blob: 74b40506c5d916324404af2087395a72c2e73fee [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>
19
20#include "io-wq.h"
21
22#define WORKER_IDLE_TIMEOUT (5 * HZ)
23
24enum {
25 IO_WORKER_F_UP = 1, /* up and active */
26 IO_WORKER_F_RUNNING = 2, /* account as running */
27 IO_WORKER_F_FREE = 4, /* worker on free list */
28 IO_WORKER_F_EXITING = 8, /* worker exiting */
29 IO_WORKER_F_FIXED = 16, /* static idle worker */
Jens Axboec5def4a2019-11-07 11:41:16 -070030 IO_WORKER_F_BOUND = 32, /* is doing bounded work */
Jens Axboe771b53d02019-10-22 10:25:58 -060031};
32
33enum {
34 IO_WQ_BIT_EXIT = 0, /* wq exiting */
35 IO_WQ_BIT_CANCEL = 1, /* cancel work on list */
Jens Axboeb60fda62019-11-19 08:37:07 -070036 IO_WQ_BIT_ERROR = 2, /* error on setup */
Jens Axboe771b53d02019-10-22 10:25:58 -060037};
38
39enum {
40 IO_WQE_FLAG_STALLED = 1, /* stalled on hash */
41};
42
43/*
44 * One for each thread in a wqe pool
45 */
46struct io_worker {
47 refcount_t ref;
48 unsigned flags;
49 struct hlist_nulls_node nulls_node;
Jens Axboee61df662019-11-13 13:54:49 -070050 struct list_head all_list;
Jens Axboe771b53d02019-10-22 10:25:58 -060051 struct task_struct *task;
52 wait_queue_head_t wait;
53 struct io_wqe *wqe;
Jens Axboe36c2f922019-11-13 09:43:34 -070054
Jens Axboe771b53d02019-10-22 10:25:58 -060055 struct io_wq_work *cur_work;
Jens Axboe36c2f922019-11-13 09:43:34 -070056 spinlock_t lock;
Jens Axboe771b53d02019-10-22 10:25:58 -060057
58 struct rcu_head rcu;
59 struct mm_struct *mm;
Jens Axboe181e4482019-11-25 08:52:30 -070060 const struct cred *creds;
Jens Axboefcb323c2019-10-24 12:39:47 -060061 struct files_struct *restore_files;
Jens Axboe771b53d02019-10-22 10:25:58 -060062};
63
Jens Axboe771b53d02019-10-22 10:25:58 -060064#if BITS_PER_LONG == 64
65#define IO_WQ_HASH_ORDER 6
66#else
67#define IO_WQ_HASH_ORDER 5
68#endif
69
Jens Axboec5def4a2019-11-07 11:41:16 -070070struct io_wqe_acct {
71 unsigned nr_workers;
72 unsigned max_workers;
73 atomic_t nr_running;
74};
75
76enum {
77 IO_WQ_ACCT_BOUND,
78 IO_WQ_ACCT_UNBOUND,
79};
80
Jens Axboe771b53d02019-10-22 10:25:58 -060081/*
82 * Per-node worker thread pool
83 */
84struct io_wqe {
85 struct {
86 spinlock_t lock;
Jens Axboe6206f0e2019-11-26 11:59:32 -070087 struct io_wq_work_list work_list;
Jens Axboe771b53d02019-10-22 10:25:58 -060088 unsigned long hash_map;
89 unsigned flags;
90 } ____cacheline_aligned_in_smp;
91
92 int node;
Jens Axboec5def4a2019-11-07 11:41:16 -070093 struct io_wqe_acct acct[2];
Jens Axboe771b53d02019-10-22 10:25:58 -060094
Jens Axboe021d1cd2019-11-14 08:00:41 -070095 struct hlist_nulls_head free_list;
96 struct hlist_nulls_head busy_list;
Jens Axboee61df662019-11-13 13:54:49 -070097 struct list_head all_list;
Jens Axboe771b53d02019-10-22 10:25:58 -060098
99 struct io_wq *wq;
100};
101
102/*
103 * Per io_wq state
104 */
105struct io_wq {
106 struct io_wqe **wqes;
107 unsigned long state;
Jens Axboe771b53d02019-10-22 10:25:58 -0600108
Jens Axboe7d723062019-11-12 22:31:31 -0700109 get_work_fn *get_work;
110 put_work_fn *put_work;
111
Jens Axboe771b53d02019-10-22 10:25:58 -0600112 struct task_struct *manager;
Jens Axboec5def4a2019-11-07 11:41:16 -0700113 struct user_struct *user;
Jens Axboe0b8c0ec2019-12-02 08:50:00 -0700114 const struct cred *creds;
Jens Axboe771b53d02019-10-22 10:25:58 -0600115 struct mm_struct *mm;
116 refcount_t refs;
117 struct completion done;
118};
119
Jens Axboe771b53d02019-10-22 10:25:58 -0600120static bool io_worker_get(struct io_worker *worker)
121{
122 return refcount_inc_not_zero(&worker->ref);
123}
124
125static void io_worker_release(struct io_worker *worker)
126{
127 if (refcount_dec_and_test(&worker->ref))
128 wake_up_process(worker->task);
129}
130
131/*
132 * Note: drops the wqe->lock if returning true! The caller must re-acquire
133 * the lock in that case. Some callers need to restart handling if this
134 * happens, so we can't just re-acquire the lock on behalf of the caller.
135 */
136static bool __io_worker_unuse(struct io_wqe *wqe, struct io_worker *worker)
137{
Jens Axboefcb323c2019-10-24 12:39:47 -0600138 bool dropped_lock = false;
139
Jens Axboe181e4482019-11-25 08:52:30 -0700140 if (worker->creds) {
141 revert_creds(worker->creds);
142 worker->creds = NULL;
143 }
144
Jens Axboefcb323c2019-10-24 12:39:47 -0600145 if (current->files != worker->restore_files) {
146 __acquire(&wqe->lock);
147 spin_unlock_irq(&wqe->lock);
148 dropped_lock = true;
149
150 task_lock(current);
151 current->files = worker->restore_files;
152 task_unlock(current);
153 }
154
Jens Axboe771b53d02019-10-22 10:25:58 -0600155 /*
156 * If we have an active mm, we need to drop the wq lock before unusing
157 * it. If we do, return true and let the caller retry the idle loop.
