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> | |
320ae51f | 19 | #include <linux/blk-mq.h> |
52abca64 | 20 | #include <linux/blk-pm.h> |
fe45e630 | 21 | #include <linux/blk-integrity.h> |
1da177e4 LT |
22 | #include <linux/highmem.h> |
23 | #include <linux/mm.h> | |
cee9a0c4 | 24 | #include <linux/pagemap.h> |
1da177e4 LT |
25 | #include <linux/kernel_stat.h> |
26 | #include <linux/string.h> | |
27 | #include <linux/init.h> | |
1da177e4 LT |
28 | #include <linux/completion.h> |
29 | #include <linux/slab.h> | |
30 | #include <linux/swap.h> | |
31 | #include <linux/writeback.h> | |
faccbd4b | 32 | #include <linux/task_io_accounting_ops.h> |
c17bb495 | 33 | #include <linux/fault-inject.h> |
73c10101 | 34 | #include <linux/list_sort.h> |
e3c78ca5 | 35 | #include <linux/delay.h> |
aaf7c680 | 36 | #include <linux/ratelimit.h> |
6c954667 | 37 | #include <linux/pm_runtime.h> |
eea8f41c | 38 | #include <linux/blk-cgroup.h> |
54d4e6ab | 39 | #include <linux/t10-pi.h> |
18fbda91 | 40 | #include <linux/debugfs.h> |
30abb3a6 | 41 | #include <linux/bpf.h> |
b8e24a93 | 42 | #include <linux/psi.h> |
71ac860a | 43 | #include <linux/sched/sysctl.h> |
a892c8d5 | 44 | #include <linux/blk-crypto.h> |
55782138 LZ |
45 | |
46 | #define CREATE_TRACE_POINTS | |
47 | #include <trace/events/block.h> | |
1da177e4 | 48 | |
8324aa91 | 49 | #include "blk.h" |
43a5e4e2 | 50 | #include "blk-mq.h" |
bd166ef1 | 51 | #include "blk-mq-sched.h" |
bca6b067 | 52 | #include "blk-pm.h" |
a7b36ee6 | 53 | #include "blk-throttle.h" |
8324aa91 | 54 | |
18fbda91 | 55 | struct dentry *blk_debugfs_root; |
18fbda91 | 56 | |
d07335e5 | 57 | EXPORT_TRACEPOINT_SYMBOL_GPL(block_bio_remap); |
b0da3f0d | 58 | EXPORT_TRACEPOINT_SYMBOL_GPL(block_rq_remap); |
0a82a8d1 | 59 | EXPORT_TRACEPOINT_SYMBOL_GPL(block_bio_complete); |
3291fa57 | 60 | EXPORT_TRACEPOINT_SYMBOL_GPL(block_split); |
cbae8d45 | 61 | EXPORT_TRACEPOINT_SYMBOL_GPL(block_unplug); |
b357e4a6 | 62 | EXPORT_TRACEPOINT_SYMBOL_GPL(block_rq_insert); |
0bfc2455 | 63 | |
a73f730d TH |
64 | DEFINE_IDA(blk_queue_ida); |
65 | ||
1da177e4 LT |
66 | /* |
67 | * For queue allocation | |
68 | */ | |
6728cb0e | 69 | struct kmem_cache *blk_requestq_cachep; |
1da177e4 | 70 | |
1da177e4 LT |
71 | /* |
72 | * Controlling structure to kblockd | |
73 | */ | |
ff856bad | 74 | static struct workqueue_struct *kblockd_workqueue; |
1da177e4 | 75 | |
8814ce8a BVA |
76 | /** |
77 | * blk_queue_flag_set - atomically set a queue flag | |
78 | * @flag: flag to be set | |
79 | * @q: request queue | |
80 | */ | |
81 | void blk_queue_flag_set(unsigned int flag, struct request_queue *q) | |
82 | { | |
57d74df9 | 83 | set_bit(flag, &q->queue_flags); |
8814ce8a BVA |
84 | } |
85 | EXPORT_SYMBOL(blk_queue_flag_set); | |
86 | ||
87 | /** | |
88 | * blk_queue_flag_clear - atomically clear a queue flag | |
89 | * @flag: flag to be cleared | |
90 | * @q: request queue | |
91 | */ | |
92 | void blk_queue_flag_clear(unsigned int flag, struct request_queue *q) | |
93 | { | |
57d74df9 | 94 | clear_bit(flag, &q->queue_flags); |
8814ce8a BVA |
95 | } |
96 | EXPORT_SYMBOL(blk_queue_flag_clear); | |
97 | ||
98 | /** | |
99 | * blk_queue_flag_test_and_set - atomically test and set a queue flag | |
100 | * @flag: flag to be set | |
101 | * @q: request queue | |
102 | * | |
103 | * Returns the previous value of @flag - 0 if the flag was not set and 1 if | |
104 | * the flag was already set. | |
105 | */ | |
106 | bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q) | |
107 | { | |
57d74df9 | 108 | return test_and_set_bit(flag, &q->queue_flags); |
8814ce8a BVA |
109 | } |
110 | EXPORT_SYMBOL_GPL(blk_queue_flag_test_and_set); | |
111 | ||
2a4aa30c | 112 | void blk_rq_init(struct request_queue *q, struct request *rq) |
1da177e4 | 113 | { |
1afb20f3 FT |
114 | memset(rq, 0, sizeof(*rq)); |
115 | ||
1da177e4 | 116 | INIT_LIST_HEAD(&rq->queuelist); |
63a71386 | 117 | rq->q = q; |
a2dec7b3 | 118 | rq->__sector = (sector_t) -1; |
2e662b65 JA |
119 | INIT_HLIST_NODE(&rq->hash); |
120 | RB_CLEAR_NODE(&rq->rb_node); | |
e44a6a23 XT |
121 | rq->tag = BLK_MQ_NO_TAG; |
122 | rq->internal_tag = BLK_MQ_NO_TAG; | |
522a7775 | 123 | rq->start_time_ns = ktime_get_ns(); |
09e099d4 | 124 | rq->part = NULL; |
a892c8d5 | 125 | blk_crypto_rq_set_defaults(rq); |
1da177e4 | 126 | } |
2a4aa30c | 127 | EXPORT_SYMBOL(blk_rq_init); |
1da177e4 | 128 | |
e47bc4ed CK |
129 | #define REQ_OP_NAME(name) [REQ_OP_##name] = #name |
130 | static const char *const blk_op_name[] = { | |
131 | REQ_OP_NAME(READ), | |
132 | REQ_OP_NAME(WRITE), | |
133 | REQ_OP_NAME(FLUSH), | |
134 | REQ_OP_NAME(DISCARD), | |
135 | REQ_OP_NAME(SECURE_ERASE), | |
136 | REQ_OP_NAME(ZONE_RESET), | |
6e33dbf2 | 137 | REQ_OP_NAME(ZONE_RESET_ALL), |
6c1b1da5 AJ |
138 | REQ_OP_NAME(ZONE_OPEN), |
139 | REQ_OP_NAME(ZONE_CLOSE), | |
140 | REQ_OP_NAME(ZONE_FINISH), | |
0512a75b | 141 | REQ_OP_NAME(ZONE_APPEND), |
e47bc4ed CK |
142 | REQ_OP_NAME(WRITE_SAME), |
143 | REQ_OP_NAME(WRITE_ZEROES), | |
e47bc4ed CK |
144 | REQ_OP_NAME(DRV_IN), |
145 | REQ_OP_NAME(DRV_OUT), | |
146 | }; | |
147 | #undef REQ_OP_NAME | |
148 | ||
149 | /** | |
150 | * blk_op_str - Return string XXX in the REQ_OP_XXX. | |
151 | * @op: REQ_OP_XXX. | |
152 | * | |
153 | * Description: Centralize block layer function to convert REQ_OP_XXX into | |
154 | * string format. Useful in the debugging and tracing bio or request. For | |
155 | * invalid REQ_OP_XXX it returns string "UNKNOWN". | |
156 | */ | |
157 | inline const char *blk_op_str(unsigned int op) | |
158 | { | |
159 | const char *op_str = "UNKNOWN"; | |
160 | ||
161 | if (op < ARRAY_SIZE(blk_op_name) && blk_op_name[op]) | |
162 | op_str = blk_op_name[op]; | |
163 | ||
164 | return op_str; | |
165 | } | |
166 | EXPORT_SYMBOL_GPL(blk_op_str); | |
167 | ||
2a842aca CH |
168 | static const struct { |
169 | int errno; | |
170 | const char *name; | |
171 | } blk_errors[] = { | |
172 | [BLK_STS_OK] = { 0, "" }, | |
173 | [BLK_STS_NOTSUPP] = { -EOPNOTSUPP, "operation not supported" }, | |
174 | [BLK_STS_TIMEOUT] = { -ETIMEDOUT, "timeout" }, | |
175 | [BLK_STS_NOSPC] = { -ENOSPC, "critical space allocation" }, | |
176 | [BLK_STS_TRANSPORT] = { -ENOLINK, "recoverable transport" }, | |
177 | [BLK_STS_TARGET] = { -EREMOTEIO, "critical target" }, | |
178 | [BLK_STS_NEXUS] = { -EBADE, "critical nexus" }, | |
179 | [BLK_STS_MEDIUM] = { -ENODATA, "critical medium" }, | |
180 | [BLK_STS_PROTECTION] = { -EILSEQ, "protection" }, | |
181 | [BLK_STS_RESOURCE] = { -ENOMEM, "kernel resource" }, | |
86ff7c2a | 182 | [BLK_STS_DEV_RESOURCE] = { -EBUSY, "device resource" }, |
03a07c92 | 183 | [BLK_STS_AGAIN] = { -EAGAIN, "nonblocking retry" }, |
2a842aca | 184 | |
4e4cbee9 CH |
185 | /* device mapper special case, should not leak out: */ |
186 | [BLK_STS_DM_REQUEUE] = { -EREMCHG, "dm internal retry" }, | |
187 | ||
3b481d91 KB |
188 | /* zone device specific errors */ |
189 | [BLK_STS_ZONE_OPEN_RESOURCE] = { -ETOOMANYREFS, "open zones exceeded" }, | |
190 | [BLK_STS_ZONE_ACTIVE_RESOURCE] = { -EOVERFLOW, "active zones exceeded" }, | |
191 | ||
2a842aca CH |
192 | /* everything else not covered above: */ |
193 | [BLK_STS_IOERR] = { -EIO, "I/O" }, | |
194 | }; | |
195 | ||
196 | blk_status_t errno_to_blk_status(int errno) | |
197 | { | |
198 | int i; | |
199 | ||
200 | for (i = 0; i < ARRAY_SIZE(blk_errors); i++) { | |
201 | if (blk_errors[i].errno == errno) | |
202 | return (__force blk_status_t)i; | |
203 | } | |
204 | ||
205 | return BLK_STS_IOERR; | |
206 | } | |
207 | EXPORT_SYMBOL_GPL(errno_to_blk_status); | |
208 | ||
209 | int blk_status_to_errno(blk_status_t status) | |
210 | { | |
211 | int idx = (__force int)status; | |
212 | ||
34bd9c1c | 213 | if (WARN_ON_ONCE(idx >= ARRAY_SIZE(blk_errors))) |
2a842aca CH |
214 | return -EIO; |
215 | return blk_errors[idx].errno; | |
216 | } | |
217 | EXPORT_SYMBOL_GPL(blk_status_to_errno); | |
218 | ||
9be3e06f | 219 | void blk_print_req_error(struct request *req, blk_status_t status) |
2a842aca CH |
220 | { |
221 | int idx = (__force int)status; | |
222 | ||
34bd9c1c | 223 | if (WARN_ON_ONCE(idx >= ARRAY_SIZE(blk_errors))) |
2a842aca CH |
224 | return; |
225 | ||
178cc590 | 226 | printk_ratelimited(KERN_ERR |
c477b797 | 227 | "%s error, dev %s, sector %llu op 0x%x:(%s) flags 0x%x " |
b0e5168a | 228 | "phys_seg %u prio class %u\n", |
