xref: /openbmc/linux/lib/sbitmap.c (revision ecefa105)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * Copyright (C) 2016 Facebook
4  * Copyright (C) 2013-2014 Jens Axboe
5  */
6 
7 #include <linux/sched.h>
8 #include <linux/random.h>
9 #include <linux/sbitmap.h>
10 #include <linux/seq_file.h>
11 
12 static int init_alloc_hint(struct sbitmap *sb, gfp_t flags)
13 {
14 	unsigned depth = sb->depth;
15 
16 	sb->alloc_hint = alloc_percpu_gfp(unsigned int, flags);
17 	if (!sb->alloc_hint)
18 		return -ENOMEM;
19 
20 	if (depth && !sb->round_robin) {
21 		int i;
22 
23 		for_each_possible_cpu(i)
24 			*per_cpu_ptr(sb->alloc_hint, i) = get_random_u32_below(depth);
25 	}
26 	return 0;
27 }
28 
29 static inline unsigned update_alloc_hint_before_get(struct sbitmap *sb,
30 						    unsigned int depth)
31 {
32 	unsigned hint;
33 
34 	hint = this_cpu_read(*sb->alloc_hint);
35 	if (unlikely(hint >= depth)) {
36 		hint = depth ? get_random_u32_below(depth) : 0;
37 		this_cpu_write(*sb->alloc_hint, hint);
38 	}
39 
40 	return hint;
41 }
42 
43 static inline void update_alloc_hint_after_get(struct sbitmap *sb,
44 					       unsigned int depth,
45 					       unsigned int hint,
46 					       unsigned int nr)
47 {
48 	if (nr == -1) {
49 		/* If the map is full, a hint won't do us much good. */
50 		this_cpu_write(*sb->alloc_hint, 0);
51 	} else if (nr == hint || unlikely(sb->round_robin)) {
52 		/* Only update the hint if we used it. */
53 		hint = nr + 1;
54 		if (hint >= depth - 1)
55 			hint = 0;
56 		this_cpu_write(*sb->alloc_hint, hint);
57 	}
58 }
59 
60 /*
61  * See if we have deferred clears that we can batch move
62  */
63 static inline bool sbitmap_deferred_clear(struct sbitmap_word *map)
64 {
65 	unsigned long mask;
66 
67 	if (!READ_ONCE(map->cleared))
68 		return false;
69 
70 	/*
71 	 * First get a stable cleared mask, setting the old mask to 0.
72 	 */
73 	mask = xchg(&map->cleared, 0);
74 
75 	/*
76 	 * Now clear the masked bits in our free word
77 	 */
78 	atomic_long_andnot(mask, (atomic_long_t *)&map->word);
79 	BUILD_BUG_ON(sizeof(atomic_long_t) != sizeof(map->word));
80 	return true;
81 }
82 
83 int sbitmap_init_node(struct sbitmap *sb, unsigned int depth, int shift,
84 		      gfp_t flags, int node, bool round_robin,
85 		      bool alloc_hint)
86 {
87 	unsigned int bits_per_word;
88 
89 	if (shift < 0)
90 		shift = sbitmap_calculate_shift(depth);
91 
92 	bits_per_word = 1U << shift;
93 	if (bits_per_word > BITS_PER_LONG)
94 		return -EINVAL;
95 
96 	sb->shift = shift;
97 	sb->depth = depth;
98 	sb->map_nr = DIV_ROUND_UP(sb->depth, bits_per_word);
99 	sb->round_robin = round_robin;
100 
101 	if (depth == 0) {
102 		sb->map = NULL;
103 		return 0;
104 	}
105 
106 	if (alloc_hint) {
107 		if (init_alloc_hint(sb, flags))
108 			return -ENOMEM;
109 	} else {
110 		sb->alloc_hint = NULL;
111 	}
112 
113 	sb->map = kvzalloc_node(sb->map_nr * sizeof(*sb->map), flags, node);
114 	if (!sb->map) {
115 		free_percpu(sb->alloc_hint);
116 		return -ENOMEM;
117 	}
118 
119 	return 0;
120 }
121 EXPORT_SYMBOL_GPL(sbitmap_init_node);
122 
123 void sbitmap_resize(struct sbitmap *sb, unsigned int depth)
124 {
125 	unsigned int bits_per_word = 1U << sb->shift;
126 	unsigned int i;
127 
128 	for (i = 0; i < sb->map_nr; i++)
129 		sbitmap_deferred_clear(&sb->map[i]);
130 
131 	sb->depth = depth;
132 	sb->map_nr = DIV_ROUND_UP(sb->depth, bits_per_word);
133 }
134 EXPORT_SYMBOL_GPL(sbitmap_resize);
135 
136 static int __sbitmap_get_word(unsigned long *word, unsigned long depth,
137 			      unsigned int hint, bool wrap)
138 {
139 	int nr;
140 
141 	/* don't wrap if starting from 0 */
142 	wrap = wrap && hint;
143 
144 	while (1) {
145 		nr = find_next_zero_bit(word, depth, hint);
146 		if (unlikely(nr >= depth)) {
147 			/*
148 			 * We started with an offset, and we didn't reset the
149 			 * offset to 0 in a failure case, so start from 0 to
150 			 * exhaust the map.