158 */
159 if (worker->mm) {
Jens Axboefcb323c2019-10-24 12:39:47 -0600160 if (!dropped_lock) {
161 __acquire(&wqe->lock);
162 spin_unlock_irq(&wqe->lock);
163 dropped_lock = true;
164 }
Jens Axboe771b53d02019-10-22 10:25:58 -0600165 __set_current_state(TASK_RUNNING);
166 set_fs(KERNEL_DS);
167 unuse_mm(worker->mm);
168 mmput(worker->mm);
169 worker->mm = NULL;
Jens Axboe771b53d02019-10-22 10:25:58 -0600170 }
171
Jens Axboefcb323c2019-10-24 12:39:47 -0600172 return dropped_lock;
Jens Axboe771b53d02019-10-22 10:25:58 -0600173}
174
Jens Axboec5def4a2019-11-07 11:41:16 -0700175static inline struct io_wqe_acct *io_work_get_acct(struct io_wqe *wqe,
176 struct io_wq_work *work)
177{
178 if (work->flags & IO_WQ_WORK_UNBOUND)
179 return &wqe->acct[IO_WQ_ACCT_UNBOUND];
180
181 return &wqe->acct[IO_WQ_ACCT_BOUND];
182}
183
184static inline struct io_wqe_acct *io_wqe_get_acct(struct io_wqe *wqe,
185 struct io_worker *worker)
186{
187 if (worker->flags & IO_WORKER_F_BOUND)
188 return &wqe->acct[IO_WQ_ACCT_BOUND];
189
190 return &wqe->acct[IO_WQ_ACCT_UNBOUND];
191}
192
Jens Axboe771b53d02019-10-22 10:25:58 -0600193static void io_worker_exit(struct io_worker *worker)
194{
195 struct io_wqe *wqe = worker->wqe;
Jens Axboec5def4a2019-11-07 11:41:16 -0700196 struct io_wqe_acct *acct = io_wqe_get_acct(wqe, worker);
197 unsigned nr_workers;
Jens Axboe771b53d02019-10-22 10:25:58 -0600198
199 /*
200 * If we're not at zero, someone else is holding a brief reference
201 * to the worker. Wait for that to go away.
202 */
203 set_current_state(TASK_INTERRUPTIBLE);
204 if (!refcount_dec_and_test(&worker->ref))
205 schedule();
206 __set_current_state(TASK_RUNNING);
207
208 preempt_disable();
209 current->flags &= ~PF_IO_WORKER;
210 if (worker->flags & IO_WORKER_F_RUNNING)
Jens Axboec5def4a2019-11-07 11:41:16 -0700211 atomic_dec(&acct->nr_running);
212 if (!(worker->flags & IO_WORKER_F_BOUND))
213 atomic_dec(&wqe->wq->user->processes);
Jens Axboe771b53d02019-10-22 10:25:58 -0600214 worker->flags = 0;
215 preempt_enable();
216
217 spin_lock_irq(&wqe->lock);
218 hlist_nulls_del_rcu(&worker->nulls_node);
Jens Axboee61df662019-11-13 13:54:49 -0700219 list_del_rcu(&worker->all_list);
Jens Axboe771b53d02019-10-22 10:25:58 -0600220 if (__io_worker_unuse(wqe, worker)) {
221 __release(&wqe->lock);
222 spin_lock_irq(&wqe->lock);
223 }
Jens Axboec5def4a2019-11-07 11:41:16 -0700224 acct->nr_workers--;
225 nr_workers = wqe->acct[IO_WQ_ACCT_BOUND].nr_workers +
226 wqe->acct[IO_WQ_ACCT_UNBOUND].nr_workers;
Jens Axboe771b53d02019-10-22 10:25:58 -0600227 spin_unlock_irq(&wqe->lock);
228
229 /* all workers gone, wq exit can proceed */
Jens Axboec5def4a2019-11-07 11:41:16 -0700230 if (!nr_workers && refcount_dec_and_test(&wqe->wq->refs))
Jens Axboe771b53d02019-10-22 10:25:58 -0600231 complete(&wqe->wq->done);
232
YueHaibing364b05f2019-11-02 15:55:01 +0800233 kfree_rcu(worker, rcu);
Jens Axboe771b53d02019-10-22 10:25:58 -0600234}
235
Jens Axboec5def4a2019-11-07 11:41:16 -0700236static inline bool io_wqe_run_queue(struct io_wqe *wqe)
237 __must_hold(wqe->lock)
238{
Jens Axboe6206f0e2019-11-26 11:59:32 -0700239 if (!wq_list_empty(&wqe->work_list) &&
240 !(wqe->flags & IO_WQE_FLAG_STALLED))
Jens Axboec5def4a2019-11-07 11:41:16 -0700241 return true;
242 return false;
243}
244
245/*
246 * Check head of free list for an available worker. If one isn't available,
247 * caller must wake up the wq manager to create one.
248 */
249static bool io_wqe_activate_free_worker(struct io_wqe *wqe)
250 __must_hold(RCU)
251{
252 struct hlist_nulls_node *n;
253 struct io_worker *worker;
254
Jens Axboe021d1cd2019-11-14 08:00:41 -0700255 n = rcu_dereference(hlist_nulls_first_rcu(&wqe->free_list));
Jens Axboec5def4a2019-11-07 11:41:16 -0700256 if (is_a_nulls(n))
257 return false;
258
259 worker = hlist_nulls_entry(n, struct io_worker, nulls_node);
260 if (io_worker_get(worker)) {
261 wake_up(&worker->wait);
262 io_worker_release(worker);
263 return true;
264 }
265
266 return false;
267}
268
269/*
270 * We need a worker. If we find a free one, we're good. If not, and we're
271 * below the max number of workers, wake up the manager to create one.
272 */
273static void io_wqe_wake_worker(struct io_wqe *wqe, struct io_wqe_acct *acct)
274{
275 bool ret;
276
277 /*
278 * Most likely an attempt to queue unbounded work on an io_wq that
279 * wasn't setup with any unbounded workers.
280 */
281 WARN_ON_ONCE(!acct->max_workers);
282
283 rcu_read_lock();
284 ret = io_wqe_activate_free_worker(wqe);
285 rcu_read_unlock();
286
287 if (!ret && acct->nr_workers < acct->max_workers)
288 wake_up_process(wqe->wq->manager);
289}
290
291static void io_wqe_inc_running(struct io_wqe *wqe, struct io_worker *worker)
292{
293 struct io_wqe_acct *acct = io_wqe_get_acct(wqe, worker);
294
295 atomic_inc(&acct->nr_running);
296}
297
298static void io_wqe_dec_running(struct io_wqe *wqe, struct io_worker *worker)
299 __must_hold(wqe->lock)
300{
301 struct io_wqe_acct *acct = io_wqe_get_acct(wqe, worker);
302
303 if (atomic_dec_and_test(&acct->nr_running) && io_wqe_run_queue(wqe))
304 io_wqe_wake_worker(wqe, acct);
305}
306
Jens Axboe771b53d02019-10-22 10:25:58 -0600307static void io_worker_start(struct io_wqe *wqe, struct io_worker *worker)
308{
309 allow_kernel_signal(SIGINT);
310
311 current->flags |= PF_IO_WORKER;
312
313 worker->flags |= (IO_WORKER_F_UP | IO_WORKER_F_RUNNING);
Jens Axboefcb323c2019-10-24 12:39:47 -0600314 worker->restore_files = current->files;
Jens Axboec5def4a2019-11-07 11:41:16 -0700315 io_wqe_inc_running(wqe, worker);
Jens Axboe771b53d02019-10-22 10:25:58 -0600316}
317
318/*
319 * Worker will start processing some work. Move it to the busy list, if
320 * it's currently on the freelist
321 */
322static void __io_worker_busy(struct io_wqe *wqe, struct io_worker *worker,
323 struct io_wq_work *work)
324 __must_hold(wqe->lock)
325{
Jens Axboec5def4a2019-11-07 11:41:16 -0700326 bool worker_bound, work_bound;
327
Jens Axboe771b53d02019-10-22 10:25:58 -0600328 if (worker->flags & IO_WORKER_F_FREE) {
329 worker->flags &= ~IO_WORKER_F_FREE;
330 hlist_nulls_del_init_rcu(&worker->nulls_node);
Jens Axboe021d1cd2019-11-14 08:00:41 -0700331 hlist_nulls_add_head_rcu(&worker->nulls_node, &wqe->busy_list);
Jens Axboe771b53d02019-10-22 10:25:58 -0600332 }
Jens Axboec5def4a2019-11-07 11:41:16 -0700333
334 /*
335 * If worker is moving from bound to unbound (or vice versa), then
336 * ensure we update the running accounting.
