1
2
3
4
5
6
7
8
9
10
11
12
13
14
15#include <linux/kernel.h>
16#include <linux/module.h>
17#include <linux/backing-dev.h>
18#include <linux/bio.h>
19#include <linux/blkdev.h>
20#include <linux/blk-mq.h>
21#include <linux/highmem.h>
22#include <linux/mm.h>
23#include <linux/kernel_stat.h>
24#include <linux/string.h>
25#include <linux/init.h>
26#include <linux/completion.h>
27#include <linux/slab.h>
28#include <linux/swap.h>
29#include <linux/writeback.h>
30#include <linux/task_io_accounting_ops.h>
31#include <linux/fault-inject.h>
32#include <linux/list_sort.h>
33#include <linux/delay.h>
34#include <linux/ratelimit.h>
35#include <linux/pm_runtime.h>
36#include <linux/blk-cgroup.h>
37#include <linux/debugfs.h>
38#include <linux/bpf.h>
39
40#define CREATE_TRACE_POINTS
41#include <trace/events/block.h>
42
43#include "blk.h"
44#include "blk-mq.h"
45#include "blk-mq-sched.h"
46#include "blk-pm.h"
47#include "blk-rq-qos.h"
48
49#ifdef CONFIG_DEBUG_FS
50struct dentry *blk_debugfs_root;
51#endif
52
53EXPORT_TRACEPOINT_SYMBOL_GPL(block_bio_remap);
54EXPORT_TRACEPOINT_SYMBOL_GPL(block_rq_remap);
55EXPORT_TRACEPOINT_SYMBOL_GPL(block_bio_complete);
56EXPORT_TRACEPOINT_SYMBOL_GPL(block_split);
57EXPORT_TRACEPOINT_SYMBOL_GPL(block_unplug);
58
59DEFINE_IDA(blk_queue_ida);
60
61
62
63
64struct kmem_cache *blk_requestq_cachep;
65
66
67
68
69static struct workqueue_struct *kblockd_workqueue;
70
71
72
73
74
75
76void blk_queue_flag_set(unsigned int flag, struct request_queue *q)
77{
78 set_bit(flag, &q->queue_flags);
79}
80EXPORT_SYMBOL(blk_queue_flag_set);
81
82
83
84
85
86
87void blk_queue_flag_clear(unsigned int flag, struct request_queue *q)
88{
89 clear_bit(flag, &q->queue_flags);
90}
91EXPORT_SYMBOL(blk_queue_flag_clear);
92
93
94
95
96
97
98
99
100
101bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q)
102{
103 return test_and_set_bit(flag, &q->queue_flags);
104}
105EXPORT_SYMBOL_GPL(blk_queue_flag_test_and_set);
106
107void blk_rq_init(struct request_queue *q, struct request *rq)
108{
109 memset(rq, 0, sizeof(*rq));
110
111 INIT_LIST_HEAD(&rq->queuelist);
112 rq->q = q;
113 rq->__sector = (sector_t) -1;
114 INIT_HLIST_NODE(&rq->hash);
115 RB_CLEAR_NODE(&rq->rb_node);
116 rq->tag = -1;
117 rq->internal_tag = -1;
118 rq->start_time_ns = ktime_get_ns();
119 rq->part = NULL;
120}
121EXPORT_SYMBOL(blk_rq_init);
122
123static const struct {
124 int errno;
125 const char *name;
126} blk_errors[] = {
127 [BLK_STS_OK] = { 0, "" },
128 [BLK_STS_NOTSUPP] = { -EOPNOTSUPP, "operation not supported" },
129 [BLK_STS_TIMEOUT] = { -ETIMEDOUT, "timeout" },
130 [BLK_STS_NOSPC] = { -ENOSPC, "critical space allocation" },
131 [BLK_STS_TRANSPORT] = { -ENOLINK, "recoverable transport" },
132 [BLK_STS_TARGET] = { -EREMOTEIO, "critical target" },
133 [BLK_STS_NEXUS] = { -EBADE, "critical nexus" },
134 [BLK_STS_MEDIUM] = { -ENODATA, "critical medium" },
135 [BLK_STS_PROTECTION] = { -EILSEQ, "protection" },
136 [BLK_STS_RESOURCE] = { -ENOMEM, "kernel resource" },
137 [BLK_STS_DEV_RESOURCE] = { -EBUSY, "device resource" },
138 [BLK_STS_AGAIN] = { -EAGAIN, "nonblocking retry" },
139
140
141 [BLK_STS_DM_REQUEUE] = { -EREMCHG, "dm internal retry" },
142
143
144 [BLK_STS_IOERR] = { -EIO, "I/O" },
145};
146
147blk_status_t errno_to_blk_status(int errno)
148{
149 int i;
150
151 for (i = 0; i < ARRAY_SIZE(blk_errors); i++) {
152 if (blk_errors[i].errno == errno)
153 return (__force blk_status_t)i;
154 }
155
156 return BLK_STS_IOERR;
157}
158EXPORT_SYMBOL_GPL(errno_to_blk_status);
159
160int blk_status_to_errno(blk_status_t status)
161{
162 int idx = (__force int)status;
163
164 if (WARN_ON_ONCE(idx >= ARRAY_SIZE(blk_errors)))
165 return -EIO;
166 return blk_errors[idx].errno;
167}
168EXPORT_SYMBOL_GPL(blk_status_to_errno);
169
170static void print_req_error(struct request *req, blk_status_t status)
171{
172 int idx = (__force int)status;
173
174 if (WARN_ON_ONCE(idx >= ARRAY_SIZE(blk_errors)))
175 return;
176
177 printk_ratelimited(KERN_ERR "%s: %s error, dev %s, sector %llu flags %x\n",
178 __func__, blk_errors[idx].name,
179 req->rq_disk ? req->rq_disk->disk_name : "?",
