blk-merge.c 14.6 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11
/*
 * 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>

#include "blk.h"

12
static unsigned int __blk_recalc_rq_segments(struct request_queue *q,
13
					     struct bio *bio)
14
{
15
	struct bio_vec bv, bvprv = { NULL };
16
	int cluster, high, highprv = 1, no_sg_merge;
17
	unsigned int seg_size, nr_phys_segs;
18
	struct bio *fbio, *bbio;
19
	struct bvec_iter iter;
20

21 22
	if (!bio)
		return 0;
23

24 25 26 27 28 29 30 31 32 33
	/*
	 * This should probably be returning 0, but blk_add_request_payload()
	 * (Christoph!!!!)
	 */
	if (bio->bi_rw & REQ_DISCARD)
		return 1;

	if (bio->bi_rw & REQ_WRITE_SAME)
		return 1;

34
	fbio = bio;
35
	cluster = blk_queue_cluster(q);
Mikulas Patocka's avatar
Mikulas Patocka committed
36
	seg_size = 0;
37
	nr_phys_segs = 0;
38 39
	no_sg_merge = test_bit(QUEUE_FLAG_NO_SG_MERGE, &q->queue_flags);
	high = 0;
40
	for_each_bio(bio) {
41
		bio_for_each_segment(bv, bio, iter) {
42 43 44 45 46 47 48
			/*
			 * If SG merging is disabled, each bio vector is
			 * a segment
			 */
			if (no_sg_merge)
				goto new_segment;

49 50
			/*
			 * the trick here is making sure that a high page is
51 52
			 * never considered part of another segment, since
			 * that might change with the bounce page.
53
			 */
54 55 56
			high = page_to_pfn(bv.bv_page) > queue_bounce_pfn(q);
			if (!high && !highprv && cluster) {
				if (seg_size + bv.bv_len
57
				    > queue_max_segment_size(q))
58
					goto new_segment;
59
				if (!BIOVEC_PHYS_MERGEABLE(&bvprv, &bv))
60
					goto new_segment;
61
				if (!BIOVEC_SEG_BOUNDARY(q, &bvprv, &bv))
62
					goto new_segment;
63

64
				seg_size += bv.bv_len;
65 66 67
				bvprv = bv;
				continue;
			}
68
new_segment:
69 70 71
			if (nr_phys_segs == 1 && seg_size >
			    fbio->bi_seg_front_size)
				fbio->bi_seg_front_size = seg_size;
72

73 74
			nr_phys_segs++;
			bvprv = bv;
75
			seg_size = bv.bv_len;
76 77
			highprv = high;
		}
78
		bbio = bio;
79 80
	}

81 82 83 84
	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;
85 86 87 88 89 90

	return nr_phys_segs;
}

void blk_recalc_rq_segments(struct request *rq)
{
91
	rq->nr_phys_segments = __blk_recalc_rq_segments(rq->q, rq->bio);
92 93 94 95
}

void blk_recount_segments(struct request_queue *q, struct bio *bio)
{
96 97 98 99 100 101 102 103 104
	if (test_bit(QUEUE_FLAG_NO_SG_MERGE, &q->queue_flags))
		bio->bi_phys_segments = bio->bi_vcnt;
	else {
		struct bio *nxt = bio->bi_next;

		bio->bi_next = NULL;
		bio->bi_phys_segments = __blk_recalc_rq_segments(q, bio);
		bio->bi_next = nxt;
	}
105

106 107 108 109 110 111 112
	bio->bi_flags |= (1 << BIO_SEG_VALID);
}
EXPORT_SYMBOL(blk_recount_segments);

static int blk_phys_contig_segment(struct request_queue *q, struct bio *bio,
				   struct bio *nxt)
{
113
	struct bio_vec end_bv = { NULL }, nxt_bv;
114 115
	struct bvec_iter iter;

116
	if (!blk_queue_cluster(q))
117 118
		return 0;

119
	if (bio->bi_seg_back_size + nxt->bi_seg_front_size >
120
	    queue_max_segment_size(q))
121 122
		return 0;

123 124 125
	if (!bio_has_data(bio))
		return 1;

126 127 128 129 130 131 132
	bio_for_each_segment(end_bv, bio, iter)
		if (end_bv.bv_len == iter.bi_size)
			break;

	nxt_bv = bio_iovec(nxt);

	if (!BIOVEC_PHYS_MERGEABLE(&end_bv, &nxt_bv))
133 134
		return 0;