c477b797 | 229 | blk_errors[idx].name, |
b0e5168a CK |
230 | req->rq_disk ? req->rq_disk->disk_name : "?", |
231 | blk_rq_pos(req), req_op(req), blk_op_str(req_op(req)), | |
232 | req->cmd_flags & ~REQ_OP_MASK, | |
233 | req->nr_phys_segments, | |
234 | IOPRIO_PRIO_CLASS(req->ioprio)); | |
2a842aca CH |
235 | } |
236 | ||
1da177e4 LT |
237 | void blk_dump_rq_flags(struct request *rq, char *msg) |
238 | { | |
aebf526b CH |
239 | printk(KERN_INFO "%s: dev %s: flags=%llx\n", msg, |
240 | rq->rq_disk ? rq->rq_disk->disk_name : "?", | |
5953316d | 241 | (unsigned long long) rq->cmd_flags); |
1da177e4 | 242 | |
83096ebf TH |
243 | printk(KERN_INFO " sector %llu, nr/cnr %u/%u\n", |
244 | (unsigned long long)blk_rq_pos(rq), | |
245 | blk_rq_sectors(rq), blk_rq_cur_sectors(rq)); | |
b4f42e28 JA |
246 | printk(KERN_INFO " bio %p, biotail %p, len %u\n", |
247 | rq->bio, rq->biotail, blk_rq_bytes(rq)); | |
1da177e4 | 248 | } |
1da177e4 LT |
249 | EXPORT_SYMBOL(blk_dump_rq_flags); |
250 | ||
1da177e4 LT |
251 | /** |
252 | * blk_sync_queue - cancel any pending callbacks on a queue | |
253 | * @q: the queue | |
254 | * | |
255 | * Description: | |
256 | * The block layer may perform asynchronous callback activity | |
257 | * on a queue, such as calling the unplug function after a timeout. | |
258 | * A block device may call blk_sync_queue to ensure that any | |
259 | * such activity is cancelled, thus allowing it to release resources | |
59c51591 | 260 | * that the callbacks might use. The caller must already have made sure |
c62b37d9 | 261 | * that its ->submit_bio will not re-add plugging prior to calling |
1da177e4 LT |
262 | * this function. |
263 | * | |
da527770 | 264 | * This function does not cancel any asynchronous activity arising |
da3dae54 | 265 | * out of elevator or throttling code. That would require elevator_exit() |
5efd6113 | 266 | * and blkcg_exit_queue() to be called with queue lock initialized. |
da527770 | 267 | * |
1da177e4 LT |
268 | */ |
269 | void blk_sync_queue(struct request_queue *q) | |
270 | { | |
70ed28b9 | 271 | del_timer_sync(&q->timeout); |
4e9b6f20 | 272 | cancel_work_sync(&q->timeout_work); |
1da177e4 LT |
273 | } |
274 | EXPORT_SYMBOL(blk_sync_queue); | |
275 | ||
c9254f2d | 276 | /** |
cd84a62e | 277 | * blk_set_pm_only - increment pm_only counter |
c9254f2d | 278 | * @q: request queue pointer |
c9254f2d | 279 | */ |
cd84a62e | 280 | void blk_set_pm_only(struct request_queue *q) |
c9254f2d | 281 | { |
cd84a62e | 282 | atomic_inc(&q->pm_only); |
c9254f2d | 283 | } |
cd84a62e | 284 | EXPORT_SYMBOL_GPL(blk_set_pm_only); |
c9254f2d | 285 | |
cd84a62e | 286 | void blk_clear_pm_only(struct request_queue *q) |
c9254f2d | 287 | { |
cd84a62e BVA |
288 | int pm_only; |
289 | ||
290 | pm_only = atomic_dec_return(&q->pm_only); | |
291 | WARN_ON_ONCE(pm_only < 0); | |
292 | if (pm_only == 0) | |
293 | wake_up_all(&q->mq_freeze_wq); | |
c9254f2d | 294 | } |
cd84a62e | 295 | EXPORT_SYMBOL_GPL(blk_clear_pm_only); |
c9254f2d | 296 | |
b5bd357c LC |
297 | /** |
298 | * blk_put_queue - decrement the request_queue refcount | |
299 | * @q: the request_queue structure to decrement the refcount for | |
300 | * | |
301 | * Decrements the refcount of the request_queue kobject. When this reaches 0 | |
302 | * we'll have blk_release_queue() called. | |
e8c7d14a LC |
303 | * |
304 | * Context: Any context, but the last reference must not be dropped from | |
305 | * atomic context. | |
b5bd357c | 306 | */ |
165125e1 | 307 | void blk_put_queue(struct request_queue *q) |
483f4afc AV |
308 | { |
309 | kobject_put(&q->kobj); | |
310 | } | |
d86e0e83 | 311 | EXPORT_SYMBOL(blk_put_queue); |
483f4afc | 312 | |
8e141f9e | 313 | void blk_queue_start_drain(struct request_queue *q) |
aed3ea94 | 314 | { |
d3cfb2a0 ML |
315 | /* |
316 | * When queue DYING flag is set, we need to block new req | |
317 | * entering queue, so we call blk_freeze_queue_start() to | |
318 | * prevent I/O from crossing blk_queue_enter(). | |
319 | */ | |
320 | blk_freeze_queue_start(q); | |
344e9ffc | 321 | if (queue_is_mq(q)) |
aed3ea94 | 322 | blk_mq_wake_waiters(q); |
055f6e18 ML |
323 | /* Make blk_queue_enter() reexamine the DYING flag. */ |
324 | wake_up_all(&q->mq_freeze_wq); | |
aed3ea94 | 325 | } |
8e141f9e CH |
326 | |
327 | void blk_set_queue_dying(struct request_queue *q) | |
328 | { | |
329 | blk_queue_flag_set(QUEUE_FLAG_DYING, q); | |
330 | blk_queue_start_drain(q); | |
331 | } | |
aed3ea94 JA |
332 | EXPORT_SYMBOL_GPL(blk_set_queue_dying); |
333 | ||
c9a929dd TH |
334 | /** |
335 | * blk_cleanup_queue - shutdown a request queue | |
336 | * @q: request queue to shutdown | |
337 | * | |
c246e80d BVA |
338 | * Mark @q DYING, drain all pending requests, mark @q DEAD, destroy and |
339 | * put it. All future requests will be failed immediately with -ENODEV. | |
e8c7d14a LC |
340 | * |
341 | * Context: can sleep | |
c94a96ac | 342 | */ |
6728cb0e | 343 | void blk_cleanup_queue(struct request_queue *q) |
483f4afc | 344 | { |
e8c7d14a LC |
345 | /* cannot be called from atomic context */ |
346 | might_sleep(); | |
347 | ||
bae85c15 BVA |
348 | WARN_ON_ONCE(blk_queue_registered(q)); |
349 | ||
3f3299d5 | 350 | /* mark @q DYING, no new request or merges will be allowed afterwards */ |
aed3ea94 | 351 | blk_set_queue_dying(q); |
6ecf23af | 352 | |
57d74df9 CH |
353 | blk_queue_flag_set(QUEUE_FLAG_NOMERGES, q); |
354 | blk_queue_flag_set(QUEUE_FLAG_NOXMERGES, q); | |
c9a929dd | 355 | |
c246e80d BVA |
356 | /* |
357 | * Drain all requests queued before DYING marking. Set DEAD flag to | |
67ed8b73 BVA |
358 | * prevent that blk_mq_run_hw_queues() accesses the hardware queues |
359 | * after draining finished. | |
c246e80d | 360 | */ |
3ef28e83 | 361 | blk_freeze_queue(q); |
c57cdf7a | 362 | |
57d74df9 | 363 | blk_queue_flag_set(QUEUE_FLAG_DEAD, q); |
c9a929dd | 364 | |
c9a929dd | 365 | blk_sync_queue(q); |
2a19b28f ML |
366 | if (queue_is_mq(q)) { |
367 | blk_mq_cancel_work_sync(q); | |
c7e2d94b | 368 | blk_mq_exit_queue(q); |
2a19b28f | 369 | } |
a1ce35fa | 370 | |
c3e22192 ML |
371 | /* |
372 | * In theory, request pool of sched_tags belongs to request queue. | |
373 | * However, the current implementation requires tag_set for freeing | |
374 | * requests, so free the pool now. | |
375 | * | |
376 | * Queue has become frozen, there can't be any in-queue requests, so | |
377 | * it is safe to free requests now. | |
378 | */ | |
379 | mutex_lock(&q->sysfs_lock); | |
380 | if (q->elevator) | |
1820f4f0 | 381 | blk_mq_sched_free_rqs(q); |
c3e22192 ML |
382 | mutex_unlock(&q->sysfs_lock); |
383 | ||
3ef28e83 | 384 | percpu_ref_exit(&q->q_usage_counter); |
45a9c9d9 | 385 | |
c9a929dd | 386 | /* @q is and will stay empty, shutdown and put */ |
483f4afc AV |
387 | blk_put_queue(q); |
388 | } | |
1da177e4 LT |
389 | EXPORT_SYMBOL(blk_cleanup_queue); |
390 | ||
3a0a5299 BVA |
391 | /** |
392 | * blk_queue_enter() - try to increase q->q_usage_counter | |
393 | * @q: request queue pointer | |
a4d34da7 | 394 | * @flags: BLK_MQ_REQ_NOWAIT and/or BLK_MQ_REQ_PM |
3a0a5299 | 395 | */ |
9a95e4ef | 396 | int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags) |
3ef28e83 | 397 | { |
a4d34da7 | 398 | const bool pm = flags & BLK_MQ_REQ_PM; |
3a0a5299 | 399 | |
1f14a098 | 400 | while (!blk_try_enter_queue(q, pm)) { |
3a0a5299 | 401 | if (flags & BLK_MQ_REQ_NOWAIT) |
3ef28e83 DW |
402 | return -EBUSY; |
403 | ||
5ed61d3f | 404 | /* |
1f14a098 CH |
405 | * read pair of barrier in blk_freeze_queue_start(), we need to |
406 | * order reading __PERCPU_REF_DEAD flag of .q_usage_counter and | |
407 | * reading .mq_freeze_depth or queue dying flag, otherwise the | |
408 | * following wait may never return if the two reads are | |
409 | * reordered. | |
5ed61d3f ML |
410 | */ |
411 | smp_rmb(); | |
1dc3039b | 412 | wait_event(q->mq_freeze_wq, |
7996a8b5 | 413 | (!q->mq_freeze_depth && |
52abca64 | 414 | blk_pm_resume_queue(pm, q)) || |
1dc3039b | 415 | blk_queue_dying(q)); |
3ef28e83 DW |
416 | if (blk_queue_dying(q)) |
417 | return -ENODEV; | |
3ef28e83 | 418 | } |
1f14a098 CH |
419 | |
420 | return 0; | |
3ef28e83 DW |
421 | } |
422 | ||
c98cb5bb | 423 | int __bio_queue_enter(struct request_queue *q, struct bio *bio) |
accea322 | 424 | { |
a6741536 | 425 | while (!blk_try_enter_queue(q, false)) { |
eab4e027 PB |
