Commit | Line | Data |
---|---|---|
3dcf60bc | 1 | // SPDX-License-Identifier: GPL-2.0 |
1da177e4 | 2 | /* |
1da177e4 LT |
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> | |
6728cb0e JA |
7 | * kernel-doc documentation started by NeilBrown <neilb@cse.unsw.edu.au> |
8 | * - July2000 | |
1da177e4 LT |
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 | */ | |
1da177e4 LT |
15 | #include <linux/kernel.h> |
16 | #include <linux/module.h> | |
1da177e4 LT |
17 | #include <linux/bio.h> |
18 | #include <linux/blkdev.h> | |
52abca64 | 19 | #include <linux/blk-pm.h> |
fe45e630 | 20 | #include <linux/blk-integrity.h> |
1da177e4 LT |
21 | #include <linux/highmem.h> |
22 | #include <linux/mm.h> | |
cee9a0c4 | 23 | #include <linux/pagemap.h> |
1da177e4 LT |
24 | #include <linux/kernel_stat.h> |
25 | #include <linux/string.h> | |
26 | #include <linux/init.h> | |
1da177e4 LT |
27 | #include <linux/completion.h> |
28 | #include <linux/slab.h> | |
29 | #include <linux/swap.h> | |
30 | #include <linux/writeback.h> | |
faccbd4b | 31 | #include <linux/task_io_accounting_ops.h> |
c17bb495 | 32 | #include <linux/fault-inject.h> |
73c10101 | 33 | #include <linux/list_sort.h> |
e3c78ca5 | 34 | #include <linux/delay.h> |
aaf7c680 | 35 | #include <linux/ratelimit.h> |
6c954667 | 36 | #include <linux/pm_runtime.h> |
54d4e6ab | 37 | #include <linux/t10-pi.h> |
18fbda91 | 38 | #include <linux/debugfs.h> |
30abb3a6 | 39 | #include <linux/bpf.h> |
82d981d4 | 40 | #include <linux/part_stat.h> |
71ac860a | 41 | #include <linux/sched/sysctl.h> |
a892c8d5 | 42 | #include <linux/blk-crypto.h> |
55782138 LZ |
43 | |
44 | #define CREATE_TRACE_POINTS | |
45 | #include <trace/events/block.h> | |
1da177e4 | 46 | |
8324aa91 | 47 | #include "blk.h" |
2aa7745b | 48 | #include "blk-mq-sched.h" |
bca6b067 | 49 | #include "blk-pm.h" |
672fdcf0 | 50 | #include "blk-cgroup.h" |
a7b36ee6 | 51 | #include "blk-throttle.h" |
8324aa91 | 52 | |
18fbda91 | 53 | struct dentry *blk_debugfs_root; |
18fbda91 | 54 | |
d07335e5 | 55 | EXPORT_TRACEPOINT_SYMBOL_GPL(block_bio_remap); |
b0da3f0d | 56 | EXPORT_TRACEPOINT_SYMBOL_GPL(block_rq_remap); |
0a82a8d1 | 57 | EXPORT_TRACEPOINT_SYMBOL_GPL(block_bio_complete); |
3291fa57 | 58 | EXPORT_TRACEPOINT_SYMBOL_GPL(block_split); |
cbae8d45 | 59 | EXPORT_TRACEPOINT_SYMBOL_GPL(block_unplug); |
b357e4a6 | 60 | EXPORT_TRACEPOINT_SYMBOL_GPL(block_rq_insert); |
0bfc2455 | 61 | |
2bd85221 | 62 | static DEFINE_IDA(blk_queue_ida); |
a73f730d | 63 | |
1da177e4 LT |
64 | /* |
65 | * For queue allocation | |
66 | */ | |
2bd85221 | 67 | static struct kmem_cache *blk_requestq_cachep; |
1da177e4 | 68 | |
1da177e4 LT |
69 | /* |
70 | * Controlling structure to kblockd | |
71 | */ | |
ff856bad | 72 | static struct workqueue_struct *kblockd_workqueue; |
1da177e4 | 73 | |
8814ce8a BVA |
74 | /** |
75 | * blk_queue_flag_set - atomically set a queue flag | |
76 | * @flag: flag to be set | |
77 | * @q: request queue | |
78 | */ | |
79 | void blk_queue_flag_set(unsigned int flag, struct request_queue *q) | |
80 | { | |
57d74df9 | 81 | set_bit(flag, &q->queue_flags); |
8814ce8a BVA |
82 | } |
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 | void blk_queue_flag_clear(unsigned int flag, struct request_queue *q) | |
91 | { | |
57d74df9 | 92 | clear_bit(flag, &q->queue_flags); |
8814ce8a BVA |
93 | } |
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 | bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q) | |
105 | { | |
57d74df9 | 106 | return test_and_set_bit(flag, &q->queue_flags); |
8814ce8a BVA |
107 | } |
108 | EXPORT_SYMBOL_GPL(blk_queue_flag_test_and_set); | |
109 | ||
e47bc4ed CK |
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), | |
6e33dbf2 | 118 | REQ_OP_NAME(ZONE_RESET_ALL), |
6c1b1da5 AJ |
119 | REQ_OP_NAME(ZONE_OPEN), |
120 | REQ_OP_NAME(ZONE_CLOSE), | |
121 | REQ_OP_NAME(ZONE_FINISH), | |
0512a75b | 122 | REQ_OP_NAME(ZONE_APPEND), |
e47bc4ed | 123 | REQ_OP_NAME(WRITE_ZEROES), |
e47bc4ed CK |
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 | */ | |
77e7ffd7 | 137 | inline const char *blk_op_str(enum req_op op) |
e47bc4ed CK |
138 | { |
139 | const char *op_str = "UNKNOWN"; | |
140 | ||
141 | if (op < ARRAY_SIZE(blk_op_name) && blk_op_name[op]) | |
142 | op_str = blk_op_name[op]; | |
143 | ||
144 | return op_str; | |
145 | } | |
146 | EXPORT_SYMBOL_GPL(blk_op_str); | |
147 | ||
2a842aca CH |
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" }, | |
7ba15083 | 158 | [BLK_STS_RESV_CONFLICT] = { -EBADE, "reservation conflict" }, |
2a842aca CH |
159 | [BLK_STS_MEDIUM] = { -ENODATA, "critical medium" }, |
160 | [BLK_STS_PROTECTION] = { -EILSEQ, "protection" }, | |
161 | [BLK_STS_RESOURCE] = { -ENOMEM, "kernel resource" }, | |
86ff7c2a | 162 | [BLK_STS_DEV_RESOURCE] = { -EBUSY, "device resource" }, |
03a07c92 | 163 | [BLK_STS_AGAIN] = { -EAGAIN, "nonblocking retry" }, |
7d32c027 | 164 | [BLK_STS_OFFLINE] = { -ENODEV, "device offline" }, |
2a842aca | 165 | |
4e4cbee9 CH |
166 | /* device mapper special case, should not leak out: */ |
167 | [BLK_STS_DM_REQUEUE] = { -EREMCHG, "dm internal retry" }, | |
168 | ||
3b481d91 KB |
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 | ||
dffc480d DLM |
173 | /* Command duration limit device-side timeout */ |
174 | [BLK_STS_DURATION_LIMIT] = { -ETIME, "duration limit exceeded" }, | |
175 | ||
2a842aca CH |
176 | /* everything else not covered above: */ |
177 | [BLK_STS_IOERR] = { -EIO, "I/O" }, | |
178 | }; | |
179 | ||
180 | blk_status_t errno_to_blk_status(int errno) | |
181 | { | |
182 | int i; | |
183 | ||
184 | for (i = 0; i < ARRAY_SIZE(blk_errors); i++) { | |
185 | if (blk_errors[i].errno == errno) | |
186 | return (__force blk_status_t)i; | |
187 | } | |
188 | ||
189 | return BLK_STS_IOERR; | |
190 | } | |
191 | EXPORT_SYMBOL_GPL(errno_to_blk_status); | |
192 | ||
193 | int blk_status_to_errno(blk_status_t status) | |
194 | { | |
195 | int idx = (__force int)status; | |
196 | ||