135
	/*
136
	 * bio and nxt are contiguous in memory; check if the queue allows
137 138
	 * these two to be merged into one
	 */
139
	if (BIOVEC_SEG_BOUNDARY(q, &end_bv, &nxt_bv))
140 141 142 143 144
		return 1;

	return 0;
}

145
static inline void
146
__blk_segment_map_sg(struct request_queue *q, struct bio_vec *bvec,
147
		     struct scatterlist *sglist, struct bio_vec *bvprv,
148 149 150 151 152
		     struct scatterlist **sg, int *nsegs, int *cluster)
{

	int nbytes = bvec->bv_len;

153
	if (*sg && *cluster) {
154 155 156
		if ((*sg)->length + nbytes > queue_max_segment_size(q))
			goto new_segment;

157
		if (!BIOVEC_PHYS_MERGEABLE(bvprv, bvec))
158
			goto new_segment;
159
		if (!BIOVEC_SEG_BOUNDARY(q, bvprv, bvec))
160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177
			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.
			 */
178
			sg_unmark_end(*sg);
179 180 181 182 183 184
			*sg = sg_next(*sg);
		}

		sg_set_page(*sg, bvec->bv_page, nbytes, bvec->bv_offset);
		(*nsegs)++;
	}
185
	*bvprv = *bvec;
186 187
}

188 189 190
static int __blk_bios_map_sg(struct request_queue *q, struct bio *bio,
			     struct scatterlist *sglist,
			     struct scatterlist **sg)
191
{
192
	struct bio_vec bvec, bvprv = { NULL };
193
	struct bvec_iter iter;
194 195 196
	int nsegs, cluster;

	nsegs = 0;
197
	cluster = blk_queue_cluster(q);
198

199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225
	if (bio->bi_rw & REQ_DISCARD) {
		/*
		 * This is a hack - drivers should be neither modifying the
		 * biovec, nor relying on bi_vcnt - but because of
		 * blk_add_request_payload(), a discard bio may or may not have
		 * a payload we need to set up here (thank you Christoph) and
		 * bi_vcnt is really the only way of telling if we need to.
		 */

		if (bio->bi_vcnt)
			goto single_segment;

		return 0;
	}

	if (bio->bi_rw & REQ_WRITE_SAME) {
single_segment:
		*sg = sglist;
		bvec = bio_iovec(bio);
		sg_set_page(*sg, bvec.bv_page, bvec.bv_len, bvec.bv_offset);
		return 1;
	}

	for_each_bio(bio)
		bio_for_each_segment(bvec, bio, iter)
			__blk_segment_map_sg(q, &bvec, sglist, &bvprv, sg,
					     &nsegs, &cluster);
226

227 228 229 230 231 232 233 234 235 236 237 238 239 240 241
	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;

	if (rq->bio)
		nsegs = __blk_bios_map_sg(q, rq->bio, sglist, &sg);
242 243

	if (unlikely(rq->cmd_flags & REQ_COPY_USER) &&
244 245 246
	    (blk_rq_bytes(rq) & q->dma_pad_mask)) {
		unsigned int pad_len =
			(q->dma_pad_mask & ~blk_rq_bytes(rq)) + 1;
247 248 249 250 251

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

252
	if (q->dma_drain_size && q->dma_drain_needed(rq)) {
253
		if (rq->cmd_flags & REQ_WRITE)
254 255
			memset(q->dma_drain_buffer, 0, q->dma_drain_size);

256 257 258 259 260 261 262
		sg->page_link &= ~0x02;
		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++;
263
		rq->extra_len += q->dma_drain_size;
264 265 266 267 268 269 270 271 272
	}

	if (sg)
		sg_mark_end(sg);

	return nsegs;
}
EXPORT_SYMBOL(blk_rq_map_sg);

273 274 275 276 277 278 279 280 281 282 283 284 285 286
/**
 * blk_bio_map_sg - map a bio to a scatterlist
 * @q: request_queue in question
 * @bio: bio being mapped
 * @sglist: scatterlist being mapped
 *
 * Note:
 *    Caller must make sure sg can hold bio->bi_phys_segments entries
 *
 * Will return the number of sg entries setup
 */
int blk_bio_map_sg(struct request_queue *q, struct bio *bio,
		   struct scatterlist *sglist)
{
287 288 289 290
	struct scatterlist *sg = NULL;
	int nsegs;
	struct bio *next = bio->bi_next;
	bio->bi_next = NULL;
291