337 */
Dan Carpenterb2e9c7d62019-11-19 09:22:16 +0300338 worker_bound = (worker->flags & IO_WORKER_F_BOUND) != 0;
339 work_bound = (work->flags & IO_WQ_WORK_UNBOUND) == 0;
340 if (worker_bound != work_bound) {
Jens Axboec5def4a2019-11-07 11:41:16 -0700341 io_wqe_dec_running(wqe, worker);
342 if (work_bound) {
343 worker->flags |= IO_WORKER_F_BOUND;
344 wqe->acct[IO_WQ_ACCT_UNBOUND].nr_workers--;
345 wqe->acct[IO_WQ_ACCT_BOUND].nr_workers++;
346 atomic_dec(&wqe->wq->user->processes);
347 } else {
348 worker->flags &= ~IO_WORKER_F_BOUND;
349 wqe->acct[IO_WQ_ACCT_UNBOUND].nr_workers++;
350 wqe->acct[IO_WQ_ACCT_BOUND].nr_workers--;
351 atomic_inc(&wqe->wq->user->processes);
352 }
353 io_wqe_inc_running(wqe, worker);
354 }
Jens Axboe771b53d02019-10-22 10:25:58 -0600355}
356
357/*
358 * No work, worker going to sleep. Move to freelist, and unuse mm if we
359 * have one attached. Dropping the mm may potentially sleep, so we drop
360 * the lock in that case and return success. Since the caller has to
361 * retry the loop in that case (we changed task state), we don't regrab
362 * the lock if we return success.
363 */
364static bool __io_worker_idle(struct io_wqe *wqe, struct io_worker *worker)
365 __must_hold(wqe->lock)
366{
367 if (!(worker->flags & IO_WORKER_F_FREE)) {
368 worker->flags |= IO_WORKER_F_FREE;
369 hlist_nulls_del_init_rcu(&worker->nulls_node);
Jens Axboe021d1cd2019-11-14 08:00:41 -0700370 hlist_nulls_add_head_rcu(&worker->nulls_node, &wqe->free_list);
Jens Axboe771b53d02019-10-22 10:25:58 -0600371 }
372
373 return __io_worker_unuse(wqe, worker);
374}
375
376static struct io_wq_work *io_get_next_work(struct io_wqe *wqe, unsigned *hash)
377 __must_hold(wqe->lock)
378{
Jens Axboe6206f0e2019-11-26 11:59:32 -0700379 struct io_wq_work_node *node, *prev;
Jens Axboe771b53d02019-10-22 10:25:58 -0600380 struct io_wq_work *work;
381
Jens Axboe6206f0e2019-11-26 11:59:32 -0700382 wq_list_for_each(node, prev, &wqe->work_list) {
383 work = container_of(node, struct io_wq_work, list);
384
Jens Axboe771b53d02019-10-22 10:25:58 -0600385 /* not hashed, can run anytime */
386 if (!(work->flags & IO_WQ_WORK_HASHED)) {
Jens Axboe6206f0e2019-11-26 11:59:32 -0700387 wq_node_del(&wqe->work_list, node, prev);
Jens Axboe771b53d02019-10-22 10:25:58 -0600388 return work;
389 }
390
391 /* hashed, can run if not already running */
392 *hash = work->flags >> IO_WQ_HASH_SHIFT;
393 if (!(wqe->hash_map & BIT_ULL(*hash))) {
394 wqe->hash_map |= BIT_ULL(*hash);
Jens Axboe6206f0e2019-11-26 11:59:32 -0700395 wq_node_del(&wqe->work_list, node, prev);
Jens Axboe771b53d02019-10-22 10:25:58 -0600396 return work;
397 }
398 }
399
400 return NULL;
401}
402
403static void io_worker_handle_work(struct io_worker *worker)
404 __releases(wqe->lock)
405{
Jens Axboe7d723062019-11-12 22:31:31 -0700406 struct io_wq_work *work, *old_work = NULL, *put_work = NULL;
Jens Axboe771b53d02019-10-22 10:25:58 -0600407 struct io_wqe *wqe = worker->wqe;
408 struct io_wq *wq = wqe->wq;
409
410 do {
411 unsigned hash = -1U;
412
413 /*
Jens Axboe771b53d02019-10-22 10:25:58 -0600414 * If we got some work, mark us as busy. If we didn't, but
415 * the list isn't empty, it means we stalled on hashed work.
416 * Mark us stalled so we don't keep looking for work when we
417 * can't make progress, any work completion or insertion will
418 * clear the stalled flag.
