blk-mq-sched.c 18.1 KB
Newer Older
1
// SPDX-License-Identifier: GPL-2.0
2 3 4 5 6 7 8 9
/*
 * blk-mq scheduling framework
 *
 * Copyright (C) 2016 Jens Axboe
 */
#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/blk-mq.h>
10
#include <linux/list_sort.h>
11 12 13 14 15

#include <trace/events/block.h>

#include "blk.h"
#include "blk-mq.h"
16
#include "blk-mq-debugfs.h"
17 18 19 20
#include "blk-mq-sched.h"
#include "blk-mq-tag.h"
#include "blk-wbt.h"

D
Damien Le Moal 已提交
21
void blk_mq_sched_assign_ioc(struct request *rq)
22
{
23
	struct request_queue *q = rq->q;
24
	struct io_context *ioc;
25 26
	struct io_cq *icq;

27 28 29 30 31 32 33
	/*
	 * May not have an IO context if it's a passthrough request
	 */
	ioc = current->io_context;
	if (!ioc)
		return;

34
	spin_lock_irq(&q->queue_lock);
35
	icq = ioc_lookup_icq(ioc, q);
36
	spin_unlock_irq(&q->queue_lock);
37 38 39 40 41 42

	if (!icq) {
		icq = ioc_create_icq(ioc, q, GFP_ATOMIC);
		if (!icq)
			return;
	}
43
	get_io_context(icq->ioc);
44
	rq->elv.icq = icq;
45 46
}

47 48 49 50
/*
 * Mark a hardware queue as needing a restart. For shared queues, maintain
 * a count of how many hardware queues are marked for restart.
 */
51
void blk_mq_sched_mark_restart_hctx(struct blk_mq_hw_ctx *hctx)
52 53 54 55
{
	if (test_bit(BLK_MQ_S_SCHED_RESTART, &hctx->state))
		return;

56
	set_bit(BLK_MQ_S_SCHED_RESTART, &hctx->state);
57
}
58
EXPORT_SYMBOL_GPL(blk_mq_sched_mark_restart_hctx);
59

60
void blk_mq_sched_restart(struct blk_mq_hw_ctx *hctx)
61 62
{
	if (!test_bit(BLK_MQ_S_SCHED_RESTART, &hctx->state))
63 64
		return;
	clear_bit(BLK_MQ_S_SCHED_RESTART, &hctx->state);
65

66 67 68 69 70 71 72 73 74
	/*
	 * Order clearing SCHED_RESTART and list_empty_careful(&hctx->dispatch)
	 * in blk_mq_run_hw_queue(). Its pair is the barrier in
	 * blk_mq_dispatch_rq_list(). So dispatch code won't see SCHED_RESTART,
	 * meantime new request added to hctx->dispatch is missed to check in
	 * blk_mq_run_hw_queue().
	 */
	smp_mb();

75
	blk_mq_run_hw_queue(hctx, true);
76 77
}

78 79
static int sched_rq_cmp(void *priv, const struct list_head *a,
			const struct list_head *b)
80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104
{
	struct request *rqa = container_of(a, struct request, queuelist);
	struct request *rqb = container_of(b, struct request, queuelist);

	return rqa->mq_hctx > rqb->mq_hctx;
}

static bool blk_mq_dispatch_hctx_list(struct list_head *rq_list)
{
	struct blk_mq_hw_ctx *hctx =
		list_first_entry(rq_list, struct request, queuelist)->mq_hctx;
	struct request *rq;
	LIST_HEAD(hctx_list);
	unsigned int count = 0;

	list_for_each_entry(rq, rq_list, queuelist) {
		if (rq->mq_hctx != hctx) {
			list_cut_before(&hctx_list, rq_list, &rq->queuelist);
			goto dispatch;
		}
		count++;
	}
	list_splice_tail_init(rq_list, &hctx_list);

dispatch:
105
	return blk_mq_dispatch_rq_list(hctx, &hctx_list, count);
106 107
}

