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
<<
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_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_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
)
151 return sprintf(page
, "%llu\n",
152 (unsigned long long)q
->limits
.max_hw_discard_sectors
<< 9);
155 static ssize_t
queue_discard_max_show(struct request_queue
*q
, char *page
)
157 return sprintf(page
, "%llu\n",
158 (unsigned long long)q
->limits
.max_discard_sectors
<< 9);
161 static ssize_t
queue_discard_max_store(struct request_queue
*q
,
162 const char *page
, size_t count
)
164 unsigned long max_discard
;
165 ssize_t ret
= queue_var_store(&max_discard
, page
, count
);
170 if (max_discard
& (q
->limits
.discard_granularity
- 1))
174 if (max_discard
> UINT_MAX
)
177 if (max_discard
> q
->limits
.max_hw_discard_sectors
)
178 max_discard
= q
->limits
.max_hw_discard_sectors
;
180 q
->limits
.max_discard_sectors
= max_discard
;
184 static ssize_t
queue_discard_zeroes_data_show(struct request_queue
*q
, char *page
)
186 return queue_var_show(queue_discard_zeroes_data(q
), page
);
189 static ssize_t
queue_write_same_max_show(struct request_queue
*q
, char *page
)
191 return sprintf(page
, "%llu\n",
192 (unsigned long long)q
->limits
.max_write_same_sectors
<< 9);
197 queue_max_sectors_store(struct request_queue
*q
, const char *page
, size_t count
)
199 unsigned long max_sectors_kb
,
200 max_hw_sectors_kb
= queue_max_hw_sectors(q
) >> 1,
201 page_kb
= 1 << (PAGE_SHIFT
- 10);
202 ssize_t ret
= queue_var_store(&max_sectors_kb
, page
, count
);
207 max_hw_sectors_kb
= min_not_zero(max_hw_sectors_kb
, (unsigned long)
208 q
->limits
.max_dev_sectors
>> 1);
210 if (max_sectors_kb
> max_hw_sectors_kb
|| max_sectors_kb
< page_kb
)
213 spin_lock_irq(q
->queue_lock
);
214 q
->limits
.max_sectors
= max_sectors_kb
<< 1;
215 spin_unlock_irq(q
->queue_lock
);
220 static ssize_t
queue_max_hw_sectors_show(struct request_queue
*q
, char *page
)
222 int max_hw_sectors_kb
= queue_max_hw_sectors(q
) >> 1;
224 return queue_var_show(max_hw_sectors_kb
, (page
));
227 #define QUEUE_SYSFS_BIT_FNS(name, flag, neg) \
229 queue_show_##name(struct request_queue *q, char *page) \
232 bit = test_bit(QUEUE_FLAG_##flag, &q->queue_flags); \
233 return queue_var_show(neg ? !bit : bit, page); \
236 queue_store_##name(struct request_queue *q, const char *page, size_t count) \
240 ret = queue_var_store(&val, page, count); \
246 spin_lock_irq(q->queue_lock); \
248 queue_flag_set(QUEUE_FLAG_##flag, q); \
250 queue_flag_clear(QUEUE_FLAG_##flag, q); \
251 spin_unlock_irq(q->queue_lock); \
255 QUEUE_SYSFS_BIT_FNS(nonrot
, NONROT
, 1);
256 QUEUE_SYSFS_BIT_FNS(random
, ADD_RANDOM
, 0);
257 QUEUE_SYSFS_BIT_FNS(iostats
, IO_STAT
, 0);
258 #undef QUEUE_SYSFS_BIT_FNS
260 static ssize_t
queue_nomerges_show(struct request_queue
*q
, char *page
)
262 return queue_var_show((blk_queue_nomerges(q
) << 1) |
263 blk_queue_noxmerges(q
), page
);
266 static ssize_t
