blk-mq.c 46.5 KB
Newer Older
1 2 3 4 5 6
/*
 * Block multiqueue core code
 *
 * Copyright (C) 2013-2014 Jens Axboe
 * Copyright (C) 2013-2014 Christoph Hellwig
 */
7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42
#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/backing-dev.h>
#include <linux/bio.h>
#include <linux/blkdev.h>
#include <linux/mm.h>
#include <linux/init.h>
#include <linux/slab.h>
#include <linux/workqueue.h>
#include <linux/smp.h>
#include <linux/llist.h>
#include <linux/list_sort.h>
#include <linux/cpu.h>
#include <linux/cache.h>
#include <linux/sched/sysctl.h>
#include <linux/delay.h>

#include <trace/events/block.h>

#include <linux/blk-mq.h>
#include "blk.h"
#include "blk-mq.h"
#include "blk-mq-tag.h"

static DEFINE_MUTEX(all_q_mutex);
static LIST_HEAD(all_q_list);

static void __blk_mq_run_hw_queue(struct blk_mq_hw_ctx *hctx);

/*
 * Check if any of the ctx's have pending work in this hardware queue
 */
static bool blk_mq_hctx_has_pending(struct blk_mq_hw_ctx *hctx)
{
	unsigned int i;

43 44
	for (i = 0; i < hctx->ctx_map.map_size; i++)
		if (hctx->ctx_map.map[i].word)
45 46 47 48 49
			return true;

	return false;
}

50 51 52 53 54 55 56 57 58
static inline struct blk_align_bitmap *get_bm(struct blk_mq_hw_ctx *hctx,
					      struct blk_mq_ctx *ctx)
{
	return &hctx->ctx_map.map[ctx->index_hw / hctx->ctx_map.bits_per_word];
}

#define CTX_TO_BIT(hctx, ctx)	\
	((ctx)->index_hw & ((hctx)->ctx_map.bits_per_word - 1))

59 60 61 62 63 64
/*
 * Mark this ctx as having pending work in this hardware queue
 */
static void blk_mq_hctx_mark_pending(struct blk_mq_hw_ctx *hctx,
				     struct blk_mq_ctx *ctx)
{
65 66 67 68 69 70 71 72 73 74 75 76
	struct blk_align_bitmap *bm = get_bm(hctx, ctx);

	if (!test_bit(CTX_TO_BIT(hctx, ctx), &bm->word))
		set_bit(CTX_TO_BIT(hctx, ctx), &bm->word);
}

static void blk_mq_hctx_clear_pending(struct blk_mq_hw_ctx *hctx,
				      struct blk_mq_ctx *ctx)
{
	struct blk_align_bitmap *bm = get_bm(hctx, ctx);

	clear_bit(CTX_TO_BIT(hctx, ctx), &bm->word);
77 78 79 80
}

static int blk_mq_queue_enter(struct request_queue *q)
{
81 82
	while (true) {
		int ret;
83

84 85
		if (percpu_ref_tryget_live(&q->mq_usage_counter))
			return 0;
86

87 88 89 90 91 92 93
		ret = wait_event_interruptible(q->mq_freeze_wq,
				!q->mq_freeze_depth || blk_queue_dying(q));
		if (blk_queue_dying(q))
			return -ENODEV;
		if (ret)
			return ret;
	}
94 95 96 97
}

static void blk_mq_queue_exit(struct request_queue *q)
{
98 99 100 101 102 103 104 105 106
	percpu_ref_put(&q->mq_usage_counter);
}

static void blk_mq_usage_counter_release(struct percpu_ref *ref)
{
	struct request_queue *q =
		container_of(ref, struct request_queue, mq_usage_counter);

	wake_up_all(&q->mq_freeze_wq);
107 108
}

109 110 111 112 113
/*
 * Guarantee no request is in use, so we can change any data structure of
 * the queue afterward.
 */
void blk_mq_freeze_queue(struct request_queue *q)
114
{
115 116
	bool freeze;

117
	spin_lock_irq(q->queue_lock);
118
	freeze = !q->mq_freeze_depth++;
119 120
	spin_unlock_irq(q->queue_lock);

121 122 123 124
	if (freeze) {
		percpu_ref_kill(&q->mq_usage_counter);
		blk_mq_run_queues(q, false);
	}
125
	wait_event(q->mq_freeze_wq, percpu_ref_is_zero(&q->mq_usage_counter));
126 127
}

128 129
static void blk_mq_unfreeze_queue(struct request_queue *q)
{
130
	bool wake;
131 132

	spin_lock_irq(q->queue_lock);
133 134
	wake = !--q->mq_freeze_depth;
	WARN_ON_ONCE(q->mq_freeze_depth < 0);
135
	spin_unlock_irq(q->queue_lock);
136 137
	if (wake) {
		percpu_ref_reinit(&q->mq_usage_counter);
138
		wake_up_all(&q->mq_freeze_wq);
139
	}
140 141 142 143 144 145 146 147
}

bool blk_mq_can_queue(struct blk_mq_hw_ctx *hctx)
{
	return blk_mq_has_free_tags(hctx->tags);
}
EXPORT_SYMBOL(blk_mq_can_queue);

148 149
static void blk_mq_rq_ctx_init(struct request_queue *q, struct blk_mq_ctx *ctx,
			       struct request *rq, unsigned int rw_flags)
150
{
151 152 153
	if (blk_queue_io_stat(q))
		rw_flags |= REQ_IO_STAT;

154 155 156
	INIT_LIST_HEAD(&rq->queuelist);
	/* csd/requeue_work/fifo_time is initialized before use */
	rq->q = q;
157
	rq->mq_ctx = ctx;
158
	rq->cmd_flags |= rw_flags;
159 160 161 162 163 164
	/* do not touch atomic flags, it needs atomic ops against the timer */
	rq->cpu = -1;
	INIT_HLIST_NODE(&rq->hash);
	RB_CLEAR_NODE(&rq->rb_node);
	rq->rq_disk = NULL;
	rq->part = NULL;
165
	rq->start_time = jiffies;
166 167
#ifdef CONFIG_BLK_CGROUP
	rq->rl = NULL;
168
	set_start_time_ns(rq);
169 170 171 172 173 174 175 176 177 178
	rq->io_start_time_ns = 0;
#endif
	rq->nr_phys_segments = 0;
#if defined(CONFIG_BLK_DEV_INTEGRITY)
	rq->nr_integrity_segments = 0;
#endif
	rq->special = NULL;
	/* tag was already set */
	rq->errors = 0;

179 180
	rq->cmd = rq->__cmd;

181 182 183 184 185 186
	rq->extra_len = 0;
	rq->sense_len = 0;
	rq->resid_len = 0;
	rq->sense = NULL;

	INIT_LIST_HEAD(&rq->timeout_list);
187 188
	rq->timeout = 0;

189 190 191 192
	rq->end_io = NULL;
	rq->end_io_data = NULL;
	rq->next_rq = NULL;

193 194 195
	ctx->rq_dispatched[rw_is_sync(rw_flags)]++;
}

196
static struct request *
197
__blk_mq_alloc_request(struct blk_mq_alloc_data *data, int rw)
198 199 200 201
{
	struct request *rq;
	unsigned int tag;

202
	tag = blk_mq_get_tag(data);
203
	if (tag != BLK_MQ_TAG_FAIL) {
204
		rq = data->hctx->tags->rqs[tag];
205 206

		rq->cmd_flags = 0;
207
		if (blk_mq_tag_busy(data->hctx)) {
208
			rq->cmd_flags = REQ_MQ_INFLIGHT;
209
			atomic_inc(&data->hctx->nr_active);
210 211 212
		}

		rq->tag = tag;
213
		blk_mq_rq_ctx_init(data->q, data->ctx, rq, rw);
214 215 216 217 218 219
		return rq;
	}

	return NULL;
}

220 221
struct request *blk_mq_alloc_request(struct request_queue *q, int rw, gfp_t gfp,
		bool reserved)
222
{
223 224
	struct blk_mq_ctx *ctx;
	struct blk_mq_hw_ctx *hctx;
225
	struct request *rq;
226
	struct blk_mq_alloc_data alloc_data;
227 228 229 230

	if (blk_mq_queue_enter(q))
		return NULL;

231 232
	ctx = blk_mq_get_ctx(q);
	hctx = q->mq_ops->map_queue(q, ctx->cpu);
233 234
	blk_mq_set_alloc_data(&alloc_data, q, gfp & ~__GFP_WAIT,
			reserved, ctx, hctx);
235

236
	rq = __blk_mq_alloc_request(&alloc_data, rw);
237 238 239 240 241 242
	if (!rq && (gfp & __GFP_WAIT)) {
		__blk_mq_run_hw_queue(hctx);
		blk_mq_put_ctx(ctx);

		ctx = blk_mq_get_ctx(q);
		hctx = q->mq_ops->map_queue(q, ctx->cpu);
243 244 245 246
		blk_mq_set_alloc_data(&alloc_data, q, gfp, reserved, ctx,
				hctx);
		rq =  __blk_mq_alloc_request(&alloc_data, rw);
		ctx = alloc_data.ctx;
247 248
	}
	blk_mq_put_ctx(ctx);
249 250
	return rq;
}
251
EXPORT_SYMBOL(blk_mq_alloc_request);
252 253 254 255 256 257 258

static void __blk_mq_free_request(struct blk_mq_hw_ctx *hctx,
				  struct blk_mq_ctx *ctx, struct request *rq)
{
	const int tag = rq->tag;
	struct request_queue *q = rq->q;

