]>
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> | |
23 | #include <linux/kernel_stat.h> | |
24 | #include <linux/string.h> | |
25 | #include <linux/init.h> | |
1da177e4 LT |
26 | #include <linux/completion.h> |
27 | #include <linux/slab.h> | |
28 | #include <linux/swap.h> | |
29 | #include <linux/writeback.h> | |
faccbd4b | 30 | #include <linux/task_io_accounting_ops.h> |
c17bb495 | 31 | #include <linux/fault-inject.h> |
73c10101 | 32 | #include <linux/list_sort.h> |
e3c78ca5 | 33 | #include <linux/delay.h> |
aaf7c680 | 34 | #include <linux/ratelimit.h> |
6c954667 | 35 | #include <linux/pm_runtime.h> |
eea8f41c | 36 | #include <linux/blk-cgroup.h> |
18fbda91 | 37 | #include <linux/debugfs.h> |
30abb3a6 | 38 | #include <linux/bpf.h> |
55782138 LZ |
39 | |
40 | #define CREATE_TRACE_POINTS | |
41 | #include <trace/events/block.h> | |
1da177e4 | 42 | |
8324aa91 | 43 | #include "blk.h" |
43a5e4e2 | 44 | #include "blk-mq.h" |
bd166ef1 | 45 | #include "blk-mq-sched.h" |
bca6b067 | 46 | #include "blk-pm.h" |
c1c80384 | 47 | #include "blk-rq-qos.h" |
8324aa91 | 48 | |
18fbda91 OS |
49 | #ifdef CONFIG_DEBUG_FS |
50 | struct dentry *blk_debugfs_root; | |
51 | #endif | |
52 | ||
d07335e5 | 53 | EXPORT_TRACEPOINT_SYMBOL_GPL(block_bio_remap); |
b0da3f0d | 54 | EXPORT_TRACEPOINT_SYMBOL_GPL(block_rq_remap); |
0a82a8d1 | 55 | EXPORT_TRACEPOINT_SYMBOL_GPL(block_bio_complete); |
3291fa57 | 56 | EXPORT_TRACEPOINT_SYMBOL_GPL(block_split); |
cbae8d45 | 57 | EXPORT_TRACEPOINT_SYMBOL_GPL(block_unplug); |
0bfc2455 | 58 | |
a73f730d TH |
59 | DEFINE_IDA(blk_queue_ida); |
60 | ||
1da177e4 LT |
61 | /* |
62 | * For queue allocation | |
63 | */ | |
6728cb0e | 64 | struct kmem_cache *blk_requestq_cachep; |
1da177e4 | 65 | |
1da177e4 LT |
66 | /* |
67 | * Controlling structure to kblockd | |
68 | */ | |
ff856bad | 69 | static struct workqueue_struct *kblockd_workqueue; |
1da177e4 | 70 | |
8814ce8a BVA |
71 | /** |
72 | * blk_queue_flag_set - atomically set a queue flag | |
73 | * @flag: flag to be set | |
74 | * @q: request queue | |
75 | */ | |
76 | void blk_queue_flag_set(unsigned int flag, struct request_queue *q) | |
77 | { | |
57d74df9 | 78 | set_bit(flag, &q->queue_flags); |
8814ce8a BVA |
79 | } |
80 | EXPORT_SYMBOL(blk_queue_flag_set); | |
81 | ||
82 | /** | |
83 | * blk_queue_flag_clear - atomically clear a queue flag | |
84 | * @flag: flag to be cleared | |
85 | * @q: request queue | |
86 | */ | |
87 | void blk_queue_flag_clear(unsigned int flag, struct request_queue *q) | |
88 | { | |
57d74df9 | 89 | clear_bit(flag, &q->queue_flags); |
8814ce8a BVA |
90 | } |
91 | EXPORT_SYMBOL(blk_queue_flag_clear); | |
92 | ||
93 | /** | |
94 | * blk_queue_flag_test_and_set - atomically test and set a queue flag | |
95 | * @flag: flag to be set | |
96 | * @q: request queue | |
97 | * | |
98 | * Returns the previous value of @flag - 0 if the flag was not set and 1 if | |
99 | * the flag was already set. | |
100 | */ | |
101 | bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q) | |
102 | { | |
57d74df9 | 103 | return test_and_set_bit(flag, &q->queue_flags); |
8814ce8a BVA |
104 | } |
105 | EXPORT_SYMBOL_GPL(blk_queue_flag_test_and_set); | |
106 | ||
2a4aa30c | 107 | void blk_rq_init(struct request_queue *q, struct request *rq) |
1da177e4 | 108 | { |
1afb20f3 FT |
109 | memset(rq, 0, sizeof(*rq)); |
110 | ||
1da177e4 | 111 | INIT_LIST_HEAD(&rq->queuelist); |
63a71386 | 112 | rq->q = q; |
a2dec7b3 | 113 | rq->__sector = (sector_t) -1; |
2e662b65 JA |
114 | INIT_HLIST_NODE(&rq->hash); |
115 | RB_CLEAR_NODE(&rq->rb_node); | |
63a71386 | 116 | rq->tag = -1; |
bd166ef1 | 117 | rq->internal_tag = -1; |
522a7775 | 118 | rq->start_time_ns = ktime_get_ns(); |
09e099d4 | 119 | rq->part = NULL; |
1da177e4 | 120 | } |
2a4aa30c | 121 | EXPORT_SYMBOL(blk_rq_init); |
1da177e4 | 122 | |
2a842aca CH |
123 | static const struct { |
124 | int errno; | |
125 | const char *name; | |
126 | } blk_errors[] = { | |
127 | [BLK_STS_OK] = { 0, "" }, | |
128 | [BLK_STS_NOTSUPP] = { -EOPNOTSUPP, "operation not supported" }, | |
129 | [BLK_STS_TIMEOUT] = { -ETIMEDOUT, "timeout" }, | |
130 | [BLK_STS_NOSPC] = { -ENOSPC, "critical space allocation" }, | |
131 | [BLK_STS_TRANSPORT] = { -ENOLINK, "recoverable transport" }, | |
132 | [BLK_STS_TARGET] = { -EREMOTEIO, "critical target" }, | |
133 | [BLK_STS_NEXUS] = { -EBADE, "critical nexus" }, | |
134 | [BLK_STS_MEDIUM] = { -ENODATA, "critical medium" }, | |
135 | [BLK_STS_PROTECTION] = { -EILSEQ, "protection" }, | |
136 | [BLK_STS_RESOURCE] = { -ENOMEM, "kernel resource" }, | |
86ff7c2a | 137 | [BLK_STS_DEV_RESOURCE] = { -EBUSY, "device resource" }, |
03a07c92 | 138 | [BLK_STS_AGAIN] = { -EAGAIN, "nonblocking retry" }, |
2a842aca | 139 | |
4e4cbee9 CH |
140 | /* device mapper special case, should not leak out: */ |
141 | [BLK_STS_DM_REQUEUE] = { -EREMCHG, "dm internal retry" }, | |
142 | ||
2a842aca CH |
143 | /* everything else not covered above: */ |
144 | [BLK_STS_IOERR] = { -EIO, "I/O" }, | |
145 | }; | |
146 | ||
147 | blk_status_t errno_to_blk_status(int errno) | |
148 | { | |
149 | int i; | |
150 | ||
151 | for (i = 0; i < ARRAY_SIZE(blk_errors); i++) { | |
152 | if (blk_errors[i].errno == errno) | |
153 | return (__force blk_status_t)i; | |
154 | } | |
155 | ||
156 | return BLK_STS_IOERR; | |
157 | } | |
158 | EXPORT_SYMBOL_GPL(errno_to_blk_status); | |
159 | ||
160 | int blk_status_to_errno(blk_status_t status) | |
161 | { | |
162 | int idx = (__force int)status; | |
163 | ||
34bd9c1c | 164 | if (WARN_ON_ONCE(idx >= ARRAY_SIZE(blk_errors))) |
2a842aca CH |
165 | return -EIO; |
166 | return blk_errors[idx].errno; | |
167 | } | |
168 | EXPORT_SYMBOL_GPL(blk_status_to_errno); | |
169 | ||
170 | static void print_req_error(struct request *req, blk_status_t status) | |
171 | { | |
172 | int idx = (__force int)status; | |
173 | ||
34bd9c1c | 174 | if (WARN_ON_ONCE(idx >= ARRAY_SIZE(blk_errors))) |
2a842aca CH |
175 | return; |
176 | ||
2149da07 BS |
177 | printk_ratelimited(KERN_ERR "%s: %s error, dev %s, sector %llu flags %x\n", |
178 | __func__, blk_errors[idx].name, | |
179 | req->rq_disk ? req->rq_disk->disk_name : "?", | |
180 | (unsigned long long)blk_rq_pos(req), | |
181 | req->cmd_flags); | |
2a842aca CH |
182 | } |
183 | ||
5bb23a68 | 184 | static void req_bio_endio(struct request *rq, struct bio *bio, |
2a842aca | 185 | unsigned int nbytes, blk_status_t error) |
1da177e4 | 186 | { |
78d8e58a | 187 | if (error) |
4e4cbee9 | 188 | bio->bi_status = error; |
797e7dbb | 189 | |
e8064021 | 190 | if (unlikely(rq->rq_flags & RQF_QUIET)) |
b7c44ed9 | 191 | bio_set_flag(bio, BIO_QUIET); |
08bafc03 | 192 | |
f79ea416 | 193 | bio_advance(bio, nbytes); |
7ba1ba12 | 194 | |
143a87f4 | 195 | /* don't actually finish bio if it's part of flush sequence */ |
e8064021 | 196 | if (bio->bi_iter.bi_size == 0 && !(rq->rq_flags & RQF_FLUSH_SEQ)) |
4246a0b6 | 197 | bio_endio(bio); |
1da177e4 | 198 | } |
1da177e4 | 199 | |
1da177e4 LT |
200 | void blk_dump_rq_flags(struct request *rq, char *msg) |
201 | { | |
aebf526b CH |
202 | printk(KERN_INFO "%s: dev %s: flags=%llx\n", msg, |
203 | rq->rq_disk ? rq->rq_disk->disk_name : "?", | |
5953316d | 204 | (unsigned long long) rq->cmd_flags); |
1da177e4 | 205 | |
83096ebf TH |
206 | printk(KERN_INFO " sector %llu, nr/cnr %u/%u\n", |
207 | (unsigned long long)blk_rq_pos(rq), | |
208 | blk_rq_sectors(rq), blk_rq_cur_sectors(rq)); | |
b4f42e28 JA |
209 | printk(KERN_INFO " bio %p, biotail %p, len %u\n", |
210 | rq->bio, rq->biotail, blk_rq_bytes(rq)); | |
1da177e4 | 211 | } |
1da177e4 LT |
212 | EXPORT_SYMBOL(blk_dump_rq_flags); |
213 | ||
1da177e4 LT |
214 | /** |
215 | * blk_sync_queue - cancel any pending callbacks on a queue | |
216 | * @q: the queue | |
217 | * | |
218 | * Description: | |
219 | * The block layer may perform asynchronous callback activity | |
220 | * on a queue, such as calling the unplug function after a timeout. | |
221 | * A block device may call blk_sync_queue to ensure that any | |
222 | * such activity is cancelled, thus allowing it to release resources | |
59c51591 | 223 | * that the callbacks might use. The caller must already have made sure |
1da177e4 LT |
224 | * that its ->make_request_fn will not re-add plugging prior to calling |
225 | * this function. | |
226 | * | |
da527770 | 227 | * This function does not cancel any asynchronous activity arising |
da3dae54 | 228 | * out of elevator or throttling code. That would require elevator_exit() |
5efd6113 | 229 | * and blkcg_exit_queue() to be called with queue lock initialized. |
da527770 | 230 | * |
1da177e4 LT |
231 | */ |
232 | void blk_sync_queue(struct request_queue *q) | |
233 | { | |
70ed28b9 | 234 | del_timer_sync(&q->timeout); |
4e9b6f20 | 235 | cancel_work_sync(&q->timeout_work); |
1da177e4 LT |
236 | } |
237 | EXPORT_SYMBOL(blk_sync_queue); | |
238 | ||
c9254f2d | 239 | /** |
cd84a62e | 240 | * blk_set_pm_only - increment pm_only counter |
c9254f2d | 241 | * @q: request queue pointer |
c9254f2d | 242 | */ |
