blk-mq-sched.c 15.9 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13
/*
 * blk-mq scheduling framework
 *
 * Copyright (C) 2016 Jens Axboe
 */
#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/blk-mq.h>

#include <trace/events/block.h>

#include "blk.h"
#include "blk-mq.h"
14
#include "blk-mq-debugfs.h"
15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33
#include "blk-mq-sched.h"
#include "blk-mq-tag.h"
#include "blk-wbt.h"

void blk_mq_sched_free_hctx_data(struct request_queue *q,
				 void (*exit)(struct blk_mq_hw_ctx *))
{
	struct blk_mq_hw_ctx *hctx;
	int i;

	queue_for_each_hw_ctx(q, hctx, i) {
		if (exit && hctx->sched_data)
			exit(hctx);
		kfree(hctx->sched_data);
		hctx->sched_data = NULL;
	}
}
EXPORT_SYMBOL_GPL(blk_mq_sched_free_hctx_data);

34
void blk_mq_sched_assign_ioc(struct request *rq, struct bio *bio)
35
{
36 37
	struct request_queue *q = rq->q;
	struct io_context *ioc = rq_ioc(bio);
38 39 40 41 42 43 44 45 46 47 48
	struct io_cq *icq;

	spin_lock_irq(q->queue_lock);
	icq = ioc_lookup_icq(ioc, q);
	spin_unlock_irq(q->queue_lock);

	if (!icq) {
		icq = ioc_create_icq(ioc, q, GFP_ATOMIC);
		if (!icq)
			return;
	}
49
	get_io_context(icq->ioc);
50
	rq->elv.icq = icq;
51 52
}

53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70
/*
 * Mark a hardware queue as needing a restart. For shared queues, maintain
 * a count of how many hardware queues are marked for restart.
 */
static void blk_mq_sched_mark_restart_hctx(struct blk_mq_hw_ctx *hctx)
{
	if (test_bit(BLK_MQ_S_SCHED_RESTART, &hctx->state))
		return;

	if (hctx->flags & BLK_MQ_F_TAG_SHARED) {
		struct request_queue *q = hctx->queue;

		if (!test_and_set_bit(BLK_MQ_S_SCHED_RESTART, &hctx->state))
			atomic_inc(&q->shared_hctx_restart);
	} else
		set_bit(BLK_MQ_S_SCHED_RESTART, &hctx->state);
}

71
static bool blk_mq_sched_restart_hctx(struct blk_mq_hw_ctx *hctx)
72 73
{
	if (!test_bit(BLK_MQ_S_SCHED_RESTART, &hctx->state))
74
		return false;
75

76 77 78 79 80 81 82
	if (hctx->flags & BLK_MQ_F_TAG_SHARED) {
		struct request_queue *q = hctx->queue;

		if (test_and_clear_bit(BLK_MQ_S_SCHED_RESTART, &hctx->state))
			atomic_dec(&q->shared_hctx_restart);
	} else
		clear_bit(BLK_MQ_S_SCHED_RESTART, &hctx->state);
83

84
	return blk_mq_run_hw_queue(hctx, true);
85 86
}

87 88 89 90 91 92
/*
 * 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.
 */
static void blk_mq_do_dispatch_sched(struct blk_mq_hw_ctx *hctx)
93 94 95 96 97 98
{
	struct request_queue *q = hctx->queue;
	struct elevator_queue *e = q->elevator;
	LIST_HEAD(rq_list);

	do {
99
		struct request *rq;
100

101 102
		if (e->type->ops.mq.has_work &&
				!e->type->ops.mq.has_work(hctx))
103
			break;
104

105
		if (!blk_mq_get_dispatch_budget(hctx))
106
			break;
107 108 109 110 111 112 113 114 115 116 117 118

		rq = e->type->ops.mq.dispatch_request(hctx);
		if (!rq) {
			blk_mq_put_dispatch_budget(hctx);
			break;
		}

		/*
		 * 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().
		 */
119
		list_add(&rq->queuelist, &rq_list);
120
	} while (blk_mq_dispatch_rq_list(q, &rq_list, true));
121 122
}