108 109
#define BLK_MQ_BUDGET_DELAY	3		/* ms units */

110 111 112 113
/*
 * Only SCSI implements .get_budget and .put_budget, and SCSI restarts
 * its queue by itself in its completion handler, so we don't need to
 * restart queue if .get_budget() returns BLK_STS_NO_RESOURCE.
114 115 116
 *
 * Returns -EAGAIN if hctx->dispatch was found non-empty and run_work has to
 * be run again.  This is necessary to avoid starving flushes.
117
 */
118
static int __blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx)
119 120 121
{
	struct request_queue *q = hctx->queue;
	struct elevator_queue *e = q->elevator;
122 123 124
	bool multi_hctxs = false, run_queue = false;
	bool dispatched = false, busy = false;
	unsigned int max_dispatch;
125
	LIST_HEAD(rq_list);
126 127 128 129 130 131
	int count = 0;

	if (hctx->dispatch_busy)
		max_dispatch = 1;
	else
		max_dispatch = hctx->queue->nr_requests;
132 133

	do {
134
		struct request *rq;
135
		int budget_token;
136

137
		if (e->type->ops.has_work && !e->type->ops.has_work(hctx))
138
			break;
139

140
		if (!list_empty_careful(&hctx->dispatch)) {
141
			busy = true;
142 143 144
			break;
		}

145 146
		budget_token = blk_mq_get_dispatch_budget(q);
		if (budget_token < 0)
147
			break;
148

149
		rq = e->type->ops.dispatch_request(hctx);
150
		if (!rq) {
151
			blk_mq_put_dispatch_budget(q, budget_token);
152 153 154 155 156 157 158
			/*
			 * We're releasing without dispatching. Holding the
			 * budget could have blocked any "hctx"s with the
			 * same queue and if we didn't dispatch then there's
			 * no guarantee anyone will kick the queue.  Kick it
			 * ourselves.
			 */
159
			run_queue = true;
160 161 162
			break;
		}

163 164
		blk_mq_set_rq_budget_token(rq, budget_token);

165 166 167 168 169
		/*
		 * Now this rq owns the budget which has to be released
		 * if this rq won't be queued to driver via .queue_rq()
		 * in blk_mq_dispatch_rq_list().
		 */
170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205
		list_add_tail(&rq->queuelist, &rq_list);
		if (rq->mq_hctx != hctx)
			multi_hctxs = true;
	} while (++count < max_dispatch);

	if (!count) {
		if (run_queue)
			blk_mq_delay_run_hw_queues(q, BLK_MQ_BUDGET_DELAY);
	} else if (multi_hctxs) {
		/*
		 * Requests from different hctx may be dequeued from some
		 * schedulers, such as bfq and deadline.
		 *
		 * Sort the requests in the list according to their hctx,
		 * dispatch batching requests from same hctx at a time.
		 */
		list_sort(NULL, &rq_list, sched_rq_cmp);
		do {
			dispatched |= blk_mq_dispatch_hctx_list(&rq_list);
		} while (!list_empty(&rq_list));
	} else {
		dispatched = blk_mq_dispatch_rq_list(hctx, &rq_list, count);
	}

	if (busy)
		return -EAGAIN;
	return !!dispatched;
}

static int blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx)
{
	int ret;

	do {
		ret = __blk_mq_do_dispatch_sched(hctx);
	} while (ret == 1);
206 207

	return ret;
208 209
}

210 211 212
static struct blk_mq_ctx *blk_mq_next_ctx(struct blk_mq_hw_ctx *hctx,
					  struct blk_mq_ctx *ctx)
{
213
	unsigned short idx = ctx->index_hw[hctx->type];
214 215 216 217 218 219 220

	if (++idx == hctx->nr_ctx)
		idx = 0;

	return hctx->ctxs[idx];
}

221 222 223 224
/*
 * Only SCSI implements .get_budget and .put_budget, and SCSI restarts
 * its queue by itself in its completion handler, so we don't need to
 * restart queue if .get_budget() returns BLK_STS_NO_RESOURCE.
225 226
 *
 * Returns -EAGAIN if hctx->dispatch was found non-empty and run_work has to
227
 * be run again.  This is necessary to avoid starving flushes.
228
 */
229
static int blk_mq_do_dispatch_ctx(struct blk_mq_hw_ctx *hctx)
230 231 232 233
{
	struct request_queue *q = hctx->queue;
	LIST_HEAD(rq_list);
	struct blk_mq_ctx *ctx = READ_ONCE(hctx->dispatch_from);
234
	int ret = 0;
235
	struct request *rq;
236 237

	do {
238 239
		int budget_token;

240 241 242 243 244
		if (!list_empty_careful(&hctx->dispatch)) {
			ret = -EAGAIN;
			break;
		}

245 246 247
		if (!sbitmap_any_bit_set(&hctx->ctx_map))
			break;

