blk-merge.c 19.8 KB
Newer Older
1 2 3 4 5 6 7 8 9
/*
 * 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>

10 11
#include <trace/events/block.h>

12 13
#include "blk.h"

14 15
static struct bio *blk_bio_discard_split(struct request_queue *q,
					 struct bio *bio,
16 17
					 struct bio_set *bs,
					 unsigned *nsegs)
18 19 20 21 22 23
{
	unsigned int max_discard_sectors, granularity;
	int alignment;
	sector_t tmp;
	unsigned split_sectors;

24 25
	*nsegs = 1;

26 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
	/* Zero-sector (unknown) and one-sector granularities are the same.  */
	granularity = max(q->limits.discard_granularity >> 9, 1U);

	max_discard_sectors = min(q->limits.max_discard_sectors, UINT_MAX >> 9);
	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);
}

57 58 59 60 61 62 63 64 65 66 67 68 69 70
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);
}

71 72
static struct bio *blk_bio_write_same_split(struct request_queue *q,
					    struct bio *bio,
73 74
					    struct bio_set *bs,
					    unsigned *nsegs)
75
{
76 77
	*nsegs = 1;

78 79 80 81 82 83 84 85 86
	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);
}

87 88 89 90 91 92 93 94 95 96 97 98
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;
}

99 100
static struct bio *blk_bio_segment_split(struct request_queue *q,
					 struct bio *bio,
101 102
					 struct bio_set *bs,
					 unsigned *segs)
103
{
104
	struct bio_vec bv, bvprv, *bvprvp = NULL;
105
	struct bvec_iter iter;
106
	unsigned seg_size = 0, nsegs = 0, sectors = 0;
M
Ming Lei 已提交
107 108 109
	unsigned front_seg_size = bio->bi_seg_front_size;
	bool do_split = true;
	struct bio *new = NULL;
110
	const unsigned max_sectors = get_max_io_size(q, bio);
111
	unsigned bvecs = 0;
112 113

	bio_for_each_segment(bv, bio, iter) {
114 115 116 117 118 119
		/*
		 * With arbitrary bio size, the incoming bio may be very
		 * big. We have to split the bio into small bios so that
		 * each holds at most BIO_MAX_PAGES bvecs because
		 * bio_clone() can fail to allocate big bvecs.
		 *
120 121 122
		 * Those drivers which will need to use bio_clone()
		 * should tell us in some way.  For now, impose the
		 * BIO_MAX_PAGES limit on all queues.
123
		 *
124
		 * TODO: handle users of bio_clone() differently.
125 126 127 128
		 */
		if (bvecs++ >= BIO_MAX_PAGES)
			goto split;

129 130 131 132
		/*
		 * If the queue doesn't support SG gaps and adding this
		 * offset would create a gap, disallow it.
		 */
133
		if (bvprvp && bvec_gap_to_prev(q, bvprvp, bv.bv_offset))
134 135
			goto split;

136
		if (sectors + (bv.bv_len >> 9) > max_sectors) {
137 138 139 140 141
			/*
			 * Consider this a new segment if we're splitting in
			 * the middle of this vector.
			 */
			if (nsegs < queue_max_segments(q) &&
142
			    sectors < max_sectors) {
143
				nsegs++;
144
				sectors = max_sectors;
145
			}
146 147 148
			if (sectors)
				goto split;
			/* Make this single bvec as the 1st segment */
149 150
		}

