[linux.git] / lib / sbitmap.c

/*
 * Copyright (C) 2016 Facebook
 * Copyright (C) 2013-2014 Jens Axboe
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public
 * License v2 as published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program.  If not, see <https://www.gnu.org/licenses/>.
 */

#include <linux/sched.h>
#include <linux/random.h>
#include <linux/sbitmap.h>
#include <linux/seq_file.h>

/*
 * See if we have deferred clears that we can batch move
 */
static inline bool sbitmap_deferred_clear(struct sbitmap *sb, int index)
{
	unsigned long mask, val;
	bool ret = false;
	unsigned long flags;

	spin_lock_irqsave(&sb->map[index].swap_lock, flags);

	if (!sb->map[index].cleared)
		goto out_unlock;

	/*
	 * First get a stable cleared mask, setting the old mask to 0.
	 */
	do {
		mask = sb->map[index].cleared;
	} while (cmpxchg(&sb->map[index].cleared, mask, 0) != mask);

	/*
	 * Now clear the masked bits in our free word
	 */
	do {
		val = sb->map[index].word;
	} while (cmpxchg(&sb->map[index].word, val, val & ~mask) != val);

	ret = true;
out_unlock:
	spin_unlock_irqrestore(&sb->map[index].swap_lock, flags);
	return ret;
}

int sbitmap_init_node(struct sbitmap *sb, unsigned int depth, int shift,
		      gfp_t flags, int node)
{
	unsigned int bits_per_word;
	unsigned int i;

	if (shift < 0) {
		shift = ilog2(BITS_PER_LONG);
		/*
		 * If the bitmap is small, shrink the number of bits per word so
		 * we spread over a few cachelines, at least. If less than 4
		 * bits, just forget about it, it's not going to work optimally
		 * anyway.
		 */
		if (depth >= 4) {
			while ((4U << shift) > depth)
				shift--;
		}
	}
	bits_per_word = 1U << shift;
	if (bits_per_word > BITS_PER_LONG)
		return -EINVAL;

	sb->shift = shift;
	sb->depth = depth;
	sb->map_nr = DIV_ROUND_UP(sb->depth, bits_per_word);

	if (depth == 0) {
		sb->map = NULL;
		return 0;
	}

	sb->map = kcalloc_node(sb->map_nr, sizeof(*sb->map), flags, node);
	if (!sb->map)
		return -ENOMEM;

	for (i = 0; i < sb->map_nr; i++) {
		sb->map[i].depth = min(depth, bits_per_word);
		depth -= sb->map[i].depth;
		spin_lock_init(&sb->map[i].swap_lock);
	}
	return 0;
}
EXPORT_SYMBOL_GPL(sbitmap_init_node);

void sbitmap_resize(struct sbitmap *sb, unsigned int depth)
{
	unsigned int bits_per_word = 1U << sb->shift;
	unsigned int i;

	for (i = 0; i < sb->map_nr; i++)
		sbitmap_deferred_clear(sb, i);

	sb->depth = depth;
	sb->map_nr = DIV_ROUND_UP(sb->depth, bits_per_word);

	for (i = 0; i < sb->map_nr; i++) {
		sb->map[i].depth = min(depth, bits_per_word);
		depth -= sb->map[i].depth;
	}
}
EXPORT_SYMBOL_GPL(sbitmap_resize);

static int __sbitmap_get_word(unsigned long *word, unsigned long depth,
			      unsigned int hint, bool wrap)
{
	unsigned int orig_hint = hint;
	int nr;

	while (1) {
		nr = find_next_zero_bit(word, depth, hint);
		if (unlikely(nr >= depth)) {
			/*
			 * We started with an offset, and we didn't reset the
			 * offset to 0 in a failure case, so start from 0 to
			 * exhaust the map.
			 */
			if (orig_hint && hint && wrap) {
				hint = orig_hint = 0;
				continue;
			}
			return -1;
		}

		if (!test_and_set_bit_lock(nr, word))
			break;

		hint = nr + 1;
		if (hint >= depth - 1)
			hint = 0;
	}

	return nr;
}

static int sbitmap_find_bit_in_index(struct sbitmap *sb, int index,
				     unsigned int alloc_hint, bool round_robin)
{
	int nr;

	do {
		nr = __sbitmap_get_word(&sb->map[index].word,
					sb->map[index].depth, alloc_hint,
					!round_robin);
		if (nr != -1)
			break;
		if (!sbitmap_deferred_clear(sb, index))
			break;
	} while (1);

	return nr;
}

int sbitmap_get(struct sbitmap *sb, unsigned int alloc_hint, bool round_robin)
{
	unsigned int i, index;
	int nr = -1;

	index = SB_NR_TO_INDEX(sb, alloc_hint);

	/*
	 * Unless we're doing round robin tag allocation, just use the
	 * alloc_hint to find the right word index. No point in looping
	 * twice in find_next_zero_bit() for that case.
	 */
	if (round_robin)
		alloc_hint = SB_NR_TO_BIT(sb, alloc_hint);
	else
		alloc_hint = 0;

	for (i = 0; i < sb->map_nr; i++) {
		nr = sbitmap_find_bit_in_index(sb, index, alloc_hint,
						round_robin);
		if (nr != -1) {
			nr += index << sb->shift;
			break;
		}

		/* Jump to next index. */
		alloc_hint = 0;
		if (++index >= sb->map_nr)
			index = 0;
	}

	return nr;
}
EXPORT_SYMBOL_GPL(sbitmap_get);

int sbitmap_get_shallow(struct sbitmap *sb, unsigned int alloc_hint,
			unsigned long shallow_depth)
{
	unsigned int i, index;
	int nr = -1;

	index = SB_NR_TO_INDEX(sb, alloc_hint);

	for (i = 0; i < sb->map_nr; i++) {
again:
		nr = __sbitmap_get_word(&sb->map[index].word,
					min(sb->map[index].depth, shallow_depth),
					SB_NR_TO_BIT(sb, alloc_hint), true);
		if (nr != -1) {
			nr += index << sb->shift;
			break;
		}

		if (sbitmap_deferred_clear(sb, index))
			goto again;