34bd9c1c | 197 | if (WARN_ON_ONCE(idx >= ARRAY_SIZE(blk_errors))) |
2a842aca CH |
198 | return -EIO; |
199 | return blk_errors[idx].errno; | |
200 | } | |
201 | EXPORT_SYMBOL_GPL(blk_status_to_errno); | |
202 | ||
0d7a29a2 | 203 | const char *blk_status_to_str(blk_status_t status) |
2a842aca CH |
204 | { |
205 | int idx = (__force int)status; | |
206 | ||
34bd9c1c | 207 | if (WARN_ON_ONCE(idx >= ARRAY_SIZE(blk_errors))) |
0d7a29a2 CH |
208 | return "<null>"; |
209 | return blk_errors[idx].name; | |
2a842aca CH |
210 | } |
211 | ||
1da177e4 LT |
212 | /** |
213 | * blk_sync_queue - cancel any pending callbacks on a queue | |
214 | * @q: the queue | |
215 | * | |
216 | * Description: | |
217 | * The block layer may perform asynchronous callback activity | |
218 | * on a queue, such as calling the unplug function after a timeout. | |
219 | * A block device may call blk_sync_queue to ensure that any | |
220 | * such activity is cancelled, thus allowing it to release resources | |
59c51591 | 221 | * that the callbacks might use. The caller must already have made sure |
c62b37d9 | 222 | * that its ->submit_bio will not re-add plugging prior to calling |
1da177e4 LT |
223 | * this function. |
224 | * | |
da527770 | 225 | * This function does not cancel any asynchronous activity arising |
da3dae54 | 226 | * out of elevator or throttling code. That would require elevator_exit() |
5efd6113 | 227 | * and blkcg_exit_queue() to be called with queue lock initialized. |
da527770 | 228 | * |
1da177e4 LT |
229 | */ |
230 | void blk_sync_queue(struct request_queue *q) | |
231 | { | |
70ed28b9 | 232 | del_timer_sync(&q->timeout); |
4e9b6f20 | 233 | cancel_work_sync(&q->timeout_work); |
1da177e4 LT |
234 | } |
235 | EXPORT_SYMBOL(blk_sync_queue); | |
236 | ||
c9254f2d | 237 | /** |
cd84a62e | 238 | * blk_set_pm_only - increment pm_only counter |
c9254f2d | 239 | * @q: request queue pointer |
c9254f2d | 240 | */ |
cd84a62e | 241 | void blk_set_pm_only(struct request_queue *q) |
c9254f2d | 242 | { |
cd84a62e | 243 | atomic_inc(&q->pm_only); |
c9254f2d | 244 | } |
cd84a62e | 245 | EXPORT_SYMBOL_GPL(blk_set_pm_only); |
c9254f2d | 246 | |
cd84a62e | 247 | void blk_clear_pm_only(struct request_queue *q) |
c9254f2d | 248 | { |
cd84a62e BVA |
249 | int pm_only; |
250 | ||
251 | pm_only = atomic_dec_return(&q->pm_only); | |
252 | WARN_ON_ONCE(pm_only < 0); | |
253 | if (pm_only == 0) | |
254 | wake_up_all(&q->mq_freeze_wq); | |
c9254f2d | 255 | } |
cd84a62e | 256 | EXPORT_SYMBOL_GPL(blk_clear_pm_only); |
c9254f2d | 257 | |
2bd85221 CH |
258 | static void blk_free_queue_rcu(struct rcu_head *rcu_head) |
259 | { | |
d36a9ea5 ML |
260 | struct request_queue *q = container_of(rcu_head, |
261 | struct request_queue, rcu_head); | |
262 | ||
263 | percpu_ref_exit(&q->q_usage_counter); | |
264 | kmem_cache_free(blk_requestq_cachep, q); | |
2bd85221 CH |
265 | } |
266 | ||
267 | static void blk_free_queue(struct request_queue *q) | |
268 | { | |
2bd85221 | 269 | blk_free_queue_stats(q->stats); |
2bd85221 CH |
270 | if (queue_is_mq(q)) |
271 | blk_mq_release(q); | |
272 | ||
273 | ida_free(&blk_queue_ida, q->id); | |
274 | call_rcu(&q->rcu_head, blk_free_queue_rcu); | |
275 | } | |
276 | ||
b5bd357c LC |
277 | /** |
278 | * blk_put_queue - decrement the request_queue refcount | |
279 | * @q: the request_queue structure to decrement the refcount for | |
280 | * | |
2bd85221 CH |
281 | * Decrements the refcount of the request_queue and free it when the refcount |
282 | * reaches 0. | |
b5bd357c | 283 | */ |
165125e1 | 284 | void blk_put_queue(struct request_queue *q) |
483f4afc | 285 | { |
2bd85221 CH |
286 | if (refcount_dec_and_test(&q->refs)) |
287 | blk_free_queue(q); | |
483f4afc | 288 | } |
d86e0e83 | 289 | EXPORT_SYMBOL(blk_put_queue); |
483f4afc | 290 | |
8e141f9e | 291 | void blk_queue_start_drain(struct request_queue *q) |
aed3ea94 | 292 | { |
d3cfb2a0 ML |
293 | /* |
294 | * When queue DYING flag is set, we need to block new req | |
295 | * entering queue, so we call blk_freeze_queue_start() to | |
296 | * prevent I/O from crossing blk_queue_enter(). | |
297 | */ | |
298 | blk_freeze_queue_start(q); | |
344e9ffc | 299 | if (queue_is_mq(q)) |
aed3ea94 | 300 | blk_mq_wake_waiters(q); |
055f6e18 ML |
301 | /* Make blk_queue_enter() reexamine the DYING flag. */ |
302 | wake_up_all(&q->mq_freeze_wq); | |
aed3ea94 | 303 | } |
8e141f9e | 304 | |
3a0a5299 BVA |
305 | /** |
306 | * blk_queue_enter() - try to increase q->q_usage_counter | |
307 | * @q: request queue pointer | |
a4d34da7 | 308 | * @flags: BLK_MQ_REQ_NOWAIT and/or BLK_MQ_REQ_PM |
3a0a5299 | 309 | */ |
9a95e4ef | 310 | int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags) |
3ef28e83 | 311 | { |
a4d34da7 | 312 | const bool pm = flags & BLK_MQ_REQ_PM; |
3a0a5299 | 313 | |
1f14a098 | 314 | while (!blk_try_enter_queue(q, pm)) { |
3a0a5299 | 315 | if (flags & BLK_MQ_REQ_NOWAIT) |
56f99b8d | 316 | return -EAGAIN; |
3ef28e83 | 317 | |
5ed61d3f | 318 | /* |
1f14a098 CH |
319 | * read pair of barrier in blk_freeze_queue_start(), we need to |
320 | * order reading __PERCPU_REF_DEAD flag of .q_usage_counter and | |
321 | * reading .mq_freeze_depth or queue dying flag, otherwise the | |
322 | * following wait may never return if the two reads are | |
323 | * reordered. | |
5ed61d3f ML |
324 | */ |
325 | smp_rmb(); | |
1dc3039b | 326 | wait_event(q->mq_freeze_wq, |
7996a8b5 | 327 | (!q->mq_freeze_depth && |
52abca64 | 328 | blk_pm_resume_queue(pm, q)) || |
1dc3039b | 329 | blk_queue_dying(q)); |
3ef28e83 DW |
330 | if (blk_queue_dying(q)) |
331 | return -ENODEV; | |
3ef28e83 | 332 | } |
1f14a098 CH |
333 | |
334 | return 0; | |
3ef28e83 DW |
335 | } |
336 | ||
c98cb5bb | 337 | int __bio_queue_enter(struct request_queue *q, struct bio *bio) |
accea322 | 338 | { |
a6741536 | 339 | while (!blk_try_enter_queue(q, false)) { |
eab4e027 PB |
340 | struct gendisk *disk = bio->bi_bdev->bd_disk; |
341 | ||
a6741536 | 342 | if (bio->bi_opf & REQ_NOWAIT) { |
