Blame view
fs/io-wq.c
27.5 KB
771b53d03
|
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 |
// 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/mmu_context.h> #include <linux/sched/mm.h> #include <linux/percpu.h> #include <linux/slab.h> #include <linux/kthread.h> #include <linux/rculist_nulls.h> |
9392a27d8
|
19 |
#include <linux/fs_struct.h> |
771b53d03
|
20 21 22 23 24 25 26 27 28 29 30 |
#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_WORKER_F_EXITING = 8, /* worker exiting */ IO_WORKER_F_FIXED = 16, /* static idle worker */ |
c5def4ab8
|
31 |
IO_WORKER_F_BOUND = 32, /* is doing bounded work */ |
771b53d03
|
32 33 34 35 36 |
}; enum { IO_WQ_BIT_EXIT = 0, /* wq exiting */ IO_WQ_BIT_CANCEL = 1, /* cancel work on list */ |
b60fda600
|
37 |
IO_WQ_BIT_ERROR = 2, /* error on setup */ |
771b53d03
|
38 39 40 41 42 43 44 45 46 47 48 49 50 |
}; 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; |
e61df66c6
|
51 |
struct list_head all_list; |
771b53d03
|
52 |
struct task_struct *task; |
771b53d03
|
53 |
struct io_wqe *wqe; |
36c2f9223
|
54 |
|
771b53d03
|
55 |
struct io_wq_work *cur_work; |
36c2f9223
|
56 |
spinlock_t lock; |
771b53d03
|
57 58 59 |
struct rcu_head rcu; struct mm_struct *mm; |
cccf0ee83
|
60 61 |
const struct cred *cur_creds; const struct cred *saved_creds; |
fcb323cc5
|
62 |
struct files_struct *restore_files; |
9392a27d8
|
63 |
struct fs_struct *restore_fs; |
771b53d03
|
64 |
}; |
771b53d03
|
65 66 67 68 69 |
#if BITS_PER_LONG == 64 #define IO_WQ_HASH_ORDER 6 #else #define IO_WQ_HASH_ORDER 5 #endif |
c5def4ab8
|
70 71 72 73 74 75 76 77 78 79 |
struct io_wqe_acct { unsigned nr_workers; unsigned max_workers; atomic_t nr_running; }; enum { IO_WQ_ACCT_BOUND, IO_WQ_ACCT_UNBOUND, }; |
771b53d03
|
80 81 82 83 84 85 |
/* * Per-node worker thread pool */ struct io_wqe { struct { spinlock_t lock; |
6206f0e18
|
86 |
struct io_wq_work_list work_list; |
771b53d03
|
87 88 89 90 91 |
unsigned long hash_map; unsigned flags; } ____cacheline_aligned_in_smp; int node; |
c5def4ab8
|
92 |
struct io_wqe_acct acct[2]; |
771b53d03
|
93 |
|
021d1cdda
|
94 |
struct hlist_nulls_head free_list; |
e61df66c6
|
95 |
struct list_head all_list; |
771b53d03
|
96 97 98 99 100 101 102 103 104 105 |
struct io_wq *wq; }; /* * Per io_wq state */ struct io_wq { struct io_wqe **wqes; unsigned long state; |
771b53d03
|
106 |
|
7d7230652
|
107 108 |
get_work_fn *get_work; put_work_fn *put_work; |
771b53d03
|
109 |
struct task_struct *manager; |
c5def4ab8
|
110 |
struct user_struct *user; |
771b53d03
|
111 112 |
refcount_t refs; struct completion done; |
848f7e188
|
113 114 |
refcount_t use_refs; |
771b53d03
|
115 |
}; |
771b53d03
|
116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 |
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) { |
fcb323cc5
|
134 |
bool dropped_lock = false; |
cccf0ee83
|
135 136 137 |
if (worker->saved_creds) { revert_creds(worker->saved_creds); worker->cur_creds = worker->saved_creds = NULL; |
181e448d8
|
138 |
} |
fcb323cc5
|
139 140 141 142 143 144 145 146 147 |
if (current->files != worker->restore_files) { __acquire(&wqe->lock); spin_unlock_irq(&wqe->lock); dropped_lock = true; task_lock(current); current->files = worker->restore_files; task_unlock(current); } |
9392a27d8
|
148 149 |
if (current->fs != worker->restore_fs) current->fs = worker->restore_fs; |
771b53d03
|
150 151 152 153 154 |
/* * 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) { |
fcb323cc5
|
155 156 157 158 159 |
if (!dropped_lock) { __acquire(&wqe->lock); spin_unlock_irq(&wqe->lock); dropped_lock = true; } |
771b53d03
|
160 161 162 163 164 |
__set_current_state(TASK_RUNNING); set_fs(KERNEL_DS); unuse_mm(worker->mm); mmput(worker->mm); worker->mm = NULL; |
771b53d03
|
165 |
} |
