]>
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> | |
17 | #include <linux/backing-dev.h> | |
18 | #include <linux/bio.h> | |
19 | #include <linux/blkdev.h> | |
320ae51f | 20 | #include <linux/blk-mq.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> |
eea8f41c | 37 | #include <linux/blk-cgroup.h> |
54d4e6ab | 38 | #include <linux/t10-pi.h> |
18fbda91 | 39 | #include <linux/debugfs.h> |
30abb3a6 | 40 | #include <linux/bpf.h> |
b8e24a93 | 41 | #include <linux/psi.h> |
71ac860a | 42 | #include <linux/sched/sysctl.h> |
a892c8d5 | 43 | #include <linux/blk-crypto.h> |
55782138 LZ |
44 | |
45 | #define CREATE_TRACE_POINTS | |
46 | #include <trace/events/block.h> | |
1da177e4 | 47 | |
8324aa91 | 48 | #include "blk.h" |
43a5e4e2 | 49 | #include "blk-mq.h" |
bd166ef1 | 50 | #include "blk-mq-sched.h" |
bca6b067 | 51 | #include "blk-pm.h" |
c1c80384 | 52 | #include "blk-rq-qos.h" |
8324aa91 | 53 | |
18fbda91 OS |
54 | #ifdef CONFIG_DEBUG_FS |
55 | struct dentry *blk_debugfs_root; | |
56 | #endif | |
57 | ||
d07335e5 | 58 | EXPORT_TRACEPOINT_SYMBOL_GPL(block_bio_remap); |
b0da3f0d | 59 | EXPORT_TRACEPOINT_SYMBOL_GPL(block_rq_remap); |
0a82a8d1 | 60 | EXPORT_TRACEPOINT_SYMBOL_GPL(block_bio_complete); |
3291fa57 | 61 | EXPORT_TRACEPOINT_SYMBOL_GPL(block_split); |
cbae8d45 | 62 | EXPORT_TRACEPOINT_SYMBOL_GPL(block_unplug); |
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); | |
63a71386 | 121 | rq->tag = -1; |
bd166ef1 | 122 | rq->internal_tag = -1; |
522a7775 | 123 | rq->start_time_ns = ktime_get_ns(); |
09e099d4 | 124 | rq->part = NULL; |
b554db14 | 125 | refcount_set(&rq->ref, 1); |
a892c8d5 | 126 | blk_crypto_rq_set_defaults(rq); |
1da177e4 | 127 | } |
2a4aa30c | 128 | EXPORT_SYMBOL(blk_rq_init); |
1da177e4 | 129 | |
e47bc4ed CK |
130 | #define REQ_OP_NAME(name) [REQ_OP_##name] = #name |
131 | static const char *const blk_op_name[] = { | |
132 | REQ_OP_NAME(READ), | |
133 | REQ_OP_NAME(WRITE), | |
134 | REQ_OP_NAME(FLUSH), | |
135 | REQ_OP_NAME(DISCARD), | |
136 | REQ_OP_NAME(SECURE_ERASE), | |
137 | REQ_OP_NAME(ZONE_RESET), | |
6e33dbf2 | 138 | REQ_OP_NAME(ZONE_RESET_ALL), |
6c1b1da5 AJ |
139 | REQ_OP_NAME(ZONE_OPEN), |
140 | REQ_OP_NAME(ZONE_CLOSE), | |
141 | REQ_OP_NAME(ZONE_FINISH), | |
0512a75b | 142 | REQ_OP_NAME(ZONE_APPEND), |
e47bc4ed CK |
143 | REQ_OP_NAME(WRITE_SAME), |
144 | REQ_OP_NAME(WRITE_ZEROES), | |
145 | REQ_OP_NAME(SCSI_IN), | |
146 | REQ_OP_NAME(SCSI_OUT), | |
147 | REQ_OP_NAME(DRV_IN), | |
148 | REQ_OP_NAME(DRV_OUT), | |
149 | }; | |
150 | #undef REQ_OP_NAME | |
151 | ||
152 | /** | |
153 | * blk_op_str - Return string XXX in the REQ_OP_XXX. | |
154 | * @op: REQ_OP_XXX. | |
155 | * | |
156 | * Description: Centralize block layer function to convert REQ_OP_XXX into | |
157 | * string format. Useful in the debugging and tracing bio or request. For | |
158 | * invalid REQ_OP_XXX it returns string "UNKNOWN". | |
159 | */ | |
160 | inline const char *blk_op_str(unsigned int op) | |
161 | { | |
162 | const char *op_str = "UNKNOWN"; | |
163 | ||
164 | if (op < ARRAY_SIZE(blk_op_name) && blk_op_name[op]) | |
165 | op_str = blk_op_name[op]; | |
166 | ||
167 | return op_str; | |
168 | } | |
169 | EXPORT_SYMBOL_GPL(blk_op_str); | |
170 | ||
2a842aca CH |
171 | static const struct { |
172 | int errno; | |
173 | const char *name; | |
174 | } blk_errors[] = { | |
175 | [BLK_STS_OK] = { 0, "" }, | |
176 | [BLK_STS_NOTSUPP] = { -EOPNOTSUPP, "operation not supported" }, | |
177 | [BLK_STS_TIMEOUT] = { -ETIMEDOUT, "timeout" }, | |
178 | [BLK_STS_NOSPC] = { -ENOSPC, "critical space allocation" }, | |
179 | [BLK_STS_TRANSPORT] = { -ENOLINK, "recoverable transport" }, | |
180 | [BLK_STS_TARGET] = { -EREMOTEIO, "critical target" }, | |
181 | [BLK_STS_NEXUS] = { -EBADE, "critical nexus" }, | |
182 | [BLK_STS_MEDIUM] = { -ENODATA, "critical medium" }, | |
183 | [BLK_STS_PROTECTION] = { -EILSEQ, "protection" }, | |
184 | [BLK_STS_RESOURCE] = { -ENOMEM, "kernel resource" }, | |
86ff7c2a | 185 | [BLK_STS_DEV_RESOURCE] = { -EBUSY, "device resource" }, |
03a07c92 | 186 | [BLK_STS_AGAIN] = { -EAGAIN, "nonblocking retry" }, |
2a842aca | 187 | |
4e4cbee9 CH |
188 | /* device mapper special case, should not leak out: */ |
189 | [BLK_STS_DM_REQUEUE] = { -EREMCHG, "dm internal retry" }, | |
190 | ||
2a842aca CH |
191 | /* everything else not covered above: */ |
192 | [BLK_STS_IOERR] = { -EIO, "I/O" }, | |
193 | }; | |
194 | ||
195 | blk_status_t errno_to_blk_status(int errno) | |
196 | { | |
197 | int i; | |
198 | ||
199 | for (i = 0; i < ARRAY_SIZE(blk_errors); i++) { | |
200 | if (blk_errors[i].errno == errno) | |
201 | return (__force blk_status_t)i; | |
202 | } | |
203 | ||
204 | return BLK_STS_IOERR; | |
205 | } | |
206 | EXPORT_SYMBOL_GPL(errno_to_blk_status); | |
207 | ||
208 | int blk_status_to_errno(blk_status_t status) | |
209 | { | |
210 | int idx = (__force int)status; | |
211 | ||
34bd9c1c | 212 | if (WARN_ON_ONCE(idx >= ARRAY_SIZE(blk_errors))) |
2a842aca CH |
213 | return -EIO; |
214 | return blk_errors[idx].errno; | |
215 | } | |
216 | EXPORT_SYMBOL_GPL(blk_status_to_errno); | |
217 | ||
178cc590 CH |
218 | static void print_req_error(struct request *req, blk_status_t status, |
219 | const char *caller) | |
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 |
b0e5168a CK |
227 | "%s: %s error, dev %s, sector %llu op 0x%x:(%s) flags 0x%x " |
228 | "phys_seg %u prio class %u\n", | |
178cc590 | 229 | caller, 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 | ||
5bb23a68 | 237 | static void req_bio_endio(struct request *rq, struct bio *bio, |
2a842aca | 238 | unsigned int nbytes, blk_status_t error) |
1da177e4 | 239 | { |
78d8e58a | 240 | if (error) |
4e4cbee9 | 241 | bio->bi_status = error; |
797e7dbb | 242 | |
e8064021 | 243 | if (unlikely(rq->rq_flags & RQF_QUIET)) |
b7c44ed9 | 244 | bio_set_flag(bio, BIO_QUIET); |
08bafc03 | 245 | |
f79ea416 | 246 | bio_advance(bio, nbytes); |
7ba1ba12 | 247 | |
0512a75b KB |
248 | if (req_op(rq) == REQ_OP_ZONE_APPEND && error == BLK_STS_OK) { |
249 | /* | |
250 | * Partial zone append completions cannot be supported as the | |
251 | * BIO fragments may end up not being written sequentially. | |
252 | */ | |
253 | if (bio->bi_iter.bi_size) | |
254 | bio->bi_status = BLK_STS_IOERR; | |
255 | else | |
256 | bio->bi_iter.bi_sector = rq->__sector; | |
257 | } | |
258 | ||
143a87f4 | 259 | /* don't actually finish bio if it's part of flush sequence */ |
e8064021 | 260 | if (bio->bi_iter.bi_size == 0 && !(rq->rq_flags & RQF_FLUSH_SEQ)) |
4246a0b6 | 261 | bio_endio(bio); |
1da177e4 | 262 | } |
1da177e4 | 263 | |
1da177e4 LT |
264 | void blk_dump_rq_flags(struct request *rq, char *msg) |
265 | { | |
aebf526b CH |
266 | printk(KERN_INFO "%s: dev %s: flags=%llx\n", msg, |
267 | rq->rq_disk ? rq->rq_disk->disk_name : "?", | |
5953316d | 268 | (unsigned long long) rq->cmd_flags); |
1da177e4 | 269 | |
83096ebf TH |
270 | printk(KERN_INFO " sector %llu, nr/cnr %u/%u\n", |
271 | (unsigned long long)blk_rq_pos(rq), | |
272 | blk_rq_sectors(rq), blk_rq_cur_sectors(rq)); | |
b4f42e28 JA |
273 | printk(KERN_INFO " bio %p, biotail %p, len %u\n", |
274 | rq->bio, rq->biotail, blk_rq_bytes(rq)); | |
1da177e4 | 275 | } |
1da177e4 LT |
276 | EXPORT_SYMBOL(blk_dump_rq_flags); |
277 | ||
1da177e4 LT |
278 | /** |
279 | * blk_sync_queue - cancel any pending callbacks on a queue | |
280 | * @q: the queue | |
281 | * | |
282 | * Description: | |
283 | * The block layer may perform asynchronous callback activity | |
284 | * on a queue, such as calling the unplug function after a timeout. | |
285 | * A block device may call blk_sync_queue to ensure that any | |
286 | * such activity is cancelled, thus allowing it to release resources | |
59c51591 | 287 | * that the callbacks might use. The caller must already have made sure |
1da177e4 LT |
288 | * that its ->make_request_fn will not re-add plugging prior to calling |
289 | * this function. | |
290 | * | |
da527770 | 291 | * This function does not cancel any asynchronous activity arising |
da3dae54 | 292 | * out of elevator or throttling code. That would require elevator_exit() |
5efd6113 | 293 | * and blkcg_exit_queue() to be called with queue lock initialized. |
da527770 | 294 | * |
1da177e4 LT |
295 | */ |
296 | void blk_sync_queue(struct request_queue *q) | |
297 | { | |
70ed28b9 | 298 | del_timer_sync(&q->timeout); |
4e9b6f20 | 299 | cancel_work_sync(&q->timeout_work); |
1da177e4 LT |
300 | } |
301 | EXPORT_SYMBOL(blk_sync_queue); | |
302 | ||
c9254f2d | 303 | /** |
cd84a62e | 304 | * blk_set_pm_only - increment pm_only counter |
c9254f2d | 305 | * @q: request queue pointer |
c9254f2d | 306 | */ |
cd84a62e | 307 | void blk_set_pm_only(struct request_queue *q) |
c9254f2d | 308 | { |
cd84a62e | 309 | atomic_inc(&q->pm_only); |
c9254f2d | 310 | } |
cd84a62e | 311 | EXPORT_SYMBOL_GPL(blk_set_pm_only); |
c9254f2d | 312 | |
cd84a62e | 313 | void blk_clear_pm_only(struct request_queue *q) |
c9254f2d | 314 | { |
cd84a62e BVA |
315 | int pm_only; |
316 | ||
317 | pm_only = atomic_dec_return(&q->pm_only); | |
318 | WARN_ON_ONCE(pm_only < 0); | |
319 | if (pm_only == 0) | |
320 | wake_up_all(&q->mq_freeze_wq); | |
c9254f2d | 321 | } |
cd84a62e | 322 | EXPORT_SYMBOL_GPL(blk_clear_pm_only); |
c9254f2d | 323 | |
b5bd357c LC |