419 */
420 work = io_get_next_work(wqe, &hash);
421 if (work)
422 __io_worker_busy(wqe, worker, work);
Jens Axboe6206f0e2019-11-26 11:59:32 -0700423 else if (!wq_list_empty(&wqe->work_list))
Jens Axboe771b53d02019-10-22 10:25:58 -0600424 wqe->flags |= IO_WQE_FLAG_STALLED;
425
426 spin_unlock_irq(&wqe->lock);
Jens Axboe7d723062019-11-12 22:31:31 -0700427 if (put_work && wq->put_work)
428 wq->put_work(old_work);
Jens Axboe771b53d02019-10-22 10:25:58 -0600429 if (!work)
430 break;
431next:
Jens Axboe36c2f922019-11-13 09:43:34 -0700432 /* flush any pending signals before assigning new work */
433 if (signal_pending(current))
434 flush_signals(current);
435
436 spin_lock_irq(&worker->lock);
437 worker->cur_work = work;
438 spin_unlock_irq(&worker->lock);
439
Jens Axboeb76da702019-11-20 13:05:32 -0700440 if (work->flags & IO_WQ_WORK_CB)
441 work->func(&work);
442
Jens Axboefcb323c2019-10-24 12:39:47 -0600443 if ((work->flags & IO_WQ_WORK_NEEDS_FILES) &&
444 current->files != work->files) {
445 task_lock(current);
446 current->files = work->files;
447 task_unlock(current);
448 }
Jens Axboe771b53d02019-10-22 10:25:58 -0600449 if ((work->flags & IO_WQ_WORK_NEEDS_USER) && !worker->mm &&
450 wq->mm && mmget_not_zero(wq->mm)) {
451 use_mm(wq->mm);
452 set_fs(USER_DS);
453 worker->mm = wq->mm;
454 }
Jens Axboe181e4482019-11-25 08:52:30 -0700455 if (!worker->creds)
456 worker->creds = override_creds(wq->creds);
Jens Axboe771b53d02019-10-22 10:25:58 -0600457 if (test_bit(IO_WQ_BIT_CANCEL, &wq->state))
458 work->flags |= IO_WQ_WORK_CANCEL;
459 if (worker->mm)
460 work->flags |= IO_WQ_WORK_HAS_MM;
461
Jens Axboe7d723062019-11-12 22:31:31 -0700462 if (wq->get_work && !(work->flags & IO_WQ_WORK_INTERNAL)) {
463 put_work = work;
464 wq->get_work(work);
465 }
466
Jens Axboe771b53d02019-10-22 10:25:58 -0600467 old_work = work;
468 work->func(&work);
469
Jens Axboe36c2f922019-11-13 09:43:34 -0700470 spin_lock_irq(&worker->lock);
Jens Axboe771b53d02019-10-22 10:25:58 -0600471 worker->cur_work = NULL;
Jens Axboe36c2f922019-11-13 09:43:34 -0700472 spin_unlock_irq(&worker->lock);
473
474 spin_lock_irq(&wqe->lock);
475
Jens Axboe771b53d02019-10-22 10:25:58 -0600476 if (hash != -1U) {
477 wqe->hash_map &= ~BIT_ULL(hash);
478 wqe->flags &= ~IO_WQE_FLAG_STALLED;
479 }
480 if (work && work != old_work) {
481 spin_unlock_irq(&wqe->lock);
Jens Axboe7d723062019-11-12 22:31:31 -0700482
483 if (put_work && wq->put_work) {
484 wq->put_work(put_work);
485 put_work = NULL;
486 }
487
Jens Axboe771b53d02019-10-22 10:25:58 -0600488 /* dependent work not hashed */
489 hash = -1U;
490 goto next;
491 }
492 } while (1);
493}
494
Jens Axboe771b53d02019-10-22 10:25:58 -0600495static int io_wqe_worker(void *data)
496{
497 struct io_worker *worker = data;
498 struct io_wqe *wqe = worker->wqe;
499 struct io_wq *wq = wqe->wq;
500 DEFINE_WAIT(wait);
501
502 io_worker_start(wqe, worker);
503
504 while (!test_bit(IO_WQ_BIT_EXIT, &wq->state)) {
505 prepare_to_wait(&worker->wait, &wait, TASK_INTERRUPTIBLE);
506
507 spin_lock_irq(&wqe->lock);
508 if (io_wqe_run_queue(wqe)) {
509 __set_current_state(TASK_RUNNING);
510 io_worker_handle_work(worker);
511 continue;
512 }
513 /* drops the lock on success, retry */
514 if (__io_worker_idle(wqe, worker)) {
515 __release(&wqe->lock);
516 continue;
517 }
518 spin_unlock_irq(&wqe->lock);
519 if (signal_pending(current))
520 flush_signals(current);
521 if (schedule_timeout(WORKER_IDLE_TIMEOUT))
522 continue;
523 /* timed out, exit unless we're the fixed worker */
524 if (test_bit(IO_WQ_BIT_EXIT, &wq->state) ||
525 !(worker->flags & IO_WORKER_F_FIXED))
526 break;
527 }
528
529 finish_wait(&worker->wait, &wait);
530
531 if (test_bit(IO_WQ_BIT_EXIT, &wq->state)) {
532 spin_lock_irq(&wqe->lock);
Jens Axboe6206f0e2019-11-26 11:59:32 -0700533 if (!wq_list_empty(&wqe->work_list))
Jens Axboe771b53d02019-10-22 10:25:58 -0600534 io_worker_handle_work(worker);
535 else
536 spin_unlock_irq(&wqe->lock);
537 }
538
539 io_worker_exit(worker);
540 return 0;
541}
542
543/*
Jens Axboe771b53d02019-10-22 10:25:58 -0600544 * Called when a worker is scheduled in. Mark us as currently running.
545 */
546void io_wq_worker_running(struct task_struct *tsk)
547{
548 struct io_worker *worker = kthread_data(tsk);
549 struct io_wqe *wqe = worker->wqe;
550
551 if (!(worker->flags & IO_WORKER_F_UP))
552 return;
553 if (worker->flags & IO_WORKER_F_RUNNING)
554 return;
555 worker->flags |= IO_WORKER_F_RUNNING;
Jens Axboec5def4a2019-11-07 11:41:16 -0700556 io_wqe_inc_running(wqe, worker);
Jens Axboe771b53d02019-10-22 10:25:58 -0600557}
558
559/*
560 * Called when worker is going to sleep. If there are no workers currently
561 * running and we have work pending, wake up a free one or have the manager
562 * set one up.
563 */
564void io_wq_worker_sleeping(struct task_struct *tsk)
565{
566 struct io_worker *worker = kthread_data(tsk);
567 struct io_wqe *wqe = worker->wqe;
568
569 if (!(worker->flags & IO_WORKER_F_UP))
570 return;
571 if (!(worker->flags & IO_WORKER_F_RUNNING))
572 return;
573
574 worker->flags &= ~IO_WORKER_F_RUNNING;
575
576 spin_lock_irq(&wqe->lock);
Jens Axboec5def4a2019-11-07 11:41:16 -0700577 io_wqe_dec_running(wqe, worker);
Jens Axboe771b53d02019-10-22 10:25:58 -0600578 spin_unlock_irq(&wqe->lock);
579}
580
Jens Axboeb60fda62019-11-19 08:37:07 -0700581static bool create_io_worker(struct io_wq *wq, struct io_wqe *wqe, int index)
Jens Axboe771b53d02019-10-22 10:25:58 -0600582{
Jens Axboec5def4a2019-11-07 11:41:16 -0700583 struct io_wqe_acct *acct =&wqe->acct[index];
Jens Axboe771b53d02019-10-22 10:25:58 -0600584 struct io_worker *worker;
585
Jann Hornad6e0052019-11-26 17:39:45 +0100586 worker = kzalloc_node(sizeof(*worker), GFP_KERNEL, wqe->node);
Jens Axboe771b53d02019-10-22 10:25:58 -0600587 if (!worker)