426 | struct gendisk *disk = bio->bi_bdev->bd_disk; |
427 | ||
a6741536 | 428 | if (bio->bi_opf & REQ_NOWAIT) { |
8e141f9e | 429 | if (test_bit(GD_DEAD, &disk->state)) |
a6741536 | 430 | goto dead; |
accea322 | 431 | bio_wouldblock_error(bio); |
a6741536 CH |
432 | return -EBUSY; |
433 | } | |
434 | ||
435 | /* | |
436 | * read pair of barrier in blk_freeze_queue_start(), we need to | |
437 | * order reading __PERCPU_REF_DEAD flag of .q_usage_counter and | |
438 | * reading .mq_freeze_depth or queue dying flag, otherwise the | |
439 | * following wait may never return if the two reads are | |
440 | * reordered. | |
441 | */ | |
442 | smp_rmb(); | |
443 | wait_event(q->mq_freeze_wq, | |
444 | (!q->mq_freeze_depth && | |
445 | blk_pm_resume_queue(false, q)) || | |
8e141f9e CH |
446 | test_bit(GD_DEAD, &disk->state)); |
447 | if (test_bit(GD_DEAD, &disk->state)) | |
a6741536 | 448 | goto dead; |
accea322 CH |
449 | } |
450 | ||
a6741536 CH |
451 | return 0; |
452 | dead: | |
453 | bio_io_error(bio); | |
454 | return -ENODEV; | |
accea322 CH |
455 | } |
456 | ||
3ef28e83 DW |
457 | void blk_queue_exit(struct request_queue *q) |
458 | { | |
459 | percpu_ref_put(&q->q_usage_counter); | |
460 | } | |
461 | ||
462 | static void blk_queue_usage_counter_release(struct percpu_ref *ref) | |
463 | { | |
464 | struct request_queue *q = | |
465 | container_of(ref, struct request_queue, q_usage_counter); | |
466 | ||
467 | wake_up_all(&q->mq_freeze_wq); | |
468 | } | |
469 | ||
bca237a5 | 470 | static void blk_rq_timed_out_timer(struct timer_list *t) |
287922eb | 471 | { |
bca237a5 | 472 | struct request_queue *q = from_timer(q, t, timeout); |
287922eb CH |
473 | |
474 | kblockd_schedule_work(&q->timeout_work); | |
475 | } | |
476 | ||
2e3c18d0 TH |
477 | static void blk_timeout_work(struct work_struct *work) |
478 | { | |
479 | } | |
480 | ||
c62b37d9 | 481 | struct request_queue *blk_alloc_queue(int node_id) |
1946089a | 482 | { |
165125e1 | 483 | struct request_queue *q; |
338aa96d | 484 | int ret; |
1946089a | 485 | |
8324aa91 | 486 | q = kmem_cache_alloc_node(blk_requestq_cachep, |
3d745ea5 | 487 | GFP_KERNEL | __GFP_ZERO, node_id); |
1da177e4 LT |
488 | if (!q) |
489 | return NULL; | |
490 | ||
cbf62af3 | 491 | q->last_merge = NULL; |
cbf62af3 | 492 | |
3d745ea5 | 493 | q->id = ida_simple_get(&blk_queue_ida, 0, 0, GFP_KERNEL); |
a73f730d | 494 | if (q->id < 0) |
3d2936f4 | 495 | goto fail_q; |
a73f730d | 496 | |
c495a176 | 497 | ret = bioset_init(&q->bio_split, BIO_POOL_SIZE, 0, 0); |
338aa96d | 498 | if (ret) |
54efd50b KO |
499 | goto fail_id; |
500 | ||
a83b576c JA |
501 | q->stats = blk_alloc_queue_stats(); |
502 | if (!q->stats) | |
edb0872f | 503 | goto fail_split; |
a83b576c | 504 | |
5151412d | 505 | q->node = node_id; |
0989a025 | 506 | |
079a2e3e | 507 | atomic_set(&q->nr_active_requests_shared_tags, 0); |
bccf5e26 | 508 | |
bca237a5 | 509 | timer_setup(&q->timeout, blk_rq_timed_out_timer, 0); |
2e3c18d0 | 510 | INIT_WORK(&q->timeout_work, blk_timeout_work); |
a612fddf | 511 | INIT_LIST_HEAD(&q->icq_list); |
4eef3049 | 512 | #ifdef CONFIG_BLK_CGROUP |
e8989fae | 513 | INIT_LIST_HEAD(&q->blkg_list); |
4eef3049 | 514 | #endif |
483f4afc | 515 | |
8324aa91 | 516 | kobject_init(&q->kobj, &blk_queue_ktype); |
1da177e4 | 517 | |
85e0cbbb | 518 | mutex_init(&q->debugfs_mutex); |
483f4afc | 519 | mutex_init(&q->sysfs_lock); |
cecf5d87 | 520 | mutex_init(&q->sysfs_dir_lock); |
0d945c1f | 521 | spin_lock_init(&q->queue_lock); |
c94a96ac | 522 | |
320ae51f | 523 | init_waitqueue_head(&q->mq_freeze_wq); |
7996a8b5 | 524 | mutex_init(&q->mq_freeze_lock); |
320ae51f | 525 | |
3ef28e83 DW |
526 | /* |
527 | * Init percpu_ref in atomic mode so that it's faster to shutdown. | |
528 | * See blk_register_queue() for details. | |
529 | */ | |
530 | if (percpu_ref_init(&q->q_usage_counter, | |
531 | blk_queue_usage_counter_release, | |
532 | PERCPU_REF_INIT_ATOMIC, GFP_KERNEL)) | |
edb0872f | 533 | goto fail_stats; |
f51b802c | 534 | |
3ef28e83 DW |
535 | if (blkcg_init_queue(q)) |
536 | goto fail_ref; | |
537 | ||
3d745ea5 CH |
538 | blk_queue_dma_alignment(q, 511); |
539 | blk_set_default_limits(&q->limits); | |
d2a27964 | 540 | q->nr_requests = BLKDEV_DEFAULT_RQ; |
3d745ea5 | 541 | |
1da177e4 | 542 | return q; |
a73f730d | 543 | |
3ef28e83 DW |
544 | fail_ref: |
545 | percpu_ref_exit(&q->q_usage_counter); | |
a83b576c | 546 | fail_stats: |
edb0872f | 547 | blk_free_queue_stats(q->stats); |
54efd50b | 548 | fail_split: |
338aa96d | 549 | bioset_exit(&q->bio_split); |
a73f730d TH |
550 | fail_id: |
551 | ida_simple_remove(&blk_queue_ida, q->id); | |
552 | fail_q: | |
553 | kmem_cache_free(blk_requestq_cachep, q); | |
554 | return NULL; | |
1da177e4 | 555 | } |
1da177e4 | 556 | |
b5bd357c LC |
557 | /** |
558 | * blk_get_queue - increment the request_queue refcount | |
559 | * @q: the request_queue structure to increment the refcount for | |
560 | * | |
561 | * Increment the refcount of the request_queue kobject. | |
763b5892 LC |
562 | * |
563 | * Context: Any context. | |
b5bd357c | 564 | */ |
09ac46c4 | 565 | bool blk_get_queue(struct request_queue *q) |
1da177e4 | 566 | { |
3f3299d5 | 567 | if (likely(!blk_queue_dying(q))) { |
09ac46c4 TH |
568 | __blk_get_queue(q); |
569 | return true; | |
1da177e4 LT |
570 | } |
571 | ||
09ac46c4 | 572 | return false; |
1da177e4 | 573 | } |
d86e0e83 | 574 | EXPORT_SYMBOL(blk_get_queue); |
1da177e4 | 575 | |
52c5e62d | 576 | static void handle_bad_sector(struct bio *bio, sector_t maxsector) |
1da177e4 LT |
577 | { |
578 | char b[BDEVNAME_SIZE]; | |
579 | ||
8a3ee677 | 580 | pr_info_ratelimited("%s: attempt to access beyond end of device\n" |
f4ac712e | 581 | "%s: rw=%d, want=%llu, limit=%llu\n", |
8a3ee677 | 582 | current->comm, |
f4ac712e TH |
583 | bio_devname(bio, b), bio->bi_opf, |
584 | bio_end_sector(bio), maxsector); | |
1da177e4 LT |
585 | } |
586 | ||
c17bb495 AM |
587 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
588 | ||
589 | static DECLARE_FAULT_ATTR(fail_make_request); | |
590 | ||
591 | static int __init setup_fail_make_request(char *str) | |
592 | { | |
593 | return setup_fault_attr(&fail_make_request, str); | |
594 | } | |
595 | __setup("fail_make_request=", setup_fail_make_request); | |
596 | ||
8446fe92 | 597 | static bool should_fail_request(struct block_device *part, unsigned int bytes) |
c17bb495 | 598 | { |
8446fe92 | 599 | return part->bd_make_it_fail && should_fail(&fail_make_request, bytes); |
c17bb495 AM |
600 | } |
601 | ||
602 | static int __init fail_make_request_debugfs(void) | |
603 | { | |
dd48c085 AM |
604 | struct dentry *dir = fault_create_debugfs_attr("fail_make_request", |
605 | NULL, &fail_make_request); | |
606 | ||
21f9fcd8 | 607 | return PTR_ERR_OR_ZERO(dir); |
c17bb495 AM |
608 | } |
609 | ||
610 | late_initcall(fail_make_request_debugfs); | |
611 | ||
612 | #else /* CONFIG_FAIL_MAKE_REQUEST */ | |
613 | ||
8446fe92 | 614 | static inline bool should_fail_request(struct block_device *part, |
b2c9cd37 | 615 | unsigned int bytes) |
c17bb495 | 616 | { |
b2c9cd37 | 617 | return false; |
c17bb495 AM |
618 | } |
619 | ||
620 | #endif /* CONFIG_FAIL_MAKE_REQUEST */ | |
621 | ||
2f9f6221 | 622 | static inline bool bio_check_ro(struct bio *bio) |
721c7fc7 | 623 | { |
2f9f6221 | 624 | if (op_is_write(bio_op(bio)) && bdev_read_only(bio->bi_bdev)) { |
721c7fc7 ID |
625 | char b[BDEVNAME_SIZE]; |
626 | ||
8b2ded1c MP |
627 | if (op_is_flush(bio->bi_opf) && !bio_sectors(bio)) |
628 | return false; | |
629 | ||
a32e236e | 630 | WARN_ONCE(1, |
c8178674 | 631 | "Trying to write to read-only block-device %s (partno %d)\n", |
2f9f6221 | 632 | bio_devname(bio, b), bio->bi_bdev->bd_partno); |
a32e236e LT |
633 | /* Older lvm-tools actually trigger this */ |
634 | return false; | |
721c7fc7 ID |
635 | } |
636 | ||
637 | return false; | |
638 | } | |
639 | ||
30abb3a6 HM |
640 | static noinline int should_fail_bio(struct bio *bio) |
641 | { | |
309dca30 | 642 | if (should_fail_request(bdev_whole(bio->bi_bdev), bio->bi_iter.bi_size)) |
30abb3a6 HM |
643 | return -EIO; |
644 | return 0; | |
645 | } | |
646 | ALLOW_ERROR_INJECTION(should_fail_bio, ERRNO); | |
647 | ||
52c5e62d CH |
648 | /* |
649 | * Check whether this bio extends beyond the end of the device or partition. | |
650 | * This may well happen - the kernel calls bread() without checking the size of | |