324 | /** |
325 | * blk_put_queue - decrement the request_queue refcount | |
326 | * @q: the request_queue structure to decrement the refcount for | |
327 | * | |
328 | * Decrements the refcount of the request_queue kobject. When this reaches 0 | |
329 | * we'll have blk_release_queue() called. | |
330 | */ | |
165125e1 | 331 | void blk_put_queue(struct request_queue *q) |
483f4afc AV |
332 | { |
333 | kobject_put(&q->kobj); | |
334 | } | |
d86e0e83 | 335 | EXPORT_SYMBOL(blk_put_queue); |
483f4afc | 336 | |
aed3ea94 JA |
337 | void blk_set_queue_dying(struct request_queue *q) |
338 | { | |
8814ce8a | 339 | blk_queue_flag_set(QUEUE_FLAG_DYING, q); |
aed3ea94 | 340 | |
d3cfb2a0 ML |
341 | /* |
342 | * When queue DYING flag is set, we need to block new req | |
343 | * entering queue, so we call blk_freeze_queue_start() to | |
344 | * prevent I/O from crossing blk_queue_enter(). | |
345 | */ | |
346 | blk_freeze_queue_start(q); | |
347 | ||
344e9ffc | 348 | if (queue_is_mq(q)) |
aed3ea94 | 349 | blk_mq_wake_waiters(q); |
055f6e18 ML |
350 | |
351 | /* Make blk_queue_enter() reexamine the DYING flag. */ | |
352 | wake_up_all(&q->mq_freeze_wq); | |
aed3ea94 JA |
353 | } |
354 | EXPORT_SYMBOL_GPL(blk_set_queue_dying); | |
355 | ||
c9a929dd TH |
356 | /** |
357 | * blk_cleanup_queue - shutdown a request queue | |
358 | * @q: request queue to shutdown | |
359 | * | |
c246e80d BVA |
360 | * Mark @q DYING, drain all pending requests, mark @q DEAD, destroy and |
361 | * put it. All future requests will be failed immediately with -ENODEV. | |
c94a96ac | 362 | */ |
6728cb0e | 363 | void blk_cleanup_queue(struct request_queue *q) |
483f4afc | 364 | { |
bae85c15 BVA |
365 | WARN_ON_ONCE(blk_queue_registered(q)); |
366 | ||
3f3299d5 | 367 | /* mark @q DYING, no new request or merges will be allowed afterwards */ |
aed3ea94 | 368 | blk_set_queue_dying(q); |
6ecf23af | 369 | |
57d74df9 CH |
370 | blk_queue_flag_set(QUEUE_FLAG_NOMERGES, q); |
371 | blk_queue_flag_set(QUEUE_FLAG_NOXMERGES, q); | |
c9a929dd | 372 | |
c246e80d BVA |
373 | /* |
374 | * Drain all requests queued before DYING marking. Set DEAD flag to | |
67ed8b73 BVA |
375 | * prevent that blk_mq_run_hw_queues() accesses the hardware queues |
376 | * after draining finished. | |
c246e80d | 377 | */ |
3ef28e83 | 378 | blk_freeze_queue(q); |
c57cdf7a ML |
379 | |
380 | rq_qos_exit(q); | |
381 | ||
57d74df9 | 382 | blk_queue_flag_set(QUEUE_FLAG_DEAD, q); |
c9a929dd | 383 | |
5a48fc14 DW |
384 | /* for synchronous bio-based driver finish in-flight integrity i/o */ |
385 | blk_flush_integrity(); | |
386 | ||
c9a929dd | 387 | /* @q won't process any more request, flush async actions */ |
dc3b17cc | 388 | del_timer_sync(&q->backing_dev_info->laptop_mode_wb_timer); |
c9a929dd TH |
389 | blk_sync_queue(q); |
390 | ||
344e9ffc | 391 | if (queue_is_mq(q)) |
c7e2d94b | 392 | blk_mq_exit_queue(q); |
a1ce35fa | 393 | |
c3e22192 ML |
394 | /* |
395 | * In theory, request pool of sched_tags belongs to request queue. | |
396 | * However, the current implementation requires tag_set for freeing | |
397 | * requests, so free the pool now. | |
398 | * | |
399 | * Queue has become frozen, there can't be any in-queue requests, so | |
400 | * it is safe to free requests now. | |
401 | */ | |
402 | mutex_lock(&q->sysfs_lock); | |
403 | if (q->elevator) | |
404 | blk_mq_sched_free_requests(q); | |
405 | mutex_unlock(&q->sysfs_lock); | |
406 | ||
3ef28e83 | 407 | percpu_ref_exit(&q->q_usage_counter); |
45a9c9d9 | 408 | |
c9a929dd | 409 | /* @q is and will stay empty, shutdown and put */ |
483f4afc AV |
410 | blk_put_queue(q); |
411 | } | |
1da177e4 LT |
412 | EXPORT_SYMBOL(blk_cleanup_queue); |
413 | ||
3a0a5299 BVA |
414 | /** |
415 | * blk_queue_enter() - try to increase q->q_usage_counter | |
416 | * @q: request queue pointer | |
417 | * @flags: BLK_MQ_REQ_NOWAIT and/or BLK_MQ_REQ_PREEMPT | |
418 | */ | |
9a95e4ef | 419 | int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags) |
3ef28e83 | 420 | { |
cd84a62e | 421 | const bool pm = flags & BLK_MQ_REQ_PREEMPT; |
3a0a5299 | 422 | |
3ef28e83 | 423 | while (true) { |
3a0a5299 | 424 | bool success = false; |
3ef28e83 | 425 | |
818e0fa2 | 426 | rcu_read_lock(); |
3a0a5299 BVA |
427 | if (percpu_ref_tryget_live(&q->q_usage_counter)) { |
428 | /* | |
cd84a62e BVA |
429 | * The code that increments the pm_only counter is |
430 | * responsible for ensuring that that counter is | |
431 | * globally visible before the queue is unfrozen. | |
3a0a5299 | 432 | */ |
cd84a62e | 433 | if (pm || !blk_queue_pm_only(q)) { |
3a0a5299 BVA |
434 | success = true; |
435 | } else { | |
436 | percpu_ref_put(&q->q_usage_counter); | |
437 | } | |
438 | } | |
818e0fa2 | 439 | rcu_read_unlock(); |
3a0a5299 BVA |
440 | |
441 | if (success) | |
3ef28e83 DW |
442 | return 0; |
443 | ||
3a0a5299 | 444 | if (flags & BLK_MQ_REQ_NOWAIT) |
3ef28e83 DW |
445 | return -EBUSY; |
446 | ||
5ed61d3f | 447 | /* |
1671d522 | 448 | * read pair of barrier in blk_freeze_queue_start(), |
5ed61d3f | 449 | * we need to order reading __PERCPU_REF_DEAD flag of |
d3cfb2a0 ML |
450 | * .q_usage_counter and reading .mq_freeze_depth or |
451 | * queue dying flag, otherwise the following wait may | |
452 | * never return if the two reads are reordered. | |
5ed61d3f ML |
453 | */ |
454 | smp_rmb(); | |
455 | ||
1dc3039b | 456 | wait_event(q->mq_freeze_wq, |
7996a8b5 | 457 | (!q->mq_freeze_depth && |
0d25bd07 BVA |
458 | (pm || (blk_pm_request_resume(q), |
459 | !blk_queue_pm_only(q)))) || | |
1dc3039b | 460 | blk_queue_dying(q)); |
3ef28e83 DW |
461 | if (blk_queue_dying(q)) |
462 | return -ENODEV; | |
3ef28e83 DW |
463 | } |
464 | } | |
465 | ||
accea322 CH |
466 | static inline int bio_queue_enter(struct bio *bio) |
467 | { | |
468 | struct request_queue *q = bio->bi_disk->queue; | |
469 | bool nowait = bio->bi_opf & REQ_NOWAIT; | |
470 | int ret; | |
471 | ||
472 | ret = blk_queue_enter(q, nowait ? BLK_MQ_REQ_NOWAIT : 0); | |
473 | if (unlikely(ret)) { | |
474 | if (nowait && !blk_queue_dying(q)) | |
475 | bio_wouldblock_error(bio); | |
476 | else | |
477 | bio_io_error(bio); | |
478 | } | |
479 | ||
480 | return ret; | |
481 | } | |
482 | ||
3ef28e83 DW |
483 | void blk_queue_exit(struct request_queue *q) |
484 | { | |
485 | percpu_ref_put(&q->q_usage_counter); | |
486 | } | |
487 | ||
488 | static void blk_queue_usage_counter_release(struct percpu_ref *ref) | |
489 | { | |
490 | struct request_queue *q = | |
491 | container_of(ref, struct request_queue, q_usage_counter); | |
492 | ||
493 | wake_up_all(&q->mq_freeze_wq); | |
494 | } | |
495 | ||
bca237a5 | 496 | static void blk_rq_timed_out_timer(struct timer_list *t) |
287922eb | 497 | { |
bca237a5 | 498 | struct request_queue *q = from_timer(q, t, timeout); |
287922eb CH |
499 | |
500 | kblockd_schedule_work(&q->timeout_work); | |
501 | } | |
502 | ||
2e3c18d0 TH |
503 | static void blk_timeout_work(struct work_struct *work) |
504 | { | |
505 | } | |
506 | ||
3d745ea5 | 507 | struct request_queue *__blk_alloc_queue(int node_id) |
1946089a | 508 | { |
165125e1 | 509 | struct request_queue *q; |
338aa96d | 510 | int ret; |
1946089a | 511 | |
8324aa91 | 512 | q = kmem_cache_alloc_node(blk_requestq_cachep, |
3d745ea5 | 513 | GFP_KERNEL | __GFP_ZERO, node_id); |
1da177e4 LT |
514 | if (!q) |
515 | return NULL; | |
516 | ||
cbf62af3 | 517 | q->last_merge = NULL; |
cbf62af3 | 518 | |
3d745ea5 | 519 | q->id = ida_simple_get(&blk_queue_ida, 0, 0, GFP_KERNEL); |
a73f730d | 520 | if (q->id < 0) |
3d2936f4 | 521 | goto fail_q; |
a73f730d | 522 | |
338aa96d KO |
523 | ret = bioset_init(&q->bio_split, BIO_POOL_SIZE, 0, BIOSET_NEED_BVECS); |
524 | if (ret) | |
54efd50b KO |
525 | goto fail_id; |
526 | ||
aef33c2f | 527 | q->backing_dev_info = bdi_alloc(node_id); |
d03f6cdc JK |
528 | if (!q->backing_dev_info) |
529 | goto fail_split; | |
530 | ||
a83b576c JA |
531 | q->stats = blk_alloc_queue_stats(); |
532 | if (!q->stats) | |
533 | goto fail_stats; | |
534 | ||
b5420237 | 535 | q->backing_dev_info->ra_pages = VM_READAHEAD_PAGES; |
dc3b17cc | 536 | q->backing_dev_info->capabilities = BDI_CAP_CGROUP_WRITEBACK; |
5151412d | 537 | q->node = node_id; |
0989a025 | 538 | |
bca237a5 KC |
539 | timer_setup(&q->backing_dev_info->laptop_mode_wb_timer, |
540 | laptop_mode_timer_fn, 0); | |
541 | timer_setup(&q->timeout, blk_rq_timed_out_timer, 0); | |
2e3c18d0 | 542 | INIT_WORK(&q->timeout_work, blk_timeout_work); |
a612fddf | 543 | INIT_LIST_HEAD(&q->icq_list); |
4eef3049 | 544 | #ifdef CONFIG_BLK_CGROUP |
e8989fae | 545 | INIT_LIST_HEAD(&q->blkg_list); |
4eef3049 | 546 | #endif |
483f4afc | 547 | |
8324aa91 | 548 | kobject_init(&q->kobj, &blk_queue_ktype); |
1da177e4 | 549 | |
5acb3cc2 WL |
550 | #ifdef CONFIG_BLK_DEV_IO_TRACE |
551 | mutex_init(&q->blk_trace_mutex); | |
552 | #endif | |
483f4afc | 553 | mutex_init(&q->sysfs_lock); |
cecf5d87 | 554 | mutex_init(&q->sysfs_dir_lock); |
0d945c1f | 555 | spin_lock_init(&q->queue_lock); |
c94a96ac | 556 | |
320ae51f | 557 | init_waitqueue_head(&q->mq_freeze_wq); |
7996a8b5 | 558 | mutex_init(&q->mq_freeze_lock); |
320ae51f | 559 | |
3ef28e83 DW |
560 | /* |
561 | * Init percpu_ref in atomic mode so that it's faster to shutdown. | |
562 | * See blk_register_queue() for details. | |
