]>
Commit | Line | Data |
---|---|---|
1da177e4 | 1 | /* |
1da177e4 LT |
2 | * Copyright (C) 1991, 1992 Linus Torvalds |
3 | * Copyright (C) 1994, Karl Keyte: Added support for disk statistics | |
4 | * Elevator latency, (C) 2000 Andrea Arcangeli <andrea@suse.de> SuSE | |
5 | * Queue request tables / lock, selectable elevator, Jens Axboe <axboe@suse.de> | |
6728cb0e JA |
6 | * kernel-doc documentation started by NeilBrown <neilb@cse.unsw.edu.au> |
7 | * - July2000 | |
1da177e4 LT |
8 | * bio rewrite, highmem i/o, etc, Jens Axboe <axboe@suse.de> - may 2001 |
9 | */ | |
10 | ||
11 | /* | |
12 | * This handles all read/write requests to block devices | |
13 | */ | |
1da177e4 LT |
14 | #include <linux/kernel.h> |
15 | #include <linux/module.h> | |
16 | #include <linux/backing-dev.h> | |
17 | #include <linux/bio.h> | |
18 | #include <linux/blkdev.h> | |
320ae51f | 19 | #include <linux/blk-mq.h> |
1da177e4 LT |
20 | #include <linux/highmem.h> |
21 | #include <linux/mm.h> | |
22 | #include <linux/kernel_stat.h> | |
23 | #include <linux/string.h> | |
24 | #include <linux/init.h> | |
1da177e4 LT |
25 | #include <linux/completion.h> |
26 | #include <linux/slab.h> | |
27 | #include <linux/swap.h> | |
28 | #include <linux/writeback.h> | |
faccbd4b | 29 | #include <linux/task_io_accounting_ops.h> |
c17bb495 | 30 | #include <linux/fault-inject.h> |
73c10101 | 31 | #include <linux/list_sort.h> |
e3c78ca5 | 32 | #include <linux/delay.h> |
aaf7c680 | 33 | #include <linux/ratelimit.h> |
6c954667 | 34 | #include <linux/pm_runtime.h> |
eea8f41c | 35 | #include <linux/blk-cgroup.h> |
18fbda91 | 36 | #include <linux/debugfs.h> |
30abb3a6 | 37 | #include <linux/bpf.h> |
55782138 LZ |
38 | |
39 | #define CREATE_TRACE_POINTS | |
40 | #include <trace/events/block.h> | |
1da177e4 | 41 | |
8324aa91 | 42 | #include "blk.h" |
43a5e4e2 | 43 | #include "blk-mq.h" |
bd166ef1 | 44 | #include "blk-mq-sched.h" |
bca6b067 | 45 | #include "blk-pm.h" |
c1c80384 | 46 | #include "blk-rq-qos.h" |
8324aa91 | 47 | |
18fbda91 OS |
48 | #ifdef CONFIG_DEBUG_FS |
49 | struct dentry *blk_debugfs_root; | |
50 | #endif | |
51 | ||
d07335e5 | 52 | EXPORT_TRACEPOINT_SYMBOL_GPL(block_bio_remap); |
b0da3f0d | 53 | EXPORT_TRACEPOINT_SYMBOL_GPL(block_rq_remap); |
0a82a8d1 | 54 | EXPORT_TRACEPOINT_SYMBOL_GPL(block_bio_complete); |
3291fa57 | 55 | EXPORT_TRACEPOINT_SYMBOL_GPL(block_split); |
cbae8d45 | 56 | EXPORT_TRACEPOINT_SYMBOL_GPL(block_unplug); |
0bfc2455 | 57 | |
a73f730d TH |
58 | DEFINE_IDA(blk_queue_ida); |
59 | ||
1da177e4 LT |
60 | /* |
61 | * For queue allocation | |
62 | */ | |
6728cb0e | 63 | struct kmem_cache *blk_requestq_cachep; |
1da177e4 | 64 | |
1da177e4 LT |
65 | /* |
66 | * Controlling structure to kblockd | |
67 | */ | |
ff856bad | 68 | static struct workqueue_struct *kblockd_workqueue; |
1da177e4 | 69 | |
8814ce8a BVA |
70 | /** |
71 | * blk_queue_flag_set - atomically set a queue flag | |
72 | * @flag: flag to be set | |
73 | * @q: request queue | |
74 | */ | |
75 | void blk_queue_flag_set(unsigned int flag, struct request_queue *q) | |
76 | { | |
57d74df9 | 77 | set_bit(flag, &q->queue_flags); |
8814ce8a BVA |
78 | } |
79 | EXPORT_SYMBOL(blk_queue_flag_set); | |
80 | ||
81 | /** | |
82 | * blk_queue_flag_clear - atomically clear a queue flag | |
83 | * @flag: flag to be cleared | |
84 | * @q: request queue | |
85 | */ | |
86 | void blk_queue_flag_clear(unsigned int flag, struct request_queue *q) | |
87 | { | |
57d74df9 | 88 | clear_bit(flag, &q->queue_flags); |
8814ce8a BVA |
89 | } |
90 | EXPORT_SYMBOL(blk_queue_flag_clear); | |
91 | ||
92 | /** | |
93 | * blk_queue_flag_test_and_set - atomically test and set a queue flag | |
94 | * @flag: flag to be set | |
95 | * @q: request queue | |
96 | * | |
97 | * Returns the previous value of @flag - 0 if the flag was not set and 1 if | |
98 | * the flag was already set. | |
99 | */ | |
100 | bool blk_queue_flag_test_and_set(unsigned int flag, struct request_queue *q) | |
101 | { | |
57d74df9 | 102 | return test_and_set_bit(flag, &q->queue_flags); |
8814ce8a BVA |
103 | } |
104 | EXPORT_SYMBOL_GPL(blk_queue_flag_test_and_set); | |
105 | ||
2a4aa30c | 106 | void blk_rq_init(struct request_queue *q, struct request *rq) |
1da177e4 | 107 | { |
1afb20f3 FT |
108 | memset(rq, 0, sizeof(*rq)); |
109 | ||
1da177e4 | 110 | INIT_LIST_HEAD(&rq->queuelist); |
63a71386 | 111 | rq->q = q; |
a2dec7b3 | 112 | rq->__sector = (sector_t) -1; |
2e662b65 JA |
113 | INIT_HLIST_NODE(&rq->hash); |
114 | RB_CLEAR_NODE(&rq->rb_node); | |
63a71386 | 115 | rq->tag = -1; |
bd166ef1 | 116 | rq->internal_tag = -1; |
522a7775 | 117 | rq->start_time_ns = ktime_get_ns(); |
09e099d4 | 118 | rq->part = NULL; |
1da177e4 | 119 | } |
2a4aa30c | 120 | EXPORT_SYMBOL(blk_rq_init); |
1da177e4 | 121 | |
2a842aca CH |
122 | static const struct { |
123 | int errno; | |
124 | const char *name; | |
125 | } blk_errors[] = { | |
126 | [BLK_STS_OK] = { 0, "" }, | |
127 | [BLK_STS_NOTSUPP] = { -EOPNOTSUPP, "operation not supported" }, | |
128 | [BLK_STS_TIMEOUT] = { -ETIMEDOUT, "timeout" }, | |
129 | [BLK_STS_NOSPC] = { -ENOSPC, "critical space allocation" }, | |
130 | [BLK_STS_TRANSPORT] = { -ENOLINK, "recoverable transport" }, | |
131 | [BLK_STS_TARGET] = { -EREMOTEIO, "critical target" }, | |
132 | [BLK_STS_NEXUS] = { -EBADE, "critical nexus" }, | |
133 | [BLK_STS_MEDIUM] = { -ENODATA, "critical medium" }, | |
134 | [BLK_STS_PROTECTION] = { -EILSEQ, "protection" }, | |
135 | [BLK_STS_RESOURCE] = { -ENOMEM, "kernel resource" }, | |
86ff7c2a | 136 | [BLK_STS_DEV_RESOURCE] = { -EBUSY, "device resource" }, |
03a07c92 | 137 | [BLK_STS_AGAIN] = { -EAGAIN, "nonblocking retry" }, |
2a842aca | 138 | |
4e4cbee9 CH |
139 | /* device mapper special case, should not leak out: */ |
140 | [BLK_STS_DM_REQUEUE] = { -EREMCHG, "dm internal retry" }, | |
141 | ||
2a842aca CH |
142 | /* everything else not covered above: */ |
143 | [BLK_STS_IOERR] = { -EIO, "I/O" }, | |
144 | }; | |
145 | ||
146 | blk_status_t errno_to_blk_status(int errno) | |
147 | { | |
148 | int i; | |
149 | ||
150 | for (i = 0; i < ARRAY_SIZE(blk_errors); i++) { | |
151 | if (blk_errors[i].errno == errno) | |
152 | return (__force blk_status_t)i; | |
153 | } | |
154 | ||
155 | return BLK_STS_IOERR; | |
156 | } | |
157 | EXPORT_SYMBOL_GPL(errno_to_blk_status); | |
158 | ||
159 | int blk_status_to_errno(blk_status_t status) | |
160 | { | |
161 | int idx = (__force int)status; | |
162 | ||
34bd9c1c | 163 | if (WARN_ON_ONCE(idx >= ARRAY_SIZE(blk_errors))) |
2a842aca CH |
164 | return -EIO; |
165 | return blk_errors[idx].errno; | |
166 | } | |
167 | EXPORT_SYMBOL_GPL(blk_status_to_errno); | |
168 | ||
169 | static void print_req_error(struct request *req, blk_status_t status) | |
170 | { | |
171 | int idx = (__force int)status; | |
172 | ||
34bd9c1c | 173 | if (WARN_ON_ONCE(idx >= ARRAY_SIZE(blk_errors))) |
2a842aca CH |
174 | return; |
175 | ||
2149da07 BS |
176 | printk_ratelimited(KERN_ERR "%s: %s error, dev %s, sector %llu flags %x\n", |
177 | __func__, blk_errors[idx].name, | |
178 | req->rq_disk ? req->rq_disk->disk_name : "?", | |
179 | (unsigned long long)blk_rq_pos(req), | |
180 | req->cmd_flags); | |
2a842aca CH |
181 | } |
182 | ||
5bb23a68 | 183 | static void req_bio_endio(struct request *rq, struct bio *bio, |
2a842aca | 184 | unsigned int nbytes, blk_status_t error) |
1da177e4 | 185 | { |
78d8e58a | 186 | if (error) |
4e4cbee9 | 187 | bio->bi_status = error; |
797e7dbb | 188 | |
e8064021 | 189 | if (unlikely(rq->rq_flags & RQF_QUIET)) |
b7c44ed9 | 190 | bio_set_flag(bio, BIO_QUIET); |
08bafc03 | 191 | |
f79ea416 | 192 | bio_advance(bio, nbytes); |
7ba1ba12 | 193 | |
143a87f4 | 194 | /* don't actually finish bio if it's part of flush sequence */ |
e8064021 | 195 | if (bio->bi_iter.bi_size == 0 && !(rq->rq_flags & RQF_FLUSH_SEQ)) |
4246a0b6 | 196 | bio_endio(bio); |
1da177e4 | 197 | } |
1da177e4 | 198 | |
1da177e4 LT |
199 | void blk_dump_rq_flags(struct request *rq, char *msg) |
200 | { | |
aebf526b CH |
201 | printk(KERN_INFO "%s: dev %s: flags=%llx\n", msg, |
202 | rq->rq_disk ? rq->rq_disk->disk_name : "?", | |
5953316d | 203 | (unsigned long long) rq->cmd_flags); |
1da177e4 | 204 | |
83096ebf TH |
205 | printk(KERN_INFO " sector %llu, nr/cnr %u/%u\n", |
206 | (unsigned long long)blk_rq_pos(rq), | |
207 | blk_rq_sectors(rq), blk_rq_cur_sectors(rq)); | |
b4f42e28 JA |
208 | printk(KERN_INFO " bio %p, biotail %p, len %u\n", |
209 | rq->bio, rq->biotail, blk_rq_bytes(rq)); | |
1da177e4 | 210 | } |
1da177e4 LT |
211 | EXPORT_SYMBOL(blk_dump_rq_flags); |
212 | ||
1da177e4 LT |
213 | /** |
214 | * blk_sync_queue - cancel any pending callbacks on a queue | |
215 | * @q: the queue | |
216 | * | |
217 | * Description: | |
218 | * The block layer may perform asynchronous callback activity | |
219 | * on a queue, such as calling the unplug function after a timeout. | |
220 | * A block device may call blk_sync_queue to ensure that any | |
221 | * such activity is cancelled, thus allowing it to release resources | |
59c51591 | 222 | * that the callbacks might use. The caller must already have made sure |
1da177e4 LT |
223 | * that its ->make_request_fn will not re-add plugging prior to calling |
224 | * this function. | |
225 | * | |
