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> | |
19 | #include <linux/highmem.h> | |
20 | #include <linux/mm.h> | |
21 | #include <linux/kernel_stat.h> | |
22 | #include <linux/string.h> | |
23 | #include <linux/init.h> | |
1da177e4 LT |
24 | #include <linux/completion.h> |
25 | #include <linux/slab.h> | |
26 | #include <linux/swap.h> | |
27 | #include <linux/writeback.h> | |
faccbd4b | 28 | #include <linux/task_io_accounting_ops.h> |
ff856bad JA |
29 | #include <linux/interrupt.h> |
30 | #include <linux/cpu.h> | |
2056a782 | 31 | #include <linux/blktrace_api.h> |
c17bb495 | 32 | #include <linux/fault-inject.h> |
1da177e4 | 33 | |
8324aa91 JA |
34 | #include "blk.h" |
35 | ||
165125e1 | 36 | static int __make_request(struct request_queue *q, struct bio *bio); |
1da177e4 LT |
37 | |
38 | /* | |
39 | * For the allocated request tables | |
40 | */ | |
5ece6c52 | 41 | static struct kmem_cache *request_cachep; |
1da177e4 LT |
42 | |
43 | /* | |
44 | * For queue allocation | |
45 | */ | |
6728cb0e | 46 | struct kmem_cache *blk_requestq_cachep; |
1da177e4 | 47 | |
1da177e4 LT |
48 | /* |
49 | * Controlling structure to kblockd | |
50 | */ | |
ff856bad | 51 | static struct workqueue_struct *kblockd_workqueue; |
1da177e4 | 52 | |
ff856bad JA |
53 | static DEFINE_PER_CPU(struct list_head, blk_cpu_done); |
54 | ||
26b8256e JA |
55 | static void drive_stat_acct(struct request *rq, int new_io) |
56 | { | |
28f13702 | 57 | struct hd_struct *part; |
26b8256e | 58 | int rw = rq_data_dir(rq); |
c9959059 | 59 | int cpu; |
26b8256e JA |
60 | |
61 | if (!blk_fs_request(rq) || !rq->rq_disk) | |
62 | return; | |
63 | ||
c9959059 | 64 | cpu = disk_stat_lock(); |
e71bf0d0 | 65 | part = disk_map_sector_rcu(rq->rq_disk, rq->sector); |
c9959059 | 66 | |
28f13702 | 67 | if (!new_io) |
c9959059 | 68 | all_stat_inc(cpu, rq->rq_disk, part, merges[rw], rq->sector); |
28f13702 | 69 | else { |
c9959059 | 70 | disk_round_stats(cpu, rq->rq_disk); |
26b8256e | 71 | rq->rq_disk->in_flight++; |
6f2576af | 72 | if (part) { |
c9959059 | 73 | part_round_stats(cpu, part); |
6f2576af JM |
74 | part->in_flight++; |
75 | } | |
26b8256e | 76 | } |
e71bf0d0 | 77 | |
c9959059 | 78 | disk_stat_unlock(); |
26b8256e JA |
79 | } |
80 | ||
8324aa91 | 81 | void blk_queue_congestion_threshold(struct request_queue *q) |
1da177e4 LT |
82 | { |
83 | int nr; | |
84 | ||
85 | nr = q->nr_requests - (q->nr_requests / 8) + 1; | |
86 | if (nr > q->nr_requests) | |
87 | nr = q->nr_requests; | |
88 | q->nr_congestion_on = nr; | |
89 | ||
90 | nr = q->nr_requests - (q->nr_requests / 8) - (q->nr_requests / 16) - 1; | |
91 | if (nr < 1) | |
92 | nr = 1; | |
93 | q->nr_congestion_off = nr; | |
94 | } | |
95 | ||
1da177e4 LT |
96 | /** |
97 | * blk_get_backing_dev_info - get the address of a queue's backing_dev_info | |
98 | * @bdev: device | |
99 | * | |
100 | * Locates the passed device's request queue and returns the address of its | |
101 | * backing_dev_info | |
102 | * | |
103 | * Will return NULL if the request queue cannot be located. | |
104 | */ | |
105 | struct backing_dev_info *blk_get_backing_dev_info(struct block_device *bdev) | |
106 | { | |
107 | struct backing_dev_info *ret = NULL; | |
165125e1 | 108 | struct request_queue *q = bdev_get_queue(bdev); |
1da177e4 LT |
109 | |
110 | if (q) | |
111 | ret = &q->backing_dev_info; | |
112 | return ret; | |
113 | } | |
1da177e4 LT |
114 | EXPORT_SYMBOL(blk_get_backing_dev_info); |
115 | ||
2a4aa30c | 116 | void blk_rq_init(struct request_queue *q, struct request *rq) |
1da177e4 | 117 | { |
1afb20f3 FT |
118 | memset(rq, 0, sizeof(*rq)); |
119 | ||
1da177e4 | 120 | INIT_LIST_HEAD(&rq->queuelist); |
ff856bad | 121 | INIT_LIST_HEAD(&rq->donelist); |
63a71386 JA |
122 | rq->q = q; |
123 | rq->sector = rq->hard_sector = (sector_t) -1; | |
2e662b65 JA |
124 | INIT_HLIST_NODE(&rq->hash); |
125 | RB_CLEAR_NODE(&rq->rb_node); | |
d7e3c324 | 126 | rq->cmd = rq->__cmd; |
63a71386 | 127 | rq->tag = -1; |
1da177e4 | 128 | rq->ref_count = 1; |
1da177e4 | 129 | } |
2a4aa30c | 130 | EXPORT_SYMBOL(blk_rq_init); |
1da177e4 | 131 | |
5bb23a68 N |
132 | static void req_bio_endio(struct request *rq, struct bio *bio, |
133 | unsigned int nbytes, int error) | |
1da177e4 | 134 | { |
165125e1 | 135 | struct request_queue *q = rq->q; |
797e7dbb | 136 | |
5bb23a68 N |
137 | if (&q->bar_rq != rq) { |
138 | if (error) | |
139 | clear_bit(BIO_UPTODATE, &bio->bi_flags); | |
140 | else if (!test_bit(BIO_UPTODATE, &bio->bi_flags)) | |
141 | error = -EIO; | |
797e7dbb | 142 | |
5bb23a68 | 143 | if (unlikely(nbytes > bio->bi_size)) { |
6728cb0e | 144 | printk(KERN_ERR "%s: want %u bytes done, %u left\n", |
24c03d47 | 145 | __func__, nbytes, bio->bi_size); |
5bb23a68 N |
146 | nbytes = bio->bi_size; |
147 | } | |
797e7dbb | 148 | |
5bb23a68 N |
149 | bio->bi_size -= nbytes; |
150 | bio->bi_sector += (nbytes >> 9); | |
7ba1ba12 MP |
151 | |
152 | if (bio_integrity(bio)) | |
153 | bio_integrity_advance(bio, nbytes); | |
154 | ||
5bb23a68 | 155 | if (bio->bi_size == 0) |
6712ecf8 | 156 | bio_endio(bio, error); |
5bb23a68 N |
157 | } else { |
158 | ||
159 | /* | |
160 | * Okay, this is the barrier request in progress, just | |
161 | * record the error; | |
162 | */ | |
163 | if (error && !q->orderr) | |
164 | q->orderr = error; | |
165 | } | |
1da177e4 | 166 | } |
1da177e4 | 167 | |
1da177e4 LT |
168 | void blk_dump_rq_flags(struct request *rq, char *msg) |
169 | { | |
170 | int bit; | |
171 | ||
6728cb0e | 172 | printk(KERN_INFO "%s: dev %s: type=%x, flags=%x\n", msg, |
4aff5e23 JA |
173 | rq->rq_disk ? rq->rq_disk->disk_name : "?", rq->cmd_type, |
174 | rq->cmd_flags); | |
1da177e4 | 175 | |
6728cb0e JA |
176 | printk(KERN_INFO " sector %llu, nr/cnr %lu/%u\n", |
177 | (unsigned long long)rq->sector, | |
178 | rq->nr_sectors, | |
179 | rq->current_nr_sectors); | |
180 | printk(KERN_INFO " bio %p, biotail %p, buffer %p, data %p, len %u\n", | |
181 | rq->bio, rq->biotail, | |
182 | rq->buffer, rq->data, | |
183 | rq->data_len); | |
1da177e4 | 184 | |
4aff5e23 | 185 | if (blk_pc_request(rq)) { |
6728cb0e | 186 | printk(KERN_INFO " cdb: "); |
d34c87e4 | 187 | for (bit = 0; bit < BLK_MAX_CDB; bit++) |
1da177e4 LT |
188 | printk("%02x ", rq->cmd[bit]); |
189 | printk("\n"); | |
190 | } | |
191 | } | |
1da177e4 LT |
192 | EXPORT_SYMBOL(blk_dump_rq_flags); |
193 | ||
1da177e4 LT |
194 | /* |
195 | * "plug" the device if there are no outstanding requests: this will | |
196 | * force the transfer to start only after we have put all the requests | |
197 | * on the list. | |
198 | * | |
199 | * This is called with interrupts off and no requests on the queue and | |
200 | * with the queue lock held. | |
201 | */ | |
165125e1 | 202 | void blk_plug_device(struct request_queue *q) |
1da177e4 LT |
203 | { |
204 | WARN_ON(!irqs_disabled()); | |
205 | ||
206 | /* | |
207 | * don't plug a stopped queue, it must be paired with blk_start_queue() | |
208 | * which will restart the queueing | |
209 | */ | |
7daac490 | 210 | if (blk_queue_stopped(q)) |
1da177e4 LT |
211 | return; |
212 | ||
e48ec690 | 213 | if (!queue_flag_test_and_set(QUEUE_FLAG_PLUGGED, q)) { |
1da177e4 | 214 | mod_timer(&q->unplug_timer, jiffies + q->unplug_delay); |
2056a782 JA |
215 | blk_add_trace_generic(q, NULL, 0, BLK_TA_PLUG); |
216 | } | |
1da177e4 | 217 | } |
1da177e4 LT |
218 | EXPORT_SYMBOL(blk_plug_device); |
219 | ||
6c5e0c4d JA |
220 | /** |
221 | * blk_plug_device_unlocked - plug a device without queue lock held | |
222 | * @q: The &struct request_queue to plug | |
223 | * | |
224 | * Description: | |
225 | * Like @blk_plug_device(), but grabs the queue lock and disables | |
226 | * interrupts. | |
227 | **/ | |
228 | void blk_plug_device_unlocked(struct request_queue *q) | |
229 | { | |
230 | unsigned long flags; | |
231 | ||
232 | spin_lock_irqsave(q->queue_lock, flags); | |
233 | blk_plug_device(q); | |
234 | spin_unlock_irqrestore(q->queue_lock, flags); | |
235 | } | |
236 | EXPORT_SYMBOL(blk_plug_device_unlocked); | |
237 | ||
1da177e4 LT |
238 | /* |
239 | * remove the queue from the plugged list, if present. called with | |
240 | * queue lock held and interrupts disabled. | |
241 | */ | |
165125e1 | 242 | int blk_remove_plug(struct request_queue *q) |
1da177e4 LT |
243 | { |
244 | WARN_ON(!irqs_disabled()); | |
245 | ||
e48ec690 | 246 | if (!queue_flag_test_and_clear(QUEUE_FLAG_PLUGGED, q)) |
1da177e4 LT |
247 | return 0; |
248 | ||
249 | del_timer(&q->unplug_timer); | |
250 | return 1; | |
251 | } | |
1da177e4 LT |
252 | EXPORT_SYMBOL(blk_remove_plug); |
253 | ||
254 | /* | |
255 | * remove the plug and let it rip.. | |
256 | */ | |
165125e1 | 257 | void __generic_unplug_device(struct request_queue *q) |
1da177e4 | 258 | { |
7daac490 | 259 | if (unlikely(blk_queue_stopped(q))) |
1da177e4 LT |
260 | return; |
261 | ||
262 | if (!blk_remove_plug(q)) | |
263 | return; | |
264 | ||
22e2c507 | 265 | q->request_fn(q); |
1da177e4 LT |
266 | } |
267 | EXPORT_SYMBOL(__generic_unplug_device); | |
268 | ||
269 | /** | |
270 | * generic_unplug_device - fire a request queue | |
165125e1 | 271 | * @q: The &struct request_queue in question |
1da177e4 LT |
272 | * |
273 | * Description: | |
274 | * Linux uses plugging to build bigger requests queues before letting | |
275 | * the device have at them. If a queue is plugged, the I/O scheduler | |
276 | * is still adding and merging requests on the queue. Once the queue | |
277 | * gets unplugged, the request_fn defined for the queue is invoked and | |
278 | * transfers started. | |
279 | **/ | |
165125e1 | 280 | void generic_unplug_device(struct request_queue *q) |
1da177e4 | 281 | { |
dbaf2c00 JA |
282 | if (blk_queue_plugged(q)) { |
283 | spin_lock_irq(q->queue_lock); | |
284 | __generic_unplug_device(q); | |
285 | spin_unlock_irq(q->queue_lock); | |
286 | } | |
1da177e4 LT |
287 | } |
288 | EXPORT_SYMBOL(generic_unplug_device); | |
289 | ||
290 | static void blk_backing_dev_unplug(struct backing_dev_info *bdi, | |
291 | struct page *page) | |
292 | { | |
165125e1 | 293 | struct request_queue *q = bdi->unplug_io_data; |
1da177e4 | 294 | |
2ad8b1ef | 295 | blk_unplug(q); |
1da177e4 LT |
296 | } |
297 | ||