651 | * the device, e.g., when mounting a file system. | |
652 | */ | |
2f9f6221 | 653 | static inline int bio_check_eod(struct bio *bio) |
52c5e62d | 654 | { |
2f9f6221 | 655 | sector_t maxsector = bdev_nr_sectors(bio->bi_bdev); |
52c5e62d CH |
656 | unsigned int nr_sectors = bio_sectors(bio); |
657 | ||
658 | if (nr_sectors && maxsector && | |
659 | (nr_sectors > maxsector || | |
660 | bio->bi_iter.bi_sector > maxsector - nr_sectors)) { | |
661 | handle_bad_sector(bio, maxsector); | |
662 | return -EIO; | |
663 | } | |
664 | return 0; | |
665 | } | |
666 | ||
74d46992 CH |
667 | /* |
668 | * Remap block n of partition p to block n+start(p) of the disk. | |
669 | */ | |
2f9f6221 | 670 | static int blk_partition_remap(struct bio *bio) |
74d46992 | 671 | { |
309dca30 | 672 | struct block_device *p = bio->bi_bdev; |
74d46992 | 673 | |
52c5e62d | 674 | if (unlikely(should_fail_request(p, bio->bi_iter.bi_size))) |
2f9f6221 | 675 | return -EIO; |
5eac3eb3 | 676 | if (bio_sectors(bio)) { |
8446fe92 | 677 | bio->bi_iter.bi_sector += p->bd_start_sect; |
1c02fca6 | 678 | trace_block_bio_remap(bio, p->bd_dev, |
29ff57c6 | 679 | bio->bi_iter.bi_sector - |
8446fe92 | 680 | p->bd_start_sect); |
52c5e62d | 681 | } |
30c5d345 | 682 | bio_set_flag(bio, BIO_REMAPPED); |
2f9f6221 | 683 | return 0; |
74d46992 CH |
684 | } |
685 | ||
0512a75b KB |
686 | /* |
687 | * Check write append to a zoned block device. | |
688 | */ | |
689 | static inline blk_status_t blk_check_zone_append(struct request_queue *q, | |
690 | struct bio *bio) | |
691 | { | |
692 | sector_t pos = bio->bi_iter.bi_sector; | |
693 | int nr_sectors = bio_sectors(bio); | |
694 | ||
695 | /* Only applicable to zoned block devices */ | |
696 | if (!blk_queue_is_zoned(q)) | |
697 | return BLK_STS_NOTSUPP; | |
698 | ||
699 | /* The bio sector must point to the start of a sequential zone */ | |
700 | if (pos & (blk_queue_zone_sectors(q) - 1) || | |
701 | !blk_queue_zone_is_seq(q, pos)) | |
702 | return BLK_STS_IOERR; | |
703 | ||
704 | /* | |
705 | * Not allowed to cross zone boundaries. Otherwise, the BIO will be | |
706 | * split and could result in non-contiguous sectors being written in | |
707 | * different zones. | |
708 | */ | |
709 | if (nr_sectors > q->limits.chunk_sectors) | |
710 | return BLK_STS_IOERR; | |
711 | ||
712 | /* Make sure the BIO is small enough and will not get split */ | |
713 | if (nr_sectors > q->limits.max_zone_append_sectors) | |
714 | return BLK_STS_IOERR; | |
715 | ||
716 | bio->bi_opf |= REQ_NOMERGE; | |
717 | ||
718 | return BLK_STS_OK; | |
719 | } | |
720 | ||
900e0807 | 721 | noinline_for_stack bool submit_bio_checks(struct bio *bio) |
1da177e4 | 722 | { |
309dca30 | 723 | struct block_device *bdev = bio->bi_bdev; |
eab4e027 | 724 | struct request_queue *q = bdev_get_queue(bdev); |
4e4cbee9 | 725 | blk_status_t status = BLK_STS_IOERR; |
5a473e83 | 726 | struct blk_plug *plug; |
1da177e4 LT |
727 | |
728 | might_sleep(); | |
1da177e4 | 729 | |
5a473e83 JA |
730 | plug = blk_mq_plug(q, bio); |
731 | if (plug && plug->nowait) | |
732 | bio->bi_opf |= REQ_NOWAIT; | |
733 | ||
03a07c92 | 734 | /* |
b0beb280 | 735 | * For a REQ_NOWAIT based request, return -EOPNOTSUPP |
021a2446 | 736 | * if queue does not support NOWAIT. |
03a07c92 | 737 | */ |
021a2446 | 738 | if ((bio->bi_opf & REQ_NOWAIT) && !blk_queue_nowait(q)) |
b0beb280 | 739 | goto not_supported; |
03a07c92 | 740 | |
30abb3a6 | 741 | if (should_fail_bio(bio)) |
5a7bbad2 | 742 | goto end_io; |
2f9f6221 CH |
743 | if (unlikely(bio_check_ro(bio))) |
744 | goto end_io; | |
3a905c37 CH |
745 | if (!bio_flagged(bio, BIO_REMAPPED)) { |
746 | if (unlikely(bio_check_eod(bio))) | |
747 | goto end_io; | |
748 | if (bdev->bd_partno && unlikely(blk_partition_remap(bio))) | |
749 | goto end_io; | |
750 | } | |
2056a782 | 751 | |
5a7bbad2 | 752 | /* |
ed00aabd CH |
753 | * Filter flush bio's early so that bio based drivers without flush |
754 | * support don't have to worry about them. | |
5a7bbad2 | 755 | */ |
f3a8ab7d | 756 | if (op_is_flush(bio->bi_opf) && |
c888a8f9 | 757 | !test_bit(QUEUE_FLAG_WC, &q->queue_flags)) { |
1eff9d32 | 758 | bio->bi_opf &= ~(REQ_PREFLUSH | REQ_FUA); |
e439ab71 | 759 | if (!bio_sectors(bio)) { |
4e4cbee9 | 760 | status = BLK_STS_OK; |
51fd77bd JA |
761 | goto end_io; |
762 | } | |
5a7bbad2 | 763 | } |
5ddfe969 | 764 | |
d04c406f | 765 | if (!test_bit(QUEUE_FLAG_POLL, &q->queue_flags)) |
6ce913fe | 766 | bio_clear_polled(bio); |
d04c406f | 767 | |
288dab8a CH |
768 | switch (bio_op(bio)) { |
769 | case REQ_OP_DISCARD: | |
770 | if (!blk_queue_discard(q)) | |
771 | goto not_supported; | |
772 | break; | |
773 | case REQ_OP_SECURE_ERASE: | |
774 | if (!blk_queue_secure_erase(q)) | |
775 | goto not_supported; | |
776 | break; | |
777 | case REQ_OP_WRITE_SAME: | |
74d46992 | 778 | if (!q->limits.max_write_same_sectors) |
288dab8a | 779 | goto not_supported; |
58886785 | 780 | break; |
0512a75b KB |
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; | |
2d253440 | 786 | case REQ_OP_ZONE_RESET: |
6c1b1da5 AJ |
787 | case REQ_OP_ZONE_OPEN: |
788 | case REQ_OP_ZONE_CLOSE: | |
789 | case REQ_OP_ZONE_FINISH: | |
74d46992 | 790 | if (!blk_queue_is_zoned(q)) |
2d253440 | 791 | goto not_supported; |
288dab8a | 792 | break; |
6e33dbf2 CK |
793 | case REQ_OP_ZONE_RESET_ALL: |
794 | if (!blk_queue_is_zoned(q) || !blk_queue_zone_resetall(q)) | |
795 | goto not_supported; | |
796 | break; | |
a6f0788e | 797 | case REQ_OP_WRITE_ZEROES: |
74d46992 | 798 | if (!q->limits.max_write_zeroes_sectors) |
a6f0788e CK |
799 | goto not_supported; |
800 | break; | |
288dab8a CH |
801 | default: |
802 | break; | |
5a7bbad2 | 803 | } |
01edede4 | 804 | |
7f4b35d1 | 805 | /* |
3e82c348 CH |
806 | * Various block parts want %current->io_context, so allocate it up |
807 | * front rather than dealing with lots of pain to allocate it only | |
808 | * where needed. This may fail and the block layer knows how to live | |
809 | * with it. | |
7f4b35d1 | 810 | */ |
3e82c348 CH |
811 | if (unlikely(!current->io_context)) |
812 | create_task_io_context(current, GFP_ATOMIC, q->node); | |
7f4b35d1 | 813 | |
b781d8db | 814 | if (blk_throtl_bio(bio)) |
ae118896 | 815 | return false; |
db18a53e CH |
816 | |
817 | blk_cgroup_bio_start(bio); | |
818 | blkcg_bio_issue_init(bio); | |
27a84d54 | 819 | |
fbbaf700 | 820 | if (!bio_flagged(bio, BIO_TRACE_COMPLETION)) { |
e8a676d6 | 821 | trace_block_bio_queue(bio); |
fbbaf700 N |
822 | /* Now that enqueuing has been traced, we need to trace |
823 | * completion as well. | |
824 | */ | |
825 | bio_set_flag(bio, BIO_TRACE_COMPLETION); | |
826 | } | |
27a84d54 | 827 | return true; |
a7384677 | 828 | |
288dab8a | 829 | not_supported: |
4e4cbee9 | 830 | status = BLK_STS_NOTSUPP; |
a7384677 | 831 | end_io: |
4e4cbee9 | 832 | bio->bi_status = status; |
4246a0b6 | 833 | bio_endio(bio); |
27a84d54 | 834 | return false; |
1da177e4 LT |
835 | } |
836 | ||
900e0807 | 837 | static void __submit_bio_fops(struct gendisk *disk, struct bio *bio) |
ac7c5675 | 838 | { |
cc9c884d | 839 | if (unlikely(bio_queue_enter(bio) != 0)) |
3e08773c | 840 | return; |
900e0807 JA |
841 | if (submit_bio_checks(bio) && blk_crypto_bio_prep(&bio)) |
842 | disk->fops->submit_bio(bio); | |
843 | blk_queue_exit(disk->queue); | |
844 | } | |
cc9c884d | 845 | |
900e0807 JA |
846 | static void __submit_bio(struct bio *bio) |
847 | { | |
848 | struct gendisk *disk = bio->bi_bdev->bd_disk; | |
cc9c884d | 849 | |
900e0807 | 850 | if (!disk->fops->submit_bio) |
3e08773c | 851 | blk_mq_submit_bio(bio); |
900e0807 JA |
852 | else |
853 | __submit_bio_fops(disk, bio); | |
ac7c5675 CH |
854 | } |
855 | ||
566acf2d CH |
856 | /* |
857 | * The loop in this function may be a bit non-obvious, and so deserves some | |
858 | * explanation: | |
859 | * | |
860 | * - Before entering the loop, bio->bi_next is NULL (as all callers ensure | |
861 | * that), so we have a list with a single bio. | |
862 | * - We pretend that we have just taken it off a longer list, so we assign | |
863 | * bio_list to a pointer to the bio_list_on_stack, thus initialising the | |
864 | * bio_list of new bios to be added. ->submit_bio() may indeed add some more | |
865 | * bios through a recursive call to submit_bio_noacct. If it did, we find a | |