259 260 261
	if (rq->cmd_flags & REQ_MQ_INFLIGHT)
		atomic_dec(&hctx->nr_active);

262
	clear_bit(REQ_ATOM_STARTED, &rq->atomic_flags);
263
	blk_mq_put_tag(hctx, tag, &ctx->last_tag);
264 265 266 267 268 269 270 271 272 273 274 275 276 277 278
	blk_mq_queue_exit(q);
}

void blk_mq_free_request(struct request *rq)
{
	struct blk_mq_ctx *ctx = rq->mq_ctx;
	struct blk_mq_hw_ctx *hctx;
	struct request_queue *q = rq->q;

	ctx->rq_completed[rq_is_sync(rq)]++;

	hctx = q->mq_ops->map_queue(q, ctx->cpu);
	__blk_mq_free_request(hctx, ctx, rq);
}

279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298
/*
 * Clone all relevant state from a request that has been put on hold in
 * the flush state machine into the preallocated flush request that hangs
 * off the request queue.
 *
 * For a driver the flush request should be invisible, that's why we are
 * impersonating the original request here.
 */
void blk_mq_clone_flush_request(struct request *flush_rq,
		struct request *orig_rq)
{
	struct blk_mq_hw_ctx *hctx =
		orig_rq->q->mq_ops->map_queue(orig_rq->q, orig_rq->mq_ctx->cpu);

	flush_rq->mq_ctx = orig_rq->mq_ctx;
	flush_rq->tag = orig_rq->tag;
	memcpy(blk_mq_rq_to_pdu(flush_rq), blk_mq_rq_to_pdu(orig_rq),
		hctx->cmd_size);
}

299
inline void __blk_mq_end_io(struct request *rq, int error)
300
{
M
Ming Lei 已提交
301 302
	blk_account_io_done(rq);

C
Christoph Hellwig 已提交
303
	if (rq->end_io) {
304
		rq->end_io(rq, error);
C
Christoph Hellwig 已提交
305 306 307
	} else {
		if (unlikely(blk_bidi_rq(rq)))
			blk_mq_free_request(rq->next_rq);
308
		blk_mq_free_request(rq);
C
Christoph Hellwig 已提交
309
	}
310
}
311 312 313 314 315 316 317 318 319
EXPORT_SYMBOL(__blk_mq_end_io);

void blk_mq_end_io(struct request *rq, int error)
{
	if (blk_update_request(rq, error, blk_rq_bytes(rq)))
		BUG();
	__blk_mq_end_io(rq, error);
}
EXPORT_SYMBOL(blk_mq_end_io);
320

321
static void __blk_mq_complete_request_remote(void *data)
322
{
323
	struct request *rq = data;
324

325
	rq->q->softirq_done_fn(rq);
326 327
}

328
static void blk_mq_ipi_complete_request(struct request *rq)
329 330
{
	struct blk_mq_ctx *ctx = rq->mq_ctx;
C
Christoph Hellwig 已提交
331
	bool shared = false;
332 333
	int cpu;

C
Christoph Hellwig 已提交
334
	if (!test_bit(QUEUE_FLAG_SAME_COMP, &rq->q->queue_flags)) {
335 336 337
		rq->q->softirq_done_fn(rq);
		return;
	}
338 339

	cpu = get_cpu();
C
Christoph Hellwig 已提交
340 341 342 343
	if (!test_bit(QUEUE_FLAG_SAME_FORCE, &rq->q->queue_flags))
		shared = cpus_share_cache(cpu, ctx->cpu);

	if (cpu != ctx->cpu && !shared && cpu_online(ctx->cpu)) {
344
		rq->csd.func = __blk_mq_complete_request_remote;
345 346
		rq->csd.info = rq;
		rq->csd.flags = 0;
347
		smp_call_function_single_async(ctx->cpu, &rq->csd);
348
	} else {
349
		rq->q->softirq_done_fn(rq);
350
	}
351 352
	put_cpu();
}
353

354 355 356 357 358 359 360 361 362 363
void __blk_mq_complete_request(struct request *rq)
{
	struct request_queue *q = rq->q;

	if (!q->softirq_done_fn)
		blk_mq_end_io(rq, rq->errors);
	else
		blk_mq_ipi_complete_request(rq);
}

364 365 366 367 368 369 370 371 372 373
/**
 * blk_mq_complete_request - end I/O on a request
 * @rq:		the request being processed
 *
 * Description:
 *	Ends all I/O on a request. It does not handle partial completions.
 *	The actual completion happens out-of-order, through a IPI handler.
 **/
void blk_mq_complete_request(struct request *rq)
{
374 375 376
	struct request_queue *q = rq->q;

	if (unlikely(blk_should_fake_timeout(q)))
377
		return;
378 379
	if (!blk_mark_rq_complete(rq))
		__blk_mq_complete_request(rq);
380 381
}
EXPORT_SYMBOL(blk_mq_complete_request);
382

383
static void blk_mq_start_request(struct request *rq, bool last)
384 385 386 387 388
{
	struct request_queue *q = rq->q;

	trace_block_rq_issue(q, rq);

C
Christoph Hellwig 已提交
389
	rq->resid_len = blk_rq_bytes(rq);
C
Christoph Hellwig 已提交
390 391
	if (unlikely(blk_bidi_rq(rq)))
		rq->next_rq->resid_len = blk_rq_bytes(rq->next_rq);
C
Christoph Hellwig 已提交
392

393
	blk_add_timer(rq);
394 395 396 397 398 399 400

	/*
	 * Mark us as started and clear complete. Complete might have been
	 * set if requeue raced with timeout, which then marked it as
	 * complete. So be sure to clear complete again when we start
	 * the request, otherwise we'll ignore the completion event.
	 */
401 402 403 404
	if (!test_bit(REQ_ATOM_STARTED, &rq->atomic_flags))
		set_bit(REQ_ATOM_STARTED, &rq->atomic_flags);
	if (test_bit(REQ_ATOM_COMPLETE, &rq->atomic_flags))
		clear_bit(REQ_ATOM_COMPLETE, &rq->atomic_flags);
405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423

	if (q->dma_drain_size && blk_rq_bytes(rq)) {
		/*
		 * Make sure space for the drain appears.  We know we can do
		 * this because max_hw_segments has been adjusted to be one
		 * fewer than the device can handle.
		 */
		rq->nr_phys_segments++;
	}

	/*
	 * Flag the last request in the series so that drivers know when IO
	 * should be kicked off, if they don't do it on a per-request basis.
	 *
	 * Note: the flag isn't the only condition drivers should do kick off.
	 * If drive is busy, the last request might not have the bit set.
	 */
	if (last)
		rq->cmd_flags |= REQ_END;
424 425
}

426
static void __blk_mq_requeue_request(struct request *rq)
427 428 429 430 431
{
	struct request_queue *q = rq->q;

	trace_block_rq_requeue(q, rq);
	clear_bit(REQ_ATOM_STARTED, &rq->atomic_flags);
432 433 434 435 436

	rq->cmd_flags &= ~REQ_END;

	if (q->dma_drain_size && blk_rq_bytes(rq))
		rq->nr_phys_segments--;
437 438
}

439 440 441 442 443 444
void blk_mq_requeue_request(struct request *rq)
{
	__blk_mq_requeue_request(rq);
	blk_clear_rq_complete(rq);

	BUG_ON(blk_queued_rq(rq));
445
	blk_mq_add_to_requeue_list(rq, true);
446 447 448
}
EXPORT_SYMBOL(blk_mq_requeue_request);

449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506
static void blk_mq_requeue_work(struct work_struct *work)
{
	struct request_queue *q =
		container_of(work, struct request_queue, requeue_work);
	LIST_HEAD(rq_list);
	struct request *rq, *next;
	unsigned long flags;

	spin_lock_irqsave(&q->requeue_lock, flags);
	list_splice_init(&q->requeue_list, &rq_list);
	spin_unlock_irqrestore(&q->requeue_lock, flags);

	list_for_each_entry_safe(rq, next, &rq_list, queuelist) {
		if (!(rq->cmd_flags & REQ_SOFTBARRIER))
			continue;

		rq->cmd_flags &= ~REQ_SOFTBARRIER;
		list_del_init(&rq->queuelist);
		blk_mq_insert_request(rq, true, false, false);
	}

	while (!list_empty(&rq_list)) {
		rq = list_entry(rq_list.next, struct request, queuelist);
		list_del_init(&rq->queuelist);
		blk_mq_insert_request(rq, false, false, false);
	}

	blk_mq_run_queues(q, false);
}

void blk_mq_add_to_requeue_list(struct request *rq, bool at_head)
{
	struct request_queue *q = rq->q;
	unsigned long flags;

