[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/random.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;
}

#define BT_ALLOC_RR(tags) (tags->alloc_policy == BLK_TAG_ALLOC_RR)

static int __bt_get(struct blk_mq_hw_ctx *hctx, struct sbitmap_queue *bt,
		    struct blk_mq_tags *tags)
{
	if (!hctx_may_queue(hctx, bt))
		return -1;
	return __sbitmap_queue_get(bt, BT_ALLOC_RR(tags));
}

static int bt_get(struct blk_mq_alloc_data *data, struct sbitmap_queue *bt,
		  struct blk_mq_hw_ctx *hctx, struct blk_mq_tags *tags)
{
	struct sbq_wait_state *ws;
	DEFINE_WAIT(wait);
	int tag;

	tag = __bt_get(hctx, bt, tags);
	if (tag != -1)
		return tag;

	if (data->flags & BLK_MQ_REQ_NOWAIT)
		return -1;

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

		tag = __bt_get(hctx, bt, tags);
		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. Note that hctx can be NULL here for
		 * reserved tag allocation.
		 */
		if (hctx)
			blk_mq_run_hw_queue(hctx, false);

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

		blk_mq_put_ctx(data->ctx);

		io_schedule();

		data->ctx = blk_mq_get_ctx(data->q);
		data->hctx = data->q->mq_ops->map_queue(data->q,
				data->ctx->cpu);
		if (data->flags & BLK_MQ_REQ_RESERVED) {
			bt = &data->hctx->tags->breserved_tags;
		} else {
			hctx = data->hctx;
			bt = &hctx->tags->bitmap_tags;
		}
		finish_wait(&ws->wait, &wait);
		ws = bt_wait_ptr(bt, hctx);
	} while (1);

	finish_wait(&ws->wait, &wait);
	return tag;
}

static unsigned int __blk_mq_get_tag(struct blk_mq_alloc_data *data)
{
	int tag;

	tag = bt_get(data, &data->hctx->tags->bitmap_tags, data->hctx,
		     data->hctx->tags);
	if (tag >= 0)
		return tag + data->hctx->tags->nr_reserved_tags;

	return BLK_MQ_TAG_FAIL;
}

static unsigned int __blk_mq_get_reserved_tag(struct blk_mq_alloc_data *data)
{
	int tag;

	if (unlikely(!data->hctx->tags->nr_reserved_tags)) {
		WARN_ON_ONCE(1);
		return BLK_MQ_TAG_FAIL;
	}

	tag = bt_get(data, &data->hctx->tags->breserved_tags, NULL,
		     data->hctx->tags);
	if (tag < 0)
		return BLK_MQ_TAG_FAIL;

	return tag;
}

unsigned int blk_mq_get_tag(struct blk_mq_alloc_data *data)
{
	if (data->flags & BLK_MQ_REQ_RESERVED)
		return __blk_mq_get_reserved_tag(data);
	return __blk_mq_get_tag(data);
}

void blk_mq_put_tag(struct blk_mq_hw_ctx *hctx, struct blk_mq_ctx *ctx,
		    unsigned int tag)
{
	struct blk_mq_tags *tags = hctx->tags;

	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,
				    BT_ALLOC_RR(tags), ctx->cpu);
	} else {
		BUG_ON(tag >= tags->nr_reserved_tags);
		sbitmap_queue_clear(&tags->breserved_tags, tag,
				    BT_ALLOC_RR(tags), 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->rqs[j])
				continue;

			ret = set->ops->reinit_request(set->driver_data,
						tags->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, int node)
{
	return sbitmap_queue_init_node(bt, depth, -1, 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;

	tags->alloc_policy = alloc_policy;

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

	if (!zalloc_cpumask_var(&tags->cpumask, GFP_KERNEL)) {
		kfree(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);
	free_cpumask_var(tags->cpumask);
	kfree(tags);
}

void blk_mq_tag_init_last_tag(struct blk_mq_tags *tags, unsigned int *tag)
{
	unsigned int depth = tags->nr_tags - tags->nr_reserved_tags;

	*tag = prandom_u32() % depth;
}

int blk_mq_tag_update_depth(struct blk_mq_tags *tags, unsigned int tdepth)
{
	tdepth -= tags->nr_reserved_tags;
	if (tdepth > tags->nr_tags)
		return -EINVAL;