86db1e29 | 298 | void blk_unplug_work(struct work_struct *work) |
1da177e4 | 299 | { |
165125e1 JA |
300 | struct request_queue *q = |
301 | container_of(work, struct request_queue, unplug_work); | |
1da177e4 | 302 | |
2056a782 JA |
303 | blk_add_trace_pdu_int(q, BLK_TA_UNPLUG_IO, NULL, |
304 | q->rq.count[READ] + q->rq.count[WRITE]); | |
305 | ||
1da177e4 LT |
306 | q->unplug_fn(q); |
307 | } | |
308 | ||
86db1e29 | 309 | void blk_unplug_timeout(unsigned long data) |
1da177e4 | 310 | { |
165125e1 | 311 | struct request_queue *q = (struct request_queue *)data; |
1da177e4 | 312 | |
2056a782 JA |
313 | blk_add_trace_pdu_int(q, BLK_TA_UNPLUG_TIMER, NULL, |
314 | q->rq.count[READ] + q->rq.count[WRITE]); | |
315 | ||
1da177e4 LT |
316 | kblockd_schedule_work(&q->unplug_work); |
317 | } | |
318 | ||
2ad8b1ef AB |
319 | void blk_unplug(struct request_queue *q) |
320 | { | |
321 | /* | |
322 | * devices don't necessarily have an ->unplug_fn defined | |
323 | */ | |
324 | if (q->unplug_fn) { | |
325 | blk_add_trace_pdu_int(q, BLK_TA_UNPLUG_IO, NULL, | |
326 | q->rq.count[READ] + q->rq.count[WRITE]); | |
327 | ||
328 | q->unplug_fn(q); | |
329 | } | |
330 | } | |
331 | EXPORT_SYMBOL(blk_unplug); | |
332 | ||
1da177e4 LT |
333 | /** |
334 | * blk_start_queue - restart a previously stopped queue | |
165125e1 | 335 | * @q: The &struct request_queue in question |
1da177e4 LT |
336 | * |
337 | * Description: | |
338 | * blk_start_queue() will clear the stop flag on the queue, and call | |
339 | * the request_fn for the queue if it was in a stopped state when | |
340 | * entered. Also see blk_stop_queue(). Queue lock must be held. | |
341 | **/ | |
165125e1 | 342 | void blk_start_queue(struct request_queue *q) |
1da177e4 | 343 | { |
a038e253 PBG |
344 | WARN_ON(!irqs_disabled()); |
345 | ||
75ad23bc | 346 | queue_flag_clear(QUEUE_FLAG_STOPPED, q); |
1da177e4 LT |
347 | |
348 | /* | |
349 | * one level of recursion is ok and is much faster than kicking | |
350 | * the unplug handling | |
351 | */ | |
e48ec690 | 352 | if (!queue_flag_test_and_set(QUEUE_FLAG_REENTER, q)) { |
1da177e4 | 353 | q->request_fn(q); |
75ad23bc | 354 | queue_flag_clear(QUEUE_FLAG_REENTER, q); |
1da177e4 LT |
355 | } else { |
356 | blk_plug_device(q); | |
357 | kblockd_schedule_work(&q->unplug_work); | |
358 | } | |
359 | } | |
1da177e4 LT |
360 | EXPORT_SYMBOL(blk_start_queue); |
361 | ||
362 | /** | |
363 | * blk_stop_queue - stop a queue | |
165125e1 | 364 | * @q: The &struct request_queue in question |
1da177e4 LT |
365 | * |
366 | * Description: | |
367 | * The Linux block layer assumes that a block driver will consume all | |
368 | * entries on the request queue when the request_fn strategy is called. | |
369 | * Often this will not happen, because of hardware limitations (queue | |
370 | * depth settings). If a device driver gets a 'queue full' response, | |
371 | * or if it simply chooses not to queue more I/O at one point, it can | |
372 | * call this function to prevent the request_fn from being called until | |
373 | * the driver has signalled it's ready to go again. This happens by calling | |
374 | * blk_start_queue() to restart queue operations. Queue lock must be held. | |
375 | **/ | |
165125e1 | 376 | void blk_stop_queue(struct request_queue *q) |
1da177e4 LT |
377 | { |
378 | blk_remove_plug(q); | |
75ad23bc | 379 | queue_flag_set(QUEUE_FLAG_STOPPED, q); |
1da177e4 LT |
380 | } |
381 | EXPORT_SYMBOL(blk_stop_queue); | |
382 | ||
383 | /** | |
384 | * blk_sync_queue - cancel any pending callbacks on a queue | |
385 | * @q: the queue | |
386 | * | |
387 | * Description: | |
388 | * The block layer may perform asynchronous callback activity | |
389 | * on a queue, such as calling the unplug function after a timeout. | |
390 | * A block device may call blk_sync_queue to ensure that any | |
391 | * such activity is cancelled, thus allowing it to release resources | |
59c51591 | 392 | * that the callbacks might use. The caller must already have made sure |
1da177e4 LT |
393 | * that its ->make_request_fn will not re-add plugging prior to calling |
394 | * this function. | |
395 | * | |
396 | */ | |
397 | void blk_sync_queue(struct request_queue *q) | |
398 | { | |
399 | del_timer_sync(&q->unplug_timer); | |
abbeb88d | 400 | kblockd_flush_work(&q->unplug_work); |
1da177e4 LT |
401 | } |
402 | EXPORT_SYMBOL(blk_sync_queue); | |
403 | ||
404 | /** | |
405 | * blk_run_queue - run a single device queue | |
406 | * @q: The queue to run | |
407 | */ | |
75ad23bc | 408 | void __blk_run_queue(struct request_queue *q) |
1da177e4 | 409 | { |
1da177e4 | 410 | blk_remove_plug(q); |
dac07ec1 JA |
411 | |
412 | /* | |
413 | * Only recurse once to avoid overrunning the stack, let the unplug | |
414 | * handling reinvoke the handler shortly if we already got there. | |
415 | */ | |
416 | if (!elv_queue_empty(q)) { | |
e48ec690 | 417 | if (!queue_flag_test_and_set(QUEUE_FLAG_REENTER, q)) { |
dac07ec1 | 418 | q->request_fn(q); |
75ad23bc | 419 | queue_flag_clear(QUEUE_FLAG_REENTER, q); |
dac07ec1 JA |
420 | } else { |
421 | blk_plug_device(q); | |
422 | kblockd_schedule_work(&q->unplug_work); | |
423 | } | |
424 | } | |
75ad23bc NP |
425 | } |
426 | EXPORT_SYMBOL(__blk_run_queue); | |
dac07ec1 | 427 | |
75ad23bc NP |
428 | /** |
429 | * blk_run_queue - run a single device queue | |
430 | * @q: The queue to run | |
431 | */ | |
432 | void blk_run_queue(struct request_queue *q) | |
433 | { | |
434 | unsigned long flags; | |
435 | ||
436 | spin_lock_irqsave(q->queue_lock, flags); | |
437 | __blk_run_queue(q); | |
1da177e4 LT |
438 | spin_unlock_irqrestore(q->queue_lock, flags); |
439 | } | |
440 | EXPORT_SYMBOL(blk_run_queue); | |
441 | ||
165125e1 | 442 | void blk_put_queue(struct request_queue *q) |
483f4afc AV |
443 | { |
444 | kobject_put(&q->kobj); | |
445 | } | |
483f4afc | 446 | |
6728cb0e | 447 | void blk_cleanup_queue(struct request_queue *q) |
483f4afc AV |
448 | { |
449 | mutex_lock(&q->sysfs_lock); | |
75ad23bc | 450 | queue_flag_set_unlocked(QUEUE_FLAG_DEAD, q); |
483f4afc AV |
451 | mutex_unlock(&q->sysfs_lock); |
452 | ||
453 | if (q->elevator) | |
454 | elevator_exit(q->elevator); | |
455 | ||
456 | blk_put_queue(q); | |
457 | } | |
1da177e4 LT |
458 | EXPORT_SYMBOL(blk_cleanup_queue); |
459 | ||
165125e1 | 460 | static int blk_init_free_list(struct request_queue *q) |
1da177e4 LT |
461 | { |
462 | struct request_list *rl = &q->rq; | |
463 | ||
464 | rl->count[READ] = rl->count[WRITE] = 0; | |
465 | rl->starved[READ] = rl->starved[WRITE] = 0; | |
cb98fc8b | 466 | rl->elvpriv = 0; |
1da177e4 LT |
467 | init_waitqueue_head(&rl->wait[READ]); |
468 | init_waitqueue_head(&rl->wait[WRITE]); | |
1da177e4 | 469 | |
1946089a CL |
470 | rl->rq_pool = mempool_create_node(BLKDEV_MIN_RQ, mempool_alloc_slab, |
471 | mempool_free_slab, request_cachep, q->node); | |
1da177e4 LT |
472 | |
473 | if (!rl->rq_pool) | |
474 | return -ENOMEM; | |
475 | ||
476 | return 0; | |
477 | } | |
478 | ||
165125e1 | 479 | struct request_queue *blk_alloc_queue(gfp_t gfp_mask) |
1da177e4 | 480 | { |
1946089a CL |
481 | return blk_alloc_queue_node(gfp_mask, -1); |
482 | } | |
483 | EXPORT_SYMBOL(blk_alloc_queue); | |
1da177e4 | 484 | |
165125e1 | 485 | struct request_queue *blk_alloc_queue_node(gfp_t gfp_mask, int node_id) |
1946089a | 486 | { |
165125e1 | 487 | struct request_queue *q; |
e0bf68dd | 488 | int err; |
1946089a | 489 | |
8324aa91 | 490 | q = kmem_cache_alloc_node(blk_requestq_cachep, |
94f6030c | 491 | gfp_mask | __GFP_ZERO, node_id); |
1da177e4 LT |
492 | if (!q) |
493 | return NULL; | |
494 | ||
e0bf68dd PZ |
495 | q->backing_dev_info.unplug_io_fn = blk_backing_dev_unplug; |
496 | q->backing_dev_info.unplug_io_data = q; | |
497 | err = bdi_init(&q->backing_dev_info); | |
498 | if (err) { | |
8324aa91 | 499 | kmem_cache_free(blk_requestq_cachep, q); |
e0bf68dd PZ |
500 | return NULL; |
501 | } | |
502 | ||
1da177e4 | 503 | init_timer(&q->unplug_timer); |
483f4afc | 504 | |
8324aa91 | 505 | kobject_init(&q->kobj, &blk_queue_ktype); |
1da177e4 | 506 | |
483f4afc | 507 | mutex_init(&q->sysfs_lock); |
e7e72bf6 | 508 | spin_lock_init(&q->__queue_lock); |
483f4afc | 509 | |
1da177e4 LT |
510 | return q; |
511 | } | |
1946089a | 512 | EXPORT_SYMBOL(blk_alloc_queue_node); |
1da177e4 LT |
513 | |
514 | /** | |
515 | * blk_init_queue - prepare a request queue for use with a block device | |
516 | * @rfn: The function to be called to process requests that have been | |
517 | * placed on the queue. | |
518 | * @lock: Request queue spin lock | |
519 | * | |
520 | * Description: | |
521 | * If a block device wishes to use the standard request handling procedures, | |
522 | * which sorts requests and coalesces adjacent requests, then it must | |
523 | * call blk_init_queue(). The function @rfn will be called when there | |
524 | * are requests on the queue that need to be processed. If the device | |
525 | * supports plugging, then @rfn may not be called immediately when requests | |
526 | * are available on the queue, but may be called at some time later instead. | |
527 | * Plugged queues are generally unplugged when a buffer belonging to one | |
528 | * of the requests on the queue is needed, or due to memory pressure. | |
529 | * | |
530 | * @rfn is not required, or even expected, to remove all requests off the | |
531 | * queue, but only as many as it can handle at a time. If it does leave | |
532 | * requests on the queue, it is responsible for arranging that the requests | |
533 | * get dealt with eventually. | |
534 | * | |
535 | * The queue spin lock must be held while manipulating the requests on the | |
a038e253 PBG |
536 | * request queue; this lock will be taken also from interrupt context, so irq |
537 | * disabling is needed for it. | |
1da177e4 | 538 | * |
710027a4 | 539 | * Function returns a pointer to the initialized request queue, or %NULL if |
1da177e4 LT |
540 | * it didn't succeed. |
541 | * | |
542 | * Note: | |
543 | * blk_init_queue() must be paired with a blk_cleanup_queue() call | |
544 | * when the block device is deactivated (such as at module unload). | |
545 | **/ | |
1946089a | 546 | |
165125e1 | 547 | struct request_queue *blk_init_queue(request_fn_proc *rfn, spinlock_t *lock) |
