blk-merge.c 20.7 KB
Newer Older
1
// SPDX-License-Identifier: GPL-2.0
2 3 4 5 6 7 8 9 10
/*
 * Functions related to segment and merge handling
 */
#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/bio.h>
#include <linux/blkdev.h>
#include <linux/scatterlist.h>

11 12
#include <trace/events/block.h>

13 14
#include "blk.h"

15 16
static struct bio *blk_bio_discard_split(struct request_queue *q,
					 struct bio *bio,
17 18
					 struct bio_set *bs,
					 unsigned *nsegs)
19 20 21 22 23 24
{
	unsigned int max_discard_sectors, granularity;
	int alignment;
	sector_t tmp;
	unsigned split_sectors;

25 26
	*nsegs = 1;

27 28 29
	/* Zero-sector (unknown) and one-sector granularities are the same.  */
	granularity = max(q->limits.discard_granularity >> 9, 1U);

30 31
	max_discard_sectors = min(q->limits.max_discard_sectors,
			bio_allowed_max_sectors(q));
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
	max_discard_sectors -= max_discard_sectors % granularity;

	if (unlikely(!max_discard_sectors)) {
		/* XXX: warn */
		return NULL;
	}

	if (bio_sectors(bio) <= max_discard_sectors)
		return NULL;

	split_sectors = max_discard_sectors;

	/*
	 * If the next starting sector would be misaligned, stop the discard at
	 * the previous aligned sector.
	 */
	alignment = (q->limits.discard_alignment >> 9) % granularity;

	tmp = bio->bi_iter.bi_sector + split_sectors - alignment;
	tmp = sector_div(tmp, granularity);

	if (split_sectors > tmp)
		split_sectors -= tmp;

	return bio_split(bio, split_sectors, GFP_NOIO, bs);
}

59 60 61 62 63 64 65 66 67 68 69 70 71 72
static struct bio *blk_bio_write_zeroes_split(struct request_queue *q,
		struct bio *bio, struct bio_set *bs, unsigned *nsegs)
{
	*nsegs = 1;

	if (!q->limits.max_write_zeroes_sectors)
		return NULL;

	if (bio_sectors(bio) <= q->limits.max_write_zeroes_sectors)
		return NULL;

	return bio_split(bio, q->limits.max_write_zeroes_sectors, GFP_NOIO, bs);
}

73 74
static struct bio *blk_bio_write_same_split(struct request_queue *q,
					    struct bio *bio,
75 76
					    struct bio_set *bs,
					    unsigned *nsegs)
77
{
78 79
	*nsegs = 1;

80 81 82 83 84 85 86 87 88
	if (!q->limits.max_write_same_sectors)
		return NULL;

	if (bio_sectors(bio) <= q->limits.max_write_same_sectors)
		return NULL;

	return bio_split(bio, q->limits.max_write_same_sectors, GFP_NOIO, bs);
}

89 90 91 92 93 94 95 96 97 98 99 100
static inline unsigned get_max_io_size(struct request_queue *q,
				       struct bio *bio)
{
	unsigned sectors = blk_max_size_offset(q, bio->bi_iter.bi_sector);
	unsigned mask = queue_logical_block_size(q) - 1;

	/* aligned to logical block size */
	sectors &= ~(mask >> 9);

	return sectors;
}

101 102
static struct bio *blk_bio_segment_split(struct request_queue *q,
					 struct bio *bio,
103 104
					 struct bio_set *bs,
					 unsigned *segs)
105
{
106
	struct bio_vec bv, bvprv, *bvprvp = NULL;
107
	struct bvec_iter iter;
108
	unsigned seg_size = 0, nsegs = 0, sectors = 0;
M
Ming Lei 已提交
109 110 111
	unsigned front_seg_size = bio->bi_seg_front_size;
	bool do_split = true;
	struct bio *new = NULL;
112
	const unsigned max_sectors = get_max_io_size(q, bio);
113 114 115 116 117 118

	bio_for_each_segment(bv, bio, iter) {
		/*
		 * If the queue doesn't support SG gaps and adding this
		 * offset would create a gap, disallow it.
		 */
119
		if (bvprvp && bvec_gap_to_prev(q, bvprvp, bv.bv_offset))
120 121
			goto split;