da527770 | 226 | * This function does not cancel any asynchronous activity arising |
da3dae54 | 227 | * out of elevator or throttling code. That would require elevator_exit() |
5efd6113 | 228 | * and blkcg_exit_queue() to be called with queue lock initialized. |
da527770 | 229 | * |
1da177e4 LT |
230 | */ |
231 | void blk_sync_queue(struct request_queue *q) | |
232 | { | |
70ed28b9 | 233 | del_timer_sync(&q->timeout); |
4e9b6f20 | 234 | cancel_work_sync(&q->timeout_work); |
f04c1fe7 | 235 | |
344e9ffc | 236 | if (queue_is_mq(q)) { |
f04c1fe7 ML |
237 | struct blk_mq_hw_ctx *hctx; |
238 | int i; | |
239 | ||
aba7afc5 | 240 | cancel_delayed_work_sync(&q->requeue_work); |
21c6e939 | 241 | queue_for_each_hw_ctx(q, hctx, i) |
9f993737 | 242 | cancel_delayed_work_sync(&hctx->run_work); |
f04c1fe7 | 243 | } |
1da177e4 LT |
244 | } |
245 | EXPORT_SYMBOL(blk_sync_queue); | |
246 | ||
c9254f2d | 247 | /** |
cd84a62e | 248 | * blk_set_pm_only - increment pm_only counter |
c9254f2d | 249 | * @q: request queue pointer |
c9254f2d | 250 | */ |
cd84a62e | 251 | void blk_set_pm_only(struct request_queue *q) |
c9254f2d | 252 | { |
cd84a62e | 253 | atomic_inc(&q->pm_only); |
c9254f2d | 254 | } |
cd84a62e | 255 | EXPORT_SYMBOL_GPL(blk_set_pm_only); |
c9254f2d | 256 | |
cd84a62e | 257 | void blk_clear_pm_only(struct request_queue *q) |
c9254f2d | 258 | { |
cd84a62e BVA |
259 | int pm_only; |
260 | ||
261 | pm_only = atomic_dec_return(&q->pm_only); | |
262 | WARN_ON_ONCE(pm_only < 0); | |
263 | if (pm_only == 0) | |
264 | wake_up_all(&q->mq_freeze_wq); | |
c9254f2d | 265 | } |
cd84a62e | 266 | EXPORT_SYMBOL_GPL(blk_clear_pm_only); |
c9254f2d | 267 | |
165125e1 | 268 | void blk_put_queue(struct request_queue *q) |
483f4afc AV |
269 | { |
270 | kobject_put(&q->kobj); | |
271 | } | |
d86e0e83 | 272 | EXPORT_SYMBOL(blk_put_queue); |
483f4afc | 273 | |
aed3ea94 JA |
274 | void blk_set_queue_dying(struct request_queue *q) |
275 | { | |
8814ce8a | 276 | blk_queue_flag_set(QUEUE_FLAG_DYING, q); |
aed3ea94 | 277 | |
d3cfb2a0 ML |
278 | /* |
279 | * When queue DYING flag is set, we need to block new req | |
280 | * entering queue, so we call blk_freeze_queue_start() to | |
281 | * prevent I/O from crossing blk_queue_enter(). | |
282 | */ | |
283 | blk_freeze_queue_start(q); | |
284 | ||
344e9ffc | 285 | if (queue_is_mq(q)) |
aed3ea94 | 286 | blk_mq_wake_waiters(q); |
055f6e18 ML |
287 | |
288 | /* Make blk_queue_enter() reexamine the DYING flag. */ | |
289 | wake_up_all(&q->mq_freeze_wq); | |
aed3ea94 JA |
290 | } |
291 | EXPORT_SYMBOL_GPL(blk_set_queue_dying); | |
292 | ||
4cf6324b BVA |
293 | /* Unconfigure the I/O scheduler and dissociate from the cgroup controller. */ |
294 | void blk_exit_queue(struct request_queue *q) | |
295 | { | |
296 | /* | |
297 | * Since the I/O scheduler exit code may access cgroup information, | |
298 | * perform I/O scheduler exit before disassociating from the block | |
299 | * cgroup controller. | |
300 | */ | |
301 | if (q->elevator) { | |
302 | ioc_clear_queue(q); | |
303 | elevator_exit(q, q->elevator); | |
304 | q->elevator = NULL; | |
305 | } | |
306 | ||
307 | /* | |
308 | * Remove all references to @q from the block cgroup controller before | |
309 | * restoring @q->queue_lock to avoid that restoring this pointer causes | |
310 | * e.g. blkcg_print_blkgs() to crash. | |
311 | */ | |
312 | blkcg_exit_queue(q); | |
313 | ||
314 | /* | |
315 | * Since the cgroup code may dereference the @q->backing_dev_info | |
316 | * pointer, only decrease its reference count after having removed the | |
317 | * association with the block cgroup controller. | |
318 | */ | |
319 | bdi_put(q->backing_dev_info); | |
320 | } | |
321 | ||
c9a929dd TH |
322 | /** |
323 | * blk_cleanup_queue - shutdown a request queue | |
324 | * @q: request queue to shutdown | |
325 | * | |
c246e80d BVA |
326 | * Mark @q DYING, drain all pending requests, mark @q DEAD, destroy and |
327 | * put it. All future requests will be failed immediately with -ENODEV. | |
c94a96ac | 328 | */ |
6728cb0e | 329 | void blk_cleanup_queue(struct request_queue *q) |
483f4afc | 330 | { |
3f3299d5 | 331 | /* mark @q DYING, no new request or merges will be allowed afterwards */ |
483f4afc | 332 | mutex_lock(&q->sysfs_lock); |
aed3ea94 | 333 | blk_set_queue_dying(q); |
6ecf23af | 334 | |
57d74df9 CH |
335 | blk_queue_flag_set(QUEUE_FLAG_NOMERGES, q); |
336 | blk_queue_flag_set(QUEUE_FLAG_NOXMERGES, q); | |
337 | blk_queue_flag_set(QUEUE_FLAG_DYING, q); | |
c9a929dd TH |
338 | mutex_unlock(&q->sysfs_lock); |
339 | ||
c246e80d BVA |
340 | /* |
341 | * Drain all requests queued before DYING marking. Set DEAD flag to | |
342 | * prevent that q->request_fn() gets invoked after draining finished. | |
343 | */ | |
3ef28e83 | 344 | blk_freeze_queue(q); |
c57cdf7a ML |
345 | |
346 | rq_qos_exit(q); | |
347 | ||
57d74df9 | 348 | blk_queue_flag_set(QUEUE_FLAG_DEAD, q); |
c9a929dd | 349 | |
c2856ae2 ML |
350 | /* |
351 | * make sure all in-progress dispatch are completed because | |
352 | * blk_freeze_queue() can only complete all requests, and | |
353 | * dispatch may still be in-progress since we dispatch requests | |
1311326c ML |
354 | * from more than one contexts. |
355 | * | |
8dc765d4 ML |
356 | * We rely on driver to deal with the race in case that queue |
357 | * initialization isn't done. | |
c2856ae2 | 358 | */ |
344e9ffc | 359 | if (queue_is_mq(q) && blk_queue_init_done(q)) |
c2856ae2 ML |
360 | blk_mq_quiesce_queue(q); |
361 | ||
5a48fc14 DW |
362 | /* for synchronous bio-based driver finish in-flight integrity i/o */ |
363 | blk_flush_integrity(); | |
364 | ||
c9a929dd | 365 | /* @q won't process any more request, flush async actions */ |
dc3b17cc | 366 | del_timer_sync(&q->backing_dev_info->laptop_mode_wb_timer); |
c9a929dd TH |
367 | blk_sync_queue(q); |
368 | ||
a063057d BVA |
369 | /* |
370 | * I/O scheduler exit is only safe after the sysfs scheduler attribute | |
371 | * has been removed. | |
372 | */ | |
373 | WARN_ON_ONCE(q->kobj.state_in_sysfs); | |
374 | ||
4cf6324b | 375 | blk_exit_queue(q); |
a063057d | 376 | |
344e9ffc | 377 | if (queue_is_mq(q)) |
45a9c9d9 | 378 | blk_mq_free_queue(q); |
a1ce35fa | 379 | |
3ef28e83 | 380 | percpu_ref_exit(&q->q_usage_counter); |
45a9c9d9 | 381 | |
c9a929dd | 382 | /* @q is and will stay empty, shutdown and put */ |
483f4afc AV |
383 | blk_put_queue(q); |
384 | } | |
1da177e4 LT |
385 | EXPORT_SYMBOL(blk_cleanup_queue); |
386 | ||
165125e1 | 387 | struct request_queue *blk_alloc_queue(gfp_t gfp_mask) |
1da177e4 | 388 | { |
6d469642 | 389 | return blk_alloc_queue_node(gfp_mask, NUMA_NO_NODE); |
1946089a CL |
390 | } |
391 | EXPORT_SYMBOL(blk_alloc_queue); | |
1da177e4 | 392 | |
3a0a5299 BVA |
393 | /** |
394 | * blk_queue_enter() - try to increase q->q_usage_counter | |
395 | * @q: request queue pointer | |
396 | * @flags: BLK_MQ_REQ_NOWAIT and/or BLK_MQ_REQ_PREEMPT | |
397 | */ | |
9a95e4ef | 398 | int blk_queue_enter(struct request_queue *q, blk_mq_req_flags_t flags) |
3ef28e83 | 399 | { |
cd84a62e | 400 | const bool pm = flags & BLK_MQ_REQ_PREEMPT; |
3a0a5299 | 401 | |
3ef28e83 | 402 | while (true) { |
3a0a5299 | 403 | bool success = false; |
3ef28e83 | 404 | |
818e0fa2 | 405 | rcu_read_lock(); |
3a0a5299 BVA |
406 | if (percpu_ref_tryget_live(&q->q_usage_counter)) { |
407 | /* | |
cd84a62e BVA |
408 | * The code that increments the pm_only counter is |
409 | * responsible for ensuring that that counter is | |
410 | * globally visible before the queue is unfrozen. | |
3a0a5299 | 411 | */ |
cd84a62e | 412 | if (pm || !blk_queue_pm_only(q)) { |
3a0a5299 BVA |
413 | success = true; |
414 | } else { | |
415 | percpu_ref_put(&q->q_usage_counter); | |
416 | } | |
417 | } | |
818e0fa2 | 418 | rcu_read_unlock(); |
3a0a5299 BVA |
419 | |
420 | if (success) | |
3ef28e83 DW |
421 | return 0; |
422 | ||
3a0a5299 | 423 | if (flags & BLK_MQ_REQ_NOWAIT) |
3ef28e83 DW |
424 | return -EBUSY; |
425 | ||
5ed61d3f | 426 | /* |
1671d522 | 427 | * read pair of barrier in blk_freeze_queue_start(), |
5ed61d3f | 428 | * we need to order reading __PERCPU_REF_DEAD flag of |
d3cfb2a0 ML |
429 | * .q_usage_counter and reading .mq_freeze_depth or |
430 | * queue dying flag, otherwise the following wait may | |
431 | * never return if the two reads are reordered. | |
5ed61d3f ML |
432 | */ |
433 | smp_rmb(); | |
434 | ||
1dc3039b AJ |
435 | wait_event(q->mq_freeze_wq, |
436 | (atomic_read(&q->mq_freeze_depth) == 0 && | |
0d25bd07 BVA |
437 | (pm || (blk_pm_request_resume(q), |
438 | !blk_queue_pm_only(q)))) || | |
1dc3039b | 439 | blk_queue_dying(q)); |
3ef28e83 DW |
440 | if (blk_queue_dying(q)) |
441 | return -ENODEV; | |
3ef28e83 DW |
442 | } |
443 | } | |
444 | ||
445 | void blk_queue_exit(struct request_queue *q) | |
446 | { | |
447 | percpu_ref_put(&q->q_usage_counter); | |
448 | } | |
449 | ||
450 | static void blk_queue_usage_counter_release(struct percpu_ref *ref) | |
451 | { | |
452 | struct request_queue *q = | |
453 | container_of(ref, struct request_queue, q_usage_counter); | |
454 | ||
455 | wake_up_all(&q->mq_freeze_wq); | |
456 | } | |
457 | ||
bca237a5 | 458 | static void blk_rq_timed_out_timer(struct timer_list *t) |
287922eb | 459 | { |
bca237a5 | 460 | struct request_queue *q = from_timer(q, t, timeout); |
