restore_fs;
161
Jens Axboe771b53d02019-10-22 10:25:58 -0600162 /*
163 * If we have an active mm, we need to drop the wq lock before unusing
164 * it. If we do, return true and let the caller retry the idle loop.
165 */
166 if (worker->mm) {
Jens Axboefcb323c2019-10-24 12:39:47 -0600167 if (!dropped_lock) {
168 __acquire(&wqe->lock);
169 spin_unlock_irq(&wqe->lock);
170 dropped_lock = true;
171 }
Jens Axboe771b53d02019-10-22 10:25:58 -0600172 __set_current_state(TASK_RUNNING);
173 set_fs(KERNEL_DS);
174 unuse_mm(worker->mm);
175 mmput(worker->mm);
176 worker->mm = NULL;
Jens Axboe771b53d02019-10-22 10:25:58 -0600177 }
178
Jens Axboefcb323c2019-10-24 12:39:47 -0600179 return dropped_lock;
Jens Axboe771b53d02019-10-22 10:25:58 -0600180}
181
Jens Axboec5def4a2019-11-07 11:41:16 -0700182static inline struct io_wqe_acct *io_work_get_acct(struct io_wqe *wqe,
183 struct io_wq_work *work)
184{
185 if (work->flags & IO_WQ_WORK_UNBOUND)
186 return &wqe->acct[IO_WQ_ACCT_UNBOUND];
187
188 return &wqe->acct[IO_WQ_ACCT_BOUND];
189}
190
191static inline struct io_wqe_acct *io_wqe_get_acct(struct io_wqe *wqe,
192 struct io_worker *worker)
193{
194 if (worker->flags & IO_WORKER_F_BOUND)
195 return &wqe->acct[IO_WQ_ACCT_BOUND];
196
197 return &wqe->acct[IO_WQ_ACCT_UNBOUND];
198}
199
Jens Axboe771b53d02019-10-22 10:25:58 -0600200static void io_worker_exit(struct io_worker *worker)
201{
202 struct io_wqe *wqe = worker->wqe;
Jens Axboec5def4a2019-11-07 11:41:16 -0700203 struct io_wqe_acct *acct = io_wqe_get_acct(wqe, worker);
204 unsigned nr_workers;
Jens Axboe771b53d02019-10-22 10:25:58 -0600205
206 /*
207 * If we're not at zero, someone else is holding a brief reference
208 * to the worker. Wait for that to go away.
209 */
210 set_current_state(TASK_INTERRUPTIBLE);
211 if (!refcount_dec_and_test(&worker->ref))
212 schedule();
213 __set_current_state(TASK_RUNNING);
214
215 preempt_disable();
216 current->flags &= ~PF_IO_WORKER;
217 if (worker->flags & IO_WORKER_F_RUNNING)
Jens Axboec5def4a2019-11-07 11:41:16 -0700218 atomic_dec(&acct->nr_running);
219 if (!(worker->flags & IO_WORKER_F_BOUND))
220 atomic_dec(&wqe->wq->user->processes);
Jens Axboe771b53d02019-10-22 10:25:58 -0600221 worker->flags = 0;
222 preempt_enable();
223
224 spin_lock_irq(&wqe->lock);
225 hlist_nulls_del_rcu(&worker->nulls_node);
Jens Axboee61df662019-11-13 13:54:49 -0700226 list_del_rcu(&worker->all_list);
Jens Axboe771b53d02019-10-22 10:25:58 -0600227 if (__io_worker_unuse(wqe, worker)) {
228 __release(&wqe->lock);
229 spin_lock_irq(&wqe->lock);
230 }
Jens Axboec5def4a2019-11-07 11:41:16 -0700231 acct->nr_workers--;
232 nr_workers = wqe->acct[IO_WQ_ACCT_BOUND].nr_workers +
233 wqe->acct[IO_WQ_ACCT_UNBOUND].nr_workers;
Jens Axboe771b53d02019-10-22 10:25:58 -0600234 spin_unlock_irq(&wqe->lock);
235
236 /* all workers gone, wq exit can proceed */
Jens Axboec5def4a2019-11-07 11:41:16 -0700237 if (!nr_workers && refcount_dec_and_test(&wqe->wq->refs))
Jens Axboe771b53d02019-10-22 10:25:58 -0600238 complete(&wqe->wq->done);
239
YueHaibing364b05f2019-11-02 15:55:01 +0800240 kfree_rcu(worker, rcu);
Jens Axboe771b53d02019-10-22 10:25:58 -0600241}
242
Jens Axboec5def4a2019-11-07 11:41:16 -0700243static inline bool io_wqe_run_queue(struct io_wqe *wqe)
244 __must_hold(wqe->lock)
245{
Jens Axboe6206f0e2019-11-26 11:59:32 -0700246 if (!wq_list_empty(&wqe->work_list) &&
247 !(wqe->flags & IO_WQE_FLAG_STALLED))
Jens Axboec5def4a2019-11-07 11:41:16 -0700248 return true;
249 return false;
250}
251
252/*
253 * Check head of free list for an available worker. If one isn't available,
254 * caller must wake up the wq manager to create one.
255 */
256static bool io_wqe_activate_free_worker(struct io_wqe *wqe)
257 __must_hold(RCU)
258{
259 struct hlist_nulls_node *n;
260 struct io_worker *worker;
261
Jens Axboe021d1cd2019-11-14 08:00:41 -0700262 n = rcu_dereference(hlist_nulls_first_rcu(&wqe->free_list));
Jens Axboec5def4a2019-11-07 11:41:16 -0700263 if (is_a_nulls(n))
264 return false;
265
266 worker = hlist_nulls_entry(n, struct io_worker, nulls_node);
267 if (io_worker_get(worker)) {
Jens Axboe506d95f2019-12-07 21:03:59 -0700268 wake_up_process(worker->task);
Jens Axboec5def4a2019-11-07 11:41:16 -0700269 io_worker_release(worker);
270 return true;
271 }
272
273 return false;
274}
275
276/*
277 * We need a worker. If we find a free one, we're good. If not, and we're
278 * below the max number of workers, wake up the manager to create one.
279 */
280static void io_wqe_wake_worker(struct io_wqe *wqe, struct io_wqe_acct *acct)
281{
282 bool ret;
283
284 /*
285 * Most likely an attempt to queue unbounded work on an io_wq that
286 * wasn't setup with any unbounded workers.