563 | */ | |
564 | if (percpu_ref_init(&q->q_usage_counter, | |
565 | blk_queue_usage_counter_release, | |
566 | PERCPU_REF_INIT_ATOMIC, GFP_KERNEL)) | |
fff4996b | 567 | goto fail_bdi; |
f51b802c | 568 | |
3ef28e83 DW |
569 | if (blkcg_init_queue(q)) |
570 | goto fail_ref; | |
571 | ||
3d745ea5 CH |
572 | blk_queue_dma_alignment(q, 511); |
573 | blk_set_default_limits(&q->limits); | |
574 | ||
1da177e4 | 575 | return q; |
a73f730d | 576 | |
3ef28e83 DW |
577 | fail_ref: |
578 | percpu_ref_exit(&q->q_usage_counter); | |
fff4996b | 579 | fail_bdi: |
a83b576c JA |
580 | blk_free_queue_stats(q->stats); |
581 | fail_stats: | |
d03f6cdc | 582 | bdi_put(q->backing_dev_info); |
54efd50b | 583 | fail_split: |
338aa96d | 584 | bioset_exit(&q->bio_split); |
a73f730d TH |
585 | fail_id: |
586 | ida_simple_remove(&blk_queue_ida, q->id); | |
587 | fail_q: | |
588 | kmem_cache_free(blk_requestq_cachep, q); | |
589 | return NULL; | |
1da177e4 | 590 | } |
3d745ea5 CH |
591 | |
592 | struct request_queue *blk_alloc_queue(make_request_fn make_request, int node_id) | |
593 | { | |
594 | struct request_queue *q; | |
595 | ||
596 | if (WARN_ON_ONCE(!make_request)) | |
654a3667 | 597 | return NULL; |
3d745ea5 CH |
598 | |
599 | q = __blk_alloc_queue(node_id); | |
600 | if (!q) | |
601 | return NULL; | |
602 | q->make_request_fn = make_request; | |
603 | q->nr_requests = BLKDEV_MAX_RQ; | |
604 | return q; | |
605 | } | |
606 | EXPORT_SYMBOL(blk_alloc_queue); | |
1da177e4 | 607 | |
b5bd357c LC |
608 | /** |
609 | * blk_get_queue - increment the request_queue refcount | |
610 | * @q: the request_queue structure to increment the refcount for | |
611 | * | |
612 | * Increment the refcount of the request_queue kobject. | |
613 | */ | |
09ac46c4 | 614 | bool blk_get_queue(struct request_queue *q) |
1da177e4 | 615 | { |
3f3299d5 | 616 | if (likely(!blk_queue_dying(q))) { |
09ac46c4 TH |
617 | __blk_get_queue(q); |
618 | return true; | |
1da177e4 LT |
619 | } |
620 | ||
09ac46c4 | 621 | return false; |
1da177e4 | 622 | } |
d86e0e83 | 623 | EXPORT_SYMBOL(blk_get_queue); |
1da177e4 | 624 | |
a1ce35fa JA |
625 | /** |
626 | * blk_get_request - allocate a request | |
627 | * @q: request queue to allocate a request for | |
628 | * @op: operation (REQ_OP_*) and REQ_* flags, e.g. REQ_SYNC. | |
629 | * @flags: BLK_MQ_REQ_* flags, e.g. BLK_MQ_REQ_NOWAIT. | |
1da177e4 | 630 | */ |
a1ce35fa JA |
631 | struct request *blk_get_request(struct request_queue *q, unsigned int op, |
632 | blk_mq_req_flags_t flags) | |
1da177e4 | 633 | { |
a1ce35fa | 634 | struct request *req; |
1da177e4 | 635 | |
a1ce35fa JA |
636 | WARN_ON_ONCE(op & REQ_NOWAIT); |
637 | WARN_ON_ONCE(flags & ~(BLK_MQ_REQ_NOWAIT | BLK_MQ_REQ_PREEMPT)); | |
1da177e4 | 638 | |
a1ce35fa JA |
639 | req = blk_mq_alloc_request(q, op, flags); |
640 | if (!IS_ERR(req) && q->mq_ops->initialize_rq_fn) | |
641 | q->mq_ops->initialize_rq_fn(req); | |
1da177e4 | 642 | |
a1ce35fa | 643 | return req; |
1da177e4 | 644 | } |
a1ce35fa | 645 | EXPORT_SYMBOL(blk_get_request); |
1da177e4 | 646 | |
1da177e4 LT |
647 | void blk_put_request(struct request *req) |
648 | { | |
a1ce35fa | 649 | blk_mq_free_request(req); |
1da177e4 | 650 | } |
1da177e4 LT |
651 | EXPORT_SYMBOL(blk_put_request); |
652 | ||
b5af37ab KK |
653 | static void blk_account_io_merge_bio(struct request *req) |
654 | { | |
655 | if (!blk_do_io_stat(req)) | |
656 | return; | |
657 | ||
658 | part_stat_lock(); | |
659 | part_stat_inc(req->part, merges[op_stat_group(req_op(req))]); | |
660 | part_stat_unlock(); | |
661 | } | |
662 | ||
14ccb66b CH |
663 | bool bio_attempt_back_merge(struct request *req, struct bio *bio, |
664 | unsigned int nr_segs) | |
73c10101 | 665 | { |
1eff9d32 | 666 | const int ff = bio->bi_opf & REQ_FAILFAST_MASK; |
73c10101 | 667 | |
14ccb66b | 668 | if (!ll_back_merge_fn(req, bio, nr_segs)) |
73c10101 JA |
669 | return false; |
670 | ||
14ccb66b | 671 | trace_block_bio_backmerge(req->q, req, bio); |
d3e65fff | 672 | rq_qos_merge(req->q, req, bio); |
73c10101 JA |
673 | |
674 | if ((req->cmd_flags & REQ_FAILFAST_MASK) != ff) | |
675 | blk_rq_set_mixed_merge(req); | |
676 | ||
677 | req->biotail->bi_next = bio; | |
678 | req->biotail = bio; | |
4f024f37 | 679 | req->__data_len += bio->bi_iter.bi_size; |
73c10101 | 680 | |
a892c8d5 ST |
681 | bio_crypt_free_ctx(bio); |
682 | ||
b5af37ab | 683 | blk_account_io_merge_bio(req); |
73c10101 JA |
684 | return true; |
685 | } | |
686 | ||
14ccb66b CH |
687 | bool bio_attempt_front_merge(struct request *req, struct bio *bio, |
688 | unsigned int nr_segs) | |
73c10101 | 689 | { |
1eff9d32 | 690 | const int ff = bio->bi_opf & REQ_FAILFAST_MASK; |
73c10101 | 691 | |
14ccb66b | 692 | if (!ll_front_merge_fn(req, bio, nr_segs)) |
73c10101 JA |
693 | return false; |
694 | ||
14ccb66b | 695 | trace_block_bio_frontmerge(req->q, req, bio); |
d3e65fff | 696 | rq_qos_merge(req->q, req, bio); |
73c10101 JA |
697 | |
698 | if ((req->cmd_flags & REQ_FAILFAST_MASK) != ff) | |
699 | blk_rq_set_mixed_merge(req); | |
700 | ||
73c10101 JA |
701 | bio->bi_next = req->bio; |
702 | req->bio = bio; | |
703 | ||
4f024f37 KO |
704 | req->__sector = bio->bi_iter.bi_sector; |
705 | req->__data_len += bio->bi_iter.bi_size; | |
73c10101 | 706 | |
a892c8d5 ST |
707 | bio_crypt_do_front_merge(req, bio); |
708 | ||
b5af37ab | 709 | blk_account_io_merge_bio(req); |
73c10101 JA |
710 | return true; |
711 | } | |
712 | ||
1e739730 CH |
713 | bool bio_attempt_discard_merge(struct request_queue *q, struct request *req, |
714 | struct bio *bio) | |
715 | { | |
716 | unsigned short segments = blk_rq_nr_discard_segments(req); | |
717 | ||
718 | if (segments >= queue_max_discard_segments(q)) | |
719 | goto no_merge; | |
720 | if (blk_rq_sectors(req) + bio_sectors(bio) > | |
721 | blk_rq_get_max_sectors(req, blk_rq_pos(req))) | |
722 | goto no_merge; | |
723 | ||
d3e65fff TH |
724 | rq_qos_merge(q, req, bio); |
725 | ||
1e739730 CH |
726 | req->biotail->bi_next = bio; |
727 | req->biotail = bio; | |
728 | req->__data_len += bio->bi_iter.bi_size; | |
1e739730 CH |
729 | req->nr_phys_segments = segments + 1; |
730 | ||
b5af37ab | 731 | blk_account_io_merge_bio(req); |
1e739730 CH |
732 | return true; |
733 | no_merge: | |
734 | req_set_nomerge(q, req); | |
735 | return false; | |
736 | } | |
737 | ||
bd87b589 | 738 | /** |
320ae51f | 739 | * blk_attempt_plug_merge - try to merge with %current's plugged list |
bd87b589 TH |
740 | * @q: request_queue new bio is being queued at |
741 | * @bio: new bio being queued | |
14ccb66b | 742 | * @nr_segs: number of segments in @bio |
ccc2600b RD |
743 | * @same_queue_rq: pointer to &struct request that gets filled in when |
744 | * another request associated with @q is found on the plug list | |
745 | * (optional, may be %NULL) | |
bd87b589 TH |
746 | * |
747 | * Determine whether @bio being queued on @q can be merged with a request | |
748 | * on %current's plugged list. Returns %true if merge was successful, | |
749 | * otherwise %false. | |
750 | * | |
07c2bd37 TH |
751 | * Plugging coalesces IOs from the same issuer for the same purpose without |
752 | * going through @q->queue_lock. As such it's more of an issuing mechanism | |
753 | * than scheduling, and the request, while may have elvpriv data, is not | |
754 | * added on the elevator at this point. In addition, we don't have | |
755 | * reliable access to the elevator outside queue lock. Only check basic | |
756 | * merging parameters without querying the elevator. | |
da41a589 RE |
757 | * |
758 | * Caller must ensure !blk_queue_nomerges(q) beforehand. | |
73c10101 | 759 | */ |
320ae51f | 760 | bool blk_attempt_plug_merge(struct request_queue *q, struct bio *bio, |
14ccb66b | 761 | unsigned int nr_segs, struct request **same_queue_rq) |
73c10101 JA |
762 | { |
763 | struct blk_plug *plug; | |
764 | struct request *rq; | |
92f399c7 | 765 | struct list_head *plug_list; |
73c10101 | 766 | |
b49773e7 | 767 | plug = blk_mq_plug(q, bio); |
73c10101 | 768 | if (!plug) |
34fe7c05 | 769 | return false; |
73c10101 | 770 | |
a1ce35fa | 771 | plug_list = &plug->mq_list; |
92f399c7 SL |
772 | |
773 | list_for_each_entry_reverse(rq, plug_list, queuelist) { | |
34fe7c05 | 774 | bool merged = false; |
73c10101 | 775 | |
5f0ed774 | 776 | if (rq->q == q && same_queue_rq) { |
5b3f341f SL |
777 | /* |
778 | * Only blk-mq multiple hardware queues case checks the | |
779 | * rq in the same queue, there should be only one such | |
780 | * rq in a queue | |
781 | **/ | |
5f0ed774 | 782 | *same_queue_rq = rq; |
5b3f341f | 783 | } |
56ebdaf2 | 784 | |
07c2bd37 | 785 | if (rq->q != q || !blk_rq_merge_ok(rq, bio)) |
73c10101 JA |
786 | continue; |
787 | ||
34fe7c05 CH |
788 | switch (blk_try_merge(rq, bio)) { |
789 | case ELEVATOR_BACK_MERGE: | |
14ccb66b | 790 | merged = bio_attempt_back_merge(rq, bio, nr_segs); |
34fe7c05 CH |
791 | break; |
792 | case ELEVATOR_FRONT_MERGE: | |
14ccb66b | 793 | merged = bio_attempt_front_merge(rq, bio, nr_segs); |
34fe7c05 | 794 | break; |
1e739730 CH |
795 | case ELEVATOR_DISCARD_MERGE: |
796 | merged = bio_attempt_discard_merge(q, rq, bio); | |
797 | break; | |
34fe7c05 CH |
798 | default: |
799 | break; | |
73c10101 | 800 | } |
34fe7c05 CH |
801 | |
802 | if (merged) | |
803 | return true; | |