8e141f9e | 343 | if (test_bit(GD_DEAD, &disk->state)) |
a6741536 | 344 | goto dead; |
accea322 | 345 | bio_wouldblock_error(bio); |
56f99b8d | 346 | return -EAGAIN; |
a6741536 CH |
347 | } |
348 | ||
349 | /* | |
350 | * read pair of barrier in blk_freeze_queue_start(), we need to | |
351 | * order reading __PERCPU_REF_DEAD flag of .q_usage_counter and | |
352 | * reading .mq_freeze_depth or queue dying flag, otherwise the | |
353 | * following wait may never return if the two reads are | |
354 | * reordered. | |
355 | */ | |
356 | smp_rmb(); | |
357 | wait_event(q->mq_freeze_wq, | |
358 | (!q->mq_freeze_depth && | |
359 | blk_pm_resume_queue(false, q)) || | |
8e141f9e CH |
360 | test_bit(GD_DEAD, &disk->state)); |
361 | if (test_bit(GD_DEAD, &disk->state)) | |
a6741536 | 362 | goto dead; |
accea322 CH |
363 | } |
364 | ||
a6741536 CH |
365 | return 0; |
366 | dead: | |
367 | bio_io_error(bio); | |
368 | return -ENODEV; | |
accea322 CH |
369 | } |
370 | ||
3ef28e83 DW |
371 | void blk_queue_exit(struct request_queue *q) |
372 | { | |
373 | percpu_ref_put(&q->q_usage_counter); | |
374 | } | |
375 | ||
376 | static void blk_queue_usage_counter_release(struct percpu_ref *ref) | |
377 | { | |
378 | struct request_queue *q = | |
379 | container_of(ref, struct request_queue, q_usage_counter); | |
380 | ||
381 | wake_up_all(&q->mq_freeze_wq); | |
382 | } | |
383 | ||
bca237a5 | 384 | static void blk_rq_timed_out_timer(struct timer_list *t) |
287922eb | 385 | { |
bca237a5 | 386 | struct request_queue *q = from_timer(q, t, timeout); |
287922eb CH |
387 | |
388 | kblockd_schedule_work(&q->timeout_work); | |
389 | } | |
390 | ||
2e3c18d0 TH |
391 | static void blk_timeout_work(struct work_struct *work) |
392 | { | |
393 | } | |
394 | ||
80bd4a7a | 395 | struct request_queue *blk_alloc_queue(int node_id) |
1946089a | 396 | { |
165125e1 | 397 | struct request_queue *q; |
1946089a | 398 | |
80bd4a7a CH |
399 | q = kmem_cache_alloc_node(blk_requestq_cachep, GFP_KERNEL | __GFP_ZERO, |
400 | node_id); | |
1da177e4 LT |
401 | if (!q) |
402 | return NULL; | |
403 | ||
cbf62af3 | 404 | q->last_merge = NULL; |
cbf62af3 | 405 | |
798f2a6f | 406 | q->id = ida_alloc(&blk_queue_ida, GFP_KERNEL); |
a73f730d | 407 | if (q->id < 0) |
80bd4a7a | 408 | goto fail_q; |
a73f730d | 409 | |
a83b576c JA |
410 | q->stats = blk_alloc_queue_stats(); |
411 | if (!q->stats) | |
46754bd0 | 412 | goto fail_id; |
a83b576c | 413 | |
5151412d | 414 | q->node = node_id; |
0989a025 | 415 | |
079a2e3e | 416 | atomic_set(&q->nr_active_requests_shared_tags, 0); |
bccf5e26 | 417 | |
bca237a5 | 418 | timer_setup(&q->timeout, blk_rq_timed_out_timer, 0); |
2e3c18d0 | 419 | INIT_WORK(&q->timeout_work, blk_timeout_work); |
a612fddf | 420 | INIT_LIST_HEAD(&q->icq_list); |
483f4afc | 421 | |
2bd85221 | 422 | refcount_set(&q->refs, 1); |
85e0cbbb | 423 | mutex_init(&q->debugfs_mutex); |
483f4afc | 424 | mutex_init(&q->sysfs_lock); |
cecf5d87 | 425 | mutex_init(&q->sysfs_dir_lock); |
a13bd91b | 426 | mutex_init(&q->rq_qos_mutex); |
0d945c1f | 427 | spin_lock_init(&q->queue_lock); |
c94a96ac | 428 | |
320ae51f | 429 | init_waitqueue_head(&q->mq_freeze_wq); |
7996a8b5 | 430 | mutex_init(&q->mq_freeze_lock); |
320ae51f | 431 | |
3ef28e83 DW |
432 | /* |
433 | * Init percpu_ref in atomic mode so that it's faster to shutdown. | |
434 | * See blk_register_queue() for details. | |
435 | */ | |
436 | if (percpu_ref_init(&q->q_usage_counter, | |
437 | blk_queue_usage_counter_release, | |
438 | PERCPU_REF_INIT_ATOMIC, GFP_KERNEL)) | |
edb0872f | 439 | goto fail_stats; |
f51b802c | 440 | |
3d745ea5 | 441 | blk_set_default_limits(&q->limits); |
d2a27964 | 442 | q->nr_requests = BLKDEV_DEFAULT_RQ; |
3d745ea5 | 443 | |
1da177e4 | 444 | return q; |
a73f730d | 445 | |
a83b576c | 446 | fail_stats: |
edb0872f | 447 | blk_free_queue_stats(q->stats); |
a73f730d | 448 | fail_id: |
798f2a6f | 449 | ida_free(&blk_queue_ida, q->id); |
a73f730d | 450 | fail_q: |
80bd4a7a | 451 | kmem_cache_free(blk_requestq_cachep, q); |
a73f730d | 452 | return NULL; |
1da177e4 | 453 | } |
1da177e4 | 454 | |
b5bd357c LC |
455 | /** |
456 | * blk_get_queue - increment the request_queue refcount | |
457 | * @q: the request_queue structure to increment the refcount for | |
458 | * | |
459 | * Increment the refcount of the request_queue kobject. | |
763b5892 LC |
460 | * |
461 | * Context: Any context. | |
b5bd357c | 462 | */ |
09ac46c4 | 463 | bool blk_get_queue(struct request_queue *q) |
1da177e4 | 464 | { |
828b5f01 CH |
465 | if (unlikely(blk_queue_dying(q))) |
466 | return false; | |
2bd85221 | 467 | refcount_inc(&q->refs); |
828b5f01 | 468 | return true; |
1da177e4 | 469 | } |
d86e0e83 | 470 | EXPORT_SYMBOL(blk_get_queue); |
1da177e4 | 471 | |
c17bb495 AM |
472 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
473 | ||
474 | static DECLARE_FAULT_ATTR(fail_make_request); | |
475 | ||
476 | static int __init setup_fail_make_request(char *str) | |
477 | { | |
478 | return setup_fault_attr(&fail_make_request, str); | |
479 | } | |
480 | __setup("fail_make_request=", setup_fail_make_request); | |
481 | ||
06c8c691 | 482 | bool should_fail_request(struct block_device *part, unsigned int bytes) |
c17bb495 | 483 | { |
8446fe92 | 484 | return part->bd_make_it_fail && should_fail(&fail_make_request, bytes); |
c17bb495 AM |
485 | } |
486 | ||
487 | static int __init fail_make_request_debugfs(void) | |
488 | { | |
dd48c085 AM |
489 | struct dentry *dir = fault_create_debugfs_attr("fail_make_request", |
490 | NULL, &fail_make_request); | |
491 | ||
21f9fcd8 | 492 | return PTR_ERR_OR_ZERO(dir); |
c17bb495 AM |
493 | } |
494 | ||
495 | late_initcall(fail_make_request_debugfs); | |
c17bb495 AM |
496 | #endif /* CONFIG_FAIL_MAKE_REQUEST */ |
497 | ||
bdb7d420 | 498 | static inline void bio_check_ro(struct bio *bio) |
721c7fc7 | 499 | { |
2f9f6221 | 500 | if (op_is_write(bio_op(bio)) && bdev_read_only(bio->bi_bdev)) { |
8b2ded1c | 501 | if (op_is_flush(bio->bi_opf) && !bio_sectors(bio)) |