248 249
		budget_token = blk_mq_get_dispatch_budget(q);
		if (budget_token < 0)
250
			break;
251 252 253

		rq = blk_mq_dequeue_from_ctx(hctx, ctx);
		if (!rq) {
254
			blk_mq_put_dispatch_budget(q, budget_token);
255 256 257 258 259 260 261 262
			/*
			 * We're releasing without dispatching. Holding the
			 * budget could have blocked any "hctx"s with the
			 * same queue and if we didn't dispatch then there's
			 * no guarantee anyone will kick the queue.  Kick it
			 * ourselves.
			 */
			blk_mq_delay_run_hw_queues(q, BLK_MQ_BUDGET_DELAY);
263 264 265
			break;
		}

266 267
		blk_mq_set_rq_budget_token(rq, budget_token);

268 269 270 271 272 273 274 275 276 277
		/*
		 * Now this rq owns the budget which has to be released
		 * if this rq won't be queued to driver via .queue_rq()
		 * in blk_mq_dispatch_rq_list().
		 */
		list_add(&rq->queuelist, &rq_list);

		/* round robin for fair dispatch */
		ctx = blk_mq_next_ctx(hctx, rq->mq_ctx);

278
	} while (blk_mq_dispatch_rq_list(rq->mq_hctx, &rq_list, 1));
279 280

	WRITE_ONCE(hctx->dispatch_from, ctx);
281
	return ret;
282 283
}

284
static int __blk_mq_sched_dispatch_requests(struct blk_mq_hw_ctx *hctx)
285
{
286 287
	struct request_queue *q = hctx->queue;
	struct elevator_queue *e = q->elevator;
288
	const bool has_sched_dispatch = e && e->type->ops.dispatch_request;
289
	int ret = 0;
290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310
	LIST_HEAD(rq_list);

	/*
	 * If we have previous entries on our dispatch list, grab them first for
	 * more fair dispatch.
	 */
	if (!list_empty_careful(&hctx->dispatch)) {
		spin_lock(&hctx->lock);
		if (!list_empty(&hctx->dispatch))
			list_splice_init(&hctx->dispatch, &rq_list);
		spin_unlock(&hctx->lock);
	}

	/*
	 * Only ask the scheduler for requests, if we didn't have residual
	 * requests from the dispatch list. This is to avoid the case where
	 * we only ever dispatch a fraction of the requests available because
	 * of low device queue depth. Once we pull requests out of the IO
	 * scheduler, we can no longer merge or sort them. So it's best to
	 * leave them there for as long as we can. Mark the hw queue as
	 * needing a restart in that case.
311 312 313 314
	 *
	 * We want to dispatch from the scheduler if there was nothing
	 * on the dispatch list or we were able to dispatch from the
	 * dispatch list.
315
	 */
316
	if (!list_empty(&rq_list)) {
317
		blk_mq_sched_mark_restart_hctx(hctx);
318
		if (blk_mq_dispatch_rq_list(hctx, &rq_list, 0)) {
319
			if (has_sched_dispatch)
320
				ret = blk_mq_do_dispatch_sched(hctx);
321
			else
322
				ret = blk_mq_do_dispatch_ctx(hctx);
323
		}
324
	} else if (has_sched_dispatch) {
325
		ret = blk_mq_do_dispatch_sched(hctx);
326 327
	} else if (hctx->dispatch_busy) {
		/* dequeue request one by one from sw queue if queue is busy */
328
		ret = blk_mq_do_dispatch_ctx(hctx);
329
	} else {
330
		blk_mq_flush_busy_ctxs(hctx, &rq_list);
331
		blk_mq_dispatch_rq_list(hctx, &rq_list, 0);
332
	}
333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354

	return ret;
}

void blk_mq_sched_dispatch_requests(struct blk_mq_hw_ctx *hctx)
{
	struct request_queue *q = hctx->queue;

	/* RCU or SRCU read lock is needed before checking quiesced flag */
	if (unlikely(blk_mq_hctx_stopped(hctx) || blk_queue_quiesced(q)))
		return;

	hctx->run++;

	/*
	 * A return of -EAGAIN is an indication that hctx->dispatch is not
	 * empty and we must run again in order to avoid starving flushes.
	 */
	if (__blk_mq_sched_dispatch_requests(hctx) == -EAGAIN) {
		if (__blk_mq_sched_dispatch_requests(hctx) == -EAGAIN)
			blk_mq_run_hw_queue(hctx, true);
	}
355 356
}