151 			 */
152 			if (hint && wrap) {
153 				hint = 0;
154 				continue;
155 			}
156 			return -1;
157 		}
158 
159 		if (!test_and_set_bit_lock(nr, word))
160 			break;
161 
162 		hint = nr + 1;
163 		if (hint >= depth - 1)
164 			hint = 0;
165 	}
166 
167 	return nr;
168 }
169 
170 static int sbitmap_find_bit_in_word(struct sbitmap_word *map,
171 				    unsigned int depth,
172 				    unsigned int alloc_hint,
173 				    bool wrap)
174 {
175 	int nr;
176 
177 	do {
178 		nr = __sbitmap_get_word(&map->word, depth,
179 					alloc_hint, wrap);
180 		if (nr != -1)
181 			break;
182 		if (!sbitmap_deferred_clear(map))
183 			break;
184 	} while (1);
185 
186 	return nr;
187 }
188 
189 static int sbitmap_find_bit(struct sbitmap *sb,
190 			    unsigned int depth,
191 			    unsigned int index,
192 			    unsigned int alloc_hint,
193 			    bool wrap)
194 {
195 	unsigned int i;
196 	int nr = -1;
197 
198 	for (i = 0; i < sb->map_nr; i++) {
199 		nr = sbitmap_find_bit_in_word(&sb->map[index],
200 					      min_t(unsigned int,
201 						    __map_depth(sb, index),
202 						    depth),
203 					      alloc_hint, wrap);
204 
205 		if (nr != -1) {
206 			nr += index << sb->shift;
207 			break;
208 		}
209 
210 		/* Jump to next index. */
211 		alloc_hint = 0;
212 		if (++index >= sb->map_nr)
213 			index = 0;
214 	}
215 
216 	return nr;
217 }
218 
219 static int __sbitmap_get(struct sbitmap *sb, unsigned int alloc_hint)
220 {
221 	unsigned int index;
222 
223 	index = SB_NR_TO_INDEX(sb, alloc_hint);
224 
225 	/*
226 	 * Unless we're doing round robin tag allocation, just use the
227 	 * alloc_hint to find the right word index. No point in looping
228 	 * twice in find_next_zero_bit() for that case.