		/* Jump to next index. */
		index++;
		alloc_hint = index << sb->shift;

		if (index >= sb->map_nr) {
			index = 0;
			alloc_hint = 0;
		}
	}

	return nr;
}
EXPORT_SYMBOL_GPL(sbitmap_get_shallow);

bool sbitmap_any_bit_set(const struct sbitmap *sb)
{
	unsigned int i;

	for (i = 0; i < sb->map_nr; i++) {
		if (sb->map[i].word & ~sb->map[i].cleared)
			return true;
	}
	return false;
}
EXPORT_SYMBOL_GPL(sbitmap_any_bit_set);

bool sbitmap_any_bit_clear(const struct sbitmap *sb)
{
	unsigned int i;

	for (i = 0; i < sb->map_nr; i++) {
		const struct sbitmap_word *word = &sb->map[i];
		unsigned long mask = word->word & ~word->cleared;
		unsigned long ret;

		ret = find_first_zero_bit(&mask, word->depth);
		if (ret < word->depth)
			return true;
	}
	return false;
}
EXPORT_SYMBOL_GPL(sbitmap_any_bit_clear);

static unsigned int __sbitmap_weight(const struct sbitmap *sb, bool set)
{
	unsigned int i, weight = 0;

	for (i = 0; i < sb->map_nr; i++) {
		const struct sbitmap_word *word = &sb->map[i];

		if (set)
			weight += bitmap_weight(&word->word, word->depth);
		else
			weight += bitmap_weight(&word->cleared, word->depth);
	}
	return weight;
}

static unsigned int sbitmap_weight(const struct sbitmap *sb)
{
	return __sbitmap_weight(sb, true);
}

static unsigned int sbitmap_cleared(const struct sbitmap *sb)
{
	return __sbitmap_weight(sb, false);
}

void sbitmap_show(struct sbitmap *sb, struct seq_file *m)
{
	seq_printf(m, "depth=%u\n", sb->depth);
	seq_printf(m, "busy=%u\n", sbitmap_weight(sb) - sbitmap_cleared(sb));
	seq_printf(m, "cleared=%u\n", sbitmap_cleared(sb));
	seq_printf(m, "bits_per_word=%u\n", 1U << sb->shift);
	seq_printf(m, "map_nr=%u\n", sb->map_nr);
}
EXPORT_SYMBOL_GPL(sbitmap_show);

static inline void emit_byte(struct seq_file *m, unsigned int offset, u8 byte)
{
	if ((offset & 0xf) == 0) {
		if (offset != 0)
			seq_putc(m, '\n');
		seq_printf(m, "%08x:", offset);
	}
	if ((offset & 0x1) == 0)
		seq_putc(m, ' ');
	seq_printf(m, "%02x", byte);
}

void sbitmap_bitmap_show(struct sbitmap *sb, struct seq_file *m)
{
	u8 byte = 0;
	unsigned int byte_bits = 0;
	unsigned int offset = 0;
	int i;

	for (i = 0; i < sb->map_nr; i++) {
		unsigned long word = READ_ONCE(sb->map[i].word);
		unsigned int word_bits = READ_ONCE(sb->map[i].depth);

		while (word_bits > 0) {
			unsigned int bits = min(8 - byte_bits, word_bits);

			byte |= (word & (BIT(bits) - 1)) << byte_bits;
			byte_bits += bits;
			if (byte_bits == 8) {
				emit_byte(m, offset, byte);
				byte = 0;
				byte_bits = 0;
				offset++;
			}
			word >>= bits;
			word_bits -= bits;
		}
	}
	if (byte_bits) {
		emit_byte(m, offset, byte);
		offset++;
	}
	if (offset)
		seq_putc(m, '\n');
}
EXPORT_SYMBOL_GPL(sbitmap_bitmap_show);

static unsigned int sbq_calc_wake_batch(struct sbitmap_queue *sbq,
					unsigned int depth)
{
	unsigned int wake_batch;
	unsigned int shallow_depth;

	/*
	 * For each batch, we wake up one queue. We need to make sure that our
	 * batch size is small enough that the full depth of the bitmap,
	 * potentially limited by a shallow depth, is enough to wake up all of
	 * the queues.
	 *
	 * Each full word of the bitmap has bits_per_word bits, and there might
	 * be a partial word. There are depth / bits_per_word full words and
	 * depth % bits_per_word bits left over. In bitwise arithmetic:
	 *
	 * bits_per_word = 1 << shift
	 * depth / bits_per_word = depth >> shift
	 * depth % bits_per_word = depth & ((1 << shift) - 1)
	 *
	 * Each word can be limited to sbq->min_shallow_depth bits.
	 */
	shallow_depth = min(1U << sbq->sb.shift, sbq->min_shallow_depth);
	depth = ((depth >> sbq->sb.shift) * shallow_depth +
		 min(depth & ((1U << sbq->sb.shift) - 1), shallow_depth));
	wake_batch = clamp_t(unsigned int, depth / SBQ_WAIT_QUEUES, 1,
			     SBQ_WAKE_BATCH);

	return wake_batch;
}

int sbitmap_queue_init_node(struct sbitmap_queue *sbq, unsigned int depth,
			    int shift, bool round_robin, gfp_t flags, int node)
{
	int ret;
	int i;

	ret = sbitmap_init_node(&sbq->sb, depth, shift, flags, node);
	if (ret)
		return ret;

	sbq->alloc_hint = alloc_percpu_gfp(unsigned int, flags);
	if (!sbq->alloc_hint) {
		sbitmap_free(&sbq->sb);
		return -ENOMEM;
	}

	if (depth && !round_robin) {
		for_each_possible_cpu(i)
			*per_cpu_ptr(sbq->alloc_hint, i) = prandom_u32() % depth;
	}

	sbq->min_shallow_depth = UINT_MAX;
	sbq->wake_batch = sbq_calc_wake_batch(sbq, depth);
	atomic_set(&sbq->wake_index, 0);
	atomic_set(&sbq->ws_active, 0);

	sbq->ws = kzalloc_node(SBQ_WAIT_QUEUES * sizeof(*sbq->ws), flags, node);
	if (!sbq->ws) {
		free_percpu(sbq->alloc_hint);
		sbitmap_free(&sbq->sb);
		return -ENOMEM;
	}

	for (i = 0; i < SBQ_WAIT_QUEUES; i++) {
		init_waitqueue_head(&sbq->ws[i].wait);
		atomic_set(&sbq->ws[i].wait_cnt, sbq->wake_batch);
	}

	sbq->round_robin = round_robin;
	return 0;
}
EXPORT_SYMBOL_GPL(sbitmap_queue_init_node);

static void sbitmap_queue_update_wake_batch(struct sbitmap_queue *sbq,
					    unsigned int depth)
{
	unsigned int wake_batch = sbq_calc_wake_batch(sbq, depth);
	int i;

	if (sbq->wake_batch != wake_batch) {
		WRITE_ONCE(sbq->wake_batch, wake_batch);
		/*
		 * Pairs with the memory barrier in sbitmap_queue_wake_up()
		 * to ensure that the batch size is updated before the wait
		 * counts.
		 */
		smp_mb__before_atomic();
		for (i = 0; i < SBQ_WAIT_QUEUES; i++)
			atomic_set(&sbq->ws[i].wait_cnt, 1);
	}
}

