dm-rq.c 21.6 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
/*
 * Copyright (C) 2016 Red Hat, Inc. All rights reserved.
 *
 * This file is released under the GPL.
 */

#include "dm-core.h"
#include "dm-rq.h"

#include <linux/elevator.h> /* for rq_end_sector() */
#include <linux/blk-mq.h>

#define DM_MSG_PREFIX "core-rq"

#define DM_MQ_NR_HW_QUEUES 1
#define DM_MQ_QUEUE_DEPTH 2048
static unsigned dm_mq_nr_hw_queues = DM_MQ_NR_HW_QUEUES;
static unsigned dm_mq_queue_depth = DM_MQ_QUEUE_DEPTH;

/*
 * Request-based DM's mempools' reserved IOs set by the user.
 */
#define RESERVED_REQUEST_BASED_IOS	256
static unsigned reserved_rq_based_ios = RESERVED_REQUEST_BASED_IOS;

26
static bool use_blk_mq = IS_ENABLED(CONFIG_DM_MQ_DEFAULT);
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58

bool dm_use_blk_mq_default(void)
{
	return use_blk_mq;
}

bool dm_use_blk_mq(struct mapped_device *md)
{
	return md->use_blk_mq;
}
EXPORT_SYMBOL_GPL(dm_use_blk_mq);

unsigned dm_get_reserved_rq_based_ios(void)
{
	return __dm_get_module_param(&reserved_rq_based_ios,
				     RESERVED_REQUEST_BASED_IOS, DM_RESERVED_MAX_IOS);
}
EXPORT_SYMBOL_GPL(dm_get_reserved_rq_based_ios);

static unsigned dm_get_blk_mq_nr_hw_queues(void)
{
	return __dm_get_module_param(&dm_mq_nr_hw_queues, 1, 32);
}

static unsigned dm_get_blk_mq_queue_depth(void)
{
	return __dm_get_module_param(&dm_mq_queue_depth,
				     DM_MQ_QUEUE_DEPTH, BLK_MQ_MAX_DEPTH);
}

int dm_request_based(struct mapped_device *md)
{
59
	return queue_is_rq_based(md->queue);
60
61
62
63
64
65
66
67
68
69
70
71
}

static void dm_old_start_queue(struct request_queue *q)
{
	unsigned long flags;

	spin_lock_irqsave(q->queue_lock, flags);
	if (blk_queue_stopped(q))
		blk_start_queue(q);
	spin_unlock_irqrestore(q->queue_lock, flags);
}

72
73
static void dm_mq_start_queue(struct request_queue *q)
{
74
	blk_mq_unquiesce_queue(q);
75
76
77
	blk_mq_kick_requeue_list(q);
}

78
79
80
81
void dm_start_queue(struct request_queue *q)
{
	if (!q->mq_ops)
		dm_old_start_queue(q);
82
83
	else
		dm_mq_start_queue(q);
84
85
86
87
88
89
90
}

static void dm_old_stop_queue(struct request_queue *q)
{
	unsigned long flags;

	spin_lock_irqsave(q->queue_lock, flags);
91
92
	if (!blk_queue_stopped(q))
		blk_stop_queue(q);
93
94
95
	spin_unlock_irqrestore(q->queue_lock, flags);
}

96
97
static void dm_mq_stop_queue(struct request_queue *q)
{
98
	if (blk_mq_queue_stopped(q))
99
		return;
100

101
	blk_mq_quiesce_queue(q);
102
103
104
105
106
107
}

void dm_stop_queue(struct request_queue *q)
{
	if (!q->mq_ops)
		dm_old_stop_queue(q);
108
109
	else
		dm_mq_stop_queue(q);
110
111
112
113
114
115
116
117
118
119
120
}

/*
 * Partial completion handling for request-based dm
 */
static void end_clone_bio(struct bio *clone)
{
	struct dm_rq_clone_bio_info *info =
		container_of(clone, struct dm_rq_clone_bio_info, clone);
	struct dm_rq_target_io *tio = info->tio;
	unsigned int nr_bytes = info->orig->bi_iter.bi_size;
121
	blk_status_t error = clone->bi_status;
122
	bool is_last = !clone->bi_next;
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139

	bio_put(clone);

	if (tio->error)
		/*
		 * An error has already been detected on the request.
		 * Once error occurred, just let clone->end_io() handle
		 * the remainder.
		 */
		return;
	else if (error) {
		/*
		 * Don't notice the error to the upper layer yet.
		 * The error handling decision is made by the target driver,
		 * when the request is completed.
		 */
		tio->error = error;
140
		goto exit;
141
142
143
144
145
146
	}