151
		if (bvprvp && blk_queue_cluster(q)) {
152 153
			if (seg_size + bv.bv_len > queue_max_segment_size(q))
				goto new_segment;
154
			if (!BIOVEC_PHYS_MERGEABLE(bvprvp, &bv))
155
				goto new_segment;
156
			if (!BIOVEC_SEG_BOUNDARY(q, bvprvp, &bv))
157 158 159 160
				goto new_segment;

			seg_size += bv.bv_len;
			bvprv = bv;
M
Ming Lei 已提交
161
			bvprvp = &bvprv;
162
			sectors += bv.bv_len >> 9;
163 164 165

			if (nsegs == 1 && seg_size > front_seg_size)
				front_seg_size = seg_size;
166 167 168 169 170 171 172 173
			continue;
		}
new_segment:
		if (nsegs == queue_max_segments(q))
			goto split;

		nsegs++;
		bvprv = bv;
M
Ming Lei 已提交
174
		bvprvp = &bvprv;
175
		seg_size = bv.bv_len;
176
		sectors += bv.bv_len >> 9;
M
Ming Lei 已提交
177 178 179

		if (nsegs == 1 && seg_size > front_seg_size)
			front_seg_size = seg_size;
180 181
	}

M
Ming Lei 已提交
182
	do_split = false;
183
split:
184
	*segs = nsegs;
M
Ming Lei 已提交
185 186 187 188 189 190 191 192 193 194 195 196

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

	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;
197 198
}

199
void blk_queue_split(struct request_queue *q, struct bio **bio)
200
{
201 202
	struct bio *split, *res;
	unsigned nsegs;
203

A
Adrian Hunter 已提交
204 205 206
	switch (bio_op(*bio)) {
	case REQ_OP_DISCARD:
	case REQ_OP_SECURE_ERASE:
207
		split = blk_bio_discard_split(q, *bio, q->bio_split, &nsegs);
A
Adrian Hunter 已提交
208
		break;
209
	case REQ_OP_WRITE_ZEROES:
210
		split = blk_bio_write_zeroes_split(q, *bio, q->bio_split, &nsegs);
211
		break;
A
Adrian Hunter 已提交
212
	case REQ_OP_WRITE_SAME:
213
		split = blk_bio_write_same_split(q, *bio, q->bio_split, &nsegs);
A
Adrian Hunter 已提交
214 215
		break;
	default:
216
		split = blk_bio_segment_split(q, *bio, q->bio_split, &nsegs);
A
Adrian Hunter 已提交
217 218
		break;
	}
219 220 221 222 223

	/* physical segments can be figured out during splitting */
	res = split ? split : *bio;
	res->bi_phys_segments = nsegs;
	bio_set_flag(res, BIO_SEG_VALID);
224 225

	if (split) {
M
Ming Lei 已提交
226
		/* there isn't chance to merge the splitted bio */
J
Jens Axboe 已提交
227
		split->bi_opf |= REQ_NOMERGE;
M
Ming Lei 已提交
228

229
		bio_chain(split, *bio);
230
		trace_block_split(q, split, (*bio)->bi_iter.bi_sector);
231 232 233 234 235 236
		generic_make_request(*bio);
		*bio = split;
	}
}
EXPORT_SYMBOL(blk_queue_split);

237
static unsigned int __blk_recalc_rq_segments(struct request_queue *q,
M
Ming Lei 已提交
238 239
					     struct bio *bio,
					     bool no_sg_merge)
240
{
241
	struct bio_vec bv, bvprv = { NULL };
242
	int cluster, prev = 0;
243
	unsigned int seg_size, nr_phys_segs;
244
	struct bio *fbio, *bbio;
245
	struct bvec_iter iter;
246

247 248
	if (!bio)
		return 0;
249

250 251 252 253
	switch (bio_op(bio)) {
	case REQ_OP_DISCARD:
	case REQ_OP_SECURE_ERASE:
	case REQ_OP_WRITE_ZEROES:
254 255
		return 0;
	case REQ_OP_WRITE_SAME:
256
		return 1;
257
	}
258

259
	fbio = bio;
260
	cluster = blk_queue_cluster(q);
M
Mikulas Patocka 已提交
261
	seg_size = 0;
262
	nr_phys_segs = 0;
263
	for_each_bio(bio) {
264
		bio_for_each_segment(bv, bio, iter) {
265 266 267 268 269 270 271
			/*
			 * If SG merging is disabled, each bio vector is
			 * a segment
			 */
			if (no_sg_merge)
				goto new_segment;

