blk-mq-tag.c 14.3 KB
Newer Older
1
/*
2 3 4
 * 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.
5 6 7
 *
 * Copyright (C) 2013-2014 Jens Axboe
 */
8 9 10 11 12 13 14 15 16 17
#include <linux/kernel.h>
#include <linux/module.h>

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

bool blk_mq_has_free_tags(struct blk_mq_tags *tags)
{
18 19 20
	if (!tags)
		return true;

21
	return sbitmap_any_bit_clear(&tags->bitmap_tags.sb);
22 23 24 25
}

/*
 * If a previously inactive queue goes active, bump the active user count.
26 27 28
 * 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.
29 30 31 32 33 34 35 36 37 38 39
 */
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;
}

/*
40
 * Wakeup all potentially sleeping on tags
41
 */
42
void blk_mq_tag_wakeup_all(struct blk_mq_tags *tags, bool include_reserve)
43
{
44 45 46
	sbitmap_queue_wake_all(&tags->bitmap_tags);
	if (include_reserve)
		sbitmap_queue_wake_all(&tags->breserved_tags);
47 48
}

49 50 51 52 53 54 55 56 57 58 59 60 61
/*
 * 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);

62
	blk_mq_tag_wakeup_all(tags, false);
63 64
}

65 66 67 68 69
/*
 * For shared tag users, we track the number of currently active users
 * and attempt to provide a fair share of the tag depth for each of them.
 */
static inline bool hctx_may_queue(struct blk_mq_hw_ctx *hctx,
70
				  struct sbitmap_queue *bt)
71 72 73 74 75 76 77 78 79 80 81
{
	unsigned int depth, users;

	if (!hctx || !(hctx->flags & BLK_MQ_F_TAG_SHARED))
		return true;
	if (!test_bit(BLK_MQ_S_TAG_ACTIVE, &hctx->state))
		return true;

	/*
	 * Don't try dividing an ant
	 */
82
	if (bt->sb.depth == 1)
83 84 85 86 87 88 89 90 91
		return true;

	users = atomic_read(&hctx->tags->active_queues);
	if (!users)
		return true;

	/*
	 * Allow at least some tags
	 */
92
	depth = max((bt->sb.depth + users - 1) / users, 4U);
93 94 95
	return atomic_read(&hctx->nr_active) < depth;
}

96 97
static int __blk_mq_get_tag(struct blk_mq_alloc_data *data,
			    struct sbitmap_queue *bt)
98
{
99 100
	if (!(data->flags & BLK_MQ_REQ_INTERNAL) &&
	    !hctx_may_queue(data->hctx, bt))
101
		return -1;
102 103 104 105
	if (data->shallow_depth)
		return __sbitmap_queue_get_shallow(bt, data->shallow_depth);
	else
		return __sbitmap_queue_get(bt);
106 107
}

108
unsigned int blk_mq_get_tag(struct blk_mq_alloc_data *data)
109
{
110 111
	struct blk_mq_tags *tags = blk_mq_tags_from_data(data);
	struct sbitmap_queue *bt;
112
	struct sbq_wait_state *ws;
113
	DEFINE_WAIT(wait);
114
	unsigned int tag_offset;
115
	bool drop_ctx;
116 117
	int tag;

118 119 120 121 122 123 124 125 126 127 128 129
	if (data->flags & BLK_MQ_REQ_RESERVED) {
		if (unlikely(!tags->nr_reserved_tags)) {
			WARN_ON_ONCE(1);
			return BLK_MQ_TAG_FAIL;
		}
		bt = &tags->breserved_tags;
		tag_offset = 0;
	} else {
		bt = &tags->bitmap_tags;
		tag_offset = tags->nr_reserved_tags;
	}

130
	tag = __blk_mq_get_tag(data, bt);
131
	if (tag != -1)
132
		goto found_tag;
133

134
	if (data->flags & BLK_MQ_REQ_NOWAIT)
135
		return BLK_MQ_TAG_FAIL;
136

137
	ws = bt_wait_ptr(bt, data->hctx);
138
	drop_ctx = data->ctx == NULL;
139
	do {
140 141
		struct sbitmap_queue *bt_prev;

B
Bart Van Assche 已提交
142 143 144
		/*
		 * We're out of tags on this hardware queue, kick any
		 * pending IO submits before going to sleep waiting for
145
		 * some to complete.
B
Bart Van Assche 已提交
146
		 */
147
		blk_mq_run_hw_queue(data->hctx, false);
B
Bart Van Assche 已提交
148

149 150 151 152
		/*
		 * Retry tag allocation after running the hardware queue,
		 * as running the queue may also have found completions.
		 */
153
		tag = __blk_mq_get_tag(data, bt);
154 155 156
		if (tag != -1)
			break;

157 158 159 160 161 162 163
		prepare_to_wait_exclusive(&ws->wait, &wait,
						TASK_UNINTERRUPTIBLE);

		tag = __blk_mq_get_tag(data, bt);
		if (tag != -1)
			break;