357 358
bool __blk_mq_sched_bio_merge(struct request_queue *q, struct bio *bio,
		unsigned int nr_segs)
359 360
{
	struct elevator_queue *e = q->elevator;
361 362
	struct blk_mq_ctx *ctx;
	struct blk_mq_hw_ctx *hctx;
363
	bool ret = false;
M
Ming Lei 已提交
364
	enum hctx_type type;
365

366
	if (e && e->type->ops.bio_merge)
367
		return e->type->ops.bio_merge(q, bio, nr_segs);
368

369 370
	ctx = blk_mq_get_ctx(q);
	hctx = blk_mq_map_queue(q, bio->bi_opf, ctx);
M
Ming Lei 已提交
371
	type = hctx->type;
372 373 374 375 376 377 378 379 380 381 382 383 384 385
	if (!(hctx->flags & BLK_MQ_F_SHOULD_MERGE) ||
	    list_empty_careful(&ctx->rq_lists[type]))
		return false;

	/* default per sw-queue merge */
	spin_lock(&ctx->lock);
	/*
	 * Reverse check our software queue for entries that we could
	 * potentially merge with. Currently includes a hand-wavy stop
	 * count of 8, to not spend too much time checking for merges.
	 */
	if (blk_bio_list_merge(q, &ctx->rq_lists[type], bio, nr_segs)) {
		ctx->rq_merged++;
		ret = true;
386 387
	}

388 389
	spin_unlock(&ctx->lock);

390
	return ret;
391 392 393 394 395 396 397 398
}

bool blk_mq_sched_try_insert_merge(struct request_queue *q, struct request *rq)
{
	return rq_mergeable(rq) && elv_attempt_insert_merge(q, rq);
}
EXPORT_SYMBOL_GPL(blk_mq_sched_try_insert_merge);

399 400
static bool blk_mq_sched_bypass_insert(struct blk_mq_hw_ctx *hctx,
				       struct request *rq)
401
{
402 403 404 405 406 407 408 409 410 411 412 413
	/*
	 * dispatch flush and passthrough rq directly
	 *
	 * passthrough request has to be added to hctx->dispatch directly.
	 * For some reason, device may be in one situation which can't
	 * handle FS request, so STS_RESOURCE is always returned and the
	 * FS request will be added to hctx->dispatch. However passthrough
	 * request may be required at that time for fixing the problem. If
	 * passthrough request is added to scheduler queue, there isn't any
	 * chance to dispatch it given we prioritize requests in hctx->dispatch.
	 */
	if ((rq->rq_flags & RQF_FLUSH_SEQ) || blk_rq_is_passthrough(rq))
414 415 416
		return true;

	return false;
417 418
}

419
void blk_mq_sched_insert_request(struct request *rq, bool at_head,
420
				 bool run_queue, bool async)
421 422 423 424
{
	struct request_queue *q = rq->q;
	struct elevator_queue *e = q->elevator;
	struct blk_mq_ctx *ctx = rq->mq_ctx;
425
	struct blk_mq_hw_ctx *hctx = rq->mq_hctx;
426

427
	WARN_ON(e && (rq->tag != BLK_MQ_NO_TAG));
428

429
	if (blk_mq_sched_bypass_insert(hctx, rq)) {
430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451
		/*
		 * Firstly normal IO request is inserted to scheduler queue or
		 * sw queue, meantime we add flush request to dispatch queue(
		 * hctx->dispatch) directly and there is at most one in-flight
		 * flush request for each hw queue, so it doesn't matter to add
		 * flush request to tail or front of the dispatch queue.
		 *
		 * Secondly in case of NCQ, flush request belongs to non-NCQ
		 * command, and queueing it will fail when there is any
		 * in-flight normal IO request(NCQ command). When adding flush
		 * rq to the front of hctx->dispatch, it is easier to introduce
		 * extra time to flush rq's latency because of S_SCHED_RESTART
		 * compared with adding to the tail of dispatch queue, then
		 * chance of flush merge is increased, and less flush requests
		 * will be issued to controller. It is observed that ~10% time
		 * is saved in blktests block/004 on disk attached to AHCI/NCQ
		 * drive when adding flush rq to the front of hctx->dispatch.
		 *
		 * Simply queue flush rq to the front of hctx->dispatch so that
		 * intensive flush workloads can benefit in case of NCQ HW.
		 */
		at_head = (rq->rq_flags & RQF_FLUSH_SEQ) ? true : at_head;
452
		blk_mq_request_bypass_insert(rq, at_head, false);
453
		goto run;
454
	}
455