180 (unsigned long long)blk_rq_pos(req),
181 req->cmd_flags);
182}
183
184static void req_bio_endio(struct request *rq, struct bio *bio,
185 unsigned int nbytes, blk_status_t error)
186{
187 if (error)
188 bio->bi_status = error;
189
190 if (unlikely(rq->rq_flags & RQF_QUIET))
191 bio_set_flag(bio, BIO_QUIET);
192
193 bio_advance(bio, nbytes);
194
195
196 if (bio->bi_iter.bi_size == 0 && !(rq->rq_flags & RQF_FLUSH_SEQ))
197 bio_endio(bio);
198}
199
200void blk_dump_rq_flags(struct request *rq, char *msg)
201{
202 printk(KERN_INFO "%s: dev %s: flags=%llx\n", msg,
203 rq->rq_disk ? rq->rq_disk->disk_name : "?",
204 (unsigned long long) rq->cmd_flags);
205
206 printk(KERN_INFO " sector %llu, nr/cnr %u/%u\n",
207 (unsigned long long)blk_rq_pos(rq),
208 blk_rq_sectors(rq), blk_rq_cur_sectors(rq));
209 printk(KERN_INFO " bio %p, biotail %p, len %u\n",
210 rq->bio, rq->biotail, blk_rq_bytes(rq));
211}
212EXPORT_SYMBOL(blk_dump_rq_flags);
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232void blk_sync_queue(struct request_queue *q)
233{
234 del_timer_sync(&q->timeout);
235 cancel_work_sync(&q->timeout_work);
236}
237EXPORT_SYMBOL(blk_sync_queue);
238
239
240
241
242
243void blk_set_pm_only(struct request_queue *q)
244{
245 atomic_inc(&q->pm_only);
246}
247EXPORT_SYMBOL_GPL(blk_set_pm_only);
248
249void blk_clear_pm_only(struct request_queue *q)
250{
251 int pm_only;
252
253 pm_only = atomic_dec_return(&q->pm_only);
254 WARN_ON_ONCE(pm_only < 0);
255 if (pm_only == 0)
256 wake_up_all(&q->mq_freeze_wq);
257}
258EXPORT_SYMBOL_GPL(blk_clear_pm_only);
259
260void blk_put_queue(struct request_queue *q)
261{
262 kobject_put(&q->kobj);
263}
264EXPORT_SYMBOL(blk_put_queue);
265
266void blk_set_queue_dying(struct request_queue *q)
267{
268 blk_queue_flag_set(QUEUE_FLAG_DYING, q);
269
270
271
272
273
274
275 blk_freeze_queue_start(q);
276
277 if (queue_is_mq(q))
278 blk_mq_wake_waiters(q);
279
280
281 wake_up_all(&q->mq_freeze_wq);
282}
283EXPORT_SYMBOL_GPL(blk_set_queue_dying);
284
285
286
287
288
289
290
291
292void blk_cleanup_queue(struct request_queue *q)
293{
294
295 mutex_lock(&q->sysfs_lock);
296 blk_set_queue_dying(q);
297
298 blk_queue_flag_set(QUEUE_FLAG_NOMERGES, q);
299 blk_queue_flag_set(QUEUE_FLAG_NOXMERGES, q);
300 blk_queue_flag_set(QUEUE_FLAG_DYING, q);
301 mutex_unlock(&q->sysfs_lock);
302
303
304
305
306
307 blk_freeze_queue(q);
308
309 rq_qos_exit(q);
310
311 blk_queue_flag_set(QUEUE_FLAG_DEAD, q);
312
313
314 blk_flush_integrity();
315
316
317 del_timer_sync(&q->backing_dev_info->laptop_mode_wb_timer);
318 blk_sync_queue(q);
319
320 if (queue_is_mq(q))
321 blk_mq_exit_queue(q);
322
323
324
325
326
327
328
329
330
331 mutex_lock(&q->sysfs_lock);
332 if (q->elevator)
333 blk_mq_sched_free_requests(q);
334 mutex_unlock(&q->sysfs_lock);
335
336 percpu_ref_exit(&q->q_usage_counter);
337
338
339 blk_put_queue(q);
340}
341EXPORT_SYMBOL(blk_cleanup_queue);
342
343struct request_queue *blk_alloc_queue(gfp_t gfp_mask)
344{
345 return blk_alloc_queue_node(gfp_mask, NUMA_NO_NODE);
346}
347EXPORT_SYMBOL(blk_alloc_queue);
348
349
350
351
352
353
354int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags)
355{
356 const bool pm = flags & BLK_MQ_REQ_PREEMPT;
357
358 while (true) {
359 bool success = false;
360
361 rcu_read_lock();
362 if (percpu_ref_tryget_live(&q->q_usage_counter)) {
363
364
365
366
367
368 if (pm || !blk_queue_pm_only(q)) {
369 success = true;
370 } else {
371 percpu_ref_put(&q->q_usage_counter);
372 }
373 }
374 rcu_read_unlock();
375
376 if (success)
377 return 0;
378
379 if (flags & BLK_MQ_REQ_NOWAIT)
380 return -EBUSY;
381
382
383
384
385
386
387
388
389 smp_rmb();
390
391 wait_event(q->mq_freeze_wq,
392 (!q->mq_freeze_depth &&
393 (pm || (blk_pm_request_resume(q),
394 !blk_queue_pm_only(q)))) ||
395 blk_queue_dying(q));
396 if (blk_queue_dying(q))
397 return -ENODEV;
398 }
399}
400
401void blk_queue_exit(struct request_queue *q)
402{
403 percpu_ref_put(&q->q_usage_counter);
404}
405
406static void blk_queue_usage_counter_release(struct percpu_ref *ref)
407{
408 struct request_queue *q =
409 container_of(ref, struct request_queue, q_usage_counter);