272
			if (prev && cluster) {
273
				if (seg_size + bv.bv_len
274
				    > queue_max_segment_size(q))
275
					goto new_segment;
276
				if (!BIOVEC_PHYS_MERGEABLE(&bvprv, &bv))
277
					goto new_segment;
278
				if (!BIOVEC_SEG_BOUNDARY(q, &bvprv, &bv))
279
					goto new_segment;
280

281
				seg_size += bv.bv_len;
282 283 284
				bvprv = bv;
				continue;
			}
285
new_segment:
286 287 288
			if (nr_phys_segs == 1 && seg_size >
			    fbio->bi_seg_front_size)
				fbio->bi_seg_front_size = seg_size;
289

290 291
			nr_phys_segs++;
			bvprv = bv;
292
			prev = 1;
293
			seg_size = bv.bv_len;
294
		}
295
		bbio = bio;
296 297
	}

298 299 300 301
	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;
302 303 304 305 306 307

	return nr_phys_segs;
}

void blk_recalc_rq_segments(struct request *rq)
{
M
Ming Lei 已提交
308 309 310 311 312
	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);
313 314 315 316
}

void blk_recount_segments(struct request_queue *q, struct bio *bio)
{
317 318 319 320 321 322 323
	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;
324

325 326 327
	if (test_bit(QUEUE_FLAG_NO_SG_MERGE, &q->queue_flags) &&
			(seg_cnt < queue_max_segments(q)))
		bio->bi_phys_segments = seg_cnt;
328 329 330 331
	else {
		struct bio *nxt = bio->bi_next;

		bio->bi_next = NULL;
332
		bio->bi_phys_segments = __blk_recalc_rq_segments(q, bio, false);
333 334
		bio->bi_next = nxt;
	}
335

336
	bio_set_flag(bio, BIO_SEG_VALID);
337 338 339 340 341 342
}
EXPORT_SYMBOL(blk_recount_segments);

static int blk_phys_contig_segment(struct request_queue *q, struct bio *bio,
				   struct bio *nxt)
{
343
	struct bio_vec end_bv = { NULL }, nxt_bv;
344

345
	if (!blk_queue_cluster(q))
346 347
		return 0;

348
	if (bio->bi_seg_back_size + nxt->bi_seg_front_size >
349
	    queue_max_segment_size(q))
350 351
		return 0;

352 353 354
	if (!bio_has_data(bio))
		return 1;

355 356
	bio_get_last_bvec(bio, &end_bv);
	bio_get_first_bvec(nxt, &nxt_bv);
357 358

	if (!BIOVEC_PHYS_MERGEABLE(&end_bv, &nxt_bv))
359 360
		return 0;

361
	/*
362
	 * bio and nxt are contiguous in memory; check if the queue allows
363 364
	 * these two to be merged into one
	 */
365
	if (BIOVEC_SEG_BOUNDARY(q, &end_bv, &nxt_bv))
366 367 368 369 370
		return 1;

	return 0;
}

371
static inline void
372
__blk_segment_map_sg(struct request_queue *q, struct bio_vec *bvec,
373
		     struct scatterlist *sglist, struct bio_vec *bvprv,
374 375 376 377 378
		     struct scatterlist **sg, int *nsegs, int *cluster)
{

	int nbytes = bvec->bv_len;

379
	if (*sg && *cluster) {
380 381 382
		if ((*sg)->length + nbytes > queue_max_segment_size(q))
			goto new_segment;

383
		if (!BIOVEC_PHYS_MERGEABLE(bvprv, bvec))
384
			goto new_segment;
385
		if (!BIOVEC_SEG_BOUNDARY(q, bvprv, bvec))
386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403
			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.
			 */
404
			sg_unmark_end(*sg);
405 406 407 408 409 410
			*sg = sg_next(*sg);
		}

		sg_set_page(*sg, bvec->bv_page, nbytes, bvec->bv_offset);
		(*nsegs)++;
	}
411
	*bvprv = *bvec;
412 413
}