123 124 125 126 127 128 129 130 131 132 133
static struct blk_mq_ctx *blk_mq_next_ctx(struct blk_mq_hw_ctx *hctx,
					  struct blk_mq_ctx *ctx)
{
	unsigned idx = ctx->index_hw;

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

	return hctx->ctxs[idx];
}

134 135 136 137 138 139
/*
 * 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.
 */
static void blk_mq_do_dispatch_ctx(struct blk_mq_hw_ctx *hctx)
140 141 142 143 144 145 146 147 148 149 150
{
	struct request_queue *q = hctx->queue;
	LIST_HEAD(rq_list);
	struct blk_mq_ctx *ctx = READ_ONCE(hctx->dispatch_from);

	do {
		struct request *rq;

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

151
		if (!blk_mq_get_dispatch_budget(hctx))
152
			break;
153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174

		rq = blk_mq_dequeue_from_ctx(hctx, ctx);
		if (!rq) {
			blk_mq_put_dispatch_budget(hctx);
			break;
		}

		/*
		 * 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);

	} while (blk_mq_dispatch_rq_list(q, &rq_list, true));

	WRITE_ONCE(hctx->dispatch_from, ctx);
}

175
/* return true if hw queue need to be run again */
176
void blk_mq_sched_dispatch_requests(struct blk_mq_hw_ctx *hctx)
177
{
178 179
	struct request_queue *q = hctx->queue;
	struct elevator_queue *e = q->elevator;
180
	const bool has_sched_dispatch = e && e->type->ops.mq.dispatch_request;
181 182
	LIST_HEAD(rq_list);

183 184
	/* RCU or SRCU read lock is needed before checking quiesced flag */
	if (unlikely(blk_mq_hctx_stopped(hctx) || blk_queue_quiesced(q)))
185
		return;
186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207

	hctx->run++;

	/*
	 * 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.
208 209 210 211
	 *
	 * 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.
212
	 */
213
	if (!list_empty(&rq_list)) {
214
		blk_mq_sched_mark_restart_hctx(hctx);
215 216
		if (blk_mq_dispatch_rq_list(q, &rq_list, false)) {
			if (has_sched_dispatch)
217
				blk_mq_do_dispatch_sched(hctx);
218
			else
219
				blk_mq_do_dispatch_ctx(hctx);
220
		}
221
	} else if (has_sched_dispatch) {
222
		blk_mq_do_dispatch_sched(hctx);
223 224 225 226 227 228 229 230 231
	} else if (q->mq_ops->get_budget) {
		/*
		 * If we need to get budget before queuing request, we
		 * dequeue request one by one from sw queue for avoiding
		 * to mess up I/O merge when dispatch runs out of resource.
		 *
		 * TODO: get more budgets, and dequeue more requests in
		 * one time.
		 */
232
		blk_mq_do_dispatch_ctx(hctx);
233
	} else {
234
		blk_mq_flush_busy_ctxs(hctx, &rq_list);
235
		blk_mq_dispatch_rq_list(q, &rq_list, false);
236
	}
237 238
}

239 240
bool blk_mq_sched_try_merge(struct request_queue *q, struct bio *bio,
			    struct request **merged_request)
241 242 243
{
	struct request *rq;

244 245
	switch (elv_merge(q, &rq, bio)) {
	case ELEVATOR_BACK_MERGE:
246 247
		if (!blk_mq_sched_allow_merge(q, rq, bio))
			return false;
248 249 250 251 252 253 254
		if (!bio_attempt_back_merge(q, rq, bio))
			return false;
		*merged_request = attempt_back_merge(q, rq);
		if (!*merged_request)
			elv_merged_request(q, rq, ELEVATOR_BACK_MERGE);
		return true;
	case ELEVATOR_FRONT_MERGE:
255 256
		if (!blk_mq_sched_allow_merge(q, rq, bio))
			return false;
257 258 259 260 261 262 263 264
		if (!bio_attempt_front_merge(q, rq, bio))
			return false;
		*merged_request = attempt_front_merge(q, rq);
		if (!*merged_request)
			elv_merged_request(q, rq, ELEVATOR_FRONT_MERGE);
		return true;
	default:
		return false;
265 266 267 268
	}
}
EXPORT_SYMBOL_GPL(blk_mq_sched_try_merge);