73c10101 | 804 | } |
34fe7c05 CH |
805 | |
806 | return false; | |
73c10101 JA |
807 | } |
808 | ||
52c5e62d | 809 | static void handle_bad_sector(struct bio *bio, sector_t maxsector) |
1da177e4 LT |
810 | { |
811 | char b[BDEVNAME_SIZE]; | |
812 | ||
813 | printk(KERN_INFO "attempt to access beyond end of device\n"); | |
6296b960 | 814 | printk(KERN_INFO "%s: rw=%d, want=%Lu, limit=%Lu\n", |
74d46992 | 815 | bio_devname(bio, b), bio->bi_opf, |
f73a1c7d | 816 | (unsigned long long)bio_end_sector(bio), |
52c5e62d | 817 | (long long)maxsector); |
1da177e4 LT |
818 | } |
819 | ||
c17bb495 AM |
820 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
821 | ||
822 | static DECLARE_FAULT_ATTR(fail_make_request); | |
823 | ||
824 | static int __init setup_fail_make_request(char *str) | |
825 | { | |
826 | return setup_fault_attr(&fail_make_request, str); | |
827 | } | |
828 | __setup("fail_make_request=", setup_fail_make_request); | |
829 | ||
b2c9cd37 | 830 | static bool should_fail_request(struct hd_struct *part, unsigned int bytes) |
c17bb495 | 831 | { |
b2c9cd37 | 832 | return part->make_it_fail && should_fail(&fail_make_request, bytes); |
c17bb495 AM |
833 | } |
834 | ||
835 | static int __init fail_make_request_debugfs(void) | |
836 | { | |
dd48c085 AM |
837 | struct dentry *dir = fault_create_debugfs_attr("fail_make_request", |
838 | NULL, &fail_make_request); | |
839 | ||
21f9fcd8 | 840 | return PTR_ERR_OR_ZERO(dir); |
c17bb495 AM |
841 | } |
842 | ||
843 | late_initcall(fail_make_request_debugfs); | |
844 | ||
845 | #else /* CONFIG_FAIL_MAKE_REQUEST */ | |
846 | ||
b2c9cd37 AM |
847 | static inline bool should_fail_request(struct hd_struct *part, |
848 | unsigned int bytes) | |
c17bb495 | 849 | { |
b2c9cd37 | 850 | return false; |
c17bb495 AM |
851 | } |
852 | ||
853 | #endif /* CONFIG_FAIL_MAKE_REQUEST */ | |
854 | ||
721c7fc7 ID |
855 | static inline bool bio_check_ro(struct bio *bio, struct hd_struct *part) |
856 | { | |
b089cfd9 JA |
857 | const int op = bio_op(bio); |
858 | ||
8b2ded1c | 859 | if (part->policy && op_is_write(op)) { |
721c7fc7 ID |
860 | char b[BDEVNAME_SIZE]; |
861 | ||
8b2ded1c MP |
862 | if (op_is_flush(bio->bi_opf) && !bio_sectors(bio)) |
863 | return false; | |
864 | ||
a32e236e | 865 | WARN_ONCE(1, |
721c7fc7 ID |
866 | "generic_make_request: Trying to write " |
867 | "to read-only block-device %s (partno %d)\n", | |
868 | bio_devname(bio, b), part->partno); | |
a32e236e LT |
869 | /* Older lvm-tools actually trigger this */ |
870 | return false; | |
721c7fc7 ID |
871 | } |
872 | ||
873 | return false; | |
874 | } | |
875 | ||
30abb3a6 HM |
876 | static noinline int should_fail_bio(struct bio *bio) |
877 | { | |
878 | if (should_fail_request(&bio->bi_disk->part0, bio->bi_iter.bi_size)) | |
879 | return -EIO; | |
880 | return 0; | |
881 | } | |
882 | ALLOW_ERROR_INJECTION(should_fail_bio, ERRNO); | |
883 | ||
52c5e62d CH |
884 | /* |
885 | * Check whether this bio extends beyond the end of the device or partition. | |
886 | * This may well happen - the kernel calls bread() without checking the size of | |
887 | * the device, e.g., when mounting a file system. | |
888 | */ | |
889 | static inline int bio_check_eod(struct bio *bio, sector_t maxsector) | |
890 | { | |
891 | unsigned int nr_sectors = bio_sectors(bio); | |
892 | ||
893 | if (nr_sectors && maxsector && | |
894 | (nr_sectors > maxsector || | |
895 | bio->bi_iter.bi_sector > maxsector - nr_sectors)) { | |
896 | handle_bad_sector(bio, maxsector); | |
897 | return -EIO; | |
898 | } | |
899 | return 0; | |
900 | } | |
901 | ||
74d46992 CH |
902 | /* |
903 | * Remap block n of partition p to block n+start(p) of the disk. | |
904 | */ | |
905 | static inline int blk_partition_remap(struct bio *bio) | |
906 | { | |
907 | struct hd_struct *p; | |
52c5e62d | 908 | int ret = -EIO; |
74d46992 | 909 | |
721c7fc7 ID |
910 | rcu_read_lock(); |
911 | p = __disk_get_part(bio->bi_disk, bio->bi_partno); | |
52c5e62d CH |
912 | if (unlikely(!p)) |
913 | goto out; | |
914 | if (unlikely(should_fail_request(p, bio->bi_iter.bi_size))) | |
915 | goto out; | |
916 | if (unlikely(bio_check_ro(bio, p))) | |
721c7fc7 | 917 | goto out; |
721c7fc7 | 918 | |
5eac3eb3 | 919 | if (bio_sectors(bio)) { |
52c5e62d CH |
920 | if (bio_check_eod(bio, part_nr_sects_read(p))) |
921 | goto out; | |
922 | bio->bi_iter.bi_sector += p->start_sect; | |
52c5e62d CH |
923 | trace_block_bio_remap(bio->bi_disk->queue, bio, part_devt(p), |
924 | bio->bi_iter.bi_sector - p->start_sect); | |
925 | } | |
c04fa44b | 926 | bio->bi_partno = 0; |
52c5e62d | 927 | ret = 0; |
721c7fc7 ID |
928 | out: |
929 | rcu_read_unlock(); | |
74d46992 CH |
930 | return ret; |
931 | } | |
932 | ||
0512a75b KB |
933 | /* |
934 | * Check write append to a zoned block device. | |
935 | */ | |
936 | static inline blk_status_t blk_check_zone_append(struct request_queue *q, | |
937 | struct bio *bio) | |
938 | { | |
939 | sector_t pos = bio->bi_iter.bi_sector; | |
940 | int nr_sectors = bio_sectors(bio); | |
941 | ||
942 | /* Only applicable to zoned block devices */ | |
943 | if (!blk_queue_is_zoned(q)) | |
944 | return BLK_STS_NOTSUPP; | |
945 | ||
946 | /* The bio sector must point to the start of a sequential zone */ | |
947 | if (pos & (blk_queue_zone_sectors(q) - 1) || | |
948 | !blk_queue_zone_is_seq(q, pos)) | |
949 | return BLK_STS_IOERR; | |
950 | ||
951 | /* | |
952 | * Not allowed to cross zone boundaries. Otherwise, the BIO will be | |
953 | * split and could result in non-contiguous sectors being written in | |
954 | * different zones. | |
955 | */ | |
956 | if (nr_sectors > q->limits.chunk_sectors) | |
957 | return BLK_STS_IOERR; | |
958 | ||
959 | /* Make sure the BIO is small enough and will not get split */ | |
960 | if (nr_sectors > q->limits.max_zone_append_sectors) | |
961 | return BLK_STS_IOERR; | |
962 | ||
963 | bio->bi_opf |= REQ_NOMERGE; | |
964 | ||
965 | return BLK_STS_OK; | |
966 | } | |
967 | ||
27a84d54 CH |
968 | static noinline_for_stack bool |
969 | generic_make_request_checks(struct bio *bio) | |
1da177e4 | 970 | { |
165125e1 | 971 | struct request_queue *q; |
5a7bbad2 | 972 | int nr_sectors = bio_sectors(bio); |
4e4cbee9 | 973 | blk_status_t status = BLK_STS_IOERR; |
5a7bbad2 | 974 | char b[BDEVNAME_SIZE]; |
1da177e4 LT |
975 | |
976 | might_sleep(); | |
1da177e4 | 977 | |
74d46992 | 978 | q = bio->bi_disk->queue; |
5a7bbad2 CH |
979 | if (unlikely(!q)) { |
980 | printk(KERN_ERR | |
981 | "generic_make_request: Trying to access " | |
982 | "nonexistent block-device %s (%Lu)\n", | |
74d46992 | 983 | bio_devname(bio, b), (long long)bio->bi_iter.bi_sector); |
5a7bbad2 CH |
984 | goto end_io; |
985 | } | |
c17bb495 | 986 | |
03a07c92 | 987 | /* |
b0beb280 JA |
988 | * For a REQ_NOWAIT based request, return -EOPNOTSUPP |
989 | * if queue is not a request based queue. | |
03a07c92 | 990 | */ |
b0beb280 JA |
991 | if ((bio->bi_opf & REQ_NOWAIT) && !queue_is_mq(q)) |
992 | goto not_supported; | |
03a07c92 | 993 | |
30abb3a6 | 994 | if (should_fail_bio(bio)) |
5a7bbad2 | 995 | goto end_io; |
2056a782 | 996 | |
52c5e62d CH |
997 | if (bio->bi_partno) { |
998 | if (unlikely(blk_partition_remap(bio))) | |
721c7fc7 ID |
999 | goto end_io; |
1000 | } else { | |
52c5e62d CH |
1001 | if (unlikely(bio_check_ro(bio, &bio->bi_disk->part0))) |
1002 | goto end_io; | |
1003 | if (unlikely(bio_check_eod(bio, get_capacity(bio->bi_disk)))) | |
721c7fc7 ID |
1004 | goto end_io; |
1005 | } | |
2056a782 | 1006 | |
5a7bbad2 CH |
1007 | /* |
1008 | * Filter flush bio's early so that make_request based | |
1009 | * drivers without flush support don't have to worry | |
1010 | * about them. | |
1011 | */ | |
f3a8ab7d | 1012 | if (op_is_flush(bio->bi_opf) && |
c888a8f9 | 1013 | !test_bit(QUEUE_FLAG_WC, &q->queue_flags)) { |
1eff9d32 | 1014 | bio->bi_opf &= ~(REQ_PREFLUSH | REQ_FUA); |
5a7bbad2 | 1015 | if (!nr_sectors) { |
4e4cbee9 | 1016 | status = BLK_STS_OK; |
51fd77bd JA |
1017 | goto end_io; |
1018 | } | |
5a7bbad2 | 1019 | } |
5ddfe969 | 1020 | |
d04c406f CH |
1021 | if (!test_bit(QUEUE_FLAG_POLL, &q->queue_flags)) |
1022 | bio->bi_opf &= ~REQ_HIPRI; | |
1023 | ||
288dab8a CH |
1024 | switch (bio_op(bio)) { |
1025 | case REQ_OP_DISCARD: | |
1026 | if (!blk_queue_discard(q)) | |
1027 | goto not_supported; | |
1028 | break; | |
1029 | case REQ_OP_SECURE_ERASE: | |
1030 | if (!blk_queue_secure_erase(q)) | |
1031 | goto not_supported; | |
1032 | break; | |
1033 | case REQ_OP_WRITE_SAME: | |
74d46992 | 1034 | if (!q->limits.max_write_same_sectors) |
288dab8a | 1035 | goto not_supported; |
58886785 | 1036 | break; |
0512a75b KB |
1037 | case REQ_OP_ZONE_APPEND: |
1038 | status = blk_check_zone_append(q, bio); | |
1039 | if (status != BLK_STS_OK) | |
1040 | goto end_io; | |
1041 | break; | |
2d253440 | 1042 | case REQ_OP_ZONE_RESET: |
6c1b1da5 AJ |
1043 | case REQ_OP_ZONE_OPEN: |
1044 | case REQ_OP_ZONE_CLOSE: | |
1045 | case REQ_OP_ZONE_FINISH: | |
74d46992 | 1046 | if (!blk_queue_is_zoned(q)) |
2d253440 | 1047 | goto not_supported; |
288dab8a | 1048 | break; |
6e33dbf2 CK |
1049 | case REQ_OP_ZONE_RESET_ALL: |
1050 | if (!blk_queue_is_zoned(q) || !blk_queue_zone_resetall(q)) | |
1051 | goto not_supported; | |
1052 | break; | |
a6f0788e | 1053 | case REQ_OP_WRITE_ZEROES: |
