blk-mq-tag.c 15.0 KB
Newer Older
1
// SPDX-License-Identifier: GPL-2.0
2
/*
3 4 5
 * Tag allocation using scalable bitmaps. Uses active queue tracking to support
 * fairer distribution of tags between multiple submitters when a shared tag map
 * is used.
6 7 8
 *
 * Copyright (C) 2013-2014 Jens Axboe
 */
9 10 11 12
#include <linux/kernel.h>
#include <linux/module.h>

#include <linux/blk-mq.h>
13
#include <linux/delay.h>
14 15 16 17
#include "blk.h"
#include "blk-mq.h"
#include "blk-mq-tag.h"

18 19
/*
 * If a previously inactive queue goes active, bump the active user count.
20 21 22
 * We need to do this before try to allocate driver tag, then even if fail
 * to get tag when first time, the other shared-tag users could reserve
 * budget for it.
23 24 25 26 27 28 29 30 31 32 33
 */
bool __blk_mq_tag_busy(struct blk_mq_hw_ctx *hctx)
{
	if (!test_bit(BLK_MQ_S_TAG_ACTIVE, &hctx->state) &&
	    !test_and_set_bit(BLK_MQ_S_TAG_ACTIVE, &hctx->state))
		atomic_inc(&hctx->tags->active_queues);

	return true;
}

/*
34
 * Wakeup all potentially sleeping on tags
35
 */
36
void blk_mq_tag_wakeup_all(struct blk_mq_tags *tags, bool include_reserve)
37
{
38 39 40
	sbitmap_queue_wake_all(&tags->bitmap_tags);
	if (include_reserve)
		sbitmap_queue_wake_all(&tags->breserved_tags);
41 42
}

43 44 45 46 47 48 49 50 51 52 53 54 55
/*
 * If a previously busy queue goes inactive, potential waiters could now
 * be allowed to queue. Wake them up and check.
 */
void __blk_mq_tag_idle(struct blk_mq_hw_ctx *hctx)
{
	struct blk_mq_tags *tags = hctx->tags;

	if (!test_and_clear_bit(BLK_MQ_S_TAG_ACTIVE, &hctx->state))
		return;

	atomic_dec(&tags->active_queues);

56
	blk_mq_tag_wakeup_all(tags, false);
57 58
}

59 60
static int __blk_mq_get_tag(struct blk_mq_alloc_data *data,
			    struct sbitmap_queue *bt)
61
{
62
	if (!data->q->elevator && !hctx_may_queue(data->hctx, bt))
63
		return BLK_MQ_NO_TAG;
64

65 66 67 68
	if (data->shallow_depth)
		return __sbitmap_queue_get_shallow(bt, data->shallow_depth);
	else
		return __sbitmap_queue_get(bt);
69 70
}

71
unsigned int blk_mq_get_tag(struct blk_mq_alloc_data *data)
72
{
73 74
	struct blk_mq_tags *tags = blk_mq_tags_from_data(data);
	struct sbitmap_queue *bt;
75
	struct sbq_wait_state *ws;
J
Jens Axboe 已提交
76
	DEFINE_SBQ_WAIT(wait);
77
	unsigned int tag_offset;
78 79
	int tag;

80 81 82
	if (data->flags & BLK_MQ_REQ_RESERVED) {
		if (unlikely(!tags->nr_reserved_tags)) {
			WARN_ON_ONCE(1);
83
			return BLK_MQ_NO_TAG;
84 85 86 87 88 89 90 91
		}
		bt = &tags->breserved_tags;
		tag_offset = 0;
	} else {
		bt = &tags->bitmap_tags;
		tag_offset = tags->nr_reserved_tags;
	}

92
	tag = __blk_mq_get_tag(data, bt);
93
	if (tag != BLK_MQ_NO_TAG)
94
		goto found_tag;
95

