Line data Source code
1 : // SPDX-License-Identifier: GPL-2.0
2 : /*
3 : * Copyright (C) 1991, 1992 Linus Torvalds
4 : * Copyright (C) 1994, Karl Keyte: Added support for disk statistics
5 : * Elevator latency, (C) 2000 Andrea Arcangeli <andrea@suse.de> SuSE
6 : * Queue request tables / lock, selectable elevator, Jens Axboe <axboe@suse.de>
7 : * kernel-doc documentation started by NeilBrown <neilb@cse.unsw.edu.au>
8 : * - July2000
9 : * bio rewrite, highmem i/o, etc, Jens Axboe <axboe@suse.de> - may 2001
10 : */
11 :
12 : /*
13 : * This handles all read/write requests to block devices
14 : */
15 : #include <linux/kernel.h>
16 : #include <linux/module.h>
17 : #include <linux/bio.h>
18 : #include <linux/blkdev.h>
19 : #include <linux/blk-pm.h>
20 : #include <linux/blk-integrity.h>
21 : #include <linux/highmem.h>
22 : #include <linux/mm.h>
23 : #include <linux/pagemap.h>
24 : #include <linux/kernel_stat.h>
25 : #include <linux/string.h>
26 : #include <linux/init.h>
27 : #include <linux/completion.h>
28 : #include <linux/slab.h>
29 : #include <linux/swap.h>
30 : #include <linux/writeback.h>
31 : #include <linux/task_io_accounting_ops.h>
32 : #include <linux/fault-inject.h>
33 : #include <linux/list_sort.h>
34 : #include <linux/delay.h>
35 : #include <linux/ratelimit.h>
36 : #include <linux/pm_runtime.h>
37 : #include <linux/t10-pi.h>
38 : #include <linux/debugfs.h>
39 : #include <linux/bpf.h>
40 : #include <linux/part_stat.h>
41 : #include <linux/sched/sysctl.h>
42 : #include <linux/blk-crypto.h>
43 :
44 : #define CREATE_TRACE_POINTS
45 : #include <trace/events/block.h>
46 :
47 : #include "blk.h"
48 : #include "blk-mq-sched.h"
49 : #include "blk-pm.h"
50 : #include "blk-cgroup.h"
51 : #include "blk-throttle.h"
52 :
53 : struct dentry *blk_debugfs_root;
54 :
55 : EXPORT_TRACEPOINT_SYMBOL_GPL(block_bio_remap);
56 : EXPORT_TRACEPOINT_SYMBOL_GPL(block_rq_remap);
57 : EXPORT_TRACEPOINT_SYMBOL_GPL(block_bio_complete);
58 : EXPORT_TRACEPOINT_SYMBOL_GPL(block_split);
59 : EXPORT_TRACEPOINT_SYMBOL_GPL(block_unplug);
60 : EXPORT_TRACEPOINT_SYMBOL_GPL(block_rq_insert);
61 :
62 : static DEFINE_IDA(blk_queue_ida);
63 :
64 : /*
65 : * For queue allocation
66 : */
67 : static struct kmem_cache *blk_requestq_cachep;
68 :
69 : /*
70 : * Controlling structure to kblockd
71 : */
72 : static struct workqueue_struct *kblockd_workqueue;
73 :
74 : /**
75 : * blk_queue_flag_set - atomically set a queue flag
76 : * @flag: flag to be set
77 : * @q: request queue
78 : */
79 0 : void blk_queue_flag_set(unsigned int flag, struct request_queue *q)
80 : {
81 0 : set_bit(flag, &q->queue_flags);
82 0 : }
83 : EXPORT_SYMBOL(blk_queue_flag_set);
84 :
85 : /**
86 : * blk_queue_flag_clear - atomically clear a queue flag
87 : * @flag: flag to be cleared
88 : * @q: request queue
89 : */
90 0 : void blk_queue_flag_clear(unsigned int flag, struct request_queue *q)
91 : {
92 0 : clear_bit(flag, &q->queue_flags);
93 0 : }
94 : EXPORT_SYMBOL(blk_queue_flag_clear);
95 :
96 : /**
97 : * blk_queue_flag_test_and_set - atomically test and set a queue flag
98 : * @flag: flag to be set
99 : * @q: request queue
100 : *
101 : * Returns the previous value of @flag - 0 if the flag was not set and 1 if
102 : * the flag was already set.
103 : */
104 0 : bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q)
105 : {
106 0 : return test_and_set_bit(flag, &q->queue_flags);
107 : }
108 : EXPORT_SYMBOL_GPL(blk_queue_flag_test_and_set);
109 :
110 : #define REQ_OP_NAME(name) [REQ_OP_##name] = #name
111 : static const char *const blk_op_name[] = {
112 : REQ_OP_NAME(READ),
113 : REQ_OP_NAME(WRITE),
114 : REQ_OP_NAME(FLUSH),
115 : REQ_OP_NAME(DISCARD),
116 : REQ_OP_NAME(SECURE_ERASE),
117 : REQ_OP_NAME(ZONE_RESET),
118 : REQ_OP_NAME(ZONE_RESET_ALL),
119 : REQ_OP_NAME(ZONE_OPEN),
120 : REQ_OP_NAME(ZONE_CLOSE),
121 : REQ_OP_NAME(ZONE_FINISH),
122 : REQ_OP_NAME(ZONE_APPEND),
123 : REQ_OP_NAME(WRITE_ZEROES),
124 : REQ_OP_NAME(DRV_IN),
125 : REQ_OP_NAME(DRV_OUT),
126 : };
127 : #undef REQ_OP_NAME
128 :
129 : /**
130 : * blk_op_str - Return string XXX in the REQ_OP_XXX.
131 : * @op: REQ_OP_XXX.
132 : *
133 : * Description: Centralize block layer function to convert REQ_OP_XXX into
134 : * string format. Useful in the debugging and tracing bio or request. For
135 : * invalid REQ_OP_XXX it returns string "UNKNOWN".