	/*
	 * 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);

	blk_mq_tag_wakeup_all(tags, false);
	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 = q->mq_ops->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>
4bb659b1	10	#include <linux/random.h>
320ae51f JA	11
	12	#include <linux/blk-mq.h>
	13	#include "blk.h"
	14	#include "blk-mq.h"
	15	#include "blk-mq-tag.h"
	16
320ae51f JA	17	bool blk_mq_has_free_tags(struct blk_mq_tags *tags)
320ae51f JA	18	{
4bb659b1 JA	19	if (!tags)
	20	return true;
	21
88459642	22	return sbitmap_any_bit_clear(&tags->bitmap_tags.sb);
0d2602ca JA	23	}
	24
	25	/*
	26	* If a previously inactive queue goes active, bump the active user count.
	27	*/
	28	bool __blk_mq_tag_busy(struct blk_mq_hw_ctx *hctx)
	29	{
	30	if (!test_bit(BLK_MQ_S_TAG_ACTIVE, &hctx->state) &&
	31	!test_and_set_bit(BLK_MQ_S_TAG_ACTIVE, &hctx->state))
	32	atomic_inc(&hctx->tags->active_queues);
	33
	34	return true;
	35	}
	36
	37	/*
aed3ea94	38	* Wakeup all potentially sleeping on tags
0d2602ca	39	*/
aed3ea94	40	void blk_mq_tag_wakeup_all(struct blk_mq_tags *tags, bool include_reserve)
0d2602ca	41	{
88459642 OS	42	sbitmap_queue_wake_all(&tags->bitmap_tags);
	43	if (include_reserve)
	44	sbitmap_queue_wake_all(&tags->breserved_tags);
0d2602ca JA	45	}
0d2602ca JA	46
e3a2b3f9 JA	47	/*
	48	* If a previously busy queue goes inactive, potential waiters could now
	49	* be allowed to queue. Wake them up and check.
	50	*/
	51	void __blk_mq_tag_idle(struct blk_mq_hw_ctx *hctx)
	52	{
	53	struct blk_mq_tags *tags = hctx->tags;
	54
	55	if (!test_and_clear_bit(BLK_MQ_S_TAG_ACTIVE, &hctx->state))
	56	return;
	57
	58	atomic_dec(&tags->active_queues);
	59
aed3ea94	60	blk_mq_tag_wakeup_all(tags, false);
e3a2b3f9 JA	61	}
e3a2b3f9 JA	62
0d2602ca JA	63	/*
	64	* For shared tag users, we track the number of currently active users
	65	* and attempt to provide a fair share of the tag depth for each of them.
	66	*/
	67	static inline bool hctx_may_queue(struct blk_mq_hw_ctx *hctx,
88459642	68	struct sbitmap_queue *bt)
0d2602ca JA	69	{
	70	unsigned int depth, users;
	71
	72	if (!hctx \|\| !(hctx->flags & BLK_MQ_F_TAG_SHARED))
	73	return true;
	74	if (!test_bit(BLK_MQ_S_TAG_ACTIVE, &hctx->state))
	75	return true;
	76
	77	/*
	78	* Don't try dividing an ant
	79	*/
88459642	80	if (bt->sb.depth == 1)
0d2602ca JA	81	return true;
	82
	83	users = atomic_read(&hctx->tags->active_queues);
	84	if (!users)
	85	return true;
	86
	87	/*
	88	* Allow at least some tags
	89	*/
88459642	90	depth = max((bt->sb.depth + users - 1) / users, 4U);
0d2602ca JA	91	return atomic_read(&hctx->nr_active) < depth;
	92	}
	93
24391c0d SL	94	#define BT_ALLOC_RR(tags) (tags->alloc_policy == BLK_TAG_ALLOC_RR)
24391c0d SL	95
88459642	96	static int __bt_get(struct blk_mq_hw_ctx hctx, struct sbitmap_queue bt,
40aabb67	97	struct blk_mq_tags *tags)
4bb659b1	98	{
0d2602ca JA	99	if (!hctx_may_queue(hctx, bt))
0d2602ca JA	100	return -1;
40aabb67	101	return __sbitmap_queue_get(bt, BT_ALLOC_RR(tags));
4bb659b1 JA	102	}
4bb659b1 JA	103
40aabb67 OS	104	static int bt_get(struct blk_mq_alloc_data data, struct sbitmap_queue bt,