	/*
	 * We abuse this flag that is otherwise used by the I/O scheduler to
	 * request head insertation from the workqueue.
	 */
	BUG_ON(rq->cmd_flags & REQ_SOFTBARRIER);

	spin_lock_irqsave(&q->requeue_lock, flags);
	if (at_head) {
		rq->cmd_flags |= REQ_SOFTBARRIER;
		list_add(&rq->queuelist, &q->requeue_list);
	} else {
		list_add_tail(&rq->queuelist, &q->requeue_list);
	}
	spin_unlock_irqrestore(&q->requeue_lock, flags);
}
EXPORT_SYMBOL(blk_mq_add_to_requeue_list);

void blk_mq_kick_requeue_list(struct request_queue *q)
{
	kblockd_schedule_work(&q->requeue_work);
}
EXPORT_SYMBOL(blk_mq_kick_requeue_list);

507
static inline bool is_flush_request(struct request *rq, unsigned int tag)
508
{
509 510 511 512 513 514 515
	return ((rq->cmd_flags & REQ_FLUSH_SEQ) &&
			rq->q->flush_rq->tag == tag);
}

struct request *blk_mq_tag_to_rq(struct blk_mq_tags *tags, unsigned int tag)
{
	struct request *rq = tags->rqs[tag];
516

517 518
	if (!is_flush_request(rq, tag))
		return rq;
519

520
	return rq->q->flush_rq;
521 522 523
}
EXPORT_SYMBOL(blk_mq_tag_to_rq);

524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544
struct blk_mq_timeout_data {
	struct blk_mq_hw_ctx *hctx;
	unsigned long *next;
	unsigned int *next_set;
};

static void blk_mq_timeout_check(void *__data, unsigned long *free_tags)
{
	struct blk_mq_timeout_data *data = __data;
	struct blk_mq_hw_ctx *hctx = data->hctx;
	unsigned int tag;

	 /* It may not be in flight yet (this is where
	 * the REQ_ATOMIC_STARTED flag comes in). The requests are
	 * statically allocated, so we know it's always safe to access the
	 * memory associated with a bit offset into ->rqs[].
	 */
	tag = 0;
	do {
		struct request *rq;

545 546
		tag = find_next_zero_bit(free_tags, hctx->tags->nr_tags, tag);
		if (tag >= hctx->tags->nr_tags)
547 548
			break;

549
		rq = blk_mq_tag_to_rq(hctx->tags, tag++);
550 551
		if (rq->q != hctx->queue)
			continue;
552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575
		if (!test_bit(REQ_ATOM_STARTED, &rq->atomic_flags))
			continue;

		blk_rq_check_expired(rq, data->next, data->next_set);
	} while (1);
}

static void blk_mq_hw_ctx_check_timeout(struct blk_mq_hw_ctx *hctx,
					unsigned long *next,
					unsigned int *next_set)
{
	struct blk_mq_timeout_data data = {
		.hctx		= hctx,
		.next		= next,
		.next_set	= next_set,
	};

	/*
	 * Ask the tagging code to iterate busy requests, so we can
	 * check them for timeout.
	 */
	blk_mq_tag_busy_iter(hctx->tags, blk_mq_timeout_check, &data);
}

576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597
static enum blk_eh_timer_return blk_mq_rq_timed_out(struct request *rq)
{
	struct request_queue *q = rq->q;

	/*
	 * We know that complete is set at this point. If STARTED isn't set
	 * anymore, then the request isn't active and the "timeout" should
	 * just be ignored. This can happen due to the bitflag ordering.
	 * Timeout first checks if STARTED is set, and if it is, assumes
	 * the request is active. But if we race with completion, then
	 * we both flags will get cleared. So check here again, and ignore
	 * a timeout event with a request that isn't active.
	 */
	if (!test_bit(REQ_ATOM_STARTED, &rq->atomic_flags))
		return BLK_EH_NOT_HANDLED;

	if (!q->mq_ops->timeout)
		return BLK_EH_RESET_TIMER;

	return q->mq_ops->timeout(rq);
}

598 599 600 601 602 603 604
static void blk_mq_rq_timer(unsigned long data)
{
	struct request_queue *q = (struct request_queue *) data;
	struct blk_mq_hw_ctx *hctx;
	unsigned long next = 0;
	int i, next_set = 0;

605 606 607 608 609 610 611 612
	queue_for_each_hw_ctx(q, hctx, i) {
		/*
		 * If not software queues are currently mapped to this
		 * hardware queue, there's nothing to check
		 */
		if (!hctx->nr_ctx || !hctx->tags)
			continue;

613
		blk_mq_hw_ctx_check_timeout(hctx, &next, &next_set);
614
	}
615

616 617 618 619 620 621 622
	if (next_set) {
		next = blk_rq_timeout(round_jiffies_up(next));
		mod_timer(&q->timeout, next);
	} else {
		queue_for_each_hw_ctx(q, hctx, i)
			blk_mq_tag_idle(hctx);
	}
623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663
}

/*
 * 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;

	list_for_each_entry_reverse(rq, &ctx->rq_list, queuelist) {
		int el_ret;

		if (!checked--)
			break;

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

		el_ret = blk_try_merge(rq, bio);
		if (el_ret == ELEVATOR_BACK_MERGE) {
			if (bio_attempt_back_merge(q, rq, bio)) {
				ctx->rq_merged++;
				return true;
			}
			break;
		} else if (el_ret == ELEVATOR_FRONT_MERGE) {
			if (bio_attempt_front_merge(q, rq, bio)) {
				ctx->rq_merged++;
				return true;
			}
			break;
		}
	}

	return false;
}

664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697
/*
 * Process software queues that have been marked busy, splicing them
 * to the for-dispatch
 */
static void flush_busy_ctxs(struct blk_mq_hw_ctx *hctx, struct list_head *list)
{
	struct blk_mq_ctx *ctx;
	int i;

	for (i = 0; i < hctx->ctx_map.map_size; i++) {
		struct blk_align_bitmap *bm = &hctx->ctx_map.map[i];
		unsigned int off, bit;

		if (!bm->word)
			continue;

		bit = 0;
		off = i * hctx->ctx_map.bits_per_word;
		do {
			bit = find_next_bit(&bm->word, bm->depth, bit);
			if (bit >= bm->depth)
				break;

			ctx = hctx->ctxs[bit + off];
			clear_bit(bit, &bm->word);
			spin_lock(&ctx->lock);
			list_splice_tail_init(&ctx->rq_list, list);
			spin_unlock(&ctx->lock);

			bit++;
		} while (1);
	}
}

698 699 700 701 702 703 704 705 706 707 708
/*
 * Run this hardware queue, pulling any software queues mapped to it in.
 * Note that this function currently has various problems around ordering
 * of IO. In particular, we'd like FIFO behaviour on handling existing
 * items on the hctx->dispatch list. Ignore that for now.
 */
static void __blk_mq_run_hw_queue(struct blk_mq_hw_ctx *hctx)
{
	struct request_queue *q = hctx->queue;
	struct request *rq;
	LIST_HEAD(rq_list);
709
	int queued;
710

711
	WARN_ON(!cpumask_test_cpu(raw_smp_processor_id(), hctx->cpumask));
712

713
	if (unlikely(test_bit(BLK_MQ_S_STOPPED, &hctx->state)))
714 715 716 717 718 719 720
		return;

	hctx->run++;

	/*
	 * Touch any software queue that has pending entries.
	 */
721
	flush_busy_ctxs(hctx, &rq_list);
722 723 724 725 726 727 728 729 730 731 732 733 734 735 736

	/*
	 * If we have previous entries on our dispatch list, grab them
	 * and stuff them at the front 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);
	}

	/*
	 * Now process all the entries, sending them to the driver.
	 */
737
	queued = 0;
738 739 740 741 742 743
	while (!list_empty(&rq_list)) {
		int ret;

		rq = list_first_entry(&rq_list, struct request, queuelist);
		list_del_init(&rq->queuelist);

744
		blk_mq_start_request(rq, list_empty(&rq_list));
745 746 747 748 749 750 751 752

		ret = q->mq_ops->queue_rq(hctx, rq);
		switch (ret) {
		case BLK_MQ_RQ_QUEUE_OK:
			queued++;
			continue;
		case BLK_MQ_RQ_QUEUE_BUSY:
			list_add(&rq->queuelist, &rq_list);
753
			__blk_mq_requeue_request(rq);
754 755 756 757
			break;
		default:
			pr_err("blk-mq: bad return on queue: %d\n", ret);
		case BLK_MQ_RQ_QUEUE_ERROR:
758
			rq->errors = -EIO;
759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782
			blk_mq_end_io(rq, rq->errors);
			break;
		}

		if (ret == BLK_MQ_RQ_QUEUE_BUSY)
			break;
	}

	if (!queued)
		hctx->dispatched[0]++;
	else if (queued < (1 << (BLK_MQ_MAX_DISPATCH_ORDER - 1)))
		hctx->dispatched[ilog2(queued) + 1]++;