fcb323cc5
|
166 |
return dropped_lock; |
771b53d03
|
167 |
} |
c5def4ab8
|
168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 |
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]; } |
771b53d03
|
185 186 187 |
static void io_worker_exit(struct io_worker *worker) { struct io_wqe *wqe = worker->wqe; |
c5def4ab8
|
188 189 |
struct io_wqe_acct *acct = io_wqe_get_acct(wqe, worker); unsigned nr_workers; |
771b53d03
|
190 191 192 193 194 195 196 197 198 199 200 201 202 |
/* * 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) |
c5def4ab8
|
203 204 205 |
atomic_dec(&acct->nr_running); if (!(worker->flags & IO_WORKER_F_BOUND)) atomic_dec(&wqe->wq->user->processes); |
771b53d03
|
206 207 208 209 210 |
worker->flags = 0; preempt_enable(); spin_lock_irq(&wqe->lock); hlist_nulls_del_rcu(&worker->nulls_node); |
e61df66c6
|
211 |
list_del_rcu(&worker->all_list); |
771b53d03
|
212 213 214 215 |
if (__io_worker_unuse(wqe, worker)) { __release(&wqe->lock); spin_lock_irq(&wqe->lock); } |
c5def4ab8
|
216 217 218 |
acct->nr_workers--; nr_workers = wqe->acct[IO_WQ_ACCT_BOUND].nr_workers + wqe->acct[IO_WQ_ACCT_UNBOUND].nr_workers; |
771b53d03
|
219 220 221 |
spin_unlock_irq(&wqe->lock); /* all workers gone, wq exit can proceed */ |
c5def4ab8
|
222 |
if (!nr_workers && refcount_dec_and_test(&wqe->wq->refs)) |
771b53d03
|
223 |
complete(&wqe->wq->done); |
364b05fd0
|
224 |
kfree_rcu(worker, rcu); |
771b53d03
|
225 |
} |
c5def4ab8
|
226 227 228 |
static inline bool io_wqe_run_queue(struct io_wqe *wqe) __must_hold(wqe->lock) { |
6206f0e18
|
229 230 |
if (!wq_list_empty(&wqe->work_list) && !(wqe->flags & IO_WQE_FLAG_STALLED)) |
c5def4ab8
|
231 232 233 234 235 236 237 238 239 240 241 242 243 |
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; |
021d1cdda
|
244 |
n = rcu_dereference(hlist_nulls_first_rcu(&wqe->free_list)); |
c5def4ab8
|
245 246 247 248 249 |
if (is_a_nulls(n)) return false; worker = hlist_nulls_entry(n, struct io_worker, nulls_node); if (io_worker_get(worker)) { |
506d95ff5
|
250 |
wake_up_process(worker->task); |
c5def4ab8
|
251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 |
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); } |
771b53d03
|
295 296 297 298 299 300 301 |
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); |
fcb323cc5
|
302 |
worker->restore_files = current->files; |
9392a27d8
|
303 |
worker->restore_fs = current->fs; |
c5def4ab8
|
304 |
io_wqe_inc_running(wqe, worker); |
771b53d03
|
305 306 307 308 309 310 311 312 313 314 |
} /* * 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) { |
c5def4ab8
|
315 |
bool worker_bound, work_bound; |
771b53d03
|
316 317 318 |
if (worker->flags & IO_WORKER_F_FREE) { worker->flags &= ~IO_WORKER_F_FREE; hlist_nulls_del_init_rcu(&worker->nulls_node); |
771b53d03
|
319 |
} |
c5def4ab8
|
320 321 322 323 324 |
/* * If worker is moving from bound to unbound (or vice versa), then * ensure we update the running accounting. */ |
b2e9c7d64
|
325 326 327 |
worker_bound = (worker->flags & IO_WORKER_F_BOUND) != 0; work_bound = (work->flags & IO_WQ_WORK_UNBOUND) == 0; if (worker_bound != work_bound) { |
c5def4ab8
|
328 329 330 331 332 333 334 335 336 337 338 339 340 341 |
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); } |
771b53d03
|
342 343 344 345 346 347 348 349 350 351 352 353 354 355 |
} /* * 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; |
021d1cdda
|
356 |
hlist_nulls_add_head_rcu(&worker->nulls_node, &wqe->free_list); |
771b53d03
|
357 358 359 360 361 362 363 364 |
} return __io_worker_unuse(wqe, worker); } static struct io_wq_work *io_get_next_work(struct io_wqe *wqe, unsigned *hash) __must_hold(wqe->lock) { |
6206f0e18
|
365 |
struct io_wq_work_node *node, *prev; |
771b53d03
|
366 |
struct io_wq_work *work; |
