mm: vmscan: do not iterate all mem cgroups for global direct reclaim
[linux/fpc-iii.git] / block / blk-sysfs.c
blob590d1ef2f961cc151b0a070cc825a16dd616c16d
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * Functions related to sysfs handling
4 */
5 #include <linux/kernel.h>
6 #include <linux/slab.h>
7 #include <linux/module.h>
8 #include <linux/bio.h>
9 #include <linux/blkdev.h>
10 #include <linux/backing-dev.h>
11 #include <linux/blktrace_api.h>
12 #include <linux/blk-mq.h>
13 #include <linux/blk-cgroup.h>
15 #include "blk.h"
16 #include "blk-mq.h"
17 #include "blk-mq-debugfs.h"
18 #include "blk-wbt.h"
20 struct queue_sysfs_entry {
21 struct attribute attr;
22 ssize_t (*show)(struct request_queue *, char *);
23 ssize_t (*store)(struct request_queue *, const char *, size_t);
26 static ssize_t
27 queue_var_show(unsigned long var, char *page)
29 return sprintf(page, "%lu\n", var);
32 static ssize_t
33 queue_var_store(unsigned long *var, const char *page, size_t count)
35 int err;
36 unsigned long v;
38 err = kstrtoul(page, 10, &v);
39 if (err || v > UINT_MAX)
40 return -EINVAL;
42 *var = v;
44 return count;
47 static ssize_t queue_var_store64(s64 *var, const char *page)
49 int err;
50 s64 v;
52 err = kstrtos64(page, 10, &v);
53 if (err < 0)
54 return err;
56 *var = v;
57 return 0;
60 static ssize_t queue_requests_show(struct request_queue *q, char *page)
62 return queue_var_show(q->nr_requests, (page));
65 static ssize_t
66 queue_requests_store(struct request_queue *q, const char *page, size_t count)
68 unsigned long nr;
69 int ret, err;
71 if (!queue_is_mq(q))
72 return -EINVAL;
74 ret = queue_var_store(&nr, page, count);
75 if (ret < 0)
76 return ret;
78 if (nr < BLKDEV_MIN_RQ)
79 nr = BLKDEV_MIN_RQ;
81 err = blk_mq_update_nr_requests(q, nr);
82 if (err)
83 return err;
85 return ret;
88 static ssize_t queue_ra_show(struct request_queue *q, char *page)
90 unsigned long ra_kb = q->backing_dev_info->ra_pages <<
91 (PAGE_SHIFT - 10);
93 return queue_var_show(ra_kb, (page));
96 static ssize_t
97 queue_ra_store(struct request_queue *q, const char *page, size_t count)
99 unsigned long ra_kb;
100 ssize_t ret = queue_var_store(&ra_kb, page, count);
102 if (ret < 0)
103 return ret;
105 q->backing_dev_info->ra_pages = ra_kb >> (PAGE_SHIFT - 10);
107 return ret;
110 static ssize_t queue_max_sectors_show(struct request_queue *q, char *page)
112 int max_sectors_kb = queue_max_sectors(q) >> 1;
114 return queue_var_show(max_sectors_kb, (page));
117 static ssize_t queue_max_segments_show(struct request_queue *q, char *page)
119 return queue_var_show(queue_max_segments(q), (page));
122 static ssize_t queue_max_discard_segments_show(struct request_queue *q,
123 char *page)
125 return queue_var_show(queue_max_discard_segments(q), (page));
128 static ssize_t queue_max_integrity_segments_show(struct request_queue *q, char *page)
130 return queue_var_show(q->limits.max_integrity_segments, (page));
133 static ssize_t queue_max_segment_size_show(struct request_queue *q, char *page)
135 return queue_var_show(queue_max_segment_size(q), (page));
138 static ssize_t queue_logical_block_size_show(struct request_queue *q, char *page)
140 return queue_var_show(queue_logical_block_size(q), page);
143 static ssize_t queue_physical_block_size_show(struct request_queue *q, char *page)
145 return queue_var_show(queue_physical_block_size(q), page);
