[mirror_ubuntu-artful-kernel.git] / block / blk-mq-tag.c

/*
 * 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.
 *
 * Copyright (C) 2013-2014 Jens Axboe
 */
#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)
{
	if (!tags)
		return true;

	return sbitmap_any_bit_clear(&tags->bitmap_tags.sb);
}

/*
 * If a previously inactive queue goes active, bump the active user count.
 */
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;
}

/*
 * Wakeup all potentially sleeping on tags
 */
void blk_mq_tag_wakeup_all(struct blk_mq_tags *tags, bool include_reserve)
{
	sbitmap_queue_wake_all(&tags->bitmap_tags);
	if (include_reserve)
		sbitmap_queue_wake_all(&tags->breserved_tags);
}

/*
 * 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);

	blk_mq_tag_wakeup_all(tags, false);
}

/*
 * 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,
				  struct sbitmap_queue *bt)
{
	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
	 */
	if (bt->sb.depth == 1)
		return true;

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

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

static int __blk_mq_get_tag(struct blk_mq_alloc_data *data,
			    struct sbitmap_queue *bt)
{
	if (!(data->flags & BLK_MQ_REQ_INTERNAL) &&
	    !hctx_may_queue(data->hctx, bt))
		return -1;
	return __sbitmap_queue_get(bt);
}

unsigned int blk_mq_get_tag(struct blk_mq_alloc_data *data)
{
	struct blk_mq_tags *tags = blk_mq_tags_from_data(data);
	struct sbitmap_queue *bt;
	struct sbq_wait_state *ws;
	DEFINE_WAIT(wait);
	unsigned int tag_offset;
	bool drop_ctx;
	int tag;

	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;
	}

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

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

	ws = bt_wait_ptr(bt, data->hctx);
	drop_ctx = data->ctx == NULL;
	do {
		prepare_to_wait(&ws->wait, &wait, TASK_UNINTERRUPTIBLE);

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

		/*
		 * We're out of tags on this hardware queue, kick any
		 * pending IO submits before going to sleep waiting for
		 * some to complete.
		 */
		blk_mq_run_hw_queue(data->hctx, false);

		/*
		 * Retry tag allocation after running the hardware queue,
		 * as running the queue may also have found completions.
		 */
		tag = __blk_mq_get_tag(data, bt);
		if (tag != -1)
			break;

		if (data->ctx)
			blk_mq_put_ctx(data->ctx);

		io_schedule();

		data->ctx = blk_mq_get_ctx(data->q);
		data->hctx = blk_mq_map_queue(data->q, data->ctx->cpu);
		tags = blk_mq_tags_from_data(data);
		if (data->flags & BLK_MQ_REQ_RESERVED)
			bt = &tags->breserved_tags;
		else
			bt = &tags->bitmap_tags;

		finish_wait(&ws->wait, &wait);
		ws = bt_wait_ptr(bt, data->hctx);
	} while (1);

	if (drop_ctx && data->ctx)
		blk_mq_put_ctx(data->ctx);

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

found_tag:
	return tag + tag_offset;
}

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

		BUG_ON(real_tag >= tags->nr_tags);
		sbitmap_queue_clear(&tags->bitmap_tags, real_tag, ctx->cpu);
	} else {
		BUG_ON(tag >= tags->nr_reserved_tags);
		sbitmap_queue_clear(&tags->breserved_tags, tag, ctx->cpu);
	}
}

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)
{
	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;
	struct request *rq;

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

	if (rq->q == hctx->queue)
		iter_data->fn(hctx, rq, iter_data->data, reserved);
	return true;
}

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);
}

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)
{
	struct bt_tags_iter_data *iter_data = data;
	struct blk_mq_tags *tags = iter_data->tags;
	bool reserved = iter_data->reserved;
	struct request *rq;

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

	iter_data->fn(rq, iter_data->data, reserved);
	return true;
}

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);
}

static void blk_mq_all_tag_busy_iter(struct blk_mq_tags *tags,
		busy_tag_iter_fn *fn, void *priv)
{
	if (tags->nr_reserved_tags)
		bt_tags_for_each(tags, &tags->breserved_tags, fn, priv, true);
	bt_tags_for_each(tags, &tags->bitmap_tags, fn, priv, false);
}