410
411 wake_up_all(&q->mq_freeze_wq);
412}
413
414static void blk_rq_timed_out_timer(struct timer_list *t)
415{
416 struct request_queue *q = from_timer(q, t, timeout);
417
418 kblockd_schedule_work(&q->timeout_work);
419}
420
421static void blk_timeout_work(struct work_struct *work)
422{
423}
424
425
426
427
428
429
430struct request_queue *blk_alloc_queue_node(gfp_t gfp_mask, int node_id)
431{
432 struct request_queue *q;
433 int ret;
434
435 q = kmem_cache_alloc_node(blk_requestq_cachep,
436 gfp_mask | __GFP_ZERO, node_id);
437 if (!q)
438 return NULL;
439
440 INIT_LIST_HEAD(&q->queue_head);
441 q->last_merge = NULL;
442
443 q->id = ida_simple_get(&blk_queue_ida, 0, 0, gfp_mask);
444 if (q->id < 0)
445 goto fail_q;
446
447 ret = bioset_init(&q->bio_split, BIO_POOL_SIZE, 0, BIOSET_NEED_BVECS);
448 if (ret)
449 goto fail_id;
450
451 q->backing_dev_info = bdi_alloc_node(gfp_mask, node_id);
452 if (!q->backing_dev_info)
453 goto fail_split;
454
455 q->stats = blk_alloc_queue_stats();
456 if (!q->stats)
457 goto fail_stats;
458
459 q->backing_dev_info->ra_pages = VM_READAHEAD_PAGES;
460 q->backing_dev_info->capabilities = BDI_CAP_CGROUP_WRITEBACK;
461 q->backing_dev_info->name = "block";
462 q->node = node_id;
463
464 timer_setup(&q->backing_dev_info->laptop_mode_wb_timer,
465 laptop_mode_timer_fn, 0);
466 timer_setup(&q->timeout, blk_rq_timed_out_timer, 0);
467 INIT_WORK(&q->timeout_work, blk_timeout_work);
468 INIT_LIST_HEAD(&q->icq_list);
469#ifdef CONFIG_BLK_CGROUP
470 INIT_LIST_HEAD(&q->blkg_list);
471#endif
472
473 kobject_init(&q->kobj, &blk_queue_ktype);
474
475#ifdef CONFIG_BLK_DEV_IO_TRACE
476 mutex_init(&q->blk_trace_mutex);
477#endif
478 mutex_init(&q->sysfs_lock);
479 spin_lock_init(&q->queue_lock);
480
481 init_waitqueue_head(&q->mq_freeze_wq);
482 mutex_init(&q->mq_freeze_lock);
483
484
485
486
487
488 if (percpu_ref_init(&q->q_usage_counter,
489 blk_queue_usage_counter_release,
490 PERCPU_REF_INIT_ATOMIC, GFP_KERNEL))
491 goto fail_bdi;
492
493 if (blkcg_init_queue(q))
494 goto fail_ref;
495
496 return q;
497
498fail_ref:
499 percpu_ref_exit(&q->q_usage_counter);
500fail_bdi:
501 blk_free_queue_stats(q->stats);
502fail_stats:
503 bdi_put(q->backing_dev_info);
504fail_split:
505 bioset_exit(&q->bio_split);
506fail_id:
507 ida_simple_remove(&blk_queue_ida, q->id);
508fail_q:
509 kmem_cache_free(blk_requestq_cachep, q);
510 return NULL;
511}
512EXPORT_SYMBOL(blk_alloc_queue_node);
513
514bool blk_get_queue(struct request_queue *q)
515{
516 if (likely(!blk_queue_dying(q))) {
517 __blk_get_queue(q);
518 return true;
519 }
520
521 return false;
522}
523EXPORT_SYMBOL(blk_get_queue);
524
525
526
527
528
529
530
531struct request *blk_get_request(struct request_queue *q, unsigned int op,
532 blk_mq_req_flags_t flags)
533{
534 struct request *req;
535
536 WARN_ON_ONCE(op & REQ_NOWAIT);
537 WARN_ON_ONCE(flags & ~(BLK_MQ_REQ_NOWAIT | BLK_MQ_REQ_PREEMPT));
538
539 req = blk_mq_alloc_request(q, op, flags);
540 if (!IS_ERR(req) && q->mq_ops->initialize_rq_fn)
541 q->mq_ops->initialize_rq_fn(req);
542
543 return req;
544}
545EXPORT_SYMBOL(blk_get_request);
546
547void blk_put_request(struct request *req)
548{
549 blk_mq_free_request(req);
550}
551EXPORT_SYMBOL(blk_put_request);
552
553bool bio_attempt_back_merge(struct request_queue *q, struct request *req,
554 struct bio *bio)
555{
556 const int ff = bio->bi_opf & REQ_FAILFAST_MASK;
557
558 if (!ll_back_merge_fn(q, req, bio))
559 return false;
560
561 trace_block_bio_backmerge(q, req, bio);
562
563 if ((req->cmd_flags & REQ_FAILFAST_MASK) != ff)
564 blk_rq_set_mixed_merge(req);
565
566 req->biotail->bi_next = bio;
567 req->biotail = bio;
568 req->__data_len += bio->bi_iter.bi_size;
569
570 blk_account_io_start(req, false);
571 return true;
572}
573
574bool bio_attempt_front_merge(struct request_queue *q, struct request *req,
575 struct bio *bio)
576{
577 const int ff = bio->bi_opf & REQ_FAILFAST_MASK;
578
579 if (!ll_front_merge_fn(q, req, bio))
580 return false;
581
582 trace_block_bio_frontmerge(q, req, bio);
583
584 if ((req->cmd_flags & REQ_FAILFAST_MASK) != ff)