136 : */
137 0 : inline const char *blk_op_str(enum req_op op)
138 : {
139 0 : const char *op_str = "UNKNOWN";
140 :
141 0 : if (op < ARRAY_SIZE(blk_op_name) && blk_op_name[op])
142 0 : op_str = blk_op_name[op];
143 :
144 0 : return op_str;
145 : }
146 : EXPORT_SYMBOL_GPL(blk_op_str);
147 :
148 : static const struct {
149 : int errno;
150 : const char *name;
151 : } blk_errors[] = {
152 : [BLK_STS_OK] = { 0, "" },
153 : [BLK_STS_NOTSUPP] = { -EOPNOTSUPP, "operation not supported" },
154 : [BLK_STS_TIMEOUT] = { -ETIMEDOUT, "timeout" },
155 : [BLK_STS_NOSPC] = { -ENOSPC, "critical space allocation" },
156 : [BLK_STS_TRANSPORT] = { -ENOLINK, "recoverable transport" },
157 : [BLK_STS_TARGET] = { -EREMOTEIO, "critical target" },
158 : [BLK_STS_NEXUS] = { -EBADE, "critical nexus" },
159 : [BLK_STS_MEDIUM] = { -ENODATA, "critical medium" },
160 : [BLK_STS_PROTECTION] = { -EILSEQ, "protection" },
161 : [BLK_STS_RESOURCE] = { -ENOMEM, "kernel resource" },
162 : [BLK_STS_DEV_RESOURCE] = { -EBUSY, "device resource" },
163 : [BLK_STS_AGAIN] = { -EAGAIN, "nonblocking retry" },
164 : [BLK_STS_OFFLINE] = { -ENODEV, "device offline" },
165 :
166 : /* device mapper special case, should not leak out: */
167 : [BLK_STS_DM_REQUEUE] = { -EREMCHG, "dm internal retry" },
168 :
169 : /* zone device specific errors */
170 : [BLK_STS_ZONE_OPEN_RESOURCE] = { -ETOOMANYREFS, "open zones exceeded" },
171 : [BLK_STS_ZONE_ACTIVE_RESOURCE] = { -EOVERFLOW, "active zones exceeded" },
172 :
173 : /* everything else not covered above: */
174 : [BLK_STS_IOERR] = { -EIO, "I/O" },
175 : };
176 :
177 0 : blk_status_t errno_to_blk_status(int errno)
178 : {
179 : int i;
180 :
181 0 : for (i = 0; i < ARRAY_SIZE(blk_errors); i++) {
182 0 : if (blk_errors[i].errno == errno)
183 0 : return (__force blk_status_t)i;
184 : }
185 :
186 : return BLK_STS_IOERR;
187 : }
188 : EXPORT_SYMBOL_GPL(errno_to_blk_status);
189 :
190 0 : int blk_status_to_errno(blk_status_t status)
191 : {
192 0 : int idx = (__force int)status;
193 :
194 0 : if (WARN_ON_ONCE(idx >= ARRAY_SIZE(blk_errors)))
195 : return -EIO;
196 0 : return blk_errors[idx].errno;
197 : }
198 : EXPORT_SYMBOL_GPL(blk_status_to_errno);
199 :
200 0 : const char *blk_status_to_str(blk_status_t status)
201 : {
202 0 : int idx = (__force int)status;
203 :
204 0 : if (WARN_ON_ONCE(idx >= ARRAY_SIZE(blk_errors)))
205 : return "<null>";
206 0 : return blk_errors[idx].name;
207 : }
208 :
209 : /**
210 : * blk_sync_queue - cancel any pending callbacks on a queue
211 : * @q: the queue
212 : *
213 : * Description:
214 : * The block layer may perform asynchronous callback activity
215 : * on a queue, such as calling the unplug function after a timeout.
216 : * A block device may call blk_sync_queue to ensure that any
217 : * such activity is cancelled, thus allowing it to release resources
218 : * that the callbacks might use. The caller must already have made sure
219 : * that its ->submit_bio will not re-add plugging prior to calling
220 : * this function.
221 : *
222 : * This function does not cancel any asynchronous activity arising
223 : * out of elevator or throttling code. That would require elevator_exit()
224 : * and blkcg_exit_queue() to be called with queue lock initialized.
225 : *
226 : */
227 0 : void blk_sync_queue(struct request_queue *q)
228 : {
229 0 : del_timer_sync(&q->timeout);
230 0 : cancel_work_sync(&q->timeout_work);
231 0 : }
232 : EXPORT_SYMBOL(blk_sync_queue);
233 :
234 : /**
235 : * blk_set_pm_only - increment pm_only counter
236 : * @q: request queue pointer
237 : */
238 0 : void blk_set_pm_only(struct request_queue *q)
239 : {
240 0 : atomic_inc(&q->pm_only);
241 0 : }
242 : EXPORT_SYMBOL_GPL(blk_set_pm_only);
243 :
244 0 : void blk_clear_pm_only(struct request_queue *q)
245 : {
246 : int pm_only;
247 :
248 0 : pm_only = atomic_dec_return(&q->pm_only);
249 0 : WARN_ON_ONCE(pm_only < 0);
250 0 : if (pm_only == 0)
251 0 : wake_up_all(&q->mq_freeze_wq);
252 0 : }
253 : EXPORT_SYMBOL_GPL(blk_clear_pm_only);
254 :
255 0 : static void blk_free_queue_rcu(struct rcu_head *rcu_head)
256 : {
257 0 : struct request_queue *q = container_of(rcu_head,
258 : struct request_queue, rcu_head);
259 :
260 0 : percpu_ref_exit(&q->q_usage_counter);
261 0 : kmem_cache_free(blk_requestq_cachep, q);
262 0 : }
263 :
264 0 : static void blk_free_queue(struct request_queue *q)
265 : {
266 0 : if (q->poll_stat)
267 0 : blk_stat_remove_callback(q, q->poll_cb);
268 0 : blk_stat_free_callback(q->poll_cb);
269 :
270 0 : blk_free_queue_stats(q->stats);
271 0 : kfree(q->poll_stat);
272 :
273 0 : if (queue_is_mq(q))
274 0 : blk_mq_release(q);
275 :
276 0 : ida_free(&blk_queue_ida, q->id);
277 0 : call_rcu(&q->rcu_head, blk_free_queue_rcu);
278 0 : }
279 :
280 : /**
281 : * blk_put_queue - decrement the request_queue refcount
282 : * @q: the request_queue structure to decrement the refcount for
283 : *
284 : * Decrements the refcount of the request_queue and free it when the refcount
285 : * reaches 0.
286 : */
287 0 : void blk_put_queue(struct request_queue *q)
288 : {
289 0 : if (refcount_dec_and_test(&q->refs))
290 0 : blk_free_queue(q);
291 0 : }
292 : EXPORT_SYMBOL(blk_put_queue);
293 :
294 0 : void blk_queue_start_drain(struct request_queue *q)
295 : {
296 : /*
297 : * When queue DYING flag is set, we need to block new req
298 : * entering queue, so we call blk_freeze_queue_start() to
299 : * prevent I/O from crossing blk_queue_enter().