866 | * non-NULL value in bio_list and re-enter the loop from the top. | |
867 | * - In this case we really did just take the bio of the top of the list (no | |
868 | * pretending) and so remove it from bio_list, and call into ->submit_bio() | |
869 | * again. | |
870 | * | |
871 | * bio_list_on_stack[0] contains bios submitted by the current ->submit_bio. | |
872 | * bio_list_on_stack[1] contains bios that were submitted before the current | |
873 | * ->submit_bio_bio, but that haven't been processed yet. | |
874 | */ | |
3e08773c | 875 | static void __submit_bio_noacct(struct bio *bio) |
566acf2d CH |
876 | { |
877 | struct bio_list bio_list_on_stack[2]; | |
566acf2d CH |
878 | |
879 | BUG_ON(bio->bi_next); | |
880 | ||
881 | bio_list_init(&bio_list_on_stack[0]); | |
882 | current->bio_list = bio_list_on_stack; | |
883 | ||
884 | do { | |
eab4e027 | 885 | struct request_queue *q = bdev_get_queue(bio->bi_bdev); |
566acf2d CH |
886 | struct bio_list lower, same; |
887 | ||
566acf2d CH |
888 | /* |
889 | * Create a fresh bio_list for all subordinate requests. | |
890 | */ | |
891 | bio_list_on_stack[1] = bio_list_on_stack[0]; | |
892 | bio_list_init(&bio_list_on_stack[0]); | |
893 | ||
3e08773c | 894 | __submit_bio(bio); |
566acf2d CH |
895 | |
896 | /* | |
897 | * Sort new bios into those for a lower level and those for the | |
898 | * same level. | |
899 | */ | |
900 | bio_list_init(&lower); | |
901 | bio_list_init(&same); | |
902 | while ((bio = bio_list_pop(&bio_list_on_stack[0])) != NULL) | |
eab4e027 | 903 | if (q == bdev_get_queue(bio->bi_bdev)) |
566acf2d CH |
904 | bio_list_add(&same, bio); |
905 | else | |
906 | bio_list_add(&lower, bio); | |
907 | ||
908 | /* | |
909 | * Now assemble so we handle the lowest level first. | |
910 | */ | |
911 | bio_list_merge(&bio_list_on_stack[0], &lower); | |
912 | bio_list_merge(&bio_list_on_stack[0], &same); | |
913 | bio_list_merge(&bio_list_on_stack[0], &bio_list_on_stack[1]); | |
914 | } while ((bio = bio_list_pop(&bio_list_on_stack[0]))); | |
915 | ||
916 | current->bio_list = NULL; | |
566acf2d CH |
917 | } |
918 | ||
3e08773c | 919 | static void __submit_bio_noacct_mq(struct bio *bio) |
ff93ea0c | 920 | { |
7c792f33 | 921 | struct bio_list bio_list[2] = { }; |
ff93ea0c | 922 | |
7c792f33 | 923 | current->bio_list = bio_list; |
ff93ea0c CH |
924 | |
925 | do { | |
3e08773c | 926 | __submit_bio(bio); |
7c792f33 | 927 | } while ((bio = bio_list_pop(&bio_list[0]))); |
ff93ea0c CH |
928 | |
929 | current->bio_list = NULL; | |
ff93ea0c CH |
930 | } |
931 | ||
27a84d54 | 932 | /** |
ed00aabd | 933 | * submit_bio_noacct - re-submit a bio to the block device layer for I/O |
27a84d54 CH |
934 | * @bio: The bio describing the location in memory and on the device. |
935 | * | |
3fdd4086 CH |
936 | * This is a version of submit_bio() that shall only be used for I/O that is |
937 | * resubmitted to lower level drivers by stacking block drivers. All file | |
938 | * systems and other upper level users of the block layer should use | |
939 | * submit_bio() instead. | |
d89d8796 | 940 | */ |
3e08773c | 941 | void submit_bio_noacct(struct bio *bio) |
d89d8796 | 942 | { |
27a84d54 | 943 | /* |
566acf2d CH |
944 | * We only want one ->submit_bio to be active at a time, else stack |
945 | * usage with stacked devices could be a problem. Use current->bio_list | |
946 | * to collect a list of requests submited by a ->submit_bio method while | |
947 | * it is active, and then process them after it returned. | |
27a84d54 | 948 | */ |
3e08773c | 949 | if (current->bio_list) |
f5fe1b51 | 950 | bio_list_add(¤t->bio_list[0], bio); |
3e08773c CH |
951 | else if (!bio->bi_bdev->bd_disk->fops->submit_bio) |
952 | __submit_bio_noacct_mq(bio); | |
953 | else | |
954 | __submit_bio_noacct(bio); | |
d89d8796 | 955 | } |
ed00aabd | 956 | EXPORT_SYMBOL(submit_bio_noacct); |
1da177e4 LT |
957 | |
958 | /** | |
710027a4 | 959 | * submit_bio - submit a bio to the block device layer for I/O |
1da177e4 LT |
960 | * @bio: The &struct bio which describes the I/O |
961 | * | |
3fdd4086 CH |
962 | * submit_bio() is used to submit I/O requests to block devices. It is passed a |
963 | * fully set up &struct bio that describes the I/O that needs to be done. The | |
309dca30 | 964 | * bio will be send to the device described by the bi_bdev field. |
1da177e4 | 965 | * |
3fdd4086 CH |
966 | * The success/failure status of the request, along with notification of |
967 | * completion, is delivered asynchronously through the ->bi_end_io() callback | |
968 | * in @bio. The bio must NOT be touched by thecaller until ->bi_end_io() has | |
969 | * been called. | |
1da177e4 | 970 | */ |
3e08773c | 971 | void submit_bio(struct bio *bio) |
1da177e4 | 972 | { |
d3f77dfd | 973 | if (blkcg_punt_bio_submit(bio)) |
3e08773c | 974 | return; |
d3f77dfd | 975 | |
bf2de6f5 JA |
976 | /* |
977 | * If it's a regular read/write or a barrier with data attached, | |
978 | * go through the normal accounting stuff before submission. | |
979 | */ | |
e2a60da7 | 980 | if (bio_has_data(bio)) { |
4363ac7c MP |
981 | unsigned int count; |
982 | ||
95fe6c1a | 983 | if (unlikely(bio_op(bio) == REQ_OP_WRITE_SAME)) |
309dca30 | 984 | count = queue_logical_block_size( |
eab4e027 | 985 | bdev_get_queue(bio->bi_bdev)) >> 9; |
4363ac7c MP |
986 | else |
987 | count = bio_sectors(bio); | |
988 | ||
a8ebb056 | 989 | if (op_is_write(bio_op(bio))) { |
bf2de6f5 JA |
990 | count_vm_events(PGPGOUT, count); |
991 | } else { | |
4f024f37 | 992 | task_io_account_read(bio->bi_iter.bi_size); |
bf2de6f5 JA |
993 | count_vm_events(PGPGIN, count); |
994 | } | |
1da177e4 LT |
995 | } |
996 | ||
b8e24a93 | 997 | /* |
760f83ea CH |
998 | * If we're reading data that is part of the userspace workingset, count |
999 | * submission time as memory stall. When the device is congested, or | |
1000 | * the submitting cgroup IO-throttled, submission can be a significant | |
1001 | * part of overall IO time. | |
b8e24a93 | 1002 | */ |
760f83ea CH |
1003 | if (unlikely(bio_op(bio) == REQ_OP_READ && |
1004 | bio_flagged(bio, BIO_WORKINGSET))) { | |
1005 | unsigned long pflags; | |
b8e24a93 | 1006 | |
760f83ea | 1007 | psi_memstall_enter(&pflags); |
3e08773c | 1008 | submit_bio_noacct(bio); |
b8e24a93 | 1009 | psi_memstall_leave(&pflags); |
3e08773c | 1010 | return; |
760f83ea CH |
1011 | } |
1012 | ||
3e08773c | 1013 | submit_bio_noacct(bio); |
1da177e4 | 1014 | } |
1da177e4 LT |
1015 | EXPORT_SYMBOL(submit_bio); |
1016 | ||
3e08773c CH |
1017 | /** |
1018 | * bio_poll - poll for BIO completions | |
1019 | * @bio: bio to poll for | |
1020 | * @flags: BLK_POLL_* flags that control the behavior | |
1021 | * | |
1022 | * Poll for completions on queue associated with the bio. Returns number of | |
1023 | * completed entries found. | |
1024 | * | |
1025 | * Note: the caller must either be the context that submitted @bio, or | |
1026 | * be in a RCU critical section to prevent freeing of @bio. | |
1027 | */ | |
5a72e899 | 1028 | int bio_poll(struct bio *bio, struct io_comp_batch *iob, unsigned int flags) |
3e08773c | 1029 | { |
859897c3 | 1030 | struct request_queue *q = bdev_get_queue(bio->bi_bdev); |
3e08773c CH |
1031 | blk_qc_t cookie = READ_ONCE(bio->bi_cookie); |
1032 | int ret; | |
1033 | ||
1034 | if (cookie == BLK_QC_T_NONE || | |
1035 | !test_bit(QUEUE_FLAG_POLL, &q->queue_flags)) | |
1036 | return 0; | |
1037 | ||
1038 | if (current->plug) | |
008f75a2 | 1039 | blk_flush_plug(current->plug, false); |
3e08773c CH |
1040 | |
1041 | if (blk_queue_enter(q, BLK_MQ_REQ_NOWAIT)) | |
1042 | return 0; | |
1043 | if (WARN_ON_ONCE(!queue_is_mq(q))) | |
1044 | ret = 0; /* not yet implemented, should not happen */ | |
1045 | else | |
5a72e899 | 1046 | ret = blk_mq_poll(q, cookie, iob, flags); |
3e08773c CH |
1047 | blk_queue_exit(q); |
1048 | return ret; | |
1049 | } | |
1050 | EXPORT_SYMBOL_GPL(bio_poll); | |
1051 | ||
1052 | /* | |
1053 | * Helper to implement file_operations.iopoll. Requires the bio to be stored | |
1054 | * in iocb->private, and cleared before freeing the bio. | |
1055 | */ | |
5a72e899 JA |
1056 | int iocb_bio_iopoll(struct kiocb *kiocb, struct io_comp_batch *iob, |
1057 | unsigned int flags) | |
3e08773c CH |
1058 | { |
1059 | struct bio *bio; | |
1060 | int ret = 0; | |
1061 | ||
1062 | /* | |
1063 | * Note: the bio cache only uses SLAB_TYPESAFE_BY_RCU, so bio can | |