cd84a62e | 243 | void blk_set_pm_only(struct request_queue *q) |
c9254f2d | 244 | { |
cd84a62e | 245 | atomic_inc(&q->pm_only); |
c9254f2d | 246 | } |
cd84a62e | 247 | EXPORT_SYMBOL_GPL(blk_set_pm_only); |
c9254f2d | 248 | |
cd84a62e | 249 | void blk_clear_pm_only(struct request_queue *q) |
c9254f2d | 250 | { |
cd84a62e BVA |
251 | int pm_only; |
252 | ||
253 | pm_only = atomic_dec_return(&q->pm_only); | |
254 | WARN_ON_ONCE(pm_only < 0); | |
255 | if (pm_only == 0) | |
256 | wake_up_all(&q->mq_freeze_wq); | |
c9254f2d | 257 | } |
cd84a62e | 258 | EXPORT_SYMBOL_GPL(blk_clear_pm_only); |
c9254f2d | 259 | |
165125e1 | 260 | void blk_put_queue(struct request_queue *q) |
483f4afc AV |
261 | { |
262 | kobject_put(&q->kobj); | |
263 | } | |
d86e0e83 | 264 | EXPORT_SYMBOL(blk_put_queue); |
483f4afc | 265 | |
aed3ea94 JA |
266 | void blk_set_queue_dying(struct request_queue *q) |
267 | { | |
8814ce8a | 268 | blk_queue_flag_set(QUEUE_FLAG_DYING, q); |
aed3ea94 | 269 | |
d3cfb2a0 ML |
270 | /* |
271 | * When queue DYING flag is set, we need to block new req | |
272 | * entering queue, so we call blk_freeze_queue_start() to | |
273 | * prevent I/O from crossing blk_queue_enter(). | |
274 | */ | |
275 | blk_freeze_queue_start(q); | |
276 | ||
344e9ffc | 277 | if (queue_is_mq(q)) |
aed3ea94 | 278 | blk_mq_wake_waiters(q); |
055f6e18 ML |
279 | |
280 | /* Make blk_queue_enter() reexamine the DYING flag. */ | |
281 | wake_up_all(&q->mq_freeze_wq); | |
aed3ea94 JA |
282 | } |
283 | EXPORT_SYMBOL_GPL(blk_set_queue_dying); | |
284 | ||
4cf6324b BVA |
285 | /* Unconfigure the I/O scheduler and dissociate from the cgroup controller. */ |
286 | void blk_exit_queue(struct request_queue *q) | |
287 | { | |
288 | /* | |
289 | * Since the I/O scheduler exit code may access cgroup information, | |
290 | * perform I/O scheduler exit before disassociating from the block | |
291 | * cgroup controller. | |
292 | */ | |
293 | if (q->elevator) { | |
294 | ioc_clear_queue(q); | |
295 | elevator_exit(q, q->elevator); | |
296 | q->elevator = NULL; | |
297 | } | |
298 | ||
299 | /* | |
300 | * Remove all references to @q from the block cgroup controller before | |
301 | * restoring @q->queue_lock to avoid that restoring this pointer causes | |
302 | * e.g. blkcg_print_blkgs() to crash. | |
303 | */ | |
304 | blkcg_exit_queue(q); | |
305 | ||
306 | /* | |
307 | * Since the cgroup code may dereference the @q->backing_dev_info | |
308 | * pointer, only decrease its reference count after having removed the | |
309 | * association with the block cgroup controller. | |
310 | */ | |
311 | bdi_put(q->backing_dev_info); | |
312 | } | |
313 | ||
c9a929dd TH |
314 | /** |
315 | * blk_cleanup_queue - shutdown a request queue | |
316 | * @q: request queue to shutdown | |
317 | * | |
c246e80d BVA |
318 | * Mark @q DYING, drain all pending requests, mark @q DEAD, destroy and |
319 | * put it. All future requests will be failed immediately with -ENODEV. | |
c94a96ac | 320 | */ |
6728cb0e | 321 | void blk_cleanup_queue(struct request_queue *q) |
483f4afc | 322 | { |
3f3299d5 | 323 | /* mark @q DYING, no new request or merges will be allowed afterwards */ |
483f4afc | 324 | mutex_lock(&q->sysfs_lock); |
aed3ea94 | 325 | blk_set_queue_dying(q); |
6ecf23af | 326 | |
57d74df9 CH |
327 | blk_queue_flag_set(QUEUE_FLAG_NOMERGES, q); |
328 | blk_queue_flag_set(QUEUE_FLAG_NOXMERGES, q); | |
329 | blk_queue_flag_set(QUEUE_FLAG_DYING, q); | |
c9a929dd TH |
330 | mutex_unlock(&q->sysfs_lock); |
331 | ||
c246e80d BVA |
332 | /* |
333 | * Drain all requests queued before DYING marking. Set DEAD flag to | |
334 | * prevent that q->request_fn() gets invoked after draining finished. | |
335 | */ | |
3ef28e83 | 336 | blk_freeze_queue(q); |
c57cdf7a ML |
337 | |
338 | rq_qos_exit(q); | |
339 | ||
57d74df9 | 340 | blk_queue_flag_set(QUEUE_FLAG_DEAD, q); |
c9a929dd | 341 | |
5a48fc14 DW |
342 | /* for synchronous bio-based driver finish in-flight integrity i/o */ |
343 | blk_flush_integrity(); | |
344 | ||
c9a929dd | 345 | /* @q won't process any more request, flush async actions */ |
dc3b17cc | 346 | del_timer_sync(&q->backing_dev_info->laptop_mode_wb_timer); |
c9a929dd TH |
347 | blk_sync_queue(q); |
348 | ||
a063057d BVA |
349 | /* |
350 | * I/O scheduler exit is only safe after the sysfs scheduler attribute | |
351 | * has been removed. | |
352 | */ | |
353 | WARN_ON_ONCE(q->kobj.state_in_sysfs); | |
354 | ||
4cf6324b | 355 | blk_exit_queue(q); |
a063057d | 356 | |
344e9ffc | 357 | if (queue_is_mq(q)) |
c7e2d94b | 358 | blk_mq_exit_queue(q); |
a1ce35fa | 359 | |
3ef28e83 | 360 | percpu_ref_exit(&q->q_usage_counter); |
45a9c9d9 | 361 | |
c9a929dd | 362 | /* @q is and will stay empty, shutdown and put */ |
483f4afc AV |
363 | blk_put_queue(q); |
364 | } | |
1da177e4 LT |
365 | EXPORT_SYMBOL(blk_cleanup_queue); |
366 | ||
165125e1 | 367 | struct request_queue *blk_alloc_queue(gfp_t gfp_mask) |
1da177e4 | 368 | { |
6d469642 | 369 | return blk_alloc_queue_node(gfp_mask, NUMA_NO_NODE); |
1946089a CL |
370 | } |
371 | EXPORT_SYMBOL(blk_alloc_queue); | |
1da177e4 | 372 | |
3a0a5299 BVA |
373 | /** |
374 | * blk_queue_enter() - try to increase q->q_usage_counter | |
375 | * @q: request queue pointer | |
376 | * @flags: BLK_MQ_REQ_NOWAIT and/or BLK_MQ_REQ_PREEMPT | |
377 | */ | |
9a95e4ef | 378 | int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags) |
3ef28e83 | 379 | { |
cd84a62e | 380 | const bool pm = flags & BLK_MQ_REQ_PREEMPT; |
3a0a5299 | 381 | |
3ef28e83 | 382 | while (true) { |
3a0a5299 | 383 | bool success = false; |
3ef28e83 | 384 | |
818e0fa2 | 385 | rcu_read_lock(); |
3a0a5299 BVA |
386 | if (percpu_ref_tryget_live(&q->q_usage_counter)) { |
387 | /* | |
cd84a62e BVA |
388 | * The code that increments the pm_only counter is |
389 | * responsible for ensuring that that counter is | |
390 | * globally visible before the queue is unfrozen. | |
3a0a5299 | 391 | */ |
cd84a62e | 392 | if (pm || !blk_queue_pm_only(q)) { |
3a0a5299 BVA |
393 | success = true; |
394 | } else { | |
395 | percpu_ref_put(&q->q_usage_counter); | |
396 | } | |
397 | } | |
818e0fa2 | 398 | rcu_read_unlock(); |
3a0a5299 BVA |
399 | |
400 | if (success) | |
3ef28e83 DW |
401 | return 0; |
402 | ||
3a0a5299 | 403 | if (flags & BLK_MQ_REQ_NOWAIT) |
3ef28e83 DW |
404 | return -EBUSY; |
405 | ||
5ed61d3f | 406 | /* |
1671d522 | 407 | * read pair of barrier in blk_freeze_queue_start(), |
5ed61d3f | 408 | * we need to order reading __PERCPU_REF_DEAD flag of |
d3cfb2a0 ML |
409 | * .q_usage_counter and reading .mq_freeze_depth or |
410 | * queue dying flag, otherwise the following wait may | |
411 | * never return if the two reads are reordered. | |
5ed61d3f ML |
412 | */ |
413 | smp_rmb(); | |
414 | ||
1dc3039b | 415 | wait_event(q->mq_freeze_wq, |
7996a8b5 | 416 | (!q->mq_freeze_depth && |
0d25bd07 BVA |
417 | (pm || (blk_pm_request_resume(q), |
418 | !blk_queue_pm_only(q)))) || | |
1dc3039b | 419 | blk_queue_dying(q)); |
3ef28e83 DW |
420 | if (blk_queue_dying(q)) |
421 | return -ENODEV; | |
3ef28e83 DW |
422 | } |
423 | } | |
424 | ||
425 | void blk_queue_exit(struct request_queue *q) | |
426 | { | |
427 | percpu_ref_put(&q->q_usage_counter); | |
428 | } | |
429 | ||
430 | static void blk_queue_usage_counter_release(struct percpu_ref *ref) | |
431 | { | |
432 | struct request_queue *q = | |
433 | container_of(ref, struct request_queue, q_usage_counter); | |
434 | ||
435 | wake_up_all(&q->mq_freeze_wq); | |
436 | } | |
437 | ||
bca237a5 | 438 | static void blk_rq_timed_out_timer(struct timer_list *t) |
287922eb | 439 | { |
bca237a5 | 440 | struct request_queue *q = from_timer(q, t, timeout); |
287922eb CH |
441 | |
442 | kblockd_schedule_work(&q->timeout_work); | |
443 | } | |
444 | ||
2e3c18d0 TH |
445 | static void blk_timeout_work(struct work_struct *work) |
446 | { | |
447 | } | |
448 | ||
498f6650 BVA |
449 | /** |
450 | * blk_alloc_queue_node - allocate a request queue | |
451 | * @gfp_mask: memory allocation flags | |
452 | * @node_id: NUMA node to allocate memory from | |
498f6650 | 453 | */ |
6d469642 | 454 | struct request_queue *blk_alloc_queue_node(gfp_t gfp_mask, int node_id) |
1946089a | 455 | { |
165125e1 | 456 | struct request_queue *q; |
338aa96d | 457 | int ret; |
1946089a | 458 | |
8324aa91 | 459 | q = kmem_cache_alloc_node(blk_requestq_cachep, |
94f6030c | 460 | gfp_mask | __GFP_ZERO, node_id); |
1da177e4 LT |
461 | if (!q) |
462 | return NULL; | |
463 | ||
cbf62af3 CH |
464 | INIT_LIST_HEAD(&q->queue_head); |
465 | q->last_merge = NULL; | |
cbf62af3 | 466 | |
00380a40 | 467 | q->id = ida_simple_get(&blk_queue_ida, 0, 0, gfp_mask); |
a73f730d | 468 | if (q->id < 0) |
3d2936f4 | 469 | goto fail_q; |
a73f730d | 470 | |
338aa96d KO |
471 | ret = bioset_init(&q->bio_split, BIO_POOL_SIZE, 0, BIOSET_NEED_BVECS); |
472 | if (ret) | |
54efd50b KO |
473 | goto fail_id; |
474 | ||
d03f6cdc JK |
475 | q->backing_dev_info = bdi_alloc_node(gfp_mask, node_id); |
476 | if (!q->backing_dev_info) | |
477 | goto fail_split; | |
