2 * Functions related to sysfs handling
4 #include <linux/kernel.h>
5 #include <linux/slab.h>
6 #include <linux/module.h>
8 #include <linux/blkdev.h>
9 #include <linux/backing-dev.h>
10 #include <linux/blktrace_api.h>
11 #include <linux/blk-mq.h>
12 #include <linux/blk-cgroup.h>
17 struct queue_sysfs_entry {
18 struct attribute attr;
19 ssize_t (*show)(struct request_queue *, char *);
20 ssize_t (*store)(struct request_queue *, const char *, size_t);
24 queue_var_show(unsigned long var, char *page)
26 return sprintf(page, "%lu\n", var);
30 queue_var_store(unsigned long *var, const char *page, size_t count)
35 err = kstrtoul(page, 10, &v);
36 if (err || v > UINT_MAX)
44 static ssize_t queue_requests_show(struct request_queue *q, char *page)
46 return queue_var_show(q->nr_requests, (page));
50 queue_requests_store(struct request_queue *q, const char *page, size_t count)
55 if (!q->request_fn && !q->mq_ops)
58 ret = queue_var_store(&nr, page, count);
62 if (nr < BLKDEV_MIN_RQ)
66 err = blk_update_nr_requests(q, nr);
68 err = blk_mq_update_nr_requests(q, nr);
76 static ssize_t queue_ra_show(struct request_queue *q, char *page)
78 unsigned long ra_kb = q->backing_dev_info->ra_pages <<
79 (PAGE_CACHE_SHIFT - 10);
81 return queue_var_show(ra_kb, (page));
85 queue_ra_store(struct request_queue *q, const char *page, size_t count)
88 ssize_t ret = queue_var_store(&ra_kb, page, count);
93 q->backing_dev_info->ra_pages = ra_kb >> (PAGE_CACHE_SHIFT - 10);
98 static ssize_t queue_max_sectors_show(struct request_queue *q, char *page)
100 int max_sectors_kb = queue_max_sectors(q) >> 1;
102 return queue_var_show(max_sectors_kb, (page));
105 static ssize_t queue_max_segments_show(struct request_queue *q, char *page)
107 return queue_var_show(queue_max_segments(q), (page));
110 static ssize_t queue_max_integrity_segments_show(struct request_queue *q, char *page)
112 return queue_var_show(q->limits.max_integrity_segments, (page));
115 static ssize_t queue_max_segment_size_show(struct request_queue *q, char *page)
117 if (blk_queue_cluster(q))
118 return queue_var_show(queue_max_segment_size(q), (page));
120 return queue_var_show(PAGE_CACHE_SIZE, (page));
123 static ssize_t queue_logical_block_size_show(struct request_queue *q, char *page)
125 return queue_var_show(queue_logical_block_size(q), page);
128 static ssize_t queue_physical_block_size_show(struct request_queue *q, char *page)
130 return queue_var_show(queue_physical_block_size(q), page);
133 static ssize_t queue_io_min_show(struct request_queue *q, char *page)
135 return queue_var_show(queue_io_min(q), page);
138 static ssize_t queue_io_opt_show(struct request_queue *q, char *page)
140 return queue_var_show(queue_io_opt(q), page);
143 static ssize_t queue_discard_granularity_show(struct request_queue *q, char *page)
145 return queue_var_show(q->limits.discard_granularity, page);
148 static ssize_t queue_discard_max_hw_show(struct request_queue *q, char *page)
150 unsigned long long val;
152 val = q->limits.max_hw_discard_sectors << 9;
153 return sprintf(page, "%llu\n", val);
156 static ssize_t queue_discard_max_show(struct request_queue *q, char *page)
158 return sprintf(page, "%llu\n",
159 (unsigned long long)q->limits.max_discard_sectors << 9);
162 static ssize_t queue_discard_max_store(struct request_queue *q,
163 const char *page, size_t count)
165 unsigned long max_discard;
166 ssize_t ret = queue_var_store(&max_discard, page, count);
171 if (max_discard & (q->limits.discard_granularity - 1))
