[mirror_ubuntu-bionic-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_hw_ctx *hctx, struct sbitmap_queue *bt)
{
	if (!hctx_may_queue(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;
	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->hctx, 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);
	do {
		prepare_to_wait(&ws->wait, &wait, TASK_UNINTERRUPTIBLE);

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

		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);

	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 (tag >= tags->nr_reserved_tags) {
		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];

		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 unsigned int bt_unused_tags(const struct sbitmap_queue *bt)
{
	return bt->sb.depth - sbitmap_weight(&bt->sb);
}

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);

ssize_t blk_mq_tag_sysfs_show(struct blk_mq_tags *tags, char *page)
{
	char *orig_page = page;
	unsigned int free, res;

	if (!tags)
		return 0;

	page += sprintf(page, "nr_tags=%u, reserved_tags=%u, "
			"bits_per_word=%u\n",
			tags->nr_tags, tags->nr_reserved_tags,
			1U << tags->bitmap_tags.sb.shift);

	free = bt_unused_tags(&tags->bitmap_tags);
	res = bt_unused_tags(&tags->breserved_tags);

	page += sprintf(page, "nr_free=%u, nr_reserved=%u\n", free, res);
	page += sprintf(page, "active_queues=%u\n", atomic_read(&tags->active_queues));

	return page - orig_page;
}
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
4941115b	93	static int __blk_mq_get_tag(struct blk_mq_hw_ctx hctx, struct sbitmap_queue bt)
4bb659b1	94	{
0d2602ca JA	95	if (!hctx_may_queue(hctx, bt))
0d2602ca JA	96	return -1;
f4a644db	97	return __sbitmap_queue_get(bt);
4bb659b1 JA	98	}
4bb659b1 JA	99
4941115b	100	unsigned int blk_mq_get_tag(struct blk_mq_alloc_data *data)
320ae51f	101	{
4941115b JA	102	struct blk_mq_tags *tags = blk_mq_tags_from_data(data);
4941115b JA	103	struct sbitmap_queue *bt;
88459642	104	struct sbq_wait_state *ws;
4bb659b1	105	DEFINE_WAIT(wait);
4941115b	106	unsigned int tag_offset;
320ae51f JA	107	int tag;
320ae51f JA	108
4941115b JA	109	if (data->flags & BLK_MQ_REQ_RESERVED) {
	110	if (unlikely(!tags->nr_reserved_tags)) {
	111	WARN_ON_ONCE(1);
	112	return BLK_MQ_TAG_FAIL;
	113	}
	114	bt = &tags->breserved_tags;
	115	tag_offset = 0;
	116	} else {
	117	bt = &tags->bitmap_tags;
	118	tag_offset = tags->nr_reserved_tags;
	119	}
	120
	121	tag = __blk_mq_get_tag(data->hctx, bt);
4bb659b1	122	if (tag != -1)
4941115b	123	goto found_tag;
4bb659b1	124
6f3b0e8b	125	if (data->flags & BLK_MQ_REQ_NOWAIT)
4941115b	126	return BLK_MQ_TAG_FAIL;
4bb659b1	127
4941115b	128	ws = bt_wait_ptr(bt, data->hctx);
4bb659b1	129	do {
88459642	130	prepare_to_wait(&ws->wait, &wait, TASK_UNINTERRUPTIBLE);
4bb659b1	131
4941115b	132	tag = __blk_mq_get_tag(data->hctx, bt);