478 | ||
a83b576c JA |
479 | q->stats = blk_alloc_queue_stats(); |
480 | if (!q->stats) | |
481 | goto fail_stats; | |
482 | ||
b5420237 | 483 | q->backing_dev_info->ra_pages = VM_READAHEAD_PAGES; |
dc3b17cc JK |
484 | q->backing_dev_info->capabilities = BDI_CAP_CGROUP_WRITEBACK; |
485 | q->backing_dev_info->name = "block"; | |
5151412d | 486 | q->node = node_id; |
0989a025 | 487 | |
bca237a5 KC |
488 | timer_setup(&q->backing_dev_info->laptop_mode_wb_timer, |
489 | laptop_mode_timer_fn, 0); | |
490 | timer_setup(&q->timeout, blk_rq_timed_out_timer, 0); | |
2e3c18d0 | 491 | INIT_WORK(&q->timeout_work, blk_timeout_work); |
a612fddf | 492 | INIT_LIST_HEAD(&q->icq_list); |
4eef3049 | 493 | #ifdef CONFIG_BLK_CGROUP |
e8989fae | 494 | INIT_LIST_HEAD(&q->blkg_list); |
4eef3049 | 495 | #endif |
483f4afc | 496 | |
8324aa91 | 497 | kobject_init(&q->kobj, &blk_queue_ktype); |
1da177e4 | 498 | |
5acb3cc2 WL |
499 | #ifdef CONFIG_BLK_DEV_IO_TRACE |
500 | mutex_init(&q->blk_trace_mutex); | |
501 | #endif | |
483f4afc | 502 | mutex_init(&q->sysfs_lock); |
0d945c1f | 503 | spin_lock_init(&q->queue_lock); |
c94a96ac | 504 | |
320ae51f | 505 | init_waitqueue_head(&q->mq_freeze_wq); |
7996a8b5 | 506 | mutex_init(&q->mq_freeze_lock); |
320ae51f | 507 | |
3ef28e83 DW |
508 | /* |
509 | * Init percpu_ref in atomic mode so that it's faster to shutdown. | |
510 | * See blk_register_queue() for details. | |
511 | */ | |
512 | if (percpu_ref_init(&q->q_usage_counter, | |
513 | blk_queue_usage_counter_release, | |
514 | PERCPU_REF_INIT_ATOMIC, GFP_KERNEL)) | |
fff4996b | 515 | goto fail_bdi; |
f51b802c | 516 | |
3ef28e83 DW |
517 | if (blkcg_init_queue(q)) |
518 | goto fail_ref; | |
519 | ||
1da177e4 | 520 | return q; |
a73f730d | 521 | |
3ef28e83 DW |
522 | fail_ref: |
523 | percpu_ref_exit(&q->q_usage_counter); | |
fff4996b | 524 | fail_bdi: |
a83b576c JA |
525 | blk_free_queue_stats(q->stats); |
526 | fail_stats: | |
d03f6cdc | 527 | bdi_put(q->backing_dev_info); |
54efd50b | 528 | fail_split: |
338aa96d | 529 | bioset_exit(&q->bio_split); |
a73f730d TH |
530 | fail_id: |
531 | ida_simple_remove(&blk_queue_ida, q->id); | |
532 | fail_q: | |
533 | kmem_cache_free(blk_requestq_cachep, q); | |
534 | return NULL; | |
1da177e4 | 535 | } |
1946089a | 536 | EXPORT_SYMBOL(blk_alloc_queue_node); |
1da177e4 | 537 | |
09ac46c4 | 538 | bool blk_get_queue(struct request_queue *q) |
1da177e4 | 539 | { |
3f3299d5 | 540 | if (likely(!blk_queue_dying(q))) { |
09ac46c4 TH |
541 | __blk_get_queue(q); |
542 | return true; | |
1da177e4 LT |
543 | } |
544 | ||
09ac46c4 | 545 | return false; |
1da177e4 | 546 | } |
d86e0e83 | 547 | EXPORT_SYMBOL(blk_get_queue); |
1da177e4 | 548 | |
a1ce35fa JA |
549 | /** |
550 | * blk_get_request - allocate a request | |
551 | * @q: request queue to allocate a request for | |
552 | * @op: operation (REQ_OP_*) and REQ_* flags, e.g. REQ_SYNC. | |
553 | * @flags: BLK_MQ_REQ_* flags, e.g. BLK_MQ_REQ_NOWAIT. | |
1da177e4 | 554 | */ |
a1ce35fa JA |
555 | struct request *blk_get_request(struct request_queue *q, unsigned int op, |
556 | blk_mq_req_flags_t flags) | |
1da177e4 | 557 | { |
a1ce35fa | 558 | struct request *req; |
1da177e4 | 559 | |
a1ce35fa JA |
560 | WARN_ON_ONCE(op & REQ_NOWAIT); |
561 | WARN_ON_ONCE(flags & ~(BLK_MQ_REQ_NOWAIT | BLK_MQ_REQ_PREEMPT)); | |
1da177e4 | 562 | |
a1ce35fa JA |
563 | req = blk_mq_alloc_request(q, op, flags); |
564 | if (!IS_ERR(req) && q->mq_ops->initialize_rq_fn) | |
565 | q->mq_ops->initialize_rq_fn(req); | |
1da177e4 | 566 | |
a1ce35fa | 567 | return req; |
1da177e4 | 568 | } |
a1ce35fa | 569 | EXPORT_SYMBOL(blk_get_request); |
1da177e4 | 570 | |
1da177e4 LT |
571 | void blk_put_request(struct request *req) |
572 | { | |
a1ce35fa | 573 | blk_mq_free_request(req); |
1da177e4 | 574 | } |
1da177e4 LT |
575 | EXPORT_SYMBOL(blk_put_request); |
576 | ||
320ae51f JA |
577 | bool bio_attempt_back_merge(struct request_queue *q, struct request *req, |
578 | struct bio *bio) | |
73c10101 | 579 | { |
1eff9d32 | 580 | const int ff = bio->bi_opf & REQ_FAILFAST_MASK; |
73c10101 | 581 | |
73c10101 JA |
582 | if (!ll_back_merge_fn(q, req, bio)) |
583 | return false; | |
584 | ||
8c1cf6bb | 585 | trace_block_bio_backmerge(q, req, bio); |
73c10101 JA |
586 | |
587 | if ((req->cmd_flags & REQ_FAILFAST_MASK) != ff) | |
588 | blk_rq_set_mixed_merge(req); | |
589 | ||
590 | req->biotail->bi_next = bio; | |
591 | req->biotail = bio; | |
4f024f37 | 592 | req->__data_len += bio->bi_iter.bi_size; |
73c10101 | 593 | |
320ae51f | 594 | blk_account_io_start(req, false); |
73c10101 JA |
595 | return true; |
596 | } | |
597 | ||
320ae51f JA |
598 | bool bio_attempt_front_merge(struct request_queue *q, struct request *req, |
599 | struct bio *bio) | |
73c10101 | 600 | { |
1eff9d32 | 601 | const int ff = bio->bi_opf & REQ_FAILFAST_MASK; |
73c10101 | 602 | |
73c10101 JA |
603 | if (!ll_front_merge_fn(q, req, bio)) |
604 | return false; | |
605 | ||
8c1cf6bb | 606 | trace_block_bio_frontmerge(q, req, bio); |
73c10101 JA |
607 | |
608 | if ((req->cmd_flags & REQ_FAILFAST_MASK) != ff) | |
609 | blk_rq_set_mixed_merge(req); | |
610 | ||
73c10101 JA |
611 | bio->bi_next = req->bio; |
612 | req->bio = bio; | |
613 | ||
4f024f37 KO |
614 | req->__sector = bio->bi_iter.bi_sector; |
615 | req->__data_len += bio->bi_iter.bi_size; | |
73c10101 | 616 | |
320ae51f | 617 | blk_account_io_start(req, false); |
73c10101 JA |
618 | return true; |
619 | } | |
620 | ||
1e739730 CH |
621 | bool bio_attempt_discard_merge(struct request_queue *q, struct request *req, |
622 | struct bio *bio) | |
623 | { | |
624 | unsigned short segments = blk_rq_nr_discard_segments(req); | |
625 | ||
626 | if (segments >= queue_max_discard_segments(q)) | |
627 | goto no_merge; | |
628 | if (blk_rq_sectors(req) + bio_sectors(bio) > | |
629 | blk_rq_get_max_sectors(req, blk_rq_pos(req))) | |
630 | goto no_merge; | |
631 | ||
632 | req->biotail->bi_next = bio; | |
633 | req->biotail = bio; | |
634 | req->__data_len += bio->bi_iter.bi_size; | |
1e739730 CH |
635 | req->nr_phys_segments = segments + 1; |
636 | ||
637 | blk_account_io_start(req, false); | |
638 | return true; | |
639 | no_merge: | |
640 | req_set_nomerge(q, req); | |
641 | return false; | |
642 | } | |
643 | ||
bd87b589 | 644 | /** |
320ae51f | 645 | * blk_attempt_plug_merge - try to merge with %current's plugged list |
bd87b589 TH |
646 | * @q: request_queue new bio is being queued at |
647 | * @bio: new bio being queued | |
ccc2600b RD |
648 | * @same_queue_rq: pointer to &struct request that gets filled in when |
649 | * another request associated with @q is found on the plug list | |
650 | * (optional, may be %NULL) | |
bd87b589 TH |
651 | * |
652 | * Determine whether @bio being queued on @q can be merged with a request | |
653 | * on %current's plugged list. Returns %true if merge was successful, | |
654 | * otherwise %false. | |
655 | * | |
07c2bd37 TH |
656 | * Plugging coalesces IOs from the same issuer for the same purpose without |
657 | * going through @q->queue_lock. As such it's more of an issuing mechanism | |
658 | * than scheduling, and the request, while may have elvpriv data, is not | |
659 | * added on the elevator at this point. In addition, we don't have | |
660 | * reliable access to the elevator outside queue lock. Only check basic | |
661 | * merging parameters without querying the elevator. | |
da41a589 RE |
662 | * |
663 | * Caller must ensure !blk_queue_nomerges(q) beforehand. | |
73c10101 | 664 | */ |
320ae51f | 665 | bool blk_attempt_plug_merge(struct request_queue *q, struct bio *bio, |
5b3f341f | 666 | struct request **same_queue_rq) |
73c10101 JA |
667 | { |
668 | struct blk_plug *plug; | |
669 | struct request *rq; | |
92f399c7 | 670 | struct list_head *plug_list; |
73c10101 | 671 | |
bd87b589 | 672 | plug = current->plug; |
73c10101 | 673 | if (!plug) |
34fe7c05 | 674 | return false; |
73c10101 | 675 | |
a1ce35fa | 676 | plug_list = &plug->mq_list; |
92f399c7 SL |
677 | |
678 | list_for_each_entry_reverse(rq, plug_list, queuelist) { | |
34fe7c05 | 679 | bool merged = false; |
73c10101 | 680 | |
5f0ed774 | 681 | if (rq->q == q && same_queue_rq) { |
5b3f341f SL |
682 | /* |
683 | * Only blk-mq multiple hardware queues case checks the | |
684 | * rq in the same queue, there should be only one such | |
685 | * rq in a queue | |
686 | **/ | |
5f0ed774 | 687 | *same_queue_rq = rq; |
5b3f341f | 688 | } |
56ebdaf2 | 689 | |
07c2bd37 | 690 | if (rq->q != q || !blk_rq_merge_ok(rq, bio)) |
73c10101 JA |
691 | continue; |
692 | ||
34fe7c05 CH |
693 | switch (blk_try_merge(rq, bio)) { |
694 | case ELEVATOR_BACK_MERGE: | |
695 | merged = bio_attempt_back_merge(q, rq, bio); | |
696 | break; | |
697 | case ELEVATOR_FRONT_MERGE: | |
698 | merged = bio_attempt_front_merge(q, rq, bio); | |
699 | break; | |
1e739730 CH |
700 | case ELEVATOR_DISCARD_MERGE: |
701 | merged = bio_attempt_discard_merge(q, rq, bio); | |
702 | break; | |
34fe7c05 CH |
703 | default: |
704 | break; | |