175 if (max_discard > UINT_MAX)
178 if (max_discard > q->limits.max_hw_discard_sectors)
179 max_discard = q->limits.max_hw_discard_sectors;
181 q->limits.max_discard_sectors = max_discard;
185 static ssize_t queue_discard_zeroes_data_show(struct request_queue *q, char *page)
187 return queue_var_show(queue_discard_zeroes_data(q), page);
190 static ssize_t queue_write_same_max_show(struct request_queue *q, char *page)
192 return sprintf(page, "%llu\n",
193 (unsigned long long)q->limits.max_write_same_sectors << 9);
198 queue_max_sectors_store(struct request_queue *q, const char *page, size_t count)
200 unsigned long max_sectors_kb,
201 max_hw_sectors_kb = queue_max_hw_sectors(q) >> 1,
202 page_kb = 1 << (PAGE_CACHE_SHIFT - 10);
203 ssize_t ret = queue_var_store(&max_sectors_kb, page, count);
208 max_hw_sectors_kb = min_not_zero(max_hw_sectors_kb, (unsigned long)
209 q->limits.max_dev_sectors >> 1);
211 if (max_sectors_kb > max_hw_sectors_kb || max_sectors_kb < page_kb)
214 spin_lock_irq(q->queue_lock);
215 q->limits.max_sectors = max_sectors_kb << 1;
216 q->backing_dev_info->io_pages = max_sectors_kb >> (PAGE_SHIFT - 10);
217 spin_unlock_irq(q->queue_lock);
222 static ssize_t queue_max_hw_sectors_show(struct request_queue *q, char *page)
224 int max_hw_sectors_kb = queue_max_hw_sectors(q) >> 1;
226 return queue_var_show(max_hw_sectors_kb, (page));
229 #define QUEUE_SYSFS_BIT_FNS(name, flag, neg) \
231 queue_show_##name(struct request_queue *q, char *page) \
234 bit = test_bit(QUEUE_FLAG_##flag, &q->queue_flags); \
235 return queue_var_show(neg ? !bit : bit, page); \
238 queue_store_##name(struct request_queue *q, const char *page, size_t count) \
242 ret = queue_var_store(&val, page, count); \
248 spin_lock_irq(q->queue_lock); \
250 queue_flag_set(QUEUE_FLAG_##flag, q); \
252 queue_flag_clear(QUEUE_FLAG_##flag, q); \
253 spin_unlock_irq(q->queue_lock); \
257 QUEUE_SYSFS_BIT_FNS(nonrot, NONROT, 1);
258 QUEUE_SYSFS_BIT_FNS(random, ADD_RANDOM, 0);
259 QUEUE_SYSFS_BIT_FNS(iostats, IO_STAT, 0);
260 #undef QUEUE_SYSFS_BIT_FNS
262 static ssize_t queue_nomerges_show(struct request_queue *q, char *page)
264 return queue_var_show((blk_queue_nomerges(q) << 1) |
265 blk_queue_noxmerges(q), page);
268 static ssize_t queue_nomerges_store(struct request_queue *q, const char *page,
272 ssize_t ret = queue_var_store(&nm, page, count);
277 spin_lock_irq(q->queue_lock);
278 queue_flag_clear(QUEUE_FLAG_NOMERGES, q);
279 queue_flag_clear(QUEUE_FLAG_NOXMERGES, q);
281 queue_flag_set(QUEUE_FLAG_NOMERGES, q);
283 queue_flag_set(QUEUE_FLAG_NOXMERGES, q);
284 spin_unlock_irq(q->queue_lock);
289 static ssize_t queue_rq_affinity_show(struct request_queue *q, char *page)
291 bool set = test_bit(QUEUE_FLAG_SAME_COMP, &q->queue_flags);
292 bool force = test_bit(QUEUE_FLAG_SAME_FORCE, &q->queue_flags);
294 return queue_var_show(set << force, page);
298 queue_rq_affinity_store(struct request_queue *q, const char *page, size_t count)
300 ssize_t ret = -EINVAL;
304 ret = queue_var_store(&val, page, count);
308 spin_lock_irq(q->queue_lock);
310 queue_flag_set(QUEUE_FLAG_SAME_COMP, q);
311 queue_flag_set(QUEUE_FLAG_SAME_FORCE, q);
312 } else if (val == 1) {
313 queue_flag_set(QUEUE_FLAG_SAME_COMP, q);
314 queue_flag_clear(QUEUE_FLAG_SAME_FORCE, q);
315 } else if (val == 0) {
316 queue_flag_clear(QUEUE_FLAG_SAME_COMP, q);
317 queue_flag_clear(QUEUE_FLAG_SAME_FORCE, q);
319 spin_unlock_irq(q->queue_lock);
324 static ssize_t queue_poll_show(struct request_queue *q, char *page)