1da177e4 | 548 | { |
1946089a CL |
549 | return blk_init_queue_node(rfn, lock, -1); |
550 | } | |
551 | EXPORT_SYMBOL(blk_init_queue); | |
552 | ||
165125e1 | 553 | struct request_queue * |
1946089a CL |
554 | blk_init_queue_node(request_fn_proc *rfn, spinlock_t *lock, int node_id) |
555 | { | |
165125e1 | 556 | struct request_queue *q = blk_alloc_queue_node(GFP_KERNEL, node_id); |
1da177e4 LT |
557 | |
558 | if (!q) | |
559 | return NULL; | |
560 | ||
1946089a | 561 | q->node = node_id; |
8669aafd | 562 | if (blk_init_free_list(q)) { |
8324aa91 | 563 | kmem_cache_free(blk_requestq_cachep, q); |
8669aafd AV |
564 | return NULL; |
565 | } | |
1da177e4 | 566 | |
152587de | 567 | /* |
568 | * if caller didn't supply a lock, they get per-queue locking with | |
569 | * our embedded lock | |
570 | */ | |
e7e72bf6 | 571 | if (!lock) |
152587de | 572 | lock = &q->__queue_lock; |
152587de | 573 | |
1da177e4 | 574 | q->request_fn = rfn; |
1da177e4 LT |
575 | q->prep_rq_fn = NULL; |
576 | q->unplug_fn = generic_unplug_device; | |
577 | q->queue_flags = (1 << QUEUE_FLAG_CLUSTER); | |
578 | q->queue_lock = lock; | |
579 | ||
580 | blk_queue_segment_boundary(q, 0xffffffff); | |
581 | ||
582 | blk_queue_make_request(q, __make_request); | |
583 | blk_queue_max_segment_size(q, MAX_SEGMENT_SIZE); | |
584 | ||
585 | blk_queue_max_hw_segments(q, MAX_HW_SEGMENTS); | |
586 | blk_queue_max_phys_segments(q, MAX_PHYS_SEGMENTS); | |
587 | ||
44ec9542 AS |
588 | q->sg_reserved_size = INT_MAX; |
589 | ||
abf54393 FT |
590 | blk_set_cmd_filter_defaults(&q->cmd_filter); |
591 | ||
1da177e4 LT |
592 | /* |
593 | * all done | |
594 | */ | |
595 | if (!elevator_init(q, NULL)) { | |
596 | blk_queue_congestion_threshold(q); | |
597 | return q; | |
598 | } | |
599 | ||
8669aafd | 600 | blk_put_queue(q); |
1da177e4 LT |
601 | return NULL; |
602 | } | |
1946089a | 603 | EXPORT_SYMBOL(blk_init_queue_node); |
1da177e4 | 604 | |
165125e1 | 605 | int blk_get_queue(struct request_queue *q) |
1da177e4 | 606 | { |
fde6ad22 | 607 | if (likely(!test_bit(QUEUE_FLAG_DEAD, &q->queue_flags))) { |
483f4afc | 608 | kobject_get(&q->kobj); |
1da177e4 LT |
609 | return 0; |
610 | } | |
611 | ||
612 | return 1; | |
613 | } | |
1da177e4 | 614 | |
165125e1 | 615 | static inline void blk_free_request(struct request_queue *q, struct request *rq) |
1da177e4 | 616 | { |
4aff5e23 | 617 | if (rq->cmd_flags & REQ_ELVPRIV) |
cb98fc8b | 618 | elv_put_request(q, rq); |
1da177e4 LT |
619 | mempool_free(rq, q->rq.rq_pool); |
620 | } | |
621 | ||
1ea25ecb | 622 | static struct request * |
165125e1 | 623 | blk_alloc_request(struct request_queue *q, int rw, int priv, gfp_t gfp_mask) |
1da177e4 LT |
624 | { |
625 | struct request *rq = mempool_alloc(q->rq.rq_pool, gfp_mask); | |
626 | ||
627 | if (!rq) | |
628 | return NULL; | |
629 | ||
2a4aa30c | 630 | blk_rq_init(q, rq); |
1afb20f3 | 631 | |
49171e5c | 632 | rq->cmd_flags = rw | REQ_ALLOCED; |
1da177e4 | 633 | |
cb98fc8b | 634 | if (priv) { |
cb78b285 | 635 | if (unlikely(elv_set_request(q, rq, gfp_mask))) { |
cb98fc8b TH |
636 | mempool_free(rq, q->rq.rq_pool); |
637 | return NULL; | |
638 | } | |
4aff5e23 | 639 | rq->cmd_flags |= REQ_ELVPRIV; |
cb98fc8b | 640 | } |
1da177e4 | 641 | |
cb98fc8b | 642 | return rq; |
1da177e4 LT |
643 | } |
644 | ||
645 | /* | |
646 | * ioc_batching returns true if the ioc is a valid batching request and | |
647 | * should be given priority access to a request. | |
648 | */ | |
165125e1 | 649 | static inline int ioc_batching(struct request_queue *q, struct io_context *ioc) |
1da177e4 LT |
650 | { |
651 | if (!ioc) | |
652 | return 0; | |
653 | ||
654 | /* | |
655 | * Make sure the process is able to allocate at least 1 request | |
656 | * even if the batch times out, otherwise we could theoretically | |
657 | * lose wakeups. | |
658 | */ | |
659 | return ioc->nr_batch_requests == q->nr_batching || | |
660 | (ioc->nr_batch_requests > 0 | |
661 | && time_before(jiffies, ioc->last_waited + BLK_BATCH_TIME)); | |
662 | } | |
663 | ||
664 | /* | |
665 | * ioc_set_batching sets ioc to be a new "batcher" if it is not one. This | |
666 | * will cause the process to be a "batcher" on all queues in the system. This | |
667 | * is the behaviour we want though - once it gets a wakeup it should be given | |
668 | * a nice run. | |
669 | */ | |
165125e1 | 670 | static void ioc_set_batching(struct request_queue *q, struct io_context *ioc) |
1da177e4 LT |
671 | { |
672 | if (!ioc || ioc_batching(q, ioc)) | |
673 | return; | |
674 | ||
675 | ioc->nr_batch_requests = q->nr_batching; | |
676 | ioc->last_waited = jiffies; | |
677 | } | |
678 | ||
165125e1 | 679 | static void __freed_request(struct request_queue *q, int rw) |
1da177e4 LT |
680 | { |
681 | struct request_list *rl = &q->rq; | |
682 | ||
683 | if (rl->count[rw] < queue_congestion_off_threshold(q)) | |
79e2de4b | 684 | blk_clear_queue_congested(q, rw); |
1da177e4 LT |
685 | |
686 | if (rl->count[rw] + 1 <= q->nr_requests) { | |
1da177e4 LT |
687 | if (waitqueue_active(&rl->wait[rw])) |
688 | wake_up(&rl->wait[rw]); | |
689 | ||
690 | blk_clear_queue_full(q, rw); | |
691 | } | |
692 | } | |
693 | ||
694 | /* | |
695 | * A request has just been released. Account for it, update the full and | |
696 | * congestion status, wake up any waiters. Called under q->queue_lock. | |
697 | */ | |
165125e1 | 698 | static void freed_request(struct request_queue *q, int rw, int priv) |
1da177e4 LT |
699 | { |
700 | struct request_list *rl = &q->rq; | |
701 | ||
702 | rl->count[rw]--; | |
cb98fc8b TH |
703 | if (priv) |
704 | rl->elvpriv--; | |
1da177e4 LT |
705 | |
706 | __freed_request(q, rw); | |
707 | ||
708 | if (unlikely(rl->starved[rw ^ 1])) | |
709 | __freed_request(q, rw ^ 1); | |
1da177e4 LT |
710 | } |
711 | ||
712 | #define blkdev_free_rq(list) list_entry((list)->next, struct request, queuelist) | |
713 | /* | |
d6344532 NP |
714 | * Get a free request, queue_lock must be held. |
715 | * Returns NULL on failure, with queue_lock held. | |
716 | * Returns !NULL on success, with queue_lock *not held*. | |
1da177e4 | 717 | */ |
165125e1 | 718 | static struct request *get_request(struct request_queue *q, int rw_flags, |
7749a8d4 | 719 | struct bio *bio, gfp_t gfp_mask) |
1da177e4 LT |
720 | { |
721 | struct request *rq = NULL; | |
722 | struct request_list *rl = &q->rq; | |
88ee5ef1 | 723 | struct io_context *ioc = NULL; |
7749a8d4 | 724 | const int rw = rw_flags & 0x01; |
88ee5ef1 JA |
725 | int may_queue, priv; |
726 | ||
7749a8d4 | 727 | may_queue = elv_may_queue(q, rw_flags); |
88ee5ef1 JA |
728 | if (may_queue == ELV_MQUEUE_NO) |
729 | goto rq_starved; | |
730 | ||
731 | if (rl->count[rw]+1 >= queue_congestion_on_threshold(q)) { | |
732 | if (rl->count[rw]+1 >= q->nr_requests) { | |
b5deef90 | 733 | ioc = current_io_context(GFP_ATOMIC, q->node); |
88ee5ef1 JA |
734 | /* |
735 | * The queue will fill after this allocation, so set | |
736 | * it as full, and mark this process as "batching". | |
737 | * This process will be allowed to complete a batch of | |
738 | * requests, others will be blocked. | |
739 | */ | |
740 | if (!blk_queue_full(q, rw)) { | |
741 | ioc_set_batching(q, ioc); | |
742 | blk_set_queue_full(q, rw); | |
743 | } else { | |
744 | if (may_queue != ELV_MQUEUE_MUST | |
745 | && !ioc_batching(q, ioc)) { | |
746 | /* | |
747 | * The queue is full and the allocating | |
748 | * process is not a "batcher", and not | |
749 | * exempted by the IO scheduler | |
750 | */ | |
751 | goto out; | |
752 | } | |
753 | } | |
1da177e4 | 754 | } |
79e2de4b | 755 | blk_set_queue_congested(q, rw); |
1da177e4 LT |
756 | } |
757 | ||
082cf69e JA |
758 | /* |
759 | * Only allow batching queuers to allocate up to 50% over the defined | |
760 | * limit of requests, otherwise we could have thousands of requests | |
761 | * allocated with any setting of ->nr_requests | |
762 | */ | |
fd782a4a | 763 | if (rl->count[rw] >= (3 * q->nr_requests / 2)) |
082cf69e | 764 | goto out; |
fd782a4a | 765 | |
1da177e4 LT |
766 | rl->count[rw]++; |
767 | rl->starved[rw] = 0; | |
cb98fc8b | 768 | |
64521d1a | 769 | priv = !test_bit(QUEUE_FLAG_ELVSWITCH, &q->queue_flags); |
cb98fc8b TH |
770 | if (priv) |
771 | rl->elvpriv++; | |
772 | ||
1da177e4 LT |
773 | spin_unlock_irq(q->queue_lock); |
774 | ||
7749a8d4 | 775 | rq = blk_alloc_request(q, rw_flags, priv, gfp_mask); |
88ee5ef1 | 776 | if (unlikely(!rq)) { |
1da177e4 LT |
777 | /* |
778 | * Allocation failed presumably due to memory. Undo anything | |
779 | * we might have messed up. | |
780 | * | |
781 | * Allocating task should really be put onto the front of the | |
782 | * wait queue, but this is pretty rare. | |
783 | */ | |
784 | spin_lock_irq(q->queue_lock); | |
cb98fc8b | 785 | freed_request(q, rw, priv); |
1da177e4 LT |
786 | |
787 | /* | |
788 | * in the very unlikely event that allocation failed and no | |
789 | * requests for this direction was pending, mark us starved | |
790 | * so that freeing of a request in the other direction will | |
791 | * notice us. another possible fix would be to split the | |
792 | * rq mempool into READ and WRITE | |
793 | */ | |
794 | rq_starved: | |
795 | if (unlikely(rl->count[rw] == 0)) | |
796 | rl->starved[rw] = 1; | |
797 | ||
1da177e4 LT |
798 | goto out; |
799 | } | |
800 | ||
88ee5ef1 JA |
801 | /* |
802 | * ioc may be NULL here, and ioc_batching will be false. That's | |
803 | * OK, if the queue is under the request limit then requests need | |
804 | * not count toward the nr_batch_requests limit. There will always | |
805 | * be some limit enforced by BLK_BATCH_TIME. | |
806 | */ | |
1da177e4 LT |
807 | if (ioc_batching(q, ioc)) |
808 | ioc->nr_batch_requests--; | |
6728cb0e | 809 | |
2056a782 | 810 | blk_add_trace_generic(q, bio, rw, BLK_TA_GETRQ); |
1da177e4 | 811 | out: |
1da177e4 LT |
812 | return rq; |
813 | } | |
814 | ||
815 | /* | |
816 | * No available requests for this queue, unplug the device and wait for some | |
817 | * requests to become available. | |
d6344532 NP |
818 | * |
819 | * Called with q->queue_lock held, and returns with it unlocked. | |
1da177e4 | 820 | */ |
165125e1 | 821 | static struct request *get_request_wait(struct request_queue *q, int rw_flags, |