148 static ssize_t queue_chunk_sectors_show(struct request_queue *q, char *page)
150 return queue_var_show(q->limits.chunk_sectors, page);
153 static ssize_t queue_io_min_show(struct request_queue *q, char *page)
155 return queue_var_show(queue_io_min(q), page);
158 static ssize_t queue_io_opt_show(struct request_queue *q, char *page)
160 return queue_var_show(queue_io_opt(q), page);
163 static ssize_t queue_discard_granularity_show(struct request_queue *q, char *page)
165 return queue_var_show(q->limits.discard_granularity, page);
168 static ssize_t queue_discard_max_hw_show(struct request_queue *q, char *page)
171 return sprintf(page, "%llu\n",
172 (unsigned long long)q->limits.max_hw_discard_sectors << 9);
175 static ssize_t queue_discard_max_show(struct request_queue *q, char *page)
177 return sprintf(page, "%llu\n",
178 (unsigned long long)q->limits.max_discard_sectors << 9);
181 static ssize_t queue_discard_max_store(struct request_queue *q,
182 const char *page, size_t count)
184 unsigned long max_discard;
185 ssize_t ret = queue_var_store(&max_discard, page, count);
187 if (ret < 0)
188 return ret;
190 if (max_discard & (q->limits.discard_granularity - 1))
191 return -EINVAL;
193 max_discard >>= 9;
194 if (max_discard > UINT_MAX)
195 return -EINVAL;
197 if (max_discard > q->limits.max_hw_discard_sectors)
198 max_discard = q->limits.max_hw_discard_sectors;
200 q->limits.max_discard_sectors = max_discard;
201 return ret;
204 static ssize_t queue_discard_zeroes_data_show(struct request_queue *q, char *page)
206 return queue_var_show(0, page);
209 static ssize_t queue_write_same_max_show(struct request_queue *q, char *page)
211 return sprintf(page, "%llu\n",
212 (unsigned long long)q->limits.max_write_same_sectors << 9);
215 static ssize_t queue_write_zeroes_max_show(struct request_queue *q, char *page)
217 return sprintf(page, "%llu\n",
218 (unsigned long long)q->limits.max_write_zeroes_sectors << 9);
221 static ssize_t
222 queue_max_sectors_store(struct request_queue *q, const char *page, size_t count)
224 unsigned long max_sectors_kb,
225 max_hw_sectors_kb = queue_max_hw_sectors(q) >> 1,
226 page_kb = 1 << (PAGE_SHIFT - 10);
227 ssize_t ret = queue_var_store(&max_sectors_kb, page, count);
229 if (ret < 0)
230 return ret;
232 max_hw_sectors_kb = min_not_zero(max_hw_sectors_kb, (unsigned long)
233 q->limits.max_dev_sectors >> 1);
235 if (max_sectors_kb > max_hw_sectors_kb || max_sectors_kb < page_kb)
236 return -EINVAL;
238 spin_lock_irq(&q->queue_lock);
239 q->limits.max_sectors = max_sectors_kb << 1;
240 q->backing_dev_info->io_pages = max_sectors_kb >> (PAGE_SHIFT - 10);
241 spin_unlock_irq(&q->queue_lock);
243 return ret;
246 static ssize_t queue_max_hw_sectors_show(struct request_queue *q, char *page)
248 int max_hw_sectors_kb = queue_max_hw_sectors(q) >> 1;
250 return queue_var_show(max_hw_sectors_kb, (page));
253 #define QUEUE_SYSFS_BIT_FNS(name, flag, neg) \
254 static ssize_t \
255 queue_show_##name(struct request_queue *q, char *page) \
257 int bit; \
258 bit = test_bit(QUEUE_FLAG_##flag, &q->queue_flags); \
259 return queue_var_show(neg ? !bit : bit, page); \
261 static ssize_t \
262 queue_store_##name(struct request_queue *q, const char *page, size_t count) \
264 unsigned long val; \
265 ssize_t ret; \
266 ret = queue_var_store(&val, page, count); \
267 if (ret < 0) \
268 return ret; \
269 if (neg) \
270 val = !val; \