1064 | * point to a freshly allocated bio at this point. If that happens | |
1065 | * we have a few cases to consider: | |
1066 | * | |
1067 | * 1) the bio is beeing initialized and bi_bdev is NULL. We can just | |
1068 | * simply nothing in this case | |
1069 | * 2) the bio points to a not poll enabled device. bio_poll will catch | |
1070 | * this and return 0 | |
1071 | * 3) the bio points to a poll capable device, including but not | |
1072 | * limited to the one that the original bio pointed to. In this | |
1073 | * case we will call into the actual poll method and poll for I/O, | |
1074 | * even if we don't need to, but it won't cause harm either. | |
1075 | * | |
1076 | * For cases 2) and 3) above the RCU grace period ensures that bi_bdev | |
1077 | * is still allocated. Because partitions hold a reference to the whole | |
1078 | * device bdev and thus disk, the disk is also still valid. Grabbing | |
1079 | * a reference to the queue in bio_poll() ensures the hctxs and requests | |
1080 | * are still valid as well. | |
1081 | */ | |
1082 | rcu_read_lock(); | |
1083 | bio = READ_ONCE(kiocb->private); | |
1084 | if (bio && bio->bi_bdev) | |
5a72e899 | 1085 | ret = bio_poll(bio, iob, flags); |
3e08773c CH |
1086 | rcu_read_unlock(); |
1087 | ||
1088 | return ret; | |
1089 | } | |
1090 | EXPORT_SYMBOL_GPL(iocb_bio_iopoll); | |
1091 | ||
82124d60 | 1092 | /** |
bf4e6b4e | 1093 | * blk_cloned_rq_check_limits - Helper function to check a cloned request |
0d720318 | 1094 | * for the new queue limits |
82124d60 KU |
1095 | * @q: the queue |
1096 | * @rq: the request being checked | |
1097 | * | |
1098 | * Description: | |
1099 | * @rq may have been made based on weaker limitations of upper-level queues | |
1100 | * in request stacking drivers, and it may violate the limitation of @q. | |
1101 | * Since the block layer and the underlying device driver trust @rq | |
1102 | * after it is inserted to @q, it should be checked against @q before | |
1103 | * the insertion using this generic function. | |
1104 | * | |
82124d60 | 1105 | * Request stacking drivers like request-based dm may change the queue |
bf4e6b4e HR |
1106 | * limits when retrying requests on other queues. Those requests need |
1107 | * to be checked against the new queue limits again during dispatch. | |
82124d60 | 1108 | */ |
143d2600 | 1109 | static blk_status_t blk_cloned_rq_check_limits(struct request_queue *q, |
bf4e6b4e | 1110 | struct request *rq) |
82124d60 | 1111 | { |
8327cce5 RS |
1112 | unsigned int max_sectors = blk_queue_get_max_sectors(q, req_op(rq)); |
1113 | ||
1114 | if (blk_rq_sectors(rq) > max_sectors) { | |
1115 | /* | |
1116 | * SCSI device does not have a good way to return if | |
1117 | * Write Same/Zero is actually supported. If a device rejects | |
1118 | * a non-read/write command (discard, write same,etc.) the | |
1119 | * low-level device driver will set the relevant queue limit to | |
1120 | * 0 to prevent blk-lib from issuing more of the offending | |
1121 | * operations. Commands queued prior to the queue limit being | |
1122 | * reset need to be completed with BLK_STS_NOTSUPP to avoid I/O | |
1123 | * errors being propagated to upper layers. | |
1124 | */ | |
1125 | if (max_sectors == 0) | |
1126 | return BLK_STS_NOTSUPP; | |
1127 | ||
61939b12 | 1128 | printk(KERN_ERR "%s: over max size limit. (%u > %u)\n", |
8327cce5 | 1129 | __func__, blk_rq_sectors(rq), max_sectors); |
143d2600 | 1130 | return BLK_STS_IOERR; |
82124d60 KU |
1131 | } |
1132 | ||
1133 | /* | |
9bb33f24 CH |
1134 | * The queue settings related to segment counting may differ from the |
1135 | * original queue. | |
82124d60 | 1136 | */ |
e9cd19c0 | 1137 | rq->nr_phys_segments = blk_recalc_rq_segments(rq); |
8a78362c | 1138 | if (rq->nr_phys_segments > queue_max_segments(q)) { |
61939b12 JP |
1139 | printk(KERN_ERR "%s: over max segments limit. (%hu > %hu)\n", |
1140 | __func__, rq->nr_phys_segments, queue_max_segments(q)); | |
143d2600 | 1141 | return BLK_STS_IOERR; |
82124d60 KU |
1142 | } |
1143 | ||
143d2600 | 1144 | return BLK_STS_OK; |
82124d60 | 1145 | } |
82124d60 KU |
1146 | |
1147 | /** | |
1148 | * blk_insert_cloned_request - Helper for stacking drivers to submit a request | |
1149 | * @q: the queue to submit the request | |
1150 | * @rq: the request being queued | |
1151 | */ | |
2a842aca | 1152 | blk_status_t blk_insert_cloned_request(struct request_queue *q, struct request *rq) |
82124d60 | 1153 | { |
8327cce5 RS |
1154 | blk_status_t ret; |
1155 | ||
1156 | ret = blk_cloned_rq_check_limits(q, rq); | |
1157 | if (ret != BLK_STS_OK) | |
1158 | return ret; | |
82124d60 | 1159 | |
b2c9cd37 | 1160 | if (rq->rq_disk && |
8446fe92 | 1161 | should_fail_request(rq->rq_disk->part0, blk_rq_bytes(rq))) |
2a842aca | 1162 | return BLK_STS_IOERR; |
82124d60 | 1163 | |
a892c8d5 ST |
1164 | if (blk_crypto_insert_cloned_request(rq)) |
1165 | return BLK_STS_IOERR; | |
1166 | ||
be6bfe36 | 1167 | blk_account_io_start(rq); |
82124d60 KU |
1168 | |
1169 | /* | |
a1ce35fa JA |
1170 | * Since we have a scheduler attached on the top device, |
1171 | * bypass a potential scheduler on the bottom device for | |
1172 | * insert. | |
82124d60 | 1173 | */ |
fd9c40f6 | 1174 | return blk_mq_request_issue_directly(rq, true); |
82124d60 KU |
1175 | } |
1176 | EXPORT_SYMBOL_GPL(blk_insert_cloned_request); | |
1177 | ||
80a761fd TH |
1178 | /** |
1179 | * blk_rq_err_bytes - determine number of bytes till the next failure boundary | |
1180 | * @rq: request to examine | |
1181 | * | |
1182 | * Description: | |
1183 | * A request could be merge of IOs which require different failure | |
1184 | * handling. This function determines the number of bytes which | |
1185 | * can be failed from the beginning of the request without | |
1186 | * crossing into area which need to be retried further. | |
1187 | * | |
1188 | * Return: | |
1189 | * The number of bytes to fail. | |
80a761fd TH |
1190 | */ |
1191 | unsigned int blk_rq_err_bytes(const struct request *rq) | |
1192 | { | |
1193 | unsigned int ff = rq->cmd_flags & REQ_FAILFAST_MASK; | |
1194 | unsigned int bytes = 0; | |
1195 | struct bio *bio; | |
1196 | ||
e8064021 | 1197 | if (!(rq->rq_flags & RQF_MIXED_MERGE)) |
80a761fd TH |
1198 | return blk_rq_bytes(rq); |
1199 | ||
1200 | /* | |
1201 | * Currently the only 'mixing' which can happen is between | |
1202 | * different fastfail types. We can safely fail portions | |
1203 | * which have all the failfast bits that the first one has - | |
1204 | * the ones which are at least as eager to fail as the first | |
1205 | * one. | |
1206 | */ | |
1207 | for (bio = rq->bio; bio; bio = bio->bi_next) { | |
1eff9d32 | 1208 | if ((bio->bi_opf & ff) != ff) |
80a761fd | 1209 | break; |
4f024f37 | 1210 | bytes += bio->bi_iter.bi_size; |
80a761fd TH |
1211 | } |
1212 | ||
1213 | /* this could lead to infinite loop */ | |
1214 | BUG_ON(blk_rq_bytes(rq) && !bytes); | |
1215 | return bytes; | |
1216 | } | |
1217 | EXPORT_SYMBOL_GPL(blk_rq_err_bytes); | |
1218 | ||
8446fe92 CH |
1219 | static void update_io_ticks(struct block_device *part, unsigned long now, |
1220 | bool end) | |
9123bf6f CH |
1221 | { |
1222 | unsigned long stamp; | |
1223 | again: | |
8446fe92 | 1224 | stamp = READ_ONCE(part->bd_stamp); |
d80c228d | 1225 | if (unlikely(time_after(now, stamp))) { |
8446fe92 | 1226 | if (likely(cmpxchg(&part->bd_stamp, stamp, now) == stamp)) |
9123bf6f CH |
1227 | __part_stat_add(part, io_ticks, end ? now - stamp : 1); |
1228 | } | |
8446fe92 CH |
1229 | if (part->bd_partno) { |
1230 | part = bdev_whole(part); | |
9123bf6f CH |
1231 | goto again; |
1232 | } | |
1233 | } | |
1234 | ||
be6bfe36 | 1235 | void __blk_account_io_done(struct request *req, u64 now) |
bc58ba94 | 1236 | { |
be6bfe36 | 1237 | const int sgrp = op_stat_group(req_op(req)); |
bc58ba94 | 1238 | |
be6bfe36 PB |
1239 | part_stat_lock(); |
1240 | update_io_ticks(req->part, jiffies, true); | |
1241 | part_stat_inc(req->part, ios[sgrp]); | |
1242 | part_stat_add(req->part, nsecs[sgrp], now - req->start_time_ns); | |
1243 | part_stat_unlock(); | |
bc58ba94 JA |
1244 | } |
1245 | ||
be6bfe36 | 1246 | void __blk_account_io_start(struct request *rq) |
320ae51f | 1247 | { |
0b6e522c CH |
1248 | /* passthrough requests can hold bios that do not have ->bi_bdev set */ |