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);

int blk_mq_reinit_tagset(struct blk_mq_tag_set *set)
{
	int i, j, ret = 0;

	if (!set->ops->reinit_request)
		goto out;

	for (i = 0; i < set->nr_hw_queues; i++) {
		struct blk_mq_tags *tags = set->tags[i];

		if (!tags)
			continue;

		for (j = 0; j < tags->nr_tags; j++) {
			if (!tags->static_rqs[j])
				continue;

			ret = set->ops->reinit_request(set->driver_data,
						tags->static_rqs[j]);
			if (ret)
				goto out;
		}
	}

out:
	return ret;
}
EXPORT_SYMBOL_GPL(blk_mq_reinit_tagset);

void blk_mq_queue_tag_busy_iter(struct request_queue *q, busy_iter_fn *fn,
		void *priv)
{
	struct blk_mq_hw_ctx *hctx;
	int i;


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

		/*
		 * If not software queues are currently mapped to this
		 * hardware queue, there's nothing to check
		 */
		if (!blk_mq_hw_queue_mapped(hctx))
			continue;

		if (tags->nr_reserved_tags)
			bt_for_each(hctx, &tags->breserved_tags, fn, priv, true);
		bt_for_each(hctx, &tags->bitmap_tags, fn, priv, false);
	}

}

static int bt_alloc(struct sbitmap_queue *bt, unsigned int depth,
		    bool round_robin, int node)
{
	return sbitmap_queue_init_node(bt, depth, -1, round_robin, GFP_KERNEL,
				       node);
}

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

	if (bt_alloc(&tags->bitmap_tags, depth, round_robin, node))
		goto free_tags;
	if (bt_alloc(&tags->breserved_tags, tags->nr_reserved_tags, round_robin,
		     node))
		goto free_bitmap_tags;

	return tags;
free_bitmap_tags:
	sbitmap_queue_free(&tags->bitmap_tags);
free_tags:
	kfree(tags);
	return NULL;
}

struct blk_mq_tags *blk_mq_init_tags(unsigned int total_tags,
				     unsigned int reserved_tags,
				     int node, int alloc_policy)
{
	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;

	return blk_mq_init_bitmap_tags(tags, node, alloc_policy);
}

void blk_mq_free_tags(struct blk_mq_tags *tags)
{
	sbitmap_queue_free(&tags->bitmap_tags);
	sbitmap_queue_free(&tags->breserved_tags);
	kfree(tags);
}

int blk_mq_tag_update_depth(struct blk_mq_hw_ctx *hctx,
			    struct blk_mq_tags **tagsptr, unsigned int tdepth,
			    bool can_grow)
{
	struct blk_mq_tags *tags = *tagsptr;

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

	tdepth -= tags->nr_reserved_tags;

	/*
	 * If we are allowed to grow beyond the original size, allocate
	 * a new set of tags before freeing the old one.
	 */
	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;

		new = blk_mq_alloc_rq_map(set, hctx->queue_num, tdepth, 0);
		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.
		 */
		sbitmap_queue_resize(&tags->bitmap_tags, tdepth);
	}

	return 0;
}

/**
 * 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)
{
	struct request_queue *q = rq->q;
	struct blk_mq_hw_ctx *hctx;
	int hwq = 0;

	if (q->mq_ops) {
		hctx = blk_mq_map_queue(q, rq->mq_ctx->cpu);
		hwq = hctx->queue_num;
	}