74d46992 | 1054 | if (!q->limits.max_write_zeroes_sectors) |
a6f0788e CK |
1055 | goto not_supported; |
1056 | break; | |
288dab8a CH |
1057 | default: |
1058 | break; | |
5a7bbad2 | 1059 | } |
01edede4 | 1060 | |
7f4b35d1 | 1061 | /* |
3e82c348 CH |
1062 | * Various block parts want %current->io_context, so allocate it up |
1063 | * front rather than dealing with lots of pain to allocate it only | |
1064 | * where needed. This may fail and the block layer knows how to live | |
1065 | * with it. | |
7f4b35d1 | 1066 | */ |
3e82c348 CH |
1067 | if (unlikely(!current->io_context)) |
1068 | create_task_io_context(current, GFP_ATOMIC, q->node); | |
7f4b35d1 | 1069 | |
ae118896 TH |
1070 | if (!blkcg_bio_issue_check(q, bio)) |
1071 | return false; | |
27a84d54 | 1072 | |
fbbaf700 N |
1073 | if (!bio_flagged(bio, BIO_TRACE_COMPLETION)) { |
1074 | trace_block_bio_queue(q, bio); | |
1075 | /* Now that enqueuing has been traced, we need to trace | |
1076 | * completion as well. | |
1077 | */ | |
1078 | bio_set_flag(bio, BIO_TRACE_COMPLETION); | |
1079 | } | |
27a84d54 | 1080 | return true; |
a7384677 | 1081 | |
288dab8a | 1082 | not_supported: |
4e4cbee9 | 1083 | status = BLK_STS_NOTSUPP; |
a7384677 | 1084 | end_io: |
4e4cbee9 | 1085 | bio->bi_status = status; |
4246a0b6 | 1086 | bio_endio(bio); |
27a84d54 | 1087 | return false; |
1da177e4 LT |
1088 | } |
1089 | ||
ac7c5675 CH |
1090 | static blk_qc_t do_make_request(struct bio *bio) |
1091 | { | |
1092 | struct request_queue *q = bio->bi_disk->queue; | |
1093 | blk_qc_t ret = BLK_QC_T_NONE; | |
1094 | ||
1095 | if (blk_crypto_bio_prep(&bio)) { | |
1096 | if (!q->make_request_fn) | |
1097 | return blk_mq_make_request(q, bio); | |
1098 | ret = q->make_request_fn(q, bio); | |
1099 | } | |
1100 | blk_queue_exit(q); | |
1101 | return ret; | |
1102 | } | |
1103 | ||
27a84d54 | 1104 | /** |
3fdd4086 | 1105 | * generic_make_request - re-submit a bio to the block device layer for I/O |
27a84d54 CH |
1106 | * @bio: The bio describing the location in memory and on the device. |
1107 | * | |
3fdd4086 CH |
1108 | * This is a version of submit_bio() that shall only be used for I/O that is |
1109 | * resubmitted to lower level drivers by stacking block drivers. All file | |
1110 | * systems and other upper level users of the block layer should use | |
1111 | * submit_bio() instead. | |
d89d8796 | 1112 | */ |
dece1635 | 1113 | blk_qc_t generic_make_request(struct bio *bio) |
d89d8796 | 1114 | { |
f5fe1b51 N |
1115 | /* |
1116 | * bio_list_on_stack[0] contains bios submitted by the current | |
1117 | * make_request_fn. | |
1118 | * bio_list_on_stack[1] contains bios that were submitted before | |
1119 | * the current make_request_fn, but that haven't been processed | |
1120 | * yet. | |
1121 | */ | |
1122 | struct bio_list bio_list_on_stack[2]; | |
dece1635 | 1123 | blk_qc_t ret = BLK_QC_T_NONE; |
bddd87c7 | 1124 | |
27a84d54 | 1125 | if (!generic_make_request_checks(bio)) |
dece1635 | 1126 | goto out; |
27a84d54 CH |
1127 | |
1128 | /* | |
1129 | * We only want one ->make_request_fn to be active at a time, else | |
1130 | * stack usage with stacked devices could be a problem. So use | |
1131 | * current->bio_list to keep a list of requests submited by a | |
1132 | * make_request_fn function. current->bio_list is also used as a | |
1133 | * flag to say if generic_make_request is currently active in this | |
1134 | * task or not. If it is NULL, then no make_request is active. If | |
1135 | * it is non-NULL, then a make_request is active, and new requests | |
1136 | * should be added at the tail | |
1137 | */ | |
bddd87c7 | 1138 | if (current->bio_list) { |
f5fe1b51 | 1139 | bio_list_add(¤t->bio_list[0], bio); |
dece1635 | 1140 | goto out; |
d89d8796 | 1141 | } |
27a84d54 | 1142 | |
d89d8796 NB |
1143 | /* following loop may be a bit non-obvious, and so deserves some |
1144 | * explanation. | |
1145 | * Before entering the loop, bio->bi_next is NULL (as all callers | |
1146 | * ensure that) so we have a list with a single bio. | |
1147 | * We pretend that we have just taken it off a longer list, so | |
bddd87c7 AM |
1148 | * we assign bio_list to a pointer to the bio_list_on_stack, |
1149 | * thus initialising the bio_list of new bios to be | |
27a84d54 | 1150 | * added. ->make_request() may indeed add some more bios |
d89d8796 NB |
1151 | * through a recursive call to generic_make_request. If it |
1152 | * did, we find a non-NULL value in bio_list and re-enter the loop | |
1153 | * from the top. In this case we really did just take the bio | |
bddd87c7 | 1154 | * of the top of the list (no pretending) and so remove it from |
27a84d54 | 1155 | * bio_list, and call into ->make_request() again. |
d89d8796 NB |
1156 | */ |
1157 | BUG_ON(bio->bi_next); | |
f5fe1b51 N |
1158 | bio_list_init(&bio_list_on_stack[0]); |
1159 | current->bio_list = bio_list_on_stack; | |
d89d8796 | 1160 | do { |
fe200864 | 1161 | struct request_queue *q = bio->bi_disk->queue; |
27a84d54 | 1162 | |
accea322 | 1163 | if (likely(bio_queue_enter(bio) == 0)) { |
79bd9959 N |
1164 | struct bio_list lower, same; |
1165 | ||
1166 | /* Create a fresh bio_list for all subordinate requests */ | |
f5fe1b51 N |
1167 | bio_list_on_stack[1] = bio_list_on_stack[0]; |
1168 | bio_list_init(&bio_list_on_stack[0]); | |
ac7c5675 | 1169 | ret = do_make_request(bio); |
fe200864 | 1170 | |
79bd9959 N |
1171 | /* sort new bios into those for a lower level |
1172 | * and those for the same level | |
1173 | */ | |
1174 | bio_list_init(&lower); | |
1175 | bio_list_init(&same); | |
f5fe1b51 | 1176 | while ((bio = bio_list_pop(&bio_list_on_stack[0])) != NULL) |
74d46992 | 1177 | if (q == bio->bi_disk->queue) |
79bd9959 N |
1178 | bio_list_add(&same, bio); |
1179 | else | |
1180 | bio_list_add(&lower, bio); | |
1181 | /* now assemble so we handle the lowest level first */ | |
f5fe1b51 N |
1182 | bio_list_merge(&bio_list_on_stack[0], &lower); |
1183 | bio_list_merge(&bio_list_on_stack[0], &same); | |
1184 | bio_list_merge(&bio_list_on_stack[0], &bio_list_on_stack[1]); | |
3ef28e83 | 1185 | } |
f5fe1b51 | 1186 | bio = bio_list_pop(&bio_list_on_stack[0]); |
d89d8796 | 1187 | } while (bio); |
bddd87c7 | 1188 | current->bio_list = NULL; /* deactivate */ |
dece1635 JA |
1189 | |
1190 | out: | |
1191 | return ret; | |
d89d8796 | 1192 | } |
1da177e4 LT |
1193 | EXPORT_SYMBOL(generic_make_request); |
1194 | ||
f421e1d9 CH |
1195 | /** |
1196 | * direct_make_request - hand a buffer directly to its device driver for I/O | |
1197 | * @bio: The bio describing the location in memory and on the device. | |
1198 | * | |
1199 | * This function behaves like generic_make_request(), but does not protect | |
1200 | * against recursion. Must only be used if the called driver is known | |
8cf7961d | 1201 | * to be blk-mq based. |
f421e1d9 CH |
1202 | */ |
1203 | blk_qc_t direct_make_request(struct bio *bio) | |
1204 | { | |
1205 | struct request_queue *q = bio->bi_disk->queue; | |
f421e1d9 | 1206 | |
accea322 CH |
1207 | if (WARN_ON_ONCE(q->make_request_fn)) { |
1208 | bio_io_error(bio); | |
f421e1d9 | 1209 | return BLK_QC_T_NONE; |
f421e1d9 | 1210 | } |
f421e1d9 CH |
1211 | if (!generic_make_request_checks(bio)) |
1212 | return BLK_QC_T_NONE; | |
accea322 CH |
1213 | if (unlikely(bio_queue_enter(bio))) |
1214 | return BLK_QC_T_NONE; | |
ac7c5675 CH |
1215 | if (!blk_crypto_bio_prep(&bio)) { |
1216 | blk_queue_exit(q); | |
f421e1d9 CH |
1217 | return BLK_QC_T_NONE; |
1218 | } | |
ac7c5675 | 1219 | return blk_mq_make_request(q, bio); |
f421e1d9 CH |
1220 | } |
1221 | EXPORT_SYMBOL_GPL(direct_make_request); | |
1222 | ||
1da177e4 | 1223 | /** |
710027a4 | 1224 | * submit_bio - submit a bio to the block device layer for I/O |
1da177e4 LT |
1225 | * @bio: The &struct bio which describes the I/O |
1226 | * | |
3fdd4086 CH |
1227 | * submit_bio() is used to submit I/O requests to block devices. It is passed a |
1228 | * fully set up &struct bio that describes the I/O that needs to be done. The | |
1229 | * bio will be send to the device described by the bi_disk and bi_partno fields. | |
1da177e4 | 1230 | * |
3fdd4086 CH |
1231 | * The success/failure status of the request, along with notification of |
1232 | * completion, is delivered asynchronously through the ->bi_end_io() callback | |
1233 | * in @bio. The bio must NOT be touched by thecaller until ->bi_end_io() has | |
1234 | * been called. | |
1da177e4 | 1235 | */ |
4e49ea4a | 1236 | blk_qc_t submit_bio(struct bio *bio) |
1da177e4 | 1237 | { |
d3f77dfd TH |
1238 | if (blkcg_punt_bio_submit(bio)) |
1239 | return BLK_QC_T_NONE; | |
1240 | ||
bf2de6f5 JA |
1241 | /* |
1242 | * If it's a regular read/write or a barrier with data attached, | |
1243 | * go through the normal accounting stuff before submission. | |
1244 | */ | |
e2a60da7 | 1245 | if (bio_has_data(bio)) { |
4363ac7c MP |
1246 | unsigned int count; |
1247 | ||
95fe6c1a | 1248 | if (unlikely(bio_op(bio) == REQ_OP_WRITE_SAME)) |
7c5a0dcf | 1249 | count = queue_logical_block_size(bio->bi_disk->queue) >> 9; |
4363ac7c MP |
1250 | else |
1251 | count = bio_sectors(bio); | |
1252 | ||
a8ebb056 | 1253 | if (op_is_write(bio_op(bio))) { |
bf2de6f5 JA |
1254 | count_vm_events(PGPGOUT, count); |
1255 | } else { | |