	/*
	 * Any items that need requeuing? Stuff them into hctx->dispatch,
	 * that is where we will continue on next queue run.
	 */
	if (!list_empty(&rq_list)) {
		spin_lock(&hctx->lock);
		list_splice(&rq_list, &hctx->dispatch);
		spin_unlock(&hctx->lock);
	}
}

783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806
/*
 * It'd be great if the workqueue API had a way to pass
 * in a mask and had some smarts for more clever placement.
 * For now we just round-robin here, switching for every
 * BLK_MQ_CPU_WORK_BATCH queued items.
 */
static int blk_mq_hctx_next_cpu(struct blk_mq_hw_ctx *hctx)
{
	int cpu = hctx->next_cpu;

	if (--hctx->next_cpu_batch <= 0) {
		int next_cpu;

		next_cpu = cpumask_next(hctx->next_cpu, hctx->cpumask);
		if (next_cpu >= nr_cpu_ids)
			next_cpu = cpumask_first(hctx->cpumask);

		hctx->next_cpu = next_cpu;
		hctx->next_cpu_batch = BLK_MQ_CPU_WORK_BATCH;
	}

	return cpu;
}

807 808
void blk_mq_run_hw_queue(struct blk_mq_hw_ctx *hctx, bool async)
{
809
	if (unlikely(test_bit(BLK_MQ_S_STOPPED, &hctx->state)))
810 811
		return;

812
	if (!async && cpumask_test_cpu(smp_processor_id(), hctx->cpumask))
813
		__blk_mq_run_hw_queue(hctx);
814
	else if (hctx->queue->nr_hw_queues == 1)
815
		kblockd_schedule_delayed_work(&hctx->run_work, 0);
816 817 818
	else {
		unsigned int cpu;

819
		cpu = blk_mq_hctx_next_cpu(hctx);
820
		kblockd_schedule_delayed_work_on(cpu, &hctx->run_work, 0);
821
	}
822 823 824 825 826 827 828 829 830 831
}

void blk_mq_run_queues(struct request_queue *q, bool async)
{
	struct blk_mq_hw_ctx *hctx;
	int i;

	queue_for_each_hw_ctx(q, hctx, i) {
		if ((!blk_mq_hctx_has_pending(hctx) &&
		    list_empty_careful(&hctx->dispatch)) ||
832
		    test_bit(BLK_MQ_S_STOPPED, &hctx->state))
833 834
			continue;

835
		preempt_disable();
836
		blk_mq_run_hw_queue(hctx, async);
837
		preempt_enable();
838 839 840 841 842 843
	}
}
EXPORT_SYMBOL(blk_mq_run_queues);

void blk_mq_stop_hw_queue(struct blk_mq_hw_ctx *hctx)
{
844 845
	cancel_delayed_work(&hctx->run_work);
	cancel_delayed_work(&hctx->delay_work);
846 847 848 849
	set_bit(BLK_MQ_S_STOPPED, &hctx->state);
}
EXPORT_SYMBOL(blk_mq_stop_hw_queue);

850 851 852 853 854 855 856 857 858 859
void blk_mq_stop_hw_queues(struct request_queue *q)
{
	struct blk_mq_hw_ctx *hctx;
	int i;

	queue_for_each_hw_ctx(q, hctx, i)
		blk_mq_stop_hw_queue(hctx);
}
EXPORT_SYMBOL(blk_mq_stop_hw_queues);

860 861 862
void blk_mq_start_hw_queue(struct blk_mq_hw_ctx *hctx)
{
	clear_bit(BLK_MQ_S_STOPPED, &hctx->state);
863 864

	preempt_disable();
865
	blk_mq_run_hw_queue(hctx, false);
866
	preempt_enable();
867 868 869
}
EXPORT_SYMBOL(blk_mq_start_hw_queue);

870 871 872 873 874 875 876 877 878 879 880
void blk_mq_start_hw_queues(struct request_queue *q)
{
	struct blk_mq_hw_ctx *hctx;
	int i;

	queue_for_each_hw_ctx(q, hctx, i)
		blk_mq_start_hw_queue(hctx);
}
EXPORT_SYMBOL(blk_mq_start_hw_queues);


881
void blk_mq_start_stopped_hw_queues(struct request_queue *q, bool async)
882 883 884 885 886 887 888 889 890
{
	struct blk_mq_hw_ctx *hctx;
	int i;

	queue_for_each_hw_ctx(q, hctx, i) {
		if (!test_bit(BLK_MQ_S_STOPPED, &hctx->state))
			continue;

		clear_bit(BLK_MQ_S_STOPPED, &hctx->state);
891
		preempt_disable();
892
		blk_mq_run_hw_queue(hctx, async);
893
		preempt_enable();
894 895 896 897
	}
}
EXPORT_SYMBOL(blk_mq_start_stopped_hw_queues);

898
static void blk_mq_run_work_fn(struct work_struct *work)
899 900 901
{
	struct blk_mq_hw_ctx *hctx;

902
	hctx = container_of(work, struct blk_mq_hw_ctx, run_work.work);
903

904 905 906
	__blk_mq_run_hw_queue(hctx);
}

907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925
static void blk_mq_delay_work_fn(struct work_struct *work)
{
	struct blk_mq_hw_ctx *hctx;

	hctx = container_of(work, struct blk_mq_hw_ctx, delay_work.work);

	if (test_and_clear_bit(BLK_MQ_S_STOPPED, &hctx->state))
		__blk_mq_run_hw_queue(hctx);
}

void blk_mq_delay_queue(struct blk_mq_hw_ctx *hctx, unsigned long msecs)
{
	unsigned long tmo = msecs_to_jiffies(msecs);

	if (hctx->queue->nr_hw_queues == 1)
		kblockd_schedule_delayed_work(&hctx->delay_work, tmo);
	else {
		unsigned int cpu;

926
		cpu = blk_mq_hctx_next_cpu(hctx);
927 928 929 930 931
		kblockd_schedule_delayed_work_on(cpu, &hctx->delay_work, tmo);
	}
}
EXPORT_SYMBOL(blk_mq_delay_queue);

932
static void __blk_mq_insert_request(struct blk_mq_hw_ctx *hctx,
933
				    struct request *rq, bool at_head)
934 935 936
{
	struct blk_mq_ctx *ctx = rq->mq_ctx;

937 938
	trace_block_rq_insert(hctx->queue, rq);

939 940 941 942
	if (at_head)
		list_add(&rq->queuelist, &ctx->rq_list);
	else
		list_add_tail(&rq->queuelist, &ctx->rq_list);
943

944 945 946
	blk_mq_hctx_mark_pending(hctx, ctx);
}

947 948
void blk_mq_insert_request(struct request *rq, bool at_head, bool run_queue,
		bool async)
949
{
950
	struct request_queue *q = rq->q;
951
	struct blk_mq_hw_ctx *hctx;
952 953 954 955 956
	struct blk_mq_ctx *ctx = rq->mq_ctx, *current_ctx;

	current_ctx = blk_mq_get_ctx(q);
	if (!cpu_online(ctx->cpu))
		rq->mq_ctx = ctx = current_ctx;
957 958 959

	hctx = q->mq_ops->map_queue(q, ctx->cpu);

960 961
	if (rq->cmd_flags & (REQ_FLUSH | REQ_FUA) &&
	    !(rq->cmd_flags & (REQ_FLUSH_SEQ))) {
962 963 964
		blk_insert_flush(rq);
	} else {
		spin_lock(&ctx->lock);
965
		__blk_mq_insert_request(hctx, rq, at_head);
966 967 968 969 970
		spin_unlock(&ctx->lock);
	}

	if (run_queue)
		blk_mq_run_hw_queue(hctx, async);
971 972

	blk_mq_put_ctx(current_ctx);
973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003
}

static void blk_mq_insert_requests(struct request_queue *q,
				     struct blk_mq_ctx *ctx,
				     struct list_head *list,
				     int depth,
				     bool from_schedule)

{
	struct blk_mq_hw_ctx *hctx;
	struct blk_mq_ctx *current_ctx;

	trace_block_unplug(q, depth, !from_schedule);

	current_ctx = blk_mq_get_ctx(q);

	if (!cpu_online(ctx->cpu))
		ctx = current_ctx;
	hctx = q->mq_ops->map_queue(q, ctx->cpu);