287 */
288 WARN_ON_ONCE(!acct->max_workers);
289
290 rcu_read_lock();
291 ret = io_wqe_activate_free_worker(wqe);
292 rcu_read_unlock();
293
294 if (!ret && acct->nr_workers < acct->max_workers)
295 wake_up_process(wqe->wq->manager);
296}
297
298static void io_wqe_inc_running(struct io_wqe *wqe, struct io_worker *worker)
299{
300 struct io_wqe_acct *acct = io_wqe_get_acct(wqe, worker);
301
302 atomic_inc(&acct->nr_running);
303}
304
305static void io_wqe_dec_running(struct io_wqe *wqe, struct io_worker *worker)
306 __must_hold(wqe->lock)
307{
308 struct io_wqe_acct *acct = io_wqe_get_acct(wqe, worker);
309
310 if (atomic_dec_and_test(&acct->nr_running) && io_wqe_run_queue(wqe))
311 io_wqe_wake_worker(wqe, acct);
312}
313
Jens Axboe771b53d02019-10-22 10:25:58 -0600314static void io_worker_start(struct io_wqe *wqe, struct io_worker *worker)
315{
316 allow_kernel_signal(SIGINT);
317
318 current->flags |= PF_IO_WORKER;
319
320 worker->flags |= (IO_WORKER_F_UP | IO_WORKER_F_RUNNING);
Jens Axboefcb323c2019-10-24 12:39:47 -0600321 worker->restore_files = current->files;
Jens Axboe9392a272020-02-06 21:42:51 -0700322 worker->restore_fs = current->fs;
Jens Axboec5def4a2019-11-07 11:41:16 -0700323 io_wqe_inc_running(wqe, worker);
Jens Axboe771b53d02019-10-22 10:25:58 -0600324}
325
326/*
327 * Worker will start processing some work. Move it to the busy list, if
328 * it's currently on the freelist
329 */
330static void __io_worker_busy(struct io_wqe *wqe, struct io_worker *worker,
331 struct io_wq_work *work)
332 __must_hold(wqe->lock)
333{
Jens Axboec5def4a2019-11-07 11:41:16 -0700334 bool worker_bound, work_bound;
335
Jens Axboe771b53d02019-10-22 10:25:58 -0600336 if (worker->flags & IO_WORKER_F_FREE) {
337 worker->flags &= ~IO_WORKER_F_FREE;
338 hlist_nulls_del_init_rcu(&worker->nulls_node);
Jens Axboe771b53d02019-10-22 10:25:58 -0600339 }
Jens Axboec5def4a2019-11-07 11:41:16 -0700340
341 /*
342 * If worker is moving from bound to unbound (or vice versa), then
343 * ensure we update the running accounting.
344 */
Dan Carpenterb2e9c7d62019-11-19 09:22:16 +0300345 worker_bound = (worker->flags & IO_WORKER_F_BOUND) != 0;
346 work_bound = (work->flags & IO_WQ_WORK_UNBOUND) == 0;
347 if (worker_bound != work_bound) {
Jens Axboec5def4a2019-11-07 11:41:16 -0700348 io_wqe_dec_running(wqe, worker);
349 if (work_bound) {
350 worker->flags |= IO_WORKER_F_BOUND;
351 wqe->acct[IO_WQ_ACCT_UNBOUND].nr_workers--;
352 wqe->acct[IO_WQ_ACCT_BOUND].nr_workers++;
353 atomic_dec(&wqe->wq->user->processes);
354 } else {
355 worker->flags &= ~IO_WORKER_F_BOUND;
356 wqe->acct[IO_WQ_ACCT_UNBOUND].nr_workers++;
357 wqe->acct[IO_WQ_ACCT_BOUND].nr_workers--;
358 atomic_inc(&wqe->wq->user->processes);
359 }
360 io_wqe_inc_running(wqe, worker);
361 }
Jens Axboe771b53d02019-10-22 10:25:58 -0600362}
363
364/*
365 * No work, worker going to sleep. Move to freelist, and unuse mm if we
366 * have one attached. Dropping the mm may potentially sleep, so we drop
367 * the lock in that case and return success. Since the caller has to
368 * retry the loop in that case (we changed task state), we don't regrab
369 * the lock if we return success.
370 */
371static bool __io_worker_idle(struct io_wqe *wqe, struct io_worker *worker)
372 __must_hold(wqe->lock)
373{
374 if (!(worker->flags & IO_WORKER_F_FREE)) {
375 worker->flags |= IO_WORKER_F_FREE;
Jens Axboe021d1cd2019-11-14 08:00:41 -0700376 hlist_nulls_add_head_rcu(&worker->nulls_node, &wqe->free_list);
Jens Axboe771b53d02019-10-22 10:25:58 -0600377 }
378
379 return __io_worker_unuse(wqe, worker);
380}
381
Pavel Begunkov60cf46a2020-03-14 00:31:05 +0300382static inline unsigned int io_get_work_hash(struct io_wq_work *work)
383{
384 return work->flags >> IO_WQ_HASH_SHIFT;
385}
386
387static struct io_wq_work *io_get_next_work(struct io_wqe *wqe)
Jens Axboe771b53d02019-10-22 10:25:58 -0600388 __must_hold(wqe->lock)
389{
Jens Axboe6206f0e2019-11-26 11:59:32 -0700390 struct io_wq_work_node *node, *prev;
Pavel Begunkov86f3cd12020-03-23 22:57:22 +0300391 struct io_wq_work *work, *tail;
Pavel Begunkov60cf46a2020-03-14 00:31:05 +0300392 unsigned int hash;
Jens Axboe771b53d02019-10-22 10:25:58 -0600393
Jens Axboe6206f0e2019-11-26 11:59:32 -0700394 wq_list_for_each(node, prev, &wqe->work_list) {
395 work = container_of(node, struct io_wq_work, list);
396
Jens Axboe771b53d02019-10-22 10:25:58 -0600397 /* not hashed, can run anytime */
Pavel Begunkov8766dd52020-03-14 00:31:04 +0300398 if (!io_wq_is_hashed(work)) {
Pavel Begunkov86f3cd12020-03-23 22:57:22 +0300399 wq_list_del(&wqe->work_list, node, prev);
Jens Axboe771b53d02019-10-22 10:25:58 -0600400 return work;
401 }
402
403 /* hashed, can run if not already running */
Pavel Begunkov60cf46a2020-03-14 00:31:05 +0300404 hash = io_get_work_hash(work);
405 if (!(wqe->hash_map & BIT(hash))) {
406 wqe->hash_map |= BIT(hash);
Pavel Begunkov86f3cd12020-03-23 22:57:22 +0300407 /* all items with this hash lie in [work, tail] */
408 tail = wqe->hash_tail[hash];
409 wqe->hash_tail[hash] = NULL;
410 wq_list_cut(&wqe->work_list, &tail->list, prev);
Jens Axboe771b53d02019-10-22 10:25:58 -0600411 return work;
412 }
413 }
414
415 return NULL;
416}
417
Jens Axboecccf0ee2020-01-27 16:34:48 -0700418static void io_wq_switch_mm(struct io_worker *worker, struct io_wq_work *work)
419{
420 if (worker->mm) {
421 unuse_mm(worker->mm);
422 mmput(worker->mm);
423 worker->mm = NULL;
424 }
425 if (!work->mm) {
426 set_fs(KERNEL_DS);
427 return;
428 }
429 if (mmget_not_zero(work->mm)) {
430 use_mm(work->mm);
431 if (!worker->mm)
432 set_fs(USER_DS);
433 worker->mm = work->mm;
434 /* hang on to this mm */
435 work->mm = NULL;
436 return;
437 }
438
439 /* failed grabbing mm, ensure work gets cancelled */
440 work->flags |= IO_WQ_WORK_CANCEL;
441}
442
443static void io_wq_switch_creds(struct io_worker *worker,
444 struct io_wq_work *work)
445{
446 const struct cred *old_creds = override_creds(work->creds);
447
448 worker->cur_creds = work->creds;
449 if (worker->saved_creds)
450 put_cred(old_creds); /* creds set by previous switch */
451 else
452 worker->saved_creds = old_creds;
453}
454
Pavel Begunkovdc026a72020-03-04 16:14:09 +0300455static void io_impersonate_work(struct io_worker *worker,
456 struct io_wq_work *work)
457{
458 if (work->files && current->files != work->files) {