326 return queue_var_show(test_bit(QUEUE_FLAG_POLL, &q->queue_flags), page);
329 static ssize_t queue_poll_store(struct request_queue *q, const char *page,
332 unsigned long poll_on;
335 if (!q->mq_ops || !q->mq_ops->poll)
338 ret = queue_var_store(&poll_on, page, count);
342 spin_lock_irq(q->queue_lock);
344 queue_flag_set(QUEUE_FLAG_POLL, q);
346 queue_flag_clear(QUEUE_FLAG_POLL, q);
347 spin_unlock_irq(q->queue_lock);
352 static struct queue_sysfs_entry queue_requests_entry = {
353 .attr = {.name = "nr_requests", .mode = S_IRUGO | S_IWUSR },
354 .show = queue_requests_show,
355 .store = queue_requests_store,
358 static struct queue_sysfs_entry queue_ra_entry = {
359 .attr = {.name = "read_ahead_kb", .mode = S_IRUGO | S_IWUSR },
360 .show = queue_ra_show,
361 .store = queue_ra_store,
364 static struct queue_sysfs_entry queue_max_sectors_entry = {
365 .attr = {.name = "max_sectors_kb", .mode = S_IRUGO | S_IWUSR },
366 .show = queue_max_sectors_show,
367 .store = queue_max_sectors_store,
370 static struct queue_sysfs_entry queue_max_hw_sectors_entry = {
371 .attr = {.name = "max_hw_sectors_kb", .mode = S_IRUGO },
372 .show = queue_max_hw_sectors_show,
375 static struct queue_sysfs_entry queue_max_segments_entry = {
376 .attr = {.name = "max_segments", .mode = S_IRUGO },
377 .show = queue_max_segments_show,
380 static struct queue_sysfs_entry queue_max_integrity_segments_entry = {
381 .attr = {.name = "max_integrity_segments", .mode = S_IRUGO },
382 .show = queue_max_integrity_segments_show,
385 static struct queue_sysfs_entry queue_max_segment_size_entry = {
386 .attr = {.name = "max_segment_size", .mode = S_IRUGO },
387 .show = queue_max_segment_size_show,
390 static struct queue_sysfs_entry queue_iosched_entry = {
391 .attr = {.name = "scheduler", .mode = S_IRUGO | S_IWUSR },
392 .show = elv_iosched_show,
393 .store = elv_iosched_store,
396 static struct queue_sysfs_entry queue_hw_sector_size_entry = {
397 .attr = {.name = "hw_sector_size", .mode = S_IRUGO },
398 .show = queue_logical_block_size_show,
401 static struct queue_sysfs_entry queue_logical_block_size_entry = {
402 .attr = {.name = "logical_block_size", .mode = S_IRUGO },
403 .show = queue_logical_block_size_show,
406 static struct queue_sysfs_entry queue_physical_block_size_entry = {
407 .attr = {.name = "physical_block_size", .mode = S_IRUGO },
408 .show = queue_physical_block_size_show,
411 static struct queue_sysfs_entry queue_io_min_entry = {
412 .attr = {.name = "minimum_io_size", .mode = S_IRUGO },
413 .show = queue_io_min_show,
416 static struct queue_sysfs_entry queue_io_opt_entry = {
417 .attr = {.name = "optimal_io_size", .mode = S_IRUGO },
418 .show = queue_io_opt_show,
421 static struct queue_sysfs_entry queue_discard_granularity_entry = {
422 .attr = {.name = "discard_granularity", .mode = S_IRUGO },
423 .show = queue_discard_granularity_show,
426 static struct queue_sysfs_entry queue_discard_max_hw_entry = {
427 .attr = {.name = "discard_max_hw_bytes", .mode = S_IRUGO },
428 .show = queue_discard_max_hw_show,
431 static struct queue_sysfs_entry queue_discard_max_entry = {
432 .attr = {.name = "discard_max_bytes", .mode = S_IRUGO | S_IWUSR },
433 .show = queue_discard_max_show,
434 .store = queue_discard_max_store,
437 static struct queue_sysfs_entry queue_discard_zeroes_data_entry = {
438 .attr = {.name = "discard_zeroes_data", .mode = S_IRUGO },
439 .show = queue_discard_zeroes_data_show,
442 static struct queue_sysfs_entry queue_write_same_max_entry = {
443 .attr = {.name = "write_same_max_bytes", .mode = S_IRUGO },