	/*
	 * I/O for the bio successfully completed.
	 * Notice the data completion to the upper layer.
	 */
147
	tio->completed += nr_bytes;
148
149
150
151
152
153

	/*
	 * Update the original request.
	 * Do not use blk_end_request() here, because it may complete
	 * the original request before the clone, and break the ordering.
	 */
154
155
156
	if (is_last)
 exit:
		blk_update_request(tio->orig, BLK_STS_OK, tio->completed);
157
158
159
160
}

static struct dm_rq_target_io *tio_from_request(struct request *rq)
{
161
	return blk_mq_rq_to_pdu(rq);
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
}

static void rq_end_stats(struct mapped_device *md, struct request *orig)
{
	if (unlikely(dm_stats_used(&md->stats))) {
		struct dm_rq_target_io *tio = tio_from_request(orig);
		tio->duration_jiffies = jiffies - tio->duration_jiffies;
		dm_stats_account_io(&md->stats, rq_data_dir(orig),
				    blk_rq_pos(orig), tio->n_sectors, true,
				    tio->duration_jiffies, &tio->stats_aux);
	}
}

/*
 * Don't touch any member of the md after calling this function because
 * the md may be freed in dm_put() at the end of this function.
 * Or do dm_get() before calling this function and dm_put() later.
 */
static void rq_completed(struct mapped_device *md, int rw, bool run_queue)
{
182
183
184
	struct request_queue *q = md->queue;
	unsigned long flags;

185
186
187
188
189
190
191
192
193
194
195
196
	atomic_dec(&md->pending[rw]);

	/* nudge anyone waiting on suspend queue */
	if (!md_in_flight(md))
		wake_up(&md->wait);

	/*
	 * Run this off this callpath, as drivers could invoke end_io while
	 * inside their request_fn (and holding the queue lock). Calling
	 * back into ->request_fn() could deadlock attempting to grab the
	 * queue lock again.
	 */
197
198
199
200
201
	if (!q->mq_ops && run_queue) {
		spin_lock_irqsave(q->queue_lock, flags);
		blk_run_queue_async(q);
		spin_unlock_irqrestore(q->queue_lock, flags);
	}
202
203
204
205
206
207
208
209
210
211
212
213

	/*
	 * dm_put() must be at the end of this function. See the comment above
	 */
	dm_put(md);
}

/*
 * Complete the clone and the original request.
 * Must be called without clone's queue lock held,
 * see end_clone_request() for more details.
 */
214
static void dm_end_request(struct request *clone, blk_status_t error)
215
216
217
218
219
220
{
	int rw = rq_data_dir(clone);
	struct dm_rq_target_io *tio = clone->end_io_data;
	struct mapped_device *md = tio->md;
	struct request *rq = tio->orig;

221
222
	blk_rq_unprep_clone(clone);
	tio->ti->type->release_clone_rq(clone);
223
224
225
226
227
228
229
230
231
232
233
234

	rq_end_stats(md, rq);
	if (!rq->q->mq_ops)
		blk_end_request_all(rq, error);
	else
		blk_mq_end_request(rq, error);
	rq_completed(md, rw, true);
}

/*
 * Requeue the original request of a clone.
 */
235
static void dm_old_requeue_request(struct request *rq, unsigned long delay_ms)
236
237
238
239
240
241
{
	struct request_queue *q = rq->q;
	unsigned long flags;

	spin_lock_irqsave(q->queue_lock, flags);
	blk_requeue_request(q, rq);
242
	blk_delay_queue(q, delay_ms);
243
244
245
	spin_unlock_irqrestore(q->queue_lock, flags);
}

246
static void __dm_mq_kick_requeue_list(struct request_queue *q, unsigned long msecs)
247
{
248
	blk_mq_delay_kick_requeue_list(q, msecs);
249
250
}

251
252
253
254
255
256
257
258
void dm_mq_kick_requeue_list(struct mapped_device *md)
{
	__dm_mq_kick_requeue_list(dm_get_md_queue(md), 0);
}
EXPORT_SYMBOL(dm_mq_kick_requeue_list);

static void dm_mq_delay_requeue_request(struct request *rq, unsigned long msecs)
{
259
	blk_mq_requeue_request(rq, false);
260
261
262
	__dm_mq_kick_requeue_list(rq->q, msecs);
}