22e2c507 | 822 | struct bio *bio) |
1da177e4 | 823 | { |
7749a8d4 | 824 | const int rw = rw_flags & 0x01; |
1da177e4 LT |
825 | struct request *rq; |
826 | ||
7749a8d4 | 827 | rq = get_request(q, rw_flags, bio, GFP_NOIO); |
450991bc NP |
828 | while (!rq) { |
829 | DEFINE_WAIT(wait); | |
05caf8db | 830 | struct io_context *ioc; |
1da177e4 LT |
831 | struct request_list *rl = &q->rq; |
832 | ||
833 | prepare_to_wait_exclusive(&rl->wait[rw], &wait, | |
834 | TASK_UNINTERRUPTIBLE); | |
835 | ||
05caf8db | 836 | blk_add_trace_generic(q, bio, rw, BLK_TA_SLEEPRQ); |
1da177e4 | 837 | |
05caf8db ZY |
838 | __generic_unplug_device(q); |
839 | spin_unlock_irq(q->queue_lock); | |
840 | io_schedule(); | |
1da177e4 | 841 | |
05caf8db ZY |
842 | /* |
843 | * After sleeping, we become a "batching" process and | |
844 | * will be able to allocate at least one request, and | |
845 | * up to a big batch of them for a small period time. | |
846 | * See ioc_batching, ioc_set_batching | |
847 | */ | |
848 | ioc = current_io_context(GFP_NOIO, q->node); | |
849 | ioc_set_batching(q, ioc); | |
d6344532 | 850 | |
05caf8db | 851 | spin_lock_irq(q->queue_lock); |
1da177e4 | 852 | finish_wait(&rl->wait[rw], &wait); |
05caf8db ZY |
853 | |
854 | rq = get_request(q, rw_flags, bio, GFP_NOIO); | |
855 | }; | |
1da177e4 LT |
856 | |
857 | return rq; | |
858 | } | |
859 | ||
165125e1 | 860 | struct request *blk_get_request(struct request_queue *q, int rw, gfp_t gfp_mask) |
1da177e4 LT |
861 | { |
862 | struct request *rq; | |
863 | ||
864 | BUG_ON(rw != READ && rw != WRITE); | |
865 | ||
d6344532 NP |
866 | spin_lock_irq(q->queue_lock); |
867 | if (gfp_mask & __GFP_WAIT) { | |
22e2c507 | 868 | rq = get_request_wait(q, rw, NULL); |
d6344532 | 869 | } else { |
22e2c507 | 870 | rq = get_request(q, rw, NULL, gfp_mask); |
d6344532 NP |
871 | if (!rq) |
872 | spin_unlock_irq(q->queue_lock); | |
873 | } | |
874 | /* q->queue_lock is unlocked at this point */ | |
1da177e4 LT |
875 | |
876 | return rq; | |
877 | } | |
1da177e4 LT |
878 | EXPORT_SYMBOL(blk_get_request); |
879 | ||
dc72ef4a JA |
880 | /** |
881 | * blk_start_queueing - initiate dispatch of requests to device | |
882 | * @q: request queue to kick into gear | |
883 | * | |
884 | * This is basically a helper to remove the need to know whether a queue | |
885 | * is plugged or not if someone just wants to initiate dispatch of requests | |
886 | * for this queue. | |
887 | * | |
888 | * The queue lock must be held with interrupts disabled. | |
889 | */ | |
165125e1 | 890 | void blk_start_queueing(struct request_queue *q) |
dc72ef4a JA |
891 | { |
892 | if (!blk_queue_plugged(q)) | |
893 | q->request_fn(q); | |
894 | else | |
895 | __generic_unplug_device(q); | |
896 | } | |
897 | EXPORT_SYMBOL(blk_start_queueing); | |
898 | ||
1da177e4 LT |
899 | /** |
900 | * blk_requeue_request - put a request back on queue | |
901 | * @q: request queue where request should be inserted | |
902 | * @rq: request to be inserted | |
903 | * | |
904 | * Description: | |
905 | * Drivers often keep queueing requests until the hardware cannot accept | |
906 | * more, when that condition happens we need to put the request back | |
907 | * on the queue. Must be called with queue lock held. | |
908 | */ | |
165125e1 | 909 | void blk_requeue_request(struct request_queue *q, struct request *rq) |
1da177e4 | 910 | { |
2056a782 JA |
911 | blk_add_trace_rq(q, rq, BLK_TA_REQUEUE); |
912 | ||
1da177e4 LT |
913 | if (blk_rq_tagged(rq)) |
914 | blk_queue_end_tag(q, rq); | |
915 | ||
916 | elv_requeue_request(q, rq); | |
917 | } | |
1da177e4 LT |
918 | EXPORT_SYMBOL(blk_requeue_request); |
919 | ||
920 | /** | |
710027a4 | 921 | * blk_insert_request - insert a special request into a request queue |
1da177e4 LT |
922 | * @q: request queue where request should be inserted |
923 | * @rq: request to be inserted | |
924 | * @at_head: insert request at head or tail of queue | |
925 | * @data: private data | |
1da177e4 LT |
926 | * |
927 | * Description: | |
928 | * Many block devices need to execute commands asynchronously, so they don't | |
929 | * block the whole kernel from preemption during request execution. This is | |
930 | * accomplished normally by inserting aritficial requests tagged as | |
710027a4 RD |
931 | * REQ_TYPE_SPECIAL in to the corresponding request queue, and letting them |
932 | * be scheduled for actual execution by the request queue. | |
1da177e4 LT |
933 | * |
934 | * We have the option of inserting the head or the tail of the queue. | |
935 | * Typically we use the tail for new ioctls and so forth. We use the head | |
936 | * of the queue for things like a QUEUE_FULL message from a device, or a | |
937 | * host that is unable to accept a particular command. | |
938 | */ | |
165125e1 | 939 | void blk_insert_request(struct request_queue *q, struct request *rq, |
867d1191 | 940 | int at_head, void *data) |
1da177e4 | 941 | { |
867d1191 | 942 | int where = at_head ? ELEVATOR_INSERT_FRONT : ELEVATOR_INSERT_BACK; |
1da177e4 LT |
943 | unsigned long flags; |
944 | ||
945 | /* | |
946 | * tell I/O scheduler that this isn't a regular read/write (ie it | |
947 | * must not attempt merges on this) and that it acts as a soft | |
948 | * barrier | |
949 | */ | |
4aff5e23 JA |
950 | rq->cmd_type = REQ_TYPE_SPECIAL; |
951 | rq->cmd_flags |= REQ_SOFTBARRIER; | |
1da177e4 LT |
952 | |
953 | rq->special = data; | |
954 | ||
955 | spin_lock_irqsave(q->queue_lock, flags); | |
956 | ||
957 | /* | |
958 | * If command is tagged, release the tag | |
959 | */ | |
867d1191 TH |
960 | if (blk_rq_tagged(rq)) |
961 | blk_queue_end_tag(q, rq); | |
1da177e4 | 962 | |
b238b3d4 | 963 | drive_stat_acct(rq, 1); |
867d1191 | 964 | __elv_add_request(q, rq, where, 0); |
dc72ef4a | 965 | blk_start_queueing(q); |
1da177e4 LT |
966 | spin_unlock_irqrestore(q->queue_lock, flags); |
967 | } | |
1da177e4 LT |
968 | EXPORT_SYMBOL(blk_insert_request); |
969 | ||
1da177e4 LT |
970 | /* |
971 | * add-request adds a request to the linked list. | |
972 | * queue lock is held and interrupts disabled, as we muck with the | |
973 | * request queue list. | |
974 | */ | |
6728cb0e | 975 | static inline void add_request(struct request_queue *q, struct request *req) |
1da177e4 | 976 | { |
b238b3d4 | 977 | drive_stat_acct(req, 1); |
1da177e4 | 978 | |
1da177e4 LT |
979 | /* |
980 | * elevator indicated where it wants this request to be | |
981 | * inserted at elevator_merge time | |
982 | */ | |
983 | __elv_add_request(q, req, ELEVATOR_INSERT_SORT, 0); | |
984 | } | |
6728cb0e | 985 | |
1da177e4 LT |
986 | /* |
987 | * disk_round_stats() - Round off the performance stats on a struct | |
988 | * disk_stats. | |
989 | * | |
990 | * The average IO queue length and utilisation statistics are maintained | |
991 | * by observing the current state of the queue length and the amount of | |
992 | * time it has been in this state for. | |
993 | * | |
994 | * Normally, that accounting is done on IO completion, but that can result | |
995 | * in more than a second's worth of IO being accounted for within any one | |
996 | * second, leading to >100% utilisation. To deal with that, we call this | |
997 | * function to do a round-off before returning the results when reading | |
998 | * /proc/diskstats. This accounts immediately for all queue usage up to | |
999 | * the current jiffies and restarts the counters again. | |
1000 | */ | |
c9959059 | 1001 | void disk_round_stats(int cpu, struct gendisk *disk) |
1da177e4 LT |
1002 | { |
1003 | unsigned long now = jiffies; | |
1004 | ||
b2982649 CK |
1005 | if (now == disk->stamp) |
1006 | return; | |
1da177e4 | 1007 | |
20e5c81f | 1008 | if (disk->in_flight) { |
c9959059 TH |
1009 | disk_stat_add(cpu, disk, time_in_queue, |
1010 | disk->in_flight * (now - disk->stamp)); | |
1011 | disk_stat_add(cpu, disk, io_ticks, (now - disk->stamp)); | |
20e5c81f | 1012 | } |
1da177e4 | 1013 | disk->stamp = now; |
1da177e4 | 1014 | } |
3eaf840e JNN |
1015 | EXPORT_SYMBOL_GPL(disk_round_stats); |
1016 | ||
c9959059 | 1017 | void part_round_stats(int cpu, struct hd_struct *part) |
6f2576af JM |
1018 | { |
1019 | unsigned long now = jiffies; | |
1020 | ||
1021 | if (now == part->stamp) | |
1022 | return; | |
1023 | ||
1024 | if (part->in_flight) { | |
c9959059 TH |
1025 | part_stat_add(cpu, part, time_in_queue, |
1026 | part->in_flight * (now - part->stamp)); | |
1027 | part_stat_add(cpu, part, io_ticks, (now - part->stamp)); | |
6f2576af JM |
1028 | } |
1029 | part->stamp = now; | |
1030 | } | |
1031 | ||
1da177e4 LT |
1032 | /* |
1033 | * queue lock must be held | |
1034 | */ | |
165125e1 | 1035 | void __blk_put_request(struct request_queue *q, struct request *req) |
1da177e4 | 1036 | { |
1da177e4 LT |
1037 | if (unlikely(!q)) |
1038 | return; | |
1039 | if (unlikely(--req->ref_count)) | |
1040 | return; | |
1041 | ||
8922e16c TH |
1042 | elv_completed_request(q, req); |
1043 | ||
1da177e4 LT |
1044 | /* |
1045 | * Request may not have originated from ll_rw_blk. if not, | |
1046 | * it didn't come out of our reserved rq pools | |
1047 | */ | |
49171e5c | 1048 | if (req->cmd_flags & REQ_ALLOCED) { |
1da177e4 | 1049 | int rw = rq_data_dir(req); |
4aff5e23 | 1050 | int priv = req->cmd_flags & REQ_ELVPRIV; |
1da177e4 | 1051 | |
1da177e4 | 1052 | BUG_ON(!list_empty(&req->queuelist)); |
9817064b | 1053 | BUG_ON(!hlist_unhashed(&req->hash)); |
1da177e4 LT |
1054 | |
1055 | blk_free_request(q, req); | |
cb98fc8b | 1056 | freed_request(q, rw, priv); |
1da177e4 LT |
1057 | } |
1058 | } | |
6e39b69e MC |
1059 | EXPORT_SYMBOL_GPL(__blk_put_request); |
1060 | ||
1da177e4 LT |
1061 | void blk_put_request(struct request *req) |
1062 | { | |
8922e16c | 1063 | unsigned long flags; |
165125e1 | 1064 | struct request_queue *q = req->q; |
8922e16c | 1065 | |
52a93ba8 FT |
1066 | spin_lock_irqsave(q->queue_lock, flags); |
1067 | __blk_put_request(q, req); | |
1068 | spin_unlock_irqrestore(q->queue_lock, flags); | |
1da177e4 | 1069 | } |
1da177e4 LT |
1070 | EXPORT_SYMBOL(blk_put_request); |
1071 | ||
86db1e29 | 1072 | void init_request_from_bio(struct request *req, struct bio *bio) |
52d9e675 | 1073 | { |