73c10101 | 705 | } |
34fe7c05 CH |
706 | |
707 | if (merged) | |
708 | return true; | |
73c10101 | 709 | } |
34fe7c05 CH |
710 | |
711 | return false; | |
73c10101 JA |
712 | } |
713 | ||
da8d7f07 | 714 | void blk_init_request_from_bio(struct request *req, struct bio *bio) |
52d9e675 | 715 | { |
1eff9d32 | 716 | if (bio->bi_opf & REQ_RAHEAD) |
a82afdfc | 717 | req->cmd_flags |= REQ_FAILFAST_MASK; |
b31dc66a | 718 | |
4f024f37 | 719 | req->__sector = bio->bi_iter.bi_sector; |
20578bdf | 720 | req->ioprio = bio_prio(bio); |
cb6934f8 | 721 | req->write_hint = bio->bi_write_hint; |
bc1c56fd | 722 | blk_rq_bio_prep(req->q, req, bio); |
52d9e675 | 723 | } |
da8d7f07 | 724 | EXPORT_SYMBOL_GPL(blk_init_request_from_bio); |
52d9e675 | 725 | |
52c5e62d | 726 | static void handle_bad_sector(struct bio *bio, sector_t maxsector) |
1da177e4 LT |
727 | { |
728 | char b[BDEVNAME_SIZE]; | |
729 | ||
730 | printk(KERN_INFO "attempt to access beyond end of device\n"); | |
6296b960 | 731 | printk(KERN_INFO "%s: rw=%d, want=%Lu, limit=%Lu\n", |
74d46992 | 732 | bio_devname(bio, b), bio->bi_opf, |
f73a1c7d | 733 | (unsigned long long)bio_end_sector(bio), |
52c5e62d | 734 | (long long)maxsector); |
1da177e4 LT |
735 | } |
736 | ||
c17bb495 AM |
737 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
738 | ||
739 | static DECLARE_FAULT_ATTR(fail_make_request); | |
740 | ||
741 | static int __init setup_fail_make_request(char *str) | |
742 | { | |
743 | return setup_fault_attr(&fail_make_request, str); | |
744 | } | |
745 | __setup("fail_make_request=", setup_fail_make_request); | |
746 | ||
b2c9cd37 | 747 | static bool should_fail_request(struct hd_struct *part, unsigned int bytes) |
c17bb495 | 748 | { |
b2c9cd37 | 749 | return part->make_it_fail && should_fail(&fail_make_request, bytes); |
c17bb495 AM |
750 | } |
751 | ||
752 | static int __init fail_make_request_debugfs(void) | |
753 | { | |
dd48c085 AM |
754 | struct dentry *dir = fault_create_debugfs_attr("fail_make_request", |
755 | NULL, &fail_make_request); | |
756 | ||
21f9fcd8 | 757 | return PTR_ERR_OR_ZERO(dir); |
c17bb495 AM |
758 | } |
759 | ||
760 | late_initcall(fail_make_request_debugfs); | |
761 | ||
762 | #else /* CONFIG_FAIL_MAKE_REQUEST */ | |
763 | ||
b2c9cd37 AM |
764 | static inline bool should_fail_request(struct hd_struct *part, |
765 | unsigned int bytes) | |
c17bb495 | 766 | { |
b2c9cd37 | 767 | return false; |
c17bb495 AM |
768 | } |
769 | ||
770 | #endif /* CONFIG_FAIL_MAKE_REQUEST */ | |
771 | ||
721c7fc7 ID |
772 | static inline bool bio_check_ro(struct bio *bio, struct hd_struct *part) |
773 | { | |
b089cfd9 JA |
774 | const int op = bio_op(bio); |
775 | ||
8b2ded1c | 776 | if (part->policy && op_is_write(op)) { |
721c7fc7 ID |
777 | char b[BDEVNAME_SIZE]; |
778 | ||
8b2ded1c MP |
779 | if (op_is_flush(bio->bi_opf) && !bio_sectors(bio)) |
780 | return false; | |
781 | ||
a32e236e | 782 | WARN_ONCE(1, |
721c7fc7 ID |
783 | "generic_make_request: Trying to write " |
784 | "to read-only block-device %s (partno %d)\n", | |
785 | bio_devname(bio, b), part->partno); | |
a32e236e LT |
786 | /* Older lvm-tools actually trigger this */ |
787 | return false; | |
721c7fc7 ID |
788 | } |
789 | ||
790 | return false; | |
791 | } | |
792 | ||
30abb3a6 HM |
793 | static noinline int should_fail_bio(struct bio *bio) |
794 | { | |
795 | if (should_fail_request(&bio->bi_disk->part0, bio->bi_iter.bi_size)) | |
796 | return -EIO; | |
797 | return 0; | |
798 | } | |
799 | ALLOW_ERROR_INJECTION(should_fail_bio, ERRNO); | |
800 | ||
52c5e62d CH |
801 | /* |
802 | * Check whether this bio extends beyond the end of the device or partition. | |
803 | * This may well happen - the kernel calls bread() without checking the size of | |
804 | * the device, e.g., when mounting a file system. | |
805 | */ | |
806 | static inline int bio_check_eod(struct bio *bio, sector_t maxsector) | |
807 | { | |
808 | unsigned int nr_sectors = bio_sectors(bio); | |
809 | ||
810 | if (nr_sectors && maxsector && | |
811 | (nr_sectors > maxsector || | |
812 | bio->bi_iter.bi_sector > maxsector - nr_sectors)) { | |
813 | handle_bad_sector(bio, maxsector); | |
814 | return -EIO; | |
815 | } | |
816 | return 0; | |
817 | } | |
818 | ||
74d46992 CH |
819 | /* |
820 | * Remap block n of partition p to block n+start(p) of the disk. | |
821 | */ | |
822 | static inline int blk_partition_remap(struct bio *bio) | |
823 | { | |
824 | struct hd_struct *p; | |
52c5e62d | 825 | int ret = -EIO; |
74d46992 | 826 | |
721c7fc7 ID |
827 | rcu_read_lock(); |
828 | p = __disk_get_part(bio->bi_disk, bio->bi_partno); | |
52c5e62d CH |
829 | if (unlikely(!p)) |
830 | goto out; | |
831 | if (unlikely(should_fail_request(p, bio->bi_iter.bi_size))) | |
832 | goto out; | |
833 | if (unlikely(bio_check_ro(bio, p))) | |
721c7fc7 | 834 | goto out; |
721c7fc7 | 835 | |
74d46992 CH |
836 | /* |
837 | * Zone reset does not include bi_size so bio_sectors() is always 0. | |
838 | * Include a test for the reset op code and perform the remap if needed. | |
839 | */ | |
52c5e62d CH |
840 | if (bio_sectors(bio) || bio_op(bio) == REQ_OP_ZONE_RESET) { |
841 | if (bio_check_eod(bio, part_nr_sects_read(p))) | |
842 | goto out; | |
843 | bio->bi_iter.bi_sector += p->start_sect; | |
52c5e62d CH |
844 | trace_block_bio_remap(bio->bi_disk->queue, bio, part_devt(p), |
845 | bio->bi_iter.bi_sector - p->start_sect); | |
846 | } | |
c04fa44b | 847 | bio->bi_partno = 0; |
52c5e62d | 848 | ret = 0; |
721c7fc7 ID |
849 | out: |
850 | rcu_read_unlock(); | |
74d46992 CH |
851 | return ret; |
852 | } | |
853 | ||
27a84d54 CH |
854 | static noinline_for_stack bool |
855 | generic_make_request_checks(struct bio *bio) | |
1da177e4 | 856 | { |
165125e1 | 857 | struct request_queue *q; |
5a7bbad2 | 858 | int nr_sectors = bio_sectors(bio); |
4e4cbee9 | 859 | blk_status_t status = BLK_STS_IOERR; |
5a7bbad2 | 860 | char b[BDEVNAME_SIZE]; |
1da177e4 LT |
861 | |
862 | might_sleep(); | |
1da177e4 | 863 | |
74d46992 | 864 | q = bio->bi_disk->queue; |
5a7bbad2 CH |
865 | if (unlikely(!q)) { |
866 | printk(KERN_ERR | |
867 | "generic_make_request: Trying to access " | |
868 | "nonexistent block-device %s (%Lu)\n", | |
74d46992 | 869 | bio_devname(bio, b), (long long)bio->bi_iter.bi_sector); |
5a7bbad2 CH |
870 | goto end_io; |
871 | } | |
c17bb495 | 872 | |
03a07c92 GR |
873 | /* |
874 | * For a REQ_NOWAIT based request, return -EOPNOTSUPP | |
875 | * if queue is not a request based queue. | |
876 | */ | |
344e9ffc | 877 | if ((bio->bi_opf & REQ_NOWAIT) && !queue_is_mq(q)) |
03a07c92 GR |
878 | goto not_supported; |
879 | ||
30abb3a6 | 880 | if (should_fail_bio(bio)) |
5a7bbad2 | 881 | goto end_io; |
2056a782 | 882 | |
52c5e62d CH |
883 | if (bio->bi_partno) { |
884 | if (unlikely(blk_partition_remap(bio))) | |
721c7fc7 ID |
885 | goto end_io; |
886 | } else { | |
52c5e62d CH |
887 | if (unlikely(bio_check_ro(bio, &bio->bi_disk->part0))) |
888 | goto end_io; | |
889 | if (unlikely(bio_check_eod(bio, get_capacity(bio->bi_disk)))) | |
721c7fc7 ID |
890 | goto end_io; |
891 | } | |
2056a782 | 892 | |
5a7bbad2 CH |
893 | /* |
894 | * Filter flush bio's early so that make_request based | |
895 | * drivers without flush support don't have to worry | |
896 | * about them. | |
897 | */ | |
f3a8ab7d | 898 | if (op_is_flush(bio->bi_opf) && |
c888a8f9 | 899 | !test_bit(QUEUE_FLAG_WC, &q->queue_flags)) { |
1eff9d32 | 900 | bio->bi_opf &= ~(REQ_PREFLUSH | REQ_FUA); |
5a7bbad2 | 901 | if (!nr_sectors) { |
4e4cbee9 | 902 | status = BLK_STS_OK; |
51fd77bd JA |
903 | goto end_io; |
904 | } | |
5a7bbad2 | 905 | } |
5ddfe969 | 906 | |
d04c406f CH |
907 | if (!test_bit(QUEUE_FLAG_POLL, &q->queue_flags)) |
908 | bio->bi_opf &= ~REQ_HIPRI; | |
909 | ||
288dab8a CH |
910 | switch (bio_op(bio)) { |
911 | case REQ_OP_DISCARD: | |
912 | if (!blk_queue_discard(q)) | |
913 | goto not_supported; | |
914 | break; | |
915 | case REQ_OP_SECURE_ERASE: | |
916 | if (!blk_queue_secure_erase(q)) | |
917 | goto not_supported; | |
918 | break; | |
919 | case REQ_OP_WRITE_SAME: | |
74d46992 | 920 | if (!q->limits.max_write_same_sectors) |
288dab8a | 921 | goto not_supported; |
58886785 | 922 | break; |
2d253440 | 923 | case REQ_OP_ZONE_RESET: |
74d46992 | 924 | if (!blk_queue_is_zoned(q)) |
2d253440 | 925 | goto not_supported; |
288dab8a | 926 | break; |
a6f0788e | 927 | case REQ_OP_WRITE_ZEROES: |
74d46992 | 928 | if (!q->limits.max_write_zeroes_sectors) |
a6f0788e CK |
929 | goto not_supported; |
930 | break; | |
288dab8a CH |
931 | default: |
932 | break; | |
5a7bbad2 | 933 | } |
01edede4 | 934 | |
7f4b35d1 TH |
935 | /* |
936 | * Various block parts want %current->io_context and lazy ioc | |
937 | * allocation ends up trading a lot of pain for a small amount of | |
938 | * memory. Just allocate it upfront. This may fail and block | |
939 | * layer knows how to live with it. | |
940 | */ | |
941 | create_io_context(GFP_ATOMIC, q->node); | |
942 | ||
ae118896 TH |
943 | if (!blkcg_bio_issue_check(q, bio)) |