229 	 */
230 	if (sb->round_robin)
231 		alloc_hint = SB_NR_TO_BIT(sb, alloc_hint);
232 	else
233 		alloc_hint = 0;
234 
235 	return sbitmap_find_bit(sb, UINT_MAX, index, alloc_hint,
236 				!sb->round_robin);
237 }
238 
239 int sbitmap_get(struct sbitmap *sb)
240 {
241 	int nr;
242 	unsigned int hint, depth;
243 
244 	if (WARN_ON_ONCE(unlikely(!sb->alloc_hint)))
245 		return -1;
246 
247 	depth = READ_ONCE(sb->depth);
248 	hint = update_alloc_hint_before_get(sb, depth);
249 	nr = __sbitmap_get(sb, hint);
250 	update_alloc_hint_after_get(sb, depth, hint, nr);
251 
252 	return nr;
253 }
254 EXPORT_SYMBOL_GPL(sbitmap_get);
255 
256 static int __sbitmap_get_shallow(struct sbitmap *sb,
257 				 unsigned int alloc_hint,
258 				 unsigned long shallow_depth)
259 {
260 	unsigned int index;
261 
262 	index = SB_NR_TO_INDEX(sb, alloc_hint);
263 	alloc_hint = SB_NR_TO_BIT(sb, alloc_hint);
264 
265 	return sbitmap_find_bit(sb, shallow_depth, index, alloc_hint, true);
266 }
267 
268 int sbitmap_get_shallow(struct sbitmap *sb, unsigned long shallow_depth)
269 {
270 	int nr;
271 	unsigned int hint, depth;
272 
273 	if (WARN_ON_ONCE(unlikely(!sb->alloc_hint)))
274 		return -1;
275 
276 	depth = READ_ONCE(sb->depth);
277 	hint = update_alloc_hint_before_get(sb, depth);
278 	nr = __sbitmap_get_shallow(sb, hint, shallow_depth);
279 	update_alloc_hint_after_get(sb, depth, hint, nr);
280 
281 	return nr;
282 }
283 EXPORT_SYMBOL_GPL(sbitmap_get_shallow);
284 
285 bool sbitmap_any_bit_set(const struct sbitmap *sb)
286 {
287 	unsigned int i;
288 
289 	for (i = 0; i < sb->map_nr; i++) {
290 		if (sb->map[i].word & ~sb->map[i].cleared)
291 			return true;
292 	}
293 	return false;
294 }
295 EXPORT_SYMBOL_GPL(sbitmap_any_bit_set);
296 
297 static unsigned int __sbitmap_weight(const struct sbitmap *sb, bool set)
298 {
299 	unsigned int i, weight = 0;
300 
301 	for (i = 0; i < sb->map_nr; i++) {
302 		const struct sbitmap_word *word = &sb->map[i];
303 		unsigned int word_depth = __map_depth(sb, i);
304 
305 		if (set)
306 			weight += bitmap_weight(&word->word, word_depth);
307 		else
308 			weight += bitmap_weight(&word->cleared, word_depth);
309 	}
310 	return weight;
311 }
312 
313 static unsigned int sbitmap_cleared(const struct sbitmap *sb)
314 {
315 	return __sbitmap_weight(sb, false);
316 }
317 
318 unsigned int sbitmap_weight(const struct sbitmap *sb)
319 {
320 	return __sbitmap_weight(sb, true) - sbitmap_cleared(sb);
321 }
322 EXPORT_SYMBOL_GPL(sbitmap_weight);
323 
324 void sbitmap_show(struct sbitmap *sb, struct seq_file *m)
325 {
326 	seq_printf(m, "depth=%u\n", sb->depth);
327 	seq_printf(m, "busy=%u\n", sbitmap_weight(sb));
328 	seq_printf(m, "cleared=%u\n", sbitmap_cleared(sb));
329 	seq_printf(m, "bits_per_word=%u\n", 1U << sb->shift);
330 	seq_printf(m, "map_nr=%u\n", sb->map_nr);
331 }
332 EXPORT_SYMBOL_GPL(sbitmap_show);
333 
334 static inline void emit_byte(struct seq_file *m, unsigned int offset, u8 byte)
335 {
336 	if ((offset & 0xf) == 0) {
337 		if (offset != 0)
338 			seq_putc(m, '\n');
339 		seq_printf(m, "%08x:", offset);
340 	}
341 	if ((offset & 0x1) == 0)
342 		seq_putc(m, ' ');
343 	seq_printf(m, "%02x", byte);
344 }
345 
346 void sbitmap_bitmap_show(struct sbitmap *sb, struct seq_file *m)
347 {
348 	u8 byte = 0;
349 	unsigned int byte_bits = 0;
350 	unsigned int offset = 0;
351 	int i;
352 
353 	for (i = 0; i < sb->map_nr; i++) {
354 		unsigned long word = READ_ONCE(sb->map[i].word);
355 		unsigned long cleared = READ_ONCE(sb->map[i].cleared);
356 		unsigned int word_bits = __map_depth(sb, i);
357 
358 		word &= ~cleared;
359 
360 		while (word_bits > 0) {
361 			unsigned int bits = min(8 - byte_bits, word_bits);
362 
363 			byte |= (word & (BIT(bits) - 1)) << byte_bits;
364 			byte_bits += bits;
365 			if (byte_bits == 8) {
366 				emit_byte(m, offset, byte);
367 				byte = 0;
368 				byte_bits = 0;
369 				offset++;
370 			}
371 			word >>= bits;
372 			word_bits -= bits;
373 		}
374 	}
375 	if (byte_bits) {
376 		emit_byte(m, offset, byte);
377 		offset++;
378 	}
379 	if (offset)
380 		seq_putc(m, '\n');
381 }
382 EXPORT_SYMBOL_GPL(sbitmap_bitmap_show);
383 
384 static unsigned int sbq_calc_wake_batch(struct sbitmap_queue *sbq,
385 					unsigned int depth)
386 {
387 	unsigned int wake_batch;
388 	unsigned int shallow_depth;
389 
390 	/*
391 	 * For each batch, we wake up one queue. We need to make sure that our
392 	 * batch size is small enough that the full depth of the bitmap,
393 	 * potentially limited by a shallow depth, is enough to wake up all of
394 	 * the queues.