444 .show = queue_write_same_max_show,
447 static struct queue_sysfs_entry queue_nonrot_entry = {
448 .attr = {.name = "rotational", .mode = S_IRUGO | S_IWUSR },
449 .show = queue_show_nonrot,
450 .store = queue_store_nonrot,
453 static struct queue_sysfs_entry queue_nomerges_entry = {
454 .attr = {.name = "nomerges", .mode = S_IRUGO | S_IWUSR },
455 .show = queue_nomerges_show,
456 .store = queue_nomerges_store,
459 static struct queue_sysfs_entry queue_rq_affinity_entry = {
460 .attr = {.name = "rq_affinity", .mode = S_IRUGO | S_IWUSR },
461 .show = queue_rq_affinity_show,
462 .store = queue_rq_affinity_store,
465 static struct queue_sysfs_entry queue_iostats_entry = {
466 .attr = {.name = "iostats", .mode = S_IRUGO | S_IWUSR },
467 .show = queue_show_iostats,
468 .store = queue_store_iostats,
471 static struct queue_sysfs_entry queue_random_entry = {
472 .attr = {.name = "add_random", .mode = S_IRUGO | S_IWUSR },
473 .show = queue_show_random,
474 .store = queue_store_random,
477 static struct queue_sysfs_entry queue_poll_entry = {
478 .attr = {.name = "io_poll", .mode = S_IRUGO | S_IWUSR },
479 .show = queue_poll_show,
480 .store = queue_poll_store,
483 static struct attribute *default_attrs[] = {
484 &queue_requests_entry.attr,
485 &queue_ra_entry.attr,
486 &queue_max_hw_sectors_entry.attr,
487 &queue_max_sectors_entry.attr,
488 &queue_max_segments_entry.attr,
489 &queue_max_integrity_segments_entry.attr,
490 &queue_max_segment_size_entry.attr,
491 &queue_iosched_entry.attr,
492 &queue_hw_sector_size_entry.attr,
493 &queue_logical_block_size_entry.attr,
494 &queue_physical_block_size_entry.attr,
495 &queue_io_min_entry.attr,
496 &queue_io_opt_entry.attr,
497 &queue_discard_granularity_entry.attr,
498 &queue_discard_max_entry.attr,
499 &queue_discard_max_hw_entry.attr,
500 &queue_discard_zeroes_data_entry.attr,
501 &queue_write_same_max_entry.attr,
502 &queue_nonrot_entry.attr,
503 &queue_nomerges_entry.attr,
504 &queue_rq_affinity_entry.attr,
505 &queue_iostats_entry.attr,
506 &queue_random_entry.attr,
507 &queue_poll_entry.attr,
511 #define to_queue(atr) container_of((atr), struct queue_sysfs_entry, attr)
514 queue_attr_show(struct kobject *kobj, struct attribute *attr, char *page)
516 struct queue_sysfs_entry *entry = to_queue(attr);
517 struct request_queue *q =
518 container_of(kobj, struct request_queue, kobj);
523 mutex_lock(&q->sysfs_lock);
524 if (blk_queue_dying(q)) {
525 mutex_unlock(&q->sysfs_lock);
528 res = entry->show(q, page);
529 mutex_unlock(&q->sysfs_lock);
534 queue_attr_store(struct kobject *kobj, struct attribute *attr,
535 const char *page, size_t length)
537 struct queue_sysfs_entry *entry = to_queue(attr);
538 struct request_queue *q;
544 q = container_of(kobj, struct request_queue, kobj);
545 mutex_lock(&q->sysfs_lock);
546 if (blk_queue_dying(q)) {
547 mutex_unlock(&q->sysfs_lock);
550 res = entry->store(q, page, length);
551 mutex_unlock(&q->sysfs_lock);
555 static void blk_free_queue_rcu(struct rcu_head *rcu_head)
557 struct request_queue *q = container_of(rcu_head, struct request_queue,
559 kmem_cache_free(blk_requestq_cachep, q);
563 * blk_release_queue: - release a &struct request_queue when it is no longer needed
564 * @kobj: the kobj belonging to the request queue to be released
567 * blk_release_queue is the pair to blk_init_queue() or
568 * blk_queue_make_request(). It should be called when a request queue is
569 * being released; typically when a block device is being de-registered.