944 | return false; | |
27a84d54 | 945 | |
fbbaf700 N |
946 | if (!bio_flagged(bio, BIO_TRACE_COMPLETION)) { |
947 | trace_block_bio_queue(q, bio); | |
948 | /* Now that enqueuing has been traced, we need to trace | |
949 | * completion as well. | |
950 | */ | |
951 | bio_set_flag(bio, BIO_TRACE_COMPLETION); | |
952 | } | |
27a84d54 | 953 | return true; |
a7384677 | 954 | |
288dab8a | 955 | not_supported: |
4e4cbee9 | 956 | status = BLK_STS_NOTSUPP; |
a7384677 | 957 | end_io: |
4e4cbee9 | 958 | bio->bi_status = status; |
4246a0b6 | 959 | bio_endio(bio); |
27a84d54 | 960 | return false; |
1da177e4 LT |
961 | } |
962 | ||
27a84d54 CH |
963 | /** |
964 | * generic_make_request - hand a buffer to its device driver for I/O | |
965 | * @bio: The bio describing the location in memory and on the device. | |
966 | * | |
967 | * generic_make_request() is used to make I/O requests of block | |
968 | * devices. It is passed a &struct bio, which describes the I/O that needs | |
969 | * to be done. | |
970 | * | |
971 | * generic_make_request() does not return any status. The | |
972 | * success/failure status of the request, along with notification of | |
973 | * completion, is delivered asynchronously through the bio->bi_end_io | |
974 | * function described (one day) else where. | |
975 | * | |
976 | * The caller of generic_make_request must make sure that bi_io_vec | |
977 | * are set to describe the memory buffer, and that bi_dev and bi_sector are | |
978 | * set to describe the device address, and the | |
979 | * bi_end_io and optionally bi_private are set to describe how | |
980 | * completion notification should be signaled. | |
981 | * | |
982 | * generic_make_request and the drivers it calls may use bi_next if this | |
983 | * bio happens to be merged with someone else, and may resubmit the bio to | |
984 | * a lower device by calling into generic_make_request recursively, which | |
985 | * means the bio should NOT be touched after the call to ->make_request_fn. | |
d89d8796 | 986 | */ |
dece1635 | 987 | blk_qc_t generic_make_request(struct bio *bio) |
d89d8796 | 988 | { |
f5fe1b51 N |
989 | /* |
990 | * bio_list_on_stack[0] contains bios submitted by the current | |
991 | * make_request_fn. | |
992 | * bio_list_on_stack[1] contains bios that were submitted before | |
993 | * the current make_request_fn, but that haven't been processed | |
994 | * yet. | |
995 | */ | |
996 | struct bio_list bio_list_on_stack[2]; | |
37f9579f BVA |
997 | blk_mq_req_flags_t flags = 0; |
998 | struct request_queue *q = bio->bi_disk->queue; | |
dece1635 | 999 | blk_qc_t ret = BLK_QC_T_NONE; |
bddd87c7 | 1000 | |
37f9579f BVA |
1001 | if (bio->bi_opf & REQ_NOWAIT) |
1002 | flags = BLK_MQ_REQ_NOWAIT; | |
cd4a4ae4 JA |
1003 | if (bio_flagged(bio, BIO_QUEUE_ENTERED)) |
1004 | blk_queue_enter_live(q); | |
1005 | else if (blk_queue_enter(q, flags) < 0) { | |
37f9579f BVA |
1006 | if (!blk_queue_dying(q) && (bio->bi_opf & REQ_NOWAIT)) |
1007 | bio_wouldblock_error(bio); | |
1008 | else | |
1009 | bio_io_error(bio); | |
1010 | return ret; | |
1011 | } | |
1012 | ||
27a84d54 | 1013 | if (!generic_make_request_checks(bio)) |
dece1635 | 1014 | goto out; |
27a84d54 CH |
1015 | |
1016 | /* | |
1017 | * We only want one ->make_request_fn to be active at a time, else | |
1018 | * stack usage with stacked devices could be a problem. So use | |
1019 | * current->bio_list to keep a list of requests submited by a | |
1020 | * make_request_fn function. current->bio_list is also used as a | |
1021 | * flag to say if generic_make_request is currently active in this | |
1022 | * task or not. If it is NULL, then no make_request is active. If | |
1023 | * it is non-NULL, then a make_request is active, and new requests | |
1024 | * should be added at the tail | |
1025 | */ | |
bddd87c7 | 1026 | if (current->bio_list) { |
f5fe1b51 | 1027 | bio_list_add(¤t->bio_list[0], bio); |
dece1635 | 1028 | goto out; |
d89d8796 | 1029 | } |
27a84d54 | 1030 | |
d89d8796 NB |
1031 | /* following loop may be a bit non-obvious, and so deserves some |
1032 | * explanation. | |
1033 | * Before entering the loop, bio->bi_next is NULL (as all callers | |
1034 | * ensure that) so we have a list with a single bio. | |
1035 | * We pretend that we have just taken it off a longer list, so | |
bddd87c7 AM |
1036 | * we assign bio_list to a pointer to the bio_list_on_stack, |
1037 | * thus initialising the bio_list of new bios to be | |
27a84d54 | 1038 | * added. ->make_request() may indeed add some more bios |
d89d8796 NB |
1039 | * through a recursive call to generic_make_request. If it |
1040 | * did, we find a non-NULL value in bio_list and re-enter the loop | |
1041 | * from the top. In this case we really did just take the bio | |
bddd87c7 | 1042 | * of the top of the list (no pretending) and so remove it from |
27a84d54 | 1043 | * bio_list, and call into ->make_request() again. |
d89d8796 NB |
1044 | */ |
1045 | BUG_ON(bio->bi_next); | |
f5fe1b51 N |
1046 | bio_list_init(&bio_list_on_stack[0]); |
1047 | current->bio_list = bio_list_on_stack; | |
d89d8796 | 1048 | do { |
37f9579f BVA |
1049 | bool enter_succeeded = true; |
1050 | ||
1051 | if (unlikely(q != bio->bi_disk->queue)) { | |
1052 | if (q) | |
1053 | blk_queue_exit(q); | |
1054 | q = bio->bi_disk->queue; | |
1055 | flags = 0; | |
1056 | if (bio->bi_opf & REQ_NOWAIT) | |
1057 | flags = BLK_MQ_REQ_NOWAIT; | |
1058 | if (blk_queue_enter(q, flags) < 0) { | |
1059 | enter_succeeded = false; | |
1060 | q = NULL; | |
1061 | } | |
1062 | } | |
27a84d54 | 1063 | |
37f9579f | 1064 | if (enter_succeeded) { |
79bd9959 N |
1065 | struct bio_list lower, same; |
1066 | ||
1067 | /* Create a fresh bio_list for all subordinate requests */ | |
f5fe1b51 N |
1068 | bio_list_on_stack[1] = bio_list_on_stack[0]; |
1069 | bio_list_init(&bio_list_on_stack[0]); | |
dece1635 | 1070 | ret = q->make_request_fn(q, bio); |
3ef28e83 | 1071 | |
79bd9959 N |
1072 | /* sort new bios into those for a lower level |
1073 | * and those for the same level | |
1074 | */ | |
1075 | bio_list_init(&lower); | |
1076 | bio_list_init(&same); | |
f5fe1b51 | 1077 | while ((bio = bio_list_pop(&bio_list_on_stack[0])) != NULL) |
74d46992 | 1078 | if (q == bio->bi_disk->queue) |
79bd9959 N |
1079 | bio_list_add(&same, bio); |
1080 | else | |
1081 | bio_list_add(&lower, bio); | |
1082 | /* now assemble so we handle the lowest level first */ | |
f5fe1b51 N |
1083 | bio_list_merge(&bio_list_on_stack[0], &lower); |
1084 | bio_list_merge(&bio_list_on_stack[0], &same); | |
1085 | bio_list_merge(&bio_list_on_stack[0], &bio_list_on_stack[1]); | |
3ef28e83 | 1086 | } else { |
03a07c92 GR |
1087 | if (unlikely(!blk_queue_dying(q) && |
1088 | (bio->bi_opf & REQ_NOWAIT))) | |
1089 | bio_wouldblock_error(bio); | |
1090 | else | |
1091 | bio_io_error(bio); | |
3ef28e83 | 1092 | } |
f5fe1b51 | 1093 | bio = bio_list_pop(&bio_list_on_stack[0]); |
d89d8796 | 1094 | } while (bio); |
bddd87c7 | 1095 | current->bio_list = NULL; /* deactivate */ |
dece1635 JA |
1096 | |
1097 | out: | |
37f9579f BVA |
1098 | if (q) |
1099 | blk_queue_exit(q); | |
dece1635 | 1100 | return ret; |
d89d8796 | 1101 | } |
1da177e4 LT |
1102 | EXPORT_SYMBOL(generic_make_request); |
1103 | ||
f421e1d9 CH |
1104 | /** |
1105 | * direct_make_request - hand a buffer directly to its device driver for I/O | |
1106 | * @bio: The bio describing the location in memory and on the device. | |
1107 | * | |
1108 | * This function behaves like generic_make_request(), but does not protect | |
1109 | * against recursion. Must only be used if the called driver is known | |
1110 | * to not call generic_make_request (or direct_make_request) again from | |
1111 | * its make_request function. (Calling direct_make_request again from | |
1112 | * a workqueue is perfectly fine as that doesn't recurse). | |
1113 | */ | |
1114 | blk_qc_t direct_make_request(struct bio *bio) | |
1115 | { | |
1116 | struct request_queue *q = bio->bi_disk->queue; | |
1117 | bool nowait = bio->bi_opf & REQ_NOWAIT; | |
1118 | blk_qc_t ret; | |
1119 | ||
1120 | if (!generic_make_request_checks(bio)) | |
1121 | return BLK_QC_T_NONE; | |
1122 | ||
3a0a5299 | 1123 | if (unlikely(blk_queue_enter(q, nowait ? BLK_MQ_REQ_NOWAIT : 0))) { |
f421e1d9 CH |
1124 | if (nowait && !blk_queue_dying(q)) |
1125 | bio->bi_status = BLK_STS_AGAIN; | |
1126 | else | |
1127 | bio->bi_status = BLK_STS_IOERR; | |
1128 | bio_endio(bio); | |
1129 | return BLK_QC_T_NONE; | |
1130 | } | |
1131 | ||
1132 | ret = q->make_request_fn(q, bio); | |
1133 | blk_queue_exit(q); | |
1134 | return ret; | |
1135 | } | |
1136 | EXPORT_SYMBOL_GPL(direct_make_request); | |
1137 | ||
1da177e4 | 1138 | /** |
710027a4 | 1139 | * submit_bio - submit a bio to the block device layer for I/O |
1da177e4 LT |
1140 | * @bio: The &struct bio which describes the I/O |
1141 | * | |
1142 | * submit_bio() is very similar in purpose to generic_make_request(), and | |
1143 | * uses that function to do most of the work. Both are fairly rough | |