287922eb CH |
461 | |
462 | kblockd_schedule_work(&q->timeout_work); | |
463 | } | |
464 | ||
2e3c18d0 TH |
465 | static void blk_timeout_work(struct work_struct *work) |
466 | { | |
467 | } | |
468 | ||
498f6650 BVA |
469 | /** |
470 | * blk_alloc_queue_node - allocate a request queue | |
471 | * @gfp_mask: memory allocation flags | |
472 | * @node_id: NUMA node to allocate memory from | |
498f6650 | 473 | */ |
6d469642 | 474 | struct request_queue *blk_alloc_queue_node(gfp_t gfp_mask, int node_id) |
1946089a | 475 | { |
165125e1 | 476 | struct request_queue *q; |
338aa96d | 477 | int ret; |
1946089a | 478 | |
8324aa91 | 479 | q = kmem_cache_alloc_node(blk_requestq_cachep, |
94f6030c | 480 | gfp_mask | __GFP_ZERO, node_id); |
1da177e4 LT |
481 | if (!q) |
482 | return NULL; | |
483 | ||
cbf62af3 CH |
484 | INIT_LIST_HEAD(&q->queue_head); |
485 | q->last_merge = NULL; | |
cbf62af3 | 486 | |
00380a40 | 487 | q->id = ida_simple_get(&blk_queue_ida, 0, 0, gfp_mask); |
a73f730d | 488 | if (q->id < 0) |
3d2936f4 | 489 | goto fail_q; |
a73f730d | 490 | |
338aa96d KO |
491 | ret = bioset_init(&q->bio_split, BIO_POOL_SIZE, 0, BIOSET_NEED_BVECS); |
492 | if (ret) | |
54efd50b KO |
493 | goto fail_id; |
494 | ||
d03f6cdc JK |
495 | q->backing_dev_info = bdi_alloc_node(gfp_mask, node_id); |
496 | if (!q->backing_dev_info) | |
497 | goto fail_split; | |
498 | ||
a83b576c JA |
499 | q->stats = blk_alloc_queue_stats(); |
500 | if (!q->stats) | |
501 | goto fail_stats; | |
502 | ||
dc3b17cc | 503 | q->backing_dev_info->ra_pages = |
09cbfeaf | 504 | (VM_MAX_READAHEAD * 1024) / PAGE_SIZE; |
dc3b17cc JK |
505 | q->backing_dev_info->capabilities = BDI_CAP_CGROUP_WRITEBACK; |
506 | q->backing_dev_info->name = "block"; | |
5151412d | 507 | q->node = node_id; |
0989a025 | 508 | |
bca237a5 KC |
509 | timer_setup(&q->backing_dev_info->laptop_mode_wb_timer, |
510 | laptop_mode_timer_fn, 0); | |
511 | timer_setup(&q->timeout, blk_rq_timed_out_timer, 0); | |
2e3c18d0 | 512 | INIT_WORK(&q->timeout_work, blk_timeout_work); |
a612fddf | 513 | INIT_LIST_HEAD(&q->icq_list); |
4eef3049 | 514 | #ifdef CONFIG_BLK_CGROUP |
e8989fae | 515 | INIT_LIST_HEAD(&q->blkg_list); |
4eef3049 | 516 | #endif |
483f4afc | 517 | |
8324aa91 | 518 | kobject_init(&q->kobj, &blk_queue_ktype); |
1da177e4 | 519 | |
5acb3cc2 WL |
520 | #ifdef CONFIG_BLK_DEV_IO_TRACE |
521 | mutex_init(&q->blk_trace_mutex); | |
522 | #endif | |
483f4afc | 523 | mutex_init(&q->sysfs_lock); |
0d945c1f | 524 | spin_lock_init(&q->queue_lock); |
c94a96ac | 525 | |
320ae51f JA |
526 | init_waitqueue_head(&q->mq_freeze_wq); |
527 | ||
3ef28e83 DW |
528 | /* |
529 | * Init percpu_ref in atomic mode so that it's faster to shutdown. | |
530 | * See blk_register_queue() for details. | |
531 | */ | |
532 | if (percpu_ref_init(&q->q_usage_counter, | |
533 | blk_queue_usage_counter_release, | |
534 | PERCPU_REF_INIT_ATOMIC, GFP_KERNEL)) | |
fff4996b | 535 | goto fail_bdi; |
f51b802c | 536 | |
3ef28e83 DW |
537 | if (blkcg_init_queue(q)) |
538 | goto fail_ref; | |
539 | ||
1da177e4 | 540 | return q; |
a73f730d | 541 | |
3ef28e83 DW |
542 | fail_ref: |
543 | percpu_ref_exit(&q->q_usage_counter); | |
fff4996b | 544 | fail_bdi: |
a83b576c JA |
545 | blk_free_queue_stats(q->stats); |
546 | fail_stats: | |
d03f6cdc | 547 | bdi_put(q->backing_dev_info); |
54efd50b | 548 | fail_split: |
338aa96d | 549 | bioset_exit(&q->bio_split); |
a73f730d TH |
550 | fail_id: |
551 | ida_simple_remove(&blk_queue_ida, q->id); | |
552 | fail_q: | |
553 | kmem_cache_free(blk_requestq_cachep, q); | |
554 | return NULL; | |
1da177e4 | 555 | } |
1946089a | 556 | EXPORT_SYMBOL(blk_alloc_queue_node); |
1da177e4 | 557 | |
09ac46c4 | 558 | bool blk_get_queue(struct request_queue *q) |
1da177e4 | 559 | { |
3f3299d5 | 560 | if (likely(!blk_queue_dying(q))) { |
09ac46c4 TH |
561 | __blk_get_queue(q); |
562 | return true; | |
1da177e4 LT |
563 | } |
564 | ||
09ac46c4 | 565 | return false; |
1da177e4 | 566 | } |
d86e0e83 | 567 | EXPORT_SYMBOL(blk_get_queue); |
1da177e4 | 568 | |
a1ce35fa JA |
569 | /** |
570 | * blk_get_request - allocate a request | |
571 | * @q: request queue to allocate a request for | |
572 | * @op: operation (REQ_OP_*) and REQ_* flags, e.g. REQ_SYNC. | |
573 | * @flags: BLK_MQ_REQ_* flags, e.g. BLK_MQ_REQ_NOWAIT. | |
1da177e4 | 574 | */ |
a1ce35fa JA |
575 | struct request *blk_get_request(struct request_queue *q, unsigned int op, |
576 | blk_mq_req_flags_t flags) | |
1da177e4 | 577 | { |
a1ce35fa | 578 | struct request *req; |
1da177e4 | 579 | |
a1ce35fa JA |
580 | WARN_ON_ONCE(op & REQ_NOWAIT); |
581 | WARN_ON_ONCE(flags & ~(BLK_MQ_REQ_NOWAIT | BLK_MQ_REQ_PREEMPT)); | |
1da177e4 | 582 | |
a1ce35fa JA |
583 | req = blk_mq_alloc_request(q, op, flags); |
584 | if (!IS_ERR(req) && q->mq_ops->initialize_rq_fn) | |
585 | q->mq_ops->initialize_rq_fn(req); | |
1da177e4 | 586 | |
a1ce35fa | 587 | return req; |
1da177e4 | 588 | } |
a1ce35fa | 589 | EXPORT_SYMBOL(blk_get_request); |
1da177e4 | 590 | |
1da177e4 LT |
591 | void blk_put_request(struct request *req) |
592 | { | |
a1ce35fa | 593 | blk_mq_free_request(req); |
1da177e4 | 594 | } |
1da177e4 LT |
595 | EXPORT_SYMBOL(blk_put_request); |
596 | ||
320ae51f JA |
597 | bool bio_attempt_back_merge(struct request_queue *q, struct request *req, |
598 | struct bio *bio) | |
73c10101 | 599 | { |
1eff9d32 | 600 | const int ff = bio->bi_opf & REQ_FAILFAST_MASK; |
73c10101 | 601 | |
73c10101 JA |
602 | if (!ll_back_merge_fn(q, req, bio)) |
603 | return false; | |
604 | ||
8c1cf6bb | 605 | trace_block_bio_backmerge(q, req, bio); |
73c10101 JA |
606 | |
607 | if ((req->cmd_flags & REQ_FAILFAST_MASK) != ff) | |
608 | blk_rq_set_mixed_merge(req); | |
609 | ||
610 | req->biotail->bi_next = bio; | |
611 | req->biotail = bio; | |
4f024f37 | 612 | req->__data_len += bio->bi_iter.bi_size; |
73c10101 | 613 | |
320ae51f | 614 | blk_account_io_start(req, false); |
73c10101 JA |
615 | return true; |
616 | } | |
617 | ||
320ae51f JA |
618 | bool bio_attempt_front_merge(struct request_queue *q, struct request *req, |
619 | struct bio *bio) | |
73c10101 | 620 | { |
1eff9d32 | 621 | const int ff = bio->bi_opf & REQ_FAILFAST_MASK; |
73c10101 | 622 | |
73c10101 JA |
623 | if (!ll_front_merge_fn(q, req, bio)) |
624 | return false; | |
625 | ||
8c1cf6bb | 626 | trace_block_bio_frontmerge(q, req, bio); |
73c10101 JA |
627 | |
628 | if ((req->cmd_flags & REQ_FAILFAST_MASK) != ff) | |
629 | blk_rq_set_mixed_merge(req); | |
630 | ||
73c10101 JA |
631 | bio->bi_next = req->bio; |
632 | req->bio = bio; | |
633 | ||
4f024f37 KO |
634 | req->__sector = bio->bi_iter.bi_sector; |
635 | req->__data_len += bio->bi_iter.bi_size; | |
73c10101 | 636 | |
320ae51f | 637 | blk_account_io_start(req, false); |
73c10101 JA |
638 | return true; |
639 | } | |
640 | ||
1e739730 CH |
641 | bool bio_attempt_discard_merge(struct request_queue *q, struct request *req, |
642 | struct bio *bio) | |
643 | { | |
644 | unsigned short segments = blk_rq_nr_discard_segments(req); | |
645 | ||
646 | if (segments >= queue_max_discard_segments(q)) | |
647 | goto no_merge; | |
648 | if (blk_rq_sectors(req) + bio_sectors(bio) > | |
649 | blk_rq_get_max_sectors(req, blk_rq_pos(req))) | |
650 | goto no_merge; | |
651 | ||
652 | req->biotail->bi_next = bio; | |
653 | req->biotail = bio; | |
654 | req->__data_len += bio->bi_iter.bi_size; | |
1e739730 CH |
655 | req->nr_phys_segments = segments + 1; |
656 | ||
657 | blk_account_io_start(req, false); | |
658 | return true; | |
659 | no_merge: | |
660 | req_set_nomerge(q, req); | |
661 | return false; | |
662 | } | |
663 | ||
bd87b589 | 664 | /** |
320ae51f | 665 | * blk_attempt_plug_merge - try to merge with %current's plugged list |
bd87b589 TH |
666 | * @q: request_queue new bio is being queued at |
667 | * @bio: new bio being queued | |
ccc2600b RD |
668 | * @same_queue_rq: pointer to &struct request that gets filled in when |
669 | * another request associated with @q is found on the plug list | |
670 | * (optional, may be %NULL) | |
bd87b589 TH |
671 | * |
672 | * Determine whether @bio being queued on @q can be merged with a request | |
673 | * on %current's plugged list. Returns %true if merge was successful, | |
674 | * otherwise %false. | |
675 | * | |
07c2bd37 TH |
676 | * Plugging coalesces IOs from the same issuer for the same purpose without |
677 | * going through @q->queue_lock. As such it's more of an issuing mechanism | |
678 | * than scheduling, and the request, while may have elvpriv data, is not | |
679 | * added on the elevator at this point. In addition, we don't have | |
680 | * reliable access to the elevator outside queue lock. Only check basic | |
681 | * merging parameters without querying the elevator. | |
da41a589 RE |
682 | * |
683 | * Caller must ensure !blk_queue_nomerges(q) beforehand. | |
73c10101 | 684 | */ |
320ae51f | 685 | bool blk_attempt_plug_merge(struct request_queue *q, struct bio *bio, |
5b3f341f | 686 | struct request **same_queue_rq) |
73c10101 JA |
687 | { |
688 | struct blk_plug *plug; | |
689 | struct request *rq; | |
92f399c7 | 690 | struct list_head *plug_list; |
73c10101 | 691 | |
bd87b589 | 692 | plug = current->plug; |
73c10101 | 693 | if (!plug) |
34fe7c05 | 694 | return false; |