269 270 271 272 273 274 275 276 277 278 279
/*
 * 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.
 */
static bool blk_mq_attempt_merge(struct request_queue *q,
				 struct blk_mq_ctx *ctx, struct bio *bio)
{
	struct request *rq;
	int checked = 8;

280 281
	lockdep_assert_held(&ctx->lock);

282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314
	list_for_each_entry_reverse(rq, &ctx->rq_list, queuelist) {
		bool merged = false;

		if (!checked--)
			break;

		if (!blk_rq_merge_ok(rq, bio))
			continue;

		switch (blk_try_merge(rq, bio)) {
		case ELEVATOR_BACK_MERGE:
			if (blk_mq_sched_allow_merge(q, rq, bio))
				merged = bio_attempt_back_merge(q, rq, bio);
			break;
		case ELEVATOR_FRONT_MERGE:
			if (blk_mq_sched_allow_merge(q, rq, bio))
				merged = bio_attempt_front_merge(q, rq, bio);
			break;
		case ELEVATOR_DISCARD_MERGE:
			merged = bio_attempt_discard_merge(q, rq, bio);
			break;
		default:
			continue;
		}

		if (merged)
			ctx->rq_merged++;
		return merged;
	}

	return false;
}

315 316 317
bool __blk_mq_sched_bio_merge(struct request_queue *q, struct bio *bio)
{
	struct elevator_queue *e = q->elevator;
318 319 320
	struct blk_mq_ctx *ctx = blk_mq_get_ctx(q);
	struct blk_mq_hw_ctx *hctx = blk_mq_map_queue(q, ctx->cpu);
	bool ret = false;
321

322
	if (e && e->type->ops.mq.bio_merge) {
323 324 325 326
		blk_mq_put_ctx(ctx);
		return e->type->ops.mq.bio_merge(hctx, bio);
	}

327 328 329 330 331 332 333 334 335
	if (hctx->flags & BLK_MQ_F_SHOULD_MERGE) {
		/* default per sw-queue merge */
		spin_lock(&ctx->lock);
		ret = blk_mq_attempt_merge(q, ctx, bio);
		spin_unlock(&ctx->lock);
	}

	blk_mq_put_ctx(ctx);
	return ret;
336 337 338 339 340 341 342 343 344 345 346 347 348 349
}

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);

void blk_mq_sched_request_inserted(struct request *rq)
{
	trace_block_rq_insert(rq->q, rq);
}
EXPORT_SYMBOL_GPL(blk_mq_sched_request_inserted);

350
static bool blk_mq_sched_bypass_insert(struct blk_mq_hw_ctx *hctx,
351
				       bool has_sched,
352
				       struct request *rq)
353
{
354 355 356 357 358 359 360 361
	/* dispatch flush rq directly */
	if (rq->rq_flags & RQF_FLUSH_SEQ) {
		spin_lock(&hctx->lock);
		list_add(&rq->queuelist, &hctx->dispatch);
		spin_unlock(&hctx->lock);
		return true;
	}

362
	if (has_sched)
363 364
		rq->rq_flags |= RQF_SORTED;

365
	return false;
366 367
}

368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429
/**
 * list_for_each_entry_rcu_rr - iterate in a round-robin fashion over rcu list
 * @pos:    loop cursor.
 * @skip:   the list element that will not be examined. Iteration starts at
 *          @skip->next.
 * @head:   head of the list to examine. This list must have at least one
 *          element, namely @skip.
 * @member: name of the list_head structure within typeof(*pos).
 */
#define list_for_each_entry_rcu_rr(pos, skip, head, member)		\
	for ((pos) = (skip);						\
	     (pos = (pos)->member.next != (head) ? list_entry_rcu(	\
			(pos)->member.next, typeof(*pos), member) :	\
	      list_entry_rcu((pos)->member.next->next, typeof(*pos), member)), \
	     (pos) != (skip); )