122
		if (sectors + (bv.bv_len >> 9) > max_sectors) {
123 124 125 126 127
			/*
			 * Consider this a new segment if we're splitting in
			 * the middle of this vector.
			 */
			if (nsegs < queue_max_segments(q) &&
128
			    sectors < max_sectors) {
129
				nsegs++;
130
				sectors = max_sectors;
131
			}
132
			goto split;
133 134
		}

135
		if (bvprvp && blk_queue_cluster(q)) {
136 137
			if (seg_size + bv.bv_len > queue_max_segment_size(q))
				goto new_segment;
138
			if (!BIOVEC_PHYS_MERGEABLE(bvprvp, &bv))
139
				goto new_segment;
140
			if (!BIOVEC_SEG_BOUNDARY(q, bvprvp, &bv))
141 142 143 144
				goto new_segment;

			seg_size += bv.bv_len;
			bvprv = bv;
M
Ming Lei 已提交
145
			bvprvp = &bvprv;
146
			sectors += bv.bv_len >> 9;
147

148 149 150 151 152 153
			continue;
		}
new_segment:
		if (nsegs == queue_max_segments(q))
			goto split;

154 155 156
		if (nsegs == 1 && seg_size > front_seg_size)
			front_seg_size = seg_size;

157 158
		nsegs++;
		bvprv = bv;
M
Ming Lei 已提交
159
		bvprvp = &bvprv;
160
		seg_size = bv.bv_len;
161
		sectors += bv.bv_len >> 9;
M
Ming Lei 已提交
162

163 164
	}

M
Ming Lei 已提交
165
	do_split = false;
166
split:
167
	*segs = nsegs;
M
Ming Lei 已提交
168 169 170 171 172 173 174

	if (do_split) {
		new = bio_split(bio, sectors, GFP_NOIO, bs);
		if (new)
			bio = new;
	}

175 176
	if (nsegs == 1 && seg_size > front_seg_size)
		front_seg_size = seg_size;
M
Ming Lei 已提交
177 178 179 180 181
	bio->bi_seg_front_size = front_seg_size;
	if (seg_size > bio->bi_seg_back_size)
		bio->bi_seg_back_size = seg_size;

	return do_split ? new : NULL;
182 183
}

184
void blk_queue_split(struct request_queue *q, struct bio **bio)
185
{
186 187
	struct bio *split, *res;
	unsigned nsegs;
188

A
Adrian Hunter 已提交
189 190 191
	switch (bio_op(*bio)) {
	case REQ_OP_DISCARD:
	case REQ_OP_SECURE_ERASE:
192
		split = blk_bio_discard_split(q, *bio, &q->bio_split, &nsegs);
A
Adrian Hunter 已提交
193
		break;
194
	case REQ_OP_WRITE_ZEROES:
195
		split = blk_bio_write_zeroes_split(q, *bio, &q->bio_split, &nsegs);
196
		break;
A
Adrian Hunter 已提交
197
	case REQ_OP_WRITE_SAME:
198
		split = blk_bio_write_same_split(q, *bio, &q->bio_split, &nsegs);
A
Adrian Hunter 已提交
199 200
		break;
	default:
201
		split = blk_bio_segment_split(q, *bio, &q->bio_split, &nsegs);
A
Adrian Hunter 已提交
202 203
		break;
	}
204 205 206 207 208

	/* physical segments can be figured out during splitting */
	res = split ? split : *bio;
	res->bi_phys_segments = nsegs;
	bio_set_flag(res, BIO_SEG_VALID);
209 210

	if (split) {
M
Ming Lei 已提交
211
		/* there isn't chance to merge the splitted bio */
J
Jens Axboe 已提交
212
		split->bi_opf |= REQ_NOMERGE;
M
Ming Lei 已提交
213