bdb7d420 | 502 | return; |
57e95e46 CH |
503 | pr_warn("Trying to write to read-only block-device %pg\n", |
504 | bio->bi_bdev); | |
a32e236e | 505 | /* Older lvm-tools actually trigger this */ |
721c7fc7 | 506 | } |
721c7fc7 ID |
507 | } |
508 | ||
30abb3a6 HM |
509 | static noinline int should_fail_bio(struct bio *bio) |
510 | { | |
309dca30 | 511 | if (should_fail_request(bdev_whole(bio->bi_bdev), bio->bi_iter.bi_size)) |
30abb3a6 HM |
512 | return -EIO; |
513 | return 0; | |
514 | } | |
515 | ALLOW_ERROR_INJECTION(should_fail_bio, ERRNO); | |
516 | ||
52c5e62d CH |
517 | /* |
518 | * Check whether this bio extends beyond the end of the device or partition. | |
519 | * This may well happen - the kernel calls bread() without checking the size of | |
520 | * the device, e.g., when mounting a file system. | |
521 | */ | |
2f9f6221 | 522 | static inline int bio_check_eod(struct bio *bio) |
52c5e62d | 523 | { |
2f9f6221 | 524 | sector_t maxsector = bdev_nr_sectors(bio->bi_bdev); |
52c5e62d CH |
525 | unsigned int nr_sectors = bio_sectors(bio); |
526 | ||
3eb96946 | 527 | if (nr_sectors && |
52c5e62d CH |
528 | (nr_sectors > maxsector || |
529 | bio->bi_iter.bi_sector > maxsector - nr_sectors)) { | |
ad740780 | 530 | pr_info_ratelimited("%s: attempt to access beyond end of device\n" |
069adbac CH |
531 | "%pg: rw=%d, sector=%llu, nr_sectors = %u limit=%llu\n", |
532 | current->comm, bio->bi_bdev, bio->bi_opf, | |
533 | bio->bi_iter.bi_sector, nr_sectors, maxsector); | |
52c5e62d CH |
534 | return -EIO; |
535 | } | |
536 | return 0; | |
537 | } | |
538 | ||
74d46992 CH |
539 | /* |
540 | * Remap block n of partition p to block n+start(p) of the disk. | |
541 | */ | |
2f9f6221 | 542 | static int blk_partition_remap(struct bio *bio) |
74d46992 | 543 | { |
309dca30 | 544 | struct block_device *p = bio->bi_bdev; |
74d46992 | 545 | |
52c5e62d | 546 | if (unlikely(should_fail_request(p, bio->bi_iter.bi_size))) |
2f9f6221 | 547 | return -EIO; |
5eac3eb3 | 548 | if (bio_sectors(bio)) { |
8446fe92 | 549 | bio->bi_iter.bi_sector += p->bd_start_sect; |
1c02fca6 | 550 | trace_block_bio_remap(bio, p->bd_dev, |
29ff57c6 | 551 | bio->bi_iter.bi_sector - |
8446fe92 | 552 | p->bd_start_sect); |
52c5e62d | 553 | } |
30c5d345 | 554 | bio_set_flag(bio, BIO_REMAPPED); |
2f9f6221 | 555 | return 0; |
74d46992 CH |
556 | } |
557 | ||
0512a75b KB |
558 | /* |
559 | * Check write append to a zoned block device. | |
560 | */ | |
561 | static inline blk_status_t blk_check_zone_append(struct request_queue *q, | |
562 | struct bio *bio) | |
563 | { | |
0512a75b KB |
564 | int nr_sectors = bio_sectors(bio); |
565 | ||
566 | /* Only applicable to zoned block devices */ | |
edd1dbc8 | 567 | if (!bdev_is_zoned(bio->bi_bdev)) |
0512a75b KB |
568 | return BLK_STS_NOTSUPP; |
569 | ||
570 | /* The bio sector must point to the start of a sequential zone */ | |
e29b2100 | 571 | if (!bdev_is_zone_start(bio->bi_bdev, bio->bi_iter.bi_sector) || |
052e545c | 572 | !bio_zone_is_seq(bio)) |
0512a75b KB |
573 | return BLK_STS_IOERR; |
574 | ||
575 | /* | |
576 | * Not allowed to cross zone boundaries. Otherwise, the BIO will be | |
577 | * split and could result in non-contiguous sectors being written in | |
578 | * different zones. | |
579 | */ | |
580 | if (nr_sectors > q->limits.chunk_sectors) | |
581 | return BLK_STS_IOERR; | |
582 | ||
583 | /* Make sure the BIO is small enough and will not get split */ | |
584 | if (nr_sectors > q->limits.max_zone_append_sectors) | |
585 | return BLK_STS_IOERR; | |
586 | ||
587 | bio->bi_opf |= REQ_NOMERGE; | |
588 | ||
589 | return BLK_STS_OK; | |
590 | } | |
591 | ||
900e0807 JA |
592 | static void __submit_bio(struct bio *bio) |
593 | { | |
7f36b7d0 ML |
594 | if (unlikely(!blk_crypto_bio_prep(&bio))) |
595 | return; | |
596 | ||
9f4107b0 | 597 | if (!bio->bi_bdev->bd_has_submit_bio) { |
3e08773c | 598 | blk_mq_submit_bio(bio); |
7f36b7d0 | 599 | } else if (likely(bio_queue_enter(bio) == 0)) { |
9f4107b0 JA |
600 | struct gendisk *disk = bio->bi_bdev->bd_disk; |
601 | ||
7f36b7d0 ML |
602 | disk->fops->submit_bio(bio); |
603 | blk_queue_exit(disk->queue); | |
604 | } | |
ac7c5675 CH |
605 | } |
606 | ||
566acf2d CH |
607 | /* |
608 | * The loop in this function may be a bit non-obvious, and so deserves some | |
609 | * explanation: | |
610 | * | |
611 | * - Before entering the loop, bio->bi_next is NULL (as all callers ensure | |
612 | * that), so we have a list with a single bio. | |
613 | * - We pretend that we have just taken it off a longer list, so we assign | |
614 | * bio_list to a pointer to the bio_list_on_stack, thus initialising the | |
615 | * bio_list of new bios to be added. ->submit_bio() may indeed add some more | |
616 | * bios through a recursive call to submit_bio_noacct. If it did, we find a | |
617 | * non-NULL value in bio_list and re-enter the loop from the top. | |
618 | * - In this case we really did just take the bio of the top of the list (no | |
619 | * pretending) and so remove it from bio_list, and call into ->submit_bio() | |
620 | * again. | |
621 | * | |
622 | * bio_list_on_stack[0] contains bios submitted by the current ->submit_bio. | |
623 | * bio_list_on_stack[1] contains bios that were submitted before the current | |
69fe0f29 | 624 | * ->submit_bio, but that haven't been processed yet. |
566acf2d | 625 | */ |
3e08773c | 626 | static void __submit_bio_noacct(struct bio *bio) |
566acf2d CH |
627 | { |
628 | struct bio_list bio_list_on_stack[2]; | |
566acf2d CH |
629 | |
630 | BUG_ON(bio->bi_next); | |
631 | ||
632 | bio_list_init(&bio_list_on_stack[0]); | |
633 | current->bio_list = bio_list_on_stack; | |
634 | ||
635 | do { | |
eab4e027 | 636 | struct request_queue *q = bdev_get_queue(bio->bi_bdev); |
566acf2d CH |
637 | struct bio_list lower, same; |
638 | ||
566acf2d CH |
639 | /* |
640 | * Create a fresh bio_list for all subordinate requests. | |
641 | */ | |
642 | bio_list_on_stack[1] = bio_list_on_stack[0]; | |