710027a4 | 1144 | * interfaces; @bio must be presetup and ready for I/O. |
1da177e4 LT |
1145 | * |
1146 | */ | |
4e49ea4a | 1147 | blk_qc_t submit_bio(struct bio *bio) |
1da177e4 | 1148 | { |
bf2de6f5 JA |
1149 | /* |
1150 | * If it's a regular read/write or a barrier with data attached, | |
1151 | * go through the normal accounting stuff before submission. | |
1152 | */ | |
e2a60da7 | 1153 | if (bio_has_data(bio)) { |
4363ac7c MP |
1154 | unsigned int count; |
1155 | ||
95fe6c1a | 1156 | if (unlikely(bio_op(bio) == REQ_OP_WRITE_SAME)) |
7c5a0dcf | 1157 | count = queue_logical_block_size(bio->bi_disk->queue) >> 9; |
4363ac7c MP |
1158 | else |
1159 | count = bio_sectors(bio); | |
1160 | ||
a8ebb056 | 1161 | if (op_is_write(bio_op(bio))) { |
bf2de6f5 JA |
1162 | count_vm_events(PGPGOUT, count); |
1163 | } else { | |
4f024f37 | 1164 | task_io_account_read(bio->bi_iter.bi_size); |
bf2de6f5 JA |
1165 | count_vm_events(PGPGIN, count); |
1166 | } | |
1167 | ||
1168 | if (unlikely(block_dump)) { | |
1169 | char b[BDEVNAME_SIZE]; | |
8dcbdc74 | 1170 | printk(KERN_DEBUG "%s(%d): %s block %Lu on %s (%u sectors)\n", |
ba25f9dc | 1171 | current->comm, task_pid_nr(current), |
a8ebb056 | 1172 | op_is_write(bio_op(bio)) ? "WRITE" : "READ", |
4f024f37 | 1173 | (unsigned long long)bio->bi_iter.bi_sector, |
74d46992 | 1174 | bio_devname(bio, b), count); |
bf2de6f5 | 1175 | } |
1da177e4 LT |
1176 | } |
1177 | ||
dece1635 | 1178 | return generic_make_request(bio); |
1da177e4 | 1179 | } |
1da177e4 LT |
1180 | EXPORT_SYMBOL(submit_bio); |
1181 | ||
82124d60 | 1182 | /** |
bf4e6b4e HR |
1183 | * blk_cloned_rq_check_limits - Helper function to check a cloned request |
1184 | * for new the queue limits | |
82124d60 KU |
1185 | * @q: the queue |
1186 | * @rq: the request being checked | |
1187 | * | |
1188 | * Description: | |
1189 | * @rq may have been made based on weaker limitations of upper-level queues | |
1190 | * in request stacking drivers, and it may violate the limitation of @q. | |
1191 | * Since the block layer and the underlying device driver trust @rq | |
1192 | * after it is inserted to @q, it should be checked against @q before | |
1193 | * the insertion using this generic function. | |
1194 | * | |
82124d60 | 1195 | * Request stacking drivers like request-based dm may change the queue |
bf4e6b4e HR |
1196 | * limits when retrying requests on other queues. Those requests need |
1197 | * to be checked against the new queue limits again during dispatch. | |
82124d60 | 1198 | */ |
bf4e6b4e HR |
1199 | static int blk_cloned_rq_check_limits(struct request_queue *q, |
1200 | struct request *rq) | |
82124d60 | 1201 | { |
8fe0d473 | 1202 | if (blk_rq_sectors(rq) > blk_queue_get_max_sectors(q, req_op(rq))) { |
82124d60 KU |
1203 | printk(KERN_ERR "%s: over max size limit.\n", __func__); |
1204 | return -EIO; | |
1205 | } | |
1206 | ||
1207 | /* | |
1208 | * queue's settings related to segment counting like q->bounce_pfn | |
1209 | * may differ from that of other stacking queues. | |
1210 | * Recalculate it to check the request correctly on this queue's | |
1211 | * limitation. | |
1212 | */ | |
1213 | blk_recalc_rq_segments(rq); | |
8a78362c | 1214 | if (rq->nr_phys_segments > queue_max_segments(q)) { |
82124d60 KU |
1215 | printk(KERN_ERR "%s: over max segments limit.\n", __func__); |
1216 | return -EIO; | |
1217 | } | |
1218 | ||
1219 | return 0; | |
1220 | } | |
82124d60 KU |
1221 | |
1222 | /** | |
1223 | * blk_insert_cloned_request - Helper for stacking drivers to submit a request | |
1224 | * @q: the queue to submit the request | |
1225 | * @rq: the request being queued | |
1226 | */ | |
2a842aca | 1227 | blk_status_t blk_insert_cloned_request(struct request_queue *q, struct request *rq) |
82124d60 | 1228 | { |
bf4e6b4e | 1229 | if (blk_cloned_rq_check_limits(q, rq)) |
2a842aca | 1230 | return BLK_STS_IOERR; |
82124d60 | 1231 | |
b2c9cd37 AM |
1232 | if (rq->rq_disk && |
1233 | should_fail_request(&rq->rq_disk->part0, blk_rq_bytes(rq))) | |
2a842aca | 1234 | return BLK_STS_IOERR; |
82124d60 | 1235 | |
a1ce35fa JA |
1236 | if (blk_queue_io_stat(q)) |
1237 | blk_account_io_start(rq, true); | |
82124d60 KU |
1238 | |
1239 | /* | |
a1ce35fa JA |
1240 | * Since we have a scheduler attached on the top device, |
1241 | * bypass a potential scheduler on the bottom device for | |
1242 | * insert. | |
82124d60 | 1243 | */ |
fd9c40f6 | 1244 | return blk_mq_request_issue_directly(rq, true); |
82124d60 KU |
1245 | } |
1246 | EXPORT_SYMBOL_GPL(blk_insert_cloned_request); | |
1247 | ||
80a761fd TH |
1248 | /** |
1249 | * blk_rq_err_bytes - determine number of bytes till the next failure boundary | |
1250 | * @rq: request to examine | |
1251 | * | |
1252 | * Description: | |
1253 | * A request could be merge of IOs which require different failure | |
1254 | * handling. This function determines the number of bytes which | |
1255 | * can be failed from the beginning of the request without | |
1256 | * crossing into area which need to be retried further. | |
1257 | * | |
1258 | * Return: | |
1259 | * The number of bytes to fail. | |
80a761fd TH |
1260 | */ |
1261 | unsigned int blk_rq_err_bytes(const struct request *rq) | |
1262 | { | |
1263 | unsigned int ff = rq->cmd_flags & REQ_FAILFAST_MASK; | |
1264 | unsigned int bytes = 0; | |
1265 | struct bio *bio; | |
1266 | ||
e8064021 | 1267 | if (!(rq->rq_flags & RQF_MIXED_MERGE)) |
80a761fd TH |
1268 | return blk_rq_bytes(rq); |
1269 | ||
1270 | /* | |
1271 | * Currently the only 'mixing' which can happen is between | |
1272 | * different fastfail types. We can safely fail portions | |
1273 | * which have all the failfast bits that the first one has - | |
1274 | * the ones which are at least as eager to fail as the first | |
1275 | * one. | |
1276 | */ | |
1277 | for (bio = rq->bio; bio; bio = bio->bi_next) { | |
1eff9d32 | 1278 | if ((bio->bi_opf & ff) != ff) |
80a761fd | 1279 | break; |
4f024f37 | 1280 | bytes += bio->bi_iter.bi_size; |
80a761fd TH |
1281 | } |
1282 | ||
1283 | /* this could lead to infinite loop */ | |
1284 | BUG_ON(blk_rq_bytes(rq) && !bytes); | |
1285 | return bytes; | |
1286 | } | |
1287 | EXPORT_SYMBOL_GPL(blk_rq_err_bytes); | |
1288 | ||
320ae51f | 1289 | void blk_account_io_completion(struct request *req, unsigned int bytes) |
bc58ba94 | 1290 | { |
c2553b58 | 1291 | if (blk_do_io_stat(req)) { |
ddcf35d3 | 1292 | const int sgrp = op_stat_group(req_op(req)); |
bc58ba94 | 1293 | struct hd_struct *part; |
bc58ba94 | 1294 | |
112f158f | 1295 | part_stat_lock(); |
09e099d4 | 1296 | part = req->part; |
112f158f | 1297 | part_stat_add(part, sectors[sgrp], bytes >> 9); |
bc58ba94 JA |
1298 | part_stat_unlock(); |
1299 | } | |
1300 | } | |
1301 | ||
522a7775 | 1302 | void blk_account_io_done(struct request *req, u64 now) |
bc58ba94 | 1303 | { |
bc58ba94 | 1304 | /* |
dd4c133f TH |
1305 | * Account IO completion. flush_rq isn't accounted as a |
1306 | * normal IO on queueing nor completion. Accounting the | |
1307 | * containing request is enough. | |
bc58ba94 | 1308 | */ |
e8064021 | 1309 | if (blk_do_io_stat(req) && !(req->rq_flags & RQF_FLUSH_SEQ)) { |
ddcf35d3 | 1310 | const int sgrp = op_stat_group(req_op(req)); |
bc58ba94 | 1311 | struct hd_struct *part; |
bc58ba94 | 1312 | |
112f158f | 1313 | part_stat_lock(); |
09e099d4 | 1314 | part = req->part; |
bc58ba94 | 1315 | |
5b18b5a7 | 1316 | update_io_ticks(part, jiffies); |
112f158f MS |
1317 | part_stat_inc(part, ios[sgrp]); |
1318 | part_stat_add(part, nsecs[sgrp], now - req->start_time_ns); | |
5b18b5a7 | 1319 | part_stat_add(part, time_in_queue, nsecs_to_jiffies64(now - req->start_time_ns)); |
ddcf35d3 | 1320 | part_dec_in_flight(req->q, part, rq_data_dir(req)); |
bc58ba94 | 1321 | |
6c23a968 | 1322 | hd_struct_put(part); |
bc58ba94 JA |
1323 | part_stat_unlock(); |
1324 | } | |
1325 | } | |
1326 | ||
320ae51f JA |
1327 | void blk_account_io_start(struct request *rq, bool new_io) |
1328 | { | |
1329 | struct hd_struct *part; | |
1330 | int rw = rq_data_dir(rq); | |
320ae51f JA |
1331 | |
1332 | if (!blk_do_io_stat(rq)) | |
1333 | return; | |
1334 | ||
112f158f | 1335 | part_stat_lock(); |
320ae51f JA |
1336 | |
1337 | if (!new_io) { | |
1338 | part = rq->part; | |
112f158f | 1339 | part_stat_inc(part, merges[rw]); |
320ae51f JA |
1340 | } else { |
1341 | part = disk_map_sector_rcu(rq->rq_disk, blk_rq_pos(rq)); | |
1342 | if (!hd_struct_try_get(part)) { | |
1343 | /* | |
1344 | * The partition is already being removed, | |
1345 | * the request will be accounted on the disk only | |
1346 | * | |
1347 | * We take a reference on disk->part0 although that | |
1348 | * partition will never be deleted, so we can treat | |
1349 | * it as any other partition. | |
1350 | */ | |
1351 | part = &rq->rq_disk->part0; | |
1352 | hd_struct_get(part); | |
1353 | } | |
d62e26b3 | 1354 | part_inc_in_flight(rq->q, part, rw); |
320ae51f JA |
1355 | rq->part = part; |
1356 | } | |
1357 | ||
5b18b5a7 MP |
1358 | update_io_ticks(part, jiffies); |
1359 | ||