96
	if (data->flags & BLK_MQ_REQ_NOWAIT)
97
		return BLK_MQ_NO_TAG;
98

99
	ws = bt_wait_ptr(bt, data->hctx);
100
	do {
101 102
		struct sbitmap_queue *bt_prev;

B
Bart Van Assche 已提交
103 104 105
		/*
		 * We're out of tags on this hardware queue, kick any
		 * pending IO submits before going to sleep waiting for
106
		 * some to complete.
B
Bart Van Assche 已提交
107
		 */
108
		blk_mq_run_hw_queue(data->hctx, false);
B
Bart Van Assche 已提交
109

110 111 112 113
		/*
		 * Retry tag allocation after running the hardware queue,
		 * as running the queue may also have found completions.
		 */
114
		tag = __blk_mq_get_tag(data, bt);
115
		if (tag != BLK_MQ_NO_TAG)
116 117
			break;

J
Jens Axboe 已提交
118
		sbitmap_prepare_to_wait(bt, ws, &wait, TASK_UNINTERRUPTIBLE);
119 120

		tag = __blk_mq_get_tag(data, bt);
121
		if (tag != BLK_MQ_NO_TAG)
122 123
			break;

124
		bt_prev = bt;
125
		io_schedule();
126

J
Jens Axboe 已提交
127 128
		sbitmap_finish_wait(bt, ws, &wait);

129
		data->ctx = blk_mq_get_ctx(data->q);
130
		data->hctx = blk_mq_map_queue(data->q, data->cmd_flags,
131
						data->ctx);
132 133 134 135 136 137
		tags = blk_mq_tags_from_data(data);
		if (data->flags & BLK_MQ_REQ_RESERVED)
			bt = &tags->breserved_tags;
		else
			bt = &tags->bitmap_tags;

138 139 140 141 142 143 144 145
		/*
		 * If destination hw queue is changed, fake wake up on
		 * previous queue for compensating the wake up miss, so
		 * other allocations on previous queue won't be starved.
		 */
		if (bt != bt_prev)
			sbitmap_queue_wake_up(bt_prev);

146
		ws = bt_wait_ptr(bt, data->hctx);
147 148
	} while (1);

J
Jens Axboe 已提交
149
	sbitmap_finish_wait(bt, ws, &wait);
150

151
found_tag:
152 153 154 155 156 157 158 159
	/*
	 * Give up this allocation if the hctx is inactive.  The caller will
	 * retry on an active hctx.
	 */
	if (unlikely(test_bit(BLK_MQ_S_INACTIVE, &data->hctx->state))) {
		blk_mq_put_tag(tags, data->ctx, tag + tag_offset);
		return BLK_MQ_NO_TAG;
	}
160
	return tag + tag_offset;
161 162
}

163 164
void blk_mq_put_tag(struct blk_mq_tags *tags, struct blk_mq_ctx *ctx,
		    unsigned int tag)
165
{
166
	if (!blk_mq_tag_is_reserved(tags, tag)) {
167 168
		const int real_tag = tag - tags->nr_reserved_tags;

J
Jens Axboe 已提交
169
		BUG_ON(real_tag >= tags->nr_tags);
170
		sbitmap_queue_clear(&tags->bitmap_tags, real_tag, ctx->cpu);
J
Jens Axboe 已提交
171 172
	} else {
		BUG_ON(tag >= tags->nr_reserved_tags);
173
		sbitmap_queue_clear(&tags->breserved_tags, tag, ctx->cpu);
J
Jens Axboe 已提交
174
	}
175 176
}

177 178 179 180 181 182 183 184
struct bt_iter_data {
	struct blk_mq_hw_ctx *hctx;
	busy_iter_fn *fn;
	void *data;
	bool reserved;
};

static bool bt_iter(struct sbitmap *bitmap, unsigned int bitnr, void *data)
185
{
186 187 188 189
	struct bt_iter_data *iter_data = data;
	struct blk_mq_hw_ctx *hctx = iter_data->hctx;
	struct blk_mq_tags *tags = hctx->tags;
	bool reserved = iter_data->reserved;
190
	struct request *rq;
191