6206f0e18
|
367 368 |
wq_list_for_each(node, prev, &wqe->work_list) { work = container_of(node, struct io_wq_work, list); |
771b53d03
|
369 370 |
/* not hashed, can run anytime */ if (!(work->flags & IO_WQ_WORK_HASHED)) { |
6206f0e18
|
371 |
wq_node_del(&wqe->work_list, node, prev); |
771b53d03
|
372 373 374 375 376 377 378 |
return work; } /* hashed, can run if not already running */ *hash = work->flags >> IO_WQ_HASH_SHIFT; if (!(wqe->hash_map & BIT_ULL(*hash))) { wqe->hash_map |= BIT_ULL(*hash); |
6206f0e18
|
379 |
wq_node_del(&wqe->work_list, node, prev); |
771b53d03
|
380 381 382 383 384 385 |
return work; } } return NULL; } |
cccf0ee83
|
386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 |
static void io_wq_switch_mm(struct io_worker *worker, struct io_wq_work *work) { if (worker->mm) { unuse_mm(worker->mm); mmput(worker->mm); worker->mm = NULL; } if (!work->mm) { set_fs(KERNEL_DS); return; } if (mmget_not_zero(work->mm)) { use_mm(work->mm); if (!worker->mm) set_fs(USER_DS); worker->mm = work->mm; /* hang on to this mm */ work->mm = NULL; return; } /* failed grabbing mm, ensure work gets cancelled */ work->flags |= IO_WQ_WORK_CANCEL; } static void io_wq_switch_creds(struct io_worker *worker, struct io_wq_work *work) { const struct cred *old_creds = override_creds(work->creds); worker->cur_creds = work->creds; if (worker->saved_creds) put_cred(old_creds); /* creds set by previous switch */ else worker->saved_creds = old_creds; } |
771b53d03
|
422 423 424 |
static void io_worker_handle_work(struct io_worker *worker) __releases(wqe->lock) { |
7d7230652
|
425 |
struct io_wq_work *work, *old_work = NULL, *put_work = NULL; |
771b53d03
|
426 427 428 429 430 431 432 |
struct io_wqe *wqe = worker->wqe; struct io_wq *wq = wqe->wq; do { unsigned hash = -1U; /* |
771b53d03
|
433 434 435 436 437 438 439 440 441 |
* 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. */ work = io_get_next_work(wqe, &hash); if (work) __io_worker_busy(wqe, worker, work); |
6206f0e18
|
442 |
else if (!wq_list_empty(&wqe->work_list)) |
771b53d03
|
443 444 445 |
wqe->flags |= IO_WQE_FLAG_STALLED; spin_unlock_irq(&wqe->lock); |
7d7230652
|
446 447 |
if (put_work && wq->put_work) wq->put_work(old_work); |
771b53d03
|
448 449 450 |
if (!work) break; next: |
36c2f9223
|
451 452 453 |
/* flush any pending signals before assigning new work */ if (signal_pending(current)) flush_signals(current); |
fd1c4bc6e
|
454 |
cond_resched(); |
36c2f9223
|
455 456 457 |
spin_lock_irq(&worker->lock); worker->cur_work = work; spin_unlock_irq(&worker->lock); |
b76da70fc
|
458 459 |
if (work->flags & IO_WQ_WORK_CB) work->func(&work); |
f86cd20c9
|
460 |
if (work->files && current->files != work->files) { |
fcb323cc5
|
461 462 463 464 |
task_lock(current); current->files = work->files; task_unlock(current); } |
9392a27d8
|
465 466 |
if (work->fs && current->fs != work->fs) current->fs = work->fs; |
cccf0ee83
|
467 468 469 470 |
if (work->mm != worker->mm) io_wq_switch_mm(worker, work); if (worker->cur_creds != work->creds) io_wq_switch_creds(worker, work); |
0c9d5ccd2
|
471 472 473 474 |
/* * OK to set IO_WQ_WORK_CANCEL even for uncancellable work, * the worker function will do the right thing. */ |
771b53d03
|
475 476 477 478 |
if (test_bit(IO_WQ_BIT_CANCEL, &wq->state)) work->flags |= IO_WQ_WORK_CANCEL; if (worker->mm) work->flags |= IO_WQ_WORK_HAS_MM; |
80ad89438
|
479 |
if (wq->get_work) { |
7d7230652
|
480 481 482 |
put_work = work; wq->get_work(work); } |
771b53d03
|
483 484 |
old_work = work; work->func(&work); |
36c2f9223
|
485 |
spin_lock_irq(&worker->lock); |
771b53d03
|
486 |
worker->cur_work = NULL; |
36c2f9223
|
487 488 489 |
spin_unlock_irq(&worker->lock); spin_lock_irq(&wqe->lock); |
771b53d03
|