/*
 * Called after a driver tag has been freed to check whether a hctx needs to
 * be restarted. Restarts @hctx if its tag set is not shared. Restarts hardware
 * queues in a round-robin fashion if the tag set of @hctx is shared with other
 * hardware queues.
 */
void blk_mq_sched_restart(struct blk_mq_hw_ctx *const hctx)
{
	struct blk_mq_tags *const tags = hctx->tags;
	struct blk_mq_tag_set *const set = hctx->queue->tag_set;
	struct request_queue *const queue = hctx->queue, *q;
	struct blk_mq_hw_ctx *hctx2;
	unsigned int i, j;

	if (set->flags & BLK_MQ_F_TAG_SHARED) {
		/*
		 * If this is 0, then we know that no hardware queues
		 * have RESTART marked. We're done.
		 */
		if (!atomic_read(&queue->shared_hctx_restart))
			return;

		rcu_read_lock();
		list_for_each_entry_rcu_rr(q, queue, &set->tag_list,
					   tag_set_list) {
			queue_for_each_hw_ctx(q, hctx2, i)
				if (hctx2->tags == tags &&
				    blk_mq_sched_restart_hctx(hctx2))
					goto done;
		}
		j = hctx->queue_num + 1;
		for (i = 0; i < queue->nr_hw_queues; i++, j++) {
			if (j == queue->nr_hw_queues)
				j = 0;
			hctx2 = queue->queue_hw_ctx[j];
			if (hctx2->tags == tags &&
			    blk_mq_sched_restart_hctx(hctx2))
				break;
		}
done:
		rcu_read_unlock();
	} else {
		blk_mq_sched_restart_hctx(hctx);
	}
}

430 431 432 433 434 435 436 437
void blk_mq_sched_insert_request(struct request *rq, bool at_head,
				 bool run_queue, bool async, bool can_block)
{
	struct request_queue *q = rq->q;
	struct elevator_queue *e = q->elevator;
	struct blk_mq_ctx *ctx = rq->mq_ctx;
	struct blk_mq_hw_ctx *hctx = blk_mq_map_queue(q, ctx->cpu);

438 439
	/* flush rq in flush machinery need to be dispatched directly */
	if (!(rq->rq_flags & RQF_FLUSH_SEQ) && op_is_flush(rq->cmd_flags)) {
440 441
		blk_insert_flush(rq);
		goto run;
442 443
	}

444 445
	WARN_ON(e && (rq->tag != -1));

446
	if (blk_mq_sched_bypass_insert(hctx, !!e, rq))
447 448
		goto run;

449 450 451 452 453 454 455 456 457 458 459
	if (e && e->type->ops.mq.insert_requests) {
		LIST_HEAD(list);

		list_add(&rq->queuelist, &list);
		e->type->ops.mq.insert_requests(hctx, &list, at_head);
	} else {
		spin_lock(&ctx->lock);
		__blk_mq_insert_request(hctx, rq, at_head);
		spin_unlock(&ctx->lock);
	}

460
run:
461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479
	if (run_queue)
		blk_mq_run_hw_queue(hctx, async);
}

void blk_mq_sched_insert_requests(struct request_queue *q,
				  struct blk_mq_ctx *ctx,
				  struct list_head *list, bool run_queue_async)
{
	struct blk_mq_hw_ctx *hctx = blk_mq_map_queue(q, ctx->cpu);
	struct elevator_queue *e = hctx->queue->elevator;

	if (e && e->type->ops.mq.insert_requests)
		e->type->ops.mq.insert_requests(hctx, list, false);
	else
		blk_mq_insert_requests(hctx, ctx, list);

	blk_mq_run_hw_queue(hctx, run_queue_async);
}

480 481 482 483 484 485 486 487 488 489 490
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);
		blk_mq_free_rq_map(hctx->sched_tags);
		hctx->sched_tags = NULL;
	}
}

491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509
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,
					       set->reserved_tags);
	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;
}

510
static void blk_mq_sched_tags_teardown(struct request_queue *q)
511 512 513
{
	struct blk_mq_tag_set *set = q->tag_set;
	struct blk_mq_hw_ctx *hctx;
514 515 516 517 518 519
	int i;

	queue_for_each_hw_ctx(q, hctx, i)
		blk_mq_sched_free_tags(set, hctx, i);
}

520 521 522 523
int blk_mq_sched_init_hctx(struct request_queue *q, struct blk_mq_hw_ctx *hctx,
			   unsigned int hctx_idx)
{
	struct elevator_queue *e = q->elevator;
524
	int ret;
525 526 527 528

	if (!e)
		return 0;