414 415 416 417 418 419 420 421
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;
}

422 423 424
static int __blk_bios_map_sg(struct request_queue *q, struct bio *bio,
			     struct scatterlist *sglist,
			     struct scatterlist **sg)
425
{
426
	struct bio_vec bvec, bvprv = { NULL };
427
	struct bvec_iter iter;
428
	int cluster = blk_queue_cluster(q), nsegs = 0;
429 430 431 432 433

	for_each_bio(bio)
		bio_for_each_segment(bvec, bio, iter)
			__blk_segment_map_sg(q, &bvec, sglist, &bvprv, sg,
					     &nsegs, &cluster);
434

435 436 437 438 439 440 441 442 443 444 445 446 447
	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;

448 449 450 451 452
	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)
453
		nsegs = __blk_bios_map_sg(q, rq->bio, sglist, &sg);
454

455
	if (unlikely(rq->rq_flags & RQF_COPY_USER) &&
456 457 458
	    (blk_rq_bytes(rq) & q->dma_pad_mask)) {
		unsigned int pad_len =
			(q->dma_pad_mask & ~blk_rq_bytes(rq)) + 1;
459 460 461 462 463

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

464
	if (q->dma_drain_size && q->dma_drain_needed(rq)) {
465
		if (op_is_write(req_op(rq)))
466 467
			memset(q->dma_drain_buffer, 0, q->dma_drain_size);

468
		sg_unmark_end(sg);
469 470 471 472 473 474
		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++;
475
		rq->extra_len += q->dma_drain_size;
476 477 478 479 480
	}

	if (sg)
		sg_mark_end(sg);

481 482 483 484
	/*
	 * Something must have been wrong if the figured number of
	 * segment is bigger than number of req's physical segments
	 */
485
	WARN_ON(nsegs > blk_rq_nr_phys_segments(rq));
486

487 488 489 490 491 492 493 494 495 496
	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);

497 498 499
	if (req->nr_phys_segments + nr_phys_segs > queue_max_segments(q))
		goto no_merge;

500
	if (blk_integrity_merge_bio(q, req, bio) == false)
501
		goto no_merge;
502 503 504 505 506 507 508

	/*
	 * This will form the start of a new hw segment.  Bump both
	 * counters.
	 */
	req->nr_phys_segments += nr_phys_segs;
	return 1;
509 510

no_merge:
R
Ritesh Harjani 已提交
511
	req_set_nomerge(q, req);
512
	return 0;
513 514 515 516 517
}

int ll_back_merge_fn(struct request_queue *q, struct request *req,
		     struct bio *bio)
{
518 519
	if (req_gap_back_merge(req, bio))
		return 0;
520 521 522
	if (blk_integrity_rq(req) &&
	    integrity_req_gap_back_merge(req, bio))
		return 0;
523
	if (blk_rq_sectors(req) + bio_sectors(bio) >
D
Damien Le Moal 已提交
524
	    blk_rq_get_max_sectors(req, blk_rq_pos(req))) {
R
Ritesh Harjani 已提交
525
		req_set_nomerge(q, req);
526 527
		return 0;
	}
528
	if (!bio_flagged(req->biotail, BIO_SEG_VALID))
529
		blk_recount_segments(q, req->biotail);
530
	if (!bio_flagged(bio, BIO_SEG_VALID))
531 532 533 534 535
		blk_recount_segments(q, bio);

	return ll_new_hw_segment(q, req, bio);
}

536
int ll_front_merge_fn(struct request_queue *q, struct request *req,
537 538
		      struct bio *bio)
{
539 540 541