4aff5e23 | 1074 | req->cmd_type = REQ_TYPE_FS; |
52d9e675 TH |
1075 | |
1076 | /* | |
1077 | * inherit FAILFAST from bio (for read-ahead, and explicit FAILFAST) | |
1078 | */ | |
1079 | if (bio_rw_ahead(bio) || bio_failfast(bio)) | |
4aff5e23 | 1080 | req->cmd_flags |= REQ_FAILFAST; |
52d9e675 TH |
1081 | |
1082 | /* | |
1083 | * REQ_BARRIER implies no merging, but lets make it explicit | |
1084 | */ | |
fb2dce86 | 1085 | if (unlikely(bio_discard(bio))) { |
e17fc0a1 DW |
1086 | req->cmd_flags |= REQ_DISCARD; |
1087 | if (bio_barrier(bio)) | |
1088 | req->cmd_flags |= REQ_SOFTBARRIER; | |
fb2dce86 | 1089 | req->q->prepare_discard_fn(req->q, req); |
e17fc0a1 DW |
1090 | } else if (unlikely(bio_barrier(bio))) |
1091 | req->cmd_flags |= (REQ_HARDBARRIER | REQ_NOMERGE); | |
52d9e675 | 1092 | |
b31dc66a | 1093 | if (bio_sync(bio)) |
4aff5e23 | 1094 | req->cmd_flags |= REQ_RW_SYNC; |
5404bc7a JA |
1095 | if (bio_rw_meta(bio)) |
1096 | req->cmd_flags |= REQ_RW_META; | |
b31dc66a | 1097 | |
52d9e675 TH |
1098 | req->errors = 0; |
1099 | req->hard_sector = req->sector = bio->bi_sector; | |
52d9e675 | 1100 | req->ioprio = bio_prio(bio); |
52d9e675 | 1101 | req->start_time = jiffies; |
bc1c56fd | 1102 | blk_rq_bio_prep(req->q, req, bio); |
52d9e675 TH |
1103 | } |
1104 | ||
165125e1 | 1105 | static int __make_request(struct request_queue *q, struct bio *bio) |
1da177e4 | 1106 | { |
450991bc | 1107 | struct request *req; |
fb2dce86 | 1108 | int el_ret, nr_sectors, barrier, discard, err; |
51da90fc JA |
1109 | const unsigned short prio = bio_prio(bio); |
1110 | const int sync = bio_sync(bio); | |
7749a8d4 | 1111 | int rw_flags; |
1da177e4 | 1112 | |
1da177e4 | 1113 | nr_sectors = bio_sectors(bio); |
1da177e4 LT |
1114 | |
1115 | /* | |
1116 | * low level driver can indicate that it wants pages above a | |
1117 | * certain limit bounced to low memory (ie for highmem, or even | |
1118 | * ISA dma in theory) | |
1119 | */ | |
1120 | blk_queue_bounce(q, &bio); | |
1121 | ||
1da177e4 | 1122 | barrier = bio_barrier(bio); |
e17fc0a1 DW |
1123 | if (unlikely(barrier) && bio_has_data(bio) && |
1124 | (q->next_ordered == QUEUE_ORDERED_NONE)) { | |
1da177e4 LT |
1125 | err = -EOPNOTSUPP; |
1126 | goto end_io; | |
1127 | } | |
1128 | ||
fb2dce86 DW |
1129 | discard = bio_discard(bio); |
1130 | if (unlikely(discard) && !q->prepare_discard_fn) { | |
1131 | err = -EOPNOTSUPP; | |
1132 | goto end_io; | |
1133 | } | |
1134 | ||
1da177e4 LT |
1135 | spin_lock_irq(q->queue_lock); |
1136 | ||
450991bc | 1137 | if (unlikely(barrier) || elv_queue_empty(q)) |
1da177e4 LT |
1138 | goto get_rq; |
1139 | ||
1140 | el_ret = elv_merge(q, &req, bio); | |
1141 | switch (el_ret) { | |
6728cb0e JA |
1142 | case ELEVATOR_BACK_MERGE: |
1143 | BUG_ON(!rq_mergeable(req)); | |
1da177e4 | 1144 | |
6728cb0e JA |
1145 | if (!ll_back_merge_fn(q, req, bio)) |
1146 | break; | |
1da177e4 | 1147 | |
6728cb0e | 1148 | blk_add_trace_bio(q, bio, BLK_TA_BACKMERGE); |
2056a782 | 1149 | |
6728cb0e JA |
1150 | req->biotail->bi_next = bio; |
1151 | req->biotail = bio; | |
1152 | req->nr_sectors = req->hard_nr_sectors += nr_sectors; | |
1153 | req->ioprio = ioprio_best(req->ioprio, prio); | |
1154 | drive_stat_acct(req, 0); | |
1155 | if (!attempt_back_merge(q, req)) | |
1156 | elv_merged_request(q, req, el_ret); | |
1157 | goto out; | |
1da177e4 | 1158 | |
6728cb0e JA |
1159 | case ELEVATOR_FRONT_MERGE: |
1160 | BUG_ON(!rq_mergeable(req)); | |
1da177e4 | 1161 | |
6728cb0e JA |
1162 | if (!ll_front_merge_fn(q, req, bio)) |
1163 | break; | |
1da177e4 | 1164 | |
6728cb0e | 1165 | blk_add_trace_bio(q, bio, BLK_TA_FRONTMERGE); |
2056a782 | 1166 | |
6728cb0e JA |
1167 | bio->bi_next = req->bio; |
1168 | req->bio = bio; | |
1da177e4 | 1169 | |
6728cb0e JA |
1170 | /* |
1171 | * may not be valid. if the low level driver said | |
1172 | * it didn't need a bounce buffer then it better | |
1173 | * not touch req->buffer either... | |
1174 | */ | |
1175 | req->buffer = bio_data(bio); | |
1176 | req->current_nr_sectors = bio_cur_sectors(bio); | |
1177 | req->hard_cur_sectors = req->current_nr_sectors; | |
1178 | req->sector = req->hard_sector = bio->bi_sector; | |
1179 | req->nr_sectors = req->hard_nr_sectors += nr_sectors; | |
1180 | req->ioprio = ioprio_best(req->ioprio, prio); | |
1181 | drive_stat_acct(req, 0); | |
1182 | if (!attempt_front_merge(q, req)) | |
1183 | elv_merged_request(q, req, el_ret); | |
1184 | goto out; | |
1185 | ||
1186 | /* ELV_NO_MERGE: elevator says don't/can't merge. */ | |
1187 | default: | |
1188 | ; | |
1da177e4 LT |
1189 | } |
1190 | ||
450991bc | 1191 | get_rq: |
7749a8d4 JA |
1192 | /* |
1193 | * This sync check and mask will be re-done in init_request_from_bio(), | |
1194 | * but we need to set it earlier to expose the sync flag to the | |
1195 | * rq allocator and io schedulers. | |
1196 | */ | |
1197 | rw_flags = bio_data_dir(bio); | |
1198 | if (sync) | |
1199 | rw_flags |= REQ_RW_SYNC; | |
1200 | ||
1da177e4 | 1201 | /* |
450991bc | 1202 | * Grab a free request. This is might sleep but can not fail. |
d6344532 | 1203 | * Returns with the queue unlocked. |
450991bc | 1204 | */ |
7749a8d4 | 1205 | req = get_request_wait(q, rw_flags, bio); |
d6344532 | 1206 | |
450991bc NP |
1207 | /* |
1208 | * After dropping the lock and possibly sleeping here, our request | |
1209 | * may now be mergeable after it had proven unmergeable (above). | |
1210 | * We don't worry about that case for efficiency. It won't happen | |
1211 | * often, and the elevators are able to handle it. | |
1da177e4 | 1212 | */ |
52d9e675 | 1213 | init_request_from_bio(req, bio); |
1da177e4 | 1214 | |
450991bc NP |
1215 | spin_lock_irq(q->queue_lock); |
1216 | if (elv_queue_empty(q)) | |
1217 | blk_plug_device(q); | |
1da177e4 LT |
1218 | add_request(q, req); |
1219 | out: | |
4a534f93 | 1220 | if (sync) |
1da177e4 LT |
1221 | __generic_unplug_device(q); |
1222 | ||
1223 | spin_unlock_irq(q->queue_lock); | |
1224 | return 0; | |
1225 | ||
1226 | end_io: | |
6712ecf8 | 1227 | bio_endio(bio, err); |
1da177e4 LT |
1228 | return 0; |
1229 | } | |
1230 | ||
1231 | /* | |
1232 | * If bio->bi_dev is a partition, remap the location | |
1233 | */ | |
1234 | static inline void blk_partition_remap(struct bio *bio) | |
1235 | { | |
1236 | struct block_device *bdev = bio->bi_bdev; | |
1237 | ||
bf2de6f5 | 1238 | if (bio_sectors(bio) && bdev != bdev->bd_contains) { |
1da177e4 LT |
1239 | struct hd_struct *p = bdev->bd_part; |
1240 | ||
1da177e4 LT |
1241 | bio->bi_sector += p->start_sect; |
1242 | bio->bi_bdev = bdev->bd_contains; | |
c7149d6b AB |
1243 | |
1244 | blk_add_trace_remap(bdev_get_queue(bio->bi_bdev), bio, | |
1245 | bdev->bd_dev, bio->bi_sector, | |
1246 | bio->bi_sector - p->start_sect); | |
1da177e4 LT |
1247 | } |
1248 | } | |
1249 | ||
1da177e4 LT |
1250 | static void handle_bad_sector(struct bio *bio) |
1251 | { | |
1252 | char b[BDEVNAME_SIZE]; | |
1253 | ||
1254 | printk(KERN_INFO "attempt to access beyond end of device\n"); | |
1255 | printk(KERN_INFO "%s: rw=%ld, want=%Lu, limit=%Lu\n", | |
1256 | bdevname(bio->bi_bdev, b), | |
1257 | bio->bi_rw, | |
1258 | (unsigned long long)bio->bi_sector + bio_sectors(bio), | |
1259 | (long long)(bio->bi_bdev->bd_inode->i_size >> 9)); | |
1260 | ||
1261 | set_bit(BIO_EOF, &bio->bi_flags); | |
1262 | } | |
1263 | ||
c17bb495 AM |
1264 | #ifdef CONFIG_FAIL_MAKE_REQUEST |
1265 | ||
1266 | static DECLARE_FAULT_ATTR(fail_make_request); | |
1267 | ||
1268 | static int __init setup_fail_make_request(char *str) | |
1269 | { | |
1270 | return setup_fault_attr(&fail_make_request, str); | |
1271 | } | |
1272 | __setup("fail_make_request=", setup_fail_make_request); | |
1273 | ||
1274 | static int should_fail_request(struct bio *bio) | |
1275 | { | |
1276 | if ((bio->bi_bdev->bd_disk->flags & GENHD_FL_FAIL) || | |
1277 | (bio->bi_bdev->bd_part && bio->bi_bdev->bd_part->make_it_fail)) | |
1278 | return should_fail(&fail_make_request, bio->bi_size); | |
1279 | ||
1280 | return 0; | |
1281 | } | |
1282 | ||
1283 | static int __init fail_make_request_debugfs(void) | |
1284 | { | |
1285 | return init_fault_attr_dentries(&fail_make_request, | |
1286 | "fail_make_request"); | |
1287 | } | |
1288 | ||
1289 | late_initcall(fail_make_request_debugfs); | |
1290 | ||
1291 | #else /* CONFIG_FAIL_MAKE_REQUEST */ | |
1292 | ||
1293 | static inline int should_fail_request(struct bio *bio) | |
1294 | { | |
1295 | return 0; | |
1296 | } | |
1297 | ||
1298 | #endif /* CONFIG_FAIL_MAKE_REQUEST */ | |
1299 | ||
c07e2b41 JA |
1300 | /* |
1301 | * Check whether this bio extends beyond the end of the device. | |
1302 | */ | |
1303 | static inline int bio_check_eod(struct bio *bio, unsigned int nr_sectors) | |
1304 | { | |
1305 | sector_t maxsector; | |
1306 | ||
1307 | if (!nr_sectors) | |
1308 | return 0; | |
1309 | ||
1310 | /* Test device or partition size, when known. */ | |
1311 | maxsector = bio->bi_bdev->bd_inode->i_size >> 9; | |
1312 | if (maxsector) { | |
1313 | sector_t sector = bio->bi_sector; | |
1314 | ||
1315 | if (maxsector < nr_sectors || maxsector - nr_sectors < sector) { | |
1316 | /* | |
1317 | * This may well happen - the kernel calls bread() | |
1318 | * without checking the size of the device, e.g., when | |
1319 | * mounting a device. | |
1320 | */ | |
1321 | handle_bad_sector(bio); | |
1322 | return 1; | |
1323 | } | |
1324 | } | |
1325 | ||
1326 | return 0; | |
1327 | } | |
1328 | ||
1da177e4 | 1329 | /** |
710027a4 | 1330 | * generic_make_request - hand a buffer to its device driver for I/O |
1da177e4 LT |
1331 | * @bio: The bio describing the location in memory and on the device. |
1332 | * | |
1333 | * generic_make_request() is used to make I/O requests of block | |
1334 | * devices. It is passed a &struct bio, which describes the I/O that needs | |
1335 | * to be done. | |
1336 | * | |
1337 | * generic_make_request() does not return any status. The | |
1338 | * success/failure status of the request, along with notification of | |
1339 | * completion, is delivered asynchronously through the bio->bi_end_io | |
1340 | * function described (one day) else where. | |
1341 | * | |
1342 | * The caller of generic_make_request must make sure that bi_io_vec | |