263
static void dm_requeue_original_request(struct dm_rq_target_io *tio, bool delay_requeue)
264
{
265
266
	struct mapped_device *md = tio->md;
	struct request *rq = tio->orig;
267
	int rw = rq_data_dir(rq);
268
	unsigned long delay_ms = delay_requeue ? 100 : 0;
269
270

	rq_end_stats(md, rq);
271
272
273
274
	if (tio->clone) {
		blk_rq_unprep_clone(tio->clone);
		tio->ti->type->release_clone_rq(tio->clone);
	}
275
276

	if (!rq->q->mq_ops)
277
		dm_old_requeue_request(rq, delay_ms);
278
	else
279
		dm_mq_delay_requeue_request(rq, delay_ms);
280
281
282
283

	rq_completed(md, rw, false);
}

284
static void dm_done(struct request *clone, blk_status_t error, bool mapped)
285
{
286
	int r = DM_ENDIO_DONE;
287
288
289
290
291
292
293
294
295
296
	struct dm_rq_target_io *tio = clone->end_io_data;
	dm_request_endio_fn rq_end_io = NULL;

	if (tio->ti) {
		rq_end_io = tio->ti->type->rq_end_io;

		if (mapped && rq_end_io)
			r = rq_end_io(tio->ti, clone, error, &tio->info);
	}

297
	if (unlikely(error == BLK_STS_TARGET)) {
298
299
300
301
302
303
304
		if (req_op(clone) == REQ_OP_WRITE_SAME &&
		    !clone->q->limits.max_write_same_sectors)
			disable_write_same(tio->md);
		if (req_op(clone) == REQ_OP_WRITE_ZEROES &&
		    !clone->q->limits.max_write_zeroes_sectors)
			disable_write_zeroes(tio->md);
	}
305

306
307
	switch (r) {
	case DM_ENDIO_DONE:
308
		/* The target wants to complete the I/O */
309
310
311
		dm_end_request(clone, error);
		break;
	case DM_ENDIO_INCOMPLETE:
312
313
		/* The target will handle the I/O */
		return;
314
	case DM_ENDIO_REQUEUE:
315
		/* The target wants to requeue the I/O */
316
		dm_requeue_original_request(tio, false);
317
		break;
318
319
320
321
	case DM_ENDIO_DELAY_REQUEUE:
		/* The target wants to requeue the I/O after a delay */
		dm_requeue_original_request(tio, true);
		break;
322
	default:
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
		DMWARN("unimplemented target endio return value: %d", r);
		BUG();
	}
}

/*
 * Request completion handler for request-based dm
 */
static void dm_softirq_done(struct request *rq)
{
	bool mapped = true;
	struct dm_rq_target_io *tio = tio_from_request(rq);
	struct request *clone = tio->clone;
	int rw;

	if (!clone) {
339
340
341
		struct mapped_device *md = tio->md;

		rq_end_stats(md, rq);
342
		rw = rq_data_dir(rq);
343
		if (!rq->q->mq_ops)
344
			blk_end_request_all(rq, tio->error);
345
		else
346
			blk_mq_end_request(rq, tio->error);
347
		rq_completed(md, rw, false);
348
349
350
		return;
	}

351
	if (rq->rq_flags & RQF_FAILED)
352
353
354
355
356
357
358
359
360
		mapped = false;

	dm_done(clone, tio->error, mapped);
}

/*
 * Complete the clone and the original request with the error status
 * through softirq context.
 */
361
static void dm_complete_request(struct request *rq, blk_status_t error)
362
363
364
365
366
367
368
{
	struct dm_rq_target_io *tio = tio_from_request(rq);

	tio->error = error;
	if (!rq->q->mq_ops)
		blk_complete_request(rq);
	else
369
		blk_mq_complete_request(rq);
370
371
372
373
374
375
376
377
}

/*
 * Complete the not-mapped clone and the original request with the error status
 * through softirq context.
 * Target's rq_end_io() function isn't called.
 * This may be used when the target's map_rq() or clone_and_map_rq() functions fail.
 */
378
static void dm_kill_unmapped_request(struct request *rq, blk_status_t error)
379
{
380
	rq->rq_flags |= RQF_FAILED;
381
382
383
384
385
386
	dm_complete_request(rq, error);
}