	return (hwq << BLK_MQ_UNIQUE_TAG_BITS) |
		(rq->tag & BLK_MQ_UNIQUE_TAG_MASK);
}
EXPORT_SYMBOL(blk_mq_unique_tag);
Commit	Line	Data
75bb4625	1	/*
88459642 OS	2	* Tag allocation using scalable bitmaps. Uses active queue tracking to support
	3	* fairer distribution of tags between multiple submitters when a shared tag map
	4	* is used.
75bb4625 JA	5	*
	6	* Copyright (C) 2013-2014 Jens Axboe
	7	*/
320ae51f JA	8	#include <linux/kernel.h>
320ae51f JA	9	#include <linux/module.h>
320ae51f JA	10
	11	#include <linux/blk-mq.h>
	12	#include "blk.h"
	13	#include "blk-mq.h"
	14	#include "blk-mq-tag.h"
	15
320ae51f JA	16	bool blk_mq_has_free_tags(struct blk_mq_tags *tags)
320ae51f JA	17	{
4bb659b1 JA	18	if (!tags)
	19	return true;
	20
88459642	21	return sbitmap_any_bit_clear(&tags->bitmap_tags.sb);
0d2602ca JA	22	}
	23
	24	/*
	25	* If a previously inactive queue goes active, bump the active user count.
	26	*/
	27	bool __blk_mq_tag_busy(struct blk_mq_hw_ctx *hctx)
	28	{
	29	if (!test_bit(BLK_MQ_S_TAG_ACTIVE, &hctx->state) &&
	30	!test_and_set_bit(BLK_MQ_S_TAG_ACTIVE, &hctx->state))
	31	atomic_inc(&hctx->tags->active_queues);
	32
	33	return true;
	34	}
	35
	36	/*
aed3ea94	37	* Wakeup all potentially sleeping on tags
0d2602ca	38	*/
aed3ea94	39	void blk_mq_tag_wakeup_all(struct blk_mq_tags *tags, bool include_reserve)
0d2602ca	40	{
88459642 OS	41	sbitmap_queue_wake_all(&tags->bitmap_tags);
	42	if (include_reserve)
	43	sbitmap_queue_wake_all(&tags->breserved_tags);
0d2602ca JA	44	}
0d2602ca JA	45
e3a2b3f9 JA	46	/*
	47	* If a previously busy queue goes inactive, potential waiters could now
	48	* be allowed to queue. Wake them up and check.
	49	*/
	50	void __blk_mq_tag_idle(struct blk_mq_hw_ctx *hctx)
	51	{
	52	struct blk_mq_tags *tags = hctx->tags;
	53
	54	if (!test_and_clear_bit(BLK_MQ_S_TAG_ACTIVE, &hctx->state))
	55	return;
	56
	57	atomic_dec(&tags->active_queues);
	58
aed3ea94	59	blk_mq_tag_wakeup_all(tags, false);
e3a2b3f9 JA	60	}
e3a2b3f9 JA	61
0d2602ca JA	62	/*
	63	* For shared tag users, we track the number of currently active users
	64	* and attempt to provide a fair share of the tag depth for each of them.
	65	*/
	66	static inline bool hctx_may_queue(struct blk_mq_hw_ctx *hctx,
88459642	67	struct sbitmap_queue *bt)
0d2602ca JA	68	{
	69	unsigned int depth, users;
	70
	71	if (!hctx \|\| !(hctx->flags & BLK_MQ_F_TAG_SHARED))
	72	return true;
	73	if (!test_bit(BLK_MQ_S_TAG_ACTIVE, &hctx->state))
	74	return true;
	75
	76	/*
	77	* Don't try dividing an ant
	78	*/
88459642	79	if (bt->sb.depth == 1)
0d2602ca JA	80	return true;
	81
	82	users = atomic_read(&hctx->tags->active_queues);
	83	if (!users)
	84	return true;
	85
	86	/*
	87	* Allow at least some tags
	88	*/
88459642	89	depth = max((bt->sb.depth + users - 1) / users, 4U);
0d2602ca JA	90	return atomic_read(&hctx->nr_active) < depth;
	91	}
	92
200e86b3 JA	93	static int __blk_mq_get_tag(struct blk_mq_alloc_data *data,
200e86b3 JA	94	struct sbitmap_queue *bt)
4bb659b1	95	{