292 293
	nsegs = __blk_bios_map_sg(q, bio, sglist, &sg);
	bio->bi_next = next;
294 295 296 297 298 299 300 301
	if (sg)
		sg_mark_end(sg);

	BUG_ON(bio->bi_phys_segments && nsegs > bio->bi_phys_segments);
	return nsegs;
}
EXPORT_SYMBOL(blk_bio_map_sg);

302 303 304 305 306 307
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);

308 309 310 311 312
	if (req->nr_phys_segments + nr_phys_segs > queue_max_segments(q))
		goto no_merge;

	if (bio_integrity(bio) && blk_integrity_merge_bio(q, req, bio))
		goto no_merge;
313 314 315 316 317 318 319

	/*
	 * This will form the start of a new hw segment.  Bump both
	 * counters.
	 */
	req->nr_phys_segments += nr_phys_segs;
	return 1;
320 321 322 323 324 325

no_merge:
	req->cmd_flags |= REQ_NOMERGE;
	if (req == q->last_merge)
		q->last_merge = NULL;
	return 0;
326 327 328 329 330
}

int ll_back_merge_fn(struct request_queue *q, struct request *req,
		     struct bio *bio)
{
331 332
	if (blk_rq_sectors(req) + bio_sectors(bio) >
	    blk_rq_get_max_sectors(req)) {
333 334 335 336 337
		req->cmd_flags |= REQ_NOMERGE;
		if (req == q->last_merge)
			q->last_merge = NULL;
		return 0;
	}
338
	if (!bio_flagged(req->biotail, BIO_SEG_VALID))
339
		blk_recount_segments(q, req->biotail);
340
	if (!bio_flagged(bio, BIO_SEG_VALID))
341 342 343 344 345
		blk_recount_segments(q, bio);

	return ll_new_hw_segment(q, req, bio);
}

346
int ll_front_merge_fn(struct request_queue *q, struct request *req,
347 348
		      struct bio *bio)
{
349 350
	if (blk_rq_sectors(req) + bio_sectors(bio) >
	    blk_rq_get_max_sectors(req)) {
351 352 353 354 355
		req->cmd_flags |= REQ_NOMERGE;
		if (req == q->last_merge)
			q->last_merge = NULL;
		return 0;
	}
356
	if (!bio_flagged(bio, BIO_SEG_VALID))
357
		blk_recount_segments(q, bio);
358
	if (!bio_flagged(req->bio, BIO_SEG_VALID))
359 360 361 362 363
		blk_recount_segments(q, req->bio);

	return ll_new_hw_segment(q, req, bio);
}

364 365 366 367 368 369 370 371 372 373 374
/*
 * 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;
}

375 376 377 378
static int ll_merge_requests_fn(struct request_queue *q, struct request *req,
				struct request *next)
{
	int total_phys_segments;
379 380
	unsigned int seg_size =
		req->biotail->bi_seg_back_size + next->bio->bi_seg_front_size;
381 382 383 384 385

	/*
	 * First check if the either of the requests are re-queued
	 * requests.  Can't merge them if they are.
	 */
386
	if (req_no_special_merge(req) || req_no_special_merge(next))
387 388 389 390 391
		return 0;

	/*
	 * Will it become too large?
	 */
392 393
	if ((blk_rq_sectors(req) + blk_rq_sectors(next)) >
	    blk_rq_get_max_sectors(req))
394 395 396
		return 0;

	total_phys_segments = req->nr_phys_segments + next->nr_phys_segments;
397 398 399 400 401
	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;
402
		total_phys_segments--;
403
	}
404

405
	if (total_phys_segments > queue_max_segments(q))
406 407
		return 0;

408 409 410
	if (blk_integrity_rq(req) && blk_integrity_merge_rq(q, req, next))
		return 0;

411 412 413 414 415
	/* Merge is OK... */
	req->nr_phys_segments = total_phys_segments;
	return 1;
}

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
/**
 * 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;

	if (rq->cmd_flags & REQ_MIXED_MERGE)
		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) {
		WARN_ON_ONCE((bio->bi_rw & REQ_FAILFAST_MASK) &&
			     (bio->bi_rw & REQ_FAILFAST_MASK) != ff);
		bio->bi_rw |= ff;
	}
	rq->cmd_flags |= REQ_MIXED_MERGE;
}

446 447 448 449 450 451 452
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();
453
		part = req->part;
454 455

		part_round_stats(cpu, part);
456
		part_dec_in_flight(part, rq_data_dir(req));
457