1343 | * are set to describe the memory buffer, and that bi_dev and bi_sector are | |
1344 | * set to describe the device address, and the | |
1345 | * bi_end_io and optionally bi_private are set to describe how | |
1346 | * completion notification should be signaled. | |
1347 | * | |
1348 | * generic_make_request and the drivers it calls may use bi_next if this | |
1349 | * bio happens to be merged with someone else, and may change bi_dev and | |
1350 | * bi_sector for remaps as it sees fit. So the values of these fields | |
1351 | * should NOT be depended on after the call to generic_make_request. | |
1352 | */ | |
d89d8796 | 1353 | static inline void __generic_make_request(struct bio *bio) |
1da177e4 | 1354 | { |
165125e1 | 1355 | struct request_queue *q; |
5ddfe969 | 1356 | sector_t old_sector; |
1da177e4 | 1357 | int ret, nr_sectors = bio_sectors(bio); |
2056a782 | 1358 | dev_t old_dev; |
51fd77bd | 1359 | int err = -EIO; |
1da177e4 LT |
1360 | |
1361 | might_sleep(); | |
1da177e4 | 1362 | |
c07e2b41 JA |
1363 | if (bio_check_eod(bio, nr_sectors)) |
1364 | goto end_io; | |
1da177e4 LT |
1365 | |
1366 | /* | |
1367 | * Resolve the mapping until finished. (drivers are | |
1368 | * still free to implement/resolve their own stacking | |
1369 | * by explicitly returning 0) | |
1370 | * | |
1371 | * NOTE: we don't repeat the blk_size check for each new device. | |
1372 | * Stacking drivers are expected to know what they are doing. | |
1373 | */ | |
5ddfe969 | 1374 | old_sector = -1; |
2056a782 | 1375 | old_dev = 0; |
1da177e4 LT |
1376 | do { |
1377 | char b[BDEVNAME_SIZE]; | |
1378 | ||
1379 | q = bdev_get_queue(bio->bi_bdev); | |
1380 | if (!q) { | |
1381 | printk(KERN_ERR | |
1382 | "generic_make_request: Trying to access " | |
1383 | "nonexistent block-device %s (%Lu)\n", | |
1384 | bdevname(bio->bi_bdev, b), | |
1385 | (long long) bio->bi_sector); | |
1386 | end_io: | |
51fd77bd | 1387 | bio_endio(bio, err); |
1da177e4 LT |
1388 | break; |
1389 | } | |
1390 | ||
4fa253f3 | 1391 | if (unlikely(nr_sectors > q->max_hw_sectors)) { |
6728cb0e | 1392 | printk(KERN_ERR "bio too big device %s (%u > %u)\n", |
1da177e4 LT |
1393 | bdevname(bio->bi_bdev, b), |
1394 | bio_sectors(bio), | |
1395 | q->max_hw_sectors); | |
1396 | goto end_io; | |
1397 | } | |
1398 | ||
fde6ad22 | 1399 | if (unlikely(test_bit(QUEUE_FLAG_DEAD, &q->queue_flags))) |
1da177e4 LT |
1400 | goto end_io; |
1401 | ||
c17bb495 AM |
1402 | if (should_fail_request(bio)) |
1403 | goto end_io; | |
1404 | ||
1da177e4 LT |
1405 | /* |
1406 | * If this device has partitions, remap block n | |
1407 | * of partition p to block n+start(p) of the disk. | |
1408 | */ | |
1409 | blk_partition_remap(bio); | |
1410 | ||
7ba1ba12 MP |
1411 | if (bio_integrity_enabled(bio) && bio_integrity_prep(bio)) |
1412 | goto end_io; | |
1413 | ||
5ddfe969 | 1414 | if (old_sector != -1) |
4fa253f3 | 1415 | blk_add_trace_remap(q, bio, old_dev, bio->bi_sector, |
5ddfe969 | 1416 | old_sector); |
2056a782 JA |
1417 | |
1418 | blk_add_trace_bio(q, bio, BLK_TA_QUEUE); | |
1419 | ||
5ddfe969 | 1420 | old_sector = bio->bi_sector; |
2056a782 JA |
1421 | old_dev = bio->bi_bdev->bd_dev; |
1422 | ||
c07e2b41 JA |
1423 | if (bio_check_eod(bio, nr_sectors)) |
1424 | goto end_io; | |
fb2dce86 DW |
1425 | if ((bio_empty_barrier(bio) && !q->prepare_flush_fn) || |
1426 | (bio_discard(bio) && !q->prepare_discard_fn)) { | |
51fd77bd JA |
1427 | err = -EOPNOTSUPP; |
1428 | goto end_io; | |
1429 | } | |
5ddfe969 | 1430 | |
1da177e4 LT |
1431 | ret = q->make_request_fn(q, bio); |
1432 | } while (ret); | |
1433 | } | |
1434 | ||
d89d8796 NB |
1435 | /* |
1436 | * We only want one ->make_request_fn to be active at a time, | |
1437 | * else stack usage with stacked devices could be a problem. | |
1438 | * So use current->bio_{list,tail} to keep a list of requests | |
1439 | * submited by a make_request_fn function. | |
1440 | * current->bio_tail is also used as a flag to say if | |
1441 | * generic_make_request is currently active in this task or not. | |
1442 | * If it is NULL, then no make_request is active. If it is non-NULL, | |
1443 | * then a make_request is active, and new requests should be added | |
1444 | * at the tail | |
1445 | */ | |
1446 | void generic_make_request(struct bio *bio) | |
1447 | { | |
1448 | if (current->bio_tail) { | |
1449 | /* make_request is active */ | |
1450 | *(current->bio_tail) = bio; | |
1451 | bio->bi_next = NULL; | |
1452 | current->bio_tail = &bio->bi_next; | |
1453 | return; | |
1454 | } | |
1455 | /* following loop may be a bit non-obvious, and so deserves some | |
1456 | * explanation. | |
1457 | * Before entering the loop, bio->bi_next is NULL (as all callers | |
1458 | * ensure that) so we have a list with a single bio. | |
1459 | * We pretend that we have just taken it off a longer list, so | |
1460 | * we assign bio_list to the next (which is NULL) and bio_tail | |
1461 | * to &bio_list, thus initialising the bio_list of new bios to be | |
1462 | * added. __generic_make_request may indeed add some more bios | |
1463 | * through a recursive call to generic_make_request. If it | |
1464 | * did, we find a non-NULL value in bio_list and re-enter the loop | |
1465 | * from the top. In this case we really did just take the bio | |
1466 | * of the top of the list (no pretending) and so fixup bio_list and | |
1467 | * bio_tail or bi_next, and call into __generic_make_request again. | |
1468 | * | |
1469 | * The loop was structured like this to make only one call to | |
1470 | * __generic_make_request (which is important as it is large and | |
1471 | * inlined) and to keep the structure simple. | |
1472 | */ | |
1473 | BUG_ON(bio->bi_next); | |
1474 | do { | |
1475 | current->bio_list = bio->bi_next; | |
1476 | if (bio->bi_next == NULL) | |
1477 | current->bio_tail = ¤t->bio_list; | |
1478 | else | |
1479 | bio->bi_next = NULL; | |
1480 | __generic_make_request(bio); | |
1481 | bio = current->bio_list; | |
1482 | } while (bio); | |
1483 | current->bio_tail = NULL; /* deactivate */ | |
1484 | } | |
1da177e4 LT |
1485 | EXPORT_SYMBOL(generic_make_request); |
1486 | ||
1487 | /** | |
710027a4 | 1488 | * submit_bio - submit a bio to the block device layer for I/O |
1da177e4 LT |
1489 | * @rw: whether to %READ or %WRITE, or maybe to %READA (read ahead) |
1490 | * @bio: The &struct bio which describes the I/O | |
1491 | * | |
1492 | * submit_bio() is very similar in purpose to generic_make_request(), and | |
1493 | * uses that function to do most of the work. Both are fairly rough | |
710027a4 | 1494 | * interfaces; @bio must be presetup and ready for I/O. |
1da177e4 LT |
1495 | * |
1496 | */ | |
1497 | void submit_bio(int rw, struct bio *bio) | |
1498 | { | |
1499 | int count = bio_sectors(bio); | |
1500 | ||
22e2c507 | 1501 | bio->bi_rw |= rw; |
1da177e4 | 1502 | |
bf2de6f5 JA |
1503 | /* |
1504 | * If it's a regular read/write or a barrier with data attached, | |
1505 | * go through the normal accounting stuff before submission. | |
1506 | */ | |
a9c701e5 | 1507 | if (bio_has_data(bio)) { |
bf2de6f5 JA |
1508 | if (rw & WRITE) { |
1509 | count_vm_events(PGPGOUT, count); | |
1510 | } else { | |
1511 | task_io_account_read(bio->bi_size); | |
1512 | count_vm_events(PGPGIN, count); | |
1513 | } | |
1514 | ||
1515 | if (unlikely(block_dump)) { | |
1516 | char b[BDEVNAME_SIZE]; | |
1517 | printk(KERN_DEBUG "%s(%d): %s block %Lu on %s\n", | |
ba25f9dc | 1518 | current->comm, task_pid_nr(current), |
bf2de6f5 JA |
1519 | (rw & WRITE) ? "WRITE" : "READ", |
1520 | (unsigned long long)bio->bi_sector, | |
6728cb0e | 1521 | bdevname(bio->bi_bdev, b)); |
bf2de6f5 | 1522 | } |
1da177e4 LT |
1523 | } |
1524 | ||
1525 | generic_make_request(bio); | |
1526 | } | |
1da177e4 LT |
1527 | EXPORT_SYMBOL(submit_bio); |
1528 | ||
3bcddeac KU |
1529 | /** |
1530 | * __end_that_request_first - end I/O on a request | |
1531 | * @req: the request being processed | |
710027a4 | 1532 | * @error: %0 for success, < %0 for error |
3bcddeac KU |
1533 | * @nr_bytes: number of bytes to complete |
1534 | * | |
1535 | * Description: | |
1536 | * Ends I/O on a number of bytes attached to @req, and sets it up | |
1537 | * for the next range of segments (if any) in the cluster. | |
1538 | * | |
1539 | * Return: | |
710027a4 RD |
1540 | * %0 - we are done with this request, call end_that_request_last() |
1541 | * %1 - still buffers pending for this request | |
3bcddeac | 1542 | **/ |
5450d3e1 | 1543 | static int __end_that_request_first(struct request *req, int error, |
1da177e4 LT |
1544 | int nr_bytes) |
1545 | { | |
5450d3e1 | 1546 | int total_bytes, bio_nbytes, next_idx = 0; |
1da177e4 LT |
1547 | struct bio *bio; |
1548 | ||
2056a782 JA |
1549 | blk_add_trace_rq(req->q, req, BLK_TA_COMPLETE); |
1550 | ||
1da177e4 | 1551 | /* |
710027a4 | 1552 | * for a REQ_TYPE_BLOCK_PC request, we want to carry any eventual |
1da177e4 LT |
1553 | * sense key with us all the way through |
1554 | */ | |
1555 | if (!blk_pc_request(req)) | |
1556 | req->errors = 0; | |
1557 | ||
6728cb0e JA |
1558 | if (error && (blk_fs_request(req) && !(req->cmd_flags & REQ_QUIET))) { |
1559 | printk(KERN_ERR "end_request: I/O error, dev %s, sector %llu\n", | |
1da177e4 LT |
1560 | req->rq_disk ? req->rq_disk->disk_name : "?", |
1561 | (unsigned long long)req->sector); | |
1562 | } | |
1563 | ||
d72d904a | 1564 | if (blk_fs_request(req) && req->rq_disk) { |
a362357b | 1565 | const int rw = rq_data_dir(req); |
e71bf0d0 | 1566 | struct hd_struct *part; |
c9959059 | 1567 | int cpu; |
a362357b | 1568 | |
c9959059 | 1569 | cpu = disk_stat_lock(); |
e71bf0d0 | 1570 | part = disk_map_sector_rcu(req->rq_disk, req->sector); |
c9959059 TH |
1571 | all_stat_add(cpu, req->rq_disk, part, sectors[rw], |
1572 | nr_bytes >> 9, req->sector); | |
1573 | disk_stat_unlock(); | |
d72d904a JA |
1574 | } |
1575 | ||
1da177e4 LT |
1576 | total_bytes = bio_nbytes = 0; |
1577 | while ((bio = req->bio) != NULL) { | |
1578 | int nbytes; | |
1579 | ||
bf2de6f5 JA |
1580 | /* |
1581 | * For an empty barrier request, the low level driver must | |
1582 | * store a potential error location in ->sector. We pass | |
1583 | * that back up in ->bi_sector. | |
1584 | */ | |