4f024f37 | 1256 | task_io_account_read(bio->bi_iter.bi_size); |
bf2de6f5 JA |
1257 | count_vm_events(PGPGIN, count); |
1258 | } | |
1259 | ||
1260 | if (unlikely(block_dump)) { | |
1261 | char b[BDEVNAME_SIZE]; | |
8dcbdc74 | 1262 | printk(KERN_DEBUG "%s(%d): %s block %Lu on %s (%u sectors)\n", |
ba25f9dc | 1263 | current->comm, task_pid_nr(current), |
a8ebb056 | 1264 | op_is_write(bio_op(bio)) ? "WRITE" : "READ", |
4f024f37 | 1265 | (unsigned long long)bio->bi_iter.bi_sector, |
74d46992 | 1266 | bio_devname(bio, b), count); |
bf2de6f5 | 1267 | } |
1da177e4 LT |
1268 | } |
1269 | ||
b8e24a93 | 1270 | /* |
760f83ea CH |
1271 | * If we're reading data that is part of the userspace workingset, count |
1272 | * submission time as memory stall. When the device is congested, or | |
1273 | * the submitting cgroup IO-throttled, submission can be a significant | |
1274 | * part of overall IO time. | |
b8e24a93 | 1275 | */ |
760f83ea CH |
1276 | if (unlikely(bio_op(bio) == REQ_OP_READ && |
1277 | bio_flagged(bio, BIO_WORKINGSET))) { | |
1278 | unsigned long pflags; | |
1279 | blk_qc_t ret; | |
b8e24a93 | 1280 | |
760f83ea CH |
1281 | psi_memstall_enter(&pflags); |
1282 | ret = generic_make_request(bio); | |
b8e24a93 JW |
1283 | psi_memstall_leave(&pflags); |
1284 | ||
760f83ea CH |
1285 | return ret; |
1286 | } | |
1287 | ||
1288 | return generic_make_request(bio); | |
1da177e4 | 1289 | } |
1da177e4 LT |
1290 | EXPORT_SYMBOL(submit_bio); |
1291 | ||
82124d60 | 1292 | /** |
bf4e6b4e | 1293 | * blk_cloned_rq_check_limits - Helper function to check a cloned request |
0d720318 | 1294 | * for the new queue limits |
82124d60 KU |
1295 | * @q: the queue |
1296 | * @rq: the request being checked | |
1297 | * | |
1298 | * Description: | |
1299 | * @rq may have been made based on weaker limitations of upper-level queues | |
1300 | * in request stacking drivers, and it may violate the limitation of @q. | |
1301 | * Since the block layer and the underlying device driver trust @rq | |
1302 | * after it is inserted to @q, it should be checked against @q before | |
1303 | * the insertion using this generic function. | |
1304 | * | |
82124d60 | 1305 | * Request stacking drivers like request-based dm may change the queue |
bf4e6b4e HR |
1306 | * limits when retrying requests on other queues. Those requests need |
1307 | * to be checked against the new queue limits again during dispatch. | |
82124d60 | 1308 | */ |
bf4e6b4e HR |
1309 | static int blk_cloned_rq_check_limits(struct request_queue *q, |
1310 | struct request *rq) | |
82124d60 | 1311 | { |
8fe0d473 | 1312 | if (blk_rq_sectors(rq) > blk_queue_get_max_sectors(q, req_op(rq))) { |
61939b12 JP |
1313 | printk(KERN_ERR "%s: over max size limit. (%u > %u)\n", |
1314 | __func__, blk_rq_sectors(rq), | |
1315 | blk_queue_get_max_sectors(q, req_op(rq))); | |
82124d60 KU |
1316 | return -EIO; |
1317 | } | |
1318 | ||
1319 | /* | |
1320 | * queue's settings related to segment counting like q->bounce_pfn | |
1321 | * may differ from that of other stacking queues. | |
1322 | * Recalculate it to check the request correctly on this queue's | |
1323 | * limitation. | |
1324 | */ | |
e9cd19c0 | 1325 | rq->nr_phys_segments = blk_recalc_rq_segments(rq); |
8a78362c | 1326 | if (rq->nr_phys_segments > queue_max_segments(q)) { |
61939b12 JP |
1327 | printk(KERN_ERR "%s: over max segments limit. (%hu > %hu)\n", |
1328 | __func__, rq->nr_phys_segments, queue_max_segments(q)); | |
82124d60 KU |
1329 | return -EIO; |
1330 | } | |
1331 | ||
1332 | return 0; | |
1333 | } | |
82124d60 KU |
1334 | |
1335 | /** | |
1336 | * blk_insert_cloned_request - Helper for stacking drivers to submit a request | |
1337 | * @q: the queue to submit the request | |
1338 | * @rq: the request being queued | |
1339 | */ | |
2a842aca | 1340 | blk_status_t blk_insert_cloned_request(struct request_queue *q, struct request *rq) |
82124d60 | 1341 | { |
bf4e6b4e | 1342 | if (blk_cloned_rq_check_limits(q, rq)) |
2a842aca | 1343 | return BLK_STS_IOERR; |
82124d60 | 1344 | |
b2c9cd37 AM |
1345 | if (rq->rq_disk && |
1346 | should_fail_request(&rq->rq_disk->part0, blk_rq_bytes(rq))) | |
2a842aca | 1347 | return BLK_STS_IOERR; |
82124d60 | 1348 | |
a892c8d5 ST |
1349 | if (blk_crypto_insert_cloned_request(rq)) |
1350 | return BLK_STS_IOERR; | |
1351 | ||
a1ce35fa | 1352 | if (blk_queue_io_stat(q)) |
b5af37ab | 1353 | blk_account_io_start(rq); |
82124d60 KU |
1354 | |
1355 | /* | |
a1ce35fa JA |
1356 | * Since we have a scheduler attached on the top device, |
1357 | * bypass a potential scheduler on the bottom device for | |
1358 | * insert. | |
82124d60 | 1359 | */ |
fd9c40f6 | 1360 | return blk_mq_request_issue_directly(rq, true); |
82124d60 KU |
1361 | } |
1362 | EXPORT_SYMBOL_GPL(blk_insert_cloned_request); | |
1363 | ||
80a761fd TH |
1364 | /** |
1365 | * blk_rq_err_bytes - determine number of bytes till the next failure boundary | |
1366 | * @rq: request to examine | |
1367 | * | |
1368 | * Description: | |
1369 | * A request could be merge of IOs which require different failure | |
1370 | * handling. This function determines the number of bytes which | |
1371 | * can be failed from the beginning of the request without | |
1372 | * crossing into area which need to be retried further. | |
1373 | * | |
1374 | * Return: | |
1375 | * The number of bytes to fail. | |
80a761fd TH |
1376 | */ |
1377 | unsigned int blk_rq_err_bytes(const struct request *rq) | |
1378 | { | |
1379 | unsigned int ff = rq->cmd_flags & REQ_FAILFAST_MASK; | |
1380 | unsigned int bytes = 0; | |
1381 | struct bio *bio; | |
1382 | ||
e8064021 | 1383 | if (!(rq->rq_flags & RQF_MIXED_MERGE)) |
80a761fd TH |
1384 | return blk_rq_bytes(rq); |
1385 | ||
1386 | /* | |
1387 | * Currently the only 'mixing' which can happen is between | |
1388 | * different fastfail types. We can safely fail portions | |
1389 | * which have all the failfast bits that the first one has - | |
1390 | * the ones which are at least as eager to fail as the first | |
1391 | * one. | |
1392 | */ | |
1393 | for (bio = rq->bio; bio; bio = bio->bi_next) { | |
1eff9d32 | 1394 | if ((bio->bi_opf & ff) != ff) |
80a761fd | 1395 | break; |
4f024f37 | 1396 | bytes += bio->bi_iter.bi_size; |
80a761fd TH |
1397 | } |
1398 | ||
1399 | /* this could lead to infinite loop */ | |
1400 | BUG_ON(blk_rq_bytes(rq) && !bytes); | |
1401 | return bytes; | |
1402 | } | |
1403 | EXPORT_SYMBOL_GPL(blk_rq_err_bytes); | |
1404 | ||
9123bf6f CH |
1405 | static void update_io_ticks(struct hd_struct *part, unsigned long now, bool end) |
1406 | { | |
1407 | unsigned long stamp; | |
1408 | again: | |
1409 | stamp = READ_ONCE(part->stamp); | |
1410 | if (unlikely(stamp != now)) { | |
1411 | if (likely(cmpxchg(&part->stamp, stamp, now) == stamp)) | |
1412 | __part_stat_add(part, io_ticks, end ? now - stamp : 1); | |
1413 | } | |
1414 | if (part->partno) { | |
1415 | part = &part_to_disk(part)->part0; | |
1416 | goto again; | |
1417 | } | |
1418 | } | |
1419 | ||
f1394b79 | 1420 | static void blk_account_io_completion(struct request *req, unsigned int bytes) |
bc58ba94 | 1421 | { |
ecb6186c | 1422 | if (req->part && blk_do_io_stat(req)) { |
ddcf35d3 | 1423 | const int sgrp = op_stat_group(req_op(req)); |
bc58ba94 | 1424 | struct hd_struct *part; |
bc58ba94 | 1425 | |
112f158f | 1426 | part_stat_lock(); |
09e099d4 | 1427 | part = req->part; |
112f158f | 1428 | part_stat_add(part, sectors[sgrp], bytes >> 9); |
bc58ba94 JA |
1429 | part_stat_unlock(); |
1430 | } | |
1431 | } | |
1432 | ||
522a7775 | 1433 | void blk_account_io_done(struct request *req, u64 now) |
bc58ba94 | 1434 | { |
bc58ba94 | 1435 | /* |
dd4c133f TH |
1436 | * Account IO completion. flush_rq isn't accounted as a |
1437 | * normal IO on queueing nor completion. Accounting the | |
1438 | * containing request is enough. | |
bc58ba94 | 1439 | */ |
ecb6186c LG |
1440 | if (req->part && blk_do_io_stat(req) && |
1441 | !(req->rq_flags & RQF_FLUSH_SEQ)) { | |
ddcf35d3 | 1442 | const int sgrp = op_stat_group(req_op(req)); |
bc58ba94 | 1443 | struct hd_struct *part; |
bc58ba94 | 1444 | |
112f158f | 1445 | part_stat_lock(); |
09e099d4 | 1446 | part = req->part; |
bc58ba94 | 1447 | |
2b8bd423 | 1448 | update_io_ticks(part, jiffies, true); |
112f158f MS |
1449 | part_stat_inc(part, ios[sgrp]); |
1450 | part_stat_add(part, nsecs[sgrp], now - req->start_time_ns); | |
524f9ffd | 1451 | part_stat_unlock(); |
bc58ba94 | 1452 | |
6c23a968 | 1453 | hd_struct_put(part); |
bc58ba94 JA |
1454 | } |
1455 | } | |
1456 | ||
b5af37ab | 1457 | void blk_account_io_start(struct request *rq) |
320ae51f | 1458 | { |
320ae51f JA |
1459 | if (!blk_do_io_stat(rq)) |
1460 | return; | |
1461 | ||
b5af37ab | 1462 | rq->part = disk_map_sector_rcu(rq->rq_disk, blk_rq_pos(rq)); |
524f9ffd | 1463 | |
112f158f | 1464 | part_stat_lock(); |
76268f3a | 1465 | update_io_ticks(rq->part, jiffies, false); |
320ae51f JA |
1466 | part_stat_unlock(); |
1467 | } | |
320ae51f | 1468 | |
956d510e CH |
1469 | unsigned long disk_start_io_acct(struct gendisk *disk, unsigned int sectors, |
1470 | unsigned int op) | |
1471 | { | |
1472 | struct hd_struct *part = &disk->part0; | |
1473 | const int sgrp = op_stat_group(op); | |
1474 | unsigned long now = READ_ONCE(jiffies); | |
1475 | ||
1476 | part_stat_lock(); | |