490 491 492 493 494 495 |
if (hash != -1U) { wqe->hash_map &= ~BIT_ULL(hash); wqe->flags &= ~IO_WQE_FLAG_STALLED; } if (work && work != old_work) { spin_unlock_irq(&wqe->lock); |
7d7230652
|
496 497 498 499 500 |
if (put_work && wq->put_work) { wq->put_work(put_work); put_work = NULL; } |
771b53d03
|
501 502 503 504 505 506 |
/* dependent work not hashed */ hash = -1U; goto next; } } while (1); } |
771b53d03
|
507 508 509 510 511 |
static int io_wqe_worker(void *data) { struct io_worker *worker = data; struct io_wqe *wqe = worker->wqe; struct io_wq *wq = wqe->wq; |
771b53d03
|
512 513 514 515 |
io_worker_start(wqe, worker); while (!test_bit(IO_WQ_BIT_EXIT, &wq->state)) { |
506d95ff5
|
516 |
set_current_state(TASK_INTERRUPTIBLE); |
e995d5123
|
517 |
loop: |
771b53d03
|
518 519 520 521 |
spin_lock_irq(&wqe->lock); if (io_wqe_run_queue(wqe)) { __set_current_state(TASK_RUNNING); io_worker_handle_work(worker); |
e995d5123
|
522 |
goto loop; |
771b53d03
|
523 524 525 526 |
} /* drops the lock on success, retry */ if (__io_worker_idle(wqe, worker)) { __release(&wqe->lock); |
e995d5123
|
527 |
goto loop; |
771b53d03
|
528 529 530 531 532 533 534 535 536 537 538 |
} spin_unlock_irq(&wqe->lock); 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; } |
771b53d03
|
539 540 |
if (test_bit(IO_WQ_BIT_EXIT, &wq->state)) { spin_lock_irq(&wqe->lock); |
6206f0e18
|
541 |
if (!wq_list_empty(&wqe->work_list)) |
771b53d03
|
542 543 544 545 546 547 548 549 550 551 |
io_worker_handle_work(worker); else spin_unlock_irq(&wqe->lock); } io_worker_exit(worker); return 0; } /* |
771b53d03
|
552 553 554 555 556 557 558 559 560 561 562 563 |
* 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; |
c5def4ab8
|
564 |
io_wqe_inc_running(wqe, worker); |
771b53d03
|
565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 |
} /* * 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; spin_lock_irq(&wqe->lock); |
c5def4ab8
|
585 |
io_wqe_dec_running(wqe, worker); |
771b53d03
|
586 587 |
spin_unlock_irq(&wqe->lock); } |
b60fda600
|
588 |
static bool create_io_worker(struct io_wq *wq, struct io_wqe *wqe, int index) |
771b53d03
|
589 |
{ |
c5def4ab8
|
590 |
struct io_wqe_acct *acct =&wqe->acct[index]; |
771b53d03
|
591 |
struct io_worker *worker; |
ad6e005ca
|
592 |
worker = kzalloc_node(sizeof(*worker), GFP_KERNEL, wqe->node); |
771b53d03
|
593 |
if (!worker) |
b60fda600
|
594 |
return false; |
771b53d03
|
595 596 597 |
refcount_set(&worker->ref, 1); worker->nulls_node.pprev = NULL; |
771b53d03
|
598 |
worker->wqe = wqe; |
36c2f9223
|
599 |
spin_lock_init(&worker->lock); |
771b53d03
|
600 601 |
worker->task = kthread_create_on_node(io_wqe_worker, worker, wqe->node, |
c5def4ab8
|
602 |
"io_wqe_worker-%d/%d", index, wqe->node); |
771b53d03
|
603 604 |
if (IS_ERR(worker->task)) { kfree(worker); |
b60fda600
|
605 |
return false; |
771b53d03
|
606 607 608 |
} spin_lock_irq(&wqe->lock); |
021d1cdda
|
609 |
hlist_nulls_add_head_rcu(&worker->nulls_node, &wqe->free_list); |
e61df66c6
|
610 |
list_add_tail_rcu(&worker->all_list, &wqe->all_list); |
771b53d03
|
611 |
worker->flags |= IO_WORKER_F_FREE; |
c5def4ab8
|
612 613 614 |
if (index == IO_WQ_ACCT_BOUND) worker->flags |= IO_WORKER_F_BOUND; if (!acct->nr_workers && (worker->flags & IO_WORKER_F_BOUND)) |
771b53d03
|
615 |
worker->flags |= IO_WORKER_F_FIXED; |
c5def4ab8
|
616 |
acct->nr_workers++; |
771b53d03
|
617 |
spin_unlock_irq(&wqe->lock); |
c5def4ab8
|
618 619 |
if (index == IO_WQ_ACCT_UNBOUND) atomic_inc(&wq->user->processes); |
771b53d03
|
620 |
wake_up_process(worker->task); |
b60fda600
|
621 |
return true; |
771b53d03
|
622 |
} |
c5def4ab8
|
623 |
static inline bool io_wqe_need_worker(struct io_wqe *wqe, int index) |
771b53d03
|
624 625 |
__must_hold(wqe->lock) { |
c5def4ab8