214 215 216 217 218 219 220 221 222 223
		/*
		 * Since we're recursing into make_request here, ensure
		 * that we mark this bio as already having entered the queue.
		 * If not, and the queue is going away, we can get stuck
		 * forever on waiting for the queue reference to drop. But
		 * that will never happen, as we're already holding a
		 * reference to it.
		 */
		bio_set_flag(*bio, BIO_QUEUE_ENTERED);

224
		bio_chain(split, *bio);
225
		trace_block_split(q, split, (*bio)->bi_iter.bi_sector);
226 227 228 229 230 231
		generic_make_request(*bio);
		*bio = split;
	}
}
EXPORT_SYMBOL(blk_queue_split);

232
static unsigned int __blk_recalc_rq_segments(struct request_queue *q,
M
Ming Lei 已提交
233 234
					     struct bio *bio,
					     bool no_sg_merge)
235
{
236
	struct bio_vec bv, bvprv = { NULL };
237
	int cluster, prev = 0;
238
	unsigned int seg_size, nr_phys_segs;
239
	struct bio *fbio, *bbio;
240
	struct bvec_iter iter;
241

242 243
	if (!bio)
		return 0;
244

245 246 247 248
	switch (bio_op(bio)) {
	case REQ_OP_DISCARD:
	case REQ_OP_SECURE_ERASE:
	case REQ_OP_WRITE_ZEROES:
249 250
		return 0;
	case REQ_OP_WRITE_SAME:
251
		return 1;
252
	}
253

254
	fbio = bio;
255
	cluster = blk_queue_cluster(q);
M
Mikulas Patocka 已提交
256
	seg_size = 0;
257
	nr_phys_segs = 0;
258
	for_each_bio(bio) {
259
		bio_for_each_segment(bv, bio, iter) {
260 261 262 263 264 265 266
			/*
			 * If SG merging is disabled, each bio vector is
			 * a segment
			 */
			if (no_sg_merge)
				goto new_segment;

267
			if (prev && cluster) {
268
				if (seg_size + bv.bv_len
269
				    > queue_max_segment_size(q))
270
					goto new_segment;
271
				if (!BIOVEC_PHYS_MERGEABLE(&bvprv, &bv))
272
					goto new_segment;
273
				if (!BIOVEC_SEG_BOUNDARY(q, &bvprv, &bv))
274
					goto new_segment;
275

276
				seg_size += bv.bv_len;
277 278 279
				bvprv = bv;
				continue;
			}
280
new_segment:
281 282 283
			if (nr_phys_segs == 1 && seg_size >
			    fbio->bi_seg_front_size)
				fbio->bi_seg_front_size = seg_size;
284

285 286
			nr_phys_segs++;
			bvprv = bv;
287
			prev = 1;
288
			seg_size = bv.bv_len;
289
		}
290
		bbio = bio;
291 292
	}

293 294 295 296
	if (nr_phys_segs == 1 && seg_size > fbio->bi_seg_front_size)
		fbio->bi_seg_front_size = seg_size;
	if (seg_size > bbio->bi_seg_back_size)
		bbio->bi_seg_back_size = seg_size;
297 298 299 300 301 302

	return nr_phys_segs;
}

void blk_recalc_rq_segments(struct request *rq)
{
M
Ming Lei 已提交
303 304 305 306 307
	bool no_sg_merge = !!test_bit(QUEUE_FLAG_NO_SG_MERGE,
			&rq->q->queue_flags);

	rq->nr_phys_segments = __blk_recalc_rq_segments(rq->q, rq->bio,
			no_sg_merge);
308 309 310 311
}

void blk_recount_segments(struct request_queue *q, struct bio *bio)
{
312 313 314 315 316 317 318
	unsigned short seg_cnt;

	/* estimate segment number by bi_vcnt for non-cloned bio */
	if (bio_flagged(bio, BIO_CLONED))
		seg_cnt = bio_segments(bio);
	else
		seg_cnt = bio->bi_vcnt;
319