200e86b3 JA	96	if (!(data->flags & BLK_MQ_REQ_INTERNAL) &&
200e86b3 JA	97	!hctx_may_queue(data->hctx, bt))
0d2602ca	98	return -1;
f4a644db	99	return __sbitmap_queue_get(bt);
4bb659b1 JA	100	}
4bb659b1 JA	101
4941115b	102	unsigned int blk_mq_get_tag(struct blk_mq_alloc_data *data)
320ae51f	103	{
4941115b JA	104	struct blk_mq_tags *tags = blk_mq_tags_from_data(data);
4941115b JA	105	struct sbitmap_queue *bt;
88459642	106	struct sbq_wait_state *ws;
4bb659b1	107	DEFINE_WAIT(wait);
4941115b	108	unsigned int tag_offset;
bd6737f1	109	bool drop_ctx;
320ae51f JA	110	int tag;
320ae51f JA	111
4941115b JA	112	if (data->flags & BLK_MQ_REQ_RESERVED) {
	113	if (unlikely(!tags->nr_reserved_tags)) {
	114	WARN_ON_ONCE(1);
	115	return BLK_MQ_TAG_FAIL;
	116	}
	117	bt = &tags->breserved_tags;
	118	tag_offset = 0;
	119	} else {
	120	bt = &tags->bitmap_tags;
	121	tag_offset = tags->nr_reserved_tags;
	122	}
	123
200e86b3	124	tag = __blk_mq_get_tag(data, bt);
4bb659b1	125	if (tag != -1)
4941115b	126	goto found_tag;
4bb659b1	127
6f3b0e8b	128	if (data->flags & BLK_MQ_REQ_NOWAIT)
4941115b	129	return BLK_MQ_TAG_FAIL;
4bb659b1	130
4941115b	131	ws = bt_wait_ptr(bt, data->hctx);
bd6737f1	132	drop_ctx = data->ctx == NULL;
4bb659b1	133	do {
88459642	134	prepare_to_wait(&ws->wait, &wait, TASK_UNINTERRUPTIBLE);
4bb659b1	135
200e86b3	136	tag = __blk_mq_get_tag(data, bt);
4bb659b1 JA	137	if (tag != -1)
	138	break;
	139
b3223207 BVA	140	/*
	141	* We're out of tags on this hardware queue, kick any
	142	* pending IO submits before going to sleep waiting for
8cecb07d	143	* some to complete.
b3223207	144	*/
8cecb07d	145	blk_mq_run_hw_queue(data->hctx, false);
b3223207	146
080ff351 JA	147	/*
	148	* Retry tag allocation after running the hardware queue,
	149	* as running the queue may also have found completions.
	150	*/
200e86b3	151	tag = __blk_mq_get_tag(data, bt);
080ff351 JA	152	if (tag != -1)
	153	break;
	154
bd6737f1 JA	155	if (data->ctx)
bd6737f1 JA	156	blk_mq_put_ctx(data->ctx);
cb96a42c	157
4bb659b1	158	io_schedule();
cb96a42c ML	159
cb96a42c ML	160	data->ctx = blk_mq_get_ctx(data->q);
7d7e0f90	161	data->hctx = blk_mq_map_queue(data->q, data->ctx->cpu);
4941115b JA	162	tags = blk_mq_tags_from_data(data);
	163	if (data->flags & BLK_MQ_REQ_RESERVED)
	164	bt = &tags->breserved_tags;
	165	else
	166	bt = &tags->bitmap_tags;
	167
88459642	168	finish_wait(&ws->wait, &wait);
4941115b	169	ws = bt_wait_ptr(bt, data->hctx);
4bb659b1 JA	170	} while (1);
4bb659b1 JA	171
bd6737f1 JA	172	if (drop_ctx && data->ctx)
	173	blk_mq_put_ctx(data->ctx);
	174
88459642	175	finish_wait(&ws->wait, &wait);
320ae51f	176
4941115b JA	177	found_tag:
4941115b JA	178	return tag + tag_offset;
320ae51f JA	179	}
320ae51f JA	180
4941115b JA	181	void blk_mq_put_tag(struct blk_mq_hw_ctx hctx, struct blk_mq_tags tags,
4941115b JA	182	struct blk_mq_ctx *ctx, unsigned int tag)
320ae51f	183	{
415b806d	184	if (!blk_mq_tag_is_reserved(tags, tag)) {
4bb659b1 JA	185	const int real_tag = tag - tags->nr_reserved_tags;
4bb659b1 JA	186