300 : */
301 0 : blk_freeze_queue_start(q);
302 0 : if (queue_is_mq(q))
303 0 : blk_mq_wake_waiters(q);
304 : /* Make blk_queue_enter() reexamine the DYING flag. */
305 0 : wake_up_all(&q->mq_freeze_wq);
306 0 : }
307 :
308 : /**
309 : * blk_queue_enter() - try to increase q->q_usage_counter
310 : * @q: request queue pointer
311 : * @flags: BLK_MQ_REQ_NOWAIT and/or BLK_MQ_REQ_PM
312 : */
313 0 : int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags)
314 : {
315 0 : const bool pm = flags & BLK_MQ_REQ_PM;
316 :
317 0 : while (!blk_try_enter_queue(q, pm)) {
318 0 : if (flags & BLK_MQ_REQ_NOWAIT)
319 : return -EAGAIN;
320 :
321 : /*
322 : * read pair of barrier in blk_freeze_queue_start(), we need to
323 : * order reading __PERCPU_REF_DEAD flag of .q_usage_counter and
324 : * reading .mq_freeze_depth or queue dying flag, otherwise the
325 : * following wait may never return if the two reads are
326 : * reordered.
327 : */
328 0 : smp_rmb();
329 0 : wait_event(q->mq_freeze_wq,
330 : (!q->mq_freeze_depth &&
331 : blk_pm_resume_queue(pm, q)) ||
332 : blk_queue_dying(q));
333 0 : if (blk_queue_dying(q))
334 : return -ENODEV;
335 : }
336 :
337 : return 0;
338 : }
339 :
340 0 : int __bio_queue_enter(struct request_queue *q, struct bio *bio)
341 : {
342 0 : while (!blk_try_enter_queue(q, false)) {
343 0 : struct gendisk *disk = bio->bi_bdev->bd_disk;
344 :
345 0 : if (bio->bi_opf & REQ_NOWAIT) {
346 0 : if (test_bit(GD_DEAD, &disk->state))
347 : goto dead;
348 0 : bio_wouldblock_error(bio);
349 0 : return -EAGAIN;
350 : }
351 :
352 : /*
353 : * read pair of barrier in blk_freeze_queue_start(), we need to
354 : * order reading __PERCPU_REF_DEAD flag of .q_usage_counter and
355 : * reading .mq_freeze_depth or queue dying flag, otherwise the
356 : * following wait may never return if the two reads are
357 : * reordered.
358 : */
359 0 : smp_rmb();
360 0 : wait_event(q->mq_freeze_wq,
361 : (!q->mq_freeze_depth &&
362 : blk_pm_resume_queue(false, q)) ||
363 : test_bit(GD_DEAD, &disk->state));
364 0 : if (test_bit(GD_DEAD, &disk->state))
365 : goto dead;
366 : }
367 :
368 : return 0;
369 : dead:
370 0 : bio_io_error(bio);
371 0 : return -ENODEV;
372 : }
373 :
374 0 : void blk_queue_exit(struct request_queue *q)
375 : {
376 0 : percpu_ref_put(&q->q_usage_counter);
377 0 : }
378 :
379 0 : static void blk_queue_usage_counter_release(struct percpu_ref *ref)
380 : {
381 0 : struct request_queue *q =
382 0 : container_of(ref, struct request_queue, q_usage_counter);
383 :
384 0 : wake_up_all(&q->mq_freeze_wq);
385 0 : }
386 :
387 0 : static void blk_rq_timed_out_timer(struct timer_list *t)
388 : {
389 0 : struct request_queue *q = from_timer(q, t, timeout);
390 :
391 0 : kblockd_schedule_work(&q->timeout_work);
392 0 : }
393 :
394 0 : static void blk_timeout_work(struct work_struct *work)
395 : {
396 0 : }
397 :
398 0 : struct request_queue *blk_alloc_queue(int node_id)
399 : {
400 : struct request_queue *q;
401 :
402 0 : q = kmem_cache_alloc_node(blk_requestq_cachep, GFP_KERNEL | __GFP_ZERO,
403 : node_id);
404 0 : if (!q)
405 : return NULL;
406 :
407 0 : q->last_merge = NULL;
408 :
409 0 : q->id = ida_alloc(&blk_queue_ida, GFP_KERNEL);
410 0 : if (q->id < 0)
411 : goto fail_q;
412 :
413 0 : q->stats = blk_alloc_queue_stats();
414 0 : if (!q->stats)
415 : goto fail_id;
416 :
417 0 : q->node = node_id;
418 :
419 0 : atomic_set(&q->nr_active_requests_shared_tags, 0);
420 :
421 0 : timer_setup(&q->timeout, blk_rq_timed_out_timer, 0);
422 0 : INIT_WORK(&q->timeout_work, blk_timeout_work);
423 0 : INIT_LIST_HEAD(&q->icq_list);
424 :
425 0 : refcount_set(&q->refs, 1);
426 0 : mutex_init(&q->debugfs_mutex);
427 0 : mutex_init(&q->sysfs_lock);
428 0 : mutex_init(&q->sysfs_dir_lock);
429 0 : spin_lock_init(&q->queue_lock);
430 :
431 0 : init_waitqueue_head(&q->mq_freeze_wq);
432 0 : mutex_init(&q->mq_freeze_lock);
433 :
434 : /*
435 : * Init percpu_ref in atomic mode so that it's faster to shutdown.
436 : * See blk_register_queue() for details.
437 : */
438 0 : if (percpu_ref_init(&q->q_usage_counter,
439 : blk_queue_usage_counter_release,
440 : PERCPU_REF_INIT_ATOMIC, GFP_KERNEL))
441 : goto fail_stats;
442 :
443 0 : blk_set_default_limits(&q->limits);
444 0 : q->nr_requests = BLKDEV_DEFAULT_RQ;
445 :
446 0 : return q;
447 :
448 : fail_stats:
449 0 : blk_free_queue_stats(q->stats);
450 : fail_id:
451 0 : ida_free(&blk_queue_ida, q->id);
452 : fail_q:
453 0 : kmem_cache_free(blk_requestq_cachep, q);
454 0 : return NULL;
455 : }
456 :
457 : /**
458 : * blk_get_queue - increment the request_queue refcount
459 : * @q: the request_queue structure to increment the refcount for
460 : *
461 : * Increment the refcount of the request_queue kobject.