	if (req_gap_front_merge(req, bio))
		return 0;
542 543 544
	if (blk_integrity_rq(req) &&
	    integrity_req_gap_front_merge(req, bio))
		return 0;
545
	if (blk_rq_sectors(req) + bio_sectors(bio) >
D
Damien Le Moal 已提交
546
	    blk_rq_get_max_sectors(req, bio->bi_iter.bi_sector)) {
R
Ritesh Harjani 已提交
547
		req_set_nomerge(q, req);
548 549
		return 0;
	}
550
	if (!bio_flagged(bio, BIO_SEG_VALID))
551
		blk_recount_segments(q, bio);
552
	if (!bio_flagged(req->bio, BIO_SEG_VALID))
553 554 555 556 557
		blk_recount_segments(q, req->bio);

	return ll_new_hw_segment(q, req, bio);
}

558 559 560 561 562 563 564 565 566 567 568
/*
 * 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;
}

569 570 571 572
static int ll_merge_requests_fn(struct request_queue *q, struct request *req,
				struct request *next)
{
	int total_phys_segments;
573 574
	unsigned int seg_size =
		req->biotail->bi_seg_back_size + next->bio->bi_seg_front_size;
575 576 577 578 579

	/*
	 * First check if the either of the requests are re-queued
	 * requests.  Can't merge them if they are.
	 */
580
	if (req_no_special_merge(req) || req_no_special_merge(next))
581 582
		return 0;

583
	if (req_gap_back_merge(req, next->bio))
584 585
		return 0;

586 587 588
	/*
	 * Will it become too large?
	 */
589
	if ((blk_rq_sectors(req) + blk_rq_sectors(next)) >
D
Damien Le Moal 已提交
590
	    blk_rq_get_max_sectors(req, blk_rq_pos(req)))
591 592 593
		return 0;

	total_phys_segments = req->nr_phys_segments + next->nr_phys_segments;
594 595 596 597 598
	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;
599
		total_phys_segments--;
600
	}
601

602
	if (total_phys_segments > queue_max_segments(q))
603 604
		return 0;

605
	if (blk_integrity_merge_rq(q, req, next) == false)
606 607
		return 0;

608 609 610 611 612
	/* Merge is OK... */
	req->nr_phys_segments = total_phys_segments;
	return 1;
}

613 614 615 616 617 618 619 620 621 622 623 624 625 626
/**
 * 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;

627
	if (rq->rq_flags & RQF_MIXED_MERGE)
628 629 630 631 632 633 634 635
		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
636 637 638
		WARN_ON_ONCE((bio->bi_opf & REQ_FAILFAST_MASK) &&
			     (bio->bi_opf & REQ_FAILFAST_MASK) != ff);
		bio->bi_opf |= ff;
639
	}
640
	rq->rq_flags |= RQF_MIXED_MERGE;
641 642
}

643 644 645 646 647 648 649
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();
650
		part = req->part;
651 652

		part_round_stats(cpu, part);
653
		part_dec_in_flight(part, rq_data_dir(req));
654

655
		hd_struct_put(part);
656 657 658 659
		part_stat_unlock();
	}
}

660
/*
661 662
 * 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.
663
 */
664 665
static struct request *attempt_merge(struct request_queue *q,
				     struct request *req, struct request *next)
666 667
{
	if (!rq_mergeable(req) || !rq_mergeable(next))
668
		return NULL;
669

670
	if (req_op(req) != req_op(next))
671
		return NULL;
672

673 674 675
	/*
	 * not contiguous
	 */
676
	if (blk_rq_pos(req) + blk_rq_sectors(req) != blk_rq_pos(next))
677
		return NULL;
678 679 680

	if (rq_data_dir(req) != rq_data_dir(next)
	    || req->rq_disk != next->rq_disk
681
	    || req_no_special_merge(next))
682
		return NULL;
683

684
	if (req_op(req) == REQ_OP_WRITE_SAME &&
685
	    !blk_write_same_mergeable(req->bio, next->bio))
686
		return NULL;
687