320 321 322
	if (test_bit(QUEUE_FLAG_NO_SG_MERGE, &q->queue_flags) &&
			(seg_cnt < queue_max_segments(q)))
		bio->bi_phys_segments = seg_cnt;
323 324 325 326
	else {
		struct bio *nxt = bio->bi_next;

		bio->bi_next = NULL;
327
		bio->bi_phys_segments = __blk_recalc_rq_segments(q, bio, false);
328 329
		bio->bi_next = nxt;
	}
330

331
	bio_set_flag(bio, BIO_SEG_VALID);
332 333 334 335 336 337
}
EXPORT_SYMBOL(blk_recount_segments);

static int blk_phys_contig_segment(struct request_queue *q, struct bio *bio,
				   struct bio *nxt)
{
338
	struct bio_vec end_bv = { NULL }, nxt_bv;
339

340
	if (!blk_queue_cluster(q))
341 342
		return 0;

343
	if (bio->bi_seg_back_size + nxt->bi_seg_front_size >
344
	    queue_max_segment_size(q))
345 346
		return 0;

347 348 349
	if (!bio_has_data(bio))
		return 1;

350 351
	bio_get_last_bvec(bio, &end_bv);
	bio_get_first_bvec(nxt, &nxt_bv);
352 353

	if (!BIOVEC_PHYS_MERGEABLE(&end_bv, &nxt_bv))
354 355
		return 0;

356
	/*
357
	 * bio and nxt are contiguous in memory; check if the queue allows
358 359
	 * these two to be merged into one
	 */
360
	if (BIOVEC_SEG_BOUNDARY(q, &end_bv, &nxt_bv))
361 362 363 364 365
		return 1;

	return 0;
}

366
static inline void
367
__blk_segment_map_sg(struct request_queue *q, struct bio_vec *bvec,
368
		     struct scatterlist *sglist, struct bio_vec *bvprv,
369 370 371 372 373
		     struct scatterlist **sg, int *nsegs, int *cluster)
{

	int nbytes = bvec->bv_len;

374
	if (*sg && *cluster) {
375 376 377
		if ((*sg)->length + nbytes > queue_max_segment_size(q))
			goto new_segment;

378
		if (!BIOVEC_PHYS_MERGEABLE(bvprv, bvec))
379
			goto new_segment;
380
		if (!BIOVEC_SEG_BOUNDARY(q, bvprv, bvec))
381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398
			goto new_segment;

		(*sg)->length += nbytes;
	} else {
new_segment:
		if (!*sg)
			*sg = sglist;
		else {
			/*
			 * If the driver previously mapped a shorter
			 * list, we could see a termination bit
			 * prematurely unless it fully inits the sg
			 * table on each mapping. We KNOW that there
			 * must be more entries here or the driver
			 * would be buggy, so force clear the
			 * termination bit to avoid doing a full
			 * sg_init_table() in drivers for each command.
			 */
399
			sg_unmark_end(*sg);
400 401 402 403 404 405
			*sg = sg_next(*sg);
		}

		sg_set_page(*sg, bvec->bv_page, nbytes, bvec->bv_offset);
		(*nsegs)++;
	}
406
	*bvprv = *bvec;
407 408
}

409 410 411 412 413 414 415 416
static inline int __blk_bvec_map_sg(struct request_queue *q, struct bio_vec bv,
		struct scatterlist *sglist, struct scatterlist **sg)
{
	*sg = sglist;
	sg_set_page(*sg, bv.bv_page, bv.bv_len, bv.bv_offset);
	return 1;
}

417 418 419
static int __blk_bios_map_sg(struct request_queue *q, struct bio *bio,
			     struct scatterlist *sglist,
			     struct scatterlist **sg)
420
{
421
	struct bio_vec bvec, bvprv = { NULL };
422
	struct bvec_iter iter;
423
	int cluster = blk_queue_cluster(q), nsegs = 0;
424 425 426 427 428

	for_each_bio(bio)
		bio_for_each_segment(bvec, bio, iter)
			__blk_segment_map_sg(q, &bvec, sglist, &bvprv, sg,
					     &nsegs, &cluster);
429