462 : *
463 : * Context: Any context.
464 : */
465 0 : bool blk_get_queue(struct request_queue *q)
466 : {
467 0 : if (unlikely(blk_queue_dying(q)))
468 : return false;
469 0 : refcount_inc(&q->refs);
470 0 : return true;
471 : }
472 : EXPORT_SYMBOL(blk_get_queue);
473 :
474 : #ifdef CONFIG_FAIL_MAKE_REQUEST
475 :
476 : static DECLARE_FAULT_ATTR(fail_make_request);
477 :
478 : static int __init setup_fail_make_request(char *str)
479 : {
480 : return setup_fault_attr(&fail_make_request, str);
481 : }
482 : __setup("fail_make_request=", setup_fail_make_request);
483 :
484 : bool should_fail_request(struct block_device *part, unsigned int bytes)
485 : {
486 : return part->bd_make_it_fail && should_fail(&fail_make_request, bytes);
487 : }
488 :
489 : static int __init fail_make_request_debugfs(void)
490 : {
491 : struct dentry *dir = fault_create_debugfs_attr("fail_make_request",
492 : NULL, &fail_make_request);
493 :
494 : return PTR_ERR_OR_ZERO(dir);
495 : }
496 :
497 : late_initcall(fail_make_request_debugfs);
498 : #endif /* CONFIG_FAIL_MAKE_REQUEST */
499 :
500 0 : static inline void bio_check_ro(struct bio *bio)
501 : {
502 0 : if (op_is_write(bio_op(bio)) && bdev_read_only(bio->bi_bdev)) {
503 0 : if (op_is_flush(bio->bi_opf) && !bio_sectors(bio))
504 : return;
505 0 : pr_warn("Trying to write to read-only block-device %pg\n",
506 : bio->bi_bdev);
507 : /* Older lvm-tools actually trigger this */
508 : }
509 : }
510 :
511 0 : static noinline int should_fail_bio(struct bio *bio)
512 : {
513 0 : if (should_fail_request(bdev_whole(bio->bi_bdev), bio->bi_iter.bi_size))
514 : return -EIO;
515 : return 0;
516 : }
517 : ALLOW_ERROR_INJECTION(should_fail_bio, ERRNO);
518 :
519 : /*
520 : * Check whether this bio extends beyond the end of the device or partition.
521 : * This may well happen - the kernel calls bread() without checking the size of
522 : * the device, e.g., when mounting a file system.
523 : */
524 0 : static inline int bio_check_eod(struct bio *bio)
525 : {
526 0 : sector_t maxsector = bdev_nr_sectors(bio->bi_bdev);
527 0 : unsigned int nr_sectors = bio_sectors(bio);
528 :
529 0 : if (nr_sectors && maxsector &&
530 0 : (nr_sectors > maxsector ||
531 0 : bio->bi_iter.bi_sector > maxsector - nr_sectors)) {
532 0 : pr_info_ratelimited("%s: attempt to access beyond end of device\n"
533 : "%pg: rw=%d, sector=%llu, nr_sectors = %u limit=%llu\n",
534 : current->comm, bio->bi_bdev, bio->bi_opf,
535 : bio->bi_iter.bi_sector, nr_sectors, maxsector);
536 : return -EIO;
537 : }
538 : return 0;
539 : }
540 :
541 : /*
542 : * Remap block n of partition p to block n+start(p) of the disk.
543 : */
544 : static int blk_partition_remap(struct bio *bio)
545 : {
546 0 : struct block_device *p = bio->bi_bdev;
547 :
548 0 : if (unlikely(should_fail_request(p, bio->bi_iter.bi_size)))
549 : return -EIO;
550 0 : if (bio_sectors(bio)) {
551 0 : bio->bi_iter.bi_sector += p->bd_start_sect;
552 0 : trace_block_bio_remap(bio, p->bd_dev,
553 : bio->bi_iter.bi_sector -
554 0 : p->bd_start_sect);
555 : }
556 0 : bio_set_flag(bio, BIO_REMAPPED);
557 : return 0;
558 : }
559 :
560 : /*
561 : * Check write append to a zoned block device.
562 : */
563 : static inline blk_status_t blk_check_zone_append(struct request_queue *q,
564 : struct bio *bio)
565 : {
566 : int nr_sectors = bio_sectors(bio);
567 :
568 : /* Only applicable to zoned block devices */
569 : if (!bdev_is_zoned(bio->bi_bdev))
570 : return BLK_STS_NOTSUPP;
571 :
572 : /* The bio sector must point to the start of a sequential zone */
573 : if (!bdev_is_zone_start(bio->bi_bdev, bio->bi_iter.bi_sector) ||
574 : !bio_zone_is_seq(bio))
575 : return BLK_STS_IOERR;
576 :
577 : /*
578 : * Not allowed to cross zone boundaries. Otherwise, the BIO will be
579 : * split and could result in non-contiguous sectors being written in
580 : * different zones.
581 : */
582 : if (nr_sectors > q->limits.chunk_sectors)
583 : return BLK_STS_IOERR;
584 :
585 : /* Make sure the BIO is small enough and will not get split */
586 : if (nr_sectors > q->limits.max_zone_append_sectors)
587 : return BLK_STS_IOERR;
588 :
589 : bio->bi_opf |= REQ_NOMERGE;
590 :
591 : return BLK_STS_OK;
592 : }
593 :
594 0 : static void __submit_bio(struct bio *bio)
595 : {
596 0 : struct gendisk *disk = bio->bi_bdev->bd_disk;
597 :
598 0 : if (unlikely(!blk_crypto_bio_prep(&bio)))
599 : return;
600 :
601 0 : if (!disk->fops->submit_bio) {
602 0 : blk_mq_submit_bio(bio);
603 0 : } else if (likely(bio_queue_enter(bio) == 0)) {
604 0 : disk->fops->submit_bio(bio);
605 0 : blk_queue_exit(disk->queue);
606 : }
607 : }
608 :
609 : /*
610 : * The loop in this function may be a bit non-obvious, and so deserves some
611 : * explanation:
612 : *
613 : * - Before entering the loop, bio->bi_next is NULL (as all callers ensure
614 : * that), so we have a list with a single bio.
615 : * - We pretend that we have just taken it off a longer list, so we assign
616 : * bio_list to a pointer to the bio_list_on_stack, thus initialising the
617 : * bio_list of new bios to be added. ->submit_bio() may indeed add some more
618 : * bios through a recursive call to submit_bio_noacct. If it did, we find a
619 : * non-NULL value in bio_list and re-enter the loop from the top.