570 * Currently, its primary task it to free all the &struct request
571 * structures that were allocated to the queue and the queue itself.
574 * The low level driver must have finished any outstanding requests first
575 * via blk_cleanup_queue().
577 static void blk_release_queue(struct kobject *kobj)
579 struct request_queue *q =
580 container_of(kobj, struct request_queue, kobj);
582 bdi_put(q->backing_dev_info);
586 spin_lock_irq(q->queue_lock);
588 spin_unlock_irq(q->queue_lock);
589 elevator_exit(q->elevator);
592 blk_exit_rl(&q->root_rl);
595 __blk_queue_free_tags(q);
598 blk_free_flush_queue(q->fq);
602 blk_trace_shutdown(q);
605 bioset_free(q->bio_split);
607 ida_simple_remove(&blk_queue_ida, q->id);
608 call_rcu(&q->rcu_head, blk_free_queue_rcu);
611 static const struct sysfs_ops queue_sysfs_ops = {
612 .show = queue_attr_show,
613 .store = queue_attr_store,
616 struct kobj_type blk_queue_ktype = {
617 .sysfs_ops = &queue_sysfs_ops,
618 .default_attrs = default_attrs,
619 .release = blk_release_queue,
622 int blk_register_queue(struct gendisk *disk)
625 struct device *dev = disk_to_dev(disk);
626 struct request_queue *q = disk->queue;
632 * SCSI probing may synchronously create and destroy a lot of
633 * request_queues for non-existent devices. Shutting down a fully
634 * functional queue takes measureable wallclock time as RCU grace
635 * periods are involved. To avoid excessive latency in these
636 * cases, a request_queue starts out in a degraded mode which is
637 * faster to shut down and is made fully functional here as
638 * request_queues for non-existent devices never get registered.
640 if (!blk_queue_init_done(q)) {
641 queue_flag_set_unlocked(QUEUE_FLAG_INIT_DONE, q);
642 percpu_ref_switch_to_percpu(&q->q_usage_counter);
643 blk_queue_bypass_end(q);
646 ret = blk_trace_init_sysfs(dev);
650 ret = kobject_add(&q->kobj, kobject_get(&dev->kobj), "%s", "queue");
652 blk_trace_remove_sysfs(dev);
656 kobject_uevent(&q->kobj, KOBJ_ADD);
659 blk_mq_register_disk(disk);
664 ret = elv_register_queue(q);
666 kobject_uevent(&q->kobj, KOBJ_REMOVE);
667 kobject_del(&q->kobj);
668 blk_trace_remove_sysfs(dev);
669 kobject_put(&dev->kobj);
676 void blk_unregister_queue(struct gendisk *disk)
678 struct request_queue *q = disk->queue;
684 blk_mq_unregister_disk(disk);
687 elv_unregister_queue(q);
689 kobject_uevent(&q->kobj, KOBJ_REMOVE);
690 kobject_del(&q->kobj);
691 blk_trace_remove_sysfs(disk_to_dev(disk));
692 kobject_put(&disk_to_dev(disk)->kobj);