272 if (val) \
273 blk_queue_flag_set(QUEUE_FLAG_##flag, q); \
274 else \
275 blk_queue_flag_clear(QUEUE_FLAG_##flag, q); \
276 return ret; \
279 QUEUE_SYSFS_BIT_FNS(nonrot, NONROT, 1);
280 QUEUE_SYSFS_BIT_FNS(random, ADD_RANDOM, 0);
281 QUEUE_SYSFS_BIT_FNS(iostats, IO_STAT, 0);
282 #undef QUEUE_SYSFS_BIT_FNS
284 static ssize_t queue_zoned_show(struct request_queue *q, char *page)
286 switch (blk_queue_zoned_model(q)) {
287 case BLK_ZONED_HA:
288 return sprintf(page, "host-aware\n");
289 case BLK_ZONED_HM:
290 return sprintf(page, "host-managed\n");
291 default:
292 return sprintf(page, "none\n");
296 static ssize_t queue_nr_zones_show(struct request_queue *q, char *page)
298 return queue_var_show(blk_queue_nr_zones(q), page);
301 static ssize_t queue_nomerges_show(struct request_queue *q, char *page)
303 return queue_var_show((blk_queue_nomerges(q) << 1) |
304 blk_queue_noxmerges(q), page);
307 static ssize_t queue_nomerges_store(struct request_queue *q, const char *page,
308 size_t count)
310 unsigned long nm;
311 ssize_t ret = queue_var_store(&nm, page, count);
313 if (ret < 0)
314 return ret;
316 blk_queue_flag_clear(QUEUE_FLAG_NOMERGES, q);
317 blk_queue_flag_clear(QUEUE_FLAG_NOXMERGES, q);
318 if (nm == 2)
319 blk_queue_flag_set(QUEUE_FLAG_NOMERGES, q);
320 else if (nm)
321 blk_queue_flag_set(QUEUE_FLAG_NOXMERGES, q);
323 return ret;
326 static ssize_t queue_rq_affinity_show(struct request_queue *q, char *page)
328 bool set = test_bit(QUEUE_FLAG_SAME_COMP, &q->queue_flags);
329 bool force = test_bit(QUEUE_FLAG_SAME_FORCE, &q->queue_flags);
331 return queue_var_show(set << force, page);
334 static ssize_t
335 queue_rq_affinity_store(struct request_queue *q, const char *page, size_t count)
337 ssize_t ret = -EINVAL;
338 #ifdef CONFIG_SMP
339 unsigned long val;
341 ret = queue_var_store(&val, page, count);
342 if (ret < 0)
343 return ret;
345 if (val == 2) {
346 blk_queue_flag_set(QUEUE_FLAG_SAME_COMP, q);
347 blk_queue_flag_set(QUEUE_FLAG_SAME_FORCE, q);
348 } else if (val == 1) {
349 blk_queue_flag_set(QUEUE_FLAG_SAME_COMP, q);
350 blk_queue_flag_clear(QUEUE_FLAG_SAME_FORCE, q);
351 } else if (val == 0) {
352 blk_queue_flag_clear(QUEUE_FLAG_SAME_COMP, q);
353 blk_queue_flag_clear(QUEUE_FLAG_SAME_FORCE, q);
355 #endif
356 return ret;
359 static ssize_t queue_poll_delay_show(struct request_queue *q, char *page)
361 int val;
363 if (q->poll_nsec == -1)
364 val = -1;
365 else
366 val = q->poll_nsec / 1000;
368 return sprintf(page, "%d\n", val);
371 static ssize_t queue_poll_delay_store(struct request_queue *q, const char *page,
372 size_t count)
374 int err, val;
376 if (!q->mq_ops || !q->mq_ops->poll)
377 return -EINVAL;
379 err = kstrtoint(page, 10, &val);
380 if (err < 0)
381 return err;
383 if (val == -1)
384 q->poll_nsec = -1;
385 else
386 q->poll_nsec = val * 1000;
388 return count;
391 static ssize_t queue_poll_show(struct request_queue *q, char *page)
393 return queue_var_show(test_bit(QUEUE_FLAG_POLL, &q->queue_flags), page);
396 static ssize_t queue_poll_store(struct request_queue *q, const char *page,
397 size_t count)
399 unsigned long poll_on;
400 ssize_t ret;
402 if (!q->tag_set || q->tag_set->nr_maps <= HCTX_TYPE_POLL ||
403 !q->tag_set->map[HCTX_TYPE_POLL].nr_queues)
404 return -EINVAL;
406 ret = queue_var_store(&poll_on, page, count);
407 if (ret < 0)
408 return ret;
410 if (poll_on)
411 blk_queue_flag_set(QUEUE_FLAG_POLL, q);