4bb659b1 JA	133	if (tag != -1)
	134	break;
	135
b3223207 BVA	136	/*
	137	* We're out of tags on this hardware queue, kick any
	138	* pending IO submits before going to sleep waiting for
8cecb07d	139	* some to complete.
b3223207	140	*/
8cecb07d	141	blk_mq_run_hw_queue(data->hctx, false);
b3223207	142
080ff351 JA	143	/*
	144	* Retry tag allocation after running the hardware queue,
	145	* as running the queue may also have found completions.
	146	*/
4941115b	147	tag = __blk_mq_get_tag(data->hctx, bt);
080ff351 JA	148	if (tag != -1)
	149	break;
	150
cb96a42c ML	151	blk_mq_put_ctx(data->ctx);
cb96a42c ML	152
4bb659b1	153	io_schedule();
cb96a42c ML	154
cb96a42c ML	155	data->ctx = blk_mq_get_ctx(data->q);
7d7e0f90	156	data->hctx = blk_mq_map_queue(data->q, data->ctx->cpu);
4941115b JA	157	tags = blk_mq_tags_from_data(data);
	158	if (data->flags & BLK_MQ_REQ_RESERVED)
	159	bt = &tags->breserved_tags;
	160	else
	161	bt = &tags->bitmap_tags;
	162
88459642	163	finish_wait(&ws->wait, &wait);
4941115b	164	ws = bt_wait_ptr(bt, data->hctx);
4bb659b1 JA	165	} while (1);
4bb659b1 JA	166
88459642	167	finish_wait(&ws->wait, &wait);
320ae51f	168
4941115b JA	169	found_tag:
4941115b JA	170	return tag + tag_offset;
320ae51f JA	171	}
320ae51f JA	172
4941115b JA	173	void blk_mq_put_tag(struct blk_mq_hw_ctx hctx, struct blk_mq_tags tags,
4941115b JA	174	struct blk_mq_ctx *ctx, unsigned int tag)
320ae51f	175	{
4bb659b1 JA	176	if (tag >= tags->nr_reserved_tags) {
	177	const int real_tag = tag - tags->nr_reserved_tags;
	178
70114c39	179	BUG_ON(real_tag >= tags->nr_tags);
f4a644db	180	sbitmap_queue_clear(&tags->bitmap_tags, real_tag, ctx->cpu);
70114c39 JA	181	} else {
70114c39 JA	182	BUG_ON(tag >= tags->nr_reserved_tags);
f4a644db	183	sbitmap_queue_clear(&tags->breserved_tags, tag, ctx->cpu);
70114c39	184	}
320ae51f JA	185	}
320ae51f JA	186
88459642 OS	187	struct bt_iter_data {
	188	struct blk_mq_hw_ctx *hctx;
	189	busy_iter_fn *fn;
	190	void *data;
	191	bool reserved;
	192	};
	193
	194	static bool bt_iter(struct sbitmap bitmap, unsigned int bitnr, void data)
320ae51f	195	{
88459642 OS	196	struct bt_iter_data *iter_data = data;
	197	struct blk_mq_hw_ctx *hctx = iter_data->hctx;
	198	struct blk_mq_tags *tags = hctx->tags;
	199	bool reserved = iter_data->reserved;
81481eb4	200	struct request *rq;
4bb659b1	201
88459642 OS	202	if (!reserved)
	203	bitnr += tags->nr_reserved_tags;
	204	rq = tags->rqs[bitnr];
4bb659b1	205
88459642 OS	206	if (rq->q == hctx->queue)
	207	iter_data->fn(hctx, rq, iter_data->data, reserved);
	208	return true;
	209	}
4bb659b1	210
88459642 OS	211	static void bt_for_each(struct blk_mq_hw_ctx hctx, struct sbitmap_queue bt,
	212	busy_iter_fn fn, void data, bool reserved)
	213	{
	214	struct bt_iter_data iter_data = {
	215	.hctx = hctx,
	216	.fn = fn,
	217	.data = data,
	218	.reserved = reserved,
	219	};
	220