|
626 |
struct io_wqe_acct *acct = &wqe->acct[index]; |
771b53d03
|
627 |
|
c5def4ab8
|
628 |
/* if we have available workers or no work, no need */ |
021d1cdda
|
629 |
if (!hlist_nulls_empty(&wqe->free_list) || !io_wqe_run_queue(wqe)) |
c5def4ab8
|
630 631 |
return false; return acct->nr_workers < acct->max_workers; |
771b53d03
|
632 633 634 635 636 637 638 639 |
} /* * Manager thread. Tasked with creating new workers, if we need them. */ static int io_wq_manager(void *data) { struct io_wq *wq = data; |
3fc50ab55
|
640 641 |
int workers_to_create = num_possible_nodes(); int node; |
771b53d03
|
642 |
|
b60fda600
|
643 |
/* create fixed workers */ |
3fc50ab55
|
644 645 |
refcount_set(&wq->refs, workers_to_create); for_each_node(node) { |
7563439ad
|
646 647 |
if (!node_online(node)) continue; |
3fc50ab55
|
648 649 650 |
if (!create_io_worker(wq, wq->wqes[node], IO_WQ_ACCT_BOUND)) goto err; workers_to_create--; |
b60fda600
|
651 |
} |
771b53d03
|
652 |
|
7563439ad
|
653 654 |
while (workers_to_create--) refcount_dec(&wq->refs); |
b60fda600
|
655 656 657 |
complete(&wq->done); while (!kthread_should_stop()) { |
3fc50ab55
|
658 659 |
for_each_node(node) { struct io_wqe *wqe = wq->wqes[node]; |
c5def4ab8
|
660 |
bool fork_worker[2] = { false, false }; |
771b53d03
|
661 |
|
7563439ad
|
662 663 |
if (!node_online(node)) continue; |
771b53d03
|
664 |
spin_lock_irq(&wqe->lock); |
c5def4ab8
|
665 666 667 668 |
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; |
771b53d03
|
669 |
spin_unlock_irq(&wqe->lock); |
c5def4ab8
|
670 671 672 673 |
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); |
771b53d03
|
674 675 676 677 678 679 |
} set_current_state(TASK_INTERRUPTIBLE); schedule_timeout(HZ); } return 0; |
b60fda600
|
680 681 682 |
err: set_bit(IO_WQ_BIT_ERROR, &wq->state); set_bit(IO_WQ_BIT_EXIT, &wq->state); |
3fc50ab55
|
683 |
if (refcount_sub_and_test(workers_to_create, &wq->refs)) |
b60fda600
|
684 685 |
complete(&wq->done); return 0; |
771b53d03
|
686 |
} |
c5def4ab8
|
687 688 689 690 691 692 693 694 695 696 697 |
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(); |
021d1cdda
|
698 |
free_worker = !hlist_nulls_empty(&wqe->free_list); |
c5def4ab8
|
699 700 701 702 703 704 705 706 707 708 |
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; } |
fc04c39ba
|
709 710 711 712 713 714 715 716 717 718 |
static void io_run_cancel(struct io_wq_work *work) { do { struct io_wq_work *old_work = work; work->flags |= IO_WQ_WORK_CANCEL; work->func(&work); work = (work == old_work) ? NULL : work; } while (work); } |
771b53d03
|
719 720 |
static void io_wqe_enqueue(struct io_wqe *wqe, struct io_wq_work *work) { |
c5def4ab8
|
721 |
struct io_wqe_acct *acct = io_work_get_acct(wqe, work); |
895e2ca0f
|
722 |
int work_flags; |
771b53d03
|
723 |
unsigned long flags; |
c5def4ab8
|
724 725 726 727 728 729 730 |
/* * 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))) { |
fc04c39ba
|
731 |
io_run_cancel(work); |
c5def4ab8
|
732 733 |
return; } |
895e2ca0f
|
734 |
work_flags = work->flags; |
771b53d03
|
735 |
spin_lock_irqsave(&wqe->lock, flags); |
6206f0e18
|
736 |
wq_list_add_tail(&work->list, &wqe->work_list); |
771b53d03
|
737 738 |
wqe->flags &= ~IO_WQE_FLAG_STALLED; spin_unlock_irqrestore(&wqe->lock, flags); |
895e2ca0f
|
739 740 |
if ((work_flags & IO_WQ_WORK_CONCURRENT) || !atomic_read(&acct->nr_running)) |
c5def4ab8
|
741 |
io_wqe_wake_worker(wqe, acct); |
771b53d03
|
742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 |