queue_nomerges_store(struct request_queue
*q
, const char *page
,
270 ssize_t ret
= queue_var_store(&nm
, page
, count
);
275 spin_lock_irq(q
->queue_lock
);
276 queue_flag_clear(QUEUE_FLAG_NOMERGES
, q
);
277 queue_flag_clear(QUEUE_FLAG_NOXMERGES
, q
);
279 queue_flag_set(QUEUE_FLAG_NOMERGES
, q
);
281 queue_flag_set(QUEUE_FLAG_NOXMERGES
, q
);
282 spin_unlock_irq(q
->queue_lock
);
287 static ssize_t
queue_rq_affinity_show(struct request_queue
*q
, char *page
)
289 bool set
= test_bit(QUEUE_FLAG_SAME_COMP
, &q
->queue_flags
);
290 bool force
= test_bit(QUEUE_FLAG_SAME_FORCE
, &q
->queue_flags
);
292 return queue_var_show(set
<< force
, page
);
296 queue_rq_affinity_store(struct request_queue
*q
, const char *page
, size_t count
)
298 ssize_t ret
= -EINVAL
;
302 ret
= queue_var_store(&val
, page
, count
);
306 spin_lock_irq(q
->queue_lock
);
308 queue_flag_set(QUEUE_FLAG_SAME_COMP
, q
);
309 queue_flag_set(QUEUE_FLAG_SAME_FORCE
, q
);
310 } else if (val
== 1) {
311 queue_flag_set(QUEUE_FLAG_SAME_COMP
, q
);
312 queue_flag_clear(QUEUE_FLAG_SAME_FORCE
, q
);
313 } else if (val
== 0) {
314 queue_flag_clear(QUEUE_FLAG_SAME_COMP
, q
);
315 queue_flag_clear(QUEUE_FLAG_SAME_FORCE
, q
);
317 spin_unlock_irq(q
->queue_lock
);
322 static ssize_t
queue_poll_show(struct request_queue
*q
, char *page
)
324 return queue_var_show(test_bit(QUEUE_FLAG_POLL
, &q
->queue_flags
), page
);
327 static ssize_t
queue_poll_store(struct request_queue
*q
, const char *page
,
330 unsigned long poll_on
;
333 if (!q
->mq_ops
|| !q
->mq_ops
->poll
)
336 ret
= queue_var_store(&poll_on
, page
, count
);
340 spin_lock_irq(q
->queue_lock
);
342 queue_flag_set(QUEUE_FLAG_POLL
, q
);
344 queue_flag_clear(QUEUE_FLAG_POLL
, q
);
345 spin_unlock_irq(q
->queue_lock
);
350 static ssize_t
queue_wc_show(struct request_queue
*q
, char *page
)
352 if (test_bit(QUEUE_FLAG_WC
, &q
->queue_flags
))
353 return sprintf(page
, "write back\n");
355 return sprintf(page
, "write through\n");
358 static ssize_t
queue_wc_store(struct request_queue
*q
, const char *page
,
363 if (!strncmp(page
, "write back", 10))
365 else if (!strncmp(page
, "write through", 13) ||
366 !strncmp(page
, "none", 4))
372 spin_lock_irq(q
->queue_lock
);
374 queue_flag_set(QUEUE_FLAG_WC
, q
);
376 queue_flag_clear(QUEUE_FLAG_WC
, q
);
377 spin_unlock_irq(q
->queue_lock
);
382 static ssize_t
queue_dax_show(struct request_queue
*q
, char *page
)
384 return queue_var_show(blk_queue_dax(q
), page
);
387 static struct queue_sysfs_entry queue_requests_entry
= {
388 .attr
= {.name
= "nr_requests", .mode
= S_IRUGO
| S_IWUSR
},
389 .show
= queue_requests_show
,
390 .store
= queue_requests_store
,
393 static struct queue_sysfs_entry queue_ra_entry
= {
394 .attr
= {.name
= "read_ahead_kb", .mode
= S_IRUGO
| S_IWUSR
},
395 .show
= queue_ra_show
,
396 .store
= queue_ra_store
,
399 static struct queue_sysfs_entry queue_max_sectors_entry
= {
400 .attr
= {.name
= "max_sectors_kb", .mode
= S_IRUGO
| S_IWUSR
},
401 .show