585 blk_rq_set_mixed_merge(req);
586
587 bio->bi_next = req->bio;
588 req->bio = bio;
589
590 req->__sector = bio->bi_iter.bi_sector;
591 req->__data_len += bio->bi_iter.bi_size;
592
593 blk_account_io_start(req, false);
594 return true;
595}
596
597bool bio_attempt_discard_merge(struct request_queue *q, struct request *req,
598 struct bio *bio)
599{
600 unsigned short segments = blk_rq_nr_discard_segments(req);
601
602 if (segments >= queue_max_discard_segments(q))
603 goto no_merge;
604 if (blk_rq_sectors(req) + bio_sectors(bio) >
605 blk_rq_get_max_sectors(req, blk_rq_pos(req)))
606 goto no_merge;
607
608 req->biotail->bi_next = bio;
609 req->biotail = bio;
610 req->__data_len += bio->bi_iter.bi_size;
611 req->nr_phys_segments = segments + 1;
612
613 blk_account_io_start(req, false);
614 return true;
615no_merge:
616 req_set_nomerge(q, req);
617 return false;
618}
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641bool blk_attempt_plug_merge(struct request_queue *q, struct bio *bio,
642 struct request **same_queue_rq)
643{
644 struct blk_plug *plug;
645 struct request *rq;
646 struct list_head *plug_list;
647
648 plug = current->plug;
649 if (!plug)
650 return false;
651
652 plug_list = &plug->mq_list;
653
654 list_for_each_entry_reverse(rq, plug_list, queuelist) {
655 bool merged = false;
656
657 if (rq->q == q && same_queue_rq) {
658
659
660
661
662
663 *same_queue_rq = rq;
664 }
665
666 if (rq->q != q || !blk_rq_merge_ok(rq, bio))
667 continue;
668
669 switch (blk_try_merge(rq, bio)) {
670 case ELEVATOR_BACK_MERGE:
671 merged = bio_attempt_back_merge(q, rq, bio);
672 break;
673 case ELEVATOR_FRONT_MERGE:
674 merged = bio_attempt_front_merge(q, rq, bio);
675 break;
676 case ELEVATOR_DISCARD_MERGE:
677 merged = bio_attempt_discard_merge(q, rq, bio);
678 break;
679 default:
680 break;
681 }
682
683 if (merged)
684 return true;
685 }
686
687 return false;
688}
689
690void blk_init_request_from_bio(struct request *req, struct bio *bio)
691{
692 if (bio->bi_opf & REQ_RAHEAD)
693 req->cmd_flags |= REQ_FAILFAST_MASK;
694
695 req->__sector = bio->bi_iter.bi_sector;
696 req->ioprio = bio_prio(bio);
697 req->write_hint = bio->bi_write_hint;
698 blk_rq_bio_prep(req->q, req, bio);
699}
700EXPORT_SYMBOL_GPL(blk_init_request_from_bio);
701
702static void handle_bad_sector(struct bio *bio, sector_t maxsector)
703{
704 char b[BDEVNAME_SIZE];
705
706 printk(KERN_INFO "attempt to access beyond end of device\n");
707 printk(KERN_INFO "%s: rw=%d, want=%Lu, limit=%Lu\n",
708 bio_devname(bio, b), bio->bi_opf,
709 (unsigned long long)bio_end_sector(bio),
710 (long long)maxsector);
711}
712
713#ifdef CONFIG_FAIL_MAKE_REQUEST
714
715static DECLARE_FAULT_ATTR(fail_make_request);
716
717static int __init setup_fail_make_request(char *str)
718{
719 return setup_fault_attr(&fail_make_request, str);
720}
721__setup("fail_make_request=", setup_fail_make_request);
722
723static bool should_fail_request(struct hd_struct *part, unsigned int bytes)
724{
725 return part->make_it_fail && should_fail(&fail_make_request, bytes);
726}
727
728static int __init fail_make_request_debugfs(void)
729{
730 struct dentry *dir = fault_create_debugfs_attr("fail_make_request",
731 NULL, &fail_make_request);
732
733 return PTR_ERR_OR_ZERO(dir);
734}
735
736late_initcall(fail_make_request_debugfs);
737
738#else
739
740static inline bool should_fail_request(struct hd_struct *part,
741 unsigned int bytes)
742{
743 return false;
744}
745
746#endif
747
748static inline bool bio_check_ro(struct bio *bio, struct hd_struct *part)
749{
750 const int op = bio_op(bio);
751
752 if (part->policy && op_is_write(op)) {
753 char b[BDEVNAME_SIZE];
754
755 if (op_is_flush(bio->bi_opf) && !bio_sectors(bio))
756 return false;
757
758 WARN_ONCE(1,
759 "generic_make_request: Trying to write "
760 "to read-only block-device %s (partno %d)\n",
761 bio_devname(bio, b), part->partno);
762
763 return false;
764 }
765
766 return false;
767}
768
769static noinline int should_fail_bio(struct bio *bio)
770{
771 if (should_fail_request(&bio->bi_disk->part0, bio->bi_iter.bi_size))
772 return -EIO;
773 return 0;
774}