320ae51f JA |
1360 | part_stat_unlock(); |
1361 | } | |
1362 | ||
ef71de8b CH |
1363 | /* |
1364 | * Steal bios from a request and add them to a bio list. | |
1365 | * The request must not have been partially completed before. | |
1366 | */ | |
1367 | void blk_steal_bios(struct bio_list *list, struct request *rq) | |
1368 | { | |
1369 | if (rq->bio) { | |
1370 | if (list->tail) | |
1371 | list->tail->bi_next = rq->bio; | |
1372 | else | |
1373 | list->head = rq->bio; | |
1374 | list->tail = rq->biotail; | |
1375 | ||
1376 | rq->bio = NULL; | |
1377 | rq->biotail = NULL; | |
1378 | } | |
1379 | ||
1380 | rq->__data_len = 0; | |
1381 | } | |
1382 | EXPORT_SYMBOL_GPL(blk_steal_bios); | |
1383 | ||
3bcddeac | 1384 | /** |
2e60e022 | 1385 | * blk_update_request - Special helper function for request stacking drivers |
8ebf9756 | 1386 | * @req: the request being processed |
2a842aca | 1387 | * @error: block status code |
8ebf9756 | 1388 | * @nr_bytes: number of bytes to complete @req |
3bcddeac KU |
1389 | * |
1390 | * Description: | |
8ebf9756 RD |
1391 | * Ends I/O on a number of bytes attached to @req, but doesn't complete |
1392 | * the request structure even if @req doesn't have leftover. | |
1393 | * If @req has leftover, sets it up for the next range of segments. | |
2e60e022 TH |
1394 | * |
1395 | * This special helper function is only for request stacking drivers | |
1396 | * (e.g. request-based dm) so that they can handle partial completion. | |
1397 | * Actual device drivers should use blk_end_request instead. | |
1398 | * | |
1399 | * Passing the result of blk_rq_bytes() as @nr_bytes guarantees | |
1400 | * %false return from this function. | |
3bcddeac | 1401 | * |
1954e9a9 BVA |
1402 | * Note: |
1403 | * The RQF_SPECIAL_PAYLOAD flag is ignored on purpose in both | |
1404 | * blk_rq_bytes() and in blk_update_request(). | |
1405 | * | |
3bcddeac | 1406 | * Return: |
2e60e022 TH |
1407 | * %false - this request doesn't have any more data |
1408 | * %true - this request has more data | |
3bcddeac | 1409 | **/ |
2a842aca CH |
1410 | bool blk_update_request(struct request *req, blk_status_t error, |
1411 | unsigned int nr_bytes) | |
1da177e4 | 1412 | { |
f79ea416 | 1413 | int total_bytes; |
1da177e4 | 1414 | |
2a842aca | 1415 | trace_block_rq_complete(req, blk_status_to_errno(error), nr_bytes); |
4a0efdc9 | 1416 | |
2e60e022 TH |
1417 | if (!req->bio) |
1418 | return false; | |
1419 | ||
2a842aca CH |
1420 | if (unlikely(error && !blk_rq_is_passthrough(req) && |
1421 | !(req->rq_flags & RQF_QUIET))) | |
1422 | print_req_error(req, error); | |
1da177e4 | 1423 | |
bc58ba94 | 1424 | blk_account_io_completion(req, nr_bytes); |
d72d904a | 1425 | |
f79ea416 KO |
1426 | total_bytes = 0; |
1427 | while (req->bio) { | |
1428 | struct bio *bio = req->bio; | |
4f024f37 | 1429 | unsigned bio_bytes = min(bio->bi_iter.bi_size, nr_bytes); |
1da177e4 | 1430 | |
9c24c10a | 1431 | if (bio_bytes == bio->bi_iter.bi_size) |
1da177e4 | 1432 | req->bio = bio->bi_next; |
1da177e4 | 1433 | |
fbbaf700 N |
1434 | /* Completion has already been traced */ |
1435 | bio_clear_flag(bio, BIO_TRACE_COMPLETION); | |
f79ea416 | 1436 | req_bio_endio(req, bio, bio_bytes, error); |
1da177e4 | 1437 | |
f79ea416 KO |
1438 | total_bytes += bio_bytes; |
1439 | nr_bytes -= bio_bytes; | |
1da177e4 | 1440 | |
f79ea416 KO |
1441 | if (!nr_bytes) |
1442 | break; | |
1da177e4 LT |
1443 | } |
1444 | ||
1445 | /* | |
1446 | * completely done | |
1447 | */ | |
2e60e022 TH |
1448 | if (!req->bio) { |
1449 | /* | |
1450 | * Reset counters so that the request stacking driver | |
1451 | * can find how many bytes remain in the request | |
1452 | * later. | |
1453 | */ | |
a2dec7b3 | 1454 | req->__data_len = 0; |
2e60e022 TH |
1455 | return false; |
1456 | } | |
1da177e4 | 1457 | |
a2dec7b3 | 1458 | req->__data_len -= total_bytes; |
2e46e8b2 TH |
1459 | |
1460 | /* update sector only for requests with clear definition of sector */ | |
57292b58 | 1461 | if (!blk_rq_is_passthrough(req)) |
a2dec7b3 | 1462 | req->__sector += total_bytes >> 9; |
2e46e8b2 | 1463 | |
80a761fd | 1464 | /* mixed attributes always follow the first bio */ |
e8064021 | 1465 | if (req->rq_flags & RQF_MIXED_MERGE) { |
80a761fd | 1466 | req->cmd_flags &= ~REQ_FAILFAST_MASK; |
1eff9d32 | 1467 | req->cmd_flags |= req->bio->bi_opf & REQ_FAILFAST_MASK; |
80a761fd TH |
1468 | } |
1469 | ||
ed6565e7 CH |
1470 | if (!(req->rq_flags & RQF_SPECIAL_PAYLOAD)) { |
1471 | /* | |
1472 | * If total number of sectors is less than the first segment | |
1473 | * size, something has gone terribly wrong. | |
1474 | */ | |
1475 | if (blk_rq_bytes(req) < blk_rq_cur_bytes(req)) { | |
1476 | blk_dump_rq_flags(req, "request botched"); | |
1477 | req->__data_len = blk_rq_cur_bytes(req); | |
1478 | } | |
2e46e8b2 | 1479 | |
ed6565e7 CH |
1480 | /* recalculate the number of segments */ |
1481 | blk_recalc_rq_segments(req); | |
1482 | } | |
2e46e8b2 | 1483 | |
2e60e022 | 1484 | return true; |
1da177e4 | 1485 | } |
2e60e022 | 1486 | EXPORT_SYMBOL_GPL(blk_update_request); |
1da177e4 | 1487 | |
86db1e29 JA |
1488 | void blk_rq_bio_prep(struct request_queue *q, struct request *rq, |
1489 | struct bio *bio) | |
1da177e4 | 1490 | { |
b4f42e28 | 1491 | if (bio_has_data(bio)) |
fb2dce86 | 1492 | rq->nr_phys_segments = bio_phys_segments(q, bio); |
445251d0 JA |
1493 | else if (bio_op(bio) == REQ_OP_DISCARD) |
1494 | rq->nr_phys_segments = 1; | |
b4f42e28 | 1495 | |
4f024f37 | 1496 | rq->__data_len = bio->bi_iter.bi_size; |
1da177e4 | 1497 | rq->bio = rq->biotail = bio; |
1da177e4 | 1498 | |
74d46992 CH |
1499 | if (bio->bi_disk) |
1500 | rq->rq_disk = bio->bi_disk; | |
66846572 | 1501 | } |
1da177e4 | 1502 | |
2d4dc890 IL |
1503 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE |
1504 | /** | |
1505 | * rq_flush_dcache_pages - Helper function to flush all pages in a request | |
1506 | * @rq: the request to be flushed | |
1507 | * | |
1508 | * Description: | |
1509 | * Flush all pages in @rq. | |
1510 | */ | |
1511 | void rq_flush_dcache_pages(struct request *rq) | |
1512 | { | |
1513 | struct req_iterator iter; | |
7988613b | 1514 | struct bio_vec bvec; |
2d4dc890 IL |
1515 | |
1516 | rq_for_each_segment(bvec, rq, iter) | |
7988613b | 1517 | flush_dcache_page(bvec.bv_page); |
2d4dc890 IL |
1518 | } |
1519 | EXPORT_SYMBOL_GPL(rq_flush_dcache_pages); | |
1520 | #endif | |
1521 | ||
ef9e3fac KU |
1522 | /** |
1523 | * blk_lld_busy - Check if underlying low-level drivers of a device are busy | |
1524 | * @q : the queue of the device being checked | |
1525 | * | |
1526 | * Description: | |
1527 | * Check if underlying low-level drivers of a device are busy. | |
1528 | * If the drivers want to export their busy state, they must set own | |
1529 | * exporting function using blk_queue_lld_busy() first. | |
1530 | * | |
1531 | * Basically, this function is used only by request stacking drivers | |
1532 | * to stop dispatching requests to underlying devices when underlying | |
1533 | * devices are busy. This behavior helps more I/O merging on the queue | |
1534 | * of the request stacking driver and prevents I/O throughput regression | |
1535 | * on burst I/O load. | |
1536 | * | |
1537 | * Return: | |
1538 | * 0 - Not busy (The request stacking driver should dispatch request) | |
1539 | * 1 - Busy (The request stacking driver should stop dispatching request) | |
1540 | */ | |
1541 | int blk_lld_busy(struct request_queue *q) | |
1542 | { | |
344e9ffc | 1543 | if (queue_is_mq(q) && q->mq_ops->busy) |
9ba20527 | 1544 | return q->mq_ops->busy(q); |
ef9e3fac KU |
1545 | |
1546 | return 0; | |
1547 | } | |
1548 | EXPORT_SYMBOL_GPL(blk_lld_busy); | |
1549 | ||
78d8e58a MS |
1550 | /** |
1551 | * blk_rq_unprep_clone - Helper function to free all bios in a cloned request | |
1552 | * @rq: the clone request to be cleaned up | |
1553 | * | |
1554 | * Description: | |
1555 | * Free all bios in @rq for a cloned request. | |
1556 | */ | |
1557 | void blk_rq_unprep_clone(struct request *rq) | |
1558 | { | |
1559 | struct bio *bio; | |
1560 | ||
1561 | while ((bio = rq->bio) != NULL) { | |
1562 | rq->bio = bio->bi_next; | |
1563 | ||
1564 | bio_put(bio); | |
1565 | } | |
1566 | } | |
1567 | EXPORT_SYMBOL_GPL(blk_rq_unprep_clone); | |
1568 | ||
1569 | /* | |
1570 | * Copy attributes of the original request to the clone request. | |
1571 | * The actual data parts (e.g. ->cmd, ->sense) are not copied. | |
1572 | */ | |
1573 | static void __blk_rq_prep_clone(struct request *dst, struct request *src) | |
b0fd271d | 1574 | { |
b0fd271d KU |
1575 | dst->__sector = blk_rq_pos(src); |
1576 | dst->__data_len = blk_rq_bytes(src); | |
297ba57d BVA |
1577 | if (src->rq_flags & RQF_SPECIAL_PAYLOAD) { |
1578 | dst->rq_flags |= RQF_SPECIAL_PAYLOAD; | |
1579 | dst->special_vec = src->special_vec; | |
1580 | } | |
b0fd271d KU |
1581 | dst->nr_phys_segments = src->nr_phys_segments; |
1582 | dst->ioprio = src->ioprio; | |
1583 | dst->extra_len = src->extra_len; | |
78d8e58a MS |
1584 | } |