459 task_lock(current);
460 current->files = work->files;
461 task_unlock(current);
462 }
463 if (work->fs && current->fs != work->fs)
464 current->fs = work->fs;
465 if (work->mm != worker->mm)
466 io_wq_switch_mm(worker, work);
467 if (worker->cur_creds != work->creds)
468 io_wq_switch_creds(worker, work);
469}
470
471static void io_assign_current_work(struct io_worker *worker,
472 struct io_wq_work *work)
473{
Pavel Begunkovd78298e2020-03-14 00:31:03 +0300474 if (work) {
475 /* flush pending signals before assigning new work */
476 if (signal_pending(current))
477 flush_signals(current);
478 cond_resched();
479 }
Pavel Begunkovdc026a72020-03-04 16:14:09 +0300480
481 spin_lock_irq(&worker->lock);
482 worker->cur_work = work;
483 spin_unlock_irq(&worker->lock);
484}
485
Pavel Begunkov60cf46a2020-03-14 00:31:05 +0300486static void io_wqe_enqueue(struct io_wqe *wqe, struct io_wq_work *work);
487
Jens Axboe771b53d02019-10-22 10:25:58 -0600488static void io_worker_handle_work(struct io_worker *worker)
489 __releases(wqe->lock)
490{
Jens Axboe771b53d02019-10-22 10:25:58 -0600491 struct io_wqe *wqe = worker->wqe;
492 struct io_wq *wq = wqe->wq;
493
494 do {
Pavel Begunkov86f3cd12020-03-23 22:57:22 +0300495 struct io_wq_work *work;
Pavel Begunkov60cf46a2020-03-14 00:31:05 +0300496 unsigned int hash;
Pavel Begunkovf462fd32020-03-04 16:14:11 +0300497get_next:
Jens Axboe771b53d02019-10-22 10:25:58 -0600498 /*
Jens Axboe771b53d02019-10-22 10:25:58 -0600499 * If we got some work, mark us as busy. If we didn't, but
500 * the list isn't empty, it means we stalled on hashed work.
501 * Mark us stalled so we don't keep looking for work when we
502 * can't make progress, any work completion or insertion will
503 * clear the stalled flag.
504 */
Pavel Begunkov60cf46a2020-03-14 00:31:05 +0300505 work = io_get_next_work(wqe);
Jens Axboe771b53d02019-10-22 10:25:58 -0600506 if (work)
507 __io_worker_busy(wqe, worker, work);
Jens Axboe6206f0e2019-11-26 11:59:32 -0700508 else if (!wq_list_empty(&wqe->work_list))
Jens Axboe771b53d02019-10-22 10:25:58 -0600509 wqe->flags |= IO_WQE_FLAG_STALLED;
510
511 spin_unlock_irq(&wqe->lock);
512 if (!work)
513 break;
Pavel Begunkov58e39312020-03-04 16:14:10 +0300514 io_assign_current_work(worker, work);
Jens Axboe36c2f922019-11-13 09:43:34 -0700515
Pavel Begunkovdc026a72020-03-04 16:14:09 +0300516 /* handle a whole dependent link */
517 do {
Pavel Begunkov86f3cd12020-03-23 22:57:22 +0300518 struct io_wq_work *old_work, *next_hashed, *linked;
Hillf Dantonfd1c4bc2019-12-24 09:14:29 -0700519
Pavel Begunkov86f3cd12020-03-23 22:57:22 +0300520 next_hashed = wq_next_work(work);
Pavel Begunkov58e39312020-03-04 16:14:10 +0300521 io_impersonate_work(worker, work);
Pavel Begunkovdc026a72020-03-04 16:14:09 +0300522 /*
523 * OK to set IO_WQ_WORK_CANCEL even for uncancellable
524 * work, the worker function will do the right thing.
525 */
526 if (test_bit(IO_WQ_BIT_CANCEL, &wq->state))
527 work->flags |= IO_WQ_WORK_CANCEL;
Jens Axboe36c2f922019-11-13 09:43:34 -0700528
Pavel Begunkov60cf46a2020-03-14 00:31:05 +0300529 hash = io_get_work_hash(work);
Pavel Begunkov86f3cd12020-03-23 22:57:22 +0300530 linked = old_work = work;
531 linked->func(&linked);
532 linked = (old_work == linked) ? NULL : linked;
Pavel Begunkovf2cf1142020-03-22 19:14:26 +0300533
Pavel Begunkov86f3cd12020-03-23 22:57:22 +0300534 work = next_hashed;
535 if (!work && linked && !io_wq_is_hashed(linked)) {
536 work = linked;
537 linked = NULL;
538 }
539 io_assign_current_work(worker, work);
Pavel Begunkove9fd9392020-03-04 16:14:12 +0300540 wq->free_work(old_work);
Pavel Begunkovdc026a72020-03-04 16:14:09 +0300541
Pavel Begunkov86f3cd12020-03-23 22:57:22 +0300542 if (linked)
543 io_wqe_enqueue(wqe, linked);
544
545 if (hash != -1U && !next_hashed) {
Pavel Begunkovdc026a72020-03-04 16:14:09 +0300546 spin_lock_irq(&wqe->lock);
547 wqe->hash_map &= ~BIT_ULL(hash);
548 wqe->flags &= ~IO_WQE_FLAG_STALLED;
Pavel Begunkovdc026a72020-03-04 16:14:09 +0300549 /* dependent work is not hashed */
550 hash = -1U;
Pavel Begunkovf462fd32020-03-04 16:14:11 +0300551 /* skip unnecessary unlock-lock wqe->lock */
552 if (!work)
553 goto get_next;
554 spin_unlock_irq(&wqe->lock);
Pavel Begunkovdc026a72020-03-04 16:14:09 +0300555 }
Pavel Begunkov58e39312020-03-04 16:14:10 +0300556 } while (work);
Jens Axboe36c2f922019-11-13 09:43:34 -0700557
558 spin_lock_irq(&wqe->lock);
Jens Axboe771b53d02019-10-22 10:25:58 -0600559 } while (1);
560}
561
Jens Axboe771b53d02019-10-22 10:25:58 -0600562static int io_wqe_worker(void *data)
563{
564 struct io_worker *worker = data;
565 struct io_wqe *wqe = worker->wqe;
566 struct io_wq *wq = wqe->wq;
Jens Axboe771b53d02019-10-22 10:25:58 -0600567
568 io_worker_start(wqe, worker);
569
570 while (!test_bit(IO_WQ_BIT_EXIT, &wq->state)) {
Jens Axboe506d95f2019-12-07 21:03:59 -0700571 set_current_state(TASK_INTERRUPTIBLE);
Jens Axboee995d512019-12-07 21:06:46 -0700572loop:
Jens Axboe771b53d02019-10-22 10:25:58 -0600573 spin_lock_irq(&wqe->lock);
574 if (io_wqe_run_queue(wqe)) {
575 __set_current_state(TASK_RUNNING);
576 io_worker_handle_work(worker);
Jens Axboee995d512019-12-07 21:06:46 -0700577 goto loop;
Jens Axboe771b53d02019-10-22 10:25:58 -0600578 }
579 /* drops the lock on success, retry */
580 if (__io_worker_idle(wqe, worker)) {
581 __release(&wqe->lock);
Jens Axboee995d512019-12-07 21:06:46 -0700582 goto loop;
Jens Axboe771b53d02019-10-22 10:25:58 -0600583 }
584 spin_unlock_irq(&wqe->lock);
585 if (signal_pending(current))
586 flush_signals(current);
587 if (schedule_timeout(WORKER_IDLE_TIMEOUT))
588 continue;
589 /* timed out, exit unless we're the fixed worker */
590 if (test_bit(IO_WQ_BIT_EXIT, &wq->state) ||
591 !(worker->flags & IO_WORKER_F_FIXED))
592 break;
593 }
594
Jens Axboe771b53d02019-10-22 10:25:58 -0600595 if (test_bit(IO_WQ_BIT_EXIT, &wq->state)) {
596 spin_lock_irq(&wqe->lock);
Jens Axboe6206f0e2019-11-26 11:59:32 -0700597 if (!wq_list_empty(&wqe->work_list))
Jens Axboe771b53d02019-10-22 10:25:58 -0600598 io_worker_handle_work(worker);
599 else
600 spin_unlock_irq(&wqe->lock);
601 }
602
603 io_worker_exit(worker);
604 return 0;
605}
606
607/*
Jens Axboe771b53d02019-10-22 10:25:58 -0600608 * Called when a worker is scheduled in. Mark us as currently running.