192 193 194
	if (!reserved)
		bitnr += tags->nr_reserved_tags;
	rq = tags->rqs[bitnr];
195

196 197
	/*
	 * We can hit rq == NULL here, because the tagging functions
198
	 * test and set the bit before assigning ->rqs[].
199 200
	 */
	if (rq && rq->q == hctx->queue)
201
		return iter_data->fn(hctx, rq, iter_data->data, reserved);
202 203
	return true;
}
204

205 206 207 208 209 210 211 212
/**
 * bt_for_each - iterate over the requests associated with a hardware queue
 * @hctx:	Hardware queue to examine.
 * @bt:		sbitmap to examine. This is either the breserved_tags member
 *		or the bitmap_tags member of struct blk_mq_tags.
 * @fn:		Pointer to the function that will be called for each request
 *		associated with @hctx that has been assigned a driver tag.
 *		@fn will be called as follows: @fn(@hctx, rq, @data, @reserved)
213 214
 *		where rq is a pointer to a request. Return true to continue
 *		iterating tags, false to stop.
215 216 217 218
 * @data:	Will be passed as third argument to @fn.
 * @reserved:	Indicates whether @bt is the breserved_tags member or the
 *		bitmap_tags member of struct blk_mq_tags.
 */
219 220 221 222 223 224 225 226 227 228 229
static void bt_for_each(struct blk_mq_hw_ctx *hctx, struct sbitmap_queue *bt,
			busy_iter_fn *fn, void *data, bool reserved)
{
	struct bt_iter_data iter_data = {
		.hctx = hctx,
		.fn = fn,
		.data = data,
		.reserved = reserved,
	};

	sbitmap_for_each_set(&bt->sb, bt_iter, &iter_data);
230 231
}

232 233 234 235
struct bt_tags_iter_data {
	struct blk_mq_tags *tags;
	busy_tag_iter_fn *fn;
	void *data;
M
Ming Lei 已提交
236
	unsigned int flags;
237 238
};

M
Ming Lei 已提交
239 240
#define BT_TAG_ITER_RESERVED		(1 << 0)
#define BT_TAG_ITER_STARTED		(1 << 1)
M
Ming Lei 已提交
241
#define BT_TAG_ITER_STATIC_RQS		(1 << 2)
M
Ming Lei 已提交
242

243
static bool bt_tags_iter(struct sbitmap *bitmap, unsigned int bitnr, void *data)
K
Keith Busch 已提交
244
{
245 246
	struct bt_tags_iter_data *iter_data = data;
	struct blk_mq_tags *tags = iter_data->tags;
M
Ming Lei 已提交
247
	bool reserved = iter_data->flags & BT_TAG_ITER_RESERVED;
K
Keith Busch 已提交
248 249
	struct request *rq;

250 251
	if (!reserved)
		bitnr += tags->nr_reserved_tags;
252 253 254

	/*
	 * We can hit rq == NULL here, because the tagging functions
M
Ming Lei 已提交
255
	 * test and set the bit before assigning ->rqs[].
256
	 */
M
Ming Lei 已提交
257 258 259 260
	if (iter_data->flags & BT_TAG_ITER_STATIC_RQS)
		rq = tags->static_rqs[bitnr];
	else
		rq = tags->rqs[bitnr];
M
Ming Lei 已提交
261 262 263 264 265 266
	if (!rq)
		return true;
	if ((iter_data->flags & BT_TAG_ITER_STARTED) &&
	    !blk_mq_request_started(rq))
		return true;
	return iter_data->fn(rq, iter_data->data, reserved);
267 268
}