1249 | if (rq->bio && rq->bio->bi_bdev) | |
1250 | rq->part = rq->bio->bi_bdev; | |
1251 | else | |
1252 | rq->part = rq->rq_disk->part0; | |
524f9ffd | 1253 | |
112f158f | 1254 | part_stat_lock(); |
76268f3a | 1255 | update_io_ticks(rq->part, jiffies, false); |
320ae51f JA |
1256 | part_stat_unlock(); |
1257 | } | |
320ae51f | 1258 | |
8446fe92 | 1259 | static unsigned long __part_start_io_acct(struct block_device *part, |
7b26410b | 1260 | unsigned int sectors, unsigned int op) |
956d510e | 1261 | { |
956d510e CH |
1262 | const int sgrp = op_stat_group(op); |
1263 | unsigned long now = READ_ONCE(jiffies); | |
1264 | ||
1265 | part_stat_lock(); | |
1266 | update_io_ticks(part, now, false); | |
1267 | part_stat_inc(part, ios[sgrp]); | |
1268 | part_stat_add(part, sectors[sgrp], sectors); | |
1269 | part_stat_local_inc(part, in_flight[op_is_write(op)]); | |
1270 | part_stat_unlock(); | |
320ae51f | 1271 | |
956d510e CH |
1272 | return now; |
1273 | } | |
7b26410b | 1274 | |
99dfc43e CH |
1275 | /** |
1276 | * bio_start_io_acct - start I/O accounting for bio based drivers | |
1277 | * @bio: bio to start account for | |
1278 | * | |
1279 | * Returns the start time that should be passed back to bio_end_io_acct(). | |
1280 | */ | |
1281 | unsigned long bio_start_io_acct(struct bio *bio) | |
7b26410b | 1282 | { |
99dfc43e | 1283 | return __part_start_io_acct(bio->bi_bdev, bio_sectors(bio), bio_op(bio)); |
7b26410b | 1284 | } |
99dfc43e | 1285 | EXPORT_SYMBOL_GPL(bio_start_io_acct); |
7b26410b SL |
1286 | |
1287 | unsigned long disk_start_io_acct(struct gendisk *disk, unsigned int sectors, | |
1288 | unsigned int op) | |
1289 | { | |
8446fe92 | 1290 | return __part_start_io_acct(disk->part0, sectors, op); |
7b26410b | 1291 | } |
956d510e CH |
1292 | EXPORT_SYMBOL(disk_start_io_acct); |
1293 | ||
8446fe92 | 1294 | static void __part_end_io_acct(struct block_device *part, unsigned int op, |
7b26410b | 1295 | unsigned long start_time) |
956d510e | 1296 | { |
956d510e CH |
1297 | const int sgrp = op_stat_group(op); |
1298 | unsigned long now = READ_ONCE(jiffies); | |
1299 | unsigned long duration = now - start_time; | |
5b18b5a7 | 1300 | |
956d510e CH |
1301 | part_stat_lock(); |
1302 | update_io_ticks(part, now, true); | |
1303 | part_stat_add(part, nsecs[sgrp], jiffies_to_nsecs(duration)); | |
1304 | part_stat_local_dec(part, in_flight[op_is_write(op)]); | |
320ae51f JA |
1305 | part_stat_unlock(); |
1306 | } | |
7b26410b | 1307 | |
99dfc43e CH |
1308 | void bio_end_io_acct_remapped(struct bio *bio, unsigned long start_time, |
1309 | struct block_device *orig_bdev) | |
7b26410b | 1310 | { |
99dfc43e | 1311 | __part_end_io_acct(orig_bdev, bio_op(bio), start_time); |
7b26410b | 1312 | } |
99dfc43e | 1313 | EXPORT_SYMBOL_GPL(bio_end_io_acct_remapped); |
7b26410b SL |
1314 | |
1315 | void disk_end_io_acct(struct gendisk *disk, unsigned int op, | |
1316 | unsigned long start_time) | |
1317 | { | |
8446fe92 | 1318 | __part_end_io_acct(disk->part0, op, start_time); |
7b26410b | 1319 | } |
956d510e | 1320 | EXPORT_SYMBOL(disk_end_io_acct); |
320ae51f | 1321 | |
ef71de8b CH |
1322 | /* |
1323 | * Steal bios from a request and add them to a bio list. | |
1324 | * The request must not have been partially completed before. | |
1325 | */ | |
1326 | void blk_steal_bios(struct bio_list *list, struct request *rq) | |
1327 | { | |
1328 | if (rq->bio) { | |
1329 | if (list->tail) | |
1330 | list->tail->bi_next = rq->bio; | |
1331 | else | |
1332 | list->head = rq->bio; | |
1333 | list->tail = rq->biotail; | |
1334 | ||
1335 | rq->bio = NULL; | |
1336 | rq->biotail = NULL; | |
1337 | } | |
1338 | ||
1339 | rq->__data_len = 0; | |
1340 | } | |
1341 | EXPORT_SYMBOL_GPL(blk_steal_bios); | |
1342 | ||
2d4dc890 IL |
1343 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE |
1344 | /** | |
1345 | * rq_flush_dcache_pages - Helper function to flush all pages in a request | |
1346 | * @rq: the request to be flushed | |
1347 | * | |
1348 | * Description: | |
1349 | * Flush all pages in @rq. | |
1350 | */ | |
1351 | void rq_flush_dcache_pages(struct request *rq) | |
1352 | { | |
1353 | struct req_iterator iter; | |
7988613b | 1354 | struct bio_vec bvec; |
2d4dc890 IL |
1355 | |
1356 | rq_for_each_segment(bvec, rq, iter) | |
7988613b | 1357 | flush_dcache_page(bvec.bv_page); |
2d4dc890 IL |
1358 | } |
1359 | EXPORT_SYMBOL_GPL(rq_flush_dcache_pages); | |
1360 | #endif | |
1361 | ||
ef9e3fac KU |
1362 | /** |
1363 | * blk_lld_busy - Check if underlying low-level drivers of a device are busy | |
1364 | * @q : the queue of the device being checked | |
1365 | * | |
1366 | * Description: | |
1367 | * Check if underlying low-level drivers of a device are busy. | |
1368 | * If the drivers want to export their busy state, they must set own | |
1369 | * exporting function using blk_queue_lld_busy() first. | |
1370 | * | |
1371 | * Basically, this function is used only by request stacking drivers | |
1372 | * to stop dispatching requests to underlying devices when underlying | |
1373 | * devices are busy. This behavior helps more I/O merging on the queue | |
1374 | * of the request stacking driver and prevents I/O throughput regression | |
1375 | * on burst I/O load. | |
1376 | * | |
1377 | * Return: | |
1378 | * 0 - Not busy (The request stacking driver should dispatch request) | |
1379 | * 1 - Busy (The request stacking driver should stop dispatching request) | |
1380 | */ | |
1381 | int blk_lld_busy(struct request_queue *q) | |
1382 | { | |
344e9ffc | 1383 | if (queue_is_mq(q) && q->mq_ops->busy) |
9ba20527 | 1384 | return q->mq_ops->busy(q); |
ef9e3fac KU |
1385 | |
1386 | return 0; | |
1387 | } | |
1388 | EXPORT_SYMBOL_GPL(blk_lld_busy); | |
1389 | ||
78d8e58a MS |
1390 | /** |
1391 | * blk_rq_unprep_clone - Helper function to free all bios in a cloned request | |
1392 | * @rq: the clone request to be cleaned up | |
1393 | * | |
1394 | * Description: | |
1395 | * Free all bios in @rq for a cloned request. | |
1396 | */ | |
1397 | void blk_rq_unprep_clone(struct request *rq) | |
1398 | { | |
1399 | struct bio *bio; | |
1400 | ||
1401 | while ((bio = rq->bio) != NULL) { | |
1402 | rq->bio = bio->bi_next; | |
1403 | ||
1404 | bio_put(bio); | |
1405 | } | |
1406 | } | |
1407 | EXPORT_SYMBOL_GPL(blk_rq_unprep_clone); | |
1408 | ||
78d8e58a MS |
1409 | /** |
1410 | * blk_rq_prep_clone - Helper function to setup clone request | |
1411 | * @rq: the request to be setup | |
1412 | * @rq_src: original request to be cloned | |
1413 | * @bs: bio_set that bios for clone are allocated from | |
1414 | * @gfp_mask: memory allocation mask for bio | |
1415 | * @bio_ctr: setup function to be called for each clone bio. | |
1416 | * Returns %0 for success, non %0 for failure. | |
1417 | * @data: private data to be passed to @bio_ctr | |
1418 | * | |
1419 | * Description: | |
1420 | * Clones bios in @rq_src to @rq, and copies attributes of @rq_src to @rq. | |
78d8e58a MS |
1421 | * Also, pages which the original bios are pointing to are not copied |
1422 | * and the cloned bios just point same pages. | |
1423 | * So cloned bios must be completed before original bios, which means | |
1424 | * the caller must complete @rq before @rq_src. | |
1425 | */ | |
1426 | int blk_rq_prep_clone(struct request *rq, struct request *rq_src, | |
1427 | struct bio_set *bs, gfp_t gfp_mask, | |
1428 | int (*bio_ctr)(struct bio *, struct bio *, void *), | |
1429 | void *data) | |
1430 | { | |
1431 | struct bio *bio, *bio_src; | |
1432 | ||
1433 | if (!bs) | |
f4f8154a | 1434 | bs = &fs_bio_set; |
78d8e58a MS |
1435 | |
1436 | __rq_for_each_bio(bio_src, rq_src) { | |
1437 | bio = bio_clone_fast(bio_src, gfp_mask, bs); | |
1438 | if (!bio) | |
1439 | goto free_and_out; | |
1440 | ||
1441 | if (bio_ctr && bio_ctr(bio, bio_src, data)) | |
1442 | goto free_and_out; | |
1443 | ||
1444 | if (rq->bio) { | |
1445 | rq->biotail->bi_next = bio; | |
1446 | rq->biotail = bio; | |
93f221ae | 1447 | } else { |
78d8e58a | 1448 | rq->bio = rq->biotail = bio; |
93f221ae EB |
1449 | } |
1450 | bio = NULL; | |
78d8e58a MS |
1451 | } |
1452 | ||
361301a2 GJ |
1453 | /* Copy attributes of the original request to the clone request. */ |
1454 | rq->__sector = blk_rq_pos(rq_src); | |
1455 | rq->__data_len = blk_rq_bytes(rq_src); | |
1456 | if (rq_src->rq_flags & RQF_SPECIAL_PAYLOAD) { | |