	/*
	 * preemption doesn't flush plug list, so it's possible ctx->cpu is
	 * offline now
	 */
	spin_lock(&ctx->lock);
	while (!list_empty(list)) {
		struct request *rq;

		rq = list_first_entry(list, struct request, queuelist);
		list_del_init(&rq->queuelist);
		rq->mq_ctx = ctx;
1004
		__blk_mq_insert_request(hctx, rq, false);
1005 1006 1007 1008
	}
	spin_unlock(&ctx->lock);

	blk_mq_run_hw_queue(hctx, from_schedule);
1009
	blk_mq_put_ctx(current_ctx);
1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071
}

static int plug_ctx_cmp(void *priv, struct list_head *a, struct list_head *b)
{
	struct request *rqa = container_of(a, struct request, queuelist);
	struct request *rqb = container_of(b, struct request, queuelist);

	return !(rqa->mq_ctx < rqb->mq_ctx ||
		 (rqa->mq_ctx == rqb->mq_ctx &&
		  blk_rq_pos(rqa) < blk_rq_pos(rqb)));
}

void blk_mq_flush_plug_list(struct blk_plug *plug, bool from_schedule)
{
	struct blk_mq_ctx *this_ctx;
	struct request_queue *this_q;
	struct request *rq;
	LIST_HEAD(list);
	LIST_HEAD(ctx_list);
	unsigned int depth;

	list_splice_init(&plug->mq_list, &list);

	list_sort(NULL, &list, plug_ctx_cmp);

	this_q = NULL;
	this_ctx = NULL;
	depth = 0;

	while (!list_empty(&list)) {
		rq = list_entry_rq(list.next);
		list_del_init(&rq->queuelist);
		BUG_ON(!rq->q);
		if (rq->mq_ctx != this_ctx) {
			if (this_ctx) {
				blk_mq_insert_requests(this_q, this_ctx,
							&ctx_list, depth,
							from_schedule);
			}

			this_ctx = rq->mq_ctx;
			this_q = rq->q;
			depth = 0;
		}

		depth++;
		list_add_tail(&rq->queuelist, &ctx_list);
	}

	/*
	 * If 'this_ctx' is set, we know we have entries to complete
	 * on 'ctx_list'. Do those.
	 */
	if (this_ctx) {
		blk_mq_insert_requests(this_q, this_ctx, &ctx_list, depth,
				       from_schedule);
	}
}

static void blk_mq_bio_to_request(struct request *rq, struct bio *bio)
{
	init_request_from_bio(rq, bio);
1072

1073
	if (blk_do_io_stat(rq))
1074
		blk_account_io_start(rq, 1);
1075 1076
}

1077 1078 1079 1080 1081 1082
static inline bool hctx_allow_merges(struct blk_mq_hw_ctx *hctx)
{
	return (hctx->flags & BLK_MQ_F_SHOULD_MERGE) &&
		!blk_queue_nomerges(hctx->queue);
}

1083 1084 1085
static inline bool blk_mq_merge_queue_io(struct blk_mq_hw_ctx *hctx,
					 struct blk_mq_ctx *ctx,
					 struct request *rq, struct bio *bio)
1086
{
1087
	if (!hctx_allow_merges(hctx)) {
1088 1089 1090 1091 1092 1093 1094
		blk_mq_bio_to_request(rq, bio);
		spin_lock(&ctx->lock);
insert_rq:
		__blk_mq_insert_request(hctx, rq, false);
		spin_unlock(&ctx->lock);
		return false;
	} else {
1095 1096
		struct request_queue *q = hctx->queue;

1097 1098 1099 1100 1101
		spin_lock(&ctx->lock);
		if (!blk_mq_attempt_merge(q, ctx, bio)) {
			blk_mq_bio_to_request(rq, bio);
			goto insert_rq;
		}
1102

1103 1104 1105
		spin_unlock(&ctx->lock);
		__blk_mq_free_request(hctx, ctx, rq);
		return true;
1106
	}
1107
}
1108

1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121
struct blk_map_ctx {
	struct blk_mq_hw_ctx *hctx;
	struct blk_mq_ctx *ctx;
};

static struct request *blk_mq_map_request(struct request_queue *q,
					  struct bio *bio,
					  struct blk_map_ctx *data)
{
	struct blk_mq_hw_ctx *hctx;
	struct blk_mq_ctx *ctx;
	struct request *rq;
	int rw = bio_data_dir(bio);
1122
	struct blk_mq_alloc_data alloc_data;
1123

1124
	if (unlikely(blk_mq_queue_enter(q))) {
1125
		bio_endio(bio, -EIO);
1126
		return NULL;
1127 1128 1129 1130 1131
	}

	ctx = blk_mq_get_ctx(q);
	hctx = q->mq_ops->map_queue(q, ctx->cpu);

1132
	if (rw_is_sync(bio->bi_rw))
S
Shaohua Li 已提交
1133
		rw |= REQ_SYNC;
1134

1135
	trace_block_getrq(q, bio, rw);
1136 1137 1138
	blk_mq_set_alloc_data(&alloc_data, q, GFP_ATOMIC, false, ctx,
			hctx);
	rq = __blk_mq_alloc_request(&alloc_data, rw);
1139
	if (unlikely(!rq)) {
1140
		__blk_mq_run_hw_queue(hctx);
1141 1142
		blk_mq_put_ctx(ctx);
		trace_block_sleeprq(q, bio, rw);
1143 1144

		ctx = blk_mq_get_ctx(q);
1145
		hctx = q->mq_ops->map_queue(q, ctx->cpu);
1146 1147 1148 1149 1150
		blk_mq_set_alloc_data(&alloc_data, q,
				__GFP_WAIT|GFP_ATOMIC, false, ctx, hctx);
		rq = __blk_mq_alloc_request(&alloc_data, rw);
		ctx = alloc_data.ctx;
		hctx = alloc_data.hctx;
1151 1152 1153
	}

	hctx->queued++;
1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256
	data->hctx = hctx;
	data->ctx = ctx;
	return rq;
}

/*
 * Multiple hardware queue variant. This will not use per-process plugs,
 * but will attempt to bypass the hctx queueing if we can go straight to
 * hardware for SYNC IO.
 */
static void blk_mq_make_request(struct request_queue *q, struct bio *bio)
{
	const int is_sync = rw_is_sync(bio->bi_rw);
	const int is_flush_fua = bio->bi_rw & (REQ_FLUSH | REQ_FUA);
	struct blk_map_ctx data;
	struct request *rq;

	blk_queue_bounce(q, &bio);

	if (bio_integrity_enabled(bio) && bio_integrity_prep(bio)) {
		bio_endio(bio, -EIO);
		return;
	}

	rq = blk_mq_map_request(q, bio, &data);
	if (unlikely(!rq))
		return;

	if (unlikely(is_flush_fua)) {
		blk_mq_bio_to_request(rq, bio);
		blk_insert_flush(rq);
		goto run_queue;
	}

	if (is_sync) {
		int ret;

		blk_mq_bio_to_request(rq, bio);
		blk_mq_start_request(rq, true);

		/*
		 * For OK queue, we are done. For error, kill it. Any other
		 * error (busy), just add it to our list as we previously
		 * would have done
		 */
		ret = q->mq_ops->queue_rq(data.hctx, rq);
		if (ret == BLK_MQ_RQ_QUEUE_OK)
			goto done;
		else {
			__blk_mq_requeue_request(rq);

			if (ret == BLK_MQ_RQ_QUEUE_ERROR) {
				rq->errors = -EIO;
				blk_mq_end_io(rq, rq->errors);
				goto done;
			}
		}
	}

	if (!blk_mq_merge_queue_io(data.hctx, data.ctx, rq, bio)) {
		/*
		 * For a SYNC request, send it to the hardware immediately. For
		 * an ASYNC request, just ensure that we run it later on. The
		 * latter allows for merging opportunities and more efficient
		 * dispatching.
		 */
run_queue:
		blk_mq_run_hw_queue(data.hctx, !is_sync || is_flush_fua);
	}
done:
	blk_mq_put_ctx(data.ctx);
}

/*
 * Single hardware queue variant. This will attempt to use any per-process
 * plug for merging and IO deferral.
 */
static void blk_sq_make_request(struct request_queue *q, struct bio *bio)
{
	const int is_sync = rw_is_sync(bio->bi_rw);
	const int is_flush_fua = bio->bi_rw & (REQ_FLUSH | REQ_FUA);
	unsigned int use_plug, request_count = 0;
	struct blk_map_ctx data;
	struct request *rq;

	/*
	 * If we have multiple hardware queues, just go directly to
	 * one of those for sync IO.
	 */
	use_plug = !is_flush_fua && !is_sync;

	blk_queue_bounce(q, &bio);

	if (bio_integrity_enabled(bio) && bio_integrity_prep(bio)) {
		bio_endio(bio, -EIO);
		return;
	}

	if (use_plug && !blk_queue_nomerges(q) &&
	    blk_attempt_plug_merge(q, bio, &request_count))
		return;

	rq = blk_mq_map_request(q, bio, &data);
1257 1258
	if (unlikely(!rq))
		return;
1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275

	if (unlikely(is_flush_fua)) {
		blk_mq_bio_to_request(rq, bio);
		blk_insert_flush(rq);
		goto run_queue;
	}

	/*
	 * A task plug currently exists. Since this is completely lockless,
	 * utilize that to temporarily store requests until the task is
	 * either done or scheduled away.
	 */
	if (use_plug) {
		struct blk_plug *plug = current->plug;

		if (plug) {
			blk_mq_bio_to_request(rq, bio);
S
Shaohua Li 已提交
1276
			if (list_empty(&plug->mq_list))
1277 1278 1279 1280 1281 1282
				trace_block_plug(q);
			else if (request_count >= BLK_MAX_REQUEST_COUNT) {
				blk_flush_plug_list(plug, false);
				trace_block_plug(q);
			}
			list_add_tail(&rq->queuelist, &plug->mq_list);
1283
			blk_mq_put_ctx(data.ctx);
1284 1285 1286 1287
			return;
		}
	}