412 else
413 blk_queue_flag_clear(QUEUE_FLAG_POLL, q);
415 return ret;
418 static ssize_t queue_io_timeout_show(struct request_queue *q, char *page)
420 return sprintf(page, "%u\n", jiffies_to_msecs(q->rq_timeout));
423 static ssize_t queue_io_timeout_store(struct request_queue *q, const char *page,
424 size_t count)
426 unsigned int val;
427 int err;
429 err = kstrtou32(page, 10, &val);
430 if (err || val == 0)
431 return -EINVAL;
433 blk_queue_rq_timeout(q, msecs_to_jiffies(val));
435 return count;
438 static ssize_t queue_wb_lat_show(struct request_queue *q, char *page)
440 if (!wbt_rq_qos(q))
441 return -EINVAL;
443 return sprintf(page, "%llu\n", div_u64(wbt_get_min_lat(q), 1000));
446 static ssize_t queue_wb_lat_store(struct request_queue *q, const char *page,
447 size_t count)
449 struct rq_qos *rqos;
450 ssize_t ret;
451 s64 val;
453 ret = queue_var_store64(&val, page);
454 if (ret < 0)
455 return ret;
456 if (val < -1)
457 return -EINVAL;
459 rqos = wbt_rq_qos(q);
460 if (!rqos) {
461 ret = wbt_init(q);
462 if (ret)
463 return ret;
466 if (val == -1)
467 val = wbt_default_latency_nsec(q);
468 else if (val >= 0)
469 val *= 1000ULL;
472 * Ensure that the queue is idled, in case the latency update
473 * ends up either enabling or disabling wbt completely. We can't
474 * have IO inflight if that happens.
476 blk_mq_freeze_queue(q);
477 blk_mq_quiesce_queue(q);
479 wbt_set_min_lat(q, val);
480 wbt_update_limits(q);
482 blk_mq_unquiesce_queue(q);
483 blk_mq_unfreeze_queue(q);
485 return count;
488 static ssize_t queue_wc_show(struct request_queue *q, char *page)
490 if (test_bit(QUEUE_FLAG_WC, &q->queue_flags))
491 return sprintf(page, "write back\n");
493 return sprintf(page, "write through\n");
496 static ssize_t queue_wc_store(struct request_queue *q, const char *page,
497 size_t count)
499 int set = -1;
501 if (!strncmp(page, "write back", 10))
502 set = 1;
503 else if (!strncmp(page, "write through", 13) ||
504 !strncmp(page, "none", 4))
505 set = 0;
507 if (set == -1)
508 return -EINVAL;
510 if (set)
511 blk_queue_flag_set(QUEUE_FLAG_WC, q);
512 else
513 blk_queue_flag_clear(QUEUE_FLAG_WC, q);
515 return count;
518 static ssize_t queue_fua_show(struct request_queue *q, char *page)
520 return sprintf(page, "%u\n", test_bit(QUEUE_FLAG_FUA, &q->queue_flags));
523 static ssize_t queue_dax_show(struct request_queue *q, char *page)
525 return queue_var_show(blk_queue_dax(q), page);
528 static struct queue_sysfs_entry queue_requests_entry = {
529 .attr = {.name = "nr_requests", .mode = 0644 },
530 .show = queue_requests_show,
531 .store = queue_requests_store,
534 static struct queue_sysfs_entry queue_ra_entry = {
535 .attr = {.name = "read_ahead_kb", .mode = 0644 },
536 .show = queue_ra_show,
537 .store = queue_ra_store,
540 static struct queue_sysfs_entry queue_max_sectors_entry = {
541 .attr = {.name = "max_sectors_kb", .mode = 0644 },
542 .show = queue_max_sectors_show,
543 .store = queue_max_sectors_store,
546 static struct queue_sysfs_entry queue_max_hw_sectors_entry = {
547 .attr = {.name = "max_hw_sectors_kb", .mode = 0444 },
548 .show = queue_max_hw_sectors_show,
551 static struct queue_sysfs_entry queue_max_segments_entry = {
552 .attr = {.name = "max_segments", .mode = 0444 },
553 .show = queue_max_segments_show,
556 static struct queue_sysfs_entry queue_max_discard_segments_entry = {