1585 | ||
1586 | /** | |
1587 | * blk_rq_prep_clone - Helper function to setup clone request | |
1588 | * @rq: the request to be setup | |
1589 | * @rq_src: original request to be cloned | |
1590 | * @bs: bio_set that bios for clone are allocated from | |
1591 | * @gfp_mask: memory allocation mask for bio | |
1592 | * @bio_ctr: setup function to be called for each clone bio. | |
1593 | * Returns %0 for success, non %0 for failure. | |
1594 | * @data: private data to be passed to @bio_ctr | |
1595 | * | |
1596 | * Description: | |
1597 | * Clones bios in @rq_src to @rq, and copies attributes of @rq_src to @rq. | |
1598 | * The actual data parts of @rq_src (e.g. ->cmd, ->sense) | |
1599 | * are not copied, and copying such parts is the caller's responsibility. | |
1600 | * Also, pages which the original bios are pointing to are not copied | |
1601 | * and the cloned bios just point same pages. | |
1602 | * So cloned bios must be completed before original bios, which means | |
1603 | * the caller must complete @rq before @rq_src. | |
1604 | */ | |
1605 | int blk_rq_prep_clone(struct request *rq, struct request *rq_src, | |
1606 | struct bio_set *bs, gfp_t gfp_mask, | |
1607 | int (*bio_ctr)(struct bio *, struct bio *, void *), | |
1608 | void *data) | |
1609 | { | |
1610 | struct bio *bio, *bio_src; | |
1611 | ||
1612 | if (!bs) | |
f4f8154a | 1613 | bs = &fs_bio_set; |
78d8e58a MS |
1614 | |
1615 | __rq_for_each_bio(bio_src, rq_src) { | |
1616 | bio = bio_clone_fast(bio_src, gfp_mask, bs); | |
1617 | if (!bio) | |
1618 | goto free_and_out; | |
1619 | ||
1620 | if (bio_ctr && bio_ctr(bio, bio_src, data)) | |
1621 | goto free_and_out; | |
1622 | ||
1623 | if (rq->bio) { | |
1624 | rq->biotail->bi_next = bio; | |
1625 | rq->biotail = bio; | |
1626 | } else | |
1627 | rq->bio = rq->biotail = bio; | |
1628 | } | |
1629 | ||
1630 | __blk_rq_prep_clone(rq, rq_src); | |
1631 | ||
1632 | return 0; | |
1633 | ||
1634 | free_and_out: | |
1635 | if (bio) | |
1636 | bio_put(bio); | |
1637 | blk_rq_unprep_clone(rq); | |
1638 | ||
1639 | return -ENOMEM; | |
b0fd271d KU |
1640 | } |
1641 | EXPORT_SYMBOL_GPL(blk_rq_prep_clone); | |
1642 | ||
59c3d45e | 1643 | int kblockd_schedule_work(struct work_struct *work) |
1da177e4 LT |
1644 | { |
1645 | return queue_work(kblockd_workqueue, work); | |
1646 | } | |
1da177e4 LT |
1647 | EXPORT_SYMBOL(kblockd_schedule_work); |
1648 | ||
ee63cfa7 JA |
1649 | int kblockd_schedule_work_on(int cpu, struct work_struct *work) |
1650 | { | |
1651 | return queue_work_on(cpu, kblockd_workqueue, work); | |
1652 | } | |
1653 | EXPORT_SYMBOL(kblockd_schedule_work_on); | |
1654 | ||
818cd1cb JA |
1655 | int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, |
1656 | unsigned long delay) | |
1657 | { | |
1658 | return mod_delayed_work_on(cpu, kblockd_workqueue, dwork, delay); | |
1659 | } | |
1660 | EXPORT_SYMBOL(kblockd_mod_delayed_work_on); | |
1661 | ||
75df7136 SJ |
1662 | /** |
1663 | * blk_start_plug - initialize blk_plug and track it inside the task_struct | |
1664 | * @plug: The &struct blk_plug that needs to be initialized | |
1665 | * | |
1666 | * Description: | |
40405851 JM |
1667 | * blk_start_plug() indicates to the block layer an intent by the caller |
1668 | * to submit multiple I/O requests in a batch. The block layer may use | |
1669 | * this hint to defer submitting I/Os from the caller until blk_finish_plug() | |
1670 | * is called. However, the block layer may choose to submit requests | |
1671 | * before a call to blk_finish_plug() if the number of queued I/Os | |
1672 | * exceeds %BLK_MAX_REQUEST_COUNT, or if the size of the I/O is larger than | |
1673 | * %BLK_PLUG_FLUSH_SIZE. The queued I/Os may also be submitted early if | |
1674 | * the task schedules (see below). | |
1675 | * | |
75df7136 SJ |
1676 | * Tracking blk_plug inside the task_struct will help with auto-flushing the |
1677 | * pending I/O should the task end up blocking between blk_start_plug() and | |
1678 | * blk_finish_plug(). This is important from a performance perspective, but | |
1679 | * also ensures that we don't deadlock. For instance, if the task is blocking | |
1680 | * for a memory allocation, memory reclaim could end up wanting to free a | |
1681 | * page belonging to that request that is currently residing in our private | |
1682 | * plug. By flushing the pending I/O when the process goes to sleep, we avoid | |
1683 | * this kind of deadlock. | |
1684 | */ | |
73c10101 JA |
1685 | void blk_start_plug(struct blk_plug *plug) |
1686 | { | |
1687 | struct task_struct *tsk = current; | |
1688 | ||
dd6cf3e1 SL |
1689 | /* |
1690 | * If this is a nested plug, don't actually assign it. | |
1691 | */ | |
1692 | if (tsk->plug) | |
1693 | return; | |
1694 | ||
320ae51f | 1695 | INIT_LIST_HEAD(&plug->mq_list); |
048c9374 | 1696 | INIT_LIST_HEAD(&plug->cb_list); |
5f0ed774 | 1697 | plug->rq_count = 0; |
ce5b009c | 1698 | plug->multiple_queues = false; |
5f0ed774 | 1699 | |
73c10101 | 1700 | /* |
dd6cf3e1 SL |
1701 | * Store ordering should not be needed here, since a potential |
1702 | * preempt will imply a full memory barrier | |
73c10101 | 1703 | */ |
dd6cf3e1 | 1704 | tsk->plug = plug; |
73c10101 JA |
1705 | } |
1706 | EXPORT_SYMBOL(blk_start_plug); | |
1707 | ||
74018dc3 | 1708 | static void flush_plug_callbacks(struct blk_plug *plug, bool from_schedule) |
048c9374 N |
1709 | { |
1710 | LIST_HEAD(callbacks); | |
1711 | ||
2a7d5559 SL |
1712 | while (!list_empty(&plug->cb_list)) { |
1713 | list_splice_init(&plug->cb_list, &callbacks); | |
048c9374 | 1714 | |
2a7d5559 SL |
1715 | while (!list_empty(&callbacks)) { |
1716 | struct blk_plug_cb *cb = list_first_entry(&callbacks, | |
048c9374 N |
1717 | struct blk_plug_cb, |
1718 | list); | |
2a7d5559 | 1719 | list_del(&cb->list); |
74018dc3 | 1720 | cb->callback(cb, from_schedule); |
2a7d5559 | 1721 | } |
048c9374 N |
1722 | } |
1723 | } | |
1724 | ||
9cbb1750 N |
1725 | struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug, void *data, |
1726 | int size) | |
1727 | { | |
1728 | struct blk_plug *plug = current->plug; | |
1729 | struct blk_plug_cb *cb; | |
1730 | ||
1731 | if (!plug) | |
1732 | return NULL; | |
1733 | ||
1734 | list_for_each_entry(cb, &plug->cb_list, list) | |
1735 | if (cb->callback == unplug && cb->data == data) | |
1736 | return cb; | |
1737 | ||
1738 | /* Not currently on the callback list */ | |
1739 | BUG_ON(size < sizeof(*cb)); | |
1740 | cb = kzalloc(size, GFP_ATOMIC); | |
1741 | if (cb) { | |
1742 | cb->data = data; | |
1743 | cb->callback = unplug; | |
1744 | list_add(&cb->list, &plug->cb_list); | |
1745 | } | |
1746 | return cb; | |
1747 | } | |
1748 | EXPORT_SYMBOL(blk_check_plugged); | |
1749 | ||
49cac01e | 1750 | void blk_flush_plug_list(struct blk_plug *plug, bool from_schedule) |
73c10101 | 1751 | { |
74018dc3 | 1752 | flush_plug_callbacks(plug, from_schedule); |
320ae51f JA |
1753 | |
1754 | if (!list_empty(&plug->mq_list)) | |
1755 | blk_mq_flush_plug_list(plug, from_schedule); | |
73c10101 | 1756 | } |
73c10101 | 1757 | |
40405851 JM |
1758 | /** |
1759 | * blk_finish_plug - mark the end of a batch of submitted I/O | |
1760 | * @plug: The &struct blk_plug passed to blk_start_plug() | |
1761 | * | |
1762 | * Description: | |
1763 | * Indicate that a batch of I/O submissions is complete. This function | |
1764 | * must be paired with an initial call to blk_start_plug(). The intent | |
1765 | * is to allow the block layer to optimize I/O submission. See the | |
1766 | * documentation for blk_start_plug() for more information. | |
1767 | */ | |
73c10101 JA |
1768 | void blk_finish_plug(struct blk_plug *plug) |
1769 | { | |
dd6cf3e1 SL |
1770 | if (plug != current->plug) |
1771 | return; | |
f6603783 | 1772 | blk_flush_plug_list(plug, false); |
73c10101 | 1773 | |
dd6cf3e1 | 1774 | current->plug = NULL; |
73c10101 | 1775 | } |
88b996cd | 1776 | EXPORT_SYMBOL(blk_finish_plug); |
73c10101 | 1777 | |
1da177e4 LT |
1778 | int __init blk_dev_init(void) |
1779 | { | |
ef295ecf CH |
1780 | BUILD_BUG_ON(REQ_OP_LAST >= (1 << REQ_OP_BITS)); |
1781 | BUILD_BUG_ON(REQ_OP_BITS + REQ_FLAG_BITS > 8 * | |
0762b23d | 1782 | FIELD_SIZEOF(struct request, cmd_flags)); |
ef295ecf CH |
1783 | BUILD_BUG_ON(REQ_OP_BITS + REQ_FLAG_BITS > 8 * |
1784 | FIELD_SIZEOF(struct bio, bi_opf)); | |
9eb55b03 | 1785 | |
89b90be2 TH |
1786 | /* used for unplugging and affects IO latency/throughput - HIGHPRI */ |
1787 | kblockd_workqueue = alloc_workqueue("kblockd", | |
28747fcd | 1788 | WQ_MEM_RECLAIM | WQ_HIGHPRI, 0); |
1da177e4 LT |
1789 | if (!kblockd_workqueue) |
1790 | panic("Failed to create kblockd\n"); | |
1791 | ||
c2789bd4 | 1792 | blk_requestq_cachep = kmem_cache_create("request_queue", |
165125e1 | 1793 | sizeof(struct request_queue), 0, SLAB_PANIC, NULL); |
1da177e4 | 1794 | |
18fbda91 OS |
1795 | #ifdef CONFIG_DEBUG_FS |
1796 | blk_debugfs_root = debugfs_create_dir("block", NULL); | |
1797 | #endif | |
1798 | ||
d38ecf93 | 1799 | return 0; |
1da177e4 | 1800 | } |