775ALLOW_ERROR_INJECTION(should_fail_bio, ERRNO);
776
777
778
779
780
781
782static inline int bio_check_eod(struct bio *bio, sector_t maxsector)
783{
784 unsigned int nr_sectors = bio_sectors(bio);
785
786 if (nr_sectors && maxsector &&
787 (nr_sectors > maxsector ||
788 bio->bi_iter.bi_sector > maxsector - nr_sectors)) {
789 handle_bad_sector(bio, maxsector);
790 return -EIO;
791 }
792 return 0;
793}
794
795
796
797
798static inline int blk_partition_remap(struct bio *bio)
799{
800 struct hd_struct *p;
801 int ret = -EIO;
802
803 rcu_read_lock();
804 p = __disk_get_part(bio->bi_disk, bio->bi_partno);
805 if (unlikely(!p))
806 goto out;
807 if (unlikely(should_fail_request(p, bio->bi_iter.bi_size)))
808 goto out;
809 if (unlikely(bio_check_ro(bio, p)))
810 goto out;
811
812
813
814
815
816 if (bio_sectors(bio) || bio_op(bio) == REQ_OP_ZONE_RESET) {
817 if (bio_check_eod(bio, part_nr_sects_read(p)))
818 goto out;
819 bio->bi_iter.bi_sector += p->start_sect;
820 trace_block_bio_remap(bio->bi_disk->queue, bio, part_devt(p),
821 bio->bi_iter.bi_sector - p->start_sect);
822 }
823 bio->bi_partno = 0;
824 ret = 0;
825out:
826 rcu_read_unlock();
827 return ret;
828}
829
830static noinline_for_stack bool
831generic_make_request_checks(struct bio *bio)
832{
833 struct request_queue *q;
834 int nr_sectors = bio_sectors(bio);
835 blk_status_t status = BLK_STS_IOERR;
836 char b[BDEVNAME_SIZE];
837
838 might_sleep();
839
840 q = bio->bi_disk->queue;
841 if (unlikely(!q)) {
842 printk(KERN_ERR
843 "generic_make_request: Trying to access "
844 "nonexistent block-device %s (%Lu)\n",
845 bio_devname(bio, b), (long long)bio->bi_iter.bi_sector);
846 goto end_io;
847 }
848
849
850
851
852
853 if ((bio->bi_opf & REQ_NOWAIT) && !queue_is_mq(q))
854 goto not_supported;
855
856 if (should_fail_bio(bio))
857 goto end_io;
858
859 if (bio->bi_partno) {
860 if (unlikely(blk_partition_remap(bio)))
861 goto end_io;
862 } else {
863 if (unlikely(bio_check_ro(bio, &bio->bi_disk->part0)))
864 goto end_io;
865 if (unlikely(bio_check_eod(bio, get_capacity(bio->bi_disk))))
866 goto end_io;
867 }
868
869
870
871
872
873
874 if (op_is_flush(bio->bi_opf) &&
875 !test_bit(QUEUE_FLAG_WC, &q->queue_flags)) {
876 bio->bi_opf &= ~(REQ_PREFLUSH | REQ_FUA);
877 if (!nr_sectors) {
878 status = BLK_STS_OK;
879 goto end_io;
880 }
881 }
882
883 if (!test_bit(QUEUE_FLAG_POLL, &q->queue_flags))
884 bio->bi_opf &= ~REQ_HIPRI;
885
886 switch (bio_op(bio)) {
887 case REQ_OP_DISCARD:
888 if (!blk_queue_discard(q))
889 goto not_supported;
890 break;
891 case REQ_OP_SECURE_ERASE:
892 if (!blk_queue_secure_erase(q))
893 goto not_supported;
894 break;
895 case REQ_OP_WRITE_SAME:
896 if (!q->limits.max_write_same_sectors)
897 goto not_supported;
898 break;
899 case REQ_OP_ZONE_RESET:
900 if (!blk_queue_is_zoned(q))
901 goto not_supported;
902 break;
903 case REQ_OP_WRITE_ZEROES:
904 if (!q->limits.max_write_zeroes_sectors)
905 goto not_supported;
906 break;
907 default:
908 break;
909 }
910
911
912
913
914
915
916
917 create_io_context(GFP_ATOMIC, q->node);
918
919 if (!blkcg_bio_issue_check(q, bio))
920 return false;
921
922 if (!bio_flagged(bio, BIO_TRACE_COMPLETION)) {
923 trace_block_bio_queue(q, bio);
924
925
926
927 bio_set_flag(bio, BIO_TRACE_COMPLETION);
928 }
929 return true;
930
931not_supported:
932 status = BLK_STS_NOTSUPP;
933end_io:
934 bio->bi_status = status;
935 bio_endio(bio);
936 return false;
937}
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963blk_qc_t generic_make_request(struct bio *bio)
964{
965
966
967
968
969
970
971
972 struct bio_list bio_list_on_stack[2];
973 blk_qc_t ret = BLK_QC_T_NONE;
974
975 if (!generic_make_request_checks(bio))
976 goto out;
977
978
979
980
981
982
983
984
985
986
987
988 if (current->bio_list) {
989 bio_list_add(¤t->bio_list[0], bio);
990 goto out;
991 }
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007 BUG_ON(bio->bi_next);
1008 bio_list_init(&bio_list_on_stack[0]);
1009 current->bio_list = bio_list_on_stack;
1010 do {
1011 struct request_queue *q = bio->bi_disk->queue;
1012 blk_mq_req_flags_t flags = bio->bi_opf & REQ_NOWAIT ?