/*
 * Called with the clone's queue lock held (in the case of .request_fn)
 */
387
static void end_clone_request(struct request *clone, blk_status_t error)
388
389
390
391
392
393
394
395
396
397
398
399
400
401
{
	struct dm_rq_target_io *tio = clone->end_io_data;

	/*
	 * Actual request completion is done in a softirq context which doesn't
	 * hold the clone's queue lock.  Otherwise, deadlock could occur because:
	 *     - another request may be submitted by the upper level driver
	 *       of the stacking during the completion
	 *     - the submission which requires queue lock may be done
	 *       against this clone's queue
	 */
	dm_complete_request(tio->orig, error);
}

402
static blk_status_t dm_dispatch_clone_request(struct request *clone, struct request *rq)
403
{
404
	blk_status_t r;
405
406

	if (blk_queue_io_stat(clone->q))
407
		clone->rq_flags |= RQF_IO_STAT;
408
409
410

	clone->start_time = jiffies;
	r = blk_insert_cloned_request(clone->q, clone);
411
	if (r != BLK_STS_OK && r != BLK_STS_RESOURCE && r != BLK_STS_DEV_RESOURCE)
412
413
		/* must complete clone in terms of original request */
		dm_complete_request(rq, r);
414
	return r;
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
}

static int dm_rq_bio_constructor(struct bio *bio, struct bio *bio_orig,
				 void *data)
{
	struct dm_rq_target_io *tio = data;
	struct dm_rq_clone_bio_info *info =
		container_of(bio, struct dm_rq_clone_bio_info, clone);

	info->orig = bio_orig;
	info->tio = tio;
	bio->bi_end_io = end_clone_bio;

	return 0;
}

static int setup_clone(struct request *clone, struct request *rq,
		       struct dm_rq_target_io *tio, gfp_t gfp_mask)
{
	int r;

	r = blk_rq_prep_clone(clone, rq, tio->md->bs, gfp_mask,
			      dm_rq_bio_constructor, tio);
	if (r)
		return r;

	clone->end_io = end_clone_request;
	clone->end_io_data = tio;

	tio->clone = clone;

	return 0;
}

static void map_tio_request(struct kthread_work *work);

static void init_tio(struct dm_rq_target_io *tio, struct request *rq,
		     struct mapped_device *md)
{
	tio->md = md;
	tio->ti = NULL;
	tio->clone = NULL;
	tio->orig = rq;
	tio->error = 0;
459
	tio->completed = 0;
460
461
462
463
464
465
466
467
	/*
	 * Avoid initializing info for blk-mq; it passes
	 * target-specific data through info.ptr
	 * (see: dm_mq_init_request)
	 */
	if (!md->init_tio_pdu)
		memset(&tio->info, 0, sizeof(tio->info));
	if (md->kworker_task)
468
		kthread_init_work(&tio->work, map_tio_request);
469
470
471
472
}

/*
 * Returns:
473
474
 * DM_MAPIO_*       : the request has been processed as indicated
 * DM_MAPIO_REQUEUE : the original request needs to be immediately requeued
475
476
 * < 0              : the request was completed due to failure
 */
477
static int map_request(struct dm_rq_target_io *tio)
478
479
480
{
	int r;
	struct dm_target *ti = tio->ti;
481
482
	struct mapped_device *md = tio->md;
	struct request *rq = tio->orig;
483
	struct request *clone = NULL;
484
	blk_status_t ret;
485

486
	r = ti->type->clone_and_map_rq(ti, rq, &tio->info, &clone);
487
check_again:
488
489
490
491
492
	switch (r) {
	case DM_MAPIO_SUBMITTED:
		/* The target has taken the I/O to submit by itself later */
		break;
	case DM_MAPIO_REMAPPED:
493
494
495
496
497
498
		if (setup_clone(clone, rq, tio, GFP_ATOMIC)) {
			/* -ENOMEM */
			ti->type->release_clone_rq(clone);
			return DM_MAPIO_REQUEUE;
		}