164 165
		if (data->ctx)
			blk_mq_put_ctx(data->ctx);
166

167
		bt_prev = bt;
168
		io_schedule();
169 170

		data->ctx = blk_mq_get_ctx(data->q);
171 172
		data->hctx = blk_mq_map_queue(data->q, data->cmd_flags,
						data->ctx->cpu);
173 174 175 176 177 178
		tags = blk_mq_tags_from_data(data);
		if (data->flags & BLK_MQ_REQ_RESERVED)
			bt = &tags->breserved_tags;
		else
			bt = &tags->bitmap_tags;

179
		finish_wait(&ws->wait, &wait);
180 181 182 183 184 185 186 187 188

		/*
		 * 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);

189
		ws = bt_wait_ptr(bt, data->hctx);
190 191
	} while (1);

192 193 194
	if (drop_ctx && data->ctx)
		blk_mq_put_ctx(data->ctx);

195
	finish_wait(&ws->wait, &wait);
196

197 198
found_tag:
	return tag + tag_offset;
199 200
}

201 202
void blk_mq_put_tag(struct blk_mq_hw_ctx *hctx, struct blk_mq_tags *tags,
		    struct blk_mq_ctx *ctx, unsigned int tag)
203
{
204
	if (!blk_mq_tag_is_reserved(tags, tag)) {
205 206
		const int real_tag = tag - tags->nr_reserved_tags;

J
Jens Axboe 已提交
207
		BUG_ON(real_tag >= tags->nr_tags);
208
		sbitmap_queue_clear(&tags->bitmap_tags, real_tag, ctx->cpu);
J
Jens Axboe 已提交
209 210
	} else {
		BUG_ON(tag >= tags->nr_reserved_tags);
211
		sbitmap_queue_clear(&tags->breserved_tags, tag, ctx->cpu);
J
Jens Axboe 已提交
212
	}
213 214
}

215 216 217 218 219 220 221 222
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)
223
{
224 225 226 227
	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;
228
	struct request *rq;
229

230 231 232
	if (!reserved)
		bitnr += tags->nr_reserved_tags;
	rq = tags->rqs[bitnr];
233

234 235
	/*
	 * We can hit rq == NULL here, because the tagging functions
236
	 * test and set the bit before assigning ->rqs[].
237 238
	 */
	if (rq && rq->q == hctx->queue)
239 240 241
		iter_data->fn(hctx, rq, iter_data->data, reserved);
	return true;
}
242

243 244 245 246 247 248 249 250 251 252 253 254 255
/**
 * 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)
 *		where rq is a pointer to a request.
 * @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.
 */
256 257 258 259 260 261 262 263 264 265 266
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);
267 268
}

269 270 271 272 273 274 275 276
struct bt_tags_iter_data {
	struct blk_mq_tags *tags;
	busy_tag_iter_fn *fn;
	void *data;
	bool reserved;
};

static bool bt_tags_iter(struct sbitmap *bitmap, unsigned int bitnr, void *data)
K
Keith Busch 已提交
277
{
278 279 280
	struct bt_tags_iter_data *iter_data = data;
	struct blk_mq_tags *tags = iter_data->tags;
	bool reserved = iter_data->reserved;
K
Keith Busch 已提交
281 282
	struct request *rq;

283 284
	if (!reserved)
		bitnr += tags->nr_reserved_tags;
285 286 287 288 289

	/*
	 * We can hit rq == NULL here, because the tagging functions
	 * test and set the bit before assining ->rqs[].
	 */
290
	rq = tags->rqs[bitnr];
M
Ming Lei 已提交
291
	if (rq && blk_mq_request_started(rq))
292
		iter_data->fn(rq, iter_data->data, reserved);
K
Keith Busch 已提交
293

294 295 296
	return true;
}

297 298 299 300 301 302 303 304 305 306 307 308
/**
 * 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,
 *		@reserved) where rq is a pointer to a request.
 * @data:	Will be passed as second argument to @fn.
 * @reserved:	Indicates whether @bt is the breserved_tags member or the
 *		bitmap_tags member of struct blk_mq_tags.
 */
309 310 311 312 313 314 315 316 317 318 319 320
static void bt_tags_for_each(struct blk_mq_tags *tags, struct sbitmap_queue *bt,
			     busy_tag_iter_fn *fn, void *data, bool reserved)
{
	struct bt_tags_iter_data iter_data = {
		.tags = tags,
		.fn = fn,
		.data = data,
		.reserved = reserved,
	};

	if (tags->rqs)
		sbitmap_for_each_set(&bt->sb, bt_tags_iter, &iter_data);
K
Keith Busch 已提交
321 322
}

323 324 325 326 327 328 329 330 331
/**
 * blk_mq_all_tag_busy_iter - iterate over all started requests in a tag map
 * @tags:	Tag map 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'
 *		indicates whether or not @rq is a reserved request.
 * @priv:	Will be passed as second argument to @fn.
 */
332 333
static void blk_mq_all_tag_busy_iter(struct blk_mq_tags *tags,
		busy_tag_iter_fn *fn, void *priv)
K
Keith Busch 已提交
334 335
{
	if (tags->nr_reserved_tags)
336 337
		bt_tags_for_each(tags, &tags->breserved_tags, fn, priv, true);
	bt_tags_for_each(tags, &tags->bitmap_tags, fn, priv, false);
K
Keith Busch 已提交
338 339
}