Jens Axboeb60fda62019-11-19 08:37:07 -0700588 return false;
Jens Axboe771b53d02019-10-22 10:25:58 -0600589
590 refcount_set(&worker->ref, 1);
591 worker->nulls_node.pprev = NULL;
592 init_waitqueue_head(&worker->wait);
593 worker->wqe = wqe;
Jens Axboe36c2f922019-11-13 09:43:34 -0700594 spin_lock_init(&worker->lock);
Jens Axboe771b53d02019-10-22 10:25:58 -0600595
596 worker->task = kthread_create_on_node(io_wqe_worker, worker, wqe->node,
Jens Axboec5def4a2019-11-07 11:41:16 -0700597 "io_wqe_worker-%d/%d", index, wqe->node);
Jens Axboe771b53d02019-10-22 10:25:58 -0600598 if (IS_ERR(worker->task)) {
599 kfree(worker);
Jens Axboeb60fda62019-11-19 08:37:07 -0700600 return false;
Jens Axboe771b53d02019-10-22 10:25:58 -0600601 }
602
603 spin_lock_irq(&wqe->lock);
Jens Axboe021d1cd2019-11-14 08:00:41 -0700604 hlist_nulls_add_head_rcu(&worker->nulls_node, &wqe->free_list);
Jens Axboee61df662019-11-13 13:54:49 -0700605 list_add_tail_rcu(&worker->all_list, &wqe->all_list);
Jens Axboe771b53d02019-10-22 10:25:58 -0600606 worker->flags |= IO_WORKER_F_FREE;
Jens Axboec5def4a2019-11-07 11:41:16 -0700607 if (index == IO_WQ_ACCT_BOUND)
608 worker->flags |= IO_WORKER_F_BOUND;
609 if (!acct->nr_workers && (worker->flags & IO_WORKER_F_BOUND))
Jens Axboe771b53d02019-10-22 10:25:58 -0600610 worker->flags |= IO_WORKER_F_FIXED;
Jens Axboec5def4a2019-11-07 11:41:16 -0700611 acct->nr_workers++;
Jens Axboe771b53d02019-10-22 10:25:58 -0600612 spin_unlock_irq(&wqe->lock);
613
Jens Axboec5def4a2019-11-07 11:41:16 -0700614 if (index == IO_WQ_ACCT_UNBOUND)
615 atomic_inc(&wq->user->processes);
616
Jens Axboe771b53d02019-10-22 10:25:58 -0600617 wake_up_process(worker->task);
Jens Axboeb60fda62019-11-19 08:37:07 -0700618 return true;
Jens Axboe771b53d02019-10-22 10:25:58 -0600619}
620
Jens Axboec5def4a2019-11-07 11:41:16 -0700621static inline bool io_wqe_need_worker(struct io_wqe *wqe, int index)
Jens Axboe771b53d02019-10-22 10:25:58 -0600622 __must_hold(wqe->lock)
623{
Jens Axboec5def4a2019-11-07 11:41:16 -0700624 struct io_wqe_acct *acct = &wqe->acct[index];
Jens Axboe771b53d02019-10-22 10:25:58 -0600625
Jens Axboec5def4a2019-11-07 11:41:16 -0700626 /* if we have available workers or no work, no need */
Jens Axboe021d1cd2019-11-14 08:00:41 -0700627 if (!hlist_nulls_empty(&wqe->free_list) || !io_wqe_run_queue(wqe))
Jens Axboec5def4a2019-11-07 11:41:16 -0700628 return false;
629 return acct->nr_workers < acct->max_workers;
Jens Axboe771b53d02019-10-22 10:25:58 -0600630}
631
632/*
633 * Manager thread. Tasked with creating new workers, if we need them.
634 */
635static int io_wq_manager(void *data)
636{
637 struct io_wq *wq = data;
Jann Horn3fc50ab2019-11-26 19:10:20 +0100638 int workers_to_create = num_possible_nodes();
639 int node;
Jens Axboeb60fda62019-11-19 08:37:07 -0700640
641 /* create fixed workers */
Jann Horn3fc50ab2019-11-26 19:10:20 +0100642 refcount_set(&wq->refs, workers_to_create);
643 for_each_node(node) {
644 if (!create_io_worker(wq, wq->wqes[node], IO_WQ_ACCT_BOUND))
645 goto err;
646 workers_to_create--;
Jens Axboeb60fda62019-11-19 08:37:07 -0700647 }
648
649 complete(&wq->done);
Jens Axboe771b53d02019-10-22 10:25:58 -0600650
651 while (!kthread_should_stop()) {
Jann Horn3fc50ab2019-11-26 19:10:20 +0100652 for_each_node(node) {
653 struct io_wqe *wqe = wq->wqes[node];
Jens Axboec5def4a2019-11-07 11:41:16 -0700654 bool fork_worker[2] = { false, false };
Jens Axboe771b53d02019-10-22 10:25:58 -0600655
656 spin_lock_irq(&wqe->lock);
Jens Axboec5def4a2019-11-07 11:41:16 -0700657 if (io_wqe_need_worker(wqe, IO_WQ_ACCT_BOUND))
658 fork_worker[IO_WQ_ACCT_BOUND] = true;
659 if (io_wqe_need_worker(wqe, IO_WQ_ACCT_UNBOUND))
660 fork_worker[IO_WQ_ACCT_UNBOUND] = true;
Jens Axboe771b53d02019-10-22 10:25:58 -0600661 spin_unlock_irq(&wqe->lock);
Jens Axboec5def4a2019-11-07 11:41:16 -0700662 if (fork_worker[IO_WQ_ACCT_BOUND])
663 create_io_worker(wq, wqe, IO_WQ_ACCT_BOUND);
664 if (fork_worker[IO_WQ_ACCT_UNBOUND])
665 create_io_worker(wq, wqe, IO_WQ_ACCT_UNBOUND);
Jens Axboe771b53d02019-10-22 10:25:58 -0600666 }
667 set_current_state(TASK_INTERRUPTIBLE);
668 schedule_timeout(HZ);
669 }
670
671 return 0;
Jens Axboeb60fda62019-11-19 08:37:07 -0700672err:
673 set_bit(IO_WQ_BIT_ERROR, &wq->state);
674 set_bit(IO_WQ_BIT_EXIT, &wq->state);
Jann Horn3fc50ab2019-11-26 19:10:20 +0100675 if (refcount_sub_and_test(workers_to_create, &wq->refs))
Jens Axboeb60fda62019-11-19 08:37:07 -0700676 complete(&wq->done);
677 return 0;
Jens Axboe771b53d02019-10-22 10:25:58 -0600678}
679
Jens Axboec5def4a2019-11-07 11:41:16 -0700680static bool io_wq_can_queue(struct io_wqe *wqe, struct io_wqe_acct *acct,
681 struct io_wq_work *work)
682{
683 bool free_worker;
684
685 if (!(work->flags & IO_WQ_WORK_UNBOUND))
686 return true;
687 if (atomic_read(&acct->nr_running))
688 return true;
689
690 rcu_read_lock();
Jens Axboe021d1cd2019-11-14 08:00:41 -0700691 free_worker = !hlist_nulls_empty(&wqe->free_list);
Jens Axboec5def4a2019-11-07 11:41:16 -0700692 rcu_read_unlock();
693 if (free_worker)
694 return true;
695
696 if (atomic_read(&wqe->wq->user->processes) >= acct->max_workers &&
697 !(capable(CAP_SYS_RESOURCE) || capable(CAP_SYS_ADMIN)))
698 return false;
699
700 return true;
701}
702
Jens Axboe771b53d02019-10-22 10:25:58 -0600703static void io_wqe_enqueue(struct io_wqe *wqe, struct io_wq_work *work)
704{
Jens Axboec5def4a2019-11-07 11:41:16 -0700705 struct io_wqe_acct *acct = io_work_get_acct(wqe, work);
Jens Axboe771b53d02019-10-22 10:25:58 -0600706 unsigned long flags;
707
Jens Axboec5def4a2019-11-07 11:41:16 -0700708 /*
709 * Do early check to see if we need a new unbound worker, and if we do,
710 * if we're allowed to do so. This isn't 100% accurate as there's a
711 * gap between this check and incrementing the value, but that's OK.