70114c39	187	BUG_ON(real_tag >= tags->nr_tags);
f4a644db	188	sbitmap_queue_clear(&tags->bitmap_tags, real_tag, ctx->cpu);
70114c39 JA	189	} else {
70114c39 JA	190	BUG_ON(tag >= tags->nr_reserved_tags);
f4a644db	191	sbitmap_queue_clear(&tags->breserved_tags, tag, ctx->cpu);
70114c39	192	}
320ae51f JA	193	}
320ae51f JA	194
88459642 OS	195	struct bt_iter_data {
	196	struct blk_mq_hw_ctx *hctx;
	197	busy_iter_fn *fn;
	198	void *data;
	199	bool reserved;
	200	};
	201
	202	static bool bt_iter(struct sbitmap bitmap, unsigned int bitnr, void data)
320ae51f	203	{
88459642 OS	204	struct bt_iter_data *iter_data = data;
	205	struct blk_mq_hw_ctx *hctx = iter_data->hctx;
	206	struct blk_mq_tags *tags = hctx->tags;
	207	bool reserved = iter_data->reserved;
81481eb4	208	struct request *rq;
4bb659b1	209
88459642 OS	210	if (!reserved)
	211	bitnr += tags->nr_reserved_tags;
	212	rq = tags->rqs[bitnr];
4bb659b1	213
88459642 OS	214	if (rq->q == hctx->queue)
	215	iter_data->fn(hctx, rq, iter_data->data, reserved);
	216	return true;
	217	}
4bb659b1	218
88459642 OS	219	static void bt_for_each(struct blk_mq_hw_ctx hctx, struct sbitmap_queue bt,
	220	busy_iter_fn fn, void data, bool reserved)
	221	{
	222	struct bt_iter_data iter_data = {
	223	.hctx = hctx,
	224	.fn = fn,
	225	.data = data,
	226	.reserved = reserved,
	227	};
	228
	229	sbitmap_for_each_set(&bt->sb, bt_iter, &iter_data);
320ae51f JA	230	}
320ae51f JA	231
88459642 OS	232	struct bt_tags_iter_data {
	233	struct blk_mq_tags *tags;
	234	busy_tag_iter_fn *fn;
	235	void *data;
	236	bool reserved;
	237	};
	238
	239	static bool bt_tags_iter(struct sbitmap bitmap, unsigned int bitnr, void data)
f26cdc85	240	{
88459642 OS	241	struct bt_tags_iter_data *iter_data = data;
	242	struct blk_mq_tags *tags = iter_data->tags;
	243	bool reserved = iter_data->reserved;
f26cdc85	244	struct request *rq;
f26cdc85	245
88459642 OS	246	if (!reserved)
	247	bitnr += tags->nr_reserved_tags;
	248	rq = tags->rqs[bitnr];
f26cdc85	249
88459642 OS	250	iter_data->fn(rq, iter_data->data, reserved);
	251	return true;
	252	}
	253
	254	static void bt_tags_for_each(struct blk_mq_tags tags, struct sbitmap_queue bt,
	255	busy_tag_iter_fn fn, void data, bool reserved)
	256	{
	257	struct bt_tags_iter_data iter_data = {
	258	.tags = tags,
	259	.fn = fn,
	260	.data = data,
	261	.reserved = reserved,
	262	};
	263
	264	if (tags->rqs)
	265	sbitmap_for_each_set(&bt->sb, bt_tags_iter, &iter_data);
f26cdc85 KB	266	}
f26cdc85 KB	267
e8f1e163 SG	268	static void blk_mq_all_tag_busy_iter(struct blk_mq_tags *tags,
e8f1e163 SG	269	busy_tag_iter_fn fn, void priv)
f26cdc85 KB	270	{
f26cdc85 KB	271	if (tags->nr_reserved_tags)
88459642 OS	272	bt_tags_for_each(tags, &tags->breserved_tags, fn, priv, true);
88459642 OS	273	bt_tags_for_each(tags, &tags->bitmap_tags, fn, priv, false);
f26cdc85	274	}
f26cdc85	275
e0489487 SG	276	void blk_mq_tagset_busy_iter(struct blk_mq_tag_set *tagset,
	277	busy_tag_iter_fn fn, void priv)
	278	{
	279	int i;
	280
	281	for (i = 0; i < tagset->nr_hw_queues; i++) {