1477 | update_io_ticks(part, now, false); | |
1478 | part_stat_inc(part, ios[sgrp]); | |
1479 | part_stat_add(part, sectors[sgrp], sectors); | |
1480 | part_stat_local_inc(part, in_flight[op_is_write(op)]); | |
1481 | part_stat_unlock(); | |
320ae51f | 1482 | |
956d510e CH |
1483 | return now; |
1484 | } | |
1485 | EXPORT_SYMBOL(disk_start_io_acct); | |
1486 | ||
1487 | void disk_end_io_acct(struct gendisk *disk, unsigned int op, | |
1488 | unsigned long start_time) | |
1489 | { | |
1490 | struct hd_struct *part = &disk->part0; | |
1491 | const int sgrp = op_stat_group(op); | |
1492 | unsigned long now = READ_ONCE(jiffies); | |
1493 | unsigned long duration = now - start_time; | |
5b18b5a7 | 1494 | |
956d510e CH |
1495 | part_stat_lock(); |
1496 | update_io_ticks(part, now, true); | |
1497 | part_stat_add(part, nsecs[sgrp], jiffies_to_nsecs(duration)); | |
1498 | part_stat_local_dec(part, in_flight[op_is_write(op)]); | |
320ae51f JA |
1499 | part_stat_unlock(); |
1500 | } | |
956d510e | 1501 | EXPORT_SYMBOL(disk_end_io_acct); |
320ae51f | 1502 | |
ef71de8b CH |
1503 | /* |
1504 | * Steal bios from a request and add them to a bio list. | |
1505 | * The request must not have been partially completed before. | |
1506 | */ | |
1507 | void blk_steal_bios(struct bio_list *list, struct request *rq) | |
1508 | { | |
1509 | if (rq->bio) { | |
1510 | if (list->tail) | |
1511 | list->tail->bi_next = rq->bio; | |
1512 | else | |
1513 | list->head = rq->bio; | |
1514 | list->tail = rq->biotail; | |
1515 | ||
1516 | rq->bio = NULL; | |
1517 | rq->biotail = NULL; | |
1518 | } | |
1519 | ||
1520 | rq->__data_len = 0; | |
1521 | } | |
1522 | EXPORT_SYMBOL_GPL(blk_steal_bios); | |
1523 | ||
3bcddeac | 1524 | /** |
2e60e022 | 1525 | * blk_update_request - Special helper function for request stacking drivers |
8ebf9756 | 1526 | * @req: the request being processed |
2a842aca | 1527 | * @error: block status code |
8ebf9756 | 1528 | * @nr_bytes: number of bytes to complete @req |
3bcddeac KU |
1529 | * |
1530 | * Description: | |
8ebf9756 RD |
1531 | * Ends I/O on a number of bytes attached to @req, but doesn't complete |
1532 | * the request structure even if @req doesn't have leftover. | |
1533 | * If @req has leftover, sets it up for the next range of segments. | |
2e60e022 TH |
1534 | * |
1535 | * This special helper function is only for request stacking drivers | |
1536 | * (e.g. request-based dm) so that they can handle partial completion. | |
3a211b71 | 1537 | * Actual device drivers should use blk_mq_end_request instead. |
2e60e022 TH |
1538 | * |
1539 | * Passing the result of blk_rq_bytes() as @nr_bytes guarantees | |
1540 | * %false return from this function. | |
3bcddeac | 1541 | * |
1954e9a9 BVA |
1542 | * Note: |
1543 | * The RQF_SPECIAL_PAYLOAD flag is ignored on purpose in both | |
1544 | * blk_rq_bytes() and in blk_update_request(). | |
1545 | * | |
3bcddeac | 1546 | * Return: |
2e60e022 TH |
1547 | * %false - this request doesn't have any more data |
1548 | * %true - this request has more data | |
3bcddeac | 1549 | **/ |
2a842aca CH |
1550 | bool blk_update_request(struct request *req, blk_status_t error, |
1551 | unsigned int nr_bytes) | |
1da177e4 | 1552 | { |
f79ea416 | 1553 | int total_bytes; |
1da177e4 | 1554 | |
2a842aca | 1555 | trace_block_rq_complete(req, blk_status_to_errno(error), nr_bytes); |
4a0efdc9 | 1556 | |
2e60e022 TH |
1557 | if (!req->bio) |
1558 | return false; | |
1559 | ||
54d4e6ab MG |
1560 | #ifdef CONFIG_BLK_DEV_INTEGRITY |
1561 | if (blk_integrity_rq(req) && req_op(req) == REQ_OP_READ && | |
1562 | error == BLK_STS_OK) | |
1563 | req->q->integrity.profile->complete_fn(req, nr_bytes); | |
1564 | #endif | |
1565 | ||
2a842aca CH |
1566 | if (unlikely(error && !blk_rq_is_passthrough(req) && |
1567 | !(req->rq_flags & RQF_QUIET))) | |
178cc590 | 1568 | print_req_error(req, error, __func__); |
1da177e4 | 1569 | |
bc58ba94 | 1570 | blk_account_io_completion(req, nr_bytes); |
d72d904a | 1571 | |
f79ea416 KO |
1572 | total_bytes = 0; |
1573 | while (req->bio) { | |
1574 | struct bio *bio = req->bio; | |
4f024f37 | 1575 | unsigned bio_bytes = min(bio->bi_iter.bi_size, nr_bytes); |
1da177e4 | 1576 | |
9c24c10a | 1577 | if (bio_bytes == bio->bi_iter.bi_size) |
1da177e4 | 1578 | req->bio = bio->bi_next; |
1da177e4 | 1579 | |
fbbaf700 N |
1580 | /* Completion has already been traced */ |
1581 | bio_clear_flag(bio, BIO_TRACE_COMPLETION); | |
f79ea416 | 1582 | req_bio_endio(req, bio, bio_bytes, error); |
1da177e4 | 1583 | |
f79ea416 KO |
1584 | total_bytes += bio_bytes; |
1585 | nr_bytes -= bio_bytes; | |
1da177e4 | 1586 | |
f79ea416 KO |
1587 | if (!nr_bytes) |
1588 | break; | |
1da177e4 LT |
1589 | } |
1590 | ||
1591 | /* | |
1592 | * completely done | |
1593 | */ | |
2e60e022 TH |
1594 | if (!req->bio) { |
1595 | /* | |
1596 | * Reset counters so that the request stacking driver | |
1597 | * can find how many bytes remain in the request | |
1598 | * later. | |
1599 | */ | |
a2dec7b3 | 1600 | req->__data_len = 0; |
2e60e022 TH |
1601 | return false; |
1602 | } | |
1da177e4 | 1603 | |
a2dec7b3 | 1604 | req->__data_len -= total_bytes; |
2e46e8b2 TH |
1605 | |
1606 | /* update sector only for requests with clear definition of sector */ | |
57292b58 | 1607 | if (!blk_rq_is_passthrough(req)) |
a2dec7b3 | 1608 | req->__sector += total_bytes >> 9; |
2e46e8b2 | 1609 | |
80a761fd | 1610 | /* mixed attributes always follow the first bio */ |
e8064021 | 1611 | if (req->rq_flags & RQF_MIXED_MERGE) { |
80a761fd | 1612 | req->cmd_flags &= ~REQ_FAILFAST_MASK; |
1eff9d32 | 1613 | req->cmd_flags |= req->bio->bi_opf & REQ_FAILFAST_MASK; |
80a761fd TH |
1614 | } |
1615 | ||
ed6565e7 CH |
1616 | if (!(req->rq_flags & RQF_SPECIAL_PAYLOAD)) { |
1617 | /* | |
1618 | * If total number of sectors is less than the first segment | |
1619 | * size, something has gone terribly wrong. | |
1620 | */ | |
1621 | if (blk_rq_bytes(req) < blk_rq_cur_bytes(req)) { | |
1622 | blk_dump_rq_flags(req, "request botched"); | |
1623 | req->__data_len = blk_rq_cur_bytes(req); | |
1624 | } | |
2e46e8b2 | 1625 | |
ed6565e7 | 1626 | /* recalculate the number of segments */ |
e9cd19c0 | 1627 | req->nr_phys_segments = blk_recalc_rq_segments(req); |
ed6565e7 | 1628 | } |
2e46e8b2 | 1629 | |
2e60e022 | 1630 | return true; |
1da177e4 | 1631 | } |
2e60e022 | 1632 | EXPORT_SYMBOL_GPL(blk_update_request); |
1da177e4 | 1633 | |
2d4dc890 IL |
1634 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE |
1635 | /** | |
1636 | * rq_flush_dcache_pages - Helper function to flush all pages in a request | |
1637 | * @rq: the request to be flushed | |
1638 | * | |
1639 | * Description: | |
1640 | * Flush all pages in @rq. | |
1641 | */ | |
1642 | void rq_flush_dcache_pages(struct request *rq) | |
1643 | { | |
1644 | struct req_iterator iter; | |
7988613b | 1645 | struct bio_vec bvec; |
2d4dc890 IL |
1646 | |
1647 | rq_for_each_segment(bvec, rq, iter) | |
7988613b | 1648 | flush_dcache_page(bvec.bv_page); |
2d4dc890 IL |
1649 | } |
1650 | EXPORT_SYMBOL_GPL(rq_flush_dcache_pages); | |
1651 | #endif | |
1652 | ||
ef9e3fac KU |
1653 | /** |
1654 | * blk_lld_busy - Check if underlying low-level drivers of a device are busy | |
1655 | * @q : the queue of the device being checked | |
1656 | * | |
1657 | * Description: | |
1658 | * Check if underlying low-level drivers of a device are busy. | |
1659 | * If the drivers want to export their busy state, they must set own | |
1660 | * exporting function using blk_queue_lld_busy() first. | |
1661 | * | |
1662 | * Basically, this function is used only by request stacking drivers | |
1663 | * to stop dispatching requests to underlying devices when underlying | |
1664 | * devices are busy. This behavior helps more I/O merging on the queue | |
1665 | * of the request stacking driver and prevents I/O throughput regression | |
1666 | * on burst I/O load. | |
1667 | * | |
1668 | * Return: | |
1669 | * 0 - Not busy (The request stacking driver should dispatch request) | |
1670 | * 1 - Busy (The request stacking driver should stop dispatching request) | |
1671 | */ | |
1672 | int blk_lld_busy(struct request_queue *q) | |
1673 | { | |
344e9ffc | 1674 | if (queue_is_mq(q) && q->mq_ops->busy) |
9ba20527 | 1675 | return q->mq_ops->busy(q); |
ef9e3fac KU |
1676 | |
1677 | return 0; | |
1678 | } | |
1679 | EXPORT_SYMBOL_GPL(blk_lld_busy); | |
1680 | ||
78d8e58a MS |
1681 | /** |
1682 | * blk_rq_unprep_clone - Helper function to free all bios in a cloned request | |
1683 | * @rq: the clone request to be cleaned up | |
1684 | * | |
1685 | * Description: | |
1686 | * Free all bios in @rq for a cloned request. | |
1687 | */ | |
1688 | void blk_rq_unprep_clone(struct request *rq) | |
1689 | { | |
1690 | struct bio *bio; | |
1691 | ||
1692 | while ((bio = rq->bio) != NULL) { | |
1693 | rq->bio = bio->bi_next; | |
1694 | ||
1695 | bio_put(bio); | |
1696 | } | |
1697 | } | |
1698 | EXPORT_SYMBOL_GPL(blk_rq_unprep_clone); | |
1699 | ||
78d8e58a MS |
1700 | /** |
1701 | * blk_rq_prep_clone - Helper function to setup clone request | |
1702 | * @rq: the request to be setup | |
1703 | * @rq_src: original request to be cloned | |
1704 | * @bs: bio_set that bios for clone are allocated from | |
1705 | * @gfp_mask: memory allocation mask for bio | |