499
500
501
		/* The target has remapped the I/O so dispatch it */
		trace_block_rq_remap(clone->q, clone, disk_devt(dm_disk(md)),
				     blk_rq_pos(rq));
502
		ret = dm_dispatch_clone_request(clone, rq);
503
		if (ret == BLK_STS_RESOURCE || ret == BLK_STS_DEV_RESOURCE) {
504
505
506
507
508
509
510
511
512
			blk_rq_unprep_clone(clone);
			tio->ti->type->release_clone_rq(clone);
			tio->clone = NULL;
			if (!rq->q->mq_ops)
				r = DM_MAPIO_DELAY_REQUEUE;
			else
				r = DM_MAPIO_REQUEUE;
			goto check_again;
		}
513
514
515
		break;
	case DM_MAPIO_REQUEUE:
		/* The target wants to requeue the I/O */
516
517
518
		break;
	case DM_MAPIO_DELAY_REQUEUE:
		/* The target wants to requeue the I/O after a delay */
519
		dm_requeue_original_request(tio, true);
520
		break;
521
	case DM_MAPIO_KILL:
522
		/* The target wants to complete the I/O */
523
		dm_kill_unmapped_request(rq, BLK_STS_IOERR);
524
		break;
525
526
527
	default:
		DMWARN("unimplemented target map return value: %d", r);
		BUG();
528
529
	}

530
	return r;
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
}

static void dm_start_request(struct mapped_device *md, struct request *orig)
{
	if (!orig->q->mq_ops)
		blk_start_request(orig);
	else
		blk_mq_start_request(orig);
	atomic_inc(&md->pending[rq_data_dir(orig)]);

	if (md->seq_rq_merge_deadline_usecs) {
		md->last_rq_pos = rq_end_sector(orig);
		md->last_rq_rw = rq_data_dir(orig);
		md->last_rq_start_time = ktime_get();
	}

	if (unlikely(dm_stats_used(&md->stats))) {
		struct dm_rq_target_io *tio = tio_from_request(orig);
		tio->duration_jiffies = jiffies;
		tio->n_sectors = blk_rq_sectors(orig);
		dm_stats_account_io(&md->stats, rq_data_dir(orig),
				    blk_rq_pos(orig), tio->n_sectors, false, 0,
				    &tio->stats_aux);
	}

	/*
	 * Hold the md reference here for the in-flight I/O.
	 * We can't rely on the reference count by device opener,
	 * because the device may be closed during the request completion
	 * when all bios are completed.
	 * See the comment in rq_completed() too.
	 */
	dm_get(md);
}

566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
static int __dm_rq_init_rq(struct mapped_device *md, struct request *rq)
{
	struct dm_rq_target_io *tio = blk_mq_rq_to_pdu(rq);

	/*
	 * Must initialize md member of tio, otherwise it won't
	 * be available in dm_mq_queue_rq.
	 */
	tio->md = md;

	if (md->init_tio_pdu) {
		/* target-specific per-io data is immediately after the tio */
		tio->info.ptr = tio + 1;
	}

	return 0;
}

static int dm_rq_init_rq(struct request_queue *q, struct request *rq, gfp_t gfp)
{
	return __dm_rq_init_rq(q->rq_alloc_data, rq);
}

589
590
591
592
static void map_tio_request(struct kthread_work *work)
{
	struct dm_rq_target_io *tio = container_of(work, struct dm_rq_target_io, work);

593
594
	if (map_request(tio) == DM_MAPIO_REQUEUE)
		dm_requeue_original_request(tio, false);
595
596
597
598
599
600
601
602
603
604
605
606
607
608
}

ssize_t dm_attr_rq_based_seq_io_merge_deadline_show(struct mapped_device *md, char *buf)
{
	return sprintf(buf, "%u\n", md->seq_rq_merge_deadline_usecs);
}

#define MAX_SEQ_RQ_MERGE_DEADLINE_USECS 100000

ssize_t dm_attr_rq_based_seq_io_merge_deadline_store(struct mapped_device *md,
						     const char *buf, size_t count)
{
	unsigned deadline;

609
	if (dm_get_md_type(md) != DM_TYPE_REQUEST_BASED)
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
		return count;

	if (kstrtouint(buf, 10, &deadline))
		return -EINVAL;

	if (deadline > MAX_SEQ_RQ_MERGE_DEADLINE_USECS)
		deadline = MAX_SEQ_RQ_MERGE_DEADLINE_USECS;

	md->seq_rq_merge_deadline_usecs = deadline;

	return count;
}

static bool dm_old_request_peeked_before_merge_deadline(struct mapped_device *md)
{
	ktime_t kt_deadline;

	if (!md->seq_rq_merge_deadline_usecs)
		return false;