529 530 531 532 533 534 535 536 537 538 539 540
	ret = blk_mq_sched_alloc_tags(q, hctx, hctx_idx);
	if (ret)
		return ret;

	if (e->type->ops.mq.init_hctx) {
		ret = e->type->ops.mq.init_hctx(hctx, hctx_idx);
		if (ret) {
			blk_mq_sched_free_tags(q->tag_set, hctx, hctx_idx);
			return ret;
		}
	}

541 542
	blk_mq_debugfs_register_sched_hctx(q, hctx);

543
	return 0;
544 545 546 547 548 549 550 551 552 553
}

void blk_mq_sched_exit_hctx(struct request_queue *q, struct blk_mq_hw_ctx *hctx,
			    unsigned int hctx_idx)
{
	struct elevator_queue *e = q->elevator;

	if (!e)
		return;

554 555
	blk_mq_debugfs_unregister_sched_hctx(hctx);

556 557 558 559 560
	if (e->type->ops.mq.exit_hctx && hctx->sched_data) {
		e->type->ops.mq.exit_hctx(hctx, hctx_idx);
		hctx->sched_data = NULL;
	}

561 562 563
	blk_mq_sched_free_tags(q->tag_set, hctx, hctx_idx);
}

564 565 566
int blk_mq_init_sched(struct request_queue *q, struct elevator_type *e)
{
	struct blk_mq_hw_ctx *hctx;
567
	struct elevator_queue *eq;
568 569 570 571 572 573 574
	unsigned int i;
	int ret;

	if (!e) {
		q->elevator = NULL;
		return 0;
	}
575 576

	/*
577 578 579
	 * 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.
580
	 */
581 582
	q->nr_requests = 2 * min_t(unsigned int, q->tag_set->queue_depth,
				   BLKDEV_MAX_RQ);
583 584

	queue_for_each_hw_ctx(q, hctx, i) {
585
		ret = blk_mq_sched_alloc_tags(q, hctx, i);
586
		if (ret)
587
			goto err;
588 589
	}

590 591 592
	ret = e->ops.mq.init_sched(q, e);
	if (ret)
		goto err;
593

594 595 596 597
	blk_mq_debugfs_register_sched(q);

	queue_for_each_hw_ctx(q, hctx, i) {
		if (e->ops.mq.init_hctx) {
598 599 600 601 602 603 604 605
			ret = e->ops.mq.init_hctx(hctx, i);
			if (ret) {
				eq = q->elevator;
				blk_mq_exit_sched(q, eq);
				kobject_put(&eq->kobj);
				return ret;
			}
		}
606
		blk_mq_debugfs_register_sched_hctx(q, hctx);
607 608
	}

609 610
	return 0;

611
err:
612 613
	blk_mq_sched_tags_teardown(q);
	q->elevator = NULL;
614
	return ret;
615
}
616

617 618
void blk_mq_exit_sched(struct request_queue *q, struct elevator_queue *e)
{
619 620 621
	struct blk_mq_hw_ctx *hctx;
	unsigned int i;

622 623 624 625 626
	queue_for_each_hw_ctx(q, hctx, i) {
		blk_mq_debugfs_unregister_sched_hctx(hctx);
		if (e->type->ops.mq.exit_hctx && hctx->sched_data) {
			e->type->ops.mq.exit_hctx(hctx, i);
			hctx->sched_data = NULL;
627 628
		}
	}
629
	blk_mq_debugfs_unregister_sched(q);
630 631 632 633 634 635
	if (e->type->ops.mq.exit_sched)
		e->type->ops.mq.exit_sched(e);
	blk_mq_sched_tags_teardown(q);
	q->elevator = NULL;
}

636 637 638 639 640 641 642 643 644 645
int blk_mq_sched_init(struct request_queue *q)
{
	int ret;

	mutex_lock(&q->sysfs_lock);
	ret = elevator_init(q, NULL);
	mutex_unlock(&q->sysfs_lock);

	return ret;
}