1706 | * @bio_ctr: setup function to be called for each clone bio. | |
1707 | * Returns %0 for success, non %0 for failure. | |
1708 | * @data: private data to be passed to @bio_ctr | |
1709 | * | |
1710 | * Description: | |
1711 | * Clones bios in @rq_src to @rq, and copies attributes of @rq_src to @rq. | |
78d8e58a MS |
1712 | * Also, pages which the original bios are pointing to are not copied |
1713 | * and the cloned bios just point same pages. | |
1714 | * So cloned bios must be completed before original bios, which means | |
1715 | * the caller must complete @rq before @rq_src. | |
1716 | */ | |
1717 | int blk_rq_prep_clone(struct request *rq, struct request *rq_src, | |
1718 | struct bio_set *bs, gfp_t gfp_mask, | |
1719 | int (*bio_ctr)(struct bio *, struct bio *, void *), | |
1720 | void *data) | |
1721 | { | |
1722 | struct bio *bio, *bio_src; | |
1723 | ||
1724 | if (!bs) | |
f4f8154a | 1725 | bs = &fs_bio_set; |
78d8e58a MS |
1726 | |
1727 | __rq_for_each_bio(bio_src, rq_src) { | |
1728 | bio = bio_clone_fast(bio_src, gfp_mask, bs); | |
1729 | if (!bio) | |
1730 | goto free_and_out; | |
1731 | ||
1732 | if (bio_ctr && bio_ctr(bio, bio_src, data)) | |
1733 | goto free_and_out; | |
1734 | ||
1735 | if (rq->bio) { | |
1736 | rq->biotail->bi_next = bio; | |
1737 | rq->biotail = bio; | |
1738 | } else | |
1739 | rq->bio = rq->biotail = bio; | |
1740 | } | |
1741 | ||
361301a2 GJ |
1742 | /* Copy attributes of the original request to the clone request. */ |
1743 | rq->__sector = blk_rq_pos(rq_src); | |
1744 | rq->__data_len = blk_rq_bytes(rq_src); | |
1745 | if (rq_src->rq_flags & RQF_SPECIAL_PAYLOAD) { | |
1746 | rq->rq_flags |= RQF_SPECIAL_PAYLOAD; | |
1747 | rq->special_vec = rq_src->special_vec; | |
1748 | } | |
1749 | rq->nr_phys_segments = rq_src->nr_phys_segments; | |
1750 | rq->ioprio = rq_src->ioprio; | |
78d8e58a | 1751 | |
a892c8d5 ST |
1752 | if (rq->bio) |
1753 | blk_crypto_rq_bio_prep(rq, rq->bio, gfp_mask); | |
78d8e58a MS |
1754 | |
1755 | return 0; | |
1756 | ||
1757 | free_and_out: | |
1758 | if (bio) | |
1759 | bio_put(bio); | |
1760 | blk_rq_unprep_clone(rq); | |
1761 | ||
1762 | return -ENOMEM; | |
b0fd271d KU |
1763 | } |
1764 | EXPORT_SYMBOL_GPL(blk_rq_prep_clone); | |
1765 | ||
59c3d45e | 1766 | int kblockd_schedule_work(struct work_struct *work) |
1da177e4 LT |
1767 | { |
1768 | return queue_work(kblockd_workqueue, work); | |
1769 | } | |
1da177e4 LT |
1770 | EXPORT_SYMBOL(kblockd_schedule_work); |
1771 | ||
818cd1cb JA |
1772 | int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, |
1773 | unsigned long delay) | |
1774 | { | |
1775 | return mod_delayed_work_on(cpu, kblockd_workqueue, dwork, delay); | |
1776 | } | |
1777 | EXPORT_SYMBOL(kblockd_mod_delayed_work_on); | |
1778 | ||
75df7136 SJ |
1779 | /** |
1780 | * blk_start_plug - initialize blk_plug and track it inside the task_struct | |
1781 | * @plug: The &struct blk_plug that needs to be initialized | |
1782 | * | |
1783 | * Description: | |
40405851 JM |
1784 | * blk_start_plug() indicates to the block layer an intent by the caller |
1785 | * to submit multiple I/O requests in a batch. The block layer may use | |
1786 | * this hint to defer submitting I/Os from the caller until blk_finish_plug() | |
1787 | * is called. However, the block layer may choose to submit requests | |
1788 | * before a call to blk_finish_plug() if the number of queued I/Os | |
1789 | * exceeds %BLK_MAX_REQUEST_COUNT, or if the size of the I/O is larger than | |
1790 | * %BLK_PLUG_FLUSH_SIZE. The queued I/Os may also be submitted early if | |
1791 | * the task schedules (see below). | |
1792 | * | |
75df7136 SJ |
1793 | * Tracking blk_plug inside the task_struct will help with auto-flushing the |
1794 | * pending I/O should the task end up blocking between blk_start_plug() and | |
1795 | * blk_finish_plug(). This is important from a performance perspective, but | |
1796 | * also ensures that we don't deadlock. For instance, if the task is blocking | |
1797 | * for a memory allocation, memory reclaim could end up wanting to free a | |
1798 | * page belonging to that request that is currently residing in our private | |
1799 | * plug. By flushing the pending I/O when the process goes to sleep, we avoid | |
1800 | * this kind of deadlock. | |
1801 | */ | |
73c10101 JA |
1802 | void blk_start_plug(struct blk_plug *plug) |
1803 | { | |
1804 | struct task_struct *tsk = current; | |
1805 | ||
dd6cf3e1 SL |
1806 | /* |
1807 | * If this is a nested plug, don't actually assign it. | |
1808 | */ | |
1809 | if (tsk->plug) | |
1810 | return; | |
1811 | ||
320ae51f | 1812 | INIT_LIST_HEAD(&plug->mq_list); |
048c9374 | 1813 | INIT_LIST_HEAD(&plug->cb_list); |
5f0ed774 | 1814 | plug->rq_count = 0; |
ce5b009c | 1815 | plug->multiple_queues = false; |
5f0ed774 | 1816 | |
73c10101 | 1817 | /* |
dd6cf3e1 SL |
1818 | * Store ordering should not be needed here, since a potential |
1819 | * preempt will imply a full memory barrier | |
73c10101 | 1820 | */ |
dd6cf3e1 | 1821 | tsk->plug = plug; |
73c10101 JA |
1822 | } |
1823 | EXPORT_SYMBOL(blk_start_plug); | |
1824 | ||
74018dc3 | 1825 | static void flush_plug_callbacks(struct blk_plug *plug, bool from_schedule) |
048c9374 N |
1826 | { |
1827 | LIST_HEAD(callbacks); | |
1828 | ||
2a7d5559 SL |
1829 | while (!list_empty(&plug->cb_list)) { |
1830 | list_splice_init(&plug->cb_list, &callbacks); | |
048c9374 | 1831 | |
2a7d5559 SL |
1832 | while (!list_empty(&callbacks)) { |
1833 | struct blk_plug_cb *cb = list_first_entry(&callbacks, | |
048c9374 N |
1834 | struct blk_plug_cb, |
1835 | list); | |
2a7d5559 | 1836 | list_del(&cb->list); |
74018dc3 | 1837 | cb->callback(cb, from_schedule); |
2a7d5559 | 1838 | } |
048c9374 N |
1839 | } |
1840 | } | |
1841 | ||
9cbb1750 N |
1842 | struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug, void *data, |
1843 | int size) | |
1844 | { | |
1845 | struct blk_plug *plug = current->plug; | |
1846 | struct blk_plug_cb *cb; | |
1847 | ||
1848 | if (!plug) | |
1849 | return NULL; | |
1850 | ||
1851 | list_for_each_entry(cb, &plug->cb_list, list) | |
1852 | if (cb->callback == unplug && cb->data == data) | |
1853 | return cb; | |
1854 | ||
1855 | /* Not currently on the callback list */ | |
1856 | BUG_ON(size < sizeof(*cb)); | |
1857 | cb = kzalloc(size, GFP_ATOMIC); | |
1858 | if (cb) { | |
1859 | cb->data = data; | |
1860 | cb->callback = unplug; | |
1861 | list_add(&cb->list, &plug->cb_list); | |
1862 | } | |
1863 | return cb; | |
1864 | } | |
1865 | EXPORT_SYMBOL(blk_check_plugged); | |
1866 | ||
49cac01e | 1867 | void blk_flush_plug_list(struct blk_plug *plug, bool from_schedule) |
73c10101 | 1868 | { |
74018dc3 | 1869 | flush_plug_callbacks(plug, from_schedule); |
320ae51f JA |
1870 | |
1871 | if (!list_empty(&plug->mq_list)) | |
1872 | blk_mq_flush_plug_list(plug, from_schedule); | |
73c10101 | 1873 | } |
73c10101 | 1874 | |
40405851 JM |
1875 | /** |
1876 | * blk_finish_plug - mark the end of a batch of submitted I/O | |
1877 | * @plug: The &struct blk_plug passed to blk_start_plug() | |
1878 | * | |
1879 | * Description: | |
1880 | * Indicate that a batch of I/O submissions is complete. This function | |
1881 | * must be paired with an initial call to blk_start_plug(). The intent | |
1882 | * is to allow the block layer to optimize I/O submission. See the | |
1883 | * documentation for blk_start_plug() for more information. | |
1884 | */ | |
73c10101 JA |
1885 | void blk_finish_plug(struct blk_plug *plug) |
1886 | { | |
dd6cf3e1 SL |
1887 | if (plug != current->plug) |
1888 | return; | |
f6603783 | 1889 | blk_flush_plug_list(plug, false); |
73c10101 | 1890 | |
dd6cf3e1 | 1891 | current->plug = NULL; |
73c10101 | 1892 | } |
88b996cd | 1893 | EXPORT_SYMBOL(blk_finish_plug); |
73c10101 | 1894 | |
71ac860a ML |
1895 | void blk_io_schedule(void) |
1896 | { | |
1897 | /* Prevent hang_check timer from firing at us during very long I/O */ | |
1898 | unsigned long timeout = sysctl_hung_task_timeout_secs * HZ / 2; | |
1899 | ||
1900 | if (timeout) | |
1901 | io_schedule_timeout(timeout); | |
1902 | else | |
1903 | io_schedule(); | |
1904 | } | |
1905 | EXPORT_SYMBOL_GPL(blk_io_schedule); | |
1906 | ||
1da177e4 LT |
1907 | int __init blk_dev_init(void) |
1908 | { | |
ef295ecf CH |
1909 | BUILD_BUG_ON(REQ_OP_LAST >= (1 << REQ_OP_BITS)); |
1910 | BUILD_BUG_ON(REQ_OP_BITS + REQ_FLAG_BITS > 8 * | |
c593642c | 1911 | sizeof_field(struct request, cmd_flags)); |
ef295ecf | 1912 | BUILD_BUG_ON(REQ_OP_BITS + REQ_FLAG_BITS > 8 * |
c593642c | 1913 | sizeof_field(struct bio, bi_opf)); |
9eb55b03 | 1914 | |
89b90be2 TH |
1915 | /* used for unplugging and affects IO latency/throughput - HIGHPRI */ |
1916 | kblockd_workqueue = alloc_workqueue("kblockd", | |
28747fcd | 1917 | WQ_MEM_RECLAIM | WQ_HIGHPRI, 0); |
1da177e4 LT |
1918 | if (!kblockd_workqueue) |
1919 | panic("Failed to create kblockd\n"); | |
1920 | ||
c2789bd4 | 1921 | blk_requestq_cachep = kmem_cache_create("request_queue", |
165125e1 | 1922 | sizeof(struct request_queue), 0, SLAB_PANIC, NULL); |
1da177e4 | 1923 | |
18fbda91 OS |
1924 | #ifdef CONFIG_DEBUG_FS |
1925 | blk_debugfs_root = debugfs_create_dir("block", NULL); | |
1926 | #endif | |
1927 | ||
d38ecf93 | 1928 | return 0; |
1da177e4 | 1929 | } |