} 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); } /* * Enqueue work, hashed by some key. Work items that hash to the same value * will not be done in parallel. Used to limit concurrent writes, generally * hashed by inode. */ void io_wq_enqueue_hashed(struct io_wq *wq, struct io_wq_work *work, void *val) { struct io_wqe *wqe = wq->wqes[numa_node_id()]; unsigned bit; bit = hash_ptr(val, IO_WQ_HASH_ORDER); work->flags |= (IO_WQ_WORK_HASHED | (bit << IO_WQ_HASH_SHIFT)); io_wqe_enqueue(wqe, work); } static bool io_wqe_worker_send_sig(struct io_worker *worker, void *data) { send_sig(SIGINT, worker->task, 1); return false; } /* * 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, |
771b53d03
|
778 779 780 |
bool (*func)(struct io_worker *, void *), void *data) { |
771b53d03
|
781 782 |
struct io_worker *worker; bool ret = false; |
e61df66c6
|
783 |
list_for_each_entry_rcu(worker, &wqe->all_list, all_list) { |
771b53d03
|
784 |
if (io_worker_get(worker)) { |
7563439ad
|
785 786 787 |
/* no task if node is/was offline */ if (worker->task) ret = func(worker, data); |
771b53d03
|
788 789 790 791 792 |
io_worker_release(worker); if (ret) break; } } |
e61df66c6
|
793 |
|
771b53d03
|
794 795 796 797 798 |
return ret; } void io_wq_cancel_all(struct io_wq *wq) { |
3fc50ab55
|
799 |
int node; |
771b53d03
|
800 801 |
set_bit(IO_WQ_BIT_CANCEL, &wq->state); |
771b53d03
|
802 |
rcu_read_lock(); |
3fc50ab55
|
803 804 |
for_each_node(node) { struct io_wqe *wqe = wq->wqes[node]; |
771b53d03
|
805 |
|
e61df66c6
|
806 |
io_wq_for_each_worker(wqe, io_wqe_worker_send_sig, NULL); |
771b53d03
|
807 808 809 |
} rcu_read_unlock(); } |
62755e35d
|
810 811 812 813 814 815 816 817 818 |
struct io_cb_cancel_data { struct io_wqe *wqe; work_cancel_fn *cancel; void *caller_data; }; static bool io_work_cancel(struct io_worker *worker, void *cancel_data) { struct io_cb_cancel_data *data = cancel_data; |
6f72653e7
|
819 |
unsigned long flags; |
62755e35d
|
820 821 822 823 |
bool ret = false; /* * Hold the lock to avoid ->cur_work going out of scope, caller |
36c2f9223
|
824 |
* may dereference the passed in work. |
62755e35d
|
825 |
*/ |
36c2f9223
|
826 |
spin_lock_irqsave(&worker->lock, flags); |
62755e35d
|
827 |
if (worker->cur_work && |
0c9d5ccd2
|
828 |
!(worker->cur_work->flags & IO_WQ_WORK_NO_CANCEL) && |
62755e35d
|
829 830 831 832 |
data->cancel(worker->cur_work, data->caller_data)) { send_sig(SIGINT, worker->task, 1); ret = true; } |
36c2f9223
|
833 |
spin_unlock_irqrestore(&worker->lock, flags); |
62755e35d
|
834 835 836 837 838 839 840 841 842 843 844 845 846 |
return ret; } static enum io_wq_cancel io_wqe_cancel_cb_work(struct io_wqe *wqe, work_cancel_fn *cancel, void *cancel_data) { struct io_cb_cancel_data data = { .wqe = wqe, .cancel = cancel, .caller_data = cancel_data, }; |
6206f0e18
|
847 |
struct io_wq_work_node *node, *prev; |
62755e35d
|
848 |
struct io_wq_work *work; |
6f72653e7
|
849 |
unsigned long flags; |
62755e35d
|
850 |
bool found = false; |
6f72653e7
|
851 |
spin_lock_irqsave(&wqe->lock, flags); |
6206f0e18
|
852 853 |
wq_list_for_each(node, prev, &wqe->work_list) { work = container_of(node, struct io_wq_work, list); |
62755e35d
|
854 |
if (cancel(work, cancel_data)) { |
6206f0e18
|
855 |
wq_node_del(&wqe->work_list, node, prev); |
62755e35d
|
856 857 858 859 |
found = true; break; } } |
6f72653e7
|
860 |
spin_unlock_irqrestore(&wqe->lock, flags); |
62755e35d
|
861 862 |
if (found) { |
fc04c39ba
|
863 |
io_run_cancel(work); |
62755e35d
|
864 865 866 867 |
return IO_WQ_CANCEL_OK; } rcu_read_lock(); |
e61df66c6
|
868 |
found = io_wq_for_each_worker(wqe, io_work_cancel, &data); |
62755e35d
|
869 870 871 872 873 874 875 876 |
rcu_read_unlock(); return found ? IO_WQ_CANCEL_RUNNING : IO_WQ_CANCEL_NOTFOUND; } enum io_wq_cancel io_wq_cancel_cb(struct io_wq *wq, work_cancel_fn *cancel, void *data) { enum io_wq_cancel ret = IO_WQ_CANCEL_NOTFOUND; |
3fc50ab55
|
877 |
int node; |
62755e35d
|
878 |
|
3fc50ab55
|
879 880 |