	282	if (tagset->tags && tagset->tags[i])
	283	blk_mq_all_tag_busy_iter(tagset->tags[i], fn, priv);
	284	}
	285	}
	286	EXPORT_SYMBOL(blk_mq_tagset_busy_iter);
	287
486cf989 SG	288	int blk_mq_reinit_tagset(struct blk_mq_tag_set *set)
	289	{
	290	int i, j, ret = 0;
	291
	292	if (!set->ops->reinit_request)
	293	goto out;
	294
	295	for (i = 0; i < set->nr_hw_queues; i++) {
	296	struct blk_mq_tags *tags = set->tags[i];
	297
0067d4b0 SG	298	if (!tags)
	299	continue;
	300
486cf989	301	for (j = 0; j < tags->nr_tags; j++) {
2af8cbe3	302	if (!tags->static_rqs[j])
486cf989 SG	303	continue;
	304
	305	ret = set->ops->reinit_request(set->driver_data,
2af8cbe3	306	tags->static_rqs[j]);
486cf989 SG	307	if (ret)
	308	goto out;
	309	}
	310	}
	311
	312	out:
	313	return ret;
	314	}
	315	EXPORT_SYMBOL_GPL(blk_mq_reinit_tagset);
	316
0bf6cd5b	317	void blk_mq_queue_tag_busy_iter(struct request_queue q, busy_iter_fn fn,
81481eb4	318	void *priv)
320ae51f	319	{
0bf6cd5b CH	320	struct blk_mq_hw_ctx *hctx;
	321	int i;
	322
	323
	324	queue_for_each_hw_ctx(q, hctx, i) {
	325	struct blk_mq_tags *tags = hctx->tags;
	326
	327	/*
	328	* If not software queues are currently mapped to this
	329	* hardware queue, there's nothing to check
	330	*/
	331	if (!blk_mq_hw_queue_mapped(hctx))
	332	continue;
	333
	334	if (tags->nr_reserved_tags)
88459642 OS	335	bt_for_each(hctx, &tags->breserved_tags, fn, priv, true);
88459642 OS	336	bt_for_each(hctx, &tags->bitmap_tags, fn, priv, false);
4bb659b1 JA	337	}
4bb659b1 JA	338
4bb659b1 JA	339	}
4bb659b1 JA	340
f4a644db OS	341	static int bt_alloc(struct sbitmap_queue *bt, unsigned int depth,
f4a644db OS	342	bool round_robin, int node)
4bb659b1	343	{
f4a644db OS	344	return sbitmap_queue_init_node(bt, depth, -1, round_robin, GFP_KERNEL,
f4a644db OS	345	node);
4bb659b1 JA	346	}
	347
	348	static struct blk_mq_tags blk_mq_init_bitmap_tags(struct blk_mq_tags tags,
24391c0d	349	int node, int alloc_policy)
4bb659b1 JA	350	{
4bb659b1 JA	351	unsigned int depth = tags->nr_tags - tags->nr_reserved_tags;
f4a644db	352	bool round_robin = alloc_policy == BLK_TAG_ALLOC_RR;
4bb659b1	353
f4a644db	354	if (bt_alloc(&tags->bitmap_tags, depth, round_robin, node))
88459642	355	goto free_tags;
f4a644db OS	356	if (bt_alloc(&tags->breserved_tags, tags->nr_reserved_tags, round_robin,
f4a644db OS	357	node))
88459642	358	goto free_bitmap_tags;
4bb659b1 JA	359
4bb659b1 JA	360	return tags;
88459642 OS	361	free_bitmap_tags:
	362	sbitmap_queue_free(&tags->bitmap_tags);
	363	free_tags:
4bb659b1 JA	364	kfree(tags);
	365	return NULL;
	366	}
	367
320ae51f	368	struct blk_mq_tags *blk_mq_init_tags(unsigned int total_tags,
24391c0d SL	369	unsigned int reserved_tags,
24391c0d SL	370	int node, int alloc_policy)
320ae51f	371	{
320ae51f	372	struct blk_mq_tags *tags;
320ae51f JA	373
	374	if (total_tags > BLK_MQ_TAG_MAX) {
	375	pr_err("blk-mq: tag depth too large\n");
	376	return NULL;
	377	}
	378
	379	tags = kzalloc_node(sizeof(*tags), GFP_KERNEL, node);