643 | bio_list_init(&bio_list_on_stack[0]); | |
644 | ||
3e08773c | 645 | __submit_bio(bio); |
566acf2d CH |
646 | |
647 | /* | |
648 | * Sort new bios into those for a lower level and those for the | |
649 | * same level. | |
650 | */ | |
651 | bio_list_init(&lower); | |
652 | bio_list_init(&same); | |
653 | while ((bio = bio_list_pop(&bio_list_on_stack[0])) != NULL) | |
eab4e027 | 654 | if (q == bdev_get_queue(bio->bi_bdev)) |
566acf2d CH |
655 | bio_list_add(&same, bio); |
656 | else | |
657 | bio_list_add(&lower, bio); | |
658 | ||
659 | /* | |
660 | * Now assemble so we handle the lowest level first. | |
661 | */ | |
662 | bio_list_merge(&bio_list_on_stack[0], &lower); | |
663 | bio_list_merge(&bio_list_on_stack[0], &same); | |
664 | bio_list_merge(&bio_list_on_stack[0], &bio_list_on_stack[1]); | |
665 | } while ((bio = bio_list_pop(&bio_list_on_stack[0]))); | |
666 | ||
667 | current->bio_list = NULL; | |
566acf2d CH |
668 | } |
669 | ||
3e08773c | 670 | static void __submit_bio_noacct_mq(struct bio *bio) |
ff93ea0c | 671 | { |
7c792f33 | 672 | struct bio_list bio_list[2] = { }; |
ff93ea0c | 673 | |
7c792f33 | 674 | current->bio_list = bio_list; |
ff93ea0c CH |
675 | |
676 | do { | |
3e08773c | 677 | __submit_bio(bio); |
7c792f33 | 678 | } while ((bio = bio_list_pop(&bio_list[0]))); |
ff93ea0c CH |
679 | |
680 | current->bio_list = NULL; | |
ff93ea0c CH |
681 | } |
682 | ||
3f98c753 | 683 | void submit_bio_noacct_nocheck(struct bio *bio) |
d89d8796 | 684 | { |
0f7c8f0f JH |
685 | blk_cgroup_bio_start(bio); |
686 | blkcg_bio_issue_init(bio); | |
687 | ||
688 | if (!bio_flagged(bio, BIO_TRACE_COMPLETION)) { | |
689 | trace_block_bio_queue(bio); | |
690 | /* | |
691 | * Now that enqueuing has been traced, we need to trace | |
692 | * completion as well. | |
693 | */ | |
694 | bio_set_flag(bio, BIO_TRACE_COMPLETION); | |
695 | } | |
696 | ||
27a84d54 | 697 | /* |
566acf2d CH |
698 | * We only want one ->submit_bio to be active at a time, else stack |
699 | * usage with stacked devices could be a problem. Use current->bio_list | |
700 | * to collect a list of requests submited by a ->submit_bio method while | |
701 | * it is active, and then process them after it returned. | |
27a84d54 | 702 | */ |
3e08773c | 703 | if (current->bio_list) |
f5fe1b51 | 704 | bio_list_add(¤t->bio_list[0], bio); |
9f4107b0 | 705 | else if (!bio->bi_bdev->bd_has_submit_bio) |
3e08773c CH |
706 | __submit_bio_noacct_mq(bio); |
707 | else | |
708 | __submit_bio_noacct(bio); | |
d89d8796 | 709 | } |
3f98c753 ML |
710 | |
711 | /** | |
712 | * submit_bio_noacct - re-submit a bio to the block device layer for I/O | |
713 | * @bio: The bio describing the location in memory and on the device. | |
714 | * | |
715 | * This is a version of submit_bio() that shall only be used for I/O that is | |
716 | * resubmitted to lower level drivers by stacking block drivers. All file | |
717 | * systems and other upper level users of the block layer should use | |
718 | * submit_bio() instead. | |
719 | */ | |
720 | void submit_bio_noacct(struct bio *bio) | |
1da177e4 | 721 | { |
309dca30 | 722 | struct block_device *bdev = bio->bi_bdev; |
eab4e027 | 723 | struct request_queue *q = bdev_get_queue(bdev); |
4e4cbee9 | 724 | blk_status_t status = BLK_STS_IOERR; |
1da177e4 LT |
725 | |
726 | might_sleep(); | |
1da177e4 | 727 | |
03a07c92 | 728 | /* |
b0beb280 | 729 | * For a REQ_NOWAIT based request, return -EOPNOTSUPP |
021a2446 | 730 | * if queue does not support NOWAIT. |
03a07c92 | 731 | */ |
568ec936 | 732 | if ((bio->bi_opf & REQ_NOWAIT) && !bdev_nowait(bdev)) |
b0beb280 | 733 | goto not_supported; |
03a07c92 | 734 | |
30abb3a6 | 735 | if (should_fail_bio(bio)) |
5a7bbad2 | 736 | goto end_io; |
bdb7d420 | 737 | bio_check_ro(bio); |
3a905c37 CH |
738 | if (!bio_flagged(bio, BIO_REMAPPED)) { |
739 | if (unlikely(bio_check_eod(bio))) | |
740 | goto end_io; | |
741 | if (bdev->bd_partno && unlikely(blk_partition_remap(bio))) | |
742 | goto end_io; | |
743 | } | |
2056a782 | 744 | |
5a7bbad2 | 745 | /* |
ed00aabd CH |
746 | * Filter flush bio's early so that bio based drivers without flush |
747 | * support don't have to worry about them. | |
5a7bbad2 | 748 | */ |
b4a6bb3a CH |
749 | if (op_is_flush(bio->bi_opf)) { |
750 | if (WARN_ON_ONCE(bio_op(bio) != REQ_OP_WRITE && | |
751 | bio_op(bio) != REQ_OP_ZONE_APPEND)) | |
51fd77bd | 752 | goto end_io; |
b4a6bb3a CH |
753 | if (!test_bit(QUEUE_FLAG_WC, &q->queue_flags)) { |
754 | bio->bi_opf &= ~(REQ_PREFLUSH | REQ_FUA); | |
755 | if (!bio_sectors(bio)) { | |
756 | status = BLK_STS_OK; | |
757 | goto end_io; | |
758 | } | |
51fd77bd | 759 | } |
5a7bbad2 | 760 | } |
5ddfe969 | 761 | |
d04c406f | 762 | if (!test_bit(QUEUE_FLAG_POLL, &q->queue_flags)) |
6ce913fe | 763 | bio_clear_polled(bio); |
d04c406f | 764 | |
288dab8a CH |
765 | switch (bio_op(bio)) { |
766 | case REQ_OP_DISCARD: | |
70200574 | 767 | if (!bdev_max_discard_sectors(bdev)) |
288dab8a CH |
768 | goto not_supported; |
769 | break; | |
770 | case REQ_OP_SECURE_ERASE: | |
44abff2c | 771 | if (!bdev_max_secure_erase_sectors(bdev)) |
288dab8a CH |
772 | goto not_supported; |
773 | break; | |
0512a75b KB |
774 | case REQ_OP_ZONE_APPEND: |
775 | status = blk_check_zone_append(q, bio); | |
776 | if (status != BLK_STS_OK) | |
777 | goto end_io; | |
778 | break; | |
2d253440 | 779 | case REQ_OP_ZONE_RESET: |
6c1b1da5 AJ |
780 | case REQ_OP_ZONE_OPEN: |
781 | case REQ_OP_ZONE_CLOSE: | |
782 | case REQ_OP_ZONE_FINISH: | |
edd1dbc8 | 783 | if (!bdev_is_zoned(bio->bi_bdev)) |
2d253440 | 784 | goto not_supported; |
288dab8a | 785 | break; |
6e33dbf2 | 786 | case REQ_OP_ZONE_RESET_ALL: |
edd1dbc8 | 787 | if (!bdev_is_zoned(bio->bi_bdev) || !blk_queue_zone_resetall(q)) |
6e33dbf2 CK |
788 | goto not_supported; |
789 | break; | |
a6f0788e | 790 | case REQ_OP_WRITE_ZEROES: |
74d46992 | 791 | if (!q->limits.max_write_zeroes_sectors) |
a6f0788e CK |
792 | goto not_supported; |