456
	if (e && e->type->ops.insert_requests) {
457 458 459
		LIST_HEAD(list);

		list_add(&rq->queuelist, &list);
460
		e->type->ops.insert_requests(hctx, &list, at_head);
461 462 463 464 465 466
	} else {
		spin_lock(&ctx->lock);
		__blk_mq_insert_request(hctx, rq, at_head);
		spin_unlock(&ctx->lock);
	}

467
run:
468 469 470 471
	if (run_queue)
		blk_mq_run_hw_queue(hctx, async);
}

472
void blk_mq_sched_insert_requests(struct blk_mq_hw_ctx *hctx,
473 474 475
				  struct blk_mq_ctx *ctx,
				  struct list_head *list, bool run_queue_async)
{
476
	struct elevator_queue *e;
477 478 479 480 481 482 483 484
	struct request_queue *q = hctx->queue;

	/*
	 * blk_mq_sched_insert_requests() is called from flush plug
	 * context only, and hold one usage counter to prevent queue
	 * from being released.
	 */
	percpu_ref_get(&q->q_usage_counter);
485

486
	e = hctx->queue->elevator;
487 488
	if (e && e->type->ops.insert_requests)
		e->type->ops.insert_requests(hctx, list, false);
489 490 491 492 493 494
	else {
		/*
		 * try to issue requests directly if the hw queue isn't
		 * busy in case of 'none' scheduler, and this way may save
		 * us one extra enqueue & dequeue to sw queue.
		 */
495
		if (!hctx->dispatch_busy && !e && !run_queue_async) {
496
			blk_mq_try_issue_list_directly(hctx, list);
497
			if (list_empty(list))
498
				goto out;
499 500
		}
		blk_mq_insert_requests(hctx, ctx, list);
501
	}
502 503

	blk_mq_run_hw_queue(hctx, run_queue_async);
504 505
 out:
	percpu_ref_put(&q->q_usage_counter);
506 507
}

508 509 510 511 512 513
static void blk_mq_sched_free_tags(struct blk_mq_tag_set *set,
				   struct blk_mq_hw_ctx *hctx,
				   unsigned int hctx_idx)
{
	if (hctx->sched_tags) {
		blk_mq_free_rqs(set, hctx->sched_tags, hctx_idx);
514
		blk_mq_free_rq_map(hctx->sched_tags, set->flags);
515 516 517 518
		hctx->sched_tags = NULL;
	}
}

519 520 521 522 523 524 525 526
static int blk_mq_sched_alloc_tags(struct request_queue *q,
				   struct blk_mq_hw_ctx *hctx,
				   unsigned int hctx_idx)
{
	struct blk_mq_tag_set *set = q->tag_set;
	int ret;

	hctx->sched_tags = blk_mq_alloc_rq_map(set, hctx_idx, q->nr_requests,
527
					       set->reserved_tags, set->flags);
528 529 530 531 532 533 534 535 536 537
	if (!hctx->sched_tags)
		return -ENOMEM;

	ret = blk_mq_alloc_rqs(set, hctx->sched_tags, hctx_idx, q->nr_requests);
	if (ret)
		blk_mq_sched_free_tags(set, hctx, hctx_idx);

	return ret;
}

538
/* called in queue's release handler, tagset has gone away */
539
static void blk_mq_sched_tags_teardown(struct request_queue *q)
540 541
{
	struct blk_mq_hw_ctx *hctx;
542 543
	int i;

544 545
	queue_for_each_hw_ctx(q, hctx, i) {
		if (hctx->sched_tags) {
546
			blk_mq_free_rq_map(hctx->sched_tags, hctx->flags);
547 548 549
			hctx->sched_tags = NULL;
		}
	}
550 551
}

552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588
static int blk_mq_init_sched_shared_sbitmap(struct request_queue *queue)
{
	struct blk_mq_tag_set *set = queue->tag_set;
	int alloc_policy = BLK_MQ_FLAG_TO_ALLOC_POLICY(set->flags);
	struct blk_mq_hw_ctx *hctx;
	int ret, i;