void sbitmap_queue_resize(struct sbitmap_queue *sbq, unsigned int depth)
{
	sbitmap_queue_update_wake_batch(sbq, depth);
	sbitmap_resize(&sbq->sb, depth);
}
EXPORT_SYMBOL_GPL(sbitmap_queue_resize);

int __sbitmap_queue_get(struct sbitmap_queue *sbq)
{
	unsigned int hint, depth;
	int nr;

	hint = this_cpu_read(*sbq->alloc_hint);
	depth = READ_ONCE(sbq->sb.depth);
	if (unlikely(hint >= depth)) {
		hint = depth ? prandom_u32() % depth : 0;
		this_cpu_write(*sbq->alloc_hint, hint);
	}
	nr = sbitmap_get(&sbq->sb, hint, sbq->round_robin);

	if (nr == -1) {
		/* If the map is full, a hint won't do us much good. */
		this_cpu_write(*sbq->alloc_hint, 0);
	} else if (nr == hint || unlikely(sbq->round_robin)) {
		/* Only update the hint if we used it. */
		hint = nr + 1;
		if (hint >= depth - 1)
			hint = 0;
		this_cpu_write(*sbq->alloc_hint, hint);
	}

	return nr;
}
EXPORT_SYMBOL_GPL(__sbitmap_queue_get);

int __sbitmap_queue_get_shallow(struct sbitmap_queue *sbq,
				unsigned int shallow_depth)
{
	unsigned int hint, depth;
	int nr;

	WARN_ON_ONCE(shallow_depth < sbq->min_shallow_depth);

	hint = this_cpu_read(*sbq->alloc_hint);
	depth = READ_ONCE(sbq->sb.depth);
	if (unlikely(hint >= depth)) {
		hint = depth ? prandom_u32() % depth : 0;
		this_cpu_write(*sbq->alloc_hint, hint);
	}
	nr = sbitmap_get_shallow(&sbq->sb, hint, shallow_depth);

	if (nr == -1) {
		/* If the map is full, a hint won't do us much good. */
		this_cpu_write(*sbq->alloc_hint, 0);
	} else if (nr == hint || unlikely(sbq->round_robin)) {
		/* Only update the hint if we used it. */
		hint = nr + 1;
		if (hint >= depth - 1)
			hint = 0;
		this_cpu_write(*sbq->alloc_hint, hint);
	}

	return nr;
}
EXPORT_SYMBOL_GPL(__sbitmap_queue_get_shallow);

void sbitmap_queue_min_shallow_depth(struct sbitmap_queue *sbq,
				     unsigned int min_shallow_depth)
{
	sbq->min_shallow_depth = min_shallow_depth;
	sbitmap_queue_update_wake_batch(sbq, sbq->sb.depth);
}
EXPORT_SYMBOL_GPL(sbitmap_queue_min_shallow_depth);

static struct sbq_wait_state *sbq_wake_ptr(struct sbitmap_queue *sbq)
{
	int i, wake_index;

	if (!atomic_read(&sbq->ws_active))
		return NULL;

	wake_index = atomic_read(&sbq->wake_index);
	for (i = 0; i < SBQ_WAIT_QUEUES; i++) {
		struct sbq_wait_state *ws = &sbq->ws[wake_index];

		if (waitqueue_active(&ws->wait)) {
			int o = atomic_read(&sbq->wake_index);

			if (wake_index != o)
				atomic_cmpxchg(&sbq->wake_index, o, wake_index);
			return ws;
		}

		wake_index = sbq_index_inc(wake_index);
	}

	return NULL;
}

static bool __sbq_wake_up(struct sbitmap_queue *sbq)
{
	struct sbq_wait_state *ws;
	unsigned int wake_batch;
	int wait_cnt;

	ws = sbq_wake_ptr(sbq);
	if (!ws)
		return false;

	wait_cnt = atomic_dec_return(&ws->wait_cnt);
	if (wait_cnt <= 0) {
		int ret;

		wake_batch = READ_ONCE(sbq->wake_batch);

		/*
		 * Pairs with the memory barrier in sbitmap_queue_resize() to
		 * ensure that we see the batch size update before the wait
		 * count is reset.
		 */
		smp_mb__before_atomic();

		/*
		 * For concurrent callers of this, the one that failed the
		 * atomic_cmpxhcg() race should call this function again
		 * to wakeup a new batch on a different 'ws'.
		 */
		ret = atomic_cmpxchg(&ws->wait_cnt, wait_cnt, wake_batch);
		if (ret == wait_cnt) {
			sbq_index_atomic_inc(&sbq->wake_index);
			wake_up_nr(&ws->wait, wake_batch);
			return false;
		}

		return true;
	}

	return false;
}

void sbitmap_queue_wake_up(struct sbitmap_queue *sbq)
{
	while (__sbq_wake_up(sbq))
		;
}
EXPORT_SYMBOL_GPL(sbitmap_queue_wake_up);

void sbitmap_queue_clear(struct sbitmap_queue *sbq, unsigned int nr,
			 unsigned int cpu)
{
	sbitmap_deferred_clear_bit(&sbq->sb, nr);

	/*
	 * Pairs with the memory barrier in set_current_state() to ensure the
	 * proper ordering of clear_bit_unlock()/waitqueue_active() in the waker
	 * and test_and_set_bit_lock()/prepare_to_wait()/finish_wait() in the
	 * waiter. See the comment on waitqueue_active().
	 */
	smp_mb__after_atomic();
	sbitmap_queue_wake_up(sbq);

	if (likely(!sbq->round_robin && nr < sbq->sb.depth))
		*per_cpu_ptr(sbq->alloc_hint, cpu) = nr;
}
EXPORT_SYMBOL_GPL(sbitmap_queue_clear);

void sbitmap_queue_wake_all(struct sbitmap_queue *sbq)
{
	int i, wake_index;