	kt_deadline = ns_to_ktime((u64)md->seq_rq_merge_deadline_usecs * NSEC_PER_USEC);
	kt_deadline = ktime_add_safe(md->last_rq_start_time, kt_deadline);

	return !ktime_after(ktime_get(), kt_deadline);
}

/*
 * q->request_fn for old request-based dm.
 * Called with the queue lock held.
 */
static void dm_old_request_fn(struct request_queue *q)
{
	struct mapped_device *md = q->queuedata;
	struct dm_target *ti = md->immutable_target;
	struct request *rq;
	struct dm_rq_target_io *tio;
	sector_t pos = 0;

	if (unlikely(!ti)) {
		int srcu_idx;
		struct dm_table *map = dm_get_live_table(md, &srcu_idx);

652
653
654
655
		if (unlikely(!map)) {
			dm_put_live_table(md, srcu_idx);
			return;
		}
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
		ti = dm_table_find_target(map, pos);
		dm_put_live_table(md, srcu_idx);
	}

	/*
	 * For suspend, check blk_queue_stopped() and increment
	 * ->pending within a single queue_lock not to increment the
	 * number of in-flight I/Os after the queue is stopped in
	 * dm_suspend().
	 */
	while (!blk_queue_stopped(q)) {
		rq = blk_peek_request(q);
		if (!rq)
			return;

		/* always use block 0 to find the target for flushes for now */
		pos = 0;
		if (req_op(rq) != REQ_OP_FLUSH)
			pos = blk_rq_pos(rq);

		if ((dm_old_request_peeked_before_merge_deadline(md) &&
677
		     md_in_flight(md) && rq->bio && !bio_multiple_segments(rq->bio) &&
678
679
		     md->last_rq_pos == pos && md->last_rq_rw == rq_data_dir(rq)) ||
		    (ti->type->busy && ti->type->busy(ti))) {
680
			blk_delay_queue(q, 10);
681
682
683
684
685
686
			return;
		}

		dm_start_request(md, rq);

		tio = tio_from_request(rq);
687
		init_tio(tio, rq, md);
688
689
		/* Establish tio->ti before queuing work (map_tio_request) */
		tio->ti = ti;
690
		kthread_queue_work(&md->kworker, &tio->work);
691
692
693
694
695
696
697
		BUG_ON(!irqs_disabled());
	}
}

/*
 * Fully initialize a .request_fn request-based queue.
 */
698
int dm_old_init_request_queue(struct mapped_device *md, struct dm_table *t)
699
{
700
701
	struct dm_target *immutable_tgt;

702
	/* Fully initialize the queue */
703
704
	md->queue->cmd_size = sizeof(struct dm_rq_target_io);
	md->queue->rq_alloc_data = md;
705
	md->queue->request_fn = dm_old_request_fn;
706
707
708
709
710
711
712
713
	md->queue->init_rq_fn = dm_rq_init_rq;

	immutable_tgt = dm_table_get_immutable_target(t);
	if (immutable_tgt && immutable_tgt->per_io_data_size) {
		/* any target-specific per-io data is immediately after the tio */
		md->queue->cmd_size += immutable_tgt->per_io_data_size;
		md->init_tio_pdu = true;
	}
714
	if (blk_init_allocated_queue(md->queue) < 0)
715
716
717
718
719
720
721
722
		return -EINVAL;

	/* disable dm_old_request_fn's merge heuristic by default */
	md->seq_rq_merge_deadline_usecs = 0;

	blk_queue_softirq_done(md->queue, dm_softirq_done);

	/* Initialize the request-based DM worker thread */
723
	kthread_init_worker(&md->kworker);
724
725
	md->kworker_task = kthread_run(kthread_worker_fn, &md->kworker,
				       "kdmwork-%s", dm_device_name(md));
726
727
728
729
730
	if (IS_ERR(md->kworker_task)) {
		int error = PTR_ERR(md->kworker_task);
		md->kworker_task = NULL;
		return error;
	}
731
732
733
734

	return 0;
}

735
736
static int dm_mq_init_request(struct blk_mq_tag_set *set, struct request *rq,
		unsigned int hctx_idx, unsigned int numa_node)
737
{
738
	return __dm_rq_init_rq(set->driver_data, rq);
739
740
}