458
		hd_struct_put(part);
459 460 461 462
		part_stat_unlock();
	}
}

463 464 465 466 467 468 469 470 471
/*
 * Has to be called with the request spinlock acquired
 */
static int attempt_merge(struct request_queue *q, struct request *req,
			  struct request *next)
{
	if (!rq_mergeable(req) || !rq_mergeable(next))
		return 0;

472 473 474
	if (!blk_check_merge_flags(req->cmd_flags, next->cmd_flags))
		return 0;

475 476 477
	/*
	 * not contiguous
	 */
478
	if (blk_rq_pos(req) + blk_rq_sectors(req) != blk_rq_pos(next))
479 480 481 482
		return 0;

	if (rq_data_dir(req) != rq_data_dir(next)
	    || req->rq_disk != next->rq_disk
483
	    || req_no_special_merge(next))
484 485
		return 0;

486 487 488 489
	if (req->cmd_flags & REQ_WRITE_SAME &&
	    !blk_write_same_mergeable(req->bio, next->bio))
		return 0;

490 491 492 493 494 495 496 497 498
	/*
	 * 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))
		return 0;

499 500 501 502 503 504 505 506 507 508 509 510 511
	/*
	 * 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.
	 */
	if ((req->cmd_flags | next->cmd_flags) & REQ_MIXED_MERGE ||
	    (req->cmd_flags & REQ_FAILFAST_MASK) !=
	    (next->cmd_flags & REQ_FAILFAST_MASK)) {
		blk_rq_set_mixed_merge(req);
		blk_rq_set_mixed_merge(next);
	}

512 513 514 515 516 517 518 519 520 521 522 523
	/*
	 * 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;

524
	req->__data_len += blk_rq_bytes(next);
525 526 527

	elv_merge_requests(q, req, next);

528 529 530 531
	/*
	 * 'next' is going away, so update stats accordingly
	 */
	blk_account_io_merge(next);
532 533

	req->ioprio = ioprio_best(req->ioprio, next->ioprio);
534 535
	if (blk_rq_cpu_valid(next))
		req->cpu = next->cpu;
536

537 538
	/* owner-ship of bio passed from next to req */
	next->bio = NULL;
539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561
	__blk_put_request(q, next);
	return 1;
}

int attempt_back_merge(struct request_queue *q, struct request *rq)
{
	struct request *next = elv_latter_request(q, rq);

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

	return 0;
}

int attempt_front_merge(struct request_queue *q, struct request *rq)
{
	struct request *prev = elv_former_request(q, rq);

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

	return 0;
}
562 563 564 565 566 567

int blk_attempt_req_merge(struct request_queue *q, struct request *rq,
			  struct request *next)
{
	return attempt_merge(q, rq, next);
}
568 569 570

bool blk_rq_merge_ok(struct request *rq, struct bio *bio)
{
571 572
	struct request_queue *q = rq->q;

573
	if (!rq_mergeable(rq) || !bio_mergeable(bio))
574 575
		return false;

576 577 578
	if (!blk_check_merge_flags(rq->cmd_flags, bio->bi_rw))
		return false;

579 580 581 582 583
	/* 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 */
584
	if (rq->rq_disk != bio->bi_bdev->bd_disk || req_no_special_merge(rq))
585 586 587 588 589 590
		return false;

	/* only merge integrity protected bio into ditto rq */
	if (bio_integrity(bio) != blk_integrity_rq(rq))
		return false;

591 592 593 594 595
	/* must be using the same buffer */
	if (rq->cmd_flags & REQ_WRITE_SAME &&
	    !blk_write_same_mergeable(rq->bio, bio))
		return false;

596 597 598 599 600 601 602 603
	if (q->queue_flags & (1 << QUEUE_FLAG_SG_GAPS)) {
		struct bio_vec *bprev;

		bprev = &rq->biotail->bi_io_vec[bio->bi_vcnt - 1];
		if (bvec_gap_to_prev(bprev, bio->bi_io_vec[0].bv_offset))
			return false;
	}

604 605 606 607 608
	return true;
}

int blk_try_merge(struct request *rq, struct bio *bio)
{
609
	if (blk_rq_pos(rq) + blk_rq_sectors(rq) == bio->bi_iter.bi_sector)
610
		return ELEVATOR_BACK_MERGE;
611
	else if (blk_rq_pos(rq) - bio_sectors(bio) == bio->bi_iter.bi_sector)
612 613 614
		return ELEVATOR_FRONT_MERGE;
	return ELEVATOR_NO_MERGE;
}