73c10101 | 695 | |
a1ce35fa | 696 | plug_list = &plug->mq_list; |
92f399c7 SL |
697 | |
698 | list_for_each_entry_reverse(rq, plug_list, queuelist) { | |
34fe7c05 | 699 | bool merged = false; |
73c10101 | 700 | |
5f0ed774 | 701 | if (rq->q == q && same_queue_rq) { |
5b3f341f SL |
702 | /* |
703 | * Only blk-mq multiple hardware queues case checks the | |
704 | * rq in the same queue, there should be only one such | |
705 | * rq in a queue | |
706 | **/ | |
5f0ed774 | 707 | *same_queue_rq = rq; |
5b3f341f | 708 | } |
56ebdaf2 | 709 | |
07c2bd37 | 710 | if (rq->q != q || !blk_rq_merge_ok(rq, bio)) |
73c10101 JA |
711 | continue; |
712 | ||
34fe7c05 CH |
713 | switch (blk_try_merge(rq, bio)) { |
714 | case ELEVATOR_BACK_MERGE: | |
715 | merged = bio_attempt_back_merge(q, rq, bio); | |
716 | break; | |
717 | case ELEVATOR_FRONT_MERGE: | |
718 | merged = bio_attempt_front_merge(q, rq, bio); | |
719 | break; | |
1e739730 CH |
720 | case ELEVATOR_DISCARD_MERGE: |
721 | merged = bio_attempt_discard_merge(q, rq, bio); | |
722 | break; | |
34fe7c05 CH |
723 | default: |
724 | break; | |
73c10101 | 725 | } |
34fe7c05 CH |
726 | |
727 | if (merged) | |
728 | return true; | |
73c10101 | 729 | } |
34fe7c05 CH |
730 | |
731 | return false; | |
73c10101 JA |
732 | } |
733 | ||
da8d7f07 | 734 | void blk_init_request_from_bio(struct request *req, struct bio *bio) |
52d9e675 | 735 | { |
1eff9d32 | 736 | if (bio->bi_opf & REQ_RAHEAD) |
a82afdfc | 737 | req->cmd_flags |= REQ_FAILFAST_MASK; |
b31dc66a | 738 | |
4f024f37 | 739 | req->__sector = bio->bi_iter.bi_sector; |
20578bdf | 740 | req->ioprio = bio_prio(bio); |
cb6934f8 | 741 | req->write_hint = bio->bi_write_hint; |
bc1c56fd | 742 | blk_rq_bio_prep(req->q, req, bio); |
52d9e675 | 743 | } |
da8d7f07 | 744 | EXPORT_SYMBOL_GPL(blk_init_request_from_bio); |
52d9e675 | 745 | |
52c5e62d | 746 | static void handle_bad_sector(struct bio *bio, sector_t maxsector) |
1da177e4 LT |
747 | { |
748 | char b[BDEVNAME_SIZE]; | |
749 | ||
750 | printk(KERN_INFO "attempt to access beyond end of device\n"); | |
6296b960 | 751 | printk(KERN_INFO "%s: rw=%d, want=%Lu, limit=%Lu\n", |
74d46992 | 752 | bio_devname(bio, b), bio->bi_opf, |
f73a1c7d | 753 | (unsigned long long)bio_end_sector(bio), |
52c5e62d | 754 | (long long)maxsector); |
1da177e4 LT |
755 | } |
756 | ||
c17bb495 AM |
757 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
758 | ||
759 | static DECLARE_FAULT_ATTR(fail_make_request); | |
760 | ||
761 | static int __init setup_fail_make_request(char *str) | |
762 | { | |
763 | return setup_fault_attr(&fail_make_request, str); | |
764 | } | |
765 | __setup("fail_make_request=", setup_fail_make_request); | |
766 | ||
b2c9cd37 | 767 | static bool should_fail_request(struct hd_struct *part, unsigned int bytes) |
c17bb495 | 768 | { |
b2c9cd37 | 769 | return part->make_it_fail && should_fail(&fail_make_request, bytes); |
c17bb495 AM |
770 | } |
771 | ||
772 | static int __init fail_make_request_debugfs(void) | |
773 | { | |
dd48c085 AM |
774 | struct dentry *dir = fault_create_debugfs_attr("fail_make_request", |
775 | NULL, &fail_make_request); | |
776 | ||
21f9fcd8 | 777 | return PTR_ERR_OR_ZERO(dir); |
c17bb495 AM |
778 | } |
779 | ||
780 | late_initcall(fail_make_request_debugfs); | |
781 | ||
782 | #else /* CONFIG_FAIL_MAKE_REQUEST */ | |
783 | ||
b2c9cd37 AM |
784 | static inline bool should_fail_request(struct hd_struct *part, |
785 | unsigned int bytes) | |
c17bb495 | 786 | { |
b2c9cd37 | 787 | return false; |
c17bb495 AM |
788 | } |
789 | ||
790 | #endif /* CONFIG_FAIL_MAKE_REQUEST */ | |
791 | ||
721c7fc7 ID |
792 | static inline bool bio_check_ro(struct bio *bio, struct hd_struct *part) |
793 | { | |
b089cfd9 JA |
794 | const int op = bio_op(bio); |
795 | ||
8b2ded1c | 796 | if (part->policy && op_is_write(op)) { |
721c7fc7 ID |
797 | char b[BDEVNAME_SIZE]; |
798 | ||
8b2ded1c MP |
799 | if (op_is_flush(bio->bi_opf) && !bio_sectors(bio)) |
800 | return false; | |
801 | ||
a32e236e | 802 | WARN_ONCE(1, |
721c7fc7 ID |
803 | "generic_make_request: Trying to write " |
804 | "to read-only block-device %s (partno %d)\n", | |
805 | bio_devname(bio, b), part->partno); | |
a32e236e LT |
806 | /* Older lvm-tools actually trigger this */ |
807 | return false; | |
721c7fc7 ID |
808 | } |
809 | ||
810 | return false; | |
811 | } | |
812 | ||
30abb3a6 HM |
813 | static noinline int should_fail_bio(struct bio *bio) |
814 | { | |
815 | if (should_fail_request(&bio->bi_disk->part0, bio->bi_iter.bi_size)) | |
816 | return -EIO; | |
817 | return 0; | |
818 | } | |
819 | ALLOW_ERROR_INJECTION(should_fail_bio, ERRNO); | |
820 | ||
52c5e62d CH |
821 | /* |
822 | * Check whether this bio extends beyond the end of the device or partition. | |
823 | * This may well happen - the kernel calls bread() without checking the size of | |
824 | * the device, e.g., when mounting a file system. | |
825 | */ | |
826 | static inline int bio_check_eod(struct bio *bio, sector_t maxsector) | |
827 | { | |
828 | unsigned int nr_sectors = bio_sectors(bio); | |
829 | ||
830 | if (nr_sectors && maxsector && | |
831 | (nr_sectors > maxsector || | |
832 | bio->bi_iter.bi_sector > maxsector - nr_sectors)) { | |
833 | handle_bad_sector(bio, maxsector); | |
834 | return -EIO; | |
835 | } | |
836 | return 0; | |
837 | } | |
838 | ||
74d46992 CH |
839 | /* |
840 | * Remap block n of partition p to block n+start(p) of the disk. | |
841 | */ | |
842 | static inline int blk_partition_remap(struct bio *bio) | |
843 | { | |
844 | struct hd_struct *p; | |
52c5e62d | 845 | int ret = -EIO; |
74d46992 | 846 | |
721c7fc7 ID |
847 | rcu_read_lock(); |
848 | p = __disk_get_part(bio->bi_disk, bio->bi_partno); | |
52c5e62d CH |
849 | if (unlikely(!p)) |
850 | goto out; | |
851 | if (unlikely(should_fail_request(p, bio->bi_iter.bi_size))) | |
852 | goto out; | |
853 | if (unlikely(bio_check_ro(bio, p))) | |
721c7fc7 | 854 | goto out; |
721c7fc7 | 855 | |
74d46992 CH |
856 | /* |
857 | * Zone reset does not include bi_size so bio_sectors() is always 0. | |
858 | * Include a test for the reset op code and perform the remap if needed. | |
859 | */ | |
52c5e62d CH |
860 | if (bio_sectors(bio) || bio_op(bio) == REQ_OP_ZONE_RESET) { |
861 | if (bio_check_eod(bio, part_nr_sects_read(p))) | |
862 | goto out; | |
863 | bio->bi_iter.bi_sector += p->start_sect; | |
52c5e62d CH |
864 | trace_block_bio_remap(bio->bi_disk->queue, bio, part_devt(p), |
865 | bio->bi_iter.bi_sector - p->start_sect); | |
866 | } | |
c04fa44b | 867 | bio->bi_partno = 0; |
52c5e62d | 868 | ret = 0; |
721c7fc7 ID |
869 | out: |
870 | rcu_read_unlock(); | |
74d46992 CH |
871 | return ret; |
872 | } | |
873 | ||
27a84d54 CH |
874 | static noinline_for_stack bool |
875 | generic_make_request_checks(struct bio *bio) | |
1da177e4 | 876 | { |
165125e1 | 877 | struct request_queue *q; |
5a7bbad2 | 878 | int nr_sectors = bio_sectors(bio); |
4e4cbee9 | 879 | blk_status_t status = BLK_STS_IOERR; |
5a7bbad2 | 880 | char b[BDEVNAME_SIZE]; |
1da177e4 LT |
881 | |
882 | might_sleep(); | |
1da177e4 | 883 | |
74d46992 | 884 | q = bio->bi_disk->queue; |
5a7bbad2 CH |
885 | if (unlikely(!q)) { |
886 | printk(KERN_ERR | |
887 | "generic_make_request: Trying to access " | |
888 | "nonexistent block-device %s (%Lu)\n", | |
74d46992 | 889 | bio_devname(bio, b), (long long)bio->bi_iter.bi_sector); |
5a7bbad2 CH |
890 | goto end_io; |
891 | } | |
c17bb495 | 892 | |
03a07c92 GR |
893 | /* |
894 | * For a REQ_NOWAIT based request, return -EOPNOTSUPP | |
895 | * if queue is not a request based queue. | |
896 | */ | |
344e9ffc | 897 | if ((bio->bi_opf & REQ_NOWAIT) && !queue_is_mq(q)) |
03a07c92 GR |
898 | goto not_supported; |
899 | ||
30abb3a6 | 900 | if (should_fail_bio(bio)) |
5a7bbad2 | 901 | goto end_io; |
2056a782 | 902 | |
52c5e62d CH |
903 | if (bio->bi_partno) { |
904 | if (unlikely(blk_partition_remap(bio))) | |
721c7fc7 ID |
905 | goto end_io; |
906 | } else { | |
52c5e62d CH |
907 | if (unlikely(bio_check_ro(bio, &bio->bi_disk->part0))) |
908 | goto end_io; | |
909 | if (unlikely(bio_check_eod(bio, get_capacity(bio->bi_disk)))) | |
721c7fc7 ID |
910 | goto end_io; |
911 | } | |
2056a782 | 912 | |
5a7bbad2 CH |
913 | /* |
914 | * Filter flush bio's early so that make_request based | |
915 | * drivers without flush support don't have to worry | |
916 | * about them. | |
917 | */ | |
f3a8ab7d | 918 | if (op_is_flush(bio->bi_opf) && |
c888a8f9 | 919 | !test_bit(QUEUE_FLAG_WC, &q->queue_flags)) { |
1eff9d32 | 920 | bio->bi_opf &= ~(REQ_PREFLUSH | REQ_FUA); |
5a7bbad2 | 921 | if (!nr_sectors) { |
4e4cbee9 | 922 | status = BLK_STS_OK; |
51fd77bd JA |
923 | goto end_io; |
924 | } | |
5a7bbad2 | 925 | } |
5ddfe969 | 926 | |
d04c406f CH |
927 | if (!test_bit(QUEUE_FLAG_POLL, &q->queue_flags)) |
928 | bio->bi_opf &= ~REQ_HIPRI; | |
929 | ||
288dab8a CH |
930 | switch (bio_op(bio)) { |
931 | case REQ_OP_DISCARD: | |
932 | if (!blk_queue_discard(q)) | |
933 | goto not_supported; | |
934 | break; | |
935 | case REQ_OP_SECURE_ERASE: | |
936 | if (!blk_queue_secure_erase(q)) | |
937 | goto not_supported; | |
938 | break; | |
939 | case REQ_OP_WRITE_SAME: | |
74d46992 | 940 | if (!q->limits.max_write_same_sectors) |
288dab8a | 941 | goto not_supported; |