557 .attr = {.name = "max_discard_segments", .mode = 0444 },
558 .show = queue_max_discard_segments_show,
561 static struct queue_sysfs_entry queue_max_integrity_segments_entry = {
562 .attr = {.name = "max_integrity_segments", .mode = 0444 },
563 .show = queue_max_integrity_segments_show,
566 static struct queue_sysfs_entry queue_max_segment_size_entry = {
567 .attr = {.name = "max_segment_size", .mode = 0444 },
568 .show = queue_max_segment_size_show,
571 static struct queue_sysfs_entry queue_iosched_entry = {
572 .attr = {.name = "scheduler", .mode = 0644 },
573 .show = elv_iosched_show,
574 .store = elv_iosched_store,
577 static struct queue_sysfs_entry queue_hw_sector_size_entry = {
578 .attr = {.name = "hw_sector_size", .mode = 0444 },
579 .show = queue_logical_block_size_show,
582 static struct queue_sysfs_entry queue_logical_block_size_entry = {
583 .attr = {.name = "logical_block_size", .mode = 0444 },
584 .show = queue_logical_block_size_show,
587 static struct queue_sysfs_entry queue_physical_block_size_entry = {
588 .attr = {.name = "physical_block_size", .mode = 0444 },
589 .show = queue_physical_block_size_show,
592 static struct queue_sysfs_entry queue_chunk_sectors_entry = {
593 .attr = {.name = "chunk_sectors", .mode = 0444 },
594 .show = queue_chunk_sectors_show,
597 static struct queue_sysfs_entry queue_io_min_entry = {
598 .attr = {.name = "minimum_io_size", .mode = 0444 },
599 .show = queue_io_min_show,
602 static struct queue_sysfs_entry queue_io_opt_entry = {
603 .attr = {.name = "optimal_io_size", .mode = 0444 },
604 .show = queue_io_opt_show,
607 static struct queue_sysfs_entry queue_discard_granularity_entry = {
608 .attr = {.name = "discard_granularity", .mode = 0444 },
609 .show = queue_discard_granularity_show,
612 static struct queue_sysfs_entry queue_discard_max_hw_entry = {
613 .attr = {.name = "discard_max_hw_bytes", .mode = 0444 },
614 .show = queue_discard_max_hw_show,
617 static struct queue_sysfs_entry queue_discard_max_entry = {
618 .attr = {.name = "discard_max_bytes", .mode = 0644 },
619 .show = queue_discard_max_show,
620 .store = queue_discard_max_store,
623 static struct queue_sysfs_entry queue_discard_zeroes_data_entry = {
624 .attr = {.name = "discard_zeroes_data", .mode = 0444 },
625 .show = queue_discard_zeroes_data_show,
628 static struct queue_sysfs_entry queue_write_same_max_entry = {
629 .attr = {.name = "write_same_max_bytes", .mode = 0444 },
630 .show = queue_write_same_max_show,
633 static struct queue_sysfs_entry queue_write_zeroes_max_entry = {
634 .attr = {.name = "write_zeroes_max_bytes", .mode = 0444 },
635 .show = queue_write_zeroes_max_show,
638 static struct queue_sysfs_entry queue_nonrot_entry = {
639 .attr = {.name = "rotational", .mode = 0644 },
640 .show = queue_show_nonrot,
641 .store = queue_store_nonrot,
644 static struct queue_sysfs_entry queue_zoned_entry = {
645 .attr = {.name = "zoned", .mode = 0444 },
646 .show = queue_zoned_show,
649 static struct queue_sysfs_entry queue_nr_zones_entry = {
650 .attr = {.name = "nr_zones", .mode = 0444 },
651 .show = queue_nr_zones_show,
654 static struct queue_sysfs_entry queue_nomerges_entry = {
655 .attr = {.name = "nomerges", .mode = 0644 },
656 .show = queue_nomerges_show,
657 .store = queue_nomerges_store,
660 static struct queue_sysfs_entry queue_rq_affinity_entry = {
661 .attr = {.name = "rq_affinity", .mode = 0644 },