793 | break; | |
288dab8a CH |
794 | default: |
795 | break; | |
5a7bbad2 | 796 | } |
01edede4 | 797 | |
b781d8db | 798 | if (blk_throtl_bio(bio)) |
3f98c753 ML |
799 | return; |
800 | submit_bio_noacct_nocheck(bio); | |
d24c670e | 801 | return; |
a7384677 | 802 | |
288dab8a | 803 | not_supported: |
4e4cbee9 | 804 | status = BLK_STS_NOTSUPP; |
a7384677 | 805 | end_io: |
4e4cbee9 | 806 | bio->bi_status = status; |
4246a0b6 | 807 | bio_endio(bio); |
d89d8796 | 808 | } |
ed00aabd | 809 | EXPORT_SYMBOL(submit_bio_noacct); |
1da177e4 LT |
810 | |
811 | /** | |
710027a4 | 812 | * submit_bio - submit a bio to the block device layer for I/O |
1da177e4 LT |
813 | * @bio: The &struct bio which describes the I/O |
814 | * | |
3fdd4086 CH |
815 | * submit_bio() is used to submit I/O requests to block devices. It is passed a |
816 | * fully set up &struct bio that describes the I/O that needs to be done. The | |
309dca30 | 817 | * bio will be send to the device described by the bi_bdev field. |
1da177e4 | 818 | * |
3fdd4086 CH |
819 | * The success/failure status of the request, along with notification of |
820 | * completion, is delivered asynchronously through the ->bi_end_io() callback | |
e8848087 | 821 | * in @bio. The bio must NOT be touched by the caller until ->bi_end_io() has |
3fdd4086 | 822 | * been called. |
1da177e4 | 823 | */ |
3e08773c | 824 | void submit_bio(struct bio *bio) |
1da177e4 | 825 | { |
a3e7689b CH |
826 | if (bio_op(bio) == REQ_OP_READ) { |
827 | task_io_account_read(bio->bi_iter.bi_size); | |
828 | count_vm_events(PGPGIN, bio_sectors(bio)); | |
829 | } else if (bio_op(bio) == REQ_OP_WRITE) { | |
830 | count_vm_events(PGPGOUT, bio_sectors(bio)); | |
1da177e4 LT |
831 | } |
832 | ||
3e08773c | 833 | submit_bio_noacct(bio); |
1da177e4 | 834 | } |
1da177e4 LT |
835 | EXPORT_SYMBOL(submit_bio); |
836 | ||
3e08773c CH |
837 | /** |
838 | * bio_poll - poll for BIO completions | |
839 | * @bio: bio to poll for | |
e30028ac | 840 | * @iob: batches of IO |
3e08773c CH |
841 | * @flags: BLK_POLL_* flags that control the behavior |
842 | * | |
843 | * Poll for completions on queue associated with the bio. Returns number of | |
844 | * completed entries found. | |
845 | * | |
846 | * Note: the caller must either be the context that submitted @bio, or | |
847 | * be in a RCU critical section to prevent freeing of @bio. | |
848 | */ | |
5a72e899 | 849 | int bio_poll(struct bio *bio, struct io_comp_batch *iob, unsigned int flags) |
3e08773c | 850 | { |
3e08773c | 851 | blk_qc_t cookie = READ_ONCE(bio->bi_cookie); |
310726c3 JA |
852 | struct block_device *bdev; |
853 | struct request_queue *q; | |
69fe0f29 | 854 | int ret = 0; |
3e08773c | 855 | |
310726c3 JA |
856 | bdev = READ_ONCE(bio->bi_bdev); |
857 | if (!bdev) | |
858 | return 0; | |
859 | ||
860 | q = bdev_get_queue(bdev); | |
3e08773c CH |
861 | if (cookie == BLK_QC_T_NONE || |
862 | !test_bit(QUEUE_FLAG_POLL, &q->queue_flags)) | |
863 | return 0; | |
864 | ||
110fdb44 PR |
865 | /* |
866 | * As the requests that require a zone lock are not plugged in the | |
867 | * first place, directly accessing the plug instead of using | |
868 | * blk_mq_plug() should not have any consequences during flushing for | |
869 | * zoned devices. | |
870 | */ | |
aa8dccca | 871 | blk_flush_plug(current->plug, false); |
3e08773c | 872 | |
33391eec JA |
873 | /* |
874 | * We need to be able to enter a frozen queue, similar to how | |
875 | * timeouts also need to do that. If that is blocked, then we can | |
876 | * have pending IO when a queue freeze is started, and then the | |
877 | * wait for the freeze to finish will wait for polled requests to | |
878 | * timeout as the poller is preventer from entering the queue and | |
879 | * completing them. As long as we prevent new IO from being queued, | |
880 | * that should be all that matters. | |
881 | */ | |
882 | if (!percpu_ref_tryget(&q->q_usage_counter)) | |
3e08773c | 883 | return 0; |
69fe0f29 | 884 | if (queue_is_mq(q)) { |
5a72e899 | 885 | ret = blk_mq_poll(q, cookie, iob, flags); |
69fe0f29 ML |
886 | } else { |
887 | struct gendisk *disk = q->disk; | |
888 | ||
889 | if (disk && disk->fops->poll_bio) | |
890 | ret = disk->fops->poll_bio(bio, iob, flags); | |
891 | } | |
3e08773c CH |
892 | blk_queue_exit(q); |
893 | return ret; | |
894 | } | |
895 | EXPORT_SYMBOL_GPL(bio_poll); | |
896 | ||
897 | /* | |
898 | * Helper to implement file_operations.iopoll. Requires the bio to be stored | |
899 | * in iocb->private, and cleared before freeing the bio. | |
900 | */ | |
5a72e899 JA |
901 | int iocb_bio_iopoll(struct kiocb *kiocb, struct io_comp_batch *iob, |
902 | unsigned int flags) | |
3e08773c CH |
903 | { |
904 | struct bio *bio; | |
905 | int ret = 0; | |
906 | ||
907 | /* | |
908 | * Note: the bio cache only uses SLAB_TYPESAFE_BY_RCU, so bio can | |
909 | * point to a freshly allocated bio at this point. If that happens | |
910 | * we have a few cases to consider: | |
911 | * | |
912 | * 1) the bio is beeing initialized and bi_bdev is NULL. We can just | |
913 | * simply nothing in this case | |
914 | * 2) the bio points to a not poll enabled device. bio_poll will catch | |
915 | * this and return 0 | |
916 | * 3) the bio points to a poll capable device, including but not | |
917 | * limited to the one that the original bio pointed to. In this | |
918 | * case we will call into the actual poll method and poll for I/O, | |
919 | * even if we don't need to, but it won't cause harm either. | |
920 | * | |
921 | * For cases 2) and 3) above the RCU grace period ensures that bi_bdev | |
922 | * is still allocated. Because partitions hold a reference to the whole | |
923 | * device bdev and thus disk, the disk is also still valid. Grabbing | |
924 | * a reference to the queue in bio_poll() ensures the hctxs and requests | |
925 | * are still valid as well. | |
926 | */ | |
927 | rcu_read_lock(); | |
928 | bio = READ_ONCE(kiocb->private); | |
310726c3 | 929 | if (bio) |