340 341 342 343 344 345 346 347 348
/**
 * 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'
 *		indicates whether or not @rq is a reserved request.
 * @priv:	Will be passed as second argument to @fn.
 */
349 350 351 352 353 354 355 356 357 358 359 360
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])
			blk_mq_all_tag_busy_iter(tagset->tags[i], fn, priv);
	}
}
EXPORT_SYMBOL(blk_mq_tagset_busy_iter);

361 362 363 364 365 366 367 368 369 370 371 372 373 374
/**
 * 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.
 */
375
void blk_mq_queue_tag_busy_iter(struct request_queue *q, busy_iter_fn *fn,
376
		void *priv)
377
{
378 379 380
	struct blk_mq_hw_ctx *hctx;
	int i;

381
	/*
382 383 384 385 386
	 * __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.
387
	 */
388
	if (!percpu_ref_tryget(&q->q_usage_counter))
389
		return;
390 391 392 393 394

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

		/*
395
		 * If no software queues are currently mapped to this
396 397 398 399 400 401
		 * hardware queue, there's nothing to check
		 */
		if (!blk_mq_hw_queue_mapped(hctx))
			continue;

		if (tags->nr_reserved_tags)
402 403
			bt_for_each(hctx, &tags->breserved_tags, fn, priv, true);
		bt_for_each(hctx, &tags->bitmap_tags, fn, priv, false);
404
	}
405
	blk_queue_exit(q);
406 407
}

408 409
static int bt_alloc(struct sbitmap_queue *bt, unsigned int depth,
		    bool round_robin, int node)
410
{
411 412
	return sbitmap_queue_init_node(bt, depth, -1, round_robin, GFP_KERNEL,
				       node);
413 414 415
}

static struct blk_mq_tags *blk_mq_init_bitmap_tags(struct blk_mq_tags *tags,
S
Shaohua Li 已提交
416
						   int node, int alloc_policy)
417 418
{
	unsigned int depth = tags->nr_tags - tags->nr_reserved_tags;
419
	bool round_robin = alloc_policy == BLK_TAG_ALLOC_RR;
420

421
	if (bt_alloc(&tags->bitmap_tags, depth, round_robin, node))
422
		goto free_tags;
423 424
	if (bt_alloc(&tags->breserved_tags, tags->nr_reserved_tags, round_robin,
		     node))
425
		goto free_bitmap_tags;
426 427

	return tags;
428 429 430
free_bitmap_tags:
	sbitmap_queue_free(&tags->bitmap_tags);
free_tags:
431 432 433 434
	kfree(tags);
	return NULL;
}

435
struct blk_mq_tags *blk_mq_init_tags(unsigned int total_tags,
S
Shaohua Li 已提交
436 437
				     unsigned int reserved_tags,
				     int node, int alloc_policy)
438 439 440 441 442 443 444 445 446 447 448 449 450 451 452
{
	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;

S
Shaohua Li 已提交
453
	return blk_mq_init_bitmap_tags(tags, node, alloc_policy);
454 455 456 457
}

void blk_mq_free_tags(struct blk_mq_tags *tags)
{
458 459
	sbitmap_queue_free(&tags->bitmap_tags);
	sbitmap_queue_free(&tags->breserved_tags);
460 461 462
	kfree(tags);
}

463 464 465
int blk_mq_tag_update_depth(struct blk_mq_hw_ctx *hctx,
			    struct blk_mq_tags **tagsptr, unsigned int tdepth,
			    bool can_grow)
466
{
467 468 469
	struct blk_mq_tags *tags = *tagsptr;

	if (tdepth <= tags->nr_reserved_tags)
470 471 472
		return -EINVAL;

	/*
473 474
	 * If we are allowed to grow beyond the original size, allocate
	 * a new set of tags before freeing the old one.
475
	 */
476 477 478 479 480 481 482 483 484 485 486 487 488 489 490
	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
491 492
		new = blk_mq_alloc_rq_map(set, hctx->queue_num, tdepth,
				tags->nr_reserved_tags);
493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508
		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
509 510
		sbitmap_queue_resize(&tags->bitmap_tags,
				tdepth - tags->nr_reserved_tags);
511
	}
512

513 514 515
	return 0;
}

B
Bart Van Assche 已提交
516 517 518 519 520 521 522 523 524 525 526 527 528 529
/**
 * 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)
{
530
	return (rq->mq_hctx->queue_num << BLK_MQ_UNIQUE_TAG_BITS) |
B
Bart Van Assche 已提交
531 532 533
		(rq->tag & BLK_MQ_UNIQUE_TAG_MASK);
}
EXPORT_SYMBOL(blk_mq_unique_tag);