40aabb67 OS	105	struct blk_mq_hw_ctx hctx, struct blk_mq_tags tags)
320ae51f	106	{
88459642	107	struct sbq_wait_state *ws;
4bb659b1	108	DEFINE_WAIT(wait);
320ae51f JA	109	int tag;
320ae51f JA	110
40aabb67	111	tag = __bt_get(hctx, bt, tags);
4bb659b1 JA	112	if (tag != -1)
	113	return tag;
	114
6f3b0e8b	115	if (data->flags & BLK_MQ_REQ_NOWAIT)
4bb659b1 JA	116	return -1;
4bb659b1 JA	117
88459642	118	ws = bt_wait_ptr(bt, hctx);
4bb659b1	119	do {
88459642	120	prepare_to_wait(&ws->wait, &wait, TASK_UNINTERRUPTIBLE);
4bb659b1	121
40aabb67	122	tag = __bt_get(hctx, bt, tags);
4bb659b1 JA	123	if (tag != -1)
	124	break;
	125
b3223207 BVA	126	/*
	127	* We're out of tags on this hardware queue, kick any
	128	* pending IO submits before going to sleep waiting for
bc188d81 SB	129	* some to complete. Note that hctx can be NULL here for
bc188d81 SB	130	* reserved tag allocation.
b3223207	131	*/
bc188d81 SB	132	if (hctx)
bc188d81 SB	133	blk_mq_run_hw_queue(hctx, false);
b3223207	134
080ff351 JA	135	/*
	136	* Retry tag allocation after running the hardware queue,
	137	* as running the queue may also have found completions.
	138	*/
40aabb67	139	tag = __bt_get(hctx, bt, tags);
080ff351 JA	140	if (tag != -1)
	141	break;
	142
cb96a42c ML	143	blk_mq_put_ctx(data->ctx);
cb96a42c ML	144
4bb659b1	145	io_schedule();
cb96a42c ML	146
	147	data->ctx = blk_mq_get_ctx(data->q);
	148	data->hctx = data->q->mq_ops->map_queue(data->q,
	149	data->ctx->cpu);
6f3b0e8b	150	if (data->flags & BLK_MQ_REQ_RESERVED) {
cb96a42c ML	151	bt = &data->hctx->tags->breserved_tags;
cb96a42c ML	152	} else {
cb96a42c ML	153	hctx = data->hctx;
	154	bt = &hctx->tags->bitmap_tags;
	155	}
88459642 OS	156	finish_wait(&ws->wait, &wait);
88459642 OS	157	ws = bt_wait_ptr(bt, hctx);
4bb659b1 JA	158	} while (1);
4bb659b1 JA	159
88459642	160	finish_wait(&ws->wait, &wait);
4bb659b1 JA	161	return tag;
	162	}
	163
cb96a42c	164	static unsigned int __blk_mq_get_tag(struct blk_mq_alloc_data *data)
4bb659b1 JA	165	{
	166	int tag;
	167
cb96a42c	168	tag = bt_get(data, &data->hctx->tags->bitmap_tags, data->hctx,
40aabb67	169	data->hctx->tags);
4bb659b1	170	if (tag >= 0)
cb96a42c	171	return tag + data->hctx->tags->nr_reserved_tags;
4bb659b1 JA	172
4bb659b1 JA	173	return BLK_MQ_TAG_FAIL;
320ae51f JA	174	}
320ae51f JA	175
cb96a42c	176	static unsigned int __blk_mq_get_reserved_tag(struct blk_mq_alloc_data *data)
320ae51f	177	{
40aabb67	178	int tag;
320ae51f	179
cb96a42c	180	if (unlikely(!data->hctx->tags->nr_reserved_tags)) {
320ae51f JA	181	WARN_ON_ONCE(1);
	182	return BLK_MQ_TAG_FAIL;
	183	}
	184
40aabb67 OS	185	tag = bt_get(data, &data->hctx->tags->breserved_tags, NULL,
40aabb67 OS	186	data->hctx->tags);
320ae51f JA	187	if (tag < 0)
320ae51f JA	188	return BLK_MQ_TAG_FAIL;
4bb659b1	189
320ae51f JA	190	return tag;
	191	}
	192
cb96a42c	193	unsigned int blk_mq_get_tag(struct blk_mq_alloc_data *data)
320ae51f	194	{
6f3b0e8b CH	195	if (data->flags & BLK_MQ_REQ_RESERVED)
	196	return __blk_mq_get_reserved_tag(data);
	197	return __blk_mq_get_tag(data);
320ae51f JA	198	}
320ae51f JA	199
40aabb67 OS	200	void blk_mq_put_tag(struct blk_mq_hw_ctx hctx, struct blk_mq_ctx ctx,