662 .show = queue_rq_affinity_show,
663 .store = queue_rq_affinity_store,
666 static struct queue_sysfs_entry queue_iostats_entry = {
667 .attr = {.name = "iostats", .mode = 0644 },
668 .show = queue_show_iostats,
669 .store = queue_store_iostats,
672 static struct queue_sysfs_entry queue_random_entry = {
673 .attr = {.name = "add_random", .mode = 0644 },
674 .show = queue_show_random,
675 .store = queue_store_random,
678 static struct queue_sysfs_entry queue_poll_entry = {
679 .attr = {.name = "io_poll", .mode = 0644 },
680 .show = queue_poll_show,
681 .store = queue_poll_store,
684 static struct queue_sysfs_entry queue_poll_delay_entry = {
685 .attr = {.name = "io_poll_delay", .mode = 0644 },
686 .show = queue_poll_delay_show,
687 .store = queue_poll_delay_store,
690 static struct queue_sysfs_entry queue_wc_entry = {
691 .attr = {.name = "write_cache", .mode = 0644 },
692 .show = queue_wc_show,
693 .store = queue_wc_store,
696 static struct queue_sysfs_entry queue_fua_entry = {
697 .attr = {.name = "fua", .mode = 0444 },
698 .show = queue_fua_show,
701 static struct queue_sysfs_entry queue_dax_entry = {
702 .attr = {.name = "dax", .mode = 0444 },
703 .show = queue_dax_show,
706 static struct queue_sysfs_entry queue_io_timeout_entry = {
707 .attr = {.name = "io_timeout", .mode = 0644 },
708 .show = queue_io_timeout_show,
709 .store = queue_io_timeout_store,
712 static struct queue_sysfs_entry queue_wb_lat_entry = {
713 .attr = {.name = "wbt_lat_usec", .mode = 0644 },
714 .show = queue_wb_lat_show,
715 .store = queue_wb_lat_store,
718 #ifdef CONFIG_BLK_DEV_THROTTLING_LOW
719 static struct queue_sysfs_entry throtl_sample_time_entry = {
720 .attr = {.name = "throttle_sample_time", .mode = 0644 },
721 .show = blk_throtl_sample_time_show,
722 .store = blk_throtl_sample_time_store,
724 #endif
726 static struct attribute *default_attrs[] = {
727 &queue_requests_entry.attr,
728 &queue_ra_entry.attr,
729 &queue_max_hw_sectors_entry.attr,
730 &queue_max_sectors_entry.attr,
731 &queue_max_segments_entry.attr,
732 &queue_max_discard_segments_entry.attr,
733 &queue_max_integrity_segments_entry.attr,
734 &queue_max_segment_size_entry.attr,
735 &queue_iosched_entry.attr,
736 &queue_hw_sector_size_entry.attr,
737 &queue_logical_block_size_entry.attr,
738 &queue_physical_block_size_entry.attr,
739 &queue_chunk_sectors_entry.attr,
740 &queue_io_min_entry.attr,
741 &queue_io_opt_entry.attr,
742 &queue_discard_granularity_entry.attr,
743 &queue_discard_max_entry.attr,
744 &queue_discard_max_hw_entry.attr,
745 &queue_discard_zeroes_data_entry.attr,
746 &queue_write_same_max_entry.attr,
747 &queue_write_zeroes_max_entry.attr,
748 &queue_nonrot_entry.attr,
749 &queue_zoned_entry.attr,
750 &queue_nr_zones_entry.attr,
751 &queue_nomerges_entry.attr,
752 &queue_rq_affinity_entry.attr,
753 &queue_iostats_entry.attr,
754 &queue_random_entry.attr,
755 &queue_poll_entry.attr,
756 &queue_wc_entry.attr,
757 &queue_fua_entry.attr,
758 &queue_dax_entry.attr,
759 &queue_wb_lat_entry.attr,
760 &queue_poll_delay_entry.attr,
761 &queue_io_timeout_entry.attr,
762 #ifdef CONFIG_BLK_DEV_THROTTLING_LOW
763 &throtl_sample_time_entry.attr,
764 #endif
765 NULL,
768 #define to_queue(atr) container_of((atr), struct queue_sysfs_entry, attr)
770 static ssize_t
771 queue_attr_show(struct kobject *kobj, struct attribute *attr, char *page)
773 struct queue_sysfs_entry *entry = to_queue(attr);