58886785 | 942 | break; |
2d253440 | 943 | case REQ_OP_ZONE_RESET: |
74d46992 | 944 | if (!blk_queue_is_zoned(q)) |
2d253440 | 945 | goto not_supported; |
288dab8a | 946 | break; |
a6f0788e | 947 | case REQ_OP_WRITE_ZEROES: |
74d46992 | 948 | if (!q->limits.max_write_zeroes_sectors) |
a6f0788e CK |
949 | goto not_supported; |
950 | break; | |
288dab8a CH |
951 | default: |
952 | break; | |
5a7bbad2 | 953 | } |
01edede4 | 954 | |
7f4b35d1 TH |
955 | /* |
956 | * Various block parts want %current->io_context and lazy ioc | |
957 | * allocation ends up trading a lot of pain for a small amount of | |
958 | * memory. Just allocate it upfront. This may fail and block | |
959 | * layer knows how to live with it. | |
960 | */ | |
961 | create_io_context(GFP_ATOMIC, q->node); | |
962 | ||
ae118896 TH |
963 | if (!blkcg_bio_issue_check(q, bio)) |
964 | return false; | |
27a84d54 | 965 | |
fbbaf700 N |
966 | if (!bio_flagged(bio, BIO_TRACE_COMPLETION)) { |
967 | trace_block_bio_queue(q, bio); | |
968 | /* Now that enqueuing has been traced, we need to trace | |
969 | * completion as well. | |
970 | */ | |
971 | bio_set_flag(bio, BIO_TRACE_COMPLETION); | |
972 | } | |
27a84d54 | 973 | return true; |
a7384677 | 974 | |
288dab8a | 975 | not_supported: |
4e4cbee9 | 976 | status = BLK_STS_NOTSUPP; |
a7384677 | 977 | end_io: |
4e4cbee9 | 978 | bio->bi_status = status; |
4246a0b6 | 979 | bio_endio(bio); |
27a84d54 | 980 | return false; |
1da177e4 LT |
981 | } |
982 | ||
27a84d54 CH |
983 | /** |
984 | * generic_make_request - hand a buffer to its device driver for I/O | |
985 | * @bio: The bio describing the location in memory and on the device. | |
986 | * | |
987 | * generic_make_request() is used to make I/O requests of block | |
988 | * devices. It is passed a &struct bio, which describes the I/O that needs | |
989 | * to be done. | |
990 | * | |
991 | * generic_make_request() does not return any status. The | |
992 | * success/failure status of the request, along with notification of | |
993 | * completion, is delivered asynchronously through the bio->bi_end_io | |
994 | * function described (one day) else where. | |
995 | * | |
996 | * The caller of generic_make_request must make sure that bi_io_vec | |
997 | * are set to describe the memory buffer, and that bi_dev and bi_sector are | |
998 | * set to describe the device address, and the | |
999 | * bi_end_io and optionally bi_private are set to describe how | |
1000 | * completion notification should be signaled. | |
1001 | * | |
1002 | * generic_make_request and the drivers it calls may use bi_next if this | |
1003 | * bio happens to be merged with someone else, and may resubmit the bio to | |
1004 | * a lower device by calling into generic_make_request recursively, which | |
1005 | * means the bio should NOT be touched after the call to ->make_request_fn. | |
d89d8796 | 1006 | */ |
dece1635 | 1007 | blk_qc_t generic_make_request(struct bio *bio) |
d89d8796 | 1008 | { |
f5fe1b51 N |
1009 | /* |
1010 | * bio_list_on_stack[0] contains bios submitted by the current | |
1011 | * make_request_fn. | |
1012 | * bio_list_on_stack[1] contains bios that were submitted before | |
1013 | * the current make_request_fn, but that haven't been processed | |
1014 | * yet. | |
1015 | */ | |
1016 | struct bio_list bio_list_on_stack[2]; | |
37f9579f BVA |
1017 | blk_mq_req_flags_t flags = 0; |
1018 | struct request_queue *q = bio->bi_disk->queue; | |
dece1635 | 1019 | blk_qc_t ret = BLK_QC_T_NONE; |
bddd87c7 | 1020 | |
37f9579f BVA |
1021 | if (bio->bi_opf & REQ_NOWAIT) |
1022 | flags = BLK_MQ_REQ_NOWAIT; | |
cd4a4ae4 JA |
1023 | if (bio_flagged(bio, BIO_QUEUE_ENTERED)) |
1024 | blk_queue_enter_live(q); | |
1025 | else if (blk_queue_enter(q, flags) < 0) { | |
37f9579f BVA |
1026 | if (!blk_queue_dying(q) && (bio->bi_opf & REQ_NOWAIT)) |
1027 | bio_wouldblock_error(bio); | |
1028 | else | |
1029 | bio_io_error(bio); | |
1030 | return ret; | |
1031 | } | |
1032 | ||
27a84d54 | 1033 | if (!generic_make_request_checks(bio)) |
dece1635 | 1034 | goto out; |
27a84d54 CH |
1035 | |
1036 | /* | |
1037 | * We only want one ->make_request_fn to be active at a time, else | |
1038 | * stack usage with stacked devices could be a problem. So use | |
1039 | * current->bio_list to keep a list of requests submited by a | |
1040 | * make_request_fn function. current->bio_list is also used as a | |
1041 | * flag to say if generic_make_request is currently active in this | |
1042 | * task or not. If it is NULL, then no make_request is active. If | |
1043 | * it is non-NULL, then a make_request is active, and new requests | |
1044 | * should be added at the tail | |
1045 | */ | |
bddd87c7 | 1046 | if (current->bio_list) { |
f5fe1b51 | 1047 | bio_list_add(¤t->bio_list[0], bio); |
dece1635 | 1048 | goto out; |
d89d8796 | 1049 | } |
27a84d54 | 1050 | |
d89d8796 NB |
1051 | /* following loop may be a bit non-obvious, and so deserves some |
1052 | * explanation. | |
1053 | * Before entering the loop, bio->bi_next is NULL (as all callers | |
1054 | * ensure that) so we have a list with a single bio. | |
1055 | * We pretend that we have just taken it off a longer list, so | |
bddd87c7 AM |
1056 | * we assign bio_list to a pointer to the bio_list_on_stack, |
1057 | * thus initialising the bio_list of new bios to be | |
27a84d54 | 1058 | * added. ->make_request() may indeed add some more bios |
d89d8796 NB |
1059 | * through a recursive call to generic_make_request. If it |
1060 | * did, we find a non-NULL value in bio_list and re-enter the loop | |
1061 | * from the top. In this case we really did just take the bio | |
bddd87c7 | 1062 | * of the top of the list (no pretending) and so remove it from |
27a84d54 | 1063 | * bio_list, and call into ->make_request() again. |
d89d8796 NB |
1064 | */ |
1065 | BUG_ON(bio->bi_next); | |
f5fe1b51 N |
1066 | bio_list_init(&bio_list_on_stack[0]); |
1067 | current->bio_list = bio_list_on_stack; | |
d89d8796 | 1068 | do { |
37f9579f BVA |
1069 | bool enter_succeeded = true; |
1070 | ||
1071 | if (unlikely(q != bio->bi_disk->queue)) { | |
1072 | if (q) | |
1073 | blk_queue_exit(q); | |
1074 | q = bio->bi_disk->queue; | |
1075 | flags = 0; | |
1076 | if (bio->bi_opf & REQ_NOWAIT) | |
1077 | flags = BLK_MQ_REQ_NOWAIT; | |
1078 | if (blk_queue_enter(q, flags) < 0) { | |
1079 | enter_succeeded = false; | |
1080 | q = NULL; | |
1081 | } | |
1082 | } | |
27a84d54 | 1083 | |
37f9579f | 1084 | if (enter_succeeded) { |
79bd9959 N |
1085 | struct bio_list lower, same; |
1086 | ||
1087 | /* Create a fresh bio_list for all subordinate requests */ | |
f5fe1b51 N |
1088 | bio_list_on_stack[1] = bio_list_on_stack[0]; |
1089 | bio_list_init(&bio_list_on_stack[0]); | |
dece1635 | 1090 | ret = q->make_request_fn(q, bio); |
3ef28e83 | 1091 | |
79bd9959 N |
1092 | /* sort new bios into those for a lower level |
1093 | * and those for the same level | |
1094 | */ | |
1095 | bio_list_init(&lower); | |
1096 | bio_list_init(&same); | |
f5fe1b51 | 1097 | while ((bio = bio_list_pop(&bio_list_on_stack[0])) != NULL) |
74d46992 | 1098 | if (q == bio->bi_disk->queue) |
79bd9959 N |
1099 | bio_list_add(&same, bio); |
1100 | else | |
1101 | bio_list_add(&lower, bio); | |
1102 | /* now assemble so we handle the lowest level first */ | |
f5fe1b51 N |
1103 | bio_list_merge(&bio_list_on_stack[0], &lower); |
1104 | bio_list_merge(&bio_list_on_stack[0], &same); | |
1105 | bio_list_merge(&bio_list_on_stack[0], &bio_list_on_stack[1]); | |
3ef28e83 | 1106 | } else { |
03a07c92 GR |
1107 | if (unlikely(!blk_queue_dying(q) && |
1108 | (bio->bi_opf & REQ_NOWAIT))) | |
1109 | bio_wouldblock_error(bio); | |
1110 | else | |
1111 | bio_io_error(bio); | |
3ef28e83 | 1112 | } |
f5fe1b51 | 1113 | bio = bio_list_pop(&bio_list_on_stack[0]); |
d89d8796 | 1114 | } while (bio); |
bddd87c7 | 1115 | current->bio_list = NULL; /* deactivate */ |
dece1635 JA |
1116 | |
1117 | out: | |
37f9579f BVA |
1118 | if (q) |
1119 | blk_queue_exit(q); | |
dece1635 | 1120 | return ret; |
d89d8796 | 1121 | } |
1da177e4 LT |
1122 | EXPORT_SYMBOL(generic_make_request); |
1123 | ||
f421e1d9 CH |
1124 | /** |
1125 | * direct_make_request - hand a buffer directly to its device driver for I/O | |
1126 | * @bio: The bio describing the location in memory and on the device. | |
1127 | * | |
1128 | * This function behaves like generic_make_request(), but does not protect | |
1129 | * against recursion. Must only be used if the called driver is known | |
1130 | * to not call generic_make_request (or direct_make_request) again from | |
1131 | * its make_request function. (Calling direct_make_request again from | |
1132 | * a workqueue is perfectly fine as that doesn't recurse). | |
1133 | */ | |
1134 | blk_qc_t direct_make_request(struct bio *bio) | |
1135 | { | |
1136 | struct request_queue *q = bio->bi_disk->queue; | |
1137 | bool nowait = bio->bi_opf & REQ_NOWAIT; | |
1138 | blk_qc_t ret; | |
1139 | ||
1140 | if (!generic_make_request_checks(bio)) | |
1141 | return BLK_QC_T_NONE; | |
1142 | ||
3a0a5299 | 1143 | if (unlikely(blk_queue_enter(q, nowait ? BLK_MQ_REQ_NOWAIT : 0))) { |
f421e1d9 CH |
1144 | if (nowait && !blk_queue_dying(q)) |
1145 | bio->bi_status = BLK_STS_AGAIN; | |
1146 | else | |