	/*
	 * Pairs with the memory barrier in set_current_state() like in
	 * sbitmap_queue_wake_up().
	 */
	smp_mb();
	wake_index = atomic_read(&sbq->wake_index);
	for (i = 0; i < SBQ_WAIT_QUEUES; i++) {
		struct sbq_wait_state *ws = &sbq->ws[wake_index];

		if (waitqueue_active(&ws->wait))
			wake_up(&ws->wait);

		wake_index = sbq_index_inc(wake_index);
	}
}
EXPORT_SYMBOL_GPL(sbitmap_queue_wake_all);

void sbitmap_queue_show(struct sbitmap_queue *sbq, struct seq_file *m)
{
	bool first;
	int i;

	sbitmap_show(&sbq->sb, m);

	seq_puts(m, "alloc_hint={");
	first = true;
	for_each_possible_cpu(i) {
		if (!first)
			seq_puts(m, ", ");
		first = false;
		seq_printf(m, "%u", *per_cpu_ptr(sbq->alloc_hint, i));
	}
	seq_puts(m, "}\n");

	seq_printf(m, "wake_batch=%u\n", sbq->wake_batch);
	seq_printf(m, "wake_index=%d\n", atomic_read(&sbq->wake_index));
	seq_printf(m, "ws_active=%d\n", atomic_read(&sbq->ws_active));

	seq_puts(m, "ws={\n");
	for (i = 0; i < SBQ_WAIT_QUEUES; i++) {
		struct sbq_wait_state *ws = &sbq->ws[i];

		seq_printf(m, "\t{.wait_cnt=%d, .wait=%s},\n",
			   atomic_read(&ws->wait_cnt),
			   waitqueue_active(&ws->wait) ? "active" : "inactive");
	}
	seq_puts(m, "}\n");

	seq_printf(m, "round_robin=%d\n", sbq->round_robin);
	seq_printf(m, "min_shallow_depth=%u\n", sbq->min_shallow_depth);
}
EXPORT_SYMBOL_GPL(sbitmap_queue_show);

void sbitmap_add_wait_queue(struct sbitmap_queue *sbq,
			    struct sbq_wait_state *ws,
			    struct sbq_wait *sbq_wait)
{
	if (!sbq_wait->sbq) {
		sbq_wait->sbq = sbq;
		atomic_inc(&sbq->ws_active);
	}
	add_wait_queue(&ws->wait, &sbq_wait->wait);
}
EXPORT_SYMBOL_GPL(sbitmap_add_wait_queue);

void sbitmap_del_wait_queue(struct sbq_wait *sbq_wait)
{
	list_del_init(&sbq_wait->wait.entry);
	if (sbq_wait->sbq) {
		atomic_dec(&sbq_wait->sbq->ws_active);
		sbq_wait->sbq = NULL;
	}
}
EXPORT_SYMBOL_GPL(sbitmap_del_wait_queue);

void sbitmap_prepare_to_wait(struct sbitmap_queue *sbq,
			     struct sbq_wait_state *ws,
			     struct sbq_wait *sbq_wait, int state)
{
	if (!sbq_wait->sbq) {
		atomic_inc(&sbq->ws_active);
		sbq_wait->sbq = sbq;
	}
	prepare_to_wait_exclusive(&ws->wait, &sbq_wait->wait, state);
}
EXPORT_SYMBOL_GPL(sbitmap_prepare_to_wait);