1013 BLK_MQ_REQ_NOWAIT : 0;
1014
1015 if (likely(blk_queue_enter(q, flags) == 0)) {
1016 struct bio_list lower, same;
1017
1018
1019 bio_list_on_stack[1] = bio_list_on_stack[0];
1020 bio_list_init(&bio_list_on_stack[0]);
1021 ret = q->make_request_fn(q, bio);
1022
1023 blk_queue_exit(q);
1024
1025
1026
1027
1028 bio_list_init(&lower);
1029 bio_list_init(&same);
1030 while ((bio = bio_list_pop(&bio_list_on_stack[0])) != NULL)
1031 if (q == bio->bi_disk->queue)
1032 bio_list_add(&same, bio);
1033 else
1034 bio_list_add(&lower, bio);
1035
1036 bio_list_merge(&bio_list_on_stack[0], &lower);
1037 bio_list_merge(&bio_list_on_stack[0], &same);
1038 bio_list_merge(&bio_list_on_stack[0], &bio_list_on_stack[1]);
1039 } else {
1040 if (unlikely(!blk_queue_dying(q) &&
1041 (bio->bi_opf & REQ_NOWAIT)))
1042 bio_wouldblock_error(bio);
1043 else
1044 bio_io_error(bio);
1045 }
1046 bio = bio_list_pop(&bio_list_on_stack[0]);
1047 } while (bio);
1048 current->bio_list = NULL;
1049
1050out:
1051 return ret;
1052}
1053EXPORT_SYMBOL(generic_make_request);
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065blk_qc_t direct_make_request(struct bio *bio)
1066{
1067 struct request_queue *q = bio->bi_disk->queue;
1068 bool nowait = bio->bi_opf & REQ_NOWAIT;
1069 blk_qc_t ret;
1070
1071 if (!generic_make_request_checks(bio))
1072 return BLK_QC_T_NONE;
1073
1074 if (unlikely(blk_queue_enter(q, nowait ? BLK_MQ_REQ_NOWAIT : 0))) {
1075 if (nowait && !blk_queue_dying(q))
1076 bio->bi_status = BLK_STS_AGAIN;
1077 else
1078 bio->bi_status = BLK_STS_IOERR;
1079 bio_endio(bio);
1080 return BLK_QC_T_NONE;
1081 }
1082
1083 ret = q->make_request_fn(q, bio);
1084 blk_queue_exit(q);
1085 return ret;
1086}
1087EXPORT_SYMBOL_GPL(direct_make_request);
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098blk_qc_t submit_bio(struct bio *bio)
1099{
1100
1101
1102
1103
1104 if (bio_has_data(bio)) {
1105 unsigned int count;
1106
1107 if (unlikely(bio_op(bio) == REQ_OP_WRITE_SAME))
1108 count = queue_logical_block_size(bio->bi_disk->queue) >> 9;
1109 else
1110 count = bio_sectors(bio);
1111
1112 if (op_is_write(bio_op(bio))) {
1113 count_vm_events(PGPGOUT, count);
1114 } else {
1115 task_io_account_read(bio->bi_iter.bi_size);
1116 count_vm_events(PGPGIN, count);
1117 }
1118
1119 if (unlikely(block_dump)) {
1120 char b[BDEVNAME_SIZE];
1121 printk(KERN_DEBUG "%s(%d): %s block %Lu on %s (%u sectors)\n",
1122 current->comm, task_pid_nr(current),
1123 op_is_write(bio_op(bio)) ? "WRITE" : "READ",
1124 (unsigned long long)bio->bi_iter.bi_sector,
1125 bio_devname(bio, b), count);
1126 }
1127 }
1128
1129 return generic_make_request(bio);
1130}
1131EXPORT_SYMBOL(submit_bio);
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150static int blk_cloned_rq_check_limits(struct request_queue *q,
1151 struct request *rq)
1152{
1153 if (blk_rq_sectors(rq) > blk_queue_get_max_sectors(q, req_op(rq))) {
1154 printk(KERN_ERR "%s: over max size limit. (%u > %u)\n",
1155 __func__, blk_rq_sectors(rq),
1156 blk_queue_get_max_sectors(q, req_op(rq)));
1157 return -EIO;
1158 }
1159
1160
1161
1162
1163
1164
1165
1166 blk_recalc_rq_segments(rq);
1167 if (rq->nr_phys_segments > queue_max_segments(q)) {
1168 printk(KERN_ERR "%s: over max segments limit. (%hu > %hu)\n",
1169 __func__, rq->nr_phys_segments, queue_max_segments(q));
1170 return -EIO;
1171 }
1172
1173 return 0;
1174}
1175
1176
1177
1178
1179
1180
1181blk_status_t blk_insert_cloned_request(struct request_queue *q, struct request *rq)
1182{
1183 if (blk_cloned_rq_check_limits(q, rq))
1184 return BLK_STS_IOERR;
1185
1186 if (rq->rq_disk &&
1187 should_fail_request(&rq->rq_disk->part0, blk_rq_bytes(rq)))
1188 return BLK_STS_IOERR;
1189
1190 if (blk_queue_io_stat(q))
1191 blk_account_io_start(rq, true);
1192
1193
1194
1195
1196
1197
1198 return blk_mq_request_issue_directly(rq, true);
1199}
1200EXPORT_SYMBOL_GPL(blk_insert_cloned_request);
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215unsigned int blk_rq_err_bytes(const struct request *rq)
1216{
1217 unsigned int ff = rq->cmd_flags & REQ_FAILFAST_MASK;
1218 unsigned int bytes = 0;
1219 struct bio *bio;
1220
1221 if (!(rq->rq_flags & RQF_MIXED_MERGE))
1222 return blk_rq_bytes(rq);
1223
1224
1225
1226
1227
1228
1229
1230
1231 for (bio = rq->bio; bio; bio = bio->bi_next) {
1232 if ((bio->bi_opf & ff) != ff)
1233 break;
1234 bytes += bio->bi_iter.bi_size;
1235 }
1236
1237
1238 BUG_ON(blk_rq_bytes(rq) && !bytes);
1239 return bytes;
1240}
1241EXPORT_SYMBOL_GPL(blk_rq_err_bytes);
1242
1243void blk_account_io_completion(struct request *req, unsigned int bytes)
1244{
1245 if (blk_do_io_stat(req)) {
1246 const int sgrp = op_stat_group(req_op(req));
1247 struct hd_struct *part;
1248