1147 | bio->bi_status = BLK_STS_IOERR; | |
1148 | bio_endio(bio); | |
1149 | return BLK_QC_T_NONE; | |
1150 | } | |
1151 | ||
1152 | ret = q->make_request_fn(q, bio); | |
1153 | blk_queue_exit(q); | |
1154 | return ret; | |
1155 | } | |
1156 | EXPORT_SYMBOL_GPL(direct_make_request); | |
1157 | ||
1da177e4 | 1158 | /** |
710027a4 | 1159 | * submit_bio - submit a bio to the block device layer for I/O |
1da177e4 LT |
1160 | * @bio: The &struct bio which describes the I/O |
1161 | * | |
1162 | * submit_bio() is very similar in purpose to generic_make_request(), and | |
1163 | * uses that function to do most of the work. Both are fairly rough | |
710027a4 | 1164 | * interfaces; @bio must be presetup and ready for I/O. |
1da177e4 LT |
1165 | * |
1166 | */ | |
4e49ea4a | 1167 | blk_qc_t submit_bio(struct bio *bio) |
1da177e4 | 1168 | { |
bf2de6f5 JA |
1169 | /* |
1170 | * If it's a regular read/write or a barrier with data attached, | |
1171 | * go through the normal accounting stuff before submission. | |
1172 | */ | |
e2a60da7 | 1173 | if (bio_has_data(bio)) { |
4363ac7c MP |
1174 | unsigned int count; |
1175 | ||
95fe6c1a | 1176 | if (unlikely(bio_op(bio) == REQ_OP_WRITE_SAME)) |
7c5a0dcf | 1177 | count = queue_logical_block_size(bio->bi_disk->queue) >> 9; |
4363ac7c MP |
1178 | else |
1179 | count = bio_sectors(bio); | |
1180 | ||
a8ebb056 | 1181 | if (op_is_write(bio_op(bio))) { |
bf2de6f5 JA |
1182 | count_vm_events(PGPGOUT, count); |
1183 | } else { | |
4f024f37 | 1184 | task_io_account_read(bio->bi_iter.bi_size); |
bf2de6f5 JA |
1185 | count_vm_events(PGPGIN, count); |
1186 | } | |
1187 | ||
1188 | if (unlikely(block_dump)) { | |
1189 | char b[BDEVNAME_SIZE]; | |
8dcbdc74 | 1190 | printk(KERN_DEBUG "%s(%d): %s block %Lu on %s (%u sectors)\n", |
ba25f9dc | 1191 | current->comm, task_pid_nr(current), |
a8ebb056 | 1192 | op_is_write(bio_op(bio)) ? "WRITE" : "READ", |
4f024f37 | 1193 | (unsigned long long)bio->bi_iter.bi_sector, |
74d46992 | 1194 | bio_devname(bio, b), count); |
bf2de6f5 | 1195 | } |
1da177e4 LT |
1196 | } |
1197 | ||
dece1635 | 1198 | return generic_make_request(bio); |
1da177e4 | 1199 | } |
1da177e4 LT |
1200 | EXPORT_SYMBOL(submit_bio); |
1201 | ||
82124d60 | 1202 | /** |
bf4e6b4e HR |
1203 | * blk_cloned_rq_check_limits - Helper function to check a cloned request |
1204 | * for new the queue limits | |
82124d60 KU |
1205 | * @q: the queue |
1206 | * @rq: the request being checked | |
1207 | * | |
1208 | * Description: | |
1209 | * @rq may have been made based on weaker limitations of upper-level queues | |
1210 | * in request stacking drivers, and it may violate the limitation of @q. | |
1211 | * Since the block layer and the underlying device driver trust @rq | |
1212 | * after it is inserted to @q, it should be checked against @q before | |
1213 | * the insertion using this generic function. | |
1214 | * | |
82124d60 | 1215 | * Request stacking drivers like request-based dm may change the queue |
bf4e6b4e HR |
1216 | * limits when retrying requests on other queues. Those requests need |
1217 | * to be checked against the new queue limits again during dispatch. | |
82124d60 | 1218 | */ |
bf4e6b4e HR |
1219 | static int blk_cloned_rq_check_limits(struct request_queue *q, |
1220 | struct request *rq) | |
82124d60 | 1221 | { |
8fe0d473 | 1222 | if (blk_rq_sectors(rq) > blk_queue_get_max_sectors(q, req_op(rq))) { |
82124d60 KU |
1223 | printk(KERN_ERR "%s: over max size limit.\n", __func__); |
1224 | return -EIO; | |
1225 | } | |
1226 | ||
1227 | /* | |
1228 | * queue's settings related to segment counting like q->bounce_pfn | |
1229 | * may differ from that of other stacking queues. | |
1230 | * Recalculate it to check the request correctly on this queue's | |
1231 | * limitation. | |
1232 | */ | |
1233 | blk_recalc_rq_segments(rq); | |
8a78362c | 1234 | if (rq->nr_phys_segments > queue_max_segments(q)) { |
82124d60 KU |
1235 | printk(KERN_ERR "%s: over max segments limit.\n", __func__); |
1236 | return -EIO; | |
1237 | } | |
1238 | ||
1239 | return 0; | |
1240 | } | |
82124d60 KU |
1241 | |
1242 | /** | |
1243 | * blk_insert_cloned_request - Helper for stacking drivers to submit a request | |
1244 | * @q: the queue to submit the request | |
1245 | * @rq: the request being queued | |
1246 | */ | |
2a842aca | 1247 | blk_status_t blk_insert_cloned_request(struct request_queue *q, struct request *rq) |
82124d60 | 1248 | { |
d6a51a97 JW |
1249 | blk_qc_t unused; |
1250 | ||
bf4e6b4e | 1251 | if (blk_cloned_rq_check_limits(q, rq)) |
2a842aca | 1252 | return BLK_STS_IOERR; |
82124d60 | 1253 | |
b2c9cd37 AM |
1254 | if (rq->rq_disk && |
1255 | should_fail_request(&rq->rq_disk->part0, blk_rq_bytes(rq))) | |
2a842aca | 1256 | return BLK_STS_IOERR; |
82124d60 | 1257 | |
a1ce35fa JA |
1258 | if (blk_queue_io_stat(q)) |
1259 | blk_account_io_start(rq, true); | |
82124d60 KU |
1260 | |
1261 | /* | |
a1ce35fa JA |
1262 | * Since we have a scheduler attached on the top device, |
1263 | * bypass a potential scheduler on the bottom device for | |
1264 | * insert. | |
82124d60 | 1265 | */ |
d6a51a97 | 1266 | return blk_mq_try_issue_directly(rq->mq_hctx, rq, &unused, true, true); |
82124d60 KU |
1267 | } |
1268 | EXPORT_SYMBOL_GPL(blk_insert_cloned_request); | |
1269 | ||
80a761fd TH |
1270 | /** |
1271 | * blk_rq_err_bytes - determine number of bytes till the next failure boundary | |
1272 | * @rq: request to examine | |
1273 | * | |
1274 | * Description: | |
1275 | * A request could be merge of IOs which require different failure | |
1276 | * handling. This function determines the number of bytes which | |
1277 | * can be failed from the beginning of the request without | |
1278 | * crossing into area which need to be retried further. | |
1279 | * | |
1280 | * Return: | |
1281 | * The number of bytes to fail. | |
80a761fd TH |
1282 | */ |
1283 | unsigned int blk_rq_err_bytes(const struct request *rq) | |
1284 | { | |
1285 | unsigned int ff = rq->cmd_flags & REQ_FAILFAST_MASK; | |
1286 | unsigned int bytes = 0; | |
1287 | struct bio *bio; | |
1288 | ||
e8064021 | 1289 | if (!(rq->rq_flags & RQF_MIXED_MERGE)) |
80a761fd TH |
1290 | return blk_rq_bytes(rq); |
1291 | ||
1292 | /* | |
1293 | * Currently the only 'mixing' which can happen is between | |
1294 | * different fastfail types. We can safely fail portions | |
1295 | * which have all the failfast bits that the first one has - | |
1296 | * the ones which are at least as eager to fail as the first | |
1297 | * one. | |
1298 | */ | |
1299 | for (bio = rq->bio; bio; bio = bio->bi_next) { | |
1eff9d32 | 1300 | if ((bio->bi_opf & ff) != ff) |
80a761fd | 1301 | break; |
4f024f37 | 1302 | bytes += bio->bi_iter.bi_size; |
80a761fd TH |
1303 | } |
1304 | ||
1305 | /* this could lead to infinite loop */ | |
1306 | BUG_ON(blk_rq_bytes(rq) && !bytes); | |
1307 | return bytes; | |
1308 | } | |
1309 | EXPORT_SYMBOL_GPL(blk_rq_err_bytes); | |
1310 | ||
320ae51f | 1311 | void blk_account_io_completion(struct request *req, unsigned int bytes) |
bc58ba94 | 1312 | { |
c2553b58 | 1313 | if (blk_do_io_stat(req)) { |
ddcf35d3 | 1314 | const int sgrp = op_stat_group(req_op(req)); |
bc58ba94 | 1315 | struct hd_struct *part; |
bc58ba94 | 1316 | |
112f158f | 1317 | part_stat_lock(); |
09e099d4 | 1318 | part = req->part; |
112f158f | 1319 | part_stat_add(part, sectors[sgrp], bytes >> 9); |
bc58ba94 JA |
1320 | part_stat_unlock(); |
1321 | } | |
1322 | } | |
1323 | ||
522a7775 | 1324 | void blk_account_io_done(struct request *req, u64 now) |
bc58ba94 | 1325 | { |
bc58ba94 | 1326 | /* |
dd4c133f TH |
1327 | * Account IO completion. flush_rq isn't accounted as a |
1328 | * normal IO on queueing nor completion. Accounting the | |
1329 | * containing request is enough. | |
bc58ba94 | 1330 | */ |
e8064021 | 1331 | if (blk_do_io_stat(req) && !(req->rq_flags & RQF_FLUSH_SEQ)) { |
ddcf35d3 | 1332 | const int sgrp = op_stat_group(req_op(req)); |
bc58ba94 | 1333 | struct hd_struct *part; |
bc58ba94 | 1334 | |
112f158f | 1335 | part_stat_lock(); |
09e099d4 | 1336 | part = req->part; |
bc58ba94 | 1337 | |
5b18b5a7 | 1338 | update_io_ticks(part, jiffies); |
112f158f MS |
1339 | part_stat_inc(part, ios[sgrp]); |
1340 | part_stat_add(part, nsecs[sgrp], now - req->start_time_ns); | |
5b18b5a7 | 1341 | part_stat_add(part, time_in_queue, nsecs_to_jiffies64(now - req->start_time_ns)); |
ddcf35d3 | 1342 | part_dec_in_flight(req->q, part, rq_data_dir(req)); |
bc58ba94 | 1343 | |
6c23a968 | 1344 | hd_struct_put(part); |
bc58ba94 JA |
1345 | part_stat_unlock(); |
1346 | } | |
1347 | } | |
1348 | ||
320ae51f JA |
1349 | void blk_account_io_start(struct request *rq, bool new_io) |
1350 | { | |
1351 | struct hd_struct *part; | |
1352 | int rw = rq_data_dir(rq); | |
320ae51f JA |
1353 | |
1354 | if (!blk_do_io_stat(rq)) | |
1355 | return; | |
1356 | ||
112f158f | 1357 | part_stat_lock(); |
320ae51f JA |
1358 | |
1359 | if (!new_io) { | |
1360 | part = rq->part; | |
112f158f | 1361 | part_stat_inc(part, merges[rw]); |
320ae51f JA |
1362 | } else { |
1363 | part = disk_map_sector_rcu(rq->rq_disk, blk_rq_pos(rq)); | |
1364 | if (!hd_struct_try_get(part)) { | |
1365 | /* | |
1366 | * The partition is already being removed, | |
1367 | * the request will be accounted on the disk only | |