1288 1289 1290 1291 1292 1293 1294 1295 1296
	if (!blk_mq_merge_queue_io(data.hctx, data.ctx, rq, bio)) {
		/*
		 * For a SYNC request, send it to the hardware immediately. For
		 * an ASYNC request, just ensure that we run it later on. The
		 * latter allows for merging opportunities and more efficient
		 * dispatching.
		 */
run_queue:
		blk_mq_run_hw_queue(data.hctx, !is_sync || is_flush_fua);
1297 1298
	}

1299
	blk_mq_put_ctx(data.ctx);
1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310
}

/*
 * Default mapping to a software queue, since we use one per CPU.
 */
struct blk_mq_hw_ctx *blk_mq_map_queue(struct request_queue *q, const int cpu)
{
	return q->queue_hw_ctx[q->mq_map[cpu]];
}
EXPORT_SYMBOL(blk_mq_map_queue);

1311 1312
static void blk_mq_free_rq_map(struct blk_mq_tag_set *set,
		struct blk_mq_tags *tags, unsigned int hctx_idx)
1313
{
1314
	struct page *page;
1315

1316
	if (tags->rqs && set->ops->exit_request) {
1317
		int i;
1318

1319 1320
		for (i = 0; i < tags->nr_tags; i++) {
			if (!tags->rqs[i])
1321
				continue;
1322 1323
			set->ops->exit_request(set->driver_data, tags->rqs[i],
						hctx_idx, i);
1324
		}
1325 1326
	}

1327 1328
	while (!list_empty(&tags->page_list)) {
		page = list_first_entry(&tags->page_list, struct page, lru);
1329
		list_del_init(&page->lru);
1330 1331 1332
		__free_pages(page, page->private);
	}

1333
	kfree(tags->rqs);
1334

1335
	blk_mq_free_tags(tags);
1336 1337 1338 1339
}

static size_t order_to_size(unsigned int order)
{
1340
	return (size_t)PAGE_SIZE << order;
1341 1342
}

1343 1344
static struct blk_mq_tags *blk_mq_init_rq_map(struct blk_mq_tag_set *set,
		unsigned int hctx_idx)
1345
{
1346
	struct blk_mq_tags *tags;
1347 1348 1349
	unsigned int i, j, entries_per_page, max_order = 4;
	size_t rq_size, left;

1350 1351 1352 1353
	tags = blk_mq_init_tags(set->queue_depth, set->reserved_tags,
				set->numa_node);
	if (!tags)
		return NULL;
1354

1355 1356 1357 1358 1359 1360 1361 1362
	INIT_LIST_HEAD(&tags->page_list);

	tags->rqs = kmalloc_node(set->queue_depth * sizeof(struct request *),
					GFP_KERNEL, set->numa_node);
	if (!tags->rqs) {
		blk_mq_free_tags(tags);
		return NULL;
	}
1363 1364 1365 1366 1367

	/*
	 * rq_size is the size of the request plus driver payload, rounded
	 * to the cacheline size
	 */
1368
	rq_size = round_up(sizeof(struct request) + set->cmd_size,
1369
				cache_line_size());
1370
	left = rq_size * set->queue_depth;
1371

1372
	for (i = 0; i < set->queue_depth; ) {
1373 1374 1375 1376 1377 1378 1379 1380 1381
		int this_order = max_order;
		struct page *page;
		int to_do;
		void *p;

		while (left < order_to_size(this_order - 1) && this_order)
			this_order--;

		do {
1382 1383
			page = alloc_pages_node(set->numa_node, GFP_KERNEL,
						this_order);
1384 1385 1386 1387 1388 1389 1390 1391 1392
			if (page)
				break;
			if (!this_order--)
				break;
			if (order_to_size(this_order) < rq_size)
				break;
		} while (1);

		if (!page)
1393
			goto fail;
1394 1395

		page->private = this_order;
1396
		list_add_tail(&page->lru, &tags->page_list);
1397 1398 1399

		p = page_address(page);
		entries_per_page = order_to_size(this_order) / rq_size;
1400
		to_do = min(entries_per_page, set->queue_depth - i);
1401 1402
		left -= to_do * rq_size;
		for (j = 0; j < to_do; j++) {
1403 1404 1405 1406 1407 1408
			tags->rqs[i] = p;
			if (set->ops->init_request) {
				if (set->ops->init_request(set->driver_data,
						tags->rqs[i], hctx_idx, i,
						set->numa_node))
					goto fail;
1409 1410
			}

1411 1412 1413 1414 1415
			p += rq_size;
			i++;
		}
	}

1416
	return tags;
1417

1418 1419 1420 1421
fail:
	pr_warn("%s: failed to allocate requests\n", __func__);
	blk_mq_free_rq_map(set, tags, hctx_idx);
	return NULL;
1422 1423
}

1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451
static void blk_mq_free_bitmap(struct blk_mq_ctxmap *bitmap)
{
	kfree(bitmap->map);
}

static int blk_mq_alloc_bitmap(struct blk_mq_ctxmap *bitmap, int node)
{
	unsigned int bpw = 8, total, num_maps, i;

	bitmap->bits_per_word = bpw;

	num_maps = ALIGN(nr_cpu_ids, bpw) / bpw;
	bitmap->map = kzalloc_node(num_maps * sizeof(struct blk_align_bitmap),
					GFP_KERNEL, node);
	if (!bitmap->map)
		return -ENOMEM;

	bitmap->map_size = num_maps;

	total = nr_cpu_ids;
	for (i = 0; i < num_maps; i++) {
		bitmap->map[i].depth = min(total, bitmap->bits_per_word);
		total -= bitmap->map[i].depth;
	}

	return 0;
}

1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522
static int blk_mq_hctx_cpu_offline(struct blk_mq_hw_ctx *hctx, int cpu)
{
	struct request_queue *q = hctx->queue;
	struct blk_mq_ctx *ctx;
	LIST_HEAD(tmp);

	/*
	 * Move ctx entries to new CPU, if this one is going away.
	 */
	ctx = __blk_mq_get_ctx(q, cpu);

	spin_lock(&ctx->lock);
	if (!list_empty(&ctx->rq_list)) {
		list_splice_init(&ctx->rq_list, &tmp);
		blk_mq_hctx_clear_pending(hctx, ctx);
	}
	spin_unlock(&ctx->lock);

	if (list_empty(&tmp))
		return NOTIFY_OK;

	ctx = blk_mq_get_ctx(q);
	spin_lock(&ctx->lock);

	while (!list_empty(&tmp)) {
		struct request *rq;

		rq = list_first_entry(&tmp, struct request, queuelist);
		rq->mq_ctx = ctx;
		list_move_tail(&rq->queuelist, &ctx->rq_list);
	}

	hctx = q->mq_ops->map_queue(q, ctx->cpu);
	blk_mq_hctx_mark_pending(hctx, ctx);

	spin_unlock(&ctx->lock);

	blk_mq_run_hw_queue(hctx, true);
	blk_mq_put_ctx(ctx);
	return NOTIFY_OK;
}

static int blk_mq_hctx_cpu_online(struct blk_mq_hw_ctx *hctx, int cpu)
{
	struct request_queue *q = hctx->queue;
	struct blk_mq_tag_set *set = q->tag_set;

	if (set->tags[hctx->queue_num])
		return NOTIFY_OK;

	set->tags[hctx->queue_num] = blk_mq_init_rq_map(set, hctx->queue_num);
	if (!set->tags[hctx->queue_num])
		return NOTIFY_STOP;

	hctx->tags = set->tags[hctx->queue_num];
	return NOTIFY_OK;
}

static int blk_mq_hctx_notify(void *data, unsigned long action,
			      unsigned int cpu)
{
	struct blk_mq_hw_ctx *hctx = data;

	if (action == CPU_DEAD || action == CPU_DEAD_FROZEN)
		return blk_mq_hctx_cpu_offline(hctx, cpu);
	else if (action == CPU_ONLINE || action == CPU_ONLINE_FROZEN)
		return blk_mq_hctx_cpu_online(hctx, cpu);

	return NOTIFY_OK;
}

M
Ming Lei 已提交
1523 1524 1525 1526 1527 1528 1529 1530 1531 1532
static void blk_mq_exit_hw_queues(struct request_queue *q,
		struct blk_mq_tag_set *set, int nr_queue)
{
	struct blk_mq_hw_ctx *hctx;
	unsigned int i;

	queue_for_each_hw_ctx(q, hctx, i) {
		if (i == nr_queue)
			break;

1533 1534
		blk_mq_tag_idle(hctx);

M
Ming Lei 已提交
1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552
		if (set->ops->exit_hctx)
			set->ops->exit_hctx(hctx, i);

		blk_mq_unregister_cpu_notifier(&hctx->cpu_notifier);
		kfree(hctx->ctxs);
		blk_mq_free_bitmap(&hctx->ctx_map);
	}

}

static void blk_mq_free_hw_queues(struct request_queue *q,
		struct blk_mq_tag_set *set)
{
	struct blk_mq_hw_ctx *hctx;
	unsigned int i;

	queue_for_each_hw_ctx(q, hctx, i) {
		free_cpumask_var(hctx->cpumask);
1553
		kfree(hctx);
M
Ming Lei 已提交
1554 1555 1556
	}
}