1249 part_stat_lock();
1250 part = req->part;
1251 part_stat_add(part, sectors[sgrp], bytes >> 9);
1252 part_stat_unlock();
1253 }
1254}
1255
1256void blk_account_io_done(struct request *req, u64 now)
1257{
1258
1259
1260
1261
1262
1263 if (blk_do_io_stat(req) && !(req->rq_flags & RQF_FLUSH_SEQ)) {
1264 const int sgrp = op_stat_group(req_op(req));
1265 struct hd_struct *part;
1266
1267 part_stat_lock();
1268 part = req->part;
1269
1270 update_io_ticks(part, jiffies);
1271 part_stat_inc(part, ios[sgrp]);
1272 part_stat_add(part, nsecs[sgrp], now - req->start_time_ns);
1273 part_stat_add(part, time_in_queue, nsecs_to_jiffies64(now - req->start_time_ns));
1274 part_dec_in_flight(req->q, part, rq_data_dir(req));
1275
1276 hd_struct_put(part);
1277 part_stat_unlock();
1278 }
1279}
1280
1281void blk_account_io_start(struct request *rq, bool new_io)
1282{
1283 struct hd_struct *part;
1284 int rw = rq_data_dir(rq);
1285
1286 if (!blk_do_io_stat(rq))
1287 return;
1288
1289 part_stat_lock();
1290
1291 if (!new_io) {
1292 part = rq->part;
1293 part_stat_inc(part, merges[rw]);
1294 } else {
1295 part = disk_map_sector_rcu(rq->rq_disk, blk_rq_pos(rq));
1296 if (!hd_struct_try_get(part)) {
1297
1298
1299
1300
1301
1302
1303
1304
1305 part = &rq->rq_disk->part0;
1306 hd_struct_get(part);
1307 }
1308 part_inc_in_flight(rq->q, part, rw);
1309 rq->part = part;
1310 }
1311
1312 update_io_ticks(part, jiffies);
1313
1314 part_stat_unlock();
1315}
1316
1317
1318
1319
1320
1321void blk_steal_bios(struct bio_list *list, struct request *rq)
1322{
1323 if (rq->bio) {
1324 if (list->tail)
1325 list->tail->bi_next = rq->bio;
1326 else
1327 list->head = rq->bio;
1328 list->tail = rq->biotail;
1329
1330 rq->bio = NULL;
1331 rq->biotail = NULL;
1332 }
1333
1334 rq->__data_len = 0;
1335}
1336EXPORT_SYMBOL_GPL(blk_steal_bios);
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364bool blk_update_request(struct request *req, blk_status_t error,
1365 unsigned int nr_bytes)
1366{
1367 int total_bytes;
1368
1369 trace_block_rq_complete(req, blk_status_to_errno(error), nr_bytes);
1370
1371 if (!req->bio)
1372 return false;
1373
1374 if (unlikely(error && !blk_rq_is_passthrough(req) &&
1375 !(req->rq_flags & RQF_QUIET)))
1376 print_req_error(req, error);
1377
1378 blk_account_io_completion(req, nr_bytes);
1379
1380 total_bytes = 0;
1381 while (req->bio) {
1382 struct bio *bio = req->bio;
1383 unsigned bio_bytes = min(bio->bi_iter.bi_size, nr_bytes);
1384
1385 if (bio_bytes == bio->bi_iter.bi_size)
1386 req->bio = bio->bi_next;
1387
1388
1389 bio_clear_flag(bio, BIO_TRACE_COMPLETION);
1390 req_bio_endio(req, bio, bio_bytes, error);
1391
1392 total_bytes += bio_bytes;
1393 nr_bytes -= bio_bytes;
1394
1395 if (!nr_bytes)
1396 break;
1397 }
1398
1399
1400
1401
1402 if (!req->bio) {
1403
1404
1405
1406
1407
1408 req->__data_len = 0;
1409 return false;
1410 }
1411
1412 req->__data_len -= total_bytes;
1413
1414
1415 if (!blk_rq_is_passthrough(req))
1416 req->__sector += total_bytes >> 9;
1417
1418
1419 if (req->rq_flags & RQF_MIXED_MERGE) {
1420 req->cmd_flags &= ~REQ_FAILFAST_MASK;
1421 req->cmd_flags |= req->bio->bi_opf & REQ_FAILFAST_MASK;
1422 }
1423
1424 if (!(req->rq_flags & RQF_SPECIAL_PAYLOAD)) {
1425
1426
1427
1428
1429 if (blk_rq_bytes(req) < blk_rq_cur_bytes(req)) {
1430 blk_dump_rq_flags(req, "request botched");
1431 req->__data_len = blk_rq_cur_bytes(req);
1432 }
1433
1434
1435 blk_recalc_rq_segments(req);
1436 }
1437
1438 return true;
1439}
1440EXPORT_SYMBOL_GPL(blk_update_request);
1441
1442void blk_rq_bio_prep(struct request_queue *q, struct request *rq,
1443 struct bio *bio)
1444{
1445 if (bio_has_data(bio))
1446 rq->nr_phys_segments = bio_phys_segments(q, bio);
1447 else if (bio_op(bio) == REQ_OP_DISCARD)
1448 rq->nr_phys_segments = 1;
1449
1450 rq->__data_len = bio->bi_iter.bi_size;
1451 rq->bio = rq->biotail = bio;
1452
1453 if (bio->bi_disk)
1454 rq->rq_disk = bio->bi_disk;
1455}
1456
1457#if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE
1458
1459
1460
1461
1462
1463
1464
1465void rq_flush_dcache_pages(struct request *rq)
1466{
1467 struct req_iterator iter;
1468 struct bio_vec bvec;
1469
1470 rq_for_each_segment(bvec, rq, iter)
1471 flush_dcache_page(bvec.bv_page);
1472}
1473EXPORT_SYMBOL_GPL(rq_flush_dcache_pages);
1474#endif
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495int blk_lld_busy(struct request_queue *q)
1496{
1497 if (queue_is_mq(q) && q->mq_ops->busy)
1498 return q->mq_ops->busy(q);
1499
1500 return 0;
1501}
1502EXPORT_SYMBOL_GPL(blk_lld_busy);
1503
1504
1505
1506
1507
1508
1509
1510
1511void blk_rq_unprep_clone(struct request *rq)
1512{
1513 struct bio *bio;
1514
1515 while ((bio = rq->bio) != NULL) {