269 270 271 272 273 274 275
/**
 * bt_tags_for_each - iterate over the requests in a tag map
 * @tags:	Tag map to iterate over.
 * @bt:		sbitmap to examine. This is either the breserved_tags member
 *		or the bitmap_tags member of struct blk_mq_tags.
 * @fn:		Pointer to the function that will be called for each started
 *		request. @fn will be called as follows: @fn(rq, @data,
276 277
 *		@reserved) where rq is a pointer to a request. Return true
 *		to continue iterating tags, false to stop.
278
 * @data:	Will be passed as second argument to @fn.
M
Ming Lei 已提交
279
 * @flags:	BT_TAG_ITER_*
280
 */
281
static void bt_tags_for_each(struct blk_mq_tags *tags, struct sbitmap_queue *bt,
M
Ming Lei 已提交
282
			     busy_tag_iter_fn *fn, void *data, unsigned int flags)
283 284 285 286 287
{
	struct bt_tags_iter_data iter_data = {
		.tags = tags,
		.fn = fn,
		.data = data,
M
Ming Lei 已提交
288
		.flags = flags,
289 290 291 292
	};

	if (tags->rqs)
		sbitmap_for_each_set(&bt->sb, bt_tags_iter, &iter_data);
K
Keith Busch 已提交
293 294
}

M
Ming Lei 已提交
295 296 297 298 299 300 301 302 303 304 305
static void __blk_mq_all_tag_iter(struct blk_mq_tags *tags,
		busy_tag_iter_fn *fn, void *priv, unsigned int flags)
{
	WARN_ON_ONCE(flags & BT_TAG_ITER_RESERVED);

	if (tags->nr_reserved_tags)
		bt_tags_for_each(tags, &tags->breserved_tags, fn, priv,
				 flags | BT_TAG_ITER_RESERVED);
	bt_tags_for_each(tags, &tags->bitmap_tags, fn, priv, flags);
}

306
/**
M
Ming Lei 已提交
307
 * blk_mq_all_tag_iter - iterate over all requests in a tag map
308
 * @tags:	Tag map to iterate over.
M
Ming Lei 已提交
309
 * @fn:		Pointer to the function that will be called for each
310 311
 *		request. @fn will be called as follows: @fn(rq, @priv,
 *		reserved) where rq is a pointer to a request. 'reserved'
312 313
 *		indicates whether or not @rq is a reserved request. Return
 *		true to continue iterating tags, false to stop.
314
 * @priv:	Will be passed as second argument to @fn.
M
Ming Lei 已提交
315 316
 *
 * Caller has to pass the tag map from which requests are allocated.
317
 */
M
Ming Lei 已提交
318 319
void blk_mq_all_tag_iter(struct blk_mq_tags *tags, busy_tag_iter_fn *fn,
		void *priv)
K
Keith Busch 已提交
320
{
321
	__blk_mq_all_tag_iter(tags, fn, priv, BT_TAG_ITER_STATIC_RQS);
K
Keith Busch 已提交
322 323
}

324 325 326 327 328 329
/**
 * blk_mq_tagset_busy_iter - iterate over all started requests in a tag set
 * @tagset:	Tag set to iterate over.
 * @fn:		Pointer to the function that will be called for each started
 *		request. @fn will be called as follows: @fn(rq, @priv,
 *		reserved) where rq is a pointer to a request. 'reserved'
330 331
 *		indicates whether or not @rq is a reserved request. Return
 *		true to continue iterating tags, false to stop.
332 333
 * @priv:	Will be passed as second argument to @fn.
 */
334 335 336 337 338 339 340
void blk_mq_tagset_busy_iter(struct blk_mq_tag_set *tagset,
		busy_tag_iter_fn *fn, void *priv)
{
	int i;

	for (i = 0; i < tagset->nr_hw_queues; i++) {
		if (tagset->tags && tagset->tags[i])
M
Ming Lei 已提交
341 342
			__blk_mq_all_tag_iter(tagset->tags[i], fn, priv,
					      BT_TAG_ITER_STARTED);
343 344 345 346
	}
}
EXPORT_SYMBOL(blk_mq_tagset_busy_iter);