620 : * - In this case we really did just take the bio of the top of the list (no
621 : * pretending) and so remove it from bio_list, and call into ->submit_bio()
622 : * again.
623 : *
624 : * bio_list_on_stack[0] contains bios submitted by the current ->submit_bio.
625 : * bio_list_on_stack[1] contains bios that were submitted before the current
626 : * ->submit_bio, but that haven't been processed yet.
627 : */
628 0 : static void __submit_bio_noacct(struct bio *bio)
629 : {
630 : struct bio_list bio_list_on_stack[2];
631 :
632 0 : BUG_ON(bio->bi_next);
633 :
634 0 : bio_list_init(&bio_list_on_stack[0]);
635 0 : current->bio_list = bio_list_on_stack;
636 :
637 : do {
638 0 : struct request_queue *q = bdev_get_queue(bio->bi_bdev);
639 : struct bio_list lower, same;
640 :
641 : /*
642 : * Create a fresh bio_list for all subordinate requests.
643 : */
644 0 : bio_list_on_stack[1] = bio_list_on_stack[0];
645 0 : bio_list_init(&bio_list_on_stack[0]);
646 :
647 0 : __submit_bio(bio);
648 :
649 : /*
650 : * Sort new bios into those for a lower level and those for the
651 : * same level.
652 : */
653 0 : bio_list_init(&lower);
654 : bio_list_init(&same);
655 0 : while ((bio = bio_list_pop(&bio_list_on_stack[0])) != NULL)
656 0 : if (q == bdev_get_queue(bio->bi_bdev))
657 : bio_list_add(&same, bio);
658 : else
659 : bio_list_add(&lower, bio);
660 :
661 : /*
662 : * Now assemble so we handle the lowest level first.
663 : */
664 0 : bio_list_merge(&bio_list_on_stack[0], &lower);
665 0 : bio_list_merge(&bio_list_on_stack[0], &same);
666 0 : bio_list_merge(&bio_list_on_stack[0], &bio_list_on_stack[1]);
667 0 : } while ((bio = bio_list_pop(&bio_list_on_stack[0])));
668 :
669 0 : current->bio_list = NULL;
670 0 : }
671 :
672 0 : static void __submit_bio_noacct_mq(struct bio *bio)
673 : {
674 0 : struct bio_list bio_list[2] = { };
675 :
676 0 : current->bio_list = bio_list;
677 :
678 : do {
679 0 : __submit_bio(bio);
680 0 : } while ((bio = bio_list_pop(&bio_list[0])));
681 :
682 0 : current->bio_list = NULL;
683 0 : }
684 :
685 0 : void submit_bio_noacct_nocheck(struct bio *bio)
686 : {
687 0 : blk_cgroup_bio_start(bio);
688 0 : blkcg_bio_issue_init(bio);
689 :
690 0 : if (!bio_flagged(bio, BIO_TRACE_COMPLETION)) {
691 0 : trace_block_bio_queue(bio);
692 : /*
693 : * Now that enqueuing has been traced, we need to trace
694 : * completion as well.
695 : */
696 : bio_set_flag(bio, BIO_TRACE_COMPLETION);
697 : }
698 :
699 : /*
700 : * We only want one ->submit_bio to be active at a time, else stack
701 : * usage with stacked devices could be a problem. Use current->bio_list
702 : * to collect a list of requests submited by a ->submit_bio method while
703 : * it is active, and then process them after it returned.
704 : */
705 0 : if (current->bio_list)
706 0 : bio_list_add(¤t->bio_list[0], bio);
707 0 : else if (!bio->bi_bdev->bd_disk->fops->submit_bio)
708 0 : __submit_bio_noacct_mq(bio);
709 : else
710 0 : __submit_bio_noacct(bio);
711 0 : }
712 :
713 : /**
714 : * submit_bio_noacct - re-submit a bio to the block device layer for I/O
715 : * @bio: The bio describing the location in memory and on the device.
716 : *
717 : * This is a version of submit_bio() that shall only be used for I/O that is
718 : * resubmitted to lower level drivers by stacking block drivers. All file
719 : * systems and other upper level users of the block layer should use
720 : * submit_bio() instead.
721 : */
722 0 : void submit_bio_noacct(struct bio *bio)
723 : {
724 0 : struct block_device *bdev = bio->bi_bdev;
725 0 : struct request_queue *q = bdev_get_queue(bdev);
726 0 : blk_status_t status = BLK_STS_IOERR;
727 : struct blk_plug *plug;
728 :
729 : might_sleep();
730 :
731 0 : plug = blk_mq_plug(bio);
732 0 : if (plug && plug->nowait)
733 0 : bio->bi_opf |= REQ_NOWAIT;
734 :
735 : /*
736 : * For a REQ_NOWAIT based request, return -EOPNOTSUPP
737 : * if queue does not support NOWAIT.
738 : */
739 0 : if ((bio->bi_opf & REQ_NOWAIT) && !bdev_nowait(bdev))
740 : goto not_supported;
741 :
742 0 : if (should_fail_bio(bio))
743 : goto end_io;
744 0 : bio_check_ro(bio);
745 0 : if (!bio_flagged(bio, BIO_REMAPPED)) {
746 0 : if (unlikely(bio_check_eod(bio)))
747 : goto end_io;
748 0 : if (bdev->bd_partno && unlikely(blk_partition_remap(bio)))
749 : goto end_io;
750 : }
751 :
752 : /*
753 : * Filter flush bio's early so that bio based drivers without flush
754 : * support don't have to worry about them.