5a72e899 | 930 | ret = bio_poll(bio, iob, flags); |
3e08773c CH |
931 | rcu_read_unlock(); |
932 | ||
933 | return ret; | |
934 | } | |
935 | EXPORT_SYMBOL_GPL(iocb_bio_iopoll); | |
936 | ||
450b7879 | 937 | void update_io_ticks(struct block_device *part, unsigned long now, bool end) |
9123bf6f CH |
938 | { |
939 | unsigned long stamp; | |
940 | again: | |
8446fe92 | 941 | stamp = READ_ONCE(part->bd_stamp); |
d80c228d | 942 | if (unlikely(time_after(now, stamp))) { |
939f9dd0 | 943 | if (likely(try_cmpxchg(&part->bd_stamp, &stamp, now))) |
9123bf6f CH |
944 | __part_stat_add(part, io_ticks, end ? now - stamp : 1); |
945 | } | |
8446fe92 CH |
946 | if (part->bd_partno) { |
947 | part = bdev_whole(part); | |
9123bf6f CH |
948 | goto again; |
949 | } | |
950 | } | |
951 | ||
5f275713 | 952 | unsigned long bdev_start_io_acct(struct block_device *bdev, enum req_op op, |
5f0614a5 | 953 | unsigned long start_time) |
956d510e | 954 | { |
956d510e | 955 | part_stat_lock(); |
5f0614a5 | 956 | update_io_ticks(bdev, start_time, false); |
5f0614a5 | 957 | part_stat_local_inc(bdev, in_flight[op_is_write(op)]); |
956d510e | 958 | part_stat_unlock(); |
320ae51f | 959 | |
e45c47d1 MS |
960 | return start_time; |
961 | } | |
5f0614a5 | 962 | EXPORT_SYMBOL(bdev_start_io_acct); |
e45c47d1 | 963 | |
99dfc43e CH |
964 | /** |
965 | * bio_start_io_acct - start I/O accounting for bio based drivers | |
966 | * @bio: bio to start account for | |
967 | * | |
968 | * Returns the start time that should be passed back to bio_end_io_acct(). | |
969 | */ | |
970 | unsigned long bio_start_io_acct(struct bio *bio) | |
7b26410b | 971 | { |
5f275713 | 972 | return bdev_start_io_acct(bio->bi_bdev, bio_op(bio), jiffies); |
7b26410b | 973 | } |
99dfc43e | 974 | EXPORT_SYMBOL_GPL(bio_start_io_acct); |
7b26410b | 975 | |
77e7ffd7 | 976 | void bdev_end_io_acct(struct block_device *bdev, enum req_op op, |
5f275713 | 977 | unsigned int sectors, unsigned long start_time) |
956d510e | 978 | { |
956d510e CH |
979 | const int sgrp = op_stat_group(op); |
980 | unsigned long now = READ_ONCE(jiffies); | |
981 | unsigned long duration = now - start_time; | |
5b18b5a7 | 982 | |
956d510e | 983 | part_stat_lock(); |
5f0614a5 | 984 | update_io_ticks(bdev, now, true); |
5f275713 YK |
985 | part_stat_inc(bdev, ios[sgrp]); |
986 | part_stat_add(bdev, sectors[sgrp], sectors); | |
5f0614a5 ML |
987 | part_stat_add(bdev, nsecs[sgrp], jiffies_to_nsecs(duration)); |
988 | part_stat_local_dec(bdev, in_flight[op_is_write(op)]); | |
320ae51f JA |
989 | part_stat_unlock(); |
990 | } | |
5f0614a5 | 991 | EXPORT_SYMBOL(bdev_end_io_acct); |
7b26410b | 992 | |
99dfc43e | 993 | void bio_end_io_acct_remapped(struct bio *bio, unsigned long start_time, |
5f0614a5 | 994 | struct block_device *orig_bdev) |
7b26410b | 995 | { |
5f275713 | 996 | bdev_end_io_acct(orig_bdev, bio_op(bio), bio_sectors(bio), start_time); |
7b26410b | 997 | } |
99dfc43e | 998 | EXPORT_SYMBOL_GPL(bio_end_io_acct_remapped); |
7b26410b | 999 | |
ef9e3fac KU |
1000 | /** |
1001 | * blk_lld_busy - Check if underlying low-level drivers of a device are busy | |
1002 | * @q : the queue of the device being checked | |
1003 | * | |
1004 | * Description: | |
1005 | * Check if underlying low-level drivers of a device are busy. | |
1006 | * If the drivers want to export their busy state, they must set own | |
1007 | * exporting function using blk_queue_lld_busy() first. | |
1008 | * | |
1009 | * Basically, this function is used only by request stacking drivers | |
1010 | * to stop dispatching requests to underlying devices when underlying | |
1011 | * devices are busy. This behavior helps more I/O merging on the queue | |
1012 | * of the request stacking driver and prevents I/O throughput regression | |
1013 | * on burst I/O load. | |
1014 | * | |
1015 | * Return: | |
1016 | * 0 - Not busy (The request stacking driver should dispatch request) | |
1017 | * 1 - Busy (The request stacking driver should stop dispatching request) | |
1018 | */ | |
1019 | int blk_lld_busy(struct request_queue *q) | |
1020 | { | |
344e9ffc | 1021 | if (queue_is_mq(q) && q->mq_ops->busy) |
9ba20527 | 1022 | return q->mq_ops->busy(q); |
ef9e3fac KU |
1023 | |
1024 | return 0; | |
1025 | } | |
1026 | EXPORT_SYMBOL_GPL(blk_lld_busy); | |
1027 | ||
59c3d45e | 1028 | int kblockd_schedule_work(struct work_struct *work) |
1da177e4 LT |
1029 | { |
1030 | return queue_work(kblockd_workqueue, work); | |
1031 | } | |
1da177e4 LT |
1032 | EXPORT_SYMBOL(kblockd_schedule_work); |
1033 | ||
818cd1cb JA |
1034 | int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, |
1035 | unsigned long delay) | |
1036 | { | |
1037 | return mod_delayed_work_on(cpu, kblockd_workqueue, dwork, delay); | |
1038 | } | |
1039 | EXPORT_SYMBOL(kblockd_mod_delayed_work_on); | |
1040 | ||
47c122e3 JA |
1041 | void blk_start_plug_nr_ios(struct blk_plug *plug, unsigned short nr_ios) |
1042 | { | |
1043 | struct task_struct *tsk = current; | |
1044 | ||
1045 | /* | |
1046 | * If this is a nested plug, don't actually assign it. | |
1047 | */ | |
1048 | if (tsk->plug) | |
1049 | return; | |
1050 | ||
bc490f81 | 1051 | plug->mq_list = NULL; |
47c122e3 JA |
1052 | plug->cached_rq = NULL; |
1053 | plug->nr_ios = min_t(unsigned short, nr_ios, BLK_MAX_REQUEST_COUNT); | |
1054 | plug->rq_count = 0; | |
1055 | plug->multiple_queues = false; | |
dc5fc361 | 1056 | plug->has_elevator = false; |
47c122e3 JA |
1057 | INIT_LIST_HEAD(&plug->cb_list); |
1058 | ||
1059 | /* | |
1060 | * Store ordering should not be needed here, since a potential | |
1061 | * preempt will imply a full memory barrier | |
1062 | */ | |
1063 | tsk->plug = plug; | |
1064 | } | |
1065 | ||
75df7136 SJ |
1066 | /** |
1067 | * blk_start_plug - initialize blk_plug and track it inside the task_struct | |
1068 | * @plug: The &struct blk_plug that needs to be initialized | |
1069 | * | |
1070 | * Description: | |
40405851 JM |
1071 | * blk_start_plug() indicates to the block layer an intent by the caller |