741
static blk_status_t dm_mq_queue_rq(struct blk_mq_hw_ctx *hctx,
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
			  const struct blk_mq_queue_data *bd)
{
	struct request *rq = bd->rq;
	struct dm_rq_target_io *tio = blk_mq_rq_to_pdu(rq);
	struct mapped_device *md = tio->md;
	struct dm_target *ti = md->immutable_target;

	if (unlikely(!ti)) {
		int srcu_idx;
		struct dm_table *map = dm_get_live_table(md, &srcu_idx);

		ti = dm_table_find_target(map, 0);
		dm_put_live_table(md, srcu_idx);
	}

	if (ti->type->busy && ti->type->busy(ti))
758
		return BLK_STS_RESOURCE;
759
760
761
762
763
764
765
766
767
768
769
770

	dm_start_request(md, rq);

	/* Init tio using md established in .init_request */
	init_tio(tio, rq, md);

	/*
	 * Establish tio->ti before calling map_request().
	 */
	tio->ti = ti;

	/* Direct call is fine since .queue_rq allows allocations */
771
	if (map_request(tio) == DM_MAPIO_REQUEUE) {
772
773
774
		/* Undo dm_start_request() before requeuing */
		rq_end_stats(md, rq);
		rq_completed(md, rq_data_dir(rq), false);
775
		return BLK_STS_RESOURCE;
776
777
	}

778
	return BLK_STS_OK;
779
780
}

781
static const struct blk_mq_ops dm_mq_ops = {
782
783
784
785
786
	.queue_rq = dm_mq_queue_rq,
	.complete = dm_softirq_done,
	.init_request = dm_mq_init_request,
};

787
int dm_mq_init_request_queue(struct mapped_device *md, struct dm_table *t)
788
789
{
	struct request_queue *q;
790
	struct dm_target *immutable_tgt;
791
792
	int err;

793
	if (!dm_table_all_blk_mq_devices(t)) {
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
		DMERR("request-based dm-mq may only be stacked on blk-mq device(s)");
		return -EINVAL;
	}

	md->tag_set = kzalloc_node(sizeof(struct blk_mq_tag_set), GFP_KERNEL, md->numa_node_id);
	if (!md->tag_set)
		return -ENOMEM;

	md->tag_set->ops = &dm_mq_ops;
	md->tag_set->queue_depth = dm_get_blk_mq_queue_depth();
	md->tag_set->numa_node = md->numa_node_id;
	md->tag_set->flags = BLK_MQ_F_SHOULD_MERGE | BLK_MQ_F_SG_MERGE;
	md->tag_set->nr_hw_queues = dm_get_blk_mq_nr_hw_queues();
	md->tag_set->driver_data = md;

	md->tag_set->cmd_size = sizeof(struct dm_rq_target_io);
810
	immutable_tgt = dm_table_get_immutable_target(t);
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
	if (immutable_tgt && immutable_tgt->per_io_data_size) {
		/* any target-specific per-io data is immediately after the tio */
		md->tag_set->cmd_size += immutable_tgt->per_io_data_size;
		md->init_tio_pdu = true;
	}

	err = blk_mq_alloc_tag_set(md->tag_set);
	if (err)
		goto out_kfree_tag_set;

	q = blk_mq_init_allocated_queue(md->tag_set, md->queue);
	if (IS_ERR(q)) {
		err = PTR_ERR(q);
		goto out_tag_set;
	}

	return 0;

out_tag_set:
	blk_mq_free_tag_set(md->tag_set);
out_kfree_tag_set:
	kfree(md->tag_set);

	return err;
}

void dm_mq_cleanup_mapped_device(struct mapped_device *md)
{
	if (md->tag_set) {
		blk_mq_free_tag_set(md->tag_set);
		kfree(md->tag_set);
	}
}

module_param(reserved_rq_based_ios, uint, S_IRUGO | S_IWUSR);
MODULE_PARM_DESC(reserved_rq_based_ios, "Reserved IOs in request-based mempools");

module_param(use_blk_mq, bool, S_IRUGO | S_IWUSR);
MODULE_PARM_DESC(use_blk_mq, "Use block multiqueue for request-based DM devices");

module_param(dm_mq_nr_hw_queues, uint, S_IRUGO | S_IWUSR);
MODULE_PARM_DESC(dm_mq_nr_hw_queues, "Number of hardware queues for request-based dm-mq devices");

module_param(dm_mq_queue_depth, uint, S_IRUGO | S_IWUSR);
MODULE_PARM_DESC(dm_mq_queue_depth, "Queue depth for request-based dm-mq devices");