609 */
610void io_wq_worker_running(struct task_struct *tsk)
611{
612 struct io_worker *worker = kthread_data(tsk);
613 struct io_wqe *wqe = worker->wqe;
614
615 if (!(worker->flags & IO_WORKER_F_UP))
616 return;
617 if (worker->flags & IO_WORKER_F_RUNNING)
618 return;
619 worker->flags |= IO_WORKER_F_RUNNING;
Jens Axboec5def4a2019-11-07 11:41:16 -0700620 io_wqe_inc_running(wqe, worker);
Jens Axboe771b53d02019-10-22 10:25:58 -0600621}
622
623/*
624 * Called when worker is going to sleep. If there are no workers currently
625 * running and we have work pending, wake up a free one or have the manager
626 * set one up.
627 */
628void io_wq_worker_sleeping(struct task_struct *tsk)
629{
630 struct io_worker *worker = kthread_data(tsk);
631 struct io_wqe *wqe = worker->wqe;
632
633 if (!(worker->flags & IO_WORKER_F_UP))
634 return;
635 if (!(worker->flags & IO_WORKER_F_RUNNING))
636 return;
637
638 worker->flags &= ~IO_WORKER_F_RUNNING;
639
640 spin_lock_irq(&wqe->lock);
Jens Axboec5def4a2019-11-07 11:41:16 -0700641 io_wqe_dec_running(wqe, worker);
Jens Axboe771b53d02019-10-22 10:25:58 -0600642 spin_unlock_irq(&wqe->lock);
643}
644
Jens Axboeb60fda62019-11-19 08:37:07 -0700645static bool create_io_worker(struct io_wq *wq, struct io_wqe *wqe, int index)
Jens Axboe771b53d02019-10-22 10:25:58 -0600646{
Jens Axboec5def4a2019-11-07 11:41:16 -0700647 struct io_wqe_acct *acct =&wqe->acct[index];
Jens Axboe771b53d02019-10-22 10:25:58 -0600648 struct io_worker *worker;
649
Jann Hornad6e0052019-11-26 17:39:45 +0100650 worker = kzalloc_node(sizeof(*worker), GFP_KERNEL, wqe->node);
Jens Axboe771b53d02019-10-22 10:25:58 -0600651 if (!worker)
Jens Axboeb60fda62019-11-19 08:37:07 -0700652 return false;
Jens Axboe771b53d02019-10-22 10:25:58 -0600653
654 refcount_set(&worker->ref, 1);
655 worker->nulls_node.pprev = NULL;
Jens Axboe771b53d02019-10-22 10:25:58 -0600656 worker->wqe = wqe;
Jens Axboe36c2f922019-11-13 09:43:34 -0700657 spin_lock_init(&worker->lock);
Jens Axboe771b53d02019-10-22 10:25:58 -0600658
659 worker->task = kthread_create_on_node(io_wqe_worker, worker, wqe->node,
Jens Axboec5def4a2019-11-07 11:41:16 -0700660 "io_wqe_worker-%d/%d", index, wqe->node);
Jens Axboe771b53d02019-10-22 10:25:58 -0600661 if (IS_ERR(worker->task)) {
662 kfree(worker);
Jens Axboeb60fda62019-11-19 08:37:07 -0700663 return false;
Jens Axboe771b53d02019-10-22 10:25:58 -0600664 }
665
666 spin_lock_irq(&wqe->lock);
Jens Axboe021d1cd2019-11-14 08:00:41 -0700667 hlist_nulls_add_head_rcu(&worker->nulls_node, &wqe->free_list);
Jens Axboee61df662019-11-13 13:54:49 -0700668 list_add_tail_rcu(&worker->all_list, &wqe->all_list);
Jens Axboe771b53d02019-10-22 10:25:58 -0600669 worker->flags |= IO_WORKER_F_FREE;
Jens Axboec5def4a2019-11-07 11:41:16 -0700670 if (index == IO_WQ_ACCT_BOUND)
671 worker->flags |= IO_WORKER_F_BOUND;
672 if (!acct->nr_workers && (worker->flags & IO_WORKER_F_BOUND))
Jens Axboe771b53d02019-10-22 10:25:58 -0600673 worker->flags |= IO_WORKER_F_FIXED;
Jens Axboec5def4a2019-11-07 11:41:16 -0700674 acct->nr_workers++;
Jens Axboe771b53d02019-10-22 10:25:58 -0600675 spin_unlock_irq(&wqe->lock);
676
Jens Axboec5def4a2019-11-07 11:41:16 -0700677 if (index == IO_WQ_ACCT_UNBOUND)
678 atomic_inc(&wq->user->processes);
679
Jens Axboe771b53d02019-10-22 10:25:58 -0600680 wake_up_process(worker->task);
Jens Axboeb60fda62019-11-19 08:37:07 -0700681 return true;
Jens Axboe771b53d02019-10-22 10:25:58 -0600682}
683
Jens Axboec5def4a2019-11-07 11:41:16 -0700684static inline bool io_wqe_need_worker(struct io_wqe *wqe, int index)
Jens Axboe771b53d02019-10-22 10:25:58 -0600685 __must_hold(wqe->lock)
686{
Jens Axboec5def4a2019-11-07 11:41:16 -0700687 struct io_wqe_acct *acct = &wqe->acct[index];
Jens Axboe771b53d02019-10-22 10:25:58 -0600688
Jens Axboec5def4a2019-11-07 11:41:16 -0700689 /* if we have available workers or no work, no need */
Jens Axboe021d1cd2019-11-14 08:00:41 -0700690 if (!hlist_nulls_empty(&wqe->free_list) || !io_wqe_run_queue(wqe))
Jens Axboec5def4a2019-11-07 11:41:16 -0700691 return false;
692 return acct->nr_workers < acct->max_workers;
Jens Axboe771b53d02019-10-22 10:25:58 -0600693}
694
695/*
696 * Manager thread. Tasked with creating new workers, if we need them.