395 	 *
396 	 * Each full word of the bitmap has bits_per_word bits, and there might
397 	 * be a partial word. There are depth / bits_per_word full words and
398 	 * depth % bits_per_word bits left over. In bitwise arithmetic:
399 	 *
400 	 * bits_per_word = 1 << shift
401 	 * depth / bits_per_word = depth >> shift
402 	 * depth % bits_per_word = depth & ((1 << shift) - 1)
403 	 *
404 	 * Each word can be limited to sbq->min_shallow_depth bits.
405 	 */
406 	shallow_depth = min(1U << sbq->sb.shift, sbq->min_shallow_depth);
407 	depth = ((depth >> sbq->sb.shift) * shallow_depth +
408 		 min(depth & ((1U << sbq->sb.shift) - 1), shallow_depth));
409 	wake_batch = clamp_t(unsigned int, depth / SBQ_WAIT_QUEUES, 1,
410 			     SBQ_WAKE_BATCH);
411 
412 	return wake_batch;
413 }
414 
415 int sbitmap_queue_init_node(struct sbitmap_queue *sbq, unsigned int depth,
416 			    int shift, bool round_robin, gfp_t flags, int node)
417 {
418 	int ret;
419 	int i;
420 
421 	ret = sbitmap_init_node(&sbq->sb, depth, shift, flags, node,
422 				round_robin, true);
423 	if (ret)
424 		return ret;
425 
426 	sbq->min_shallow_depth = UINT_MAX;
427 	sbq->wake_batch = sbq_calc_wake_batch(sbq, depth);
428 	atomic_set(&sbq->wake_index, 0);
429 	atomic_set(&sbq->ws_active, 0);
430 	atomic_set(&sbq->completion_cnt, 0);
431 	atomic_set(&sbq->wakeup_cnt, 0);
432 
433 	sbq->ws = kzalloc_node(SBQ_WAIT_QUEUES * sizeof(*sbq->ws), flags, node);
434 	if (!sbq->ws) {
435 		sbitmap_free(&sbq->sb);
436 		return -ENOMEM;
437 	}
438 
439 	for (i = 0; i < SBQ_WAIT_QUEUES; i++)
440 		init_waitqueue_head(&sbq->ws[i].wait);
441 
442 	return 0;
443 }
444 EXPORT_SYMBOL_GPL(sbitmap_queue_init_node);
445 
446 static void sbitmap_queue_update_wake_batch(struct sbitmap_queue *sbq,
447 					    unsigned int depth)
448 {
449 	unsigned int wake_batch;
450 
451 	wake_batch = sbq_calc_wake_batch(sbq, depth);
452 	if (sbq->wake_batch != wake_batch)
453 		WRITE_ONCE(sbq->wake_batch, wake_batch);
454 }
455 
456 void sbitmap_queue_recalculate_wake_batch(struct sbitmap_queue *sbq,
457 					    unsigned int users)
458 {
459 	unsigned int wake_batch;
460 	unsigned int depth = (sbq->sb.depth + users - 1) / users;
461 
462 	wake_batch = clamp_val(depth / SBQ_WAIT_QUEUES,
463 			1, SBQ_WAKE_BATCH);
464 
465 	WRITE_ONCE(sbq->wake_batch, wake_batch);
466 }
467 EXPORT_SYMBOL_GPL(sbitmap_queue_recalculate_wake_batch);
468 
469 void sbitmap_queue_resize(struct sbitmap_queue *sbq, unsigned int depth)
470 {
471 	sbitmap_queue_update_wake_batch(sbq, depth);
472 	sbitmap_resize(&sbq->sb, depth);
473 }
474 EXPORT_SYMBOL_GPL(sbitmap_queue_resize);
475 
476 int __sbitmap_queue_get(struct sbitmap_queue *sbq)
477 {
478 	return sbitmap_get(&sbq->sb);
479 }
480 EXPORT_SYMBOL_GPL(__sbitmap_queue_get);