1585 | if (blk_empty_barrier(req)) | |
1586 | bio->bi_sector = req->sector; | |
1587 | ||
1da177e4 LT |
1588 | if (nr_bytes >= bio->bi_size) { |
1589 | req->bio = bio->bi_next; | |
1590 | nbytes = bio->bi_size; | |
5bb23a68 | 1591 | req_bio_endio(req, bio, nbytes, error); |
1da177e4 LT |
1592 | next_idx = 0; |
1593 | bio_nbytes = 0; | |
1594 | } else { | |
1595 | int idx = bio->bi_idx + next_idx; | |
1596 | ||
1597 | if (unlikely(bio->bi_idx >= bio->bi_vcnt)) { | |
1598 | blk_dump_rq_flags(req, "__end_that"); | |
6728cb0e | 1599 | printk(KERN_ERR "%s: bio idx %d >= vcnt %d\n", |
24c03d47 | 1600 | __func__, bio->bi_idx, bio->bi_vcnt); |
1da177e4 LT |
1601 | break; |
1602 | } | |
1603 | ||
1604 | nbytes = bio_iovec_idx(bio, idx)->bv_len; | |
1605 | BIO_BUG_ON(nbytes > bio->bi_size); | |
1606 | ||
1607 | /* | |
1608 | * not a complete bvec done | |
1609 | */ | |
1610 | if (unlikely(nbytes > nr_bytes)) { | |
1611 | bio_nbytes += nr_bytes; | |
1612 | total_bytes += nr_bytes; | |
1613 | break; | |
1614 | } | |
1615 | ||
1616 | /* | |
1617 | * advance to the next vector | |
1618 | */ | |
1619 | next_idx++; | |
1620 | bio_nbytes += nbytes; | |
1621 | } | |
1622 | ||
1623 | total_bytes += nbytes; | |
1624 | nr_bytes -= nbytes; | |
1625 | ||
6728cb0e JA |
1626 | bio = req->bio; |
1627 | if (bio) { | |
1da177e4 LT |
1628 | /* |
1629 | * end more in this run, or just return 'not-done' | |
1630 | */ | |
1631 | if (unlikely(nr_bytes <= 0)) | |
1632 | break; | |
1633 | } | |
1634 | } | |
1635 | ||
1636 | /* | |
1637 | * completely done | |
1638 | */ | |
1639 | if (!req->bio) | |
1640 | return 0; | |
1641 | ||
1642 | /* | |
1643 | * if the request wasn't completed, update state | |
1644 | */ | |
1645 | if (bio_nbytes) { | |
5bb23a68 | 1646 | req_bio_endio(req, bio, bio_nbytes, error); |
1da177e4 LT |
1647 | bio->bi_idx += next_idx; |
1648 | bio_iovec(bio)->bv_offset += nr_bytes; | |
1649 | bio_iovec(bio)->bv_len -= nr_bytes; | |
1650 | } | |
1651 | ||
1652 | blk_recalc_rq_sectors(req, total_bytes >> 9); | |
1653 | blk_recalc_rq_segments(req); | |
1654 | return 1; | |
1655 | } | |
1656 | ||
ff856bad JA |
1657 | /* |
1658 | * splice the completion data to a local structure and hand off to | |
1659 | * process_completion_queue() to complete the requests | |
1660 | */ | |
1661 | static void blk_done_softirq(struct softirq_action *h) | |
1662 | { | |
626ab0e6 | 1663 | struct list_head *cpu_list, local_list; |
ff856bad JA |
1664 | |
1665 | local_irq_disable(); | |
1666 | cpu_list = &__get_cpu_var(blk_cpu_done); | |
626ab0e6 | 1667 | list_replace_init(cpu_list, &local_list); |
ff856bad JA |
1668 | local_irq_enable(); |
1669 | ||
1670 | while (!list_empty(&local_list)) { | |
6728cb0e | 1671 | struct request *rq; |
ff856bad | 1672 | |
6728cb0e | 1673 | rq = list_entry(local_list.next, struct request, donelist); |
ff856bad JA |
1674 | list_del_init(&rq->donelist); |
1675 | rq->q->softirq_done_fn(rq); | |
1676 | } | |
1677 | } | |
1678 | ||
6728cb0e JA |
1679 | static int __cpuinit blk_cpu_notify(struct notifier_block *self, |
1680 | unsigned long action, void *hcpu) | |
ff856bad JA |
1681 | { |
1682 | /* | |
1683 | * If a CPU goes away, splice its entries to the current CPU | |
1684 | * and trigger a run of the softirq | |
1685 | */ | |
8bb78442 | 1686 | if (action == CPU_DEAD || action == CPU_DEAD_FROZEN) { |
ff856bad JA |
1687 | int cpu = (unsigned long) hcpu; |
1688 | ||
1689 | local_irq_disable(); | |
1690 | list_splice_init(&per_cpu(blk_cpu_done, cpu), | |
1691 | &__get_cpu_var(blk_cpu_done)); | |
1692 | raise_softirq_irqoff(BLOCK_SOFTIRQ); | |
1693 | local_irq_enable(); | |
1694 | } | |
1695 | ||
1696 | return NOTIFY_OK; | |
1697 | } | |
1698 | ||
1699 | ||
db47d475 | 1700 | static struct notifier_block blk_cpu_notifier __cpuinitdata = { |
ff856bad JA |
1701 | .notifier_call = blk_cpu_notify, |
1702 | }; | |
1703 | ||
ff856bad JA |
1704 | /** |
1705 | * blk_complete_request - end I/O on a request | |
1706 | * @req: the request being processed | |
1707 | * | |
1708 | * Description: | |
1709 | * Ends all I/O on a request. It does not handle partial completions, | |
d6e05edc | 1710 | * unless the driver actually implements this in its completion callback |
4fa253f3 | 1711 | * through requeueing. The actual completion happens out-of-order, |
ff856bad JA |
1712 | * through a softirq handler. The user must have registered a completion |
1713 | * callback through blk_queue_softirq_done(). | |
1714 | **/ | |
1715 | ||
1716 | void blk_complete_request(struct request *req) | |
1717 | { | |
1718 | struct list_head *cpu_list; | |
1719 | unsigned long flags; | |
1720 | ||
1721 | BUG_ON(!req->q->softirq_done_fn); | |
6728cb0e | 1722 | |
ff856bad JA |
1723 | local_irq_save(flags); |
1724 | ||
1725 | cpu_list = &__get_cpu_var(blk_cpu_done); | |
1726 | list_add_tail(&req->donelist, cpu_list); | |
1727 | raise_softirq_irqoff(BLOCK_SOFTIRQ); | |
1728 | ||
1729 | local_irq_restore(flags); | |
1730 | } | |
ff856bad | 1731 | EXPORT_SYMBOL(blk_complete_request); |
6728cb0e | 1732 | |
1da177e4 LT |
1733 | /* |
1734 | * queue lock must be held | |
1735 | */ | |
5450d3e1 | 1736 | static void end_that_request_last(struct request *req, int error) |
1da177e4 LT |
1737 | { |
1738 | struct gendisk *disk = req->rq_disk; | |
8ffdc655 | 1739 | |
b8286239 KU |
1740 | if (blk_rq_tagged(req)) |
1741 | blk_queue_end_tag(req->q, req); | |
1742 | ||
1743 | if (blk_queued_rq(req)) | |
1744 | blkdev_dequeue_request(req); | |
1da177e4 LT |
1745 | |
1746 | if (unlikely(laptop_mode) && blk_fs_request(req)) | |
1747 | laptop_io_completion(); | |
1748 | ||
fd0ff8aa JA |
1749 | /* |
1750 | * Account IO completion. bar_rq isn't accounted as a normal | |
1751 | * IO on queueing nor completion. Accounting the containing | |
1752 | * request is enough. | |
1753 | */ | |
1754 | if (disk && blk_fs_request(req) && req != &req->q->bar_rq) { | |
1da177e4 | 1755 | unsigned long duration = jiffies - req->start_time; |
a362357b | 1756 | const int rw = rq_data_dir(req); |
e71bf0d0 | 1757 | struct hd_struct *part; |
c9959059 | 1758 | int cpu; |
e71bf0d0 | 1759 | |
c9959059 | 1760 | cpu = disk_stat_lock(); |
e71bf0d0 | 1761 | part = disk_map_sector_rcu(disk, req->sector); |
a362357b | 1762 | |
c9959059 TH |
1763 | all_stat_inc(cpu, disk, part, ios[rw], req->sector); |
1764 | all_stat_add(cpu, disk, part, ticks[rw], duration, req->sector); | |
1765 | disk_round_stats(cpu, disk); | |
1da177e4 | 1766 | disk->in_flight--; |
6f2576af | 1767 | if (part) { |
c9959059 | 1768 | part_round_stats(cpu, part); |
6f2576af JM |
1769 | part->in_flight--; |
1770 | } | |
e71bf0d0 | 1771 | |
c9959059 | 1772 | disk_stat_unlock(); |
1da177e4 | 1773 | } |
b8286239 | 1774 | |
1da177e4 | 1775 | if (req->end_io) |
8ffdc655 | 1776 | req->end_io(req, error); |
b8286239 KU |
1777 | else { |
1778 | if (blk_bidi_rq(req)) | |
1779 | __blk_put_request(req->next_rq->q, req->next_rq); | |
1780 | ||
1da177e4 | 1781 | __blk_put_request(req->q, req); |
b8286239 | 1782 | } |
1da177e4 LT |
1783 | } |
1784 | ||
a0cd1285 | 1785 | static inline void __end_request(struct request *rq, int uptodate, |
9e6e39f2 | 1786 | unsigned int nr_bytes) |
1da177e4 | 1787 | { |
9e6e39f2 KU |
1788 | int error = 0; |
1789 | ||
1790 | if (uptodate <= 0) | |
1791 | error = uptodate ? uptodate : -EIO; | |
1792 | ||
1793 | __blk_end_request(rq, error, nr_bytes); | |
1da177e4 LT |
1794 | } |
1795 | ||
3b11313a KU |
1796 | /** |
1797 | * blk_rq_bytes - Returns bytes left to complete in the entire request | |
5d87a052 | 1798 | * @rq: the request being processed |
3b11313a KU |
1799 | **/ |
1800 | unsigned int blk_rq_bytes(struct request *rq) | |
a0cd1285 JA |
1801 | { |
1802 | if (blk_fs_request(rq)) | |
1803 | return rq->hard_nr_sectors << 9; | |
1804 | ||
1805 | return rq->data_len; | |
1806 | } | |
3b11313a KU |
1807 | EXPORT_SYMBOL_GPL(blk_rq_bytes); |
1808 | ||
1809 | /** | |
1810 | * blk_rq_cur_bytes - Returns bytes left to complete in the current segment | |
5d87a052 | 1811 | * @rq: the request being processed |
3b11313a KU |
1812 | **/ |
1813 | unsigned int blk_rq_cur_bytes(struct request *rq) | |
1814 | { | |
1815 | if (blk_fs_request(rq)) | |
1816 | return rq->current_nr_sectors << 9; | |
1817 | ||
1818 | if (rq->bio) | |
1819 | return rq->bio->bi_size; | |
1820 | ||
1821 | return rq->data_len; | |
1822 | } | |
1823 | EXPORT_SYMBOL_GPL(blk_rq_cur_bytes); | |
a0cd1285 JA |
1824 | |
1825 | /** | |
1826 | * end_queued_request - end all I/O on a queued request | |
1827 | * @rq: the request being processed | |
710027a4 | 1828 | * @uptodate: error value or %0/%1 uptodate flag |
a0cd1285 JA |
1829 | * |
1830 | * Description: | |
1831 | * Ends all I/O on a request, and removes it from the block layer queues. | |
710027a4 | 1832 | * Not suitable for normal I/O completion, unless the driver still has |
a0cd1285 JA |
1833 | * the request attached to the block layer. |
1834 | * | |
1835 | **/ | |
1836 | void end_queued_request(struct request *rq, int uptodate) | |
1837 | { | |
9e6e39f2 | 1838 | __end_request(rq, uptodate, blk_rq_bytes(rq)); |
a0cd1285 JA |
1839 | } |
1840 | EXPORT_SYMBOL(end_queued_request); | |
1841 | ||
1842 | /** | |
1843 | * end_dequeued_request - end all I/O on a dequeued request | |
1844 | * @rq: the request being processed | |
710027a4 | 1845 | * @uptodate: error value or %0/%1 uptodate flag |
a0cd1285 JA |
1846 | * |
1847 | * Description: | |
1848 | * Ends all I/O on a request. The request must already have been | |
1849 | * dequeued using blkdev_dequeue_request(), as is normally the case | |
1850 | * for most drivers. | |
1851 | * | |
1852 | **/ | |
1853 | void end_dequeued_request(struct request *rq, int uptodate) | |
1854 | { | |
9e6e39f2 | 1855 | __end_request(rq, uptodate, blk_rq_bytes(rq)); |
a0cd1285 JA |
1856 | } |
1857 | EXPORT_SYMBOL(end_dequeued_request); | |
1858 | ||
1859 | ||
1860 | /** | |
1861 | * end_request - end I/O on the current segment of the request | |
8f731f7d | 1862 | * @req: the request being processed |
710027a4 | 1863 | * @uptodate: error value or %0/%1 uptodate flag |
a0cd1285 JA |
1864 | * |
1865 | * Description: | |
1866 | * Ends I/O on the current segment of a request. If that is the only | |