	/*
	 * Set initial depth at max so that we don't need to reallocate for
	 * updating nr_requests.
	 */
	ret = blk_mq_init_bitmaps(&queue->sched_bitmap_tags,
				  &queue->sched_breserved_tags,
				  MAX_SCHED_RQ, set->reserved_tags,
				  set->numa_node, alloc_policy);
	if (ret)
		return ret;

	queue_for_each_hw_ctx(queue, hctx, i) {
		hctx->sched_tags->bitmap_tags =
					&queue->sched_bitmap_tags;
		hctx->sched_tags->breserved_tags =
					&queue->sched_breserved_tags;
	}

	sbitmap_queue_resize(&queue->sched_bitmap_tags,
			     queue->nr_requests - set->reserved_tags);

	return 0;
}

static void blk_mq_exit_sched_shared_sbitmap(struct request_queue *queue)
{
	sbitmap_queue_free(&queue->sched_bitmap_tags);
	sbitmap_queue_free(&queue->sched_breserved_tags);
}

589 590 591
int blk_mq_init_sched(struct request_queue *q, struct elevator_type *e)
{
	struct blk_mq_hw_ctx *hctx;
592
	struct elevator_queue *eq;
593 594 595 596 597
	unsigned int i;
	int ret;

	if (!e) {
		q->elevator = NULL;
598
		q->nr_requests = q->tag_set->queue_depth;
599 600
		return 0;
	}
601 602

	/*
603 604 605
	 * Default to double of smaller one between hw queue_depth and 128,
	 * since we don't split into sync/async like the old code did.
	 * Additionally, this is a per-hw queue depth.
606
	 */
607 608
	q->nr_requests = 2 * min_t(unsigned int, q->tag_set->queue_depth,
				   BLKDEV_MAX_RQ);
609 610

	queue_for_each_hw_ctx(q, hctx, i) {
611
		ret = blk_mq_sched_alloc_tags(q, hctx, i);
612
		if (ret)
613 614 615 616 617 618 619
			goto err_free_tags;
	}

	if (blk_mq_is_sbitmap_shared(q->tag_set->flags)) {
		ret = blk_mq_init_sched_shared_sbitmap(q);
		if (ret)
			goto err_free_tags;
620 621
	}

622
	ret = e->ops.init_sched(q, e);
623
	if (ret)
624
		goto err_free_sbitmap;
625

626 627 628
	blk_mq_debugfs_register_sched(q);

	queue_for_each_hw_ctx(q, hctx, i) {
629 630
		if (e->ops.init_hctx) {
			ret = e->ops.init_hctx(hctx, i);
631 632
			if (ret) {
				eq = q->elevator;
633
				blk_mq_sched_free_requests(q);
634 635 636 637 638
				blk_mq_exit_sched(q, eq);
				kobject_put(&eq->kobj);
				return ret;
			}
		}
639
		blk_mq_debugfs_register_sched_hctx(q, hctx);
640 641
	}

642 643
	return 0;

644 645 646 647
err_free_sbitmap:
	if (blk_mq_is_sbitmap_shared(q->tag_set->flags))
		blk_mq_exit_sched_shared_sbitmap(q);
err_free_tags:
648
	blk_mq_sched_free_requests(q);
649 650
	blk_mq_sched_tags_teardown(q);
	q->elevator = NULL;
651
	return ret;
652
}
653

654 655 656 657 658 659 660 661 662 663 664 665 666 667 668
/*
 * called in either blk_queue_cleanup or elevator_switch, tagset
 * is required for freeing requests
 */
void blk_mq_sched_free_requests(struct request_queue *q)
{
	struct blk_mq_hw_ctx *hctx;
	int i;

	queue_for_each_hw_ctx(q, hctx, i) {
		if (hctx->sched_tags)
			blk_mq_free_rqs(q->tag_set, hctx->sched_tags, i);
	}
}

669 670
void blk_mq_exit_sched(struct request_queue *q, struct elevator_queue *e)
{
671 672 673
	struct blk_mq_hw_ctx *hctx;
	unsigned int i;

674 675
	queue_for_each_hw_ctx(q, hctx, i) {
		blk_mq_debugfs_unregister_sched_hctx(hctx);
676 677
		if (e->type->ops.exit_hctx && hctx->sched_data) {
			e->type->ops.exit_hctx(hctx, i);
678
			hctx->sched_data = NULL;
679 680
		}
	}
681
	blk_mq_debugfs_unregister_sched(q);
682 683
	if (e->type->ops.exit_sched)
		e->type->ops.exit_sched(e);
684
	blk_mq_sched_tags_teardown(q);
685 686
	if (blk_mq_is_sbitmap_shared(q->tag_set->flags))
		blk_mq_exit_sched_shared_sbitmap(q);
687 688
	q->elevator = NULL;
}