712 * It's close enough to not be an issue, fork() has the same delay.
713 */
714 if (unlikely(!io_wq_can_queue(wqe, acct, work))) {
715 work->flags |= IO_WQ_WORK_CANCEL;
716 work->func(&work);
717 return;
718 }
719
Jens Axboe771b53d02019-10-22 10:25:58 -0600720 spin_lock_irqsave(&wqe->lock, flags);
Jens Axboe6206f0e2019-11-26 11:59:32 -0700721 wq_list_add_tail(&work->list, &wqe->work_list);
Jens Axboe771b53d02019-10-22 10:25:58 -0600722 wqe->flags &= ~IO_WQE_FLAG_STALLED;
723 spin_unlock_irqrestore(&wqe->lock, flags);
724
Jens Axboec5def4a2019-11-07 11:41:16 -0700725 if (!atomic_read(&acct->nr_running))
726 io_wqe_wake_worker(wqe, acct);
Jens Axboe771b53d02019-10-22 10:25:58 -0600727}
728
729void io_wq_enqueue(struct io_wq *wq, struct io_wq_work *work)
730{
731 struct io_wqe *wqe = wq->wqes[numa_node_id()];
732
733 io_wqe_enqueue(wqe, work);
734}
735
736/*
737 * Enqueue work, hashed by some key. Work items that hash to the same value
738 * will not be done in parallel. Used to limit concurrent writes, generally
739 * hashed by inode.
740 */
741void io_wq_enqueue_hashed(struct io_wq *wq, struct io_wq_work *work, void *val)
742{
743 struct io_wqe *wqe = wq->wqes[numa_node_id()];
744 unsigned bit;
745
746
747 bit = hash_ptr(val, IO_WQ_HASH_ORDER);
748 work->flags |= (IO_WQ_WORK_HASHED | (bit << IO_WQ_HASH_SHIFT));
749 io_wqe_enqueue(wqe, work);
750}
751
752static bool io_wqe_worker_send_sig(struct io_worker *worker, void *data)
753{
754 send_sig(SIGINT, worker->task, 1);
755 return false;
756}
757
758/*
759 * Iterate the passed in list and call the specific function for each
760 * worker that isn't exiting
761 */
762static bool io_wq_for_each_worker(struct io_wqe *wqe,
Jens Axboe771b53d02019-10-22 10:25:58 -0600763 bool (*func)(struct io_worker *, void *),
764 void *data)
765{
Jens Axboe771b53d02019-10-22 10:25:58 -0600766 struct io_worker *worker;
767 bool ret = false;
768
Jens Axboee61df662019-11-13 13:54:49 -0700769 list_for_each_entry_rcu(worker, &wqe->all_list, all_list) {
Jens Axboe771b53d02019-10-22 10:25:58 -0600770 if (io_worker_get(worker)) {
771 ret = func(worker, data);
772 io_worker_release(worker);
773 if (ret)
774 break;
775 }
776 }
Jens Axboee61df662019-11-13 13:54:49 -0700777
Jens Axboe771b53d02019-10-22 10:25:58 -0600778 return ret;
779}
780
781void io_wq_cancel_all(struct io_wq *wq)
782{
Jann Horn3fc50ab2019-11-26 19:10:20 +0100783 int node;
Jens Axboe771b53d02019-10-22 10:25:58 -0600784
785 set_bit(IO_WQ_BIT_CANCEL, &wq->state);
786
787 /*
788 * Browse both lists, as there's a gap between handing work off
789 * to a worker and the worker putting itself on the busy_list
790 */
791 rcu_read_lock();
Jann Horn3fc50ab2019-11-26 19:10:20 +0100792 for_each_node(node) {
793 struct io_wqe *wqe = wq->wqes[node];
Jens Axboe771b53d02019-10-22 10:25:58 -0600794
Jens Axboee61df662019-11-13 13:54:49 -0700795 io_wq_for_each_worker(wqe, io_wqe_worker_send_sig, NULL);
Jens Axboe771b53d02019-10-22 10:25:58 -0600796 }
797 rcu_read_unlock();
798}
799
Jens Axboe62755e32019-10-28 21:49:21 -0600800struct io_cb_cancel_data {
801 struct io_wqe *wqe;
802 work_cancel_fn *cancel;
803 void *caller_data;
804};
805
806static bool io_work_cancel(struct io_worker *worker, void *cancel_data)
807{
808 struct io_cb_cancel_data *data = cancel_data;
Jens Axboe6f726532019-11-05 13:51:51 -0700809 unsigned long flags;
Jens Axboe62755e32019-10-28 21:49:21 -0600810 bool ret = false;
811
812 /*
813 * Hold the lock to avoid ->cur_work going out of scope, caller
Jens Axboe36c2f922019-11-13 09:43:34 -0700814 * may dereference the passed in work.
Jens Axboe62755e32019-10-28 21:49:21 -0600815 */
Jens Axboe36c2f922019-11-13 09:43:34 -0700816 spin_lock_irqsave(&worker->lock, flags);
Jens Axboe62755e32019-10-28 21:49:21 -0600817 if (worker->cur_work &&
818 data->cancel(worker->cur_work, data->caller_data)) {
819 send_sig(SIGINT, worker->task, 1);
820 ret = true;
821 }
Jens Axboe36c2f922019-11-13 09:43:34 -0700822 spin_unlock_irqrestore(&worker->lock, flags);
Jens Axboe62755e32019-10-28 21:49:21 -0600823
824 return ret;
825}
826
827static enum io_wq_cancel io_wqe_cancel_cb_work(struct io_wqe *wqe,
828 work_cancel_fn *cancel,
829 void *cancel_data)
830{
831 struct io_cb_cancel_data data = {
832 .wqe = wqe,
833 .cancel = cancel,
834 .caller_data = cancel_data,
835 };
Jens Axboe6206f0e2019-11-26 11:59:32 -0700836 struct io_wq_work_node *node, *prev;
Jens Axboe62755e32019-10-28 21:49:21 -0600837 struct io_wq_work *work;
Jens Axboe6f726532019-11-05 13:51:51 -0700838 unsigned long flags;
Jens Axboe62755e32019-10-28 21:49:21 -0600839 bool found = false;
840
Jens Axboe6f726532019-11-05 13:51:51 -0700841 spin_lock_irqsave(&wqe->lock, flags);
Jens Axboe6206f0e2019-11-26 11:59:32 -0700842 wq_list_for_each(node, prev, &wqe->work_list) {
843 work = container_of(node, struct io_wq_work, list);
844
Jens Axboe62755e32019-10-28 21:49:21 -0600845 if (cancel(work, cancel_data)) {
Jens Axboe6206f0e2019-11-26 11:59:32 -0700846 wq_node_del(&wqe->work_list, node, prev);