430 431 432 433 434 435 436 437 438 439 440 441 442
	return nsegs;
}

/*
 * map a request to scatterlist, return number of sg entries setup. Caller
 * must make sure sg can hold rq->nr_phys_segments entries
 */
int blk_rq_map_sg(struct request_queue *q, struct request *rq,
		  struct scatterlist *sglist)
{
	struct scatterlist *sg = NULL;
	int nsegs = 0;

443 444 445 446 447
	if (rq->rq_flags & RQF_SPECIAL_PAYLOAD)
		nsegs = __blk_bvec_map_sg(q, rq->special_vec, sglist, &sg);
	else if (rq->bio && bio_op(rq->bio) == REQ_OP_WRITE_SAME)
		nsegs = __blk_bvec_map_sg(q, bio_iovec(rq->bio), sglist, &sg);
	else if (rq->bio)
448
		nsegs = __blk_bios_map_sg(q, rq->bio, sglist, &sg);
449

450
	if (unlikely(rq->rq_flags & RQF_COPY_USER) &&
451 452 453
	    (blk_rq_bytes(rq) & q->dma_pad_mask)) {
		unsigned int pad_len =
			(q->dma_pad_mask & ~blk_rq_bytes(rq)) + 1;
454 455 456 457 458

		sg->length += pad_len;
		rq->extra_len += pad_len;
	}

459
	if (q->dma_drain_size && q->dma_drain_needed(rq)) {
460
		if (op_is_write(req_op(rq)))
461 462
			memset(q->dma_drain_buffer, 0, q->dma_drain_size);

463
		sg_unmark_end(sg);
464 465 466 467 468 469
		sg = sg_next(sg);
		sg_set_page(sg, virt_to_page(q->dma_drain_buffer),
			    q->dma_drain_size,
			    ((unsigned long)q->dma_drain_buffer) &
			    (PAGE_SIZE - 1));
		nsegs++;
470
		rq->extra_len += q->dma_drain_size;
471 472 473 474 475
	}

	if (sg)
		sg_mark_end(sg);

476 477 478 479
	/*
	 * Something must have been wrong if the figured number of
	 * segment is bigger than number of req's physical segments
	 */
480
	WARN_ON(nsegs > blk_rq_nr_phys_segments(rq));
481

482 483 484 485 486 487 488 489 490 491
	return nsegs;
}
EXPORT_SYMBOL(blk_rq_map_sg);

static inline int ll_new_hw_segment(struct request_queue *q,
				    struct request *req,
				    struct bio *bio)
{
	int nr_phys_segs = bio_phys_segments(q, bio);

492 493 494
	if (req->nr_phys_segments + nr_phys_segs > queue_max_segments(q))
		goto no_merge;

495
	if (blk_integrity_merge_bio(q, req, bio) == false)
496
		goto no_merge;
497 498 499 500 501 502 503

	/*
	 * This will form the start of a new hw segment.  Bump both
	 * counters.
	 */
	req->nr_phys_segments += nr_phys_segs;
	return 1;
504 505

no_merge:
R
Ritesh Harjani 已提交
506
	req_set_nomerge(q, req);
507
	return 0;
508 509 510 511 512
}

int ll_back_merge_fn(struct request_queue *q, struct request *req,
		     struct bio *bio)
{
513 514
	if (req_gap_back_merge(req, bio))
		return 0;
515 516 517
	if (blk_integrity_rq(req) &&
	    integrity_req_gap_back_merge(req, bio))
		return 0;
518
	if (blk_rq_sectors(req) + bio_sectors(bio) >
D
Damien Le Moal 已提交
519
	    blk_rq_get_max_sectors(req, blk_rq_pos(req))) {
R
Ritesh Harjani 已提交
520
		req_set_nomerge(q, req);
521 522
		return 0;
	}
523
	if (!bio_flagged(req->biotail, BIO_SEG_VALID))
524
		blk_recount_segments(q, req->biotail);
525
	if (!bio_flagged(bio, BIO_SEG_VALID))
526 527 528 529 530
		blk_recount_segments(q, bio);