1368 | * | |
1369 | * We take a reference on disk->part0 although that | |
1370 | * partition will never be deleted, so we can treat | |
1371 | * it as any other partition. | |
1372 | */ | |
1373 | part = &rq->rq_disk->part0; | |
1374 | hd_struct_get(part); | |
1375 | } | |
d62e26b3 | 1376 | part_inc_in_flight(rq->q, part, rw); |
320ae51f JA |
1377 | rq->part = part; |
1378 | } | |
1379 | ||
5b18b5a7 MP |
1380 | update_io_ticks(part, jiffies); |
1381 | ||
320ae51f JA |
1382 | part_stat_unlock(); |
1383 | } | |
1384 | ||
ef71de8b CH |
1385 | /* |
1386 | * Steal bios from a request and add them to a bio list. | |
1387 | * The request must not have been partially completed before. | |
1388 | */ | |
1389 | void blk_steal_bios(struct bio_list *list, struct request *rq) | |
1390 | { | |
1391 | if (rq->bio) { | |
1392 | if (list->tail) | |
1393 | list->tail->bi_next = rq->bio; | |
1394 | else | |
1395 | list->head = rq->bio; | |
1396 | list->tail = rq->biotail; | |
1397 | ||
1398 | rq->bio = NULL; | |
1399 | rq->biotail = NULL; | |
1400 | } | |
1401 | ||
1402 | rq->__data_len = 0; | |
1403 | } | |
1404 | EXPORT_SYMBOL_GPL(blk_steal_bios); | |
1405 | ||
3bcddeac | 1406 | /** |
2e60e022 | 1407 | * blk_update_request - Special helper function for request stacking drivers |
8ebf9756 | 1408 | * @req: the request being processed |
2a842aca | 1409 | * @error: block status code |
8ebf9756 | 1410 | * @nr_bytes: number of bytes to complete @req |
3bcddeac KU |
1411 | * |
1412 | * Description: | |
8ebf9756 RD |
1413 | * Ends I/O on a number of bytes attached to @req, but doesn't complete |
1414 | * the request structure even if @req doesn't have leftover. | |
1415 | * If @req has leftover, sets it up for the next range of segments. | |
2e60e022 TH |
1416 | * |
1417 | * This special helper function is only for request stacking drivers | |
1418 | * (e.g. request-based dm) so that they can handle partial completion. | |
1419 | * Actual device drivers should use blk_end_request instead. | |
1420 | * | |
1421 | * Passing the result of blk_rq_bytes() as @nr_bytes guarantees | |
1422 | * %false return from this function. | |
3bcddeac | 1423 | * |
1954e9a9 BVA |
1424 | * Note: |
1425 | * The RQF_SPECIAL_PAYLOAD flag is ignored on purpose in both | |
1426 | * blk_rq_bytes() and in blk_update_request(). | |
1427 | * | |
3bcddeac | 1428 | * Return: |
2e60e022 TH |
1429 | * %false - this request doesn't have any more data |
1430 | * %true - this request has more data | |
3bcddeac | 1431 | **/ |
2a842aca CH |
1432 | bool blk_update_request(struct request *req, blk_status_t error, |
1433 | unsigned int nr_bytes) | |
1da177e4 | 1434 | { |
f79ea416 | 1435 | int total_bytes; |
1da177e4 | 1436 | |
2a842aca | 1437 | trace_block_rq_complete(req, blk_status_to_errno(error), nr_bytes); |
4a0efdc9 | 1438 | |
2e60e022 TH |
1439 | if (!req->bio) |
1440 | return false; | |
1441 | ||
2a842aca CH |
1442 | if (unlikely(error && !blk_rq_is_passthrough(req) && |
1443 | !(req->rq_flags & RQF_QUIET))) | |
1444 | print_req_error(req, error); | |
1da177e4 | 1445 | |
bc58ba94 | 1446 | blk_account_io_completion(req, nr_bytes); |
d72d904a | 1447 | |
f79ea416 KO |
1448 | total_bytes = 0; |
1449 | while (req->bio) { | |
1450 | struct bio *bio = req->bio; | |
4f024f37 | 1451 | unsigned bio_bytes = min(bio->bi_iter.bi_size, nr_bytes); |
1da177e4 | 1452 | |
9c24c10a | 1453 | if (bio_bytes == bio->bi_iter.bi_size) |
1da177e4 | 1454 | req->bio = bio->bi_next; |
1da177e4 | 1455 | |
fbbaf700 N |
1456 | /* Completion has already been traced */ |
1457 | bio_clear_flag(bio, BIO_TRACE_COMPLETION); | |
f79ea416 | 1458 | req_bio_endio(req, bio, bio_bytes, error); |
1da177e4 | 1459 | |
f79ea416 KO |
1460 | total_bytes += bio_bytes; |
1461 | nr_bytes -= bio_bytes; | |
1da177e4 | 1462 | |
f79ea416 KO |
1463 | if (!nr_bytes) |
1464 | break; | |
1da177e4 LT |
1465 | } |
1466 | ||
1467 | /* | |
1468 | * completely done | |
1469 | */ | |
2e60e022 TH |
1470 | if (!req->bio) { |
1471 | /* | |
1472 | * Reset counters so that the request stacking driver | |
1473 | * can find how many bytes remain in the request | |
1474 | * later. | |
1475 | */ | |
a2dec7b3 | 1476 | req->__data_len = 0; |
2e60e022 TH |
1477 | return false; |
1478 | } | |
1da177e4 | 1479 | |
a2dec7b3 | 1480 | req->__data_len -= total_bytes; |
2e46e8b2 TH |
1481 | |
1482 | /* update sector only for requests with clear definition of sector */ | |
57292b58 | 1483 | if (!blk_rq_is_passthrough(req)) |
a2dec7b3 | 1484 | req->__sector += total_bytes >> 9; |
2e46e8b2 | 1485 | |
80a761fd | 1486 | /* mixed attributes always follow the first bio */ |
e8064021 | 1487 | if (req->rq_flags & RQF_MIXED_MERGE) { |
80a761fd | 1488 | req->cmd_flags &= ~REQ_FAILFAST_MASK; |
1eff9d32 | 1489 | req->cmd_flags |= req->bio->bi_opf & REQ_FAILFAST_MASK; |
80a761fd TH |
1490 | } |
1491 | ||
ed6565e7 CH |
1492 | if (!(req->rq_flags & RQF_SPECIAL_PAYLOAD)) { |
1493 | /* | |
1494 | * If total number of sectors is less than the first segment | |
1495 | * size, something has gone terribly wrong. | |
1496 | */ | |
1497 | if (blk_rq_bytes(req) < blk_rq_cur_bytes(req)) { | |
1498 | blk_dump_rq_flags(req, "request botched"); | |
1499 | req->__data_len = blk_rq_cur_bytes(req); | |
1500 | } | |
2e46e8b2 | 1501 | |
ed6565e7 CH |
1502 | /* recalculate the number of segments */ |
1503 | blk_recalc_rq_segments(req); | |
1504 | } | |
2e46e8b2 | 1505 | |
2e60e022 | 1506 | return true; |
1da177e4 | 1507 | } |
2e60e022 | 1508 | EXPORT_SYMBOL_GPL(blk_update_request); |
1da177e4 | 1509 | |
86db1e29 JA |
1510 | void blk_rq_bio_prep(struct request_queue *q, struct request *rq, |
1511 | struct bio *bio) | |
1da177e4 | 1512 | { |
b4f42e28 | 1513 | if (bio_has_data(bio)) |
fb2dce86 | 1514 | rq->nr_phys_segments = bio_phys_segments(q, bio); |
445251d0 JA |
1515 | else if (bio_op(bio) == REQ_OP_DISCARD) |
1516 | rq->nr_phys_segments = 1; | |
b4f42e28 | 1517 | |
4f024f37 | 1518 | rq->__data_len = bio->bi_iter.bi_size; |
1da177e4 | 1519 | rq->bio = rq->biotail = bio; |
1da177e4 | 1520 | |
74d46992 CH |
1521 | if (bio->bi_disk) |
1522 | rq->rq_disk = bio->bi_disk; | |
66846572 | 1523 | } |
1da177e4 | 1524 | |
2d4dc890 IL |
1525 | #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE |
1526 | /** | |
1527 | * rq_flush_dcache_pages - Helper function to flush all pages in a request | |
1528 | * @rq: the request to be flushed | |
1529 | * | |
1530 | * Description: | |
1531 | * Flush all pages in @rq. | |
1532 | */ | |
1533 | void rq_flush_dcache_pages(struct request *rq) | |
1534 | { | |
1535 | struct req_iterator iter; | |
7988613b | 1536 | struct bio_vec bvec; |
2d4dc890 IL |
1537 | |
1538 | rq_for_each_segment(bvec, rq, iter) | |
7988613b | 1539 | flush_dcache_page(bvec.bv_page); |
2d4dc890 IL |
1540 | } |
1541 | EXPORT_SYMBOL_GPL(rq_flush_dcache_pages); | |
1542 | #endif | |
1543 | ||
ef9e3fac KU |
1544 | /** |
1545 | * blk_lld_busy - Check if underlying low-level drivers of a device are busy | |
1546 | * @q : the queue of the device being checked | |
1547 | * | |
1548 | * Description: | |
1549 | * Check if underlying low-level drivers of a device are busy. | |
1550 | * If the drivers want to export their busy state, they must set own | |
1551 | * exporting function using blk_queue_lld_busy() first. | |
1552 | * | |
1553 | * Basically, this function is used only by request stacking drivers | |
1554 | * to stop dispatching requests to underlying devices when underlying | |
1555 | * devices are busy. This behavior helps more I/O merging on the queue | |
1556 | * of the request stacking driver and prevents I/O throughput regression | |
1557 | * on burst I/O load. | |
1558 | * | |
1559 | * Return: | |
1560 | * 0 - Not busy (The request stacking driver should dispatch request) | |
1561 | * 1 - Busy (The request stacking driver should stop dispatching request) | |
1562 | */ | |
1563 | int blk_lld_busy(struct request_queue *q) | |
1564 | { | |
344e9ffc | 1565 | if (queue_is_mq(q) && q->mq_ops->busy) |
9ba20527 | 1566 | return q->mq_ops->busy(q); |
ef9e3fac KU |
1567 | |
1568 | return 0; | |
1569 | } | |
1570 | EXPORT_SYMBOL_GPL(blk_lld_busy); | |
1571 | ||
78d8e58a MS |
1572 | /** |
1573 | * blk_rq_unprep_clone - Helper function to free all bios in a cloned request | |
1574 | * @rq: the clone request to be cleaned up | |
1575 | * | |
1576 | * Description: | |
1577 | * Free all bios in @rq for a cloned request. | |
1578 | */ | |
1579 | void blk_rq_unprep_clone(struct request *rq) | |
1580 | { | |
1581 | struct bio *bio; | |
1582 | ||
1583 | while ((bio = rq->bio) != NULL) { | |
1584 | rq->bio = bio->bi_next; | |
1585 | ||
1586 | bio_put(bio); | |
1587 | } | |
1588 | } | |
1589 | EXPORT_SYMBOL_GPL(blk_rq_unprep_clone); | |
1590 | ||
1591 | /* | |
1592 | * Copy attributes of the original request to the clone request. | |
1593 | * The actual data parts (e.g. ->cmd, ->sense) are not copied. | |
1594 | */ | |
1595 | static void __blk_rq_prep_clone(struct request *dst, struct request *src) | |
b0fd271d | 1596 | { |
b0fd271d KU |
1597 | dst->__sector = blk_rq_pos(src); |
1598 | dst->__data_len = blk_rq_bytes(src); | |
297ba57d BVA |
1599 | if (src->rq_flags & RQF_SPECIAL_PAYLOAD) { |
1600 | dst->rq_flags |= RQF_SPECIAL_PAYLOAD; | |