	221	sbitmap_for_each_set(&bt->sb, bt_iter, &iter_data);
320ae51f JA	222	}
320ae51f JA	223
88459642 OS	224	struct bt_tags_iter_data {
	225	struct blk_mq_tags *tags;
	226	busy_tag_iter_fn *fn;
	227	void *data;
	228	bool reserved;
	229	};
	230
	231	static bool bt_tags_iter(struct sbitmap bitmap, unsigned int bitnr, void data)
f26cdc85	232	{
88459642 OS	233	struct bt_tags_iter_data *iter_data = data;
	234	struct blk_mq_tags *tags = iter_data->tags;
	235	bool reserved = iter_data->reserved;
f26cdc85	236	struct request *rq;
f26cdc85	237
88459642 OS	238	if (!reserved)
	239	bitnr += tags->nr_reserved_tags;
	240	rq = tags->rqs[bitnr];
f26cdc85	241
88459642 OS	242	iter_data->fn(rq, iter_data->data, reserved);
	243	return true;
	244	}
	245
	246	static void bt_tags_for_each(struct blk_mq_tags tags, struct sbitmap_queue bt,
	247	busy_tag_iter_fn fn, void data, bool reserved)
	248	{
	249	struct bt_tags_iter_data iter_data = {
	250	.tags = tags,
	251	.fn = fn,
	252	.data = data,
	253	.reserved = reserved,
	254	};
	255
	256	if (tags->rqs)
	257	sbitmap_for_each_set(&bt->sb, bt_tags_iter, &iter_data);
f26cdc85 KB	258	}
f26cdc85 KB	259
e8f1e163 SG	260	static void blk_mq_all_tag_busy_iter(struct blk_mq_tags *tags,
e8f1e163 SG	261	busy_tag_iter_fn fn, void priv)
f26cdc85 KB	262	{
f26cdc85 KB	263	if (tags->nr_reserved_tags)
88459642 OS	264	bt_tags_for_each(tags, &tags->breserved_tags, fn, priv, true);
88459642 OS	265	bt_tags_for_each(tags, &tags->bitmap_tags, fn, priv, false);
f26cdc85	266	}
f26cdc85	267
e0489487 SG	268	void blk_mq_tagset_busy_iter(struct blk_mq_tag_set *tagset,
	269	busy_tag_iter_fn fn, void priv)
	270	{
	271	int i;
	272
	273	for (i = 0; i < tagset->nr_hw_queues; i++) {
	274	if (tagset->tags && tagset->tags[i])
	275	blk_mq_all_tag_busy_iter(tagset->tags[i], fn, priv);
	276	}
	277	}
	278	EXPORT_SYMBOL(blk_mq_tagset_busy_iter);
	279
486cf989 SG	280	int blk_mq_reinit_tagset(struct blk_mq_tag_set *set)
	281	{
	282	int i, j, ret = 0;
	283
	284	if (!set->ops->reinit_request)
	285	goto out;
	286
	287	for (i = 0; i < set->nr_hw_queues; i++) {
	288	struct blk_mq_tags *tags = set->tags[i];
	289
	290	for (j = 0; j < tags->nr_tags; j++) {
2af8cbe3	291	if (!tags->static_rqs[j])
486cf989 SG	292	continue;
	293
	294	ret = set->ops->reinit_request(set->driver_data,
2af8cbe3	295	tags->static_rqs[j]);
486cf989 SG	296	if (ret)
	297	goto out;
	298	}
	299	}
	300
	301	out:
	302	return ret;
	303	}
	304	EXPORT_SYMBOL_GPL(blk_mq_reinit_tagset);
	305
0bf6cd5b	306	void blk_mq_queue_tag_busy_iter(struct request_queue q, busy_iter_fn fn,
81481eb4	307	void *priv)
320ae51f	308	{
0bf6cd5b CH	309	struct blk_mq_hw_ctx *hctx;
	310	int i;
	311
	312
	313	queue_for_each_hw_ctx(q, hctx, i) {
	314	struct blk_mq_tags *tags = hctx->tags;
	315
	316	/*
	317	* If not software queues are currently mapped to this
	318	* hardware queue, there's nothing to check