void sbitmap_finish_wait(struct sbitmap_queue *sbq, struct sbq_wait_state *ws,
			 struct sbq_wait *sbq_wait)
{
	finish_wait(&ws->wait, &sbq_wait->wait);
	if (sbq_wait->sbq) {
		atomic_dec(&sbq->ws_active);
		sbq_wait->sbq = NULL;
	}
}
EXPORT_SYMBOL_GPL(sbitmap_finish_wait);
Commit	Line	Data
88459642 OS	1	/*
	2	* Copyright (C) 2016 Facebook
	3	* Copyright (C) 2013-2014 Jens Axboe
	4	*
	5	* This program is free software; you can redistribute it and/or
	6	* modify it under the terms of the GNU General Public
	7	* License v2 as published by the Free Software Foundation.
	8	*
	9	* This program is distributed in the hope that it will be useful,
	10	* but WITHOUT ANY WARRANTY; without even the implied warranty of
	11	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	12	* General Public License for more details.
	13	*
	14	* You should have received a copy of the GNU General Public License
	15	* along with this program. If not, see <https://www.gnu.org/licenses/>.
	16	*/
	17
af8601ad	18	#include <linux/sched.h>
98d95416	19	#include <linux/random.h>
88459642	20	#include <linux/sbitmap.h>
24af1ccf	21	#include <linux/seq_file.h>
88459642	22
b2dbff1b JA	23	/*
	24	* See if we have deferred clears that we can batch move
	25	*/
	26	static inline bool sbitmap_deferred_clear(struct sbitmap *sb, int index)
	27	{
	28	unsigned long mask, val;
b2dbff1b	29	bool ret = false;
fe76fc6a	30	unsigned long flags;
b2dbff1b	31
fe76fc6a	32	spin_lock_irqsave(&sb->map[index].swap_lock, flags);
b2dbff1b JA	33
	34	if (!sb->map[index].cleared)
	35	goto out_unlock;
	36
	37	/*
	38	* First get a stable cleared mask, setting the old mask to 0.
	39	*/
	40	do {
	41	mask = sb->map[index].cleared;
	42	} while (cmpxchg(&sb->map[index].cleared, mask, 0) != mask);
	43
	44	/*
	45	* Now clear the masked bits in our free word
	46	*/
	47	do {
	48	val = sb->map[index].word;
	49	} while (cmpxchg(&sb->map[index].word, val, val & ~mask) != val);
	50
	51	ret = true;
	52	out_unlock:
fe76fc6a	53	spin_unlock_irqrestore(&sb->map[index].swap_lock, flags);
b2dbff1b JA	54	return ret;
	55	}
	56
88459642 OS	57	int sbitmap_init_node(struct sbitmap *sb, unsigned int depth, int shift,
	58	gfp_t flags, int node)
	59	{
	60	unsigned int bits_per_word;
	61	unsigned int i;
	62
	63	if (shift < 0) {
	64	shift = ilog2(BITS_PER_LONG);
	65	/*
	66	* If the bitmap is small, shrink the number of bits per word so
	67	* we spread over a few cachelines, at least. If less than 4
	68	* bits, just forget about it, it's not going to work optimally
	69	* anyway.
	70	*/
	71	if (depth >= 4) {
	72	while ((4U << shift) > depth)
	73	shift--;
	74	}
	75	}
	76	bits_per_word = 1U << shift;
	77	if (bits_per_word > BITS_PER_LONG)
	78	return -EINVAL;
	79
	80	sb->shift = shift;
	81	sb->depth = depth;
	82	sb->map_nr = DIV_ROUND_UP(sb->depth, bits_per_word);
	83
	84	if (depth == 0) {
	85	sb->map = NULL;
	86	return 0;
	87	}
	88
590b5b7d	89	sb->map = kcalloc_node(sb->map_nr, sizeof(*sb->map), flags, node);
88459642 OS	90	if (!sb->map)
	91	return -ENOMEM;
	92
	93	for (i = 0; i < sb->map_nr; i++) {
	94	sb->map[i].depth = min(depth, bits_per_word);
	95	depth -= sb->map[i].depth;
ea86ea2c	96	spin_lock_init(&sb->map[i].swap_lock);
88459642 OS	97	}
	98	return 0;
	99	}
	100	EXPORT_SYMBOL_GPL(sbitmap_init_node);
	101
	102	void sbitmap_resize(struct sbitmap *sb, unsigned int depth)
	103	{
	104	unsigned int bits_per_word = 1U << sb->shift;
	105	unsigned int i;
	106
b2dbff1b JA	107	for (i = 0; i < sb->map_nr; i++)
	108	sbitmap_deferred_clear(sb, i);
	109
88459642 OS	110	sb->depth = depth;
	111	sb->map_nr = DIV_ROUND_UP(sb->depth, bits_per_word);
	112
	113	for (i = 0; i < sb->map_nr; i++) {
	114	sb->map[i].depth = min(depth, bits_per_word);
	115	depth -= sb->map[i].depth;
	116	}
	117	}
	118	EXPORT_SYMBOL_GPL(sbitmap_resize);
	119
c05e6673 OS	120	static int __sbitmap_get_word(unsigned long *word, unsigned long depth,
c05e6673 OS	121	unsigned int hint, bool wrap)
88459642 OS	122	{
	123	unsigned int orig_hint = hint;
	124	int nr;
	125
	126	while (1) {
c05e6673 OS	127	nr = find_next_zero_bit(word, depth, hint);
c05e6673 OS	128	if (unlikely(nr >= depth)) {
88459642 OS	129	/*
	130	* We started with an offset, and we didn't reset the
	131	* offset to 0 in a failure case, so start from 0 to
	132	* exhaust the map.
	133	*/
	134	if (orig_hint && hint && wrap) {
	135	hint = orig_hint = 0;
	136	continue;
	137	}
	138	return -1;
	139	}
	140
4ace53f1	141	if (!test_and_set_bit_lock(nr, word))
88459642 OS	142	break;
	143
	144	hint = nr + 1;
c05e6673	145	if (hint >= depth - 1)
88459642 OS	146	hint = 0;
	147	}
	148
	149	return nr;
	150	}
	151
ea86ea2c JA	152	static int sbitmap_find_bit_in_index(struct sbitmap *sb, int index,
	153	unsigned int alloc_hint, bool round_robin)
	154	{
	155	int nr;
	156
	157	do {
	158	nr = __sbitmap_get_word(&sb->map[index].word,
	159	sb->map[index].depth, alloc_hint,
	160	!round_robin);
	161	if (nr != -1)
	162	break;
	163	if (!sbitmap_deferred_clear(sb, index))
	164	break;
	165	} while (1);
	166
	167	return nr;
	168	}
	169
88459642 OS	170	int sbitmap_get(struct sbitmap *sb, unsigned int alloc_hint, bool round_robin)
	171	{
	172	unsigned int i, index;
	173	int nr = -1;
	174
	175	index = SB_NR_TO_INDEX(sb, alloc_hint);
	176
27fae429 JA	177	/*
	178	* Unless we're doing round robin tag allocation, just use the
	179	* alloc_hint to find the right word index. No point in looping
	180	* twice in find_next_zero_bit() for that case.
	181	*/