481 
482 unsigned long __sbitmap_queue_get_batch(struct sbitmap_queue *sbq, int nr_tags,
483 					unsigned int *offset)
484 {
485 	struct sbitmap *sb = &sbq->sb;
486 	unsigned int hint, depth;
487 	unsigned long index, nr;
488 	int i;
489 
490 	if (unlikely(sb->round_robin))
491 		return 0;
492 
493 	depth = READ_ONCE(sb->depth);
494 	hint = update_alloc_hint_before_get(sb, depth);
495 
496 	index = SB_NR_TO_INDEX(sb, hint);
497 
498 	for (i = 0; i < sb->map_nr; i++) {
499 		struct sbitmap_word *map = &sb->map[index];
500 		unsigned long get_mask;
501 		unsigned int map_depth = __map_depth(sb, index);
502 
503 		sbitmap_deferred_clear(map);
504 		if (map->word == (1UL << (map_depth - 1)) - 1)
505 			goto next;
506 
507 		nr = find_first_zero_bit(&map->word, map_depth);
508 		if (nr + nr_tags <= map_depth) {
509 			atomic_long_t *ptr = (atomic_long_t *) &map->word;
510 			unsigned long val;
511 
512 			get_mask = ((1UL << nr_tags) - 1) << nr;
513 			val = READ_ONCE(map->word);
514 			while (!atomic_long_try_cmpxchg(ptr, &val,
515 							  get_mask | val))
516 				;
517 			get_mask = (get_mask & ~val) >> nr;
518 			if (get_mask) {
519 				*offset = nr + (index << sb->shift);
520 				update_alloc_hint_after_get(sb, depth, hint,
521 							*offset + nr_tags - 1);
522 				return get_mask;
523 			}
524 		}
525 next:
526 		/* Jump to next index. */
527 		if (++index >= sb->map_nr)
528 			index = 0;
529 	}
530 
531 	return 0;
532 }
533 
534 int sbitmap_queue_get_shallow(struct sbitmap_queue *sbq,
535 			      unsigned int shallow_depth)
536 {
537 	WARN_ON_ONCE(shallow_depth < sbq->min_shallow_depth);
538 
539 	return sbitmap_get_shallow(&sbq->sb, shallow_depth);
540 }
541 EXPORT_SYMBOL_GPL(sbitmap_queue_get_shallow);
542 
543 void sbitmap_queue_min_shallow_depth(struct sbitmap_queue *sbq,
544 				     unsigned int min_shallow_depth)
545 {
546 	sbq->min_shallow_depth = min_shallow_depth;
547 	sbitmap_queue_update_wake_batch(sbq, sbq->sb.depth);
548 }
549 EXPORT_SYMBOL_GPL(sbitmap_queue_min_shallow_depth);
550 
551 static void __sbitmap_queue_wake_up(struct sbitmap_queue *sbq, int nr)
552 {
553 	int i, wake_index;
554 
555 	if (!atomic_read(&sbq->ws_active))
556 		return;
557 
558 	wake_index = atomic_read(&sbq->wake_index);
559 	for (i = 0; i < SBQ_WAIT_QUEUES; i++) {
560 		struct sbq_wait_state *ws = &sbq->ws[wake_index];
561 
562 		/*
563 		 * Advance the index before checking the current queue.
564 		 * It improves fairness, by ensuring the queue doesn't
565 		 * need to be fully emptied before trying to wake up
566 		 * from the next one.
567 		 */
568 		wake_index = sbq_index_inc(wake_index);
569 
570 		/*
571 		 * It is sufficient to wake up at least one waiter to
572 		 * guarantee forward progress.