1072 | * to submit multiple I/O requests in a batch. The block layer may use | |
1073 | * this hint to defer submitting I/Os from the caller until blk_finish_plug() | |
1074 | * is called. However, the block layer may choose to submit requests | |
1075 | * before a call to blk_finish_plug() if the number of queued I/Os | |
1076 | * exceeds %BLK_MAX_REQUEST_COUNT, or if the size of the I/O is larger than | |
1077 | * %BLK_PLUG_FLUSH_SIZE. The queued I/Os may also be submitted early if | |
1078 | * the task schedules (see below). | |
1079 | * | |
75df7136 SJ |
1080 | * Tracking blk_plug inside the task_struct will help with auto-flushing the |
1081 | * pending I/O should the task end up blocking between blk_start_plug() and | |
1082 | * blk_finish_plug(). This is important from a performance perspective, but | |
1083 | * also ensures that we don't deadlock. For instance, if the task is blocking | |
1084 | * for a memory allocation, memory reclaim could end up wanting to free a | |
1085 | * page belonging to that request that is currently residing in our private | |
1086 | * plug. By flushing the pending I/O when the process goes to sleep, we avoid | |
1087 | * this kind of deadlock. | |
1088 | */ | |
73c10101 JA |
1089 | void blk_start_plug(struct blk_plug *plug) |
1090 | { | |
47c122e3 | 1091 | blk_start_plug_nr_ios(plug, 1); |
73c10101 JA |
1092 | } |
1093 | EXPORT_SYMBOL(blk_start_plug); | |
1094 | ||
74018dc3 | 1095 | static void flush_plug_callbacks(struct blk_plug *plug, bool from_schedule) |
048c9374 N |
1096 | { |
1097 | LIST_HEAD(callbacks); | |
1098 | ||
2a7d5559 SL |
1099 | while (!list_empty(&plug->cb_list)) { |
1100 | list_splice_init(&plug->cb_list, &callbacks); | |
048c9374 | 1101 | |
2a7d5559 SL |
1102 | while (!list_empty(&callbacks)) { |
1103 | struct blk_plug_cb *cb = list_first_entry(&callbacks, | |
048c9374 N |
1104 | struct blk_plug_cb, |
1105 | list); | |
2a7d5559 | 1106 | list_del(&cb->list); |
74018dc3 | 1107 | cb->callback(cb, from_schedule); |
2a7d5559 | 1108 | } |
048c9374 N |
1109 | } |
1110 | } | |
1111 | ||
9cbb1750 N |
1112 | struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug, void *data, |
1113 | int size) | |
1114 | { | |
1115 | struct blk_plug *plug = current->plug; | |
1116 | struct blk_plug_cb *cb; | |
1117 | ||
1118 | if (!plug) | |
1119 | return NULL; | |
1120 | ||
1121 | list_for_each_entry(cb, &plug->cb_list, list) | |
1122 | if (cb->callback == unplug && cb->data == data) | |
1123 | return cb; | |
1124 | ||
1125 | /* Not currently on the callback list */ | |
1126 | BUG_ON(size < sizeof(*cb)); | |
1127 | cb = kzalloc(size, GFP_ATOMIC); | |
1128 | if (cb) { | |
1129 | cb->data = data; | |
1130 | cb->callback = unplug; | |
1131 | list_add(&cb->list, &plug->cb_list); | |
1132 | } | |
1133 | return cb; | |
1134 | } | |
1135 | EXPORT_SYMBOL(blk_check_plugged); | |
1136 | ||
aa8dccca | 1137 | void __blk_flush_plug(struct blk_plug *plug, bool from_schedule) |
73c10101 | 1138 | { |
b600455d PB |
1139 | if (!list_empty(&plug->cb_list)) |
1140 | flush_plug_callbacks(plug, from_schedule); | |
70904263 | 1141 | blk_mq_flush_plug_list(plug, from_schedule); |
c5fc7b93 JA |
1142 | /* |
1143 | * Unconditionally flush out cached requests, even if the unplug | |
1144 | * event came from schedule. Since we know hold references to the | |
1145 | * queue for cached requests, we don't want a blocked task holding | |
1146 | * up a queue freeze/quiesce event. | |
1147 | */ | |
1148 | if (unlikely(!rq_list_empty(plug->cached_rq))) | |
47c122e3 | 1149 | blk_mq_free_plug_rqs(plug); |
73c10101 | 1150 | } |
73c10101 | 1151 | |
40405851 JM |
1152 | /** |
1153 | * blk_finish_plug - mark the end of a batch of submitted I/O | |
1154 | * @plug: The &struct blk_plug passed to blk_start_plug() | |
1155 | * | |
1156 | * Description: | |
1157 | * Indicate that a batch of I/O submissions is complete. This function | |
1158 | * must be paired with an initial call to blk_start_plug(). The intent | |
1159 | * is to allow the block layer to optimize I/O submission. See the | |
1160 | * documentation for blk_start_plug() for more information. | |
1161 | */ | |
73c10101 JA |
1162 | void blk_finish_plug(struct blk_plug *plug) |
1163 | { | |
008f75a2 | 1164 | if (plug == current->plug) { |
aa8dccca | 1165 | __blk_flush_plug(plug, false); |
008f75a2 CH |
1166 | current->plug = NULL; |
1167 | } | |
73c10101 | 1168 | } |
88b996cd | 1169 | EXPORT_SYMBOL(blk_finish_plug); |
73c10101 | 1170 | |
71ac860a ML |
1171 | void blk_io_schedule(void) |
1172 | { | |
1173 | /* Prevent hang_check timer from firing at us during very long I/O */ | |
1174 | unsigned long timeout = sysctl_hung_task_timeout_secs * HZ / 2; | |
1175 | ||
1176 | if (timeout) | |
1177 | io_schedule_timeout(timeout); | |
1178 | else | |
1179 | io_schedule(); | |
1180 | } | |
1181 | EXPORT_SYMBOL_GPL(blk_io_schedule); | |
1182 | ||
1da177e4 LT |
1183 | int __init blk_dev_init(void) |
1184 | { | |
16458cf3 | 1185 | BUILD_BUG_ON((__force u32)REQ_OP_LAST >= (1 << REQ_OP_BITS)); |
ef295ecf | 1186 | BUILD_BUG_ON(REQ_OP_BITS + REQ_FLAG_BITS > 8 * |
c593642c | 1187 | sizeof_field(struct request, cmd_flags)); |
ef295ecf | 1188 | BUILD_BUG_ON(REQ_OP_BITS + REQ_FLAG_BITS > 8 * |
c593642c | 1189 | sizeof_field(struct bio, bi_opf)); |
9eb55b03 | 1190 | |
89b90be2 TH |
1191 | /* used for unplugging and affects IO latency/throughput - HIGHPRI */ |
1192 | kblockd_workqueue = alloc_workqueue("kblockd", | |
28747fcd | 1193 | WQ_MEM_RECLAIM | WQ_HIGHPRI, 0); |
1da177e4 LT |
1194 | if (!kblockd_workqueue) |
1195 | panic("Failed to create kblockd\n"); | |
1196 | ||
c2789bd4 | 1197 | blk_requestq_cachep = kmem_cache_create("request_queue", |
165125e1 | 1198 | sizeof(struct request_queue), 0, SLAB_PANIC, NULL); |
1da177e4 | 1199 | |
18fbda91 | 1200 | blk_debugfs_root = debugfs_create_dir("block", NULL); |
18fbda91 | 1201 | |
d38ecf93 | 1202 | return 0; |
1da177e4 | 1203 | } |