755 : */
756 0 : if (op_is_flush(bio->bi_opf)) {
757 0 : if (WARN_ON_ONCE(bio_op(bio) != REQ_OP_WRITE &&
758 : bio_op(bio) != REQ_OP_ZONE_APPEND))
759 : goto end_io;
760 0 : if (!test_bit(QUEUE_FLAG_WC, &q->queue_flags)) {
761 0 : bio->bi_opf &= ~(REQ_PREFLUSH | REQ_FUA);
762 0 : if (!bio_sectors(bio)) {
763 : status = BLK_STS_OK;
764 : goto end_io;
765 : }
766 : }
767 : }
768 :
769 0 : if (!test_bit(QUEUE_FLAG_POLL, &q->queue_flags))
770 : bio_clear_polled(bio);
771 :
772 0 : switch (bio_op(bio)) {
773 : case REQ_OP_DISCARD:
774 0 : if (!bdev_max_discard_sectors(bdev))
775 : goto not_supported;
776 : break;
777 : case REQ_OP_SECURE_ERASE:
778 0 : if (!bdev_max_secure_erase_sectors(bdev))
779 : goto not_supported;
780 : break;
781 : case REQ_OP_ZONE_APPEND:
782 : status = blk_check_zone_append(q, bio);
783 : if (status != BLK_STS_OK)
784 : goto end_io;
785 : break;
786 : case REQ_OP_ZONE_RESET:
787 : case REQ_OP_ZONE_OPEN:
788 : case REQ_OP_ZONE_CLOSE:
789 : case REQ_OP_ZONE_FINISH:
790 : if (!bdev_is_zoned(bio->bi_bdev))
791 : goto not_supported;
792 : break;
793 : case REQ_OP_ZONE_RESET_ALL:
794 : if (!bdev_is_zoned(bio->bi_bdev) || !blk_queue_zone_resetall(q))
795 : goto not_supported;
796 : break;
797 : case REQ_OP_WRITE_ZEROES:
798 0 : if (!q->limits.max_write_zeroes_sectors)
799 : goto not_supported;
800 : break;
801 : default:
802 : break;
803 : }
804 :
805 0 : if (blk_throtl_bio(bio))
806 : return;
807 0 : submit_bio_noacct_nocheck(bio);
808 0 : return;
809 :
810 : not_supported:
811 : status = BLK_STS_NOTSUPP;
812 : end_io:
813 0 : bio->bi_status = status;
814 0 : bio_endio(bio);
815 : }
816 : EXPORT_SYMBOL(submit_bio_noacct);
817 :
818 : /**
819 : * submit_bio - submit a bio to the block device layer for I/O
820 : * @bio: The &struct bio which describes the I/O
821 : *
822 : * submit_bio() is used to submit I/O requests to block devices. It is passed a
823 : * fully set up &struct bio that describes the I/O that needs to be done. The
824 : * bio will be send to the device described by the bi_bdev field.
825 : *
826 : * The success/failure status of the request, along with notification of
827 : * completion, is delivered asynchronously through the ->bi_end_io() callback
828 : * in @bio. The bio must NOT be touched by the caller until ->bi_end_io() has
829 : * been called.
830 : */
831 0 : void submit_bio(struct bio *bio)
832 : {
833 0 : if (blkcg_punt_bio_submit(bio))
834 : return;
835 :
836 0 : if (bio_op(bio) == REQ_OP_READ) {
837 0 : task_io_account_read(bio->bi_iter.bi_size);
838 0 : count_vm_events(PGPGIN, bio_sectors(bio));
839 0 : } else if (bio_op(bio) == REQ_OP_WRITE) {
840 0 : count_vm_events(PGPGOUT, bio_sectors(bio));
841 : }
842 :
843 0 : submit_bio_noacct(bio);
844 : }
845 : EXPORT_SYMBOL(submit_bio);
846 :
847 : /**
848 : * bio_poll - poll for BIO completions
849 : * @bio: bio to poll for
850 : * @iob: batches of IO
851 : * @flags: BLK_POLL_* flags that control the behavior
852 : *
853 : * Poll for completions on queue associated with the bio. Returns number of
854 : * completed entries found.
855 : *
856 : * Note: the caller must either be the context that submitted @bio, or
857 : * be in a RCU critical section to prevent freeing of @bio.
858 : */
859 0 : int bio_poll(struct bio *bio, struct io_comp_batch *iob, unsigned int flags)
860 : {
861 0 : blk_qc_t cookie = READ_ONCE(bio->bi_cookie);
862 : struct block_device *bdev;
863 : struct request_queue *q;
864 0 : int ret = 0;
865 :
866 0 : bdev = READ_ONCE(bio->bi_bdev);
867 0 : if (!bdev)
868 : return 0;
869 :
870 0 : q = bdev_get_queue(bdev);
871 0 : if (cookie == BLK_QC_T_NONE ||
872 0 : !test_bit(QUEUE_FLAG_POLL, &q->queue_flags))
873 : return 0;
874 :
875 : /*
876 : * As the requests that require a zone lock are not plugged in the
877 : * first place, directly accessing the plug instead of using
878 : * blk_mq_plug() should not have any consequences during flushing for
879 : * zoned devices.
880 : */
881 0 : blk_flush_plug(current->plug, false);
882 :
883 : /*
884 : * We need to be able to enter a frozen queue, similar to how
885 : * timeouts also need to do that. If that is blocked, then we can
886 : * have pending IO when a queue freeze is started, and then the
887 : * wait for the freeze to finish will wait for polled requests to
888 : * timeout as the poller is preventer from entering the queue and
889 : * completing them. As long as we prevent new IO from being queued,
890 : * that should be all that matters.
891 : */
892 0 : if (!percpu_ref_tryget(&q->q_usage_counter))
893 : return 0;
894 0 : if (queue_is_mq(q)) {
895 0 : ret = blk_mq_poll(q, cookie, iob, flags);
896 : } else {
897 0 : struct gendisk *disk = q->disk;
898 :
899 0 : if (disk && disk->fops->poll_bio)
900 0 : ret = disk->fops->poll_bio(bio, iob, flags);
901 : }
902 0 : blk_queue_exit(q);
903 0 : return ret;
904 : }
905 : EXPORT_SYMBOL_GPL(bio_poll);
906 :
907 : /*
908 : * Helper to implement file_operations.iopoll. Requires the bio to be stored
909 : * in iocb->private, and cleared before freeing the bio.
910 : */
911 0 : int iocb_bio_iopoll(struct kiocb *kiocb, struct io_comp_batch *iob,
912 : unsigned int flags)
913 : {
914 : struct bio *bio;
915 0 : int ret = 0;
916 :
917 : /*
918 : * Note: the bio cache only uses SLAB_TYPESAFE_BY_RCU, so bio can
919 : * point to a freshly allocated bio at this point. If that happens
920 : * we have a few cases to consider:
921 : *
922 : * 1) the bio is beeing initialized and bi_bdev is NULL. We can just
923 : * simply nothing in this case
924 : * 2) the bio points to a not poll enabled device. bio_poll will catch
925 : * this and return 0
926 : * 3) the bio points to a poll capable device, including but not
927 : * limited to the one that the original bio pointed to. In this
928 : * case we will call into the actual poll method and poll for I/O,
929 : * even if we don't need to, but it won't cause harm either.
