[linux.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;
	if (data->shallow_depth)
		return __sbitmap_queue_get_shallow(bt, data->shallow_depth);
	else
		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];

	/*
	 * We can hit rq == NULL here, because the tagging functions
	 * test and set the bit before assining ->rqs[].
	 */
	if (rq && 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;

	/*
	 * We can hit rq == NULL here, because the tagging functions
	 * test and set the bit before assining ->rqs[].
	 */
	rq = tags->rqs[bitnr];
	if (rq)
		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_tagset_iter(struct blk_mq_tag_set *set, void *data,
			 int (fn)(void *, struct request *))
{
	int i, j, ret = 0;

	if (WARN_ON_ONCE(!fn))
		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 = fn(data, tags->static_rqs[j]);
			if (ret)
				goto out;
		}
	}

out:
	return ret;
}
EXPORT_SYMBOL_GPL(blk_mq_tagset_iter);

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;
229a9287 OS	99	if (data->shallow_depth)
	100	return __sbitmap_queue_get_shallow(bt, data->shallow_depth);
	101	else
	102	return __sbitmap_queue_get(bt);
4bb659b1 JA	103	}
4bb659b1 JA	104
4941115b	105	unsigned int blk_mq_get_tag(struct blk_mq_alloc_data *data)
320ae51f	106	{
4941115b JA	107	struct blk_mq_tags *tags = blk_mq_tags_from_data(data);
4941115b JA	108	struct sbitmap_queue *bt;
88459642	109	struct sbq_wait_state *ws;
4bb659b1	110	DEFINE_WAIT(wait);
4941115b	111	unsigned int tag_offset;
bd6737f1	112	bool drop_ctx;
320ae51f JA	113	int tag;
320ae51f JA	114
4941115b JA	115	if (data->flags & BLK_MQ_REQ_RESERVED) {
	116	if (unlikely(!tags->nr_reserved_tags)) {
	117	WARN_ON_ONCE(1);
	118	return BLK_MQ_TAG_FAIL;
	119	}
	120	bt = &tags->breserved_tags;
	121	tag_offset = 0;
	122	} else {
	123	bt = &tags->bitmap_tags;
	124	tag_offset = tags->nr_reserved_tags;
	125	}
	126
200e86b3	127	tag = __blk_mq_get_tag(data, bt);
4bb659b1	128	if (tag != -1)
4941115b	129	goto found_tag;
4bb659b1	130
6f3b0e8b	131	if (data->flags & BLK_MQ_REQ_NOWAIT)
4941115b	132	return BLK_MQ_TAG_FAIL;
4bb659b1	133
4941115b	134	ws = bt_wait_ptr(bt, data->hctx);
bd6737f1	135	drop_ctx = data->ctx == NULL;
4bb659b1	136	do {
88459642	137	prepare_to_wait(&ws->wait, &wait, TASK_UNINTERRUPTIBLE);
4bb659b1	138
200e86b3	139	tag = __blk_mq_get_tag(data, bt);
4bb659b1 JA	140	if (tag != -1)
	141	break;
	142
b3223207 BVA	143	/*
	144	* We're out of tags on this hardware queue, kick any
	145	* pending IO submits before going to sleep waiting for
8cecb07d	146	* some to complete.
b3223207	147	*/
8cecb07d	148	blk_mq_run_hw_queue(data->hctx, false);
b3223207	149
080ff351 JA	150	/*
	151	* Retry tag allocation after running the hardware queue,
	152	* as running the queue may also have found completions.