	182	if (round_robin)
	183	alloc_hint = SB_NR_TO_BIT(sb, alloc_hint);
	184	else
	185	alloc_hint = 0;
	186
88459642	187	for (i = 0; i < sb->map_nr; i++) {
ea86ea2c JA	188	nr = sbitmap_find_bit_in_index(sb, index, alloc_hint,
ea86ea2c JA	189	round_robin);
88459642 OS	190	if (nr != -1) {
	191	nr += index << sb->shift;
	192	break;
	193	}
	194
	195	/* Jump to next index. */
27fae429 JA	196	alloc_hint = 0;
27fae429 JA	197	if (++index >= sb->map_nr)
88459642	198	index = 0;
88459642 OS	199	}
	200
	201	return nr;
	202	}
	203	EXPORT_SYMBOL_GPL(sbitmap_get);
	204
c05e6673 OS	205	int sbitmap_get_shallow(struct sbitmap *sb, unsigned int alloc_hint,
	206	unsigned long shallow_depth)
	207	{
	208	unsigned int i, index;
	209	int nr = -1;
	210
	211	index = SB_NR_TO_INDEX(sb, alloc_hint);
	212
	213	for (i = 0; i < sb->map_nr; i++) {
b2dbff1b	214	again:
c05e6673 OS	215	nr = __sbitmap_get_word(&sb->map[index].word,
	216	min(sb->map[index].depth, shallow_depth),
	217	SB_NR_TO_BIT(sb, alloc_hint), true);
	218	if (nr != -1) {
	219	nr += index << sb->shift;
	220	break;
	221	}
	222
b2dbff1b JA	223	if (sbitmap_deferred_clear(sb, index))
	224	goto again;
	225
c05e6673 OS	226	/* Jump to next index. */
	227	index++;
	228	alloc_hint = index << sb->shift;
	229
	230	if (index >= sb->map_nr) {
	231	index = 0;
	232	alloc_hint = 0;
	233	}
	234	}
	235
	236	return nr;
	237	}
	238	EXPORT_SYMBOL_GPL(sbitmap_get_shallow);
	239
88459642 OS	240	bool sbitmap_any_bit_set(const struct sbitmap *sb)
	241	{
	242	unsigned int i;
	243
	244	for (i = 0; i < sb->map_nr; i++) {
b2dbff1b	245	if (sb->map[i].word & ~sb->map[i].cleared)
88459642 OS	246	return true;
	247	}
	248	return false;
	249	}
	250	EXPORT_SYMBOL_GPL(sbitmap_any_bit_set);
	251
	252	bool sbitmap_any_bit_clear(const struct sbitmap *sb)
	253	{
	254	unsigned int i;
	255
	256	for (i = 0; i < sb->map_nr; i++) {
	257	const struct sbitmap_word *word = &sb->map[i];
b2dbff1b	258	unsigned long mask = word->word & ~word->cleared;
88459642 OS	259	unsigned long ret;
88459642 OS	260
b2dbff1b	261	ret = find_first_zero_bit(&mask, word->depth);
88459642 OS	262	if (ret < word->depth)
	263	return true;
	264	}
	265	return false;
	266	}
	267	EXPORT_SYMBOL_GPL(sbitmap_any_bit_clear);
	268
ea86ea2c	269	static unsigned int __sbitmap_weight(const struct sbitmap *sb, bool set)
88459642	270	{
60658e0d	271	unsigned int i, weight = 0;
88459642 OS	272
	273	for (i = 0; i < sb->map_nr; i++) {
	274	const struct sbitmap_word *word = &sb->map[i];
	275
ea86ea2c JA	276	if (set)
	277	weight += bitmap_weight(&word->word, word->depth);
	278	else
	279	weight += bitmap_weight(&word->cleared, word->depth);
88459642 OS	280	}
	281	return weight;
	282	}
ea86ea2c JA	283
	284	static unsigned int sbitmap_weight(const struct sbitmap *sb)
	285	{
	286	return __sbitmap_weight(sb, true);
	287	}
	288
	289	static unsigned int sbitmap_cleared(const struct sbitmap *sb)
	290	{
	291	return __sbitmap_weight(sb, false);
	292	}
88459642	293
24af1ccf OS	294	void sbitmap_show(struct sbitmap sb, struct seq_file m)
	295	{
	296	seq_printf(m, "depth=%u\n", sb->depth);
ea86ea2c JA	297	seq_printf(m, "busy=%u\n", sbitmap_weight(sb) - sbitmap_cleared(sb));
ea86ea2c JA	298	seq_printf(m, "cleared=%u\n", sbitmap_cleared(sb));
24af1ccf OS	299	seq_printf(m, "bits_per_word=%u\n", 1U << sb->shift);
	300	seq_printf(m, "map_nr=%u\n", sb->map_nr);
	301	}
	302	EXPORT_SYMBOL_GPL(sbitmap_show);
	303
	304	static inline void emit_byte(struct seq_file *m, unsigned int offset, u8 byte)
	305	{
	306	if ((offset & 0xf) == 0) {
	307	if (offset != 0)
	308	seq_putc(m, '\n');
	309	seq_printf(m, "%08x:", offset);
	310	}
	311	if ((offset & 0x1) == 0)
	312	seq_putc(m, ' ');
	313	seq_printf(m, "%02x", byte);
	314	}
	315
	316	void sbitmap_bitmap_show(struct sbitmap sb, struct seq_file m)
	317	{
	318	u8 byte = 0;
	319	unsigned int byte_bits = 0;
	320	unsigned int offset = 0;
	321	int i;
	322
	323	for (i = 0; i < sb->map_nr; i++) {
	324	unsigned long word = READ_ONCE(sb->map[i].word);
	325	unsigned int word_bits = READ_ONCE(sb->map[i].depth);
	326
	327	while (word_bits > 0) {
	328	unsigned int bits = min(8 - byte_bits, word_bits);
	329
	330	byte \|= (word & (BIT(bits) - 1)) << byte_bits;
	331	byte_bits += bits;
	332	if (byte_bits == 8) {
	333	emit_byte(m, offset, byte);
	334	byte = 0;
	335	byte_bits = 0;
	336	offset++;
	337	}
	338	word >>= bits;
	339	word_bits -= bits;
	340	}
	341	}
	342	if (byte_bits) {
	343	emit_byte(m, offset, byte);
	344	offset++;
	345	}
	346	if (offset)
	347	seq_putc(m, '\n');
	348	}
	349	EXPORT_SYMBOL_GPL(sbitmap_bitmap_show);
	350
a3275539 OS	351	static unsigned int sbq_calc_wake_batch(struct sbitmap_queue *sbq,
a3275539 OS	352	unsigned int depth)
88459642 OS	353	{
88459642 OS	354	unsigned int wake_batch;
a3275539	355	unsigned int shallow_depth;
88459642 OS	356
	357	/*
	358	* For each batch, we wake up one queue. We need to make sure that our
a3275539 OS	359	* batch size is small enough that the full depth of the bitmap,
	360	* potentially limited by a shallow depth, is enough to wake up all of
	361	* the queues.
	362	*
	363	* Each full word of the bitmap has bits_per_word bits, and there might
	364	* be a partial word. There are depth / bits_per_word full words and
	365	* depth % bits_per_word bits left over. In bitwise arithmetic:
	366	*
	367	* bits_per_word = 1 << shift
	368	* depth / bits_per_word = depth >> shift
	369	* depth % bits_per_word = depth & ((1 << shift) - 1)
	370	*