930 : *
931 : * For cases 2) and 3) above the RCU grace period ensures that bi_bdev
932 : * is still allocated. Because partitions hold a reference to the whole
933 : * device bdev and thus disk, the disk is also still valid. Grabbing
934 : * a reference to the queue in bio_poll() ensures the hctxs and requests
935 : * are still valid as well.
936 : */
937 : rcu_read_lock();
938 0 : bio = READ_ONCE(kiocb->private);
939 0 : if (bio)
940 0 : ret = bio_poll(bio, iob, flags);
941 : rcu_read_unlock();
942 :
943 0 : return ret;
944 : }
945 : EXPORT_SYMBOL_GPL(iocb_bio_iopoll);
946 :
947 0 : void update_io_ticks(struct block_device *part, unsigned long now, bool end)
948 : {
949 : unsigned long stamp;
950 : again:
951 0 : stamp = READ_ONCE(part->bd_stamp);
952 0 : if (unlikely(time_after(now, stamp))) {
953 0 : if (likely(try_cmpxchg(&part->bd_stamp, &stamp, now)))
954 0 : __part_stat_add(part, io_ticks, end ? now - stamp : 1);
955 : }
956 0 : if (part->bd_partno) {
957 0 : part = bdev_whole(part);
958 0 : goto again;
959 : }
960 0 : }
961 :
962 0 : unsigned long bdev_start_io_acct(struct block_device *bdev,
963 : unsigned int sectors, enum req_op op,
964 : unsigned long start_time)
965 : {
966 0 : const int sgrp = op_stat_group(op);
967 :
968 0 : part_stat_lock();
969 0 : update_io_ticks(bdev, start_time, false);
970 0 : part_stat_inc(bdev, ios[sgrp]);
971 0 : part_stat_add(bdev, sectors[sgrp], sectors);
972 0 : part_stat_local_inc(bdev, in_flight[op_is_write(op)]);
973 0 : part_stat_unlock();
974 :
975 0 : return start_time;
976 : }
977 : EXPORT_SYMBOL(bdev_start_io_acct);
978 :
979 : /**
980 : * bio_start_io_acct - start I/O accounting for bio based drivers
981 : * @bio: bio to start account for
982 : *
983 : * Returns the start time that should be passed back to bio_end_io_acct().
984 : */
985 0 : unsigned long bio_start_io_acct(struct bio *bio)
986 : {
987 0 : return bdev_start_io_acct(bio->bi_bdev, bio_sectors(bio),
988 : bio_op(bio), jiffies);
989 : }
990 : EXPORT_SYMBOL_GPL(bio_start_io_acct);
991 :
992 0 : void bdev_end_io_acct(struct block_device *bdev, enum req_op op,
993 : unsigned long start_time)
994 : {
995 0 : const int sgrp = op_stat_group(op);
996 0 : unsigned long now = READ_ONCE(jiffies);
997 0 : unsigned long duration = now - start_time;
998 :
999 0 : part_stat_lock();
1000 0 : update_io_ticks(bdev, now, true);
1001 0 : part_stat_add(bdev, nsecs[sgrp], jiffies_to_nsecs(duration));
1002 0 : part_stat_local_dec(bdev, in_flight[op_is_write(op)]);
1003 0 : part_stat_unlock();
1004 0 : }
1005 : EXPORT_SYMBOL(bdev_end_io_acct);
1006 :
1007 0 : void bio_end_io_acct_remapped(struct bio *bio, unsigned long start_time,
1008 : struct block_device *orig_bdev)
1009 : {
1010 0 : bdev_end_io_acct(orig_bdev, bio_op(bio), start_time);
1011 0 : }
1012 : EXPORT_SYMBOL_GPL(bio_end_io_acct_remapped);
1013 :
1014 : /**
1015 : * blk_lld_busy - Check if underlying low-level drivers of a device are busy
1016 : * @q : the queue of the device being checked
1017 : *
1018 : * Description:
1019 : * Check if underlying low-level drivers of a device are busy.
1020 : * If the drivers want to export their busy state, they must set own
1021 : * exporting function using blk_queue_lld_busy() first.
1022 : *
1023 : * Basically, this function is used only by request stacking drivers
1024 : * to stop dispatching requests to underlying devices when underlying
1025 : * devices are busy. This behavior helps more I/O merging on the queue
1026 : * of the request stacking driver and prevents I/O throughput regression
1027 : * on burst I/O load.
1028 : *
1029 : * Return:
1030 : * 0 - Not busy (The request stacking driver should dispatch request)
1031 : * 1 - Busy (The request stacking driver should stop dispatching request)
1032 : */
1033 0 : int blk_lld_busy(struct request_queue *q)
1034 : {
1035 0 : if (queue_is_mq(q) && q->mq_ops->busy)
1036 0 : return q->mq_ops->busy(q);
1037 :
1038 : return 0;
1039 : }
1040 : EXPORT_SYMBOL_GPL(blk_lld_busy);
1041 :
1042 0 : int kblockd_schedule_work(struct work_struct *work)
1043 : {
1044 0 : return queue_work(kblockd_workqueue, work);
1045 : }
1046 : EXPORT_SYMBOL(kblockd_schedule_work);
1047 :
1048 0 : int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork,
1049 : unsigned long delay)
1050 : {
1051 0 : return mod_delayed_work_on(cpu, kblockd_workqueue, dwork, delay);
1052 : }
1053 : EXPORT_SYMBOL(kblockd_mod_delayed_work_on);
1054 :
1055 0 : void blk_start_plug_nr_ios(struct blk_plug *plug, unsigned short nr_ios)
1056 : {
1057 0 : struct task_struct *tsk = current;
1058 :
1059 : /*
1060 : * If this is a nested plug, don't actually assign it.
1061 : */
1062 0 : if (tsk->plug)
1063 : return;
1064 :
1065 0 : plug->mq_list = NULL;
1066 0 : plug->cached_rq = NULL;
1067 0 : plug->nr_ios = min_t(unsigned short, nr_ios, BLK_MAX_REQUEST_COUNT);
1068 0 : plug->rq_count = 0;
1069 0 : plug->multiple_queues = false;
1070 0 : plug->has_elevator = false;
1071 0 : plug->nowait = false;
1072 0 : INIT_LIST_HEAD(&plug->cb_list);
1073 :
1074 : /*
1075 : * Store ordering should not be needed here, since a potential
1076 : * preempt will imply a full memory barrier
1077 : */
1078 0 : tsk->plug = plug;
1079 : }
1080 :
1081 : /**
1082 : * blk_start_plug - initialize blk_plug and track it inside the task_struct
1083 : * @plug: The &struct blk_plug that needs to be initialized
1084 : *
1085 : * Description:
1086 : * blk_start_plug() indicates to the block layer an intent by the caller
1087 : * to submit multiple I/O requests in a batch. The block layer may use
1088 : * this hint to defer submitting I/Os from the caller until blk_finish_plug()
1089 : * is called. However, the block layer may choose to submit requests
1090 : * before a call to blk_finish_plug() if the number of queued I/Os
1091 : * exceeds %BLK_MAX_REQUEST_COUNT, or if the size of the I/O is larger than
1092 : * %BLK_PLUG_FLUSH_SIZE. The queued I/Os may also be submitted early if
1093 : * the task schedules (see below).