	319	*/
	320	if (!blk_mq_hw_queue_mapped(hctx))
	321	continue;
	322
	323	if (tags->nr_reserved_tags)
88459642 OS	324	bt_for_each(hctx, &tags->breserved_tags, fn, priv, true);
88459642 OS	325	bt_for_each(hctx, &tags->bitmap_tags, fn, priv, false);
4bb659b1 JA	326	}
4bb659b1 JA	327
4bb659b1 JA	328	}
4bb659b1 JA	329
88459642	330	static unsigned int bt_unused_tags(const struct sbitmap_queue *bt)
e3a2b3f9	331	{
88459642	332	return bt->sb.depth - sbitmap_weight(&bt->sb);
e3a2b3f9 JA	333	}
e3a2b3f9 JA	334
f4a644db OS	335	static int bt_alloc(struct sbitmap_queue *bt, unsigned int depth,
f4a644db OS	336	bool round_robin, int node)
4bb659b1	337	{
f4a644db OS	338	return sbitmap_queue_init_node(bt, depth, -1, round_robin, GFP_KERNEL,
f4a644db OS	339	node);
4bb659b1 JA	340	}
	341
	342	static struct blk_mq_tags blk_mq_init_bitmap_tags(struct blk_mq_tags tags,
24391c0d	343	int node, int alloc_policy)
4bb659b1 JA	344	{
4bb659b1 JA	345	unsigned int depth = tags->nr_tags - tags->nr_reserved_tags;
f4a644db	346	bool round_robin = alloc_policy == BLK_TAG_ALLOC_RR;
4bb659b1	347
f4a644db	348	if (bt_alloc(&tags->bitmap_tags, depth, round_robin, node))
88459642	349	goto free_tags;
f4a644db OS	350	if (bt_alloc(&tags->breserved_tags, tags->nr_reserved_tags, round_robin,
f4a644db OS	351	node))
88459642	352	goto free_bitmap_tags;
4bb659b1 JA	353
4bb659b1 JA	354	return tags;
88459642 OS	355	free_bitmap_tags:
	356	sbitmap_queue_free(&tags->bitmap_tags);
	357	free_tags:
4bb659b1 JA	358	kfree(tags);
	359	return NULL;
	360	}
	361
320ae51f	362	struct blk_mq_tags *blk_mq_init_tags(unsigned int total_tags,
24391c0d SL	363	unsigned int reserved_tags,
24391c0d SL	364	int node, int alloc_policy)
320ae51f	365	{
320ae51f	366	struct blk_mq_tags *tags;
320ae51f JA	367
	368	if (total_tags > BLK_MQ_TAG_MAX) {
	369	pr_err("blk-mq: tag depth too large\n");
	370	return NULL;
	371	}
	372
	373	tags = kzalloc_node(sizeof(*tags), GFP_KERNEL, node);
	374	if (!tags)
	375	return NULL;
	376
320ae51f JA	377	tags->nr_tags = total_tags;
320ae51f JA	378	tags->nr_reserved_tags = reserved_tags;
320ae51f	379
24391c0d	380	return blk_mq_init_bitmap_tags(tags, node, alloc_policy);
320ae51f JA	381	}
	382
	383	void blk_mq_free_tags(struct blk_mq_tags *tags)
	384	{
88459642 OS	385	sbitmap_queue_free(&tags->bitmap_tags);
88459642 OS	386	sbitmap_queue_free(&tags->breserved_tags);
320ae51f JA	387	kfree(tags);
	388	}
	389
70f36b60 JA	390	int blk_mq_tag_update_depth(struct blk_mq_hw_ctx *hctx,
	391	struct blk_mq_tags **tagsptr, unsigned int tdepth,
	392	bool can_grow)
e3a2b3f9	393	{
70f36b60 JA	394	struct blk_mq_tags tags = tagsptr;
	395
	396	if (tdepth <= tags->nr_reserved_tags)
e3a2b3f9 JA	397	return -EINVAL;
e3a2b3f9 JA	398
70f36b60 JA	399	tdepth -= tags->nr_reserved_tags;
70f36b60 JA	400
e3a2b3f9	401	/*
70f36b60 JA	402	* If we are allowed to grow beyond the original size, allocate