573 		 */
574 		if (waitqueue_active(&ws->wait) &&
575 		    wake_up_nr(&ws->wait, nr))
576 			break;
577 	}
578 
579 	if (wake_index != atomic_read(&sbq->wake_index))
580 		atomic_set(&sbq->wake_index, wake_index);
581 }
582 
583 void sbitmap_queue_wake_up(struct sbitmap_queue *sbq, int nr)
584 {
585 	unsigned int wake_batch = READ_ONCE(sbq->wake_batch);
586 	unsigned int wakeups;
587 
588 	if (!atomic_read(&sbq->ws_active))
589 		return;
590 
591 	atomic_add(nr, &sbq->completion_cnt);
592 	wakeups = atomic_read(&sbq->wakeup_cnt);
593 
594 	do {
595 		if (atomic_read(&sbq->completion_cnt) - wakeups < wake_batch)
596 			return;
597 	} while (!atomic_try_cmpxchg(&sbq->wakeup_cnt,
598 				     &wakeups, wakeups + wake_batch));
599 
600 	__sbitmap_queue_wake_up(sbq, wake_batch);
601 }
602 EXPORT_SYMBOL_GPL(sbitmap_queue_wake_up);
603 
604 static inline void sbitmap_update_cpu_hint(struct sbitmap *sb, int cpu, int tag)
605 {
606 	if (likely(!sb->round_robin && tag < sb->depth))
607 		data_race(*per_cpu_ptr(sb->alloc_hint, cpu) = tag);
608 }
609 
610 void sbitmap_queue_clear_batch(struct sbitmap_queue *sbq, int offset,
611 				int *tags, int nr_tags)
612 {
613 	struct sbitmap *sb = &sbq->sb;
614 	unsigned long *addr = NULL;
615 	unsigned long mask = 0;
616 	int i;
617 
618 	smp_mb__before_atomic();
619 	for (i = 0; i < nr_tags; i++) {
620 		const int tag = tags[i] - offset;
621 		unsigned long *this_addr;
622 
623 		/* since we're clearing a batch, skip the deferred map */
624 		this_addr = &sb->map[SB_NR_TO_INDEX(sb, tag)].word;
625 		if (!addr) {
626 			addr = this_addr;
627 		} else if (addr != this_addr) {
628 			atomic_long_andnot(mask, (atomic_long_t *) addr);
629 			mask = 0;
630 			addr = this_addr;
631 		}
632 		mask |= (1UL << SB_NR_TO_BIT(sb, tag));
633 	}
634 
635 	if (mask)
636 		atomic_long_andnot(mask, (atomic_long_t *) addr);
637 
638 	smp_mb__after_atomic();
639 	sbitmap_queue_wake_up(sbq, nr_tags);
640 	sbitmap_update_cpu_hint(&sbq->sb, raw_smp_processor_id(),
641 					tags[nr_tags - 1] - offset);
642 }
643 
644 void sbitmap_queue_clear(struct sbitmap_queue *sbq, unsigned int nr,
645 			 unsigned int cpu)
646 {
647 	/*
648 	 * Once the clear bit is set, the bit may be allocated out.
649 	 *
650 	 * Orders READ/WRITE on the associated instance(such as request
651 	 * of blk_mq) by this bit for avoiding race with re-allocation,
652 	 * and its pair is the memory barrier implied in __sbitmap_get_word.
653 	 *
654 	 * One invariant is that the clear bit has to be zero when the bit
655 	 * is in use.
656 	 */
657 	smp_mb__before_atomic();
658 	sbitmap_deferred_clear_bit(&sbq->sb, nr);
659 
660 	/*
661 	 * Pairs with the memory barrier in set_current_state() to ensure the
662 	 * proper ordering of clear_bit_unlock()/waitqueue_active() in the waker
663 	 * and test_and_set_bit_lock()/prepare_to_wait()/finish_wait() in the
664 	 * waiter. See the comment on waitqueue_active().
665 	 */
666 	smp_mb__after_atomic();
667 	sbitmap_queue_wake_up(sbq, 1);
668 	sbitmap_update_cpu_hint(&sbq->sb, cpu, nr);
669 }
670 EXPORT_SYMBOL_GPL(sbitmap_queue_clear);
671 
672 void sbitmap_queue_wake_all(struct sbitmap_queue *sbq)
673 {
674 	int i, wake_index;
675 
676 	/*
677 	 * Pairs with the memory barrier in set_current_state() like in
678 	 * sbitmap_queue_wake_up().