	153	*/
200e86b3	154	tag = __blk_mq_get_tag(data, bt);
080ff351 JA	155	if (tag != -1)
	156	break;
	157
bd6737f1 JA	158	if (data->ctx)
bd6737f1 JA	159	blk_mq_put_ctx(data->ctx);
cb96a42c	160
4bb659b1	161	io_schedule();
cb96a42c ML	162
cb96a42c ML	163	data->ctx = blk_mq_get_ctx(data->q);
7d7e0f90	164	data->hctx = blk_mq_map_queue(data->q, data->ctx->cpu);
4941115b JA	165	tags = blk_mq_tags_from_data(data);
	166	if (data->flags & BLK_MQ_REQ_RESERVED)
	167	bt = &tags->breserved_tags;
	168	else
	169	bt = &tags->bitmap_tags;
	170
88459642	171	finish_wait(&ws->wait, &wait);
4941115b	172	ws = bt_wait_ptr(bt, data->hctx);
4bb659b1 JA	173	} while (1);
4bb659b1 JA	174
bd6737f1 JA	175	if (drop_ctx && data->ctx)
	176	blk_mq_put_ctx(data->ctx);
	177
88459642	178	finish_wait(&ws->wait, &wait);
320ae51f	179
4941115b JA	180	found_tag:
4941115b JA	181	return tag + tag_offset;
320ae51f JA	182	}
320ae51f JA	183
4941115b JA	184	void blk_mq_put_tag(struct blk_mq_hw_ctx hctx, struct blk_mq_tags tags,
4941115b JA	185	struct blk_mq_ctx *ctx, unsigned int tag)
320ae51f	186	{
415b806d	187	if (!blk_mq_tag_is_reserved(tags, tag)) {
4bb659b1 JA	188	const int real_tag = tag - tags->nr_reserved_tags;
4bb659b1 JA	189
70114c39	190	BUG_ON(real_tag >= tags->nr_tags);
f4a644db	191	sbitmap_queue_clear(&tags->bitmap_tags, real_tag, ctx->cpu);
70114c39 JA	192	} else {
70114c39 JA	193	BUG_ON(tag >= tags->nr_reserved_tags);
f4a644db	194	sbitmap_queue_clear(&tags->breserved_tags, tag, ctx->cpu);
70114c39	195	}
320ae51f JA	196	}
320ae51f JA	197
88459642 OS	198	struct bt_iter_data {
	199	struct blk_mq_hw_ctx *hctx;
	200	busy_iter_fn *fn;
	201	void *data;
	202	bool reserved;
	203	};
	204
	205	static bool bt_iter(struct sbitmap bitmap, unsigned int bitnr, void data)
320ae51f	206	{
88459642 OS	207	struct bt_iter_data *iter_data = data;
	208	struct blk_mq_hw_ctx *hctx = iter_data->hctx;
	209	struct blk_mq_tags *tags = hctx->tags;
	210	bool reserved = iter_data->reserved;
81481eb4	211	struct request *rq;
4bb659b1	212
88459642 OS	213	if (!reserved)
	214	bitnr += tags->nr_reserved_tags;
	215	rq = tags->rqs[bitnr];
4bb659b1	216
7f5562d5 JA	217	/*
	218	* We can hit rq == NULL here, because the tagging functions
	219	* test and set the bit before assining ->rqs[].
	220	*/
	221	if (rq && rq->q == hctx->queue)
88459642 OS	222	iter_data->fn(hctx, rq, iter_data->data, reserved);
	223	return true;
	224	}
4bb659b1	225
88459642 OS	226	static void bt_for_each(struct blk_mq_hw_ctx hctx, struct sbitmap_queue bt,
	227	busy_iter_fn fn, void data, bool reserved)
	228	{
	229	struct bt_iter_data iter_data = {
	230	.hctx = hctx,
	231	.fn = fn,
	232	.data = data,
	233	.reserved = reserved,
	234	};
	235
	236	sbitmap_for_each_set(&bt->sb, bt_iter, &iter_data);
320ae51f JA	237	}
320ae51f JA	238
88459642 OS	239	struct bt_tags_iter_data {
	240	struct blk_mq_tags *tags;
	241	busy_tag_iter_fn *fn;
	242	void *data;
	243	bool reserved;
	244	};
	245
	246	static bool bt_tags_iter(struct sbitmap bitmap, unsigned int bitnr, void data)
f26cdc85	247	{
88459642 OS	248	struct bt_tags_iter_data *iter_data = data;
	249	struct blk_mq_tags *tags = iter_data->tags;
	250	bool reserved = iter_data->reserved;
f26cdc85	251	struct request *rq;
f26cdc85	252
88459642 OS	253	if (!reserved)
88459642 OS	254	bitnr += tags->nr_reserved_tags;
7f5562d5 JA	255
	256	/*
	257	* We can hit rq == NULL here, because the tagging functions
	258	* test and set the bit before assining ->rqs[].