1557
static int blk_mq_init_hw_queues(struct request_queue *q,
1558
		struct blk_mq_tag_set *set)
1559 1560
{
	struct blk_mq_hw_ctx *hctx;
M
Ming Lei 已提交
1561
	unsigned int i;
1562 1563 1564 1565 1566 1567 1568 1569 1570

	/*
	 * Initialize hardware queues
	 */
	queue_for_each_hw_ctx(q, hctx, i) {
		int node;

		node = hctx->numa_node;
		if (node == NUMA_NO_NODE)
1571
			node = hctx->numa_node = set->numa_node;
1572

1573 1574
		INIT_DELAYED_WORK(&hctx->run_work, blk_mq_run_work_fn);
		INIT_DELAYED_WORK(&hctx->delay_work, blk_mq_delay_work_fn);
1575 1576 1577 1578
		spin_lock_init(&hctx->lock);
		INIT_LIST_HEAD(&hctx->dispatch);
		hctx->queue = q;
		hctx->queue_num = i;
1579 1580
		hctx->flags = set->flags;
		hctx->cmd_size = set->cmd_size;
1581 1582 1583 1584 1585

		blk_mq_init_cpu_notifier(&hctx->cpu_notifier,
						blk_mq_hctx_notify, hctx);
		blk_mq_register_cpu_notifier(&hctx->cpu_notifier);

1586
		hctx->tags = set->tags[i];
1587 1588

		/*
1589
		 * Allocate space for all possible cpus to avoid allocation at
1590 1591 1592 1593 1594 1595 1596
		 * runtime
		 */
		hctx->ctxs = kmalloc_node(nr_cpu_ids * sizeof(void *),
						GFP_KERNEL, node);
		if (!hctx->ctxs)
			break;

1597
		if (blk_mq_alloc_bitmap(&hctx->ctx_map, node))
1598 1599 1600 1601
			break;

		hctx->nr_ctx = 0;

1602 1603
		if (set->ops->init_hctx &&
		    set->ops->init_hctx(hctx, set->driver_data, i))
1604 1605 1606 1607 1608 1609 1610 1611 1612
			break;
	}

	if (i == q->nr_hw_queues)
		return 0;

	/*
	 * Init failed
	 */
M
Ming Lei 已提交
1613
	blk_mq_exit_hw_queues(q, set, i);
1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636

	return 1;
}

static void blk_mq_init_cpu_queues(struct request_queue *q,
				   unsigned int nr_hw_queues)
{
	unsigned int i;

	for_each_possible_cpu(i) {
		struct blk_mq_ctx *__ctx = per_cpu_ptr(q->queue_ctx, i);
		struct blk_mq_hw_ctx *hctx;

		memset(__ctx, 0, sizeof(*__ctx));
		__ctx->cpu = i;
		spin_lock_init(&__ctx->lock);
		INIT_LIST_HEAD(&__ctx->rq_list);
		__ctx->queue = q;

		/* If the cpu isn't online, the cpu is mapped to first hctx */
		if (!cpu_online(i))
			continue;

1637 1638 1639 1640
		hctx = q->mq_ops->map_queue(q, i);
		cpumask_set_cpu(i, hctx->cpumask);
		hctx->nr_ctx++;

1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656
		/*
		 * Set local node, IFF we have more than one hw queue. If
		 * not, we remain on the home node of the device
		 */
		if (nr_hw_queues > 1 && hctx->numa_node == NUMA_NO_NODE)
			hctx->numa_node = cpu_to_node(i);
	}
}

static void blk_mq_map_swqueue(struct request_queue *q)
{
	unsigned int i;
	struct blk_mq_hw_ctx *hctx;
	struct blk_mq_ctx *ctx;

	queue_for_each_hw_ctx(q, hctx, i) {
1657
		cpumask_clear(hctx->cpumask);
1658 1659 1660 1661 1662 1663 1664 1665
		hctx->nr_ctx = 0;
	}

	/*
	 * Map software to hardware queues
	 */
	queue_for_each_ctx(q, ctx, i) {
		/* If the cpu isn't online, the cpu is mapped to first hctx */
1666 1667 1668
		if (!cpu_online(i))
			continue;

1669
		hctx = q->mq_ops->map_queue(q, i);
1670
		cpumask_set_cpu(i, hctx->cpumask);
1671 1672 1673
		ctx->index_hw = hctx->nr_ctx;
		hctx->ctxs[hctx->nr_ctx++] = ctx;
	}
1674 1675

	queue_for_each_hw_ctx(q, hctx, i) {
1676
		/*
1677 1678
		 * If no software queues are mapped to this hardware queue,
		 * disable it and free the request entries.
1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693
		 */
		if (!hctx->nr_ctx) {
			struct blk_mq_tag_set *set = q->tag_set;

			if (set->tags[i]) {
				blk_mq_free_rq_map(set, set->tags[i], i);
				set->tags[i] = NULL;
				hctx->tags = NULL;
			}
			continue;
		}

		/*
		 * Initialize batch roundrobin counts
		 */
1694 1695 1696
		hctx->next_cpu = cpumask_first(hctx->cpumask);
		hctx->next_cpu_batch = BLK_MQ_CPU_WORK_BATCH;
	}
1697 1698
}

1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744
static void blk_mq_update_tag_set_depth(struct blk_mq_tag_set *set)
{
	struct blk_mq_hw_ctx *hctx;
	struct request_queue *q;
	bool shared;
	int i;

	if (set->tag_list.next == set->tag_list.prev)
		shared = false;
	else
		shared = true;

	list_for_each_entry(q, &set->tag_list, tag_set_list) {
		blk_mq_freeze_queue(q);

		queue_for_each_hw_ctx(q, hctx, i) {
			if (shared)
				hctx->flags |= BLK_MQ_F_TAG_SHARED;
			else
				hctx->flags &= ~BLK_MQ_F_TAG_SHARED;
		}
		blk_mq_unfreeze_queue(q);
	}
}

static void blk_mq_del_queue_tag_set(struct request_queue *q)
{
	struct blk_mq_tag_set *set = q->tag_set;

	mutex_lock(&set->tag_list_lock);
	list_del_init(&q->tag_set_list);
	blk_mq_update_tag_set_depth(set);
	mutex_unlock(&set->tag_list_lock);
}

static void blk_mq_add_queue_tag_set(struct blk_mq_tag_set *set,
				     struct request_queue *q)
{
	q->tag_set = set;

	mutex_lock(&set->tag_list_lock);
	list_add_tail(&q->tag_set_list, &set->tag_list);
	blk_mq_update_tag_set_depth(set);
	mutex_unlock(&set->tag_list_lock);
}

1745
struct request_queue *blk_mq_init_queue(struct blk_mq_tag_set *set)
1746 1747
{
	struct blk_mq_hw_ctx **hctxs;
1748
	struct blk_mq_ctx __percpu *ctx;
1749
	struct request_queue *q;
1750
	unsigned int *map;
1751 1752 1753 1754 1755 1756
	int i;

	ctx = alloc_percpu(struct blk_mq_ctx);
	if (!ctx)
		return ERR_PTR(-ENOMEM);

1757 1758
	hctxs = kmalloc_node(set->nr_hw_queues * sizeof(*hctxs), GFP_KERNEL,
			set->numa_node);
1759 1760 1761 1762

	if (!hctxs)
		goto err_percpu;

1763 1764 1765 1766
	map = blk_mq_make_queue_map(set);
	if (!map)
		goto err_map;

1767
	for (i = 0; i < set->nr_hw_queues; i++) {
1768 1769
		int node = blk_mq_hw_queue_to_node(map, i);

1770 1771
		hctxs[i] = kzalloc_node(sizeof(struct blk_mq_hw_ctx),
					GFP_KERNEL, node);
1772 1773 1774
		if (!hctxs[i])
			goto err_hctxs;

1775 1776 1777
		if (!zalloc_cpumask_var(&hctxs[i]->cpumask, GFP_KERNEL))
			goto err_hctxs;

1778
		atomic_set(&hctxs[i]->nr_active, 0);
1779
		hctxs[i]->numa_node = node;
1780 1781 1782
		hctxs[i]->queue_num = i;
	}

1783
	q = blk_alloc_queue_node(GFP_KERNEL, set->numa_node);
1784 1785 1786
	if (!q)
		goto err_hctxs;

1787
	if (percpu_ref_init(&q->mq_usage_counter, blk_mq_usage_counter_release))
1788 1789
		goto err_map;

1790 1791 1792 1793
	setup_timer(&q->timeout, blk_mq_rq_timer, (unsigned long) q);
	blk_queue_rq_timeout(q, 30000);

	q->nr_queues = nr_cpu_ids;
1794
	q->nr_hw_queues = set->nr_hw_queues;
1795
	q->mq_map = map;
1796 1797 1798 1799

	q->queue_ctx = ctx;
	q->queue_hw_ctx = hctxs;