1601 | dst->special_vec = src->special_vec; | |
1602 | } | |
b0fd271d KU |
1603 | dst->nr_phys_segments = src->nr_phys_segments; |
1604 | dst->ioprio = src->ioprio; | |
1605 | dst->extra_len = src->extra_len; | |
78d8e58a MS |
1606 | } |
1607 | ||
1608 | /** | |
1609 | * blk_rq_prep_clone - Helper function to setup clone request | |
1610 | * @rq: the request to be setup | |
1611 | * @rq_src: original request to be cloned | |
1612 | * @bs: bio_set that bios for clone are allocated from | |
1613 | * @gfp_mask: memory allocation mask for bio | |
1614 | * @bio_ctr: setup function to be called for each clone bio. | |
1615 | * Returns %0 for success, non %0 for failure. | |
1616 | * @data: private data to be passed to @bio_ctr | |
1617 | * | |
1618 | * Description: | |
1619 | * Clones bios in @rq_src to @rq, and copies attributes of @rq_src to @rq. | |
1620 | * The actual data parts of @rq_src (e.g. ->cmd, ->sense) | |
1621 | * are not copied, and copying such parts is the caller's responsibility. | |
1622 | * Also, pages which the original bios are pointing to are not copied | |
1623 | * and the cloned bios just point same pages. | |
1624 | * So cloned bios must be completed before original bios, which means | |
1625 | * the caller must complete @rq before @rq_src. | |
1626 | */ | |
1627 | int blk_rq_prep_clone(struct request *rq, struct request *rq_src, | |
1628 | struct bio_set *bs, gfp_t gfp_mask, | |
1629 | int (*bio_ctr)(struct bio *, struct bio *, void *), | |
1630 | void *data) | |
1631 | { | |
1632 | struct bio *bio, *bio_src; | |
1633 | ||
1634 | if (!bs) | |
f4f8154a | 1635 | bs = &fs_bio_set; |
78d8e58a MS |
1636 | |
1637 | __rq_for_each_bio(bio_src, rq_src) { | |
1638 | bio = bio_clone_fast(bio_src, gfp_mask, bs); | |
1639 | if (!bio) | |
1640 | goto free_and_out; | |
1641 | ||
1642 | if (bio_ctr && bio_ctr(bio, bio_src, data)) | |
1643 | goto free_and_out; | |
1644 | ||
1645 | if (rq->bio) { | |
1646 | rq->biotail->bi_next = bio; | |
1647 | rq->biotail = bio; | |
1648 | } else | |
1649 | rq->bio = rq->biotail = bio; | |
1650 | } | |
1651 | ||
1652 | __blk_rq_prep_clone(rq, rq_src); | |
1653 | ||
1654 | return 0; | |
1655 | ||
1656 | free_and_out: | |
1657 | if (bio) | |
1658 | bio_put(bio); | |
1659 | blk_rq_unprep_clone(rq); | |
1660 | ||
1661 | return -ENOMEM; | |
b0fd271d KU |
1662 | } |
1663 | EXPORT_SYMBOL_GPL(blk_rq_prep_clone); | |
1664 | ||
59c3d45e | 1665 | int kblockd_schedule_work(struct work_struct *work) |
1da177e4 LT |
1666 | { |
1667 | return queue_work(kblockd_workqueue, work); | |
1668 | } | |
1da177e4 LT |
1669 | EXPORT_SYMBOL(kblockd_schedule_work); |
1670 | ||
ee63cfa7 JA |
1671 | int kblockd_schedule_work_on(int cpu, struct work_struct *work) |
1672 | { | |
1673 | return queue_work_on(cpu, kblockd_workqueue, work); | |
1674 | } | |
1675 | EXPORT_SYMBOL(kblockd_schedule_work_on); | |
1676 | ||
818cd1cb JA |
1677 | int kblockd_mod_delayed_work_on(int cpu, struct delayed_work *dwork, |
1678 | unsigned long delay) | |
1679 | { | |
1680 | return mod_delayed_work_on(cpu, kblockd_workqueue, dwork, delay); | |
1681 | } | |
1682 | EXPORT_SYMBOL(kblockd_mod_delayed_work_on); | |
1683 | ||
75df7136 SJ |
1684 | /** |
1685 | * blk_start_plug - initialize blk_plug and track it inside the task_struct | |
1686 | * @plug: The &struct blk_plug that needs to be initialized | |
1687 | * | |
1688 | * Description: | |
40405851 JM |
1689 | * blk_start_plug() indicates to the block layer an intent by the caller |
1690 | * to submit multiple I/O requests in a batch. The block layer may use | |
1691 | * this hint to defer submitting I/Os from the caller until blk_finish_plug() | |
1692 | * is called. However, the block layer may choose to submit requests | |
1693 | * before a call to blk_finish_plug() if the number of queued I/Os | |
1694 | * exceeds %BLK_MAX_REQUEST_COUNT, or if the size of the I/O is larger than | |
1695 | * %BLK_PLUG_FLUSH_SIZE. The queued I/Os may also be submitted early if | |
1696 | * the task schedules (see below). | |
1697 | * | |
75df7136 SJ |
1698 | * Tracking blk_plug inside the task_struct will help with auto-flushing the |
1699 | * pending I/O should the task end up blocking between blk_start_plug() and | |
1700 | * blk_finish_plug(). This is important from a performance perspective, but | |
1701 | * also ensures that we don't deadlock. For instance, if the task is blocking | |
1702 | * for a memory allocation, memory reclaim could end up wanting to free a | |
1703 | * page belonging to that request that is currently residing in our private | |
1704 | * plug. By flushing the pending I/O when the process goes to sleep, we avoid | |
1705 | * this kind of deadlock. | |
1706 | */ | |
73c10101 JA |
1707 | void blk_start_plug(struct blk_plug *plug) |
1708 | { | |
1709 | struct task_struct *tsk = current; | |
1710 | ||
dd6cf3e1 SL |
1711 | /* |
1712 | * If this is a nested plug, don't actually assign it. | |
1713 | */ | |
1714 | if (tsk->plug) | |
1715 | return; | |
1716 | ||
320ae51f | 1717 | INIT_LIST_HEAD(&plug->mq_list); |
048c9374 | 1718 | INIT_LIST_HEAD(&plug->cb_list); |
5f0ed774 | 1719 | plug->rq_count = 0; |
ce5b009c | 1720 | plug->multiple_queues = false; |
5f0ed774 | 1721 | |
73c10101 | 1722 | /* |
dd6cf3e1 SL |
1723 | * Store ordering should not be needed here, since a potential |
1724 | * preempt will imply a full memory barrier | |
73c10101 | 1725 | */ |
dd6cf3e1 | 1726 | tsk->plug = plug; |
73c10101 JA |
1727 | } |
1728 | EXPORT_SYMBOL(blk_start_plug); | |
1729 | ||
74018dc3 | 1730 | static void flush_plug_callbacks(struct blk_plug *plug, bool from_schedule) |
048c9374 N |
1731 | { |
1732 | LIST_HEAD(callbacks); | |
1733 | ||
2a7d5559 SL |
1734 | while (!list_empty(&plug->cb_list)) { |
1735 | list_splice_init(&plug->cb_list, &callbacks); | |
048c9374 | 1736 | |
2a7d5559 SL |
1737 | while (!list_empty(&callbacks)) { |
1738 | struct blk_plug_cb *cb = list_first_entry(&callbacks, | |
048c9374 N |
1739 | struct blk_plug_cb, |
1740 | list); | |
2a7d5559 | 1741 | list_del(&cb->list); |
74018dc3 | 1742 | cb->callback(cb, from_schedule); |
2a7d5559 | 1743 | } |
048c9374 N |
1744 | } |
1745 | } | |
1746 | ||
9cbb1750 N |
1747 | struct blk_plug_cb *blk_check_plugged(blk_plug_cb_fn unplug, void *data, |
1748 | int size) | |
1749 | { | |
1750 | struct blk_plug *plug = current->plug; | |
1751 | struct blk_plug_cb *cb; | |
1752 | ||
1753 | if (!plug) | |
1754 | return NULL; | |
1755 | ||
1756 | list_for_each_entry(cb, &plug->cb_list, list) | |
1757 | if (cb->callback == unplug && cb->data == data) | |
1758 | return cb; | |
1759 | ||
1760 | /* Not currently on the callback list */ | |
1761 | BUG_ON(size < sizeof(*cb)); | |
1762 | cb = kzalloc(size, GFP_ATOMIC); | |
1763 | if (cb) { | |
1764 | cb->data = data; | |
1765 | cb->callback = unplug; | |
1766 | list_add(&cb->list, &plug->cb_list); | |
1767 | } | |
1768 | return cb; | |
1769 | } | |
1770 | EXPORT_SYMBOL(blk_check_plugged); | |
1771 | ||
49cac01e | 1772 | void blk_flush_plug_list(struct blk_plug *plug, bool from_schedule) |
73c10101 | 1773 | { |
74018dc3 | 1774 | flush_plug_callbacks(plug, from_schedule); |
320ae51f JA |
1775 | |
1776 | if (!list_empty(&plug->mq_list)) | |
1777 | blk_mq_flush_plug_list(plug, from_schedule); | |
73c10101 | 1778 | } |
73c10101 | 1779 | |
40405851 JM |
1780 | /** |
1781 | * blk_finish_plug - mark the end of a batch of submitted I/O | |
1782 | * @plug: The &struct blk_plug passed to blk_start_plug() | |
1783 | * | |
1784 | * Description: | |
1785 | * Indicate that a batch of I/O submissions is complete. This function | |
1786 | * must be paired with an initial call to blk_start_plug(). The intent | |
1787 | * is to allow the block layer to optimize I/O submission. See the | |
1788 | * documentation for blk_start_plug() for more information. | |
1789 | */ | |
73c10101 JA |
1790 | void blk_finish_plug(struct blk_plug *plug) |
1791 | { | |
dd6cf3e1 SL |
1792 | if (plug != current->plug) |
1793 | return; | |
f6603783 | 1794 | blk_flush_plug_list(plug, false); |
73c10101 | 1795 | |
dd6cf3e1 | 1796 | current->plug = NULL; |
73c10101 | 1797 | } |
88b996cd | 1798 | EXPORT_SYMBOL(blk_finish_plug); |
73c10101 | 1799 | |
1da177e4 LT |
1800 | int __init blk_dev_init(void) |
1801 | { | |
ef295ecf CH |
1802 | BUILD_BUG_ON(REQ_OP_LAST >= (1 << REQ_OP_BITS)); |
1803 | BUILD_BUG_ON(REQ_OP_BITS + REQ_FLAG_BITS > 8 * | |
0762b23d | 1804 | FIELD_SIZEOF(struct request, cmd_flags)); |
ef295ecf CH |
1805 | BUILD_BUG_ON(REQ_OP_BITS + REQ_FLAG_BITS > 8 * |
1806 | FIELD_SIZEOF(struct bio, bi_opf)); | |
9eb55b03 | 1807 | |
89b90be2 TH |
1808 | /* used for unplugging and affects IO latency/throughput - HIGHPRI */ |
1809 | kblockd_workqueue = alloc_workqueue("kblockd", | |
28747fcd | 1810 | WQ_MEM_RECLAIM | WQ_HIGHPRI, 0); |
1da177e4 LT |
1811 | if (!kblockd_workqueue) |
1812 | panic("Failed to create kblockd\n"); | |
1813 | ||
c2789bd4 | 1814 | blk_requestq_cachep = kmem_cache_create("request_queue", |
165125e1 | 1815 | sizeof(struct request_queue), 0, SLAB_PANIC, NULL); |
1da177e4 | 1816 | |
18fbda91 OS |
1817 | #ifdef CONFIG_DEBUG_FS |
1818 | blk_debugfs_root = debugfs_create_dir("block", NULL); | |
1819 | #endif | |
1820 | ||
d38ecf93 | 1821 | return 0; |
1da177e4 | 1822 | } |