1094 : *
1095 : * Tracking blk_plug inside the task_struct will help with auto-flushing the
1096 : * pending I/O should the task end up blocking between blk_start_plug() and
1097 : * blk_finish_plug(). This is important from a performance perspective, but
1098 : * also ensures that we don't deadlock. For instance, if the task is blocking
1099 : * for a memory allocation, memory reclaim could end up wanting to free a
1100 : * page belonging to that request that is currently residing in our private
1101 : * plug. By flushing the pending I/O when the process goes to sleep, we avoid
1102 : * this kind of deadlock.
1103 : */
1104 0 : void blk_start_plug(struct blk_plug *plug)
1105 : {
1106 0 : blk_start_plug_nr_ios(plug, 1);
1107 0 : }
1108 : EXPORT_SYMBOL(blk_start_plug);
1109 :
1110 0 : static void flush_plug_callbacks(struct blk_plug *plug, bool from_schedule)
1111 : {
1112 0 : LIST_HEAD(callbacks);
1113 :
1114 0 : while (!list_empty(&plug->cb_list)) {
1115 0 : list_splice_init(&plug->cb_list, &callbacks);
1116 :
1117 0 : while (!list_empty(&callbacks)) {
1118 0 : struct blk_plug_cb *cb = list_first_entry(&callbacks,
1119 : struct blk_plug_cb,
1120 : list);
1121 0 : list_del(&cb->list);
1122 0 : cb->callback(cb, from_schedule);
1123 : }
1124 : }
1125 0 : }
1126 :
1127 0 : struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug, void *data,
1128 : int size)
1129 : {
1130 0 : struct blk_plug *plug = current->plug;
1131 : struct blk_plug_cb *cb;
1132 :
1133 0 : if (!plug)
1134 : return NULL;
1135 :
1136 0 : list_for_each_entry(cb, &plug->cb_list, list)
1137 0 : if (cb->callback == unplug && cb->data == data)
1138 : return cb;
1139 :
1140 : /* Not currently on the callback list */
1141 0 : BUG_ON(size < sizeof(*cb));
1142 0 : cb = kzalloc(size, GFP_ATOMIC);
1143 0 : if (cb) {
1144 0 : cb->data = data;
1145 0 : cb->callback = unplug;
1146 0 : list_add(&cb->list, &plug->cb_list);
1147 : }
1148 : return cb;
1149 : }
1150 : EXPORT_SYMBOL(blk_check_plugged);
1151 :
1152 0 : void __blk_flush_plug(struct blk_plug *plug, bool from_schedule)
1153 : {
1154 0 : if (!list_empty(&plug->cb_list))
1155 0 : flush_plug_callbacks(plug, from_schedule);
1156 0 : if (!rq_list_empty(plug->mq_list))
1157 0 : blk_mq_flush_plug_list(plug, from_schedule);
1158 : /*
1159 : * Unconditionally flush out cached requests, even if the unplug
1160 : * event came from schedule. Since we know hold references to the
1161 : * queue for cached requests, we don't want a blocked task holding
1162 : * up a queue freeze/quiesce event.
1163 : */
1164 0 : if (unlikely(!rq_list_empty(plug->cached_rq)))
1165 0 : blk_mq_free_plug_rqs(plug);
1166 0 : }
1167 :
1168 : /**
1169 : * blk_finish_plug - mark the end of a batch of submitted I/O
1170 : * @plug: The &struct blk_plug passed to blk_start_plug()
1171 : *
1172 : * Description:
1173 : * Indicate that a batch of I/O submissions is complete. This function
1174 : * must be paired with an initial call to blk_start_plug(). The intent
1175 : * is to allow the block layer to optimize I/O submission. See the
1176 : * documentation for blk_start_plug() for more information.
1177 : */
1178 0 : void blk_finish_plug(struct blk_plug *plug)
1179 : {
1180 0 : if (plug == current->plug) {
1181 0 : __blk_flush_plug(plug, false);
1182 0 : current->plug = NULL;
1183 : }
1184 0 : }
1185 : EXPORT_SYMBOL(blk_finish_plug);
1186 :
1187 0 : void blk_io_schedule(void)
1188 : {
1189 : /* Prevent hang_check timer from firing at us during very long I/O */
1190 0 : unsigned long timeout = sysctl_hung_task_timeout_secs * HZ / 2;
1191 :
1192 : if (timeout)
1193 : io_schedule_timeout(timeout);
1194 : else
1195 0 : io_schedule();
1196 0 : }
1197 : EXPORT_SYMBOL_GPL(blk_io_schedule);
1198 :
1199 1 : int __init blk_dev_init(void)
1200 : {
1201 : BUILD_BUG_ON((__force u32)REQ_OP_LAST >= (1 << REQ_OP_BITS));
1202 : BUILD_BUG_ON(REQ_OP_BITS + REQ_FLAG_BITS > 8 *
1203 : sizeof_field(struct request, cmd_flags));
1204 : BUILD_BUG_ON(REQ_OP_BITS + REQ_FLAG_BITS > 8 *
1205 : sizeof_field(struct bio, bi_opf));
1206 :
1207 : /* used for unplugging and affects IO latency/throughput - HIGHPRI */
1208 1 : kblockd_workqueue = alloc_workqueue("kblockd",
1209 : WQ_MEM_RECLAIM | WQ_HIGHPRI, 0);
1210 1 : if (!kblockd_workqueue)
1211 0 : panic("Failed to create kblockd\n");
1212 :
1213 1 : blk_requestq_cachep = kmem_cache_create("request_queue",
1214 : sizeof(struct request_queue), 0, SLAB_PANIC, NULL);
1215 :
1216 2 : blk_debugfs_root = debugfs_create_dir("block", NULL);
1217 :
1218 1 : return 0;
1219 : }
|