697 */
698static int io_wq_manager(void *data)
699{
700 struct io_wq *wq = data;
Jann Horn3fc50ab2019-11-26 19:10:20 +0100701 int workers_to_create = num_possible_nodes();
702 int node;
Jens Axboeb60fda62019-11-19 08:37:07 -0700703
704 /* create fixed workers */
Jann Horn3fc50ab2019-11-26 19:10:20 +0100705 refcount_set(&wq->refs, workers_to_create);
706 for_each_node(node) {
Jens Axboe75634392020-02-11 06:30:06 -0700707 if (!node_online(node))
708 continue;
Jann Horn3fc50ab2019-11-26 19:10:20 +0100709 if (!create_io_worker(wq, wq->wqes[node], IO_WQ_ACCT_BOUND))
710 goto err;
711 workers_to_create--;
Jens Axboeb60fda62019-11-19 08:37:07 -0700712 }
713
Jens Axboe75634392020-02-11 06:30:06 -0700714 while (workers_to_create--)
715 refcount_dec(&wq->refs);
716
Jens Axboeb60fda62019-11-19 08:37:07 -0700717 complete(&wq->done);
Jens Axboe771b53d02019-10-22 10:25:58 -0600718
719 while (!kthread_should_stop()) {
Jens Axboeaa96bf82020-04-03 11:26:26 -0600720 if (current->task_works)
721 task_work_run();
722
Jann Horn3fc50ab2019-11-26 19:10:20 +0100723 for_each_node(node) {
724 struct io_wqe *wqe = wq->wqes[node];
Jens Axboec5def4a2019-11-07 11:41:16 -0700725 bool fork_worker[2] = { false, false };
Jens Axboe771b53d02019-10-22 10:25:58 -0600726
Jens Axboe75634392020-02-11 06:30:06 -0700727 if (!node_online(node))
728 continue;
729
Jens Axboe771b53d02019-10-22 10:25:58 -0600730 spin_lock_irq(&wqe->lock);
Jens Axboec5def4a2019-11-07 11:41:16 -0700731 if (io_wqe_need_worker(wqe, IO_WQ_ACCT_BOUND))
732 fork_worker[IO_WQ_ACCT_BOUND] = true;
733 if (io_wqe_need_worker(wqe, IO_WQ_ACCT_UNBOUND))
734 fork_worker[IO_WQ_ACCT_UNBOUND] = true;
Jens Axboe771b53d02019-10-22 10:25:58 -0600735 spin_unlock_irq(&wqe->lock);
Jens Axboec5def4a2019-11-07 11:41:16 -0700736 if (fork_worker[IO_WQ_ACCT_BOUND])
737 create_io_worker(wq, wqe, IO_WQ_ACCT_BOUND);
738 if (fork_worker[IO_WQ_ACCT_UNBOUND])
739 create_io_worker(wq, wqe, IO_WQ_ACCT_UNBOUND);
Jens Axboe771b53d02019-10-22 10:25:58 -0600740 }
741 set_current_state(TASK_INTERRUPTIBLE);
742 schedule_timeout(HZ);
743 }
744
Jens Axboeaa96bf82020-04-03 11:26:26 -0600745 if (current->task_works)
746 task_work_run();
747
Jens Axboe771b53d02019-10-22 10:25:58 -0600748 return 0;
Jens Axboeb60fda62019-11-19 08:37:07 -0700749err:
750 set_bit(IO_WQ_BIT_ERROR, &wq->state);
751 set_bit(IO_WQ_BIT_EXIT, &wq->state);
Jann Horn3fc50ab2019-11-26 19:10:20 +0100752 if (refcount_sub_and_test(workers_to_create, &wq->refs))
Jens Axboeb60fda62019-11-19 08:37:07 -0700753 complete(&wq->done);
754 return 0;
Jens Axboe771b53d02019-10-22 10:25:58 -0600755}
756
Jens Axboec5def4a2019-11-07 11:41:16 -0700757static bool io_wq_can_queue(struct io_wqe *wqe, struct io_wqe_acct *acct,
758 struct io_wq_work *work)
759{
760 bool free_worker;
761
762 if (!(work->flags & IO_WQ_WORK_UNBOUND))
763 return true;
764 if (atomic_read(&acct->nr_running))
765 return true;
766
767 rcu_read_lock();
Jens Axboe021d1cd2019-11-14 08:00:41 -0700768 free_worker = !hlist_nulls_empty(&wqe->free_list);
Jens Axboec5def4a2019-11-07 11:41:16 -0700769 rcu_read_unlock();
770 if (free_worker)
771 return true;
772
773 if (atomic_read(&wqe->wq->user->processes) >= acct->max_workers &&
774 !(capable(CAP_SYS_RESOURCE) || capable(CAP_SYS_ADMIN)))
775 return false;
776
777 return true;
778}
779
Pavel Begunkove9fd9392020-03-04 16:14:12 +0300780static void io_run_cancel(struct io_wq_work *work, struct io_wqe *wqe)
Pavel Begunkovfc04c392020-03-01 19:18:19 +0300781{
Pavel Begunkove9fd9392020-03-04 16:14:12 +0300782 struct io_wq *wq = wqe->wq;
783
Pavel Begunkovfc04c392020-03-01 19:18:19 +0300784 do {
785 struct io_wq_work *old_work = work;
786
787 work->flags |= IO_WQ_WORK_CANCEL;
788 work->func(&work);
789 work = (work == old_work) ? NULL : work;
Pavel Begunkove9fd9392020-03-04 16:14:12 +0300790 wq->free_work(old_work);
Pavel Begunkovfc04c392020-03-01 19:18:19 +0300791 } while (work);
792}
793
Pavel Begunkov86f3cd12020-03-23 22:57:22 +0300794static void io_wqe_insert_work(struct io_wqe *wqe, struct io_wq_work *work)
795{
796 unsigned int hash;
797 struct io_wq_work *tail;
798
799 if (!io_wq_is_hashed(work)) {
800append:
801 wq_list_add_tail(&work->list, &wqe->work_list);
802 return;
803 }
804
805 hash = io_get_work_hash(work);
806 tail = wqe->hash_tail[hash];
807 wqe->hash_tail[hash] = work;
808 if (!tail)
809 goto append;
810
811 wq_list_add_after(&work->list, &tail->list, &wqe->work_list);
812}
813
Jens Axboe771b53d02019-10-22 10:25:58 -0600814static void io_wqe_enqueue(struct io_wqe *wqe, struct io_wq_work *work)
815{
Jens Axboec5def4a2019-11-07 11:41:16 -0700816 struct io_wqe_acct *acct = io_work_get_acct(wqe, work);
Jens Axboe895e2ca2019-12-17 08:46:33 -0700817 int work_flags;
Jens Axboe771b53d02019-10-22 10:25:58 -0600818 unsigned long flags;
819
Jens Axboec5def4a2019-11-07 11:41:16 -0700820 /*
821 * Do early check to see if we need a new unbound worker, and if we do,
822 * if we're allowed to do so. This isn't 100% accurate as there's a
823 * gap between this check and incrementing the value, but that's OK.