1800
	q->mq_ops = set->ops;
1801
	q->queue_flags |= QUEUE_FLAG_MQ_DEFAULT;
1802

1803 1804 1805
	if (!(set->flags & BLK_MQ_F_SG_MERGE))
		q->queue_flags |= 1 << QUEUE_FLAG_NO_SG_MERGE;

1806 1807
	q->sg_reserved_size = INT_MAX;

1808 1809 1810 1811
	INIT_WORK(&q->requeue_work, blk_mq_requeue_work);
	INIT_LIST_HEAD(&q->requeue_list);
	spin_lock_init(&q->requeue_lock);

1812 1813 1814 1815 1816
	if (q->nr_hw_queues > 1)
		blk_queue_make_request(q, blk_mq_make_request);
	else
		blk_queue_make_request(q, blk_sq_make_request);

1817
	blk_queue_rq_timed_out(q, blk_mq_rq_timed_out);
1818 1819
	if (set->timeout)
		blk_queue_rq_timeout(q, set->timeout);
1820

1821 1822 1823 1824 1825
	/*
	 * Do this after blk_queue_make_request() overrides it...
	 */
	q->nr_requests = set->queue_depth;

1826 1827
	if (set->ops->complete)
		blk_queue_softirq_done(q, set->ops->complete);
1828

1829
	blk_mq_init_flush(q);
1830
	blk_mq_init_cpu_queues(q, set->nr_hw_queues);
1831

1832 1833 1834
	q->flush_rq = kzalloc(round_up(sizeof(struct request) +
				set->cmd_size, cache_line_size()),
				GFP_KERNEL);
1835
	if (!q->flush_rq)
1836 1837
		goto err_hw;

1838
	if (blk_mq_init_hw_queues(q, set))
1839 1840
		goto err_flush_rq;

1841 1842 1843 1844
	mutex_lock(&all_q_mutex);
	list_add_tail(&q->all_q_node, &all_q_list);
	mutex_unlock(&all_q_mutex);

1845 1846
	blk_mq_add_queue_tag_set(set, q);

1847 1848
	blk_mq_map_swqueue(q);

1849
	return q;
1850 1851 1852

err_flush_rq:
	kfree(q->flush_rq);
1853 1854 1855
err_hw:
	blk_cleanup_queue(q);
err_hctxs:
1856
	kfree(map);
1857
	for (i = 0; i < set->nr_hw_queues; i++) {
1858 1859
		if (!hctxs[i])
			break;
1860
		free_cpumask_var(hctxs[i]->cpumask);
1861
		kfree(hctxs[i]);
1862
	}
1863
err_map:
1864 1865 1866 1867 1868 1869 1870 1871 1872
	kfree(hctxs);
err_percpu:
	free_percpu(ctx);
	return ERR_PTR(-ENOMEM);
}
EXPORT_SYMBOL(blk_mq_init_queue);

void blk_mq_free_queue(struct request_queue *q)
{
M
Ming Lei 已提交
1873
	struct blk_mq_tag_set	*set = q->tag_set;
1874

1875 1876
	blk_mq_del_queue_tag_set(q);

M
Ming Lei 已提交
1877 1878
	blk_mq_exit_hw_queues(q, set, set->nr_hw_queues);
	blk_mq_free_hw_queues(q, set);
1879

1880
	percpu_ref_exit(&q->mq_usage_counter);
1881

1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895
	free_percpu(q->queue_ctx);
	kfree(q->queue_hw_ctx);
	kfree(q->mq_map);

	q->queue_ctx = NULL;
	q->queue_hw_ctx = NULL;
	q->mq_map = NULL;

	mutex_lock(&all_q_mutex);
	list_del_init(&q->all_q_node);
	mutex_unlock(&all_q_mutex);
}

/* Basically redo blk_mq_init_queue with queue frozen */
1896
static void blk_mq_queue_reinit(struct request_queue *q)
1897 1898 1899
{
	blk_mq_freeze_queue(q);

1900 1901
	blk_mq_sysfs_unregister(q);

1902 1903 1904 1905 1906 1907 1908 1909 1910 1911
	blk_mq_update_queue_map(q->mq_map, q->nr_hw_queues);

	/*
	 * redo blk_mq_init_cpu_queues and blk_mq_init_hw_queues. FIXME: maybe
	 * we should change hctx numa_node according to new topology (this
	 * involves free and re-allocate memory, worthy doing?)
	 */

	blk_mq_map_swqueue(q);

1912 1913
	blk_mq_sysfs_register(q);

1914 1915 1916
	blk_mq_unfreeze_queue(q);
}

1917 1918
static int blk_mq_queue_reinit_notify(struct notifier_block *nb,
				      unsigned long action, void *hcpu)
1919 1920 1921 1922
{
	struct request_queue *q;

	/*
1923 1924 1925 1926
	 * Before new mappings are established, hotadded cpu might already
	 * start handling requests. This doesn't break anything as we map
	 * offline CPUs to first hardware queue. We will re-init the queue
	 * below to get optimal settings.
1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938
	 */
	if (action != CPU_DEAD && action != CPU_DEAD_FROZEN &&
	    action != CPU_ONLINE && action != CPU_ONLINE_FROZEN)
		return NOTIFY_OK;

	mutex_lock(&all_q_mutex);
	list_for_each_entry(q, &all_q_list, all_q_node)
		blk_mq_queue_reinit(q);
	mutex_unlock(&all_q_mutex);
	return NOTIFY_OK;
}

1939 1940 1941 1942 1943 1944
/*
 * Alloc a tag set to be associated with one or more request queues.
 * May fail with EINVAL for various error conditions. May adjust the
 * requested depth down, if if it too large. In that case, the set
 * value will be stored in set->queue_depth.
 */
1945 1946 1947 1948 1949 1950
int blk_mq_alloc_tag_set(struct blk_mq_tag_set *set)
{
	int i;

	if (!set->nr_hw_queues)
		return -EINVAL;
1951
	if (!set->queue_depth)
1952 1953 1954 1955
		return -EINVAL;
	if (set->queue_depth < set->reserved_tags + BLK_MQ_TAG_MIN)
		return -EINVAL;

1956
	if (!set->nr_hw_queues || !set->ops->queue_rq || !set->ops->map_queue)
1957 1958
		return -EINVAL;

1959 1960 1961 1962 1963
	if (set->queue_depth > BLK_MQ_MAX_DEPTH) {
		pr_info("blk-mq: reduced tag depth to %u\n",
			BLK_MQ_MAX_DEPTH);
		set->queue_depth = BLK_MQ_MAX_DEPTH;
	}
1964

M
Ming Lei 已提交
1965 1966
	set->tags = kmalloc_node(set->nr_hw_queues *
				 sizeof(struct blk_mq_tags *),
1967 1968 1969 1970 1971 1972 1973 1974 1975 1976
				 GFP_KERNEL, set->numa_node);
	if (!set->tags)
		goto out;

	for (i = 0; i < set->nr_hw_queues; i++) {
		set->tags[i] = blk_mq_init_rq_map(set, i);
		if (!set->tags[i])
			goto out_unwind;
	}

1977 1978 1979
	mutex_init(&set->tag_list_lock);
	INIT_LIST_HEAD(&set->tag_list);

1980 1981 1982 1983 1984
	return 0;

out_unwind:
	while (--i >= 0)
		blk_mq_free_rq_map(set, set->tags[i], i);
1985 1986
	kfree(set->tags);
	set->tags = NULL;
1987 1988 1989 1990 1991 1992 1993 1994 1995
out:
	return -ENOMEM;
}
EXPORT_SYMBOL(blk_mq_alloc_tag_set);

void blk_mq_free_tag_set(struct blk_mq_tag_set *set)
{
	int i;

1996 1997 1998 1999 2000
	for (i = 0; i < set->nr_hw_queues; i++) {
		if (set->tags[i])
			blk_mq_free_rq_map(set, set->tags[i], i);
	}

M
Ming Lei 已提交
2001
	kfree(set->tags);
2002
	set->tags = NULL;
2003 2004 2005
}
EXPORT_SYMBOL(blk_mq_free_tag_set);

2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027
int blk_mq_update_nr_requests(struct request_queue *q, unsigned int nr)
{
	struct blk_mq_tag_set *set = q->tag_set;
	struct blk_mq_hw_ctx *hctx;
	int i, ret;

	if (!set || nr > set->queue_depth)
		return -EINVAL;

	ret = 0;
	queue_for_each_hw_ctx(q, hctx, i) {
		ret = blk_mq_tag_update_depth(hctx->tags, nr);
		if (ret)
			break;
	}

	if (!ret)
		q->nr_requests = nr;

	return ret;
}

2028 2029 2030 2031 2032 2033 2034 2035 2036 2037
void blk_mq_disable_hotplug(void)
{
	mutex_lock(&all_q_mutex);
}

void blk_mq_enable_hotplug(void)
{
	mutex_unlock(&all_q_mutex);
}

2038 2039 2040 2041
static int __init blk_mq_init(void)
{
	blk_mq_cpu_init();

2042
	hotcpu_notifier(blk_mq_queue_reinit_notify, 0);
2043 2044 2045 2046

	return 0;
}
subsys_initcall(blk_mq_init);