Jens Axboe62755e32019-10-28 21:49:21 -0600847 found = true;
848 break;
849 }
850 }
Jens Axboe6f726532019-11-05 13:51:51 -0700851 spin_unlock_irqrestore(&wqe->lock, flags);
Jens Axboe62755e32019-10-28 21:49:21 -0600852
853 if (found) {
854 work->flags |= IO_WQ_WORK_CANCEL;
855 work->func(&work);
856 return IO_WQ_CANCEL_OK;
857 }
858
859 rcu_read_lock();
Jens Axboee61df662019-11-13 13:54:49 -0700860 found = io_wq_for_each_worker(wqe, io_work_cancel, &data);
Jens Axboe62755e32019-10-28 21:49:21 -0600861 rcu_read_unlock();
862 return found ? IO_WQ_CANCEL_RUNNING : IO_WQ_CANCEL_NOTFOUND;
863}
864
865enum io_wq_cancel io_wq_cancel_cb(struct io_wq *wq, work_cancel_fn *cancel,
866 void *data)
867{
868 enum io_wq_cancel ret = IO_WQ_CANCEL_NOTFOUND;
Jann Horn3fc50ab2019-11-26 19:10:20 +0100869 int node;
Jens Axboe62755e32019-10-28 21:49:21 -0600870
Jann Horn3fc50ab2019-11-26 19:10:20 +0100871 for_each_node(node) {
872 struct io_wqe *wqe = wq->wqes[node];
Jens Axboe62755e32019-10-28 21:49:21 -0600873
874 ret = io_wqe_cancel_cb_work(wqe, cancel, data);
875 if (ret != IO_WQ_CANCEL_NOTFOUND)
876 break;
877 }
878
879 return ret;
880}
881
Jens Axboe771b53d02019-10-22 10:25:58 -0600882static bool io_wq_worker_cancel(struct io_worker *worker, void *data)
883{
884 struct io_wq_work *work = data;
Jens Axboe36c2f922019-11-13 09:43:34 -0700885 unsigned long flags;
886 bool ret = false;
Jens Axboe771b53d02019-10-22 10:25:58 -0600887
Jens Axboe36c2f922019-11-13 09:43:34 -0700888 if (worker->cur_work != work)
889 return false;
890
891 spin_lock_irqsave(&worker->lock, flags);
Jens Axboe771b53d02019-10-22 10:25:58 -0600892 if (worker->cur_work == work) {
893 send_sig(SIGINT, worker->task, 1);
Jens Axboe36c2f922019-11-13 09:43:34 -0700894 ret = true;
Jens Axboe771b53d02019-10-22 10:25:58 -0600895 }
Jens Axboe36c2f922019-11-13 09:43:34 -0700896 spin_unlock_irqrestore(&worker->lock, flags);
Jens Axboe771b53d02019-10-22 10:25:58 -0600897
Jens Axboe36c2f922019-11-13 09:43:34 -0700898 return ret;
Jens Axboe771b53d02019-10-22 10:25:58 -0600899}
900
901static enum io_wq_cancel io_wqe_cancel_work(struct io_wqe *wqe,
902 struct io_wq_work *cwork)
903{
Jens Axboe6206f0e2019-11-26 11:59:32 -0700904 struct io_wq_work_node *node, *prev;
Jens Axboe771b53d02019-10-22 10:25:58 -0600905 struct io_wq_work *work;
Jens Axboe6f726532019-11-05 13:51:51 -0700906 unsigned long flags;
Jens Axboe771b53d02019-10-22 10:25:58 -0600907 bool found = false;
908
909 cwork->flags |= IO_WQ_WORK_CANCEL;
910
911 /*
912 * First check pending list, if we're lucky we can just remove it
913 * from there. CANCEL_OK means that the work is returned as-new,
914 * no completion will be posted for it.
915 */
Jens Axboe6f726532019-11-05 13:51:51 -0700916 spin_lock_irqsave(&wqe->lock, flags);
Jens Axboe6206f0e2019-11-26 11:59:32 -0700917 wq_list_for_each(node, prev, &wqe->work_list) {
918 work = container_of(node, struct io_wq_work, list);
919
Jens Axboe771b53d02019-10-22 10:25:58 -0600920 if (work == cwork) {
Jens Axboe6206f0e2019-11-26 11:59:32 -0700921 wq_node_del(&wqe->work_list, node, prev);
Jens Axboe771b53d02019-10-22 10:25:58 -0600922 found = true;
923 break;
924 }
925 }
Jens Axboe6f726532019-11-05 13:51:51 -0700926 spin_unlock_irqrestore(&wqe->lock, flags);
Jens Axboe771b53d02019-10-22 10:25:58 -0600927
928 if (found) {
929 work->flags |= IO_WQ_WORK_CANCEL;
930 work->func(&work);
931 return IO_WQ_CANCEL_OK;
932 }
933
934 /*
935 * Now check if a free (going busy) or busy worker has the work
936 * currently running. If we find it there, we'll return CANCEL_RUNNING
937 * as an indication that we attempte to signal cancellation. The
938 * completion will run normally in this case.
939 */
940 rcu_read_lock();
Jens Axboee61df662019-11-13 13:54:49 -0700941 found = io_wq_for_each_worker(wqe, io_wq_worker_cancel, cwork);
Jens Axboe771b53d02019-10-22 10:25:58 -0600942 rcu_read_unlock();
943 return found ? IO_WQ_CANCEL_RUNNING : IO_WQ_CANCEL_NOTFOUND;
944}
945
946enum io_wq_cancel io_wq_cancel_work(struct io_wq *wq, struct io_wq_work *cwork)
947{
948 enum io_wq_cancel ret = IO_WQ_CANCEL_NOTFOUND;
Jann Horn3fc50ab2019-11-26 19:10:20 +0100949 int node;
Jens Axboe771b53d02019-10-22 10:25:58 -0600950
Jann Horn3fc50ab2019-11-26 19:10:20 +0100951 for_each_node(node) {
952 struct io_wqe *wqe = wq->wqes[node];
Jens Axboe771b53d02019-10-22 10:25:58 -0600953
954 ret = io_wqe_cancel_work(wqe, cwork);
955 if (ret != IO_WQ_CANCEL_NOTFOUND)
956 break;
957 }
958
959 return ret;
960}
961
962struct io_wq_flush_data {
963 struct io_wq_work work;
964 struct completion done;
965};
966
967static void io_wq_flush_func(struct io_wq_work **workptr)
968{
969 struct io_wq_work *work = *workptr;
970 struct io_wq_flush_data *data;
971
972 data = container_of(work, struct io_wq_flush_data, work);
973 complete(&data->done);
974}
975
976/*
977 * Doesn't wait for previously queued work to finish. When this completes,
978 * it just means that previously queued work was started.