1867 | * remaining segment, the request is also completed and freed. | |
1868 | * | |
710027a4 RD |
1869 | * This is a remnant of how older block drivers handled I/O completions. |
1870 | * Modern drivers typically end I/O on the full request in one go, unless | |
a0cd1285 JA |
1871 | * they have a residual value to account for. For that case this function |
1872 | * isn't really useful, unless the residual just happens to be the | |
1873 | * full current segment. In other words, don't use this function in new | |
1874 | * code. Either use end_request_completely(), or the | |
1875 | * end_that_request_chunk() (along with end_that_request_last()) for | |
1876 | * partial completions. | |
1877 | * | |
1878 | **/ | |
1879 | void end_request(struct request *req, int uptodate) | |
1880 | { | |
9e6e39f2 | 1881 | __end_request(req, uptodate, req->hard_cur_sectors << 9); |
a0cd1285 | 1882 | } |
1da177e4 LT |
1883 | EXPORT_SYMBOL(end_request); |
1884 | ||
336cdb40 | 1885 | /** |
e19a3ab0 KU |
1886 | * blk_end_io - Generic end_io function to complete a request. |
1887 | * @rq: the request being processed | |
710027a4 | 1888 | * @error: %0 for success, < %0 for error |
e3a04fe3 KU |
1889 | * @nr_bytes: number of bytes to complete @rq |
1890 | * @bidi_bytes: number of bytes to complete @rq->next_rq | |
e19a3ab0 KU |
1891 | * @drv_callback: function called between completion of bios in the request |
1892 | * and completion of the request. | |
710027a4 | 1893 | * If the callback returns non %0, this helper returns without |
e19a3ab0 | 1894 | * completion of the request. |
336cdb40 KU |
1895 | * |
1896 | * Description: | |
e3a04fe3 | 1897 | * Ends I/O on a number of bytes attached to @rq and @rq->next_rq. |
336cdb40 KU |
1898 | * If @rq has leftover, sets it up for the next range of segments. |
1899 | * | |
1900 | * Return: | |
710027a4 RD |
1901 | * %0 - we are done with this request |
1902 | * %1 - this request is not freed yet, it still has pending buffers. | |
336cdb40 | 1903 | **/ |
22b13210 JA |
1904 | static int blk_end_io(struct request *rq, int error, unsigned int nr_bytes, |
1905 | unsigned int bidi_bytes, | |
1906 | int (drv_callback)(struct request *)) | |
336cdb40 KU |
1907 | { |
1908 | struct request_queue *q = rq->q; | |
1909 | unsigned long flags = 0UL; | |
336cdb40 | 1910 | |
fb2dce86 | 1911 | if (bio_has_data(rq->bio) || blk_discard_rq(rq)) { |
5450d3e1 | 1912 | if (__end_that_request_first(rq, error, nr_bytes)) |
336cdb40 | 1913 | return 1; |
e3a04fe3 KU |
1914 | |
1915 | /* Bidi request must be completed as a whole */ | |
1916 | if (blk_bidi_rq(rq) && | |
5450d3e1 | 1917 | __end_that_request_first(rq->next_rq, error, bidi_bytes)) |
e3a04fe3 | 1918 | return 1; |
336cdb40 KU |
1919 | } |
1920 | ||
e19a3ab0 KU |
1921 | /* Special feature for tricky drivers */ |
1922 | if (drv_callback && drv_callback(rq)) | |
1923 | return 1; | |
1924 | ||
336cdb40 KU |
1925 | add_disk_randomness(rq->rq_disk); |
1926 | ||
1927 | spin_lock_irqsave(q->queue_lock, flags); | |
b8286239 | 1928 | end_that_request_last(rq, error); |
336cdb40 KU |
1929 | spin_unlock_irqrestore(q->queue_lock, flags); |
1930 | ||
1931 | return 0; | |
1932 | } | |
e19a3ab0 KU |
1933 | |
1934 | /** | |
1935 | * blk_end_request - Helper function for drivers to complete the request. | |
1936 | * @rq: the request being processed | |
710027a4 | 1937 | * @error: %0 for success, < %0 for error |
e19a3ab0 KU |
1938 | * @nr_bytes: number of bytes to complete |
1939 | * | |
1940 | * Description: | |
1941 | * Ends I/O on a number of bytes attached to @rq. | |
1942 | * If @rq has leftover, sets it up for the next range of segments. | |
1943 | * | |
1944 | * Return: | |
710027a4 RD |
1945 | * %0 - we are done with this request |
1946 | * %1 - still buffers pending for this request | |
e19a3ab0 | 1947 | **/ |
22b13210 | 1948 | int blk_end_request(struct request *rq, int error, unsigned int nr_bytes) |
e19a3ab0 | 1949 | { |
e3a04fe3 | 1950 | return blk_end_io(rq, error, nr_bytes, 0, NULL); |
e19a3ab0 | 1951 | } |
336cdb40 KU |
1952 | EXPORT_SYMBOL_GPL(blk_end_request); |
1953 | ||
1954 | /** | |
1955 | * __blk_end_request - Helper function for drivers to complete the request. | |
1956 | * @rq: the request being processed | |
710027a4 | 1957 | * @error: %0 for success, < %0 for error |
336cdb40 KU |
1958 | * @nr_bytes: number of bytes to complete |
1959 | * | |
1960 | * Description: | |
1961 | * Must be called with queue lock held unlike blk_end_request(). | |
1962 | * | |
1963 | * Return: | |
710027a4 RD |
1964 | * %0 - we are done with this request |
1965 | * %1 - still buffers pending for this request | |
336cdb40 | 1966 | **/ |
22b13210 | 1967 | int __blk_end_request(struct request *rq, int error, unsigned int nr_bytes) |
336cdb40 | 1968 | { |
fb2dce86 | 1969 | if ((bio_has_data(rq->bio) || blk_discard_rq(rq)) && |
051cc395 JA |
1970 | __end_that_request_first(rq, error, nr_bytes)) |
1971 | return 1; | |
336cdb40 KU |
1972 | |
1973 | add_disk_randomness(rq->rq_disk); | |
1974 | ||
b8286239 | 1975 | end_that_request_last(rq, error); |
336cdb40 KU |
1976 | |
1977 | return 0; | |
1978 | } | |
1979 | EXPORT_SYMBOL_GPL(__blk_end_request); | |
1980 | ||
e3a04fe3 KU |
1981 | /** |
1982 | * blk_end_bidi_request - Helper function for drivers to complete bidi request. | |
1983 | * @rq: the bidi request being processed | |
710027a4 | 1984 | * @error: %0 for success, < %0 for error |
e3a04fe3 KU |
1985 | * @nr_bytes: number of bytes to complete @rq |
1986 | * @bidi_bytes: number of bytes to complete @rq->next_rq | |
1987 | * | |
1988 | * Description: | |
1989 | * Ends I/O on a number of bytes attached to @rq and @rq->next_rq. | |
1990 | * | |
1991 | * Return: | |
710027a4 RD |
1992 | * %0 - we are done with this request |
1993 | * %1 - still buffers pending for this request | |
e3a04fe3 | 1994 | **/ |
22b13210 JA |
1995 | int blk_end_bidi_request(struct request *rq, int error, unsigned int nr_bytes, |
1996 | unsigned int bidi_bytes) | |
e3a04fe3 KU |
1997 | { |
1998 | return blk_end_io(rq, error, nr_bytes, bidi_bytes, NULL); | |
1999 | } | |
2000 | EXPORT_SYMBOL_GPL(blk_end_bidi_request); | |
2001 | ||
e19a3ab0 KU |
2002 | /** |
2003 | * blk_end_request_callback - Special helper function for tricky drivers | |
2004 | * @rq: the request being processed | |
710027a4 | 2005 | * @error: %0 for success, < %0 for error |
e19a3ab0 KU |
2006 | * @nr_bytes: number of bytes to complete |
2007 | * @drv_callback: function called between completion of bios in the request | |
2008 | * and completion of the request. | |
710027a4 | 2009 | * If the callback returns non %0, this helper returns without |
e19a3ab0 KU |
2010 | * completion of the request. |
2011 | * | |
2012 | * Description: | |
2013 | * Ends I/O on a number of bytes attached to @rq. | |
2014 | * If @rq has leftover, sets it up for the next range of segments. | |
2015 | * | |
2016 | * This special helper function is used only for existing tricky drivers. | |
2017 | * (e.g. cdrom_newpc_intr() of ide-cd) | |
2018 | * This interface will be removed when such drivers are rewritten. | |
2019 | * Don't use this interface in other places anymore. | |
2020 | * | |
2021 | * Return: | |
710027a4 RD |
2022 | * %0 - we are done with this request |
2023 | * %1 - this request is not freed yet. | |
2024 | * this request still has pending buffers or | |
2025 | * the driver doesn't want to finish this request yet. | |
e19a3ab0 | 2026 | **/ |
22b13210 JA |
2027 | int blk_end_request_callback(struct request *rq, int error, |
2028 | unsigned int nr_bytes, | |
e19a3ab0 KU |
2029 | int (drv_callback)(struct request *)) |
2030 | { | |
e3a04fe3 | 2031 | return blk_end_io(rq, error, nr_bytes, 0, drv_callback); |
e19a3ab0 KU |
2032 | } |
2033 | EXPORT_SYMBOL_GPL(blk_end_request_callback); | |
2034 | ||
86db1e29 JA |
2035 | void blk_rq_bio_prep(struct request_queue *q, struct request *rq, |
2036 | struct bio *bio) | |
1da177e4 | 2037 | { |
d628eaef DW |
2038 | /* Bit 0 (R/W) is identical in rq->cmd_flags and bio->bi_rw, and |
2039 | we want BIO_RW_AHEAD (bit 1) to imply REQ_FAILFAST (bit 1). */ | |
4aff5e23 | 2040 | rq->cmd_flags |= (bio->bi_rw & 3); |
1da177e4 | 2041 | |
fb2dce86 DW |
2042 | if (bio_has_data(bio)) { |
2043 | rq->nr_phys_segments = bio_phys_segments(q, bio); | |
fb2dce86 DW |
2044 | rq->buffer = bio_data(bio); |
2045 | } | |
1da177e4 LT |
2046 | rq->current_nr_sectors = bio_cur_sectors(bio); |
2047 | rq->hard_cur_sectors = rq->current_nr_sectors; | |
2048 | rq->hard_nr_sectors = rq->nr_sectors = bio_sectors(bio); | |
0e75f906 | 2049 | rq->data_len = bio->bi_size; |
1da177e4 LT |
2050 | |
2051 | rq->bio = rq->biotail = bio; | |
1da177e4 | 2052 | |
66846572 N |
2053 | if (bio->bi_bdev) |
2054 | rq->rq_disk = bio->bi_bdev->bd_disk; | |
2055 | } | |
1da177e4 LT |
2056 | |
2057 | int kblockd_schedule_work(struct work_struct *work) | |
2058 | { | |
2059 | return queue_work(kblockd_workqueue, work); | |
2060 | } | |
1da177e4 LT |
2061 | EXPORT_SYMBOL(kblockd_schedule_work); |
2062 | ||
19a75d83 | 2063 | void kblockd_flush_work(struct work_struct *work) |
1da177e4 | 2064 | { |
28e53bdd | 2065 | cancel_work_sync(work); |
1da177e4 | 2066 | } |
19a75d83 | 2067 | EXPORT_SYMBOL(kblockd_flush_work); |
1da177e4 LT |
2068 | |
2069 | int __init blk_dev_init(void) | |
2070 | { | |
ff856bad JA |
2071 | int i; |
2072 | ||
1da177e4 LT |
2073 | kblockd_workqueue = create_workqueue("kblockd"); |
2074 | if (!kblockd_workqueue) | |
2075 | panic("Failed to create kblockd\n"); | |
2076 | ||
2077 | request_cachep = kmem_cache_create("blkdev_requests", | |
20c2df83 | 2078 | sizeof(struct request), 0, SLAB_PANIC, NULL); |
1da177e4 | 2079 | |
8324aa91 | 2080 | blk_requestq_cachep = kmem_cache_create("blkdev_queue", |
165125e1 | 2081 | sizeof(struct request_queue), 0, SLAB_PANIC, NULL); |
1da177e4 | 2082 | |
0a945022 | 2083 | for_each_possible_cpu(i) |
ff856bad JA |
2084 | INIT_LIST_HEAD(&per_cpu(blk_cpu_done, i)); |
2085 | ||
962cf36c | 2086 | open_softirq(BLOCK_SOFTIRQ, blk_done_softirq); |
5a67e4c5 | 2087 | register_hotcpu_notifier(&blk_cpu_notifier); |
ff856bad | 2088 | |
d38ecf93 | 2089 | return 0; |
1da177e4 | 2090 | } |
1da177e4 | 2091 |