70f36b60 JA	403	* a new set of tags before freeing the old one.
e3a2b3f9	404	*/
70f36b60 JA	405	if (tdepth > tags->nr_tags) {
	406	struct blk_mq_tag_set *set = hctx->queue->tag_set;
	407	struct blk_mq_tags *new;
	408	bool ret;
	409
	410	if (!can_grow)
	411	return -EINVAL;
	412
	413	/*
	414	* We need some sort of upper limit, set it high enough that
	415	* no valid use cases should require more.
	416	*/
	417	if (tdepth > 16 * BLKDEV_MAX_RQ)
	418	return -EINVAL;
	419
	420	new = blk_mq_alloc_rq_map(set, hctx->queue_num, tdepth, 0);
	421	if (!new)
	422	return -ENOMEM;
	423	ret = blk_mq_alloc_rqs(set, new, hctx->queue_num, tdepth);
	424	if (ret) {
	425	blk_mq_free_rq_map(new);
	426	return -ENOMEM;
	427	}
	428
	429	blk_mq_free_rqs(set, *tagsptr, hctx->queue_num);
	430	blk_mq_free_rq_map(*tagsptr);
	431	*tagsptr = new;
	432	} else {
	433	/*
	434	* Don't need (or can't) update reserved tags here, they
	435	* remain static and should never need resizing.
	436	*/
	437	sbitmap_queue_resize(&tags->bitmap_tags, tdepth);
	438	}
88459642	439
e3a2b3f9 JA	440	return 0;
	441	}
	442
205fb5f5 BVA	443	/**
	444	* blk_mq_unique_tag() - return a tag that is unique queue-wide
	445	* @rq: request for which to compute a unique tag
	446	*
	447	* The tag field in struct request is unique per hardware queue but not over
	448	* all hardware queues. Hence this function that returns a tag with the
	449	* hardware context index in the upper bits and the per hardware queue tag in
	450	* the lower bits.
	451	*
	452	* Note: When called for a request that is queued on a non-multiqueue request
	453	* queue, the hardware context index is set to zero.
	454	*/
	455	u32 blk_mq_unique_tag(struct request *rq)
	456	{
	457	struct request_queue *q = rq->q;
	458	struct blk_mq_hw_ctx *hctx;
	459	int hwq = 0;
	460
	461	if (q->mq_ops) {
7d7e0f90	462	hctx = blk_mq_map_queue(q, rq->mq_ctx->cpu);
205fb5f5 BVA	463	hwq = hctx->queue_num;
	464	}
	465
	466	return (hwq << BLK_MQ_UNIQUE_TAG_BITS) \|
	467	(rq->tag & BLK_MQ_UNIQUE_TAG_MASK);
	468	}
	469	EXPORT_SYMBOL(blk_mq_unique_tag);
	470
320ae51f JA	471	ssize_t blk_mq_tag_sysfs_show(struct blk_mq_tags tags, char page)
	472	{
	473	char *orig_page = page;
4bb659b1	474	unsigned int free, res;
320ae51f JA	475
	476	if (!tags)
	477	return 0;
	478
59d13bf5 JA	479	page += sprintf(page, "nr_tags=%u, reserved_tags=%u, "
	480	"bits_per_word=%u\n",
	481	tags->nr_tags, tags->nr_reserved_tags,
88459642	482	1U << tags->bitmap_tags.sb.shift);
320ae51f	483
4bb659b1 JA	484	free = bt_unused_tags(&tags->bitmap_tags);
4bb659b1 JA	485	res = bt_unused_tags(&tags->breserved_tags);
320ae51f	486
4bb659b1	487	page += sprintf(page, "nr_free=%u, nr_reserved=%u\n", free, res);
0d2602ca	488	page += sprintf(page, "active_queues=%u\n", atomic_read(&tags->active_queues));
320ae51f JA	489
	490	return page - orig_page;
	491	}