979 */
980void io_wq_flush(struct io_wq *wq)
981{
982 struct io_wq_flush_data data;
Jann Horn3fc50ab2019-11-26 19:10:20 +0100983 int node;
Jens Axboe771b53d02019-10-22 10:25:58 -0600984
Jann Horn3fc50ab2019-11-26 19:10:20 +0100985 for_each_node(node) {
986 struct io_wqe *wqe = wq->wqes[node];
Jens Axboe771b53d02019-10-22 10:25:58 -0600987
988 init_completion(&data.done);
989 INIT_IO_WORK(&data.work, io_wq_flush_func);
Jens Axboe7d723062019-11-12 22:31:31 -0700990 data.work.flags |= IO_WQ_WORK_INTERNAL;
Jens Axboe771b53d02019-10-22 10:25:58 -0600991 io_wqe_enqueue(wqe, &data.work);
992 wait_for_completion(&data.done);
993 }
994}
995
Jens Axboe576a3472019-11-25 08:49:20 -0700996struct io_wq *io_wq_create(unsigned bounded, struct io_wq_data *data)
Jens Axboe771b53d02019-10-22 10:25:58 -0600997{
Jann Horn3fc50ab2019-11-26 19:10:20 +0100998 int ret = -ENOMEM, node;
Jens Axboe771b53d02019-10-22 10:25:58 -0600999 struct io_wq *wq;
1000
Jann Hornad6e0052019-11-26 17:39:45 +01001001 wq = kzalloc(sizeof(*wq), GFP_KERNEL);
Jens Axboe771b53d02019-10-22 10:25:58 -06001002 if (!wq)
1003 return ERR_PTR(-ENOMEM);
1004
Jann Horn3fc50ab2019-11-26 19:10:20 +01001005 wq->wqes = kcalloc(nr_node_ids, sizeof(struct io_wqe *), GFP_KERNEL);
Jens Axboe771b53d02019-10-22 10:25:58 -06001006 if (!wq->wqes) {
1007 kfree(wq);
1008 return ERR_PTR(-ENOMEM);
1009 }
1010
Jens Axboe576a3472019-11-25 08:49:20 -07001011 wq->get_work = data->get_work;
1012 wq->put_work = data->put_work;
Jens Axboe7d723062019-11-12 22:31:31 -07001013
Jens Axboec5def4a2019-11-07 11:41:16 -07001014 /* caller must already hold a reference to this */
Jens Axboe576a3472019-11-25 08:49:20 -07001015 wq->user = data->user;
Jens Axboe181e4482019-11-25 08:52:30 -07001016 wq->creds = data->creds;
Jens Axboec5def4a2019-11-07 11:41:16 -07001017
Jann Horn3fc50ab2019-11-26 19:10:20 +01001018 for_each_node(node) {
Jens Axboe771b53d02019-10-22 10:25:58 -06001019 struct io_wqe *wqe;
1020
Jann Hornad6e0052019-11-26 17:39:45 +01001021 wqe = kzalloc_node(sizeof(struct io_wqe), GFP_KERNEL, node);
Jens Axboe771b53d02019-10-22 10:25:58 -06001022 if (!wqe)
Jann Horn3fc50ab2019-11-26 19:10:20 +01001023 goto err;
1024 wq->wqes[node] = wqe;
Jens Axboe771b53d02019-10-22 10:25:58 -06001025 wqe->node = node;
Jens Axboec5def4a2019-11-07 11:41:16 -07001026 wqe->acct[IO_WQ_ACCT_BOUND].max_workers = bounded;
1027 atomic_set(&wqe->acct[IO_WQ_ACCT_BOUND].nr_running, 0);
Jens Axboe576a3472019-11-25 08:49:20 -07001028 if (wq->user) {
Jens Axboec5def4a2019-11-07 11:41:16 -07001029 wqe->acct[IO_WQ_ACCT_UNBOUND].max_workers =
1030 task_rlimit(current, RLIMIT_NPROC);
1031 }
1032 atomic_set(&wqe->acct[IO_WQ_ACCT_UNBOUND].nr_running, 0);
Jens Axboe771b53d02019-10-22 10:25:58 -06001033 wqe->node = node;
1034 wqe->wq = wq;
1035 spin_lock_init(&wqe->lock);
Jens Axboe6206f0e2019-11-26 11:59:32 -07001036 INIT_WQ_LIST(&wqe->work_list);
Jens Axboe021d1cd2019-11-14 08:00:41 -07001037 INIT_HLIST_NULLS_HEAD(&wqe->free_list, 0);
1038 INIT_HLIST_NULLS_HEAD(&wqe->busy_list, 1);
Jens Axboee61df662019-11-13 13:54:49 -07001039 INIT_LIST_HEAD(&wqe->all_list);
Jens Axboe771b53d02019-10-22 10:25:58 -06001040 }
1041
1042 init_completion(&wq->done);
1043
Jens Axboe771b53d02019-10-22 10:25:58 -06001044 /* caller must have already done mmgrab() on this mm */
Jens Axboe576a3472019-11-25 08:49:20 -07001045 wq->mm = data->mm;
Jens Axboe771b53d02019-10-22 10:25:58 -06001046
1047 wq->manager = kthread_create(io_wq_manager, wq, "io_wq_manager");
1048 if (!IS_ERR(wq->manager)) {
1049 wake_up_process(wq->manager);
Jens Axboeb60fda62019-11-19 08:37:07 -07001050 wait_for_completion(&wq->done);
1051 if (test_bit(IO_WQ_BIT_ERROR, &wq->state)) {
1052 ret = -ENOMEM;
1053 goto err;
1054 }
1055 reinit_completion(&wq->done);
Jens Axboe771b53d02019-10-22 10:25:58 -06001056 return wq;
1057 }
1058
1059 ret = PTR_ERR(wq->manager);
Jens Axboe771b53d02019-10-22 10:25:58 -06001060 complete(&wq->done);
Jens Axboeb60fda62019-11-19 08:37:07 -07001061err:
Jann Horn3fc50ab2019-11-26 19:10:20 +01001062 for_each_node(node)
1063 kfree(wq->wqes[node]);
Jens Axboeb60fda62019-11-19 08:37:07 -07001064 kfree(wq->wqes);
1065 kfree(wq);
Jens Axboe771b53d02019-10-22 10:25:58 -06001066 return ERR_PTR(ret);
1067}
1068
1069static bool io_wq_worker_wake(struct io_worker *worker, void *data)
1070{
1071 wake_up_process(worker->task);
1072 return false;
1073}
1074
1075void io_wq_destroy(struct io_wq *wq)
1076{
Jann Horn3fc50ab2019-11-26 19:10:20 +01001077 int node;
Jens Axboe771b53d02019-10-22 10:25:58 -06001078
Jens Axboeb60fda62019-11-19 08:37:07 -07001079 set_bit(IO_WQ_BIT_EXIT, &wq->state);
1080 if (wq->manager)
Jens Axboe771b53d02019-10-22 10:25:58 -06001081 kthread_stop(wq->manager);
Jens Axboe771b53d02019-10-22 10:25:58 -06001082
1083 rcu_read_lock();
Jann Horn3fc50ab2019-11-26 19:10:20 +01001084 for_each_node(node)
1085 io_wq_for_each_worker(wq->wqes[node], io_wq_worker_wake, NULL);
Jens Axboe771b53d02019-10-22 10:25:58 -06001086 rcu_read_unlock();
1087
1088 wait_for_completion(&wq->done);
1089
Jann Horn3fc50ab2019-11-26 19:10:20 +01001090 for_each_node(node)
1091 kfree(wq->wqes[node]);
Jens Axboe771b53d02019-10-22 10:25:58 -06001092 kfree(wq->wqes);
1093 kfree(wq);
1094}