	return ll_new_hw_segment(q, req, bio);
}

531
int ll_front_merge_fn(struct request_queue *q, struct request *req,
532 533
		      struct bio *bio)
{
534 535 536

	if (req_gap_front_merge(req, bio))
		return 0;
537 538 539
	if (blk_integrity_rq(req) &&
	    integrity_req_gap_front_merge(req, bio))
		return 0;
540
	if (blk_rq_sectors(req) + bio_sectors(bio) >
D
Damien Le Moal 已提交
541
	    blk_rq_get_max_sectors(req, bio->bi_iter.bi_sector)) {
R
Ritesh Harjani 已提交
542
		req_set_nomerge(q, req);
543 544
		return 0;
	}
545
	if (!bio_flagged(bio, BIO_SEG_VALID))
546
		blk_recount_segments(q, bio);
547
	if (!bio_flagged(req->bio, BIO_SEG_VALID))
548 549 550 551 552
		blk_recount_segments(q, req->bio);

	return ll_new_hw_segment(q, req, bio);
}

553 554 555 556 557 558 559 560 561 562 563
/*
 * blk-mq uses req->special to carry normal driver per-request payload, it
 * does not indicate a prepared command that we cannot merge with.
 */
static bool req_no_special_merge(struct request *req)
{
	struct request_queue *q = req->q;

	return !q->mq_ops && req->special;
}

564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581
static bool req_attempt_discard_merge(struct request_queue *q, struct request *req,
		struct request *next)
{
	unsigned short segments = blk_rq_nr_discard_segments(req);

	if (segments >= queue_max_discard_segments(q))
		goto no_merge;
	if (blk_rq_sectors(req) + bio_sectors(next->bio) >
	    blk_rq_get_max_sectors(req, blk_rq_pos(req)))
		goto no_merge;

	req->nr_phys_segments = segments + blk_rq_nr_discard_segments(next);
	return true;
no_merge:
	req_set_nomerge(q, req);
	return false;
}

582 583 584 585
static int ll_merge_requests_fn(struct request_queue *q, struct request *req,
				struct request *next)
{
	int total_phys_segments;
586 587
	unsigned int seg_size =
		req->biotail->bi_seg_back_size + next->bio->bi_seg_front_size;
588 589 590 591 592

	/*
	 * First check if the either of the requests are re-queued
	 * requests.  Can't merge them if they are.
	 */
593
	if (req_no_special_merge(req) || req_no_special_merge(next))
594 595
		return 0;

596
	if (req_gap_back_merge(req, next->bio))
597 598
		return 0;

599 600 601
	/*
	 * Will it become too large?
	 */
602
	if ((blk_rq_sectors(req) + blk_rq_sectors(next)) >
D
Damien Le Moal 已提交
603
	    blk_rq_get_max_sectors(req, blk_rq_pos(req)))
604 605 606
		return 0;

	total_phys_segments = req->nr_phys_segments + next->nr_phys_segments;
607 608 609 610 611
	if (blk_phys_contig_segment(q, req->biotail, next->bio)) {
		if (req->nr_phys_segments == 1)
			req->bio->bi_seg_front_size = seg_size;
		if (next->nr_phys_segments == 1)
			next->biotail->bi_seg_back_size = seg_size;
612
		total_phys_segments--;
613
	}
614

615
	if (total_phys_segments > queue_max_segments(q))
616 617
		return 0;

618
	if (blk_integrity_merge_rq(q, req, next) == false)
619 620
		return 0;

621 622 623 624 625
	/* Merge is OK... */
	req->nr_phys_segments = total_phys_segments;
	return 1;
}

626 627 628 629 630 631 632 633 634 635 636 637 638 639
/**
 * blk_rq_set_mixed_merge - mark a request as mixed merge
 * @rq: request to mark as mixed merge
 *
 * Description:
 *     @rq is about to be mixed merged.  Make sure the attributes
 *     which can be mixed are set in each bio and mark @rq as mixed
 *     merged.
 */
void blk_rq_set_mixed_merge(struct request *rq)
{
	unsigned int ff = rq->cmd_flags & REQ_FAILFAST_MASK;
	struct bio *bio;