	380	if (!tags)
	381	return NULL;
	382
320ae51f JA	383	tags->nr_tags = total_tags;
320ae51f JA	384	tags->nr_reserved_tags = reserved_tags;
320ae51f	385
24391c0d	386	return blk_mq_init_bitmap_tags(tags, node, alloc_policy);
320ae51f JA	387	}
	388
	389	void blk_mq_free_tags(struct blk_mq_tags *tags)
	390	{
88459642 OS	391	sbitmap_queue_free(&tags->bitmap_tags);
88459642 OS	392	sbitmap_queue_free(&tags->breserved_tags);
320ae51f JA	393	kfree(tags);
	394	}
	395
70f36b60 JA	396	int blk_mq_tag_update_depth(struct blk_mq_hw_ctx *hctx,
	397	struct blk_mq_tags **tagsptr, unsigned int tdepth,
	398	bool can_grow)
e3a2b3f9	399	{
70f36b60 JA	400	struct blk_mq_tags tags = tagsptr;
	401
	402	if (tdepth <= tags->nr_reserved_tags)
e3a2b3f9 JA	403	return -EINVAL;
e3a2b3f9 JA	404
70f36b60 JA	405	tdepth -= tags->nr_reserved_tags;
70f36b60 JA	406
e3a2b3f9	407	/*
70f36b60 JA	408	* If we are allowed to grow beyond the original size, allocate
70f36b60 JA	409	* a new set of tags before freeing the old one.
e3a2b3f9	410	*/
70f36b60 JA	411	if (tdepth > tags->nr_tags) {
	412	struct blk_mq_tag_set *set = hctx->queue->tag_set;
	413	struct blk_mq_tags *new;
	414	bool ret;
	415
	416	if (!can_grow)
	417	return -EINVAL;
	418
	419	/*
	420	* We need some sort of upper limit, set it high enough that
	421	* no valid use cases should require more.
	422	*/
	423	if (tdepth > 16 * BLKDEV_MAX_RQ)
	424	return -EINVAL;
	425
	426	new = blk_mq_alloc_rq_map(set, hctx->queue_num, tdepth, 0);
	427	if (!new)
	428	return -ENOMEM;
	429	ret = blk_mq_alloc_rqs(set, new, hctx->queue_num, tdepth);
	430	if (ret) {
	431	blk_mq_free_rq_map(new);
	432	return -ENOMEM;
	433	}
	434
	435	blk_mq_free_rqs(set, *tagsptr, hctx->queue_num);
	436	blk_mq_free_rq_map(*tagsptr);
	437	*tagsptr = new;
	438	} else {
	439	/*
	440	* Don't need (or can't) update reserved tags here, they
	441	* remain static and should never need resizing.
	442	*/
	443	sbitmap_queue_resize(&tags->bitmap_tags, tdepth);
	444	}
88459642	445
e3a2b3f9 JA	446	return 0;
	447	}
	448
205fb5f5 BVA	449	/**
	450	* blk_mq_unique_tag() - return a tag that is unique queue-wide
	451	* @rq: request for which to compute a unique tag
	452	*
	453	* The tag field in struct request is unique per hardware queue but not over
	454	* all hardware queues. Hence this function that returns a tag with the
	455	* hardware context index in the upper bits and the per hardware queue tag in
	456	* the lower bits.
	457	*
	458	* Note: When called for a request that is queued on a non-multiqueue request
	459	* queue, the hardware context index is set to zero.
	460	*/
	461	u32 blk_mq_unique_tag(struct request *rq)
	462	{
	463	struct request_queue *q = rq->q;
	464	struct blk_mq_hw_ctx *hctx;
	465	int hwq = 0;
	466
	467	if (q->mq_ops) {
7d7e0f90	468	hctx = blk_mq_map_queue(q, rq->mq_ctx->cpu);
205fb5f5 BVA	469	hwq = hctx->queue_num;
	470	}
	471
	472	return (hwq << BLK_MQ_UNIQUE_TAG_BITS) \|
	473	(rq->tag & BLK_MQ_UNIQUE_TAG_MASK);
	474	}
	475	EXPORT_SYMBOL(blk_mq_unique_tag);