40aabb67 OS	201	unsigned int tag)
320ae51f	202	{
0d2602ca JA	203	struct blk_mq_tags *tags = hctx->tags;
0d2602ca JA	204
4bb659b1 JA	205	if (tag >= tags->nr_reserved_tags) {
	206	const int real_tag = tag - tags->nr_reserved_tags;
	207
70114c39	208	BUG_ON(real_tag >= tags->nr_tags);
40aabb67 OS	209	sbitmap_queue_clear(&tags->bitmap_tags, real_tag,
40aabb67 OS	210	BT_ALLOC_RR(tags), ctx->cpu);
70114c39 JA	211	} else {
70114c39 JA	212	BUG_ON(tag >= tags->nr_reserved_tags);
40aabb67 OS	213	sbitmap_queue_clear(&tags->breserved_tags, tag,
40aabb67 OS	214	BT_ALLOC_RR(tags), ctx->cpu);
70114c39	215	}
320ae51f JA	216	}
320ae51f JA	217
88459642 OS	218	struct bt_iter_data {
	219	struct blk_mq_hw_ctx *hctx;
	220	busy_iter_fn *fn;
	221	void *data;
	222	bool reserved;
	223	};
	224
	225	static bool bt_iter(struct sbitmap bitmap, unsigned int bitnr, void data)
320ae51f	226	{
88459642 OS	227	struct bt_iter_data *iter_data = data;
	228	struct blk_mq_hw_ctx *hctx = iter_data->hctx;
	229	struct blk_mq_tags *tags = hctx->tags;
	230	bool reserved = iter_data->reserved;
81481eb4	231	struct request *rq;
4bb659b1	232
88459642 OS	233	if (!reserved)
	234	bitnr += tags->nr_reserved_tags;
	235	rq = tags->rqs[bitnr];
4bb659b1	236
88459642 OS	237	if (rq->q == hctx->queue)
	238	iter_data->fn(hctx, rq, iter_data->data, reserved);
	239	return true;
	240	}
4bb659b1	241
88459642 OS	242	static void bt_for_each(struct blk_mq_hw_ctx hctx, struct sbitmap_queue bt,
	243	busy_iter_fn fn, void data, bool reserved)
	244	{
	245	struct bt_iter_data iter_data = {
	246	.hctx = hctx,
	247	.fn = fn,
	248	.data = data,
	249	.reserved = reserved,
	250	};
	251
	252	sbitmap_for_each_set(&bt->sb, bt_iter, &iter_data);
320ae51f JA	253	}
320ae51f JA	254
88459642 OS	255	struct bt_tags_iter_data {
	256	struct blk_mq_tags *tags;
	257	busy_tag_iter_fn *fn;
	258	void *data;
	259	bool reserved;
	260	};
	261
	262	static bool bt_tags_iter(struct sbitmap bitmap, unsigned int bitnr, void data)
f26cdc85	263	{
88459642 OS	264	struct bt_tags_iter_data *iter_data = data;
	265	struct blk_mq_tags *tags = iter_data->tags;
	266	bool reserved = iter_data->reserved;
f26cdc85	267	struct request *rq;
f26cdc85	268
88459642 OS	269	if (!reserved)
	270	bitnr += tags->nr_reserved_tags;
	271	rq = tags->rqs[bitnr];
f26cdc85	272
88459642 OS	273	iter_data->fn(rq, iter_data->data, reserved);
	274	return true;
	275	}
	276
	277	static void bt_tags_for_each(struct blk_mq_tags tags, struct sbitmap_queue bt,
	278	busy_tag_iter_fn fn, void data, bool reserved)
	279	{
	280	struct bt_tags_iter_data iter_data = {
	281	.tags = tags,
	282	.fn = fn,
	283	.data = data,
	284	.reserved = reserved,
	285	};
	286
	287	if (tags->rqs)
	288	sbitmap_for_each_set(&bt->sb, bt_tags_iter, &iter_data);
f26cdc85 KB	289	}
f26cdc85 KB	290
e8f1e163 SG	291	static void blk_mq_all_tag_busy_iter(struct blk_mq_tags *tags,
e8f1e163 SG	292	busy_tag_iter_fn fn, void priv)
f26cdc85 KB	293	{
f26cdc85 KB	294	if (tags->nr_reserved_tags)
88459642 OS	295	bt_tags_for_each(tags, &tags->breserved_tags, fn, priv, true);
88459642 OS	296	bt_tags_for_each(tags, &tags->bitmap_tags, fn, priv, false);