	371	* Each word can be limited to sbq->min_shallow_depth bits.
88459642	372	*/
a3275539 OS	373	shallow_depth = min(1U << sbq->sb.shift, sbq->min_shallow_depth);
	374	depth = ((depth >> sbq->sb.shift) * shallow_depth +
	375	min(depth & ((1U << sbq->sb.shift) - 1), shallow_depth));
	376	wake_batch = clamp_t(unsigned int, depth / SBQ_WAIT_QUEUES, 1,
	377	SBQ_WAKE_BATCH);
88459642 OS	378
	379	return wake_batch;
	380	}
	381
	382	int sbitmap_queue_init_node(struct sbitmap_queue *sbq, unsigned int depth,
f4a644db	383	int shift, bool round_robin, gfp_t flags, int node)
88459642 OS	384	{
	385	int ret;
	386	int i;
	387
	388	ret = sbitmap_init_node(&sbq->sb, depth, shift, flags, node);
	389	if (ret)
	390	return ret;
	391
40aabb67 OS	392	sbq->alloc_hint = alloc_percpu_gfp(unsigned int, flags);
	393	if (!sbq->alloc_hint) {
	394	sbitmap_free(&sbq->sb);
	395	return -ENOMEM;
	396	}
	397
98d95416 OS	398	if (depth && !round_robin) {
	399	for_each_possible_cpu(i)
	400	*per_cpu_ptr(sbq->alloc_hint, i) = prandom_u32() % depth;
	401	}
	402
a3275539 OS	403	sbq->min_shallow_depth = UINT_MAX;
a3275539 OS	404	sbq->wake_batch = sbq_calc_wake_batch(sbq, depth);
88459642	405	atomic_set(&sbq->wake_index, 0);
5d2ee712	406	atomic_set(&sbq->ws_active, 0);
88459642	407
48e28166	408	sbq->ws = kzalloc_node(SBQ_WAIT_QUEUES * sizeof(*sbq->ws), flags, node);
88459642	409	if (!sbq->ws) {
40aabb67	410	free_percpu(sbq->alloc_hint);
88459642 OS	411	sbitmap_free(&sbq->sb);
	412	return -ENOMEM;
	413	}
	414
	415	for (i = 0; i < SBQ_WAIT_QUEUES; i++) {
	416	init_waitqueue_head(&sbq->ws[i].wait);
	417	atomic_set(&sbq->ws[i].wait_cnt, sbq->wake_batch);
	418	}
f4a644db OS	419
f4a644db OS	420	sbq->round_robin = round_robin;
88459642 OS	421	return 0;
	422	}
	423	EXPORT_SYMBOL_GPL(sbitmap_queue_init_node);
	424
a3275539 OS	425	static void sbitmap_queue_update_wake_batch(struct sbitmap_queue *sbq,
a3275539 OS	426	unsigned int depth)
88459642	427	{
a3275539	428	unsigned int wake_batch = sbq_calc_wake_batch(sbq, depth);
6c0ca7ae OS	429	int i;
	430
	431	if (sbq->wake_batch != wake_batch) {
	432	WRITE_ONCE(sbq->wake_batch, wake_batch);
	433	/*
e6fc4649 ML	434	* Pairs with the memory barrier in sbitmap_queue_wake_up()
	435	* to ensure that the batch size is updated before the wait
	436	* counts.
6c0ca7ae OS	437	*/
	438	smp_mb__before_atomic();
	439	for (i = 0; i < SBQ_WAIT_QUEUES; i++)
	440	atomic_set(&sbq->ws[i].wait_cnt, 1);
	441	}
a3275539 OS	442	}
	443
	444	void sbitmap_queue_resize(struct sbitmap_queue *sbq, unsigned int depth)
	445	{
	446	sbitmap_queue_update_wake_batch(sbq, depth);
88459642 OS	447	sbitmap_resize(&sbq->sb, depth);
	448	}
	449	EXPORT_SYMBOL_GPL(sbitmap_queue_resize);
	450
f4a644db	451	int __sbitmap_queue_get(struct sbitmap_queue *sbq)
40aabb67	452	{
05fd095d	453	unsigned int hint, depth;
40aabb67 OS	454	int nr;
	455
	456	hint = this_cpu_read(*sbq->alloc_hint);
05fd095d OS	457	depth = READ_ONCE(sbq->sb.depth);
	458	if (unlikely(hint >= depth)) {
	459	hint = depth ? prandom_u32() % depth : 0;
	460	this_cpu_write(*sbq->alloc_hint, hint);
	461	}
f4a644db	462	nr = sbitmap_get(&sbq->sb, hint, sbq->round_robin);
40aabb67 OS	463
	464	if (nr == -1) {
	465	/* If the map is full, a hint won't do us much good. */
	466	this_cpu_write(*sbq->alloc_hint, 0);
f4a644db	467	} else if (nr == hint \|\| unlikely(sbq->round_robin)) {
40aabb67 OS	468	/* Only update the hint if we used it. */
40aabb67 OS	469	hint = nr + 1;
05fd095d	470	if (hint >= depth - 1)
40aabb67 OS	471	hint = 0;
	472	this_cpu_write(*sbq->alloc_hint, hint);
	473	}
	474
	475	return nr;
	476	}
	477	EXPORT_SYMBOL_GPL(__sbitmap_queue_get);
	478
c05e6673 OS	479	int __sbitmap_queue_get_shallow(struct sbitmap_queue *sbq,
	480	unsigned int shallow_depth)
	481	{
	482	unsigned int hint, depth;
	483	int nr;
	484
61445b56 OS	485	WARN_ON_ONCE(shallow_depth < sbq->min_shallow_depth);
61445b56 OS	486
c05e6673 OS	487	hint = this_cpu_read(*sbq->alloc_hint);
	488	depth = READ_ONCE(sbq->sb.depth);
	489	if (unlikely(hint >= depth)) {
	490	hint = depth ? prandom_u32() % depth : 0;
	491	this_cpu_write(*sbq->alloc_hint, hint);
	492	}
	493	nr = sbitmap_get_shallow(&sbq->sb, hint, shallow_depth);
	494
	495	if (nr == -1) {
	496	/* If the map is full, a hint won't do us much good. */
	497	this_cpu_write(*sbq->alloc_hint, 0);
	498	} else if (nr == hint \|\| unlikely(sbq->round_robin)) {
	499	/* Only update the hint if we used it. */
	500	hint = nr + 1;
	501	if (hint >= depth - 1)
	502	hint = 0;
	503	this_cpu_write(*sbq->alloc_hint, hint);
	504	}
	505
	506	return nr;
	507	}
	508	EXPORT_SYMBOL_GPL(__sbitmap_queue_get_shallow);
	509
a3275539 OS	510	void sbitmap_queue_min_shallow_depth(struct sbitmap_queue *sbq,
	511	unsigned int min_shallow_depth)
	512	{
	513	sbq->min_shallow_depth = min_shallow_depth;
	514	sbitmap_queue_update_wake_batch(sbq, sbq->sb.depth);
	515	}
	516	EXPORT_SYMBOL_GPL(sbitmap_queue_min_shallow_depth);
	517
88459642 OS	518	static struct sbq_wait_state sbq_wake_ptr(struct sbitmap_queue sbq)
	519	{
	520	int i, wake_index;
	521
5d2ee712 JA	522	if (!atomic_read(&sbq->ws_active))
	523	return NULL;
	524
88459642 OS	525	wake_index = atomic_read(&sbq->wake_index);
	526	for (i = 0; i < SBQ_WAIT_QUEUES; i++) {
	527	struct sbq_wait_state *ws = &sbq->ws[wake_index];
	528
	529	if (waitqueue_active(&ws->wait)) {
	530	int o = atomic_read(&sbq->wake_index);
	531
	532	if (wake_index != o)
	533	atomic_cmpxchg(&sbq->wake_index, o, wake_index);
	534	return ws;
	535	}
	536