= queue_max_sectors_show
,
402 .store
= queue_max_sectors_store
,
405 static struct queue_sysfs_entry queue_max_hw_sectors_entry
= {
406 .attr
= {.name
= "max_hw_sectors_kb", .mode
= S_IRUGO
},
407 .show
= queue_max_hw_sectors_show
,
410 static struct queue_sysfs_entry queue_max_segments_entry
= {
411 .attr
= {.name
= "max_segments", .mode
= S_IRUGO
},
412 .show
= queue_max_segments_show
,
415 static struct queue_sysfs_entry queue_max_integrity_segments_entry
= {
416 .attr
= {.name
= "max_integrity_segments", .mode
= S_IRUGO
},
417 .show
= queue_max_integrity_segments_show
,
420 static struct queue_sysfs_entry queue_max_segment_size_entry
= {
421 .attr
= {.name
= "max_segment_size", .mode
= S_IRUGO
},
422 .show
= queue_max_segment_size_show
,
425 static struct queue_sysfs_entry queue_iosched_entry
= {
426 .attr
= {.name
= "scheduler", .mode
= S_IRUGO
| S_IWUSR
},
427 .show
= elv_iosched_show
,
428 .store
= elv_iosched_store
,
431 static struct queue_sysfs_entry queue_hw_sector_size_entry
= {
432 .attr
= {.name
= "hw_sector_size", .mode
= S_IRUGO
},
433 .show
= queue_logical_block_size_show
,
436 static struct queue_sysfs_entry queue_logical_block_size_entry
= {
437 .attr
= {.name
= "logical_block_size", .mode
= S_IRUGO
},
438 .show
= queue_logical_block_size_show
,
441 static struct queue_sysfs_entry queue_physical_block_size_entry
= {
442 .attr
= {.name
= "physical_block_size", .mode
= S_IRUGO
},
443 .show
= queue_physical_block_size_show
,
446 static struct queue_sysfs_entry queue_io_min_entry
= {
447 .attr
= {.name
= "minimum_io_size", .mode
= S_IRUGO
},
448 .show
= queue_io_min_show
,
451 static struct queue_sysfs_entry queue_io_opt_entry
= {
452 .attr
= {.name
= "optimal_io_size", .mode
= S_IRUGO
},
453 .show
= queue_io_opt_show
,
456 static struct queue_sysfs_entry queue_discard_granularity_entry
= {
457 .attr
= {.name
= "discard_granularity", .mode
= S_IRUGO
},
458 .show
= queue_discard_granularity_show
,
461 static struct queue_sysfs_entry queue_discard_max_hw_entry
= {
462 .attr
= {.name
= "discard_max_hw_bytes", .mode
= S_IRUGO
},
463 .show
= queue_discard_max_hw_show
,
466 static struct queue_sysfs_entry queue_discard_max_entry
= {
467 .attr
= {.name
= "discard_max_bytes", .mode
= S_IRUGO
| S_IWUSR
},
468 .show
= queue_discard_max_show
,
469 .store
= queue_discard_max_store
,
472 static struct queue_sysfs_entry queue_discard_zeroes_data_entry
= {
473 .attr
= {.name
= "discard_zeroes_data", .mode
= S_IRUGO
},
474 .show
= queue_discard_zeroes_data_show
,
477 static struct queue_sysfs_entry queue_write_same_max_entry
= {
478 .attr
= {.name
= "write_same_max_bytes", .mode
= S_IRUGO
},
479 .show
= queue_write_same_max_show
,
482 static struct queue_sysfs_entry queue_nonrot_entry
= {
483 .attr
= {.name
= "rotational", .mode
= S_IRUGO
| S_IWUSR
},
484 .show
= queue_show_nonrot
,
485 .store
= queue_store_nonrot
,
488 static struct queue_sysfs_entry queue_nomerges_entry
= {
489 .attr
= {.name
= "nomerges", .mode
= S_IRUGO
| S_IWUSR