774 struct request_queue *q =
775 container_of(kobj, struct request_queue, kobj);
776 ssize_t res;
778 if (!entry->show)
779 return -EIO;
780 mutex_lock(&q->sysfs_lock);
781 if (blk_queue_dying(q)) {
782 mutex_unlock(&q->sysfs_lock);
783 return -ENOENT;
785 res = entry->show(q, page);
786 mutex_unlock(&q->sysfs_lock);
787 return res;
790 static ssize_t
791 queue_attr_store(struct kobject *kobj, struct attribute *attr,
792 const char *page, size_t length)
794 struct queue_sysfs_entry *entry = to_queue(attr);
795 struct request_queue *q;
796 ssize_t res;
798 if (!entry->store)
799 return -EIO;
801 q = container_of(kobj, struct request_queue, kobj);
802 mutex_lock(&q->sysfs_lock);
803 if (blk_queue_dying(q)) {
804 mutex_unlock(&q->sysfs_lock);
805 return -ENOENT;
807 res = entry->store(q, page, length);
808 mutex_unlock(&q->sysfs_lock);
809 return res;
812 static void blk_free_queue_rcu(struct rcu_head *rcu_head)
814 struct request_queue *q = container_of(rcu_head, struct request_queue,
815 rcu_head);
816 kmem_cache_free(blk_requestq_cachep, q);
820 * __blk_release_queue - release a request queue when it is no longer needed
821 * @work: pointer to the release_work member of the request queue to be released
823 * Description:
824 * blk_release_queue is the counterpart of blk_init_queue(). It should be
825 * called when a request queue is being released; typically when a block
826 * device is being de-registered. Its primary task it to free the queue
827 * itself.
829 * Notes:
830 * The low level driver must have finished any outstanding requests first
831 * via blk_cleanup_queue().
833 * Although blk_release_queue() may be called with preemption disabled,
834 * __blk_release_queue() may sleep.
836 static void __blk_release_queue(struct work_struct *work)
838 struct request_queue *q = container_of(work, typeof(*q), release_work);
840 if (test_bit(QUEUE_FLAG_POLL_STATS, &q->queue_flags))
841 blk_stat_remove_callback(q, q->poll_cb);
842 blk_stat_free_callback(q->poll_cb);
844 if (!blk_queue_dead(q)) {
846 * Last reference was dropped without having called
847 * blk_cleanup_queue().
849 WARN_ONCE(blk_queue_init_done(q),
850 "request queue %p has been registered but blk_cleanup_queue() has not been called for that queue\n",
852 blk_exit_queue(q);
855 WARN(blk_queue_root_blkg(q),
856 "request queue %p is being released but it has not yet been removed from the blkcg controller\n",
859 blk_free_queue_stats(q->stats);
861 blk_queue_free_zone_bitmaps(q);
863 if (queue_is_mq(q))
864 blk_mq_release(q);
866 blk_trace_shutdown(q);
868 if (queue_is_mq(q))
869 blk_mq_debugfs_unregister(q);
871 bioset_exit(&q->bio_split);
873 ida_simple_remove(&blk_queue_ida, q->id);
874 call_rcu(&q->rcu_head, blk_free_queue_rcu);
877 static void blk_release_queue(struct kobject *kobj)
879 struct request_queue *q =
880 container_of(kobj, struct request_queue, kobj);
882 INIT_WORK(&q->release_work, __blk_release_queue);
883 schedule_work(&q->release_work);
886 static const struct sysfs_ops queue_sysfs_ops = {
887 .show = queue_attr_show,
888 .store = queue_attr_store,
891 struct kobj_type blk_queue_ktype = {
892 .sysfs_ops = &queue_sysfs_ops,
893 .default_attrs = default_attrs,
894 .release = blk_release_queue,
898 * blk_register_queue - register a block layer queue with sysfs
899 * @disk: Disk of which the request queue should be registered with sysfs.