824 * It's close enough to not be an issue, fork() has the same delay.
825 */
826 if (unlikely(!io_wq_can_queue(wqe, acct, work))) {
Pavel Begunkove9fd9392020-03-04 16:14:12 +0300827 io_run_cancel(work, wqe);
Jens Axboec5def4a2019-11-07 11:41:16 -0700828 return;
829 }
830
Jens Axboe895e2ca2019-12-17 08:46:33 -0700831 work_flags = work->flags;
Jens Axboe771b53d02019-10-22 10:25:58 -0600832 spin_lock_irqsave(&wqe->lock, flags);
Pavel Begunkov86f3cd12020-03-23 22:57:22 +0300833 io_wqe_insert_work(wqe, work);
Jens Axboe771b53d02019-10-22 10:25:58 -0600834 wqe->flags &= ~IO_WQE_FLAG_STALLED;
835 spin_unlock_irqrestore(&wqe->lock, flags);
836
Jens Axboe895e2ca2019-12-17 08:46:33 -0700837 if ((work_flags & IO_WQ_WORK_CONCURRENT) ||
838 !atomic_read(&acct->nr_running))
Jens Axboec5def4a2019-11-07 11:41:16 -0700839 io_wqe_wake_worker(wqe, acct);
Jens Axboe771b53d02019-10-22 10:25:58 -0600840}
841
842void io_wq_enqueue(struct io_wq *wq, struct io_wq_work *work)
843{
844 struct io_wqe *wqe = wq->wqes[numa_node_id()];
845
846 io_wqe_enqueue(wqe, work);
847}
848
849/*
Pavel Begunkov8766dd52020-03-14 00:31:04 +0300850 * Work items that hash to the same value will not be done in parallel.
851 * Used to limit concurrent writes, generally hashed by inode.
Jens Axboe771b53d02019-10-22 10:25:58 -0600852 */
Pavel Begunkov8766dd52020-03-14 00:31:04 +0300853void io_wq_hash_work(struct io_wq_work *work, void *val)
Jens Axboe771b53d02019-10-22 10:25:58 -0600854{
Pavel Begunkov8766dd52020-03-14 00:31:04 +0300855 unsigned int bit;
Jens Axboe771b53d02019-10-22 10:25:58 -0600856
857 bit = hash_ptr(val, IO_WQ_HASH_ORDER);
858 work->flags |= (IO_WQ_WORK_HASHED | (bit << IO_WQ_HASH_SHIFT));
Jens Axboe771b53d02019-10-22 10:25:58 -0600859}
860
861static bool io_wqe_worker_send_sig(struct io_worker *worker, void *data)
862{
863 send_sig(SIGINT, worker->task, 1);
864 return false;
865}
866
867/*
868 * Iterate the passed in list and call the specific function for each
869 * worker that isn't exiting
870 */
871static bool io_wq_for_each_worker(struct io_wqe *wqe,
Jens Axboe771b53d02019-10-22 10:25:58 -0600872 bool (*func)(struct io_worker *, void *),
873 void *data)
874{
Jens Axboe771b53d02019-10-22 10:25:58 -0600875 struct io_worker *worker;
876 bool ret = false;
877
Jens Axboee61df662019-11-13 13:54:49 -0700878 list_for_each_entry_rcu(worker, &wqe->all_list, all_list) {
Jens Axboe771b53d02019-10-22 10:25:58 -0600879 if (io_worker_get(worker)) {
Jens Axboe75634392020-02-11 06:30:06 -0700880 /* no task if node is/was offline */
881 if (worker->task)
882 ret = func(worker, data);
Jens Axboe771b53d02019-10-22 10:25:58 -0600883 io_worker_release(worker);
884 if (ret)
885 break;
886 }
887 }
Jens Axboee61df662019-11-13 13:54:49 -0700888
Jens Axboe771b53d02019-10-22 10:25:58 -0600889 return ret;
890}
891
892void io_wq_cancel_all(struct io_wq *wq)
893{
Jann Horn3fc50ab2019-11-26 19:10:20 +0100894 int node;
Jens Axboe771b53d02019-10-22 10:25:58 -0600895
896 set_bit(IO_WQ_BIT_CANCEL, &wq->state);
897
Jens Axboe771b53d02019-10-22 10:25:58 -0600898 rcu_read_lock();
Jann Horn3fc50ab2019-11-26 19:10:20 +0100899 for_each_node(node) {
900 struct io_wqe *wqe = wq->wqes[node];
Jens Axboe771b53d02019-10-22 10:25:58 -0600901
Jens Axboee61df662019-11-13 13:54:49 -0700902 io_wq_for_each_worker(wqe, io_wqe_worker_send_sig, NULL);
Jens Axboe771b53d02019-10-22 10:25:58 -0600903 }
904 rcu_read_unlock();
905}
906
Jens Axboe62755e32019-10-28 21:49:21 -0600907struct io_cb_cancel_data {
Pavel Begunkov2293b412020-03-07 01:15:39 +0300908 work_cancel_fn *fn;
909 void *data;
Jens Axboe62755e32019-10-28 21:49:21 -0600910};
911
Pavel Begunkov2293b412020-03-07 01:15:39 +0300912static bool io_wq_worker_cancel(struct io_worker *worker, void *data)
Jens Axboe62755e32019-10-28 21:49:21 -0600913{
Pavel Begunkov2293b412020-03-07 01:15:39 +0300914 struct io_cb_cancel_data *match = data;
Jens Axboe6f726532019-11-05 13:51:51 -0700915 unsigned long flags;
Jens Axboe62755e32019-10-28 21:49:21 -0600916 bool ret = false;
917
918 /*
919 * Hold the lock to avoid ->cur_work going out of scope, caller
Jens Axboe36c2f922019-11-13 09:43:34 -0700920 * may dereference the passed in work.