347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377
static bool blk_mq_tagset_count_completed_rqs(struct request *rq,
		void *data, bool reserved)
{
	unsigned *count = data;

	if (blk_mq_request_completed(rq))
		(*count)++;
	return true;
}

/**
 * blk_mq_tagset_wait_completed_request - wait until all completed req's
 * complete funtion is run
 * @tagset:	Tag set to drain completed request
 *
 * Note: This function has to be run after all IO queues are shutdown
 */
void blk_mq_tagset_wait_completed_request(struct blk_mq_tag_set *tagset)
{
	while (true) {
		unsigned count = 0;

		blk_mq_tagset_busy_iter(tagset,
				blk_mq_tagset_count_completed_rqs, &count);
		if (!count)
			break;
		msleep(5);
	}
}
EXPORT_SYMBOL(blk_mq_tagset_wait_completed_request);

378 379 380 381 382 383 384 385 386 387 388 389 390 391
/**
 * blk_mq_queue_tag_busy_iter - iterate over all requests with a driver tag
 * @q:		Request queue to examine.
 * @fn:		Pointer to the function that will be called for each request
 *		on @q. @fn will be called as follows: @fn(hctx, rq, @priv,
 *		reserved) where rq is a pointer to a request and hctx points
 *		to the hardware queue associated with the request. 'reserved'
 *		indicates whether or not @rq is a reserved request.
 * @priv:	Will be passed as third argument to @fn.
 *
 * Note: if @q->tag_set is shared with other request queues then @fn will be
 * called for all requests on all queues that share that tag set and not only
 * for requests associated with @q.
 */
392
void blk_mq_queue_tag_busy_iter(struct request_queue *q, busy_iter_fn *fn,
393
		void *priv)
394
{
395 396 397
	struct blk_mq_hw_ctx *hctx;
	int i;

398
	/*
399 400 401 402 403
	 * __blk_mq_update_nr_hw_queues() updates nr_hw_queues and queue_hw_ctx
	 * while the queue is frozen. So we can use q_usage_counter to avoid
	 * racing with it. __blk_mq_update_nr_hw_queues() uses
	 * synchronize_rcu() to ensure this function left the critical section
	 * below.
404
	 */
405
	if (!percpu_ref_tryget(&q->q_usage_counter))
406
		return;
407 408 409 410 411

	queue_for_each_hw_ctx(q, hctx, i) {
		struct blk_mq_tags *tags = hctx->tags;

		/*
412
		 * If no software queues are currently mapped to this
413 414 415 416 417 418
		 * hardware queue, there's nothing to check
		 */
		if (!blk_mq_hw_queue_mapped(hctx))
			continue;

		if (tags->nr_reserved_tags)
419 420
			bt_for_each(hctx, &tags->breserved_tags, fn, priv, true);
		bt_for_each(hctx, &tags->bitmap_tags, fn, priv, false);
421
	}
422
	blk_queue_exit(q);
423 424
}

425 426
static int bt_alloc(struct sbitmap_queue *bt, unsigned int depth,
		    bool round_robin, int node)
427
{
428 429
	return sbitmap_queue_init_node(bt, depth, -1, round_robin, GFP_KERNEL,
				       node);
430 431
}

432 433
static int blk_mq_init_bitmap_tags(struct blk_mq_tags *tags,
				   int node, int alloc_policy)
434 435
{
	unsigned int depth = tags->nr_tags - tags->nr_reserved_tags;
436
	bool round_robin = alloc_policy == BLK_TAG_ALLOC_RR;
437

438
	if (bt_alloc(&tags->bitmap_tags, depth, round_robin, node))
439
		return -ENOMEM;
440 441
	if (bt_alloc(&tags->breserved_tags, tags->nr_reserved_tags, round_robin,
		     node))
442
		goto free_bitmap_tags;
443

444
	return 0;
445 446
free_bitmap_tags:
	sbitmap_queue_free(&tags->bitmap_tags);
447
	return -ENOMEM;
448 449
}