f26cdc85	297	}
f26cdc85	298
e0489487 SG	299	void blk_mq_tagset_busy_iter(struct blk_mq_tag_set *tagset,
	300	busy_tag_iter_fn fn, void priv)
	301	{
	302	int i;
	303
	304	for (i = 0; i < tagset->nr_hw_queues; i++) {
	305	if (tagset->tags && tagset->tags[i])
	306	blk_mq_all_tag_busy_iter(tagset->tags[i], fn, priv);
	307	}
	308	}
	309	EXPORT_SYMBOL(blk_mq_tagset_busy_iter);
	310
486cf989 SG	311	int blk_mq_reinit_tagset(struct blk_mq_tag_set *set)
	312	{
	313	int i, j, ret = 0;
	314
	315	if (!set->ops->reinit_request)
	316	goto out;
	317
	318	for (i = 0; i < set->nr_hw_queues; i++) {
	319	struct blk_mq_tags *tags = set->tags[i];
	320
	321	for (j = 0; j < tags->nr_tags; j++) {
	322	if (!tags->rqs[j])
	323	continue;
	324
	325	ret = set->ops->reinit_request(set->driver_data,
	326	tags->rqs[j]);
	327	if (ret)
	328	goto out;
	329	}
	330	}
	331
	332	out:
	333	return ret;
	334	}
	335	EXPORT_SYMBOL_GPL(blk_mq_reinit_tagset);
	336
0bf6cd5b	337	void blk_mq_queue_tag_busy_iter(struct request_queue q, busy_iter_fn fn,
81481eb4	338	void *priv)
320ae51f	339	{
0bf6cd5b CH	340	struct blk_mq_hw_ctx *hctx;
	341	int i;
	342
	343
	344	queue_for_each_hw_ctx(q, hctx, i) {
	345	struct blk_mq_tags *tags = hctx->tags;
	346
	347	/*
	348	* If not software queues are currently mapped to this
	349	* hardware queue, there's nothing to check
	350	*/
	351	if (!blk_mq_hw_queue_mapped(hctx))
	352	continue;
	353
	354	if (tags->nr_reserved_tags)
88459642 OS	355	bt_for_each(hctx, &tags->breserved_tags, fn, priv, true);
88459642 OS	356	bt_for_each(hctx, &tags->bitmap_tags, fn, priv, false);
4bb659b1 JA	357	}
4bb659b1 JA	358
4bb659b1 JA	359	}
4bb659b1 JA	360
88459642	361	static unsigned int bt_unused_tags(const struct sbitmap_queue *bt)
e3a2b3f9	362	{
88459642	363	return bt->sb.depth - sbitmap_weight(&bt->sb);
e3a2b3f9 JA	364	}
e3a2b3f9 JA	365
88459642	366	static int bt_alloc(struct sbitmap_queue *bt, unsigned int depth, int node)
4bb659b1	367	{
88459642	368	return sbitmap_queue_init_node(bt, depth, -1, GFP_KERNEL, node);
4bb659b1 JA	369	}
	370
	371	static struct blk_mq_tags blk_mq_init_bitmap_tags(struct blk_mq_tags tags,
24391c0d	372	int node, int alloc_policy)
4bb659b1 JA	373	{
	374	unsigned int depth = tags->nr_tags - tags->nr_reserved_tags;
	375
24391c0d SL	376	tags->alloc_policy = alloc_policy;
24391c0d SL	377
88459642 OS	378	if (bt_alloc(&tags->bitmap_tags, depth, node))
	379	goto free_tags;
	380	if (bt_alloc(&tags->breserved_tags, tags->nr_reserved_tags, node))
	381	goto free_bitmap_tags;
4bb659b1 JA	382
4bb659b1 JA	383	return tags;
88459642 OS	384	free_bitmap_tags:
	385	sbitmap_queue_free(&tags->bitmap_tags);
	386	free_tags:
4bb659b1 JA	387	kfree(tags);
	388	return NULL;
	389	}
	390
320ae51f	391	struct blk_mq_tags *blk_mq_init_tags(unsigned int total_tags,
24391c0d SL	392	unsigned int reserved_tags,
24391c0d SL	393	int node, int alloc_policy)
320ae51f	394	{
320ae51f	395	struct blk_mq_tags *tags;
320ae51f JA	396
	397	if (total_tags > BLK_MQ_TAG_MAX) {
	398	pr_err("blk-mq: tag depth too large\n");
	399	return NULL;
	400	}
	401
	402	tags = kzalloc_node(sizeof(*tags), GFP_KERNEL, node);