1516 rq->bio = bio->bi_next;
1517
1518 bio_put(bio);
1519 }
1520}
1521EXPORT_SYMBOL_GPL(blk_rq_unprep_clone);
1522
1523
1524
1525
1526
1527static void __blk_rq_prep_clone(struct request *dst, struct request *src)
1528{
1529 dst->__sector = blk_rq_pos(src);
1530 dst->__data_len = blk_rq_bytes(src);
1531 if (src->rq_flags & RQF_SPECIAL_PAYLOAD) {
1532 dst->rq_flags |= RQF_SPECIAL_PAYLOAD;
1533 dst->special_vec = src->special_vec;
1534 }
1535 dst->nr_phys_segments = src->nr_phys_segments;
1536 dst->ioprio = src->ioprio;
1537 dst->extra_len = src->extra_len;
1538}
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559int blk_rq_prep_clone(struct request *rq, struct request *rq_src,
1560 struct bio_set *bs, gfp_t gfp_mask,
1561 int (*bio_ctr)(struct bio *, struct bio *, void *),
1562 void *data)
1563{
1564 struct bio *bio, *bio_src;
1565
1566 if (!bs)
1567 bs = &fs_bio_set;
1568
1569 __rq_for_each_bio(bio_src, rq_src) {
1570 bio = bio_clone_fast(bio_src, gfp_mask, bs);
1571 if (!bio)
1572 goto free_and_out;
1573
1574 if (bio_ctr && bio_ctr(bio, bio_src, data))
1575 goto free_and_out;
1576
1577 if (rq->bio) {
1578 rq->biotail->bi_next = bio;
1579 rq->biotail = bio;
1580 } else
1581 rq->bio = rq->biotail = bio;
1582 }
1583
1584 __blk_rq_prep_clone(rq, rq_src);
1585
1586 return 0;
1587
1588free_and_out:
1589 if (bio)
1590 bio_put(bio);
1591 blk_rq_unprep_clone(rq);
1592
1593 return -ENOMEM;
1594}
1595EXPORT_SYMBOL_GPL(blk_rq_prep_clone);
1596
1597int kblockd_schedule_work(struct work_struct *work)
1598{
1599 return queue_work(kblockd_workqueue, work);
1600}
1601EXPORT_SYMBOL(kblockd_schedule_work);
1602
1603int kblockd_schedule_work_on(int cpu, struct work_struct *work)
1604{
1605 return queue_work_on(cpu, kblockd_workqueue, work);
1606}
1607EXPORT_SYMBOL(kblockd_schedule_work_on);
1608
1609int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork,
1610 unsigned long delay)
1611{
1612 return mod_delayed_work_on(cpu, kblockd_workqueue, dwork, delay);
1613}
1614EXPORT_SYMBOL(kblockd_mod_delayed_work_on);
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639void blk_start_plug(struct blk_plug *plug)
1640{
1641 struct task_struct *tsk = current;
1642
1643
1644
1645
1646 if (tsk->plug)
1647 return;
1648
1649 INIT_LIST_HEAD(&plug->mq_list);
1650 INIT_LIST_HEAD(&plug->cb_list);
1651 plug->rq_count = 0;
1652 plug->multiple_queues = false;
1653
1654
1655
1656
1657
1658 tsk->plug = plug;
1659}
1660EXPORT_SYMBOL(blk_start_plug);
1661
1662static void flush_plug_callbacks(struct blk_plug *plug, bool from_schedule)
1663{
1664 LIST_HEAD(callbacks);
1665
1666 while (!list_empty(&plug->cb_list)) {
1667 list_splice_init(&plug->cb_list, &callbacks);
1668
1669 while (!list_empty(&callbacks)) {
1670 struct blk_plug_cb *cb = list_first_entry(&callbacks,
1671 struct blk_plug_cb,
1672 list);
1673 list_del(&cb->list);
1674 cb->callback(cb, from_schedule);
1675 }
1676 }
1677}
1678
1679struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug, void *data,
1680 int size)
1681{
1682 struct blk_plug *plug = current->plug;
1683 struct blk_plug_cb *cb;
1684
1685 if (!plug)
1686 return NULL;
1687
1688 list_for_each_entry(cb, &plug->cb_list, list)
1689 if (cb->callback == unplug && cb->data == data)
1690 return cb;
1691
1692
1693 BUG_ON(size < sizeof(*cb));
1694 cb = kzalloc(size, GFP_ATOMIC);
1695 if (cb) {
1696 cb->data = data;
1697 cb->callback = unplug;
1698 list_add(&cb->list, &plug->cb_list);
1699 }
1700 return cb;
1701}
1702EXPORT_SYMBOL(blk_check_plugged);
1703
1704void blk_flush_plug_list(struct blk_plug *plug, bool from_schedule)
1705{
1706 flush_plug_callbacks(plug, from_schedule);
1707
1708 if (!list_empty(&plug->mq_list))
1709 blk_mq_flush_plug_list(plug, from_schedule);
1710}
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722void blk_finish_plug(struct blk_plug *plug)
1723{
1724 if (plug != current->plug)
1725 return;
1726 blk_flush_plug_list(plug, false);
1727
1728 current->plug = NULL;
1729}
1730EXPORT_SYMBOL(blk_finish_plug);
1731
1732int __init blk_dev_init(void)
1733{
1734 BUILD_BUG_ON(REQ_OP_LAST >= (1 << REQ_OP_BITS));
1735 BUILD_BUG_ON(REQ_OP_BITS + REQ_FLAG_BITS > 8 *
1736 FIELD_SIZEOF(struct request, cmd_flags));
1737 BUILD_BUG_ON(REQ_OP_BITS + REQ_FLAG_BITS > 8 *
1738 FIELD_SIZEOF(struct bio, bi_opf));
1739
1740
1741 kblockd_workqueue = alloc_workqueue("kblockd",
1742 WQ_MEM_RECLAIM | WQ_HIGHPRI, 0);
1743 if (!kblockd_workqueue)
1744 panic("Failed to create kblockd\n");
1745
1746 blk_requestq_cachep = kmem_cache_create("request_queue",
1747 sizeof(struct request_queue), 0, SLAB_PANIC, NULL);
1748
1749#ifdef CONFIG_DEBUG_FS
1750 blk_debugfs_root = debugfs_create_dir("block", NULL);
1751#endif
1752
1753 return 0;
1754}
1755