679 	 */
680 	smp_mb();
681 	wake_index = atomic_read(&sbq->wake_index);
682 	for (i = 0; i < SBQ_WAIT_QUEUES; i++) {
683 		struct sbq_wait_state *ws = &sbq->ws[wake_index];
684 
685 		if (waitqueue_active(&ws->wait))
686 			wake_up(&ws->wait);
687 
688 		wake_index = sbq_index_inc(wake_index);
689 	}
690 }
691 EXPORT_SYMBOL_GPL(sbitmap_queue_wake_all);
692 
693 void sbitmap_queue_show(struct sbitmap_queue *sbq, struct seq_file *m)
694 {
695 	bool first;
696 	int i;
697 
698 	sbitmap_show(&sbq->sb, m);
699 
700 	seq_puts(m, "alloc_hint={");
701 	first = true;
702 	for_each_possible_cpu(i) {
703 		if (!first)
704 			seq_puts(m, ", ");
705 		first = false;
706 		seq_printf(m, "%u", *per_cpu_ptr(sbq->sb.alloc_hint, i));
707 	}
708 	seq_puts(m, "}\n");
709 
710 	seq_printf(m, "wake_batch=%u\n", sbq->wake_batch);
711 	seq_printf(m, "wake_index=%d\n", atomic_read(&sbq->wake_index));
712 	seq_printf(m, "ws_active=%d\n", atomic_read(&sbq->ws_active));
713 
714 	seq_puts(m, "ws={\n");
715 	for (i = 0; i < SBQ_WAIT_QUEUES; i++) {
716 		struct sbq_wait_state *ws = &sbq->ws[i];
717 		seq_printf(m, "\t{.wait=%s},\n",
718 			   waitqueue_active(&ws->wait) ? "active" : "inactive");
719 	}
720 	seq_puts(m, "}\n");
721 
722 	seq_printf(m, "round_robin=%d\n", sbq->sb.round_robin);
723 	seq_printf(m, "min_shallow_depth=%u\n", sbq->min_shallow_depth);
724 }
725 EXPORT_SYMBOL_GPL(sbitmap_queue_show);
726 
727 void sbitmap_add_wait_queue(struct sbitmap_queue *sbq,
728 			    struct sbq_wait_state *ws,
729 			    struct sbq_wait *sbq_wait)
730 {
731 	if (!sbq_wait->sbq) {
732 		sbq_wait->sbq = sbq;
733 		atomic_inc(&sbq->ws_active);
734 		add_wait_queue(&ws->wait, &sbq_wait->wait);
735 	}
736 }
737 EXPORT_SYMBOL_GPL(sbitmap_add_wait_queue);
738 
739 void sbitmap_del_wait_queue(struct sbq_wait *sbq_wait)
740 {
741 	list_del_init(&sbq_wait->wait.entry);
742 	if (sbq_wait->sbq) {
743 		atomic_dec(&sbq_wait->sbq->ws_active);
744 		sbq_wait->sbq = NULL;
745 	}
746 }
747 EXPORT_SYMBOL_GPL(sbitmap_del_wait_queue);
748 
749 void sbitmap_prepare_to_wait(struct sbitmap_queue *sbq,
750 			     struct sbq_wait_state *ws,
751 			     struct sbq_wait *sbq_wait, int state)
752 {
753 	if (!sbq_wait->sbq) {
754 		atomic_inc(&sbq->ws_active);
755 		sbq_wait->sbq = sbq;
756 	}
757 	prepare_to_wait_exclusive(&ws->wait, &sbq_wait->wait, state);
758 }
759 EXPORT_SYMBOL_GPL(sbitmap_prepare_to_wait);
760 
761 void sbitmap_finish_wait(struct sbitmap_queue *sbq, struct sbq_wait_state *ws,
762 			 struct sbq_wait *sbq_wait)
763 {
764 	finish_wait(&ws->wait, &sbq_wait->wait);
765 	if (sbq_wait->sbq) {
766 		atomic_dec(&sbq->ws_active);
767 		sbq_wait->sbq = NULL;
768 	}
769 }
770 EXPORT_SYMBOL_GPL(sbitmap_finish_wait);
771