688 689 690 691 692 693 694
	/*
	 * 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
	 * counts here.
	 */
	if (!ll_merge_requests_fn(q, req, next))
695
		return NULL;
696

697 698 699 700 701 702
	/*
	 * 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.
	 */
703
	if (((req->rq_flags | next->rq_flags) & RQF_MIXED_MERGE) ||
704 705 706 707 708 709
	    (req->cmd_flags & REQ_FAILFAST_MASK) !=
	    (next->cmd_flags & REQ_FAILFAST_MASK)) {
		blk_rq_set_mixed_merge(req);
		blk_rq_set_mixed_merge(next);
	}

710 711 712 713 714 715 716 717 718 719 720 721
	/*
	 * At this point we have either done a back merge
	 * or front merge. We need the smaller start_time of
	 * the merged requests to be the current request
	 * for accounting purposes.
	 */
	if (time_after(req->start_time, next->start_time))
		req->start_time = next->start_time;

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

722
	req->__data_len += blk_rq_bytes(next);
723 724 725

	elv_merge_requests(q, req, next);

726 727 728 729
	/*
	 * 'next' is going away, so update stats accordingly
	 */
	blk_account_io_merge(next);
730 731

	req->ioprio = ioprio_best(req->ioprio, next->ioprio);
732 733
	if (blk_rq_cpu_valid(next))
		req->cpu = next->cpu;
734

735 736 737 738
	/*
	 * ownership of bio passed from next to req, return 'next' for
	 * the caller to free
	 */
739
	next->bio = NULL;
740
	return next;
741 742
}

743
struct request *attempt_back_merge(struct request_queue *q, struct request *rq)
744 745 746 747 748 749
{
	struct request *next = elv_latter_request(q, rq);

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

750
	return NULL;
751 752
}

753
struct request *attempt_front_merge(struct request_queue *q, struct request *rq)
754 755 756 757 758 759
{
	struct request *prev = elv_former_request(q, rq);

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

760
	return NULL;
761
}
762 763 764 765

int blk_attempt_req_merge(struct request_queue *q, struct request *rq,
			  struct request *next)
{
766
	struct elevator_queue *e = q->elevator;
767
	struct request *free;
768

769
	if (!e->uses_mq && e->type->ops.sq.elevator_allow_rq_merge_fn)
770
		if (!e->type->ops.sq.elevator_allow_rq_merge_fn(q, rq, next))
771 772
			return 0;

773 774 775 776 777 778 779
	free = attempt_merge(q, rq, next);
	if (free) {
		__blk_put_request(q, free);
		return 1;
	}

	return 0;
780
}
781 782 783

bool blk_rq_merge_ok(struct request *rq, struct bio *bio)
{
784
	if (!rq_mergeable(rq) || !bio_mergeable(bio))
785 786
		return false;

787
	if (req_op(rq) != bio_op(bio))
788 789
		return false;

790 791 792 793 794
	/* 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 */
795
	if (rq->rq_disk != bio->bi_bdev->bd_disk || req_no_special_merge(rq))
796 797 798
		return false;

	/* only merge integrity protected bio into ditto rq */
799
	if (blk_integrity_merge_bio(rq->q, rq, bio) == false)
800 801
		return false;

802
	/* must be using the same buffer */
803
	if (req_op(rq) == REQ_OP_WRITE_SAME &&
804 805 806
	    !blk_write_same_mergeable(rq->bio, bio))
		return false;

807 808 809
	return true;
}

810
enum elv_merge blk_try_merge(struct request *rq, struct bio *bio)
811
{
812 813 814 815
	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)
816
		return ELEVATOR_BACK_MERGE;
817
	else if (blk_rq_pos(rq) - bio_sectors(bio) == bio->bi_iter.bi_sector)
818 819 820
		return ELEVATOR_FRONT_MERGE;
	return ELEVATOR_NO_MERGE;
}