for_each_node(node) { struct io_wqe *wqe = wq->wqes[node]; |
62755e35d
|
881 882 883 884 885 886 887 888 |
ret = io_wqe_cancel_cb_work(wqe, cancel, data); if (ret != IO_WQ_CANCEL_NOTFOUND) break; } return ret; } |
00bcda13d
|
889 890 891 892 |
struct work_match { bool (*fn)(struct io_wq_work *, void *data); void *data; }; |
771b53d03
|
893 894 |
static bool io_wq_worker_cancel(struct io_worker *worker, void *data) { |
00bcda13d
|
895 |
struct work_match *match = data; |
36c2f9223
|
896 897 |
unsigned long flags; bool ret = false; |
771b53d03
|
898 |
|
36c2f9223
|
899 |
spin_lock_irqsave(&worker->lock, flags); |
00bcda13d
|
900 |
if (match->fn(worker->cur_work, match->data) && |
0c9d5ccd2
|
901 |
!(worker->cur_work->flags & IO_WQ_WORK_NO_CANCEL)) { |
771b53d03
|
902 |
send_sig(SIGINT, worker->task, 1); |
36c2f9223
|
903 |
ret = true; |
771b53d03
|
904 |
} |
36c2f9223
|
905 |
spin_unlock_irqrestore(&worker->lock, flags); |
771b53d03
|
906 |
|
36c2f9223
|
907 |
return ret; |
771b53d03
|
908 909 910 |
} static enum io_wq_cancel io_wqe_cancel_work(struct io_wqe *wqe, |
00bcda13d
|
911 |
struct work_match *match) |
771b53d03
|
912 |
{ |
6206f0e18
|
913 |
struct io_wq_work_node *node, *prev; |
771b53d03
|
914 |
struct io_wq_work *work; |
6f72653e7
|
915 |
unsigned long flags; |
771b53d03
|
916 |
bool found = false; |
771b53d03
|
917 918 919 920 921 |
/* * 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. */ |
6f72653e7
|
922 |
spin_lock_irqsave(&wqe->lock, flags); |
6206f0e18
|
923 924 |
wq_list_for_each(node, prev, &wqe->work_list) { work = container_of(node, struct io_wq_work, list); |
00bcda13d
|
925 |
if (match->fn(work, match->data)) { |
6206f0e18
|
926 |
wq_node_del(&wqe->work_list, node, prev); |
771b53d03
|
927 928 929 930 |
found = true; break; } } |
6f72653e7
|
931 |
spin_unlock_irqrestore(&wqe->lock, flags); |
771b53d03
|
932 933 |
if (found) { |
fc04c39ba
|
934 |
io_run_cancel(work); |
771b53d03
|
935 936 937 938 939 940 |
return IO_WQ_CANCEL_OK; } /* * 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 |
d195a66e3
|
941 |
* as an indication that we attempt to signal cancellation. The |
771b53d03
|
942 943 944 |
* completion will run normally in this case. */ rcu_read_lock(); |
00bcda13d
|
945 |
found = io_wq_for_each_worker(wqe, io_wq_worker_cancel, match); |
771b53d03
|
946 947 948 |
rcu_read_unlock(); return found ? IO_WQ_CANCEL_RUNNING : IO_WQ_CANCEL_NOTFOUND; } |
00bcda13d
|
949 950 951 952 |
static bool io_wq_work_match(struct io_wq_work *work, void *data) { return work == data; } |
771b53d03
|
953 954 |
enum io_wq_cancel io_wq_cancel_work(struct io_wq *wq, struct io_wq_work *cwork) { |
00bcda13d
|
955 956 957 958 |
struct work_match match = { .fn = io_wq_work_match, .data = cwork }; |
771b53d03
|
959 |
enum io_wq_cancel ret = IO_WQ_CANCEL_NOTFOUND; |
3fc50ab55
|
960 |
int node; |
771b53d03
|
961 |
|
00bcda13d
|
962 |
cwork->flags |= IO_WQ_WORK_CANCEL; |
3fc50ab55
|
963 964 |
for_each_node(node) { struct io_wqe *wqe = wq->wqes[node]; |
771b53d03
|
965 |
|
00bcda13d
|
966 |
ret = io_wqe_cancel_work(wqe, &match); |
771b53d03
|
967 968 969 970 971 972 |
if (ret != IO_WQ_CANCEL_NOTFOUND) break; } return ret; } |
36282881a
|
973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 |
static bool io_wq_pid_match(struct io_wq_work *work, void *data) { pid_t pid = (pid_t) (unsigned long) data; if (work) return work->task_pid == pid; return false; } enum io_wq_cancel io_wq_cancel_pid(struct io_wq *wq, pid_t pid) { struct work_match match = { .fn = io_wq_pid_match, .data = (void *) (unsigned long) pid }; enum io_wq_cancel ret = IO_WQ_CANCEL_NOTFOUND; int node; for_each_node(node) { struct io_wqe *wqe = wq->wqes[node]; ret = io_wqe_cancel_work(wqe, &match); if (ret != IO_WQ_CANCEL_NOTFOUND) break; } return ret; } |