	259	*/
88459642	260	rq = tags->rqs[bitnr];
7f5562d5 JA	261	if (rq)
7f5562d5 JA	262	iter_data->fn(rq, iter_data->data, reserved);
f26cdc85	263
88459642 OS	264	return true;
	265	}
	266
	267	static void bt_tags_for_each(struct blk_mq_tags tags, struct sbitmap_queue bt,
	268	busy_tag_iter_fn fn, void data, bool reserved)
	269	{
	270	struct bt_tags_iter_data iter_data = {
	271	.tags = tags,
	272	.fn = fn,
	273	.data = data,
	274	.reserved = reserved,
	275	};
	276
	277	if (tags->rqs)
	278	sbitmap_for_each_set(&bt->sb, bt_tags_iter, &iter_data);
f26cdc85 KB	279	}
f26cdc85 KB	280
e8f1e163 SG	281	static void blk_mq_all_tag_busy_iter(struct blk_mq_tags *tags,
e8f1e163 SG	282	busy_tag_iter_fn fn, void priv)
f26cdc85 KB	283	{
f26cdc85 KB	284	if (tags->nr_reserved_tags)
88459642 OS	285	bt_tags_for_each(tags, &tags->breserved_tags, fn, priv, true);
88459642 OS	286	bt_tags_for_each(tags, &tags->bitmap_tags, fn, priv, false);
f26cdc85	287	}
f26cdc85	288
e0489487 SG	289	void blk_mq_tagset_busy_iter(struct blk_mq_tag_set *tagset,
	290	busy_tag_iter_fn fn, void priv)
	291	{
	292	int i;
	293
	294	for (i = 0; i < tagset->nr_hw_queues; i++) {
	295	if (tagset->tags && tagset->tags[i])
	296	blk_mq_all_tag_busy_iter(tagset->tags[i], fn, priv);
	297	}
	298	}
	299	EXPORT_SYMBOL(blk_mq_tagset_busy_iter);
	300
149e10f8 SG	301	int blk_mq_tagset_iter(struct blk_mq_tag_set set, void data,
149e10f8 SG	302	int (fn)(void , struct request ))
486cf989 SG	303	{
	304	int i, j, ret = 0;
	305
149e10f8	306	if (WARN_ON_ONCE(!fn))
486cf989 SG	307	goto out;
	308
	309	for (i = 0; i < set->nr_hw_queues; i++) {
	310	struct blk_mq_tags *tags = set->tags[i];
	311
0067d4b0 SG	312	if (!tags)
	313	continue;
	314
486cf989	315	for (j = 0; j < tags->nr_tags; j++) {
2af8cbe3	316	if (!tags->static_rqs[j])
486cf989 SG	317	continue;
486cf989 SG	318
149e10f8	319	ret = fn(data, tags->static_rqs[j]);
486cf989 SG	320	if (ret)
	321	goto out;
	322	}
	323	}
	324
	325	out:
	326	return ret;
	327	}
149e10f8 SG	328	EXPORT_SYMBOL_GPL(blk_mq_tagset_iter);
149e10f8 SG	329
0bf6cd5b	330	void blk_mq_queue_tag_busy_iter(struct request_queue q, busy_iter_fn fn,
81481eb4	331	void *priv)
320ae51f	332	{
0bf6cd5b CH	333	struct blk_mq_hw_ctx *hctx;
	334	int i;
	335
	336
	337	queue_for_each_hw_ctx(q, hctx, i) {
	338	struct blk_mq_tags *tags = hctx->tags;
	339
	340	/*
	341	* If not software queues are currently mapped to this
	342	* hardware queue, there's nothing to check
	343	*/
	344	if (!blk_mq_hw_queue_mapped(hctx))
	345	continue;
	346
	347	if (tags->nr_reserved_tags)
88459642 OS	348	bt_for_each(hctx, &tags->breserved_tags, fn, priv, true);
88459642 OS	349	bt_for_each(hctx, &tags->bitmap_tags, fn, priv, false);
4bb659b1 JA	350	}
4bb659b1 JA	351
4bb659b1 JA	352	}
4bb659b1 JA	353
f4a644db OS	354	static int bt_alloc(struct sbitmap_queue *bt, unsigned int depth,
f4a644db OS	355	bool round_robin, int node)
4bb659b1	356	{
f4a644db OS	357	return sbitmap_queue_init_node(bt, depth, -1, round_robin, GFP_KERNEL,
f4a644db OS	358	node);
4bb659b1 JA	359	}
	360
	361	static struct blk_mq_tags blk_mq_init_bitmap_tags(struct blk_mq_tags tags,
24391c0d	362	int node, int alloc_policy)
4bb659b1 JA	363	{
4bb659b1 JA	364	unsigned int depth = tags->nr_tags - tags->nr_reserved_tags;
f4a644db	365	bool round_robin = alloc_policy == BLK_TAG_ALLOC_RR;
4bb659b1	366
f4a644db	367	if (bt_alloc(&tags->bitmap_tags, depth, round_robin, node))
88459642	368	goto free_tags;
f4a644db OS	369	if (bt_alloc(&tags->breserved_tags, tags->nr_reserved_tags, round_robin,
f4a644db OS	370	node))
88459642	371	goto free_bitmap_tags;
4bb659b1 JA	372