	403	if (!tags)
	404	return NULL;
	405
f26cdc85 KB	406	if (!zalloc_cpumask_var(&tags->cpumask, GFP_KERNEL)) {
	407	kfree(tags);
	408	return NULL;
	409	}
	410
320ae51f JA	411	tags->nr_tags = total_tags;
320ae51f JA	412	tags->nr_reserved_tags = reserved_tags;
320ae51f	413
24391c0d	414	return blk_mq_init_bitmap_tags(tags, node, alloc_policy);
320ae51f JA	415	}
	416
	417	void blk_mq_free_tags(struct blk_mq_tags *tags)
	418	{
88459642 OS	419	sbitmap_queue_free(&tags->bitmap_tags);
88459642 OS	420	sbitmap_queue_free(&tags->breserved_tags);
f42d79ab	421	free_cpumask_var(tags->cpumask);
320ae51f JA	422	kfree(tags);
	423	}
	424
4bb659b1 JA	425	void blk_mq_tag_init_last_tag(struct blk_mq_tags tags, unsigned int tag)
	426	{
	427	unsigned int depth = tags->nr_tags - tags->nr_reserved_tags;
	428
9d3d21ae	429	*tag = prandom_u32() % depth;
4bb659b1 JA	430	}
4bb659b1 JA	431
e3a2b3f9 JA	432	int blk_mq_tag_update_depth(struct blk_mq_tags *tags, unsigned int tdepth)
	433	{
	434	tdepth -= tags->nr_reserved_tags;
	435	if (tdepth > tags->nr_tags)
	436	return -EINVAL;
	437
	438	/*
	439	* Don't need (or can't) update reserved tags here, they remain
	440	* static and should never need resizing.
	441	*/
88459642 OS	442	sbitmap_queue_resize(&tags->bitmap_tags, tdepth);
88459642 OS	443
aed3ea94	444	blk_mq_tag_wakeup_all(tags, false);
e3a2b3f9 JA	445	return 0;
	446	}
	447
205fb5f5 BVA	448	/**
	449	* blk_mq_unique_tag() - return a tag that is unique queue-wide
	450	* @rq: request for which to compute a unique tag
	451	*
	452	* The tag field in struct request is unique per hardware queue but not over
	453	* all hardware queues. Hence this function that returns a tag with the
	454	* hardware context index in the upper bits and the per hardware queue tag in
	455	* the lower bits.
	456	*
	457	* Note: When called for a request that is queued on a non-multiqueue request
	458	* queue, the hardware context index is set to zero.
	459	*/
	460	u32 blk_mq_unique_tag(struct request *rq)
	461	{
	462	struct request_queue *q = rq->q;
	463	struct blk_mq_hw_ctx *hctx;
	464	int hwq = 0;
	465
	466	if (q->mq_ops) {
	467	hctx = q->mq_ops->map_queue(q, rq->mq_ctx->cpu);
	468	hwq = hctx->queue_num;
	469	}
	470
	471	return (hwq << BLK_MQ_UNIQUE_TAG_BITS) \|
	472	(rq->tag & BLK_MQ_UNIQUE_TAG_MASK);
	473	}
	474	EXPORT_SYMBOL(blk_mq_unique_tag);
	475
320ae51f JA	476	ssize_t blk_mq_tag_sysfs_show(struct blk_mq_tags tags, char page)
	477	{
	478	char *orig_page = page;
4bb659b1	479	unsigned int free, res;
320ae51f JA	480
	481	if (!tags)
	482	return 0;
	483
59d13bf5 JA	484	page += sprintf(page, "nr_tags=%u, reserved_tags=%u, "
	485	"bits_per_word=%u\n",
	486	tags->nr_tags, tags->nr_reserved_tags,
88459642	487	1U << tags->bitmap_tags.sb.shift);
320ae51f	488
4bb659b1 JA	489	free = bt_unused_tags(&tags->bitmap_tags);
4bb659b1 JA	490	res = bt_unused_tags(&tags->breserved_tags);
320ae51f	491
4bb659b1	492	page += sprintf(page, "nr_free=%u, nr_reserved=%u\n", free, res);
0d2602ca	493	page += sprintf(page, "active_queues=%u\n", atomic_read(&tags->active_queues));
320ae51f JA	494
	495	return page - orig_page;
	496	}