Jens Axboe62755e32019-10-28 21:49:21 -0600921 */
Jens Axboe36c2f922019-11-13 09:43:34 -0700922 spin_lock_irqsave(&worker->lock, flags);
Jens Axboe62755e32019-10-28 21:49:21 -0600923 if (worker->cur_work &&
Jens Axboe0c9d5cc2019-12-11 19:29:43 -0700924 !(worker->cur_work->flags & IO_WQ_WORK_NO_CANCEL) &&
Pavel Begunkov2293b412020-03-07 01:15:39 +0300925 match->fn(worker->cur_work, match->data)) {
Jens Axboe771b53d02019-10-22 10:25:58 -0600926 send_sig(SIGINT, worker->task, 1);
Jens Axboe36c2f922019-11-13 09:43:34 -0700927 ret = true;
Jens Axboe771b53d02019-10-22 10:25:58 -0600928 }
Jens Axboe36c2f922019-11-13 09:43:34 -0700929 spin_unlock_irqrestore(&worker->lock, flags);
Jens Axboe771b53d02019-10-22 10:25:58 -0600930
Jens Axboe36c2f922019-11-13 09:43:34 -0700931 return ret;
Jens Axboe771b53d02019-10-22 10:25:58 -0600932}
933
934static enum io_wq_cancel io_wqe_cancel_work(struct io_wqe *wqe,
Pavel Begunkov2293b412020-03-07 01:15:39 +0300935 struct io_cb_cancel_data *match)
Jens Axboe771b53d02019-10-22 10:25:58 -0600936{
Jens Axboe6206f0e2019-11-26 11:59:32 -0700937 struct io_wq_work_node *node, *prev;
Jens Axboe771b53d02019-10-22 10:25:58 -0600938 struct io_wq_work *work;
Jens Axboe6f726532019-11-05 13:51:51 -0700939 unsigned long flags;
Jens Axboe771b53d02019-10-22 10:25:58 -0600940 bool found = false;
941
Jens Axboe771b53d02019-10-22 10:25:58 -0600942 /*
943 * First check pending list, if we're lucky we can just remove it
944 * from there. CANCEL_OK means that the work is returned as-new,
945 * no completion will be posted for it.
946 */
Jens Axboe6f726532019-11-05 13:51:51 -0700947 spin_lock_irqsave(&wqe->lock, flags);
Jens Axboe6206f0e2019-11-26 11:59:32 -0700948 wq_list_for_each(node, prev, &wqe->work_list) {
949 work = container_of(node, struct io_wq_work, list);
950
Jens Axboe00bcda12020-02-08 19:13:32 -0700951 if (match->fn(work, match->data)) {
Pavel Begunkov86f3cd12020-03-23 22:57:22 +0300952 wq_list_del(&wqe->work_list, node, prev);
Jens Axboe771b53d02019-10-22 10:25:58 -0600953 found = true;
954 break;
955 }
956 }
Jens Axboe6f726532019-11-05 13:51:51 -0700957 spin_unlock_irqrestore(&wqe->lock, flags);
Jens Axboe771b53d02019-10-22 10:25:58 -0600958
959 if (found) {
Pavel Begunkove9fd9392020-03-04 16:14:12 +0300960 io_run_cancel(work, wqe);
Jens Axboe771b53d02019-10-22 10:25:58 -0600961 return IO_WQ_CANCEL_OK;
962 }
963
964 /*
965 * Now check if a free (going busy) or busy worker has the work
966 * currently running. If we find it there, we'll return CANCEL_RUNNING
Brian Gianforcarod195a662019-12-13 03:09:50 -0800967 * as an indication that we attempt to signal cancellation. The
Jens Axboe771b53d02019-10-22 10:25:58 -0600968 * completion will run normally in this case.
969 */
970 rcu_read_lock();
Jens Axboe00bcda12020-02-08 19:13:32 -0700971 found = io_wq_for_each_worker(wqe, io_wq_worker_cancel, match);
Jens Axboe771b53d02019-10-22 10:25:58 -0600972 rcu_read_unlock();
973 return found ? IO_WQ_CANCEL_RUNNING : IO_WQ_CANCEL_NOTFOUND;
974}
975
Pavel Begunkov2293b412020-03-07 01:15:39 +0300976enum io_wq_cancel io_wq_cancel_cb(struct io_wq *wq, work_cancel_fn *cancel,
977 void *data)
978{
979 struct io_cb_cancel_data match = {
980 .fn = cancel,
981 .data = data,
982 };
983 enum io_wq_cancel ret = IO_WQ_CANCEL_NOTFOUND;
984 int node;
985
986 for_each_node(node) {
987 struct io_wqe *wqe = wq->wqes[node];
988
989 ret = io_wqe_cancel_work(wqe, &match);
990 if (ret != IO_WQ_CANCEL_NOTFOUND)
991 break;
992 }
993
994 return ret;
995}
996
997static bool io_wq_io_cb_cancel_data(struct io_wq_work *work, void *data)
Jens Axboe00bcda12020-02-08 19:13:32 -0700998{
999 return work == data;
1000}
1001
Jens Axboe771b53d02019-10-22 10:25:58 -06001002enum io_wq_cancel io_wq_cancel_work(struct io_wq *wq, struct io_wq_work *cwork)
1003{
Pavel Begunkov2293b412020-03-07 01:15:39 +03001004 return io_wq_cancel_cb(wq, io_wq_io_cb_cancel_data, (void *)cwork);
Jens Axboe771b53d02019-10-22 10:25:58 -06001005}
1006
Jens Axboe36282882020-02-08 19:16:39 -07001007static bool io_wq_pid_match(struct io_wq_work *work, void *data)
1008{
1009 pid_t pid = (pid_t) (unsigned long) data;
1010
Pavel Begunkov2293b412020-03-07 01:15:39 +03001011 return work->task_pid == pid;
Jens Axboe36282882020-02-08 19:16:39 -07001012}
1013
1014enum io_wq_cancel io_wq_cancel_pid(struct io_wq *wq, pid_t pid)
1015{
Pavel Begunkov2293b412020-03-07 01:15:39 +03001016 void *data = (void *) (unsigned long) pid;
Jens Axboe36282882020-02-08 19:16:39 -07001017
Pavel Begunkov2293b412020-03-07 01:15:39 +03001018 return io_wq_cancel_cb(wq, io_wq_pid_match, data);
Jens Axboe36282882020-02-08 19:16:39 -07001019}
1020
Jens Axboe576a3472019-11-25 08:49:20 -07001021struct io_wq *io_wq_create(unsigned bounded, struct io_wq_data *data)
Jens Axboe771b53d02019-10-22 10:25:58 -06001022{
Jann Horn3fc50ab2019-11-26 19:10:20 +01001023 int ret = -ENOMEM, node;
Jens Axboe771b53d02019-10-22 10:25:58 -06001024 struct io_wq *wq;
1025
Pavel Begunkove9fd9392020-03-04 16:14:12 +03001026 if (WARN_ON_ONCE(!data->free_work))