1457 | rq->rq_flags |= RQF_SPECIAL_PAYLOAD; | |
1458 | rq->special_vec = rq_src->special_vec; | |
1459 | } | |
1460 | rq->nr_phys_segments = rq_src->nr_phys_segments; | |
1461 | rq->ioprio = rq_src->ioprio; | |
78d8e58a | 1462 | |
93f221ae EB |
1463 | if (rq->bio && blk_crypto_rq_bio_prep(rq, rq->bio, gfp_mask) < 0) |
1464 | goto free_and_out; | |
78d8e58a MS |
1465 | |
1466 | return 0; | |
1467 | ||
1468 | free_and_out: | |
1469 | if (bio) | |
1470 | bio_put(bio); | |
1471 | blk_rq_unprep_clone(rq); | |
1472 | ||
1473 | return -ENOMEM; | |
b0fd271d KU |
1474 | } |
1475 | EXPORT_SYMBOL_GPL(blk_rq_prep_clone); | |
1476 | ||
59c3d45e | 1477 | int kblockd_schedule_work(struct work_struct *work) |
1da177e4 LT |
1478 | { |
1479 | return queue_work(kblockd_workqueue, work); | |
1480 | } | |
1da177e4 LT |
1481 | EXPORT_SYMBOL(kblockd_schedule_work); |
1482 | ||
818cd1cb JA |
1483 | int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, |
1484 | unsigned long delay) | |
1485 | { | |
1486 | return mod_delayed_work_on(cpu, kblockd_workqueue, dwork, delay); | |
1487 | } | |
1488 | EXPORT_SYMBOL(kblockd_mod_delayed_work_on); | |
1489 | ||
47c122e3 JA |
1490 | void blk_start_plug_nr_ios(struct blk_plug *plug, unsigned short nr_ios) |
1491 | { | |
1492 | struct task_struct *tsk = current; | |
1493 | ||
1494 | /* | |
1495 | * If this is a nested plug, don't actually assign it. | |
1496 | */ | |
1497 | if (tsk->plug) | |
1498 | return; | |
1499 | ||
bc490f81 | 1500 | plug->mq_list = NULL; |
47c122e3 JA |
1501 | plug->cached_rq = NULL; |
1502 | plug->nr_ios = min_t(unsigned short, nr_ios, BLK_MAX_REQUEST_COUNT); | |
1503 | plug->rq_count = 0; | |
1504 | plug->multiple_queues = false; | |
dc5fc361 | 1505 | plug->has_elevator = false; |
47c122e3 JA |
1506 | plug->nowait = false; |
1507 | INIT_LIST_HEAD(&plug->cb_list); | |
1508 | ||
1509 | /* | |
1510 | * Store ordering should not be needed here, since a potential | |
1511 | * preempt will imply a full memory barrier | |
1512 | */ | |
1513 | tsk->plug = plug; | |
1514 | } | |
1515 | ||
75df7136 SJ |
1516 | /** |
1517 | * blk_start_plug - initialize blk_plug and track it inside the task_struct | |
1518 | * @plug: The &struct blk_plug that needs to be initialized | |
1519 | * | |
1520 | * Description: | |
40405851 JM |
1521 | * blk_start_plug() indicates to the block layer an intent by the caller |
1522 | * to submit multiple I/O requests in a batch. The block layer may use | |
1523 | * this hint to defer submitting I/Os from the caller until blk_finish_plug() | |
1524 | * is called. However, the block layer may choose to submit requests | |
1525 | * before a call to blk_finish_plug() if the number of queued I/Os | |
1526 | * exceeds %BLK_MAX_REQUEST_COUNT, or if the size of the I/O is larger than | |
1527 | * %BLK_PLUG_FLUSH_SIZE. The queued I/Os may also be submitted early if | |
1528 | * the task schedules (see below). | |
1529 | * | |
75df7136 SJ |
1530 | * Tracking blk_plug inside the task_struct will help with auto-flushing the |
1531 | * pending I/O should the task end up blocking between blk_start_plug() and | |
1532 | * blk_finish_plug(). This is important from a performance perspective, but | |
1533 | * also ensures that we don't deadlock. For instance, if the task is blocking | |
1534 | * for a memory allocation, memory reclaim could end up wanting to free a | |
1535 | * page belonging to that request that is currently residing in our private | |
1536 | * plug. By flushing the pending I/O when the process goes to sleep, we avoid | |
1537 | * this kind of deadlock. | |
1538 | */ | |
73c10101 JA |
1539 | void blk_start_plug(struct blk_plug *plug) |
1540 | { | |
47c122e3 | 1541 | blk_start_plug_nr_ios(plug, 1); |
73c10101 JA |
1542 | } |
1543 | EXPORT_SYMBOL(blk_start_plug); | |
1544 | ||
74018dc3 | 1545 | static void flush_plug_callbacks(struct blk_plug *plug, bool from_schedule) |
048c9374 N |
1546 | { |
1547 | LIST_HEAD(callbacks); | |
1548 | ||
2a7d5559 SL |
1549 | while (!list_empty(&plug->cb_list)) { |
1550 | list_splice_init(&plug->cb_list, &callbacks); | |
048c9374 | 1551 | |
2a7d5559 SL |
1552 | while (!list_empty(&callbacks)) { |
1553 | struct blk_plug_cb *cb = list_first_entry(&callbacks, | |
048c9374 N |
1554 | struct blk_plug_cb, |
1555 | list); | |
2a7d5559 | 1556 | list_del(&cb->list); |
74018dc3 | 1557 | cb->callback(cb, from_schedule); |
2a7d5559 | 1558 | } |
048c9374 N |
1559 | } |
1560 | } | |
1561 | ||
9cbb1750 N |
1562 | struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug, void *data, |
1563 | int size) | |
1564 | { | |
1565 | struct blk_plug *plug = current->plug; | |
1566 | struct blk_plug_cb *cb; | |
1567 | ||
1568 | if (!plug) | |
1569 | return NULL; | |
1570 | ||
1571 | list_for_each_entry(cb, &plug->cb_list, list) | |
1572 | if (cb->callback == unplug && cb->data == data) | |
1573 | return cb; | |
1574 | ||
1575 | /* Not currently on the callback list */ | |
1576 | BUG_ON(size < sizeof(*cb)); | |
1577 | cb = kzalloc(size, GFP_ATOMIC); | |
1578 | if (cb) { | |
1579 | cb->data = data; | |
1580 | cb->callback = unplug; | |
1581 | list_add(&cb->list, &plug->cb_list); | |
1582 | } | |
1583 | return cb; | |
1584 | } | |
1585 | EXPORT_SYMBOL(blk_check_plugged); | |
1586 | ||
008f75a2 | 1587 | void blk_flush_plug(struct blk_plug *plug, bool from_schedule) |
73c10101 | 1588 | { |
b600455d PB |
1589 | if (!list_empty(&plug->cb_list)) |
1590 | flush_plug_callbacks(plug, from_schedule); | |
bc490f81 | 1591 | if (!rq_list_empty(plug->mq_list)) |
320ae51f | 1592 | blk_mq_flush_plug_list(plug, from_schedule); |
c5fc7b93 JA |
1593 | /* |
1594 | * Unconditionally flush out cached requests, even if the unplug | |
1595 | * event came from schedule. Since we know hold references to the | |
1596 | * queue for cached requests, we don't want a blocked task holding | |
1597 | * up a queue freeze/quiesce event. | |
1598 | */ | |
1599 | if (unlikely(!rq_list_empty(plug->cached_rq))) | |
47c122e3 | 1600 | blk_mq_free_plug_rqs(plug); |
73c10101 | 1601 | } |
73c10101 | 1602 | |
40405851 JM |
1603 | /** |
1604 | * blk_finish_plug - mark the end of a batch of submitted I/O | |
1605 | * @plug: The &struct blk_plug passed to blk_start_plug() | |
1606 | * | |
1607 | * Description: | |
1608 | * Indicate that a batch of I/O submissions is complete. This function | |
1609 | * must be paired with an initial call to blk_start_plug(). The intent | |
1610 | * is to allow the block layer to optimize I/O submission. See the | |
1611 | * documentation for blk_start_plug() for more information. | |
1612 | */ | |
73c10101 JA |
1613 | void blk_finish_plug(struct blk_plug *plug) |
1614 | { | |
008f75a2 CH |
1615 | if (plug == current->plug) { |
1616 | blk_flush_plug(plug, false); | |
1617 | current->plug = NULL; | |
1618 | } | |
73c10101 | 1619 | } |
88b996cd | 1620 | EXPORT_SYMBOL(blk_finish_plug); |
73c10101 | 1621 | |
71ac860a ML |
1622 | void blk_io_schedule(void) |
1623 | { | |
1624 | /* Prevent hang_check timer from firing at us during very long I/O */ | |
1625 | unsigned long timeout = sysctl_hung_task_timeout_secs * HZ / 2; | |
1626 | ||
1627 | if (timeout) | |
1628 | io_schedule_timeout(timeout); | |
1629 | else | |
1630 | io_schedule(); | |
1631 | } | |
1632 | EXPORT_SYMBOL_GPL(blk_io_schedule); | |
1633 | ||
1da177e4 LT |
1634 | int __init blk_dev_init(void) |
1635 | { | |
ef295ecf CH |
1636 | BUILD_BUG_ON(REQ_OP_LAST >= (1 << REQ_OP_BITS)); |
1637 | BUILD_BUG_ON(REQ_OP_BITS + REQ_FLAG_BITS > 8 * | |
c593642c | 1638 | sizeof_field(struct request, cmd_flags)); |
ef295ecf | 1639 | BUILD_BUG_ON(REQ_OP_BITS + REQ_FLAG_BITS > 8 * |
c593642c | 1640 | sizeof_field(struct bio, bi_opf)); |
9eb55b03 | 1641 | |
89b90be2 TH |
1642 | /* used for unplugging and affects IO latency/throughput - HIGHPRI */ |
1643 | kblockd_workqueue = alloc_workqueue("kblockd", | |
28747fcd | 1644 | WQ_MEM_RECLAIM | WQ_HIGHPRI, 0); |
1da177e4 LT |
1645 | if (!kblockd_workqueue) |
1646 | panic("Failed to create kblockd\n"); | |
1647 | ||
c2789bd4 | 1648 | blk_requestq_cachep = kmem_cache_create("request_queue", |
165125e1 | 1649 | sizeof(struct request_queue), 0, SLAB_PANIC, NULL); |
1da177e4 | 1650 | |
18fbda91 | 1651 | blk_debugfs_root = debugfs_create_dir("block", NULL); |
18fbda91 | 1652 | |
d38ecf93 | 1653 | return 0; |
1da177e4 | 1654 | } |