	537	wake_index = sbq_index_inc(wake_index);
	538	}
	539
	540	return NULL;
	541	}
	542
c854ab57	543	static bool __sbq_wake_up(struct sbitmap_queue *sbq)
88459642 OS	544	{
88459642 OS	545	struct sbq_wait_state *ws;
6c0ca7ae	546	unsigned int wake_batch;
88459642 OS	547	int wait_cnt;
88459642 OS	548
88459642 OS	549	ws = sbq_wake_ptr(sbq);
88459642 OS	550	if (!ws)
c854ab57	551	return false;
88459642 OS	552
88459642 OS	553	wait_cnt = atomic_dec_return(&ws->wait_cnt);
6c0ca7ae	554	if (wait_cnt <= 0) {
c854ab57 JA	555	int ret;
c854ab57 JA	556
6c0ca7ae	557	wake_batch = READ_ONCE(sbq->wake_batch);
c854ab57	558
6c0ca7ae OS	559	/*
	560	* Pairs with the memory barrier in sbitmap_queue_resize() to
	561	* ensure that we see the batch size update before the wait
	562	* count is reset.
	563	*/
	564	smp_mb__before_atomic();
c854ab57	565
6c0ca7ae	566	/*
c854ab57 JA	567	* For concurrent callers of this, the one that failed the
	568	* atomic_cmpxhcg() race should call this function again
	569	* to wakeup a new batch on a different 'ws'.
6c0ca7ae	570	*/
c854ab57 JA	571	ret = atomic_cmpxchg(&ws->wait_cnt, wait_cnt, wake_batch);
	572	if (ret == wait_cnt) {
	573	sbq_index_atomic_inc(&sbq->wake_index);
	574	wake_up_nr(&ws->wait, wake_batch);
	575	return false;
	576	}
	577
	578	return true;
88459642	579	}
c854ab57 JA	580
	581	return false;
	582	}
	583
e6fc4649	584	void sbitmap_queue_wake_up(struct sbitmap_queue *sbq)
c854ab57 JA	585	{
	586	while (__sbq_wake_up(sbq))
	587	;
88459642	588	}
e6fc4649	589	EXPORT_SYMBOL_GPL(sbitmap_queue_wake_up);
88459642	590
40aabb67	591	void sbitmap_queue_clear(struct sbitmap_queue *sbq, unsigned int nr,
f4a644db	592	unsigned int cpu)
88459642	593	{
ea86ea2c JA	594	sbitmap_deferred_clear_bit(&sbq->sb, nr);
ea86ea2c JA	595
e6fc4649 ML	596	/*
	597	* Pairs with the memory barrier in set_current_state() to ensure the
	598	* proper ordering of clear_bit_unlock()/waitqueue_active() in the waker
	599	* and test_and_set_bit_lock()/prepare_to_wait()/finish_wait() in the
	600	* waiter. See the comment on waitqueue_active().
	601	*/
	602	smp_mb__after_atomic();
	603	sbitmap_queue_wake_up(sbq);
	604
5c64a8df	605	if (likely(!sbq->round_robin && nr < sbq->sb.depth))
40aabb67	606	*per_cpu_ptr(sbq->alloc_hint, cpu) = nr;
88459642 OS	607	}
	608	EXPORT_SYMBOL_GPL(sbitmap_queue_clear);
	609
	610	void sbitmap_queue_wake_all(struct sbitmap_queue *sbq)
	611	{
	612	int i, wake_index;
	613
	614	/*
f66227de	615	* Pairs with the memory barrier in set_current_state() like in
e6fc4649	616	* sbitmap_queue_wake_up().
88459642 OS	617	*/
	618	smp_mb();
	619	wake_index = atomic_read(&sbq->wake_index);
	620	for (i = 0; i < SBQ_WAIT_QUEUES; i++) {
	621	struct sbq_wait_state *ws = &sbq->ws[wake_index];
	622
	623	if (waitqueue_active(&ws->wait))
	624	wake_up(&ws->wait);
	625
	626	wake_index = sbq_index_inc(wake_index);
	627	}
	628	}
	629	EXPORT_SYMBOL_GPL(sbitmap_queue_wake_all);
24af1ccf OS	630
	631	void sbitmap_queue_show(struct sbitmap_queue sbq, struct seq_file m)
	632	{
	633	bool first;
	634	int i;
	635
	636	sbitmap_show(&sbq->sb, m);
	637
	638	seq_puts(m, "alloc_hint={");
	639	first = true;
	640	for_each_possible_cpu(i) {
	641	if (!first)
	642	seq_puts(m, ", ");
	643	first = false;
	644	seq_printf(m, "%u", *per_cpu_ptr(sbq->alloc_hint, i));
	645	}
	646	seq_puts(m, "}\n");
	647
	648	seq_printf(m, "wake_batch=%u\n", sbq->wake_batch);
	649	seq_printf(m, "wake_index=%d\n", atomic_read(&sbq->wake_index));
5d2ee712	650	seq_printf(m, "ws_active=%d\n", atomic_read(&sbq->ws_active));
24af1ccf OS	651
	652	seq_puts(m, "ws={\n");
	653	for (i = 0; i < SBQ_WAIT_QUEUES; i++) {
	654	struct sbq_wait_state *ws = &sbq->ws[i];
	655
	656	seq_printf(m, "\t{.wait_cnt=%d, .wait=%s},\n",
	657	atomic_read(&ws->wait_cnt),
	658	waitqueue_active(&ws->wait) ? "active" : "inactive");
	659	}
	660	seq_puts(m, "}\n");
	661
	662	seq_printf(m, "round_robin=%d\n", sbq->round_robin);
a3275539	663	seq_printf(m, "min_shallow_depth=%u\n", sbq->min_shallow_depth);
24af1ccf OS	664	}
24af1ccf OS	665	EXPORT_SYMBOL_GPL(sbitmap_queue_show);
5d2ee712	666
9f6b7ef6 JA	667	void sbitmap_add_wait_queue(struct sbitmap_queue *sbq,
	668	struct sbq_wait_state *ws,
	669	struct sbq_wait *sbq_wait)
	670	{
	671	if (!sbq_wait->sbq) {
	672	sbq_wait->sbq = sbq;
	673	atomic_inc(&sbq->ws_active);
	674	}
	675	add_wait_queue(&ws->wait, &sbq_wait->wait);
	676	}
	677	EXPORT_SYMBOL_GPL(sbitmap_add_wait_queue);
	678
	679	void sbitmap_del_wait_queue(struct sbq_wait *sbq_wait)
	680	{
	681	list_del_init(&sbq_wait->wait.entry);
	682	if (sbq_wait->sbq) {
	683	atomic_dec(&sbq_wait->sbq->ws_active);
	684	sbq_wait->sbq = NULL;
	685	}
	686	}
	687	EXPORT_SYMBOL_GPL(sbitmap_del_wait_queue);
	688
5d2ee712 JA	689	void sbitmap_prepare_to_wait(struct sbitmap_queue *sbq,
	690	struct sbq_wait_state *ws,
	691	struct sbq_wait *sbq_wait, int state)
	692	{
9f6b7ef6	693	if (!sbq_wait->sbq) {
5d2ee712	694	atomic_inc(&sbq->ws_active);
9f6b7ef6	695	sbq_wait->sbq = sbq;
5d2ee712 JA	696	}
	697	prepare_to_wait_exclusive(&ws->wait, &sbq_wait->wait, state);
	698	}
	699	EXPORT_SYMBOL_GPL(sbitmap_prepare_to_wait);
	700
	701	void sbitmap_finish_wait(struct sbitmap_queue sbq, struct sbq_wait_state ws,
	702	struct sbq_wait *sbq_wait)
	703	{
	704	finish_wait(&ws->wait, &sbq_wait->wait);
9f6b7ef6	705	if (sbq_wait->sbq) {
5d2ee712	706	atomic_dec(&sbq->ws_active);
9f6b7ef6	707	sbq_wait->sbq = NULL;
5d2ee712 JA	708	}
	709	}
	710	EXPORT_SYMBOL_GPL(sbitmap_finish_wait);