1027 return ERR_PTR(-EINVAL);
1028
Jann Hornad6e0052019-11-26 17:39:45 +01001029 wq = kzalloc(sizeof(*wq), GFP_KERNEL);
Jens Axboe771b53d02019-10-22 10:25:58 -06001030 if (!wq)
1031 return ERR_PTR(-ENOMEM);
1032
Jann Horn3fc50ab2019-11-26 19:10:20 +01001033 wq->wqes = kcalloc(nr_node_ids, sizeof(struct io_wqe *), GFP_KERNEL);
Jens Axboe771b53d02019-10-22 10:25:58 -06001034 if (!wq->wqes) {
1035 kfree(wq);
1036 return ERR_PTR(-ENOMEM);
1037 }
1038
Pavel Begunkove9fd9392020-03-04 16:14:12 +03001039 wq->free_work = data->free_work;
Jens Axboe7d723062019-11-12 22:31:31 -07001040
Jens Axboec5def4a2019-11-07 11:41:16 -07001041 /* caller must already hold a reference to this */
Jens Axboe576a3472019-11-25 08:49:20 -07001042 wq->user = data->user;
Jens Axboec5def4a2019-11-07 11:41:16 -07001043
Jann Horn3fc50ab2019-11-26 19:10:20 +01001044 for_each_node(node) {
Jens Axboe771b53d02019-10-22 10:25:58 -06001045 struct io_wqe *wqe;
Jens Axboe75634392020-02-11 06:30:06 -07001046 int alloc_node = node;
Jens Axboe771b53d02019-10-22 10:25:58 -06001047
Jens Axboe75634392020-02-11 06:30:06 -07001048 if (!node_online(alloc_node))
1049 alloc_node = NUMA_NO_NODE;
1050 wqe = kzalloc_node(sizeof(struct io_wqe), GFP_KERNEL, alloc_node);
Jens Axboe771b53d02019-10-22 10:25:58 -06001051 if (!wqe)
Jann Horn3fc50ab2019-11-26 19:10:20 +01001052 goto err;
1053 wq->wqes[node] = wqe;
Jens Axboe75634392020-02-11 06:30:06 -07001054 wqe->node = alloc_node;
Jens Axboec5def4a2019-11-07 11:41:16 -07001055 wqe->acct[IO_WQ_ACCT_BOUND].max_workers = bounded;
1056 atomic_set(&wqe->acct[IO_WQ_ACCT_BOUND].nr_running, 0);
Jens Axboe576a3472019-11-25 08:49:20 -07001057 if (wq->user) {
Jens Axboec5def4a2019-11-07 11:41:16 -07001058 wqe->acct[IO_WQ_ACCT_UNBOUND].max_workers =
1059 task_rlimit(current, RLIMIT_NPROC);
1060 }
1061 atomic_set(&wqe->acct[IO_WQ_ACCT_UNBOUND].nr_running, 0);
Jens Axboe771b53d02019-10-22 10:25:58 -06001062 wqe->wq = wq;
1063 spin_lock_init(&wqe->lock);
Jens Axboe6206f0e2019-11-26 11:59:32 -07001064 INIT_WQ_LIST(&wqe->work_list);
Jens Axboe021d1cd2019-11-14 08:00:41 -07001065 INIT_HLIST_NULLS_HEAD(&wqe->free_list, 0);
Jens Axboee61df662019-11-13 13:54:49 -07001066 INIT_LIST_HEAD(&wqe->all_list);
Jens Axboe771b53d02019-10-22 10:25:58 -06001067 }
1068
1069 init_completion(&wq->done);
1070
Jens Axboe771b53d02019-10-22 10:25:58 -06001071 wq->manager = kthread_create(io_wq_manager, wq, "io_wq_manager");
1072 if (!IS_ERR(wq->manager)) {
1073 wake_up_process(wq->manager);
Jens Axboeb60fda62019-11-19 08:37:07 -07001074 wait_for_completion(&wq->done);
1075 if (test_bit(IO_WQ_BIT_ERROR, &wq->state)) {
1076 ret = -ENOMEM;
1077 goto err;
1078 }
Jens Axboe848f7e12020-01-23 15:33:32 -07001079 refcount_set(&wq->use_refs, 1);
Jens Axboeb60fda62019-11-19 08:37:07 -07001080 reinit_completion(&wq->done);
Jens Axboe771b53d02019-10-22 10:25:58 -06001081 return wq;
1082 }
1083
1084 ret = PTR_ERR(wq->manager);
Jens Axboe771b53d02019-10-22 10:25:58 -06001085 complete(&wq->done);
Jens Axboeb60fda62019-11-19 08:37:07 -07001086err:
Jann Horn3fc50ab2019-11-26 19:10:20 +01001087 for_each_node(node)
1088 kfree(wq->wqes[node]);
Jens Axboeb60fda62019-11-19 08:37:07 -07001089 kfree(wq->wqes);
1090 kfree(wq);
Jens Axboe771b53d02019-10-22 10:25:58 -06001091 return ERR_PTR(ret);
1092}
1093
Pavel Begunkoveba6f5a2020-01-28 03:15:47 +03001094bool io_wq_get(struct io_wq *wq, struct io_wq_data *data)
1095{
Pavel Begunkove9fd9392020-03-04 16:14:12 +03001096 if (data->free_work != wq->free_work)
Pavel Begunkoveba6f5a2020-01-28 03:15:47 +03001097 return false;
1098
1099 return refcount_inc_not_zero(&wq->use_refs);
1100}
1101
Jens Axboe771b53d02019-10-22 10:25:58 -06001102static bool io_wq_worker_wake(struct io_worker *worker, void *data)
1103{
1104 wake_up_process(worker->task);
1105 return false;
1106}
1107
Jens Axboe848f7e12020-01-23 15:33:32 -07001108static void __io_wq_destroy(struct io_wq *wq)
Jens Axboe771b53d02019-10-22 10:25:58 -06001109{
Jann Horn3fc50ab2019-11-26 19:10:20 +01001110 int node;
Jens Axboe771b53d02019-10-22 10:25:58 -06001111
Jens Axboeb60fda62019-11-19 08:37:07 -07001112 set_bit(IO_WQ_BIT_EXIT, &wq->state);
1113 if (wq->manager)
Jens Axboe771b53d02019-10-22 10:25:58 -06001114 kthread_stop(wq->manager);
Jens Axboe771b53d02019-10-22 10:25:58 -06001115
1116 rcu_read_lock();
Jann Horn3fc50ab2019-11-26 19:10:20 +01001117 for_each_node(node)
1118 io_wq_for_each_worker(wq->wqes[node], io_wq_worker_wake, NULL);
Jens Axboe771b53d02019-10-22 10:25:58 -06001119 rcu_read_unlock();
1120
1121 wait_for_completion(&wq->done);
1122
Jann Horn3fc50ab2019-11-26 19:10:20 +01001123 for_each_node(node)
1124 kfree(wq->wqes[node]);
Jens Axboe771b53d02019-10-22 10:25:58 -06001125 kfree(wq->wqes);
1126 kfree(wq);
1127}
Jens Axboe848f7e12020-01-23 15:33:32 -07001128
1129void io_wq_destroy(struct io_wq *wq)
1130{
1131 if (refcount_dec_and_test(&wq->use_refs))
1132 __io_wq_destroy(wq);
1133}
Jens Axboeaa96bf82020-04-03 11:26:26 -06001134
1135struct task_struct *io_wq_get_task(struct io_wq *wq)
1136{
1137 return wq->manager;
1138}