640
	if (rq->rq_flags & RQF_MIXED_MERGE)
641 642 643 644 645 646 647 648
		return;

	/*
	 * @rq will no longer represent mixable attributes for all the
	 * contained bios.  It will just track those of the first one.
	 * Distributes the attributs to each bio.
	 */
	for (bio = rq->bio; bio; bio = bio->bi_next) {
J
Jens Axboe 已提交
649 650 651
		WARN_ON_ONCE((bio->bi_opf & REQ_FAILFAST_MASK) &&
			     (bio->bi_opf & REQ_FAILFAST_MASK) != ff);
		bio->bi_opf |= ff;
652
	}
653
	rq->rq_flags |= RQF_MIXED_MERGE;
654 655
}

656 657 658 659 660 661 662
static void blk_account_io_merge(struct request *req)
{
	if (blk_do_io_stat(req)) {
		struct hd_struct *part;
		int cpu;

		cpu = part_stat_lock();
663
		part = req->part;
664

665 666
		part_round_stats(req->q, cpu, part);
		part_dec_in_flight(req->q, part, rq_data_dir(req));
667

668
		hd_struct_put(part);
669 670 671 672
		part_stat_unlock();
	}
}

673
/*
674 675
 * For non-mq, this has to be called with the request spinlock acquired.
 * For mq with scheduling, the appropriate queue wide lock should be held.
676
 */
677 678
static struct request *attempt_merge(struct request_queue *q,
				     struct request *req, struct request *next)
679
{
680 681 682
	if (!q->mq_ops)
		lockdep_assert_held(q->queue_lock);

683
	if (!rq_mergeable(req) || !rq_mergeable(next))
684
		return NULL;
685

686
	if (req_op(req) != req_op(next))
687
		return NULL;
688

689 690 691
	/*
	 * not contiguous
	 */
692
	if (blk_rq_pos(req) + blk_rq_sectors(req) != blk_rq_pos(next))
693
		return NULL;
694 695 696

	if (rq_data_dir(req) != rq_data_dir(next)
	    || req->rq_disk != next->rq_disk
697
	    || req_no_special_merge(next))
698
		return NULL;
699

700
	if (req_op(req) == REQ_OP_WRITE_SAME &&
701
	    !blk_write_same_mergeable(req->bio, next->bio))
702
		return NULL;
703

704 705 706 707 708 709 710
	/*
	 * Don't allow merge of different write hints, or for a hint with
	 * non-hint IO.
	 */
	if (req->write_hint != next->write_hint)
		return NULL;

711 712 713 714
	/*
	 * If we are allowed to merge, then append bio list
	 * from next to rq and release next. merge_requests_fn
	 * will have updated segment counts, update sector
715 716
	 * counts here. Handle DISCARDs separately, as they
	 * have separate settings.
717
	 */
718 719 720 721
	if (req_op(req) == REQ_OP_DISCARD) {
		if (!req_attempt_discard_merge(q, req, next))
			return NULL;
	} else if (!ll_merge_requests_fn(q, req, next))
722
		return NULL;
723

724 725 726 727 728 729
	/*
	 * If failfast settings disagree or any of the two is already
	 * a mixed merge, mark both as mixed before proceeding.  This
	 * makes sure that all involved bios have mixable attributes
	 * set properly.
	 */
730
	if (((req->rq_flags | next->rq_flags) & RQF_MIXED_MERGE) ||
731 732 733 734 735 736
	    (req->cmd_flags & REQ_FAILFAST_MASK) !=
	    (next->cmd_flags & REQ_FAILFAST_MASK)) {
		blk_rq_set_mixed_merge(req);
		blk_rq_set_mixed_merge(next);
	}