901 int blk_register_queue(struct gendisk *disk)
903 int ret;
904 struct device *dev = disk_to_dev(disk);
905 struct request_queue *q = disk->queue;
907 if (WARN_ON(!q))
908 return -ENXIO;
910 WARN_ONCE(test_bit(QUEUE_FLAG_REGISTERED, &q->queue_flags),
911 "%s is registering an already registered queue\n",
912 kobject_name(&dev->kobj));
913 blk_queue_flag_set(QUEUE_FLAG_REGISTERED, q);
916 * SCSI probing may synchronously create and destroy a lot of
917 * request_queues for non-existent devices. Shutting down a fully
918 * functional queue takes measureable wallclock time as RCU grace
919 * periods are involved. To avoid excessive latency in these
920 * cases, a request_queue starts out in a degraded mode which is
921 * faster to shut down and is made fully functional here as
922 * request_queues for non-existent devices never get registered.
924 if (!blk_queue_init_done(q)) {
925 blk_queue_flag_set(QUEUE_FLAG_INIT_DONE, q);
926 percpu_ref_switch_to_percpu(&q->q_usage_counter);
929 ret = blk_trace_init_sysfs(dev);
930 if (ret)
931 return ret;
933 /* Prevent changes through sysfs until registration is completed. */
934 mutex_lock(&q->sysfs_lock);
936 ret = kobject_add(&q->kobj, kobject_get(&dev->kobj), "%s", "queue");
937 if (ret < 0) {
938 blk_trace_remove_sysfs(dev);
939 goto unlock;
942 if (queue_is_mq(q)) {
943 __blk_mq_register_dev(dev, q);
944 blk_mq_debugfs_register(q);
947 kobject_uevent(&q->kobj, KOBJ_ADD);
949 wbt_enable_default(q);
951 blk_throtl_register_queue(q);
953 if (q->elevator) {
954 ret = elv_register_queue(q);
955 if (ret) {
956 mutex_unlock(&q->sysfs_lock);
957 kobject_uevent(&q->kobj, KOBJ_REMOVE);
958 kobject_del(&q->kobj);
959 blk_trace_remove_sysfs(dev);
960 kobject_put(&dev->kobj);
961 return ret;
964 ret = 0;
965 unlock:
966 mutex_unlock(&q->sysfs_lock);
967 return ret;
969 EXPORT_SYMBOL_GPL(blk_register_queue);
972 * blk_unregister_queue - counterpart of blk_register_queue()
973 * @disk: Disk of which the request queue should be unregistered from sysfs.
975 * Note: the caller is responsible for guaranteeing that this function is called
976 * after blk_register_queue() has finished.
978 void blk_unregister_queue(struct gendisk *disk)
980 struct request_queue *q = disk->queue;
982 if (WARN_ON(!q))
983 return;
985 /* Return early if disk->queue was never registered. */
986 if (!test_bit(QUEUE_FLAG_REGISTERED, &q->queue_flags))
987 return;
990 * Since sysfs_remove_dir() prevents adding new directory entries
991 * before removal of existing entries starts, protect against
992 * concurrent elv_iosched_store() calls.
994 mutex_lock(&q->sysfs_lock);
996 blk_queue_flag_clear(QUEUE_FLAG_REGISTERED, q);
999 * Remove the sysfs attributes before unregistering the queue data
1000 * structures that can be modified through sysfs.
1002 if (queue_is_mq(q))
1003 blk_mq_unregister_dev(disk_to_dev(disk), q);
1004 mutex_unlock(&q->sysfs_lock);
1006 kobject_uevent(&q->kobj, KOBJ_REMOVE);
1007 kobject_del(&q->kobj);
1008 blk_trace_remove_sysfs(disk_to_dev(disk));
1010 mutex_lock(&q->sysfs_lock);
1011 if (q->elevator)
1012 elv_unregister_queue(q);
1013 mutex_unlock(&q->sysfs_lock);
1015 kobject_put(&disk_to_dev(disk)->kobj);