450
struct blk_mq_tags *blk_mq_init_tags(unsigned int total_tags,
S
Shaohua Li 已提交
451 452
				     unsigned int reserved_tags,
				     int node, int alloc_policy)
453 454 455 456 457 458 459 460 461 462 463 464 465 466 467
{
	struct blk_mq_tags *tags;

	if (total_tags > BLK_MQ_TAG_MAX) {
		pr_err("blk-mq: tag depth too large\n");
		return NULL;
	}

	tags = kzalloc_node(sizeof(*tags), GFP_KERNEL, node);
	if (!tags)
		return NULL;

	tags->nr_tags = total_tags;
	tags->nr_reserved_tags = reserved_tags;

468 469 470 471 472
	if (blk_mq_init_bitmap_tags(tags, node, alloc_policy) < 0) {
		kfree(tags);
		return NULL;
	}
	return tags;
473 474 475 476
}

void blk_mq_free_tags(struct blk_mq_tags *tags)
{
477 478
	sbitmap_queue_free(&tags->bitmap_tags);
	sbitmap_queue_free(&tags->breserved_tags);
479 480 481
	kfree(tags);
}

482 483 484
int blk_mq_tag_update_depth(struct blk_mq_hw_ctx *hctx,
			    struct blk_mq_tags **tagsptr, unsigned int tdepth,
			    bool can_grow)
485
{
486 487 488
	struct blk_mq_tags *tags = *tagsptr;

	if (tdepth <= tags->nr_reserved_tags)
489 490 491
		return -EINVAL;

	/*
492 493
	 * If we are allowed to grow beyond the original size, allocate
	 * a new set of tags before freeing the old one.
494
	 */
495 496 497 498 499 500 501 502 503 504 505 506 507 508 509
	if (tdepth > tags->nr_tags) {
		struct blk_mq_tag_set *set = hctx->queue->tag_set;
		struct blk_mq_tags *new;
		bool ret;

		if (!can_grow)
			return -EINVAL;

		/*
		 * We need some sort of upper limit, set it high enough that
		 * no valid use cases should require more.
		 */
		if (tdepth > 16 * BLKDEV_MAX_RQ)
			return -EINVAL;

M
Ming Lei 已提交
510 511
		new = blk_mq_alloc_rq_map(set, hctx->queue_num, tdepth,
				tags->nr_reserved_tags);
512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527
		if (!new)
			return -ENOMEM;
		ret = blk_mq_alloc_rqs(set, new, hctx->queue_num, tdepth);
		if (ret) {
			blk_mq_free_rq_map(new);
			return -ENOMEM;
		}

		blk_mq_free_rqs(set, *tagsptr, hctx->queue_num);
		blk_mq_free_rq_map(*tagsptr);
		*tagsptr = new;
	} else {
		/*
		 * Don't need (or can't) update reserved tags here, they
		 * remain static and should never need resizing.
		 */
M
Ming Lei 已提交
528 529
		sbitmap_queue_resize(&tags->bitmap_tags,
				tdepth - tags->nr_reserved_tags);
530
	}
531

532 533 534
	return 0;
}

B
Bart Van Assche 已提交
535 536 537 538 539 540 541 542 543 544 545 546 547 548
/**
 * blk_mq_unique_tag() - return a tag that is unique queue-wide
 * @rq: request for which to compute a unique tag
 *
 * The tag field in struct request is unique per hardware queue but not over
 * all hardware queues. Hence this function that returns a tag with the
 * hardware context index in the upper bits and the per hardware queue tag in
 * the lower bits.
 *
 * Note: When called for a request that is queued on a non-multiqueue request
 * queue, the hardware context index is set to zero.
 */
u32 blk_mq_unique_tag(struct request *rq)
{
549
	return (rq->mq_hctx->queue_num << BLK_MQ_UNIQUE_TAG_BITS) |
B
Bart Van Assche 已提交
550 551 552
		(rq->tag & BLK_MQ_UNIQUE_TAG_MASK);
}
EXPORT_SYMBOL(blk_mq_unique_tag);