737
	/*
738 739 740
	 * At this point we have either done a back merge or front merge. We
	 * need the smaller start_time_ns of the merged requests to be the
	 * current request for accounting purposes.
741
	 */
742 743
	if (next->start_time_ns < req->start_time_ns)
		req->start_time_ns = next->start_time_ns;
744 745 746 747

	req->biotail->bi_next = next->bio;
	req->biotail = next->biotail;

748
	req->__data_len += blk_rq_bytes(next);
749

750 751
	if (req_op(req) != REQ_OP_DISCARD)
		elv_merge_requests(q, req, next);
752

753 754 755 756
	/*
	 * 'next' is going away, so update stats accordingly
	 */
	blk_account_io_merge(next);
757 758

	req->ioprio = ioprio_best(req->ioprio, next->ioprio);
759 760
	if (blk_rq_cpu_valid(next))
		req->cpu = next->cpu;
761

762 763 764 765
	/*
	 * ownership of bio passed from next to req, return 'next' for
	 * the caller to free
	 */
766
	next->bio = NULL;
767
	return next;
768 769
}

770
struct request *attempt_back_merge(struct request_queue *q, struct request *rq)
771 772 773 774 775 776
{
	struct request *next = elv_latter_request(q, rq);

	if (next)
		return attempt_merge(q, rq, next);

777
	return NULL;
778 779
}

780
struct request *attempt_front_merge(struct request_queue *q, struct request *rq)
781 782 783 784 785 786
{
	struct request *prev = elv_former_request(q, rq);

	if (prev)
		return attempt_merge(q, prev, rq);

787
	return NULL;
788
}
789 790 791 792

int blk_attempt_req_merge(struct request_queue *q, struct request *rq,
			  struct request *next)
{
793
	struct elevator_queue *e = q->elevator;
794
	struct request *free;
795

796
	if (!e->uses_mq && e->type->ops.sq.elevator_allow_rq_merge_fn)
797
		if (!e->type->ops.sq.elevator_allow_rq_merge_fn(q, rq, next))
798 799
			return 0;

800 801 802 803 804 805 806
	free = attempt_merge(q, rq, next);
	if (free) {
		__blk_put_request(q, free);
		return 1;
	}

	return 0;
807
}
808 809 810

bool blk_rq_merge_ok(struct request *rq, struct bio *bio)
{
811
	if (!rq_mergeable(rq) || !bio_mergeable(bio))
812 813
		return false;

814
	if (req_op(rq) != bio_op(bio))
815 816
		return false;

817 818 819 820 821
	/* different data direction or already started, don't merge */
	if (bio_data_dir(bio) != rq_data_dir(rq))
		return false;

	/* must be same device and not a special request */
822
	if (rq->rq_disk != bio->bi_disk || req_no_special_merge(rq))
823 824 825
		return false;

	/* only merge integrity protected bio into ditto rq */
826
	if (blk_integrity_merge_bio(rq->q, rq, bio) == false)
827 828
		return false;

829
	/* must be using the same buffer */
830
	if (req_op(rq) == REQ_OP_WRITE_SAME &&
831 832 833
	    !blk_write_same_mergeable(rq->bio, bio))
		return false;

834 835 836 837 838 839 840
	/*
	 * Don't allow merge of different write hints, or for a hint with
	 * non-hint IO.
	 */
	if (rq->write_hint != bio->bi_write_hint)
		return false;

841 842 843
	return true;
}

844
enum elv_merge blk_try_merge(struct request *rq, struct bio *bio)
845
{
846 847 848 849
	if (req_op(rq) == REQ_OP_DISCARD &&
	    queue_max_discard_segments(rq->q) > 1)
		return ELEVATOR_DISCARD_MERGE;
	else if (blk_rq_pos(rq) + blk_rq_sectors(rq) == bio->bi_iter.bi_sector)
850
		return ELEVATOR_BACK_MERGE;
851
	else if (blk_rq_pos(rq) - bio_sectors(bio) == bio->bi_iter.bi_sector)
852 853 854
		return ELEVATOR_FRONT_MERGE;
	return ELEVATOR_NO_MERGE;
}