576a347b7
|
1001 |
struct io_wq *io_wq_create(unsigned bounded, struct io_wq_data *data) |
771b53d03
|
1002 |
{ |
3fc50ab55
|
1003 |
int ret = -ENOMEM, node; |
771b53d03
|
1004 |
struct io_wq *wq; |
ad6e005ca
|
1005 |
wq = kzalloc(sizeof(*wq), GFP_KERNEL); |
771b53d03
|
1006 1007 |
if (!wq) return ERR_PTR(-ENOMEM); |
3fc50ab55
|
1008 |
wq->wqes = kcalloc(nr_node_ids, sizeof(struct io_wqe *), GFP_KERNEL); |
771b53d03
|
1009 1010 1011 1012 |
if (!wq->wqes) { kfree(wq); return ERR_PTR(-ENOMEM); } |
576a347b7
|
1013 1014 |
wq->get_work = data->get_work; wq->put_work = data->put_work; |
7d7230652
|
1015 |
|
c5def4ab8
|
1016 |
/* caller must already hold a reference to this */ |
576a347b7
|
1017 |
wq->user = data->user; |
c5def4ab8
|
1018 |
|
3fc50ab55
|
1019 |
for_each_node(node) { |
771b53d03
|
1020 |
struct io_wqe *wqe; |
7563439ad
|
1021 |
int alloc_node = node; |
771b53d03
|
1022 |
|
7563439ad
|
1023 1024 1025 |
if (!node_online(alloc_node)) alloc_node = NUMA_NO_NODE; wqe = kzalloc_node(sizeof(struct io_wqe), GFP_KERNEL, alloc_node); |
771b53d03
|
1026 |
if (!wqe) |
3fc50ab55
|
1027 1028 |
goto err; wq->wqes[node] = wqe; |
7563439ad
|
1029 |
wqe->node = alloc_node; |
c5def4ab8
|
1030 1031 |
wqe->acct[IO_WQ_ACCT_BOUND].max_workers = bounded; atomic_set(&wqe->acct[IO_WQ_ACCT_BOUND].nr_running, 0); |
576a347b7
|
1032 |
if (wq->user) { |
c5def4ab8
|
1033 1034 1035 1036 |
wqe->acct[IO_WQ_ACCT_UNBOUND].max_workers = task_rlimit(current, RLIMIT_NPROC); } atomic_set(&wqe->acct[IO_WQ_ACCT_UNBOUND].nr_running, 0); |
771b53d03
|
1037 1038 |
wqe->wq = wq; spin_lock_init(&wqe->lock); |
6206f0e18
|
1039 |
INIT_WQ_LIST(&wqe->work_list); |
021d1cdda
|
1040 |
INIT_HLIST_NULLS_HEAD(&wqe->free_list, 0); |
e61df66c6
|
1041 |
INIT_LIST_HEAD(&wqe->all_list); |
771b53d03
|
1042 1043 1044 |
} init_completion(&wq->done); |
771b53d03
|
1045 1046 1047 |
wq->manager = kthread_create(io_wq_manager, wq, "io_wq_manager"); if (!IS_ERR(wq->manager)) { wake_up_process(wq->manager); |
b60fda600
|
1048 1049 1050 1051 1052 |
wait_for_completion(&wq->done); if (test_bit(IO_WQ_BIT_ERROR, &wq->state)) { ret = -ENOMEM; goto err; } |
848f7e188
|
1053 |
refcount_set(&wq->use_refs, 1); |
b60fda600
|
1054 |
reinit_completion(&wq->done); |
771b53d03
|
1055 1056 1057 1058 |
return wq; } ret = PTR_ERR(wq->manager); |
771b53d03
|
1059 |
complete(&wq->done); |
b60fda600
|
1060 |
err: |
3fc50ab55
|
1061 1062 |
for_each_node(node) kfree(wq->wqes[node]); |
b60fda600
|
1063 1064 |
kfree(wq->wqes); kfree(wq); |
771b53d03
|
1065 1066 |
return ERR_PTR(ret); } |
eba6f5a33
|
1067 1068 1069 1070 1071 1072 1073 |
bool io_wq_get(struct io_wq *wq, struct io_wq_data *data) { if (data->get_work != wq->get_work || data->put_work != wq->put_work) return false; return refcount_inc_not_zero(&wq->use_refs); } |
771b53d03
|
1074 1075 1076 1077 1078 |
static bool io_wq_worker_wake(struct io_worker *worker, void *data) { wake_up_process(worker->task); return false; } |
848f7e188
|
1079 |
static void __io_wq_destroy(struct io_wq *wq) |
771b53d03
|
1080 |
{ |
3fc50ab55
|
1081 |
int node; |
771b53d03
|
1082 |
|
b60fda600
|
1083 1084 |
set_bit(IO_WQ_BIT_EXIT, &wq->state); if (wq->manager) |
771b53d03
|
1085 |
kthread_stop(wq->manager); |
771b53d03
|
1086 1087 |
rcu_read_lock(); |
3fc50ab55
|
1088 1089 |
for_each_node(node) io_wq_for_each_worker(wq->wqes[node], io_wq_worker_wake, NULL); |
771b53d03
|
1090 1091 1092 |
rcu_read_unlock(); wait_for_completion(&wq->done); |
3fc50ab55
|
1093 1094 |
for_each_node(node) kfree(wq->wqes[node]); |
771b53d03
|
1095 1096 1097 |
kfree(wq->wqes); kfree(wq); } |
848f7e188
|
1098 1099 1100 1101 1102 1103 |
void io_wq_destroy(struct io_wq *wq) { if (refcount_dec_and_test(&wq->use_refs)) __io_wq_destroy(wq); } |