4bb659b1 JA	373	return tags;
88459642 OS	374	free_bitmap_tags:
	375	sbitmap_queue_free(&tags->bitmap_tags);
	376	free_tags:
4bb659b1 JA	377	kfree(tags);
	378	return NULL;
	379	}
	380
320ae51f	381	struct blk_mq_tags *blk_mq_init_tags(unsigned int total_tags,
24391c0d SL	382	unsigned int reserved_tags,
24391c0d SL	383	int node, int alloc_policy)
320ae51f	384	{
320ae51f	385	struct blk_mq_tags *tags;
320ae51f JA	386
	387	if (total_tags > BLK_MQ_TAG_MAX) {
	388	pr_err("blk-mq: tag depth too large\n");
	389	return NULL;
	390	}
	391
	392	tags = kzalloc_node(sizeof(*tags), GFP_KERNEL, node);
	393	if (!tags)
	394	return NULL;
	395
320ae51f JA	396	tags->nr_tags = total_tags;
320ae51f JA	397	tags->nr_reserved_tags = reserved_tags;
320ae51f	398
24391c0d	399	return blk_mq_init_bitmap_tags(tags, node, alloc_policy);
320ae51f JA	400	}
	401
	402	void blk_mq_free_tags(struct blk_mq_tags *tags)
	403	{
88459642 OS	404	sbitmap_queue_free(&tags->bitmap_tags);
88459642 OS	405	sbitmap_queue_free(&tags->breserved_tags);
320ae51f JA	406	kfree(tags);
	407	}
	408
70f36b60 JA	409	int blk_mq_tag_update_depth(struct blk_mq_hw_ctx *hctx,
	410	struct blk_mq_tags **tagsptr, unsigned int tdepth,
	411	bool can_grow)
e3a2b3f9	412	{
70f36b60 JA	413	struct blk_mq_tags tags = tagsptr;
	414
	415	if (tdepth <= tags->nr_reserved_tags)
e3a2b3f9 JA	416	return -EINVAL;
e3a2b3f9 JA	417
70f36b60 JA	418	tdepth -= tags->nr_reserved_tags;
70f36b60 JA	419
e3a2b3f9	420	/*
70f36b60 JA	421	* If we are allowed to grow beyond the original size, allocate
70f36b60 JA	422	* a new set of tags before freeing the old one.
e3a2b3f9	423	*/
70f36b60 JA	424	if (tdepth > tags->nr_tags) {
	425	struct blk_mq_tag_set *set = hctx->queue->tag_set;
	426	struct blk_mq_tags *new;
	427	bool ret;
	428
	429	if (!can_grow)
	430	return -EINVAL;
	431
	432	/*
	433	* We need some sort of upper limit, set it high enough that
	434	* no valid use cases should require more.
	435	*/
	436	if (tdepth > 16 * BLKDEV_MAX_RQ)
	437	return -EINVAL;
	438
	439	new = blk_mq_alloc_rq_map(set, hctx->queue_num, tdepth, 0);
	440	if (!new)
	441	return -ENOMEM;
	442	ret = blk_mq_alloc_rqs(set, new, hctx->queue_num, tdepth);
	443	if (ret) {
	444	blk_mq_free_rq_map(new);
	445	return -ENOMEM;
	446	}
	447
	448	blk_mq_free_rqs(set, *tagsptr, hctx->queue_num);
	449	blk_mq_free_rq_map(*tagsptr);
	450	*tagsptr = new;
	451	} else {
	452	/*
	453	* Don't need (or can't) update reserved tags here, they
	454	* remain static and should never need resizing.
	455	*/
	456	sbitmap_queue_resize(&tags->bitmap_tags, tdepth);
	457	}
88459642	458
e3a2b3f9 JA	459	return 0;
	460	}
	461
205fb5f5 BVA	462	/**
	463	* blk_mq_unique_tag() - return a tag that is unique queue-wide
	464	* @rq: request for which to compute a unique tag
	465	*
	466	* The tag field in struct request is unique per hardware queue but not over
	467	* all hardware queues. Hence this function that returns a tag with the
	468	* hardware context index in the upper bits and the per hardware queue tag in
	469	* the lower bits.
	470	*
	471	* Note: When called for a request that is queued on a non-multiqueue request
	472	* queue, the hardware context index is set to zero.
	473	*/
	474	u32 blk_mq_unique_tag(struct request *rq)
	475	{
	476	struct request_queue *q = rq->q;
	477	struct blk_mq_hw_ctx *hctx;
	478	int hwq = 0;
	479
	480	if (q->mq_ops) {
7d7e0f90	481	hctx = blk_mq_map_queue(q, rq->mq_ctx->cpu);
205fb5f5 BVA	482	hwq = hctx->queue_num;
	483	}
	484
	485	return (hwq << BLK_MQ_UNIQUE_TAG_BITS) \|
	486	(rq->tag & BLK_MQ_UNIQUE_TAG_MASK);
	487	}
	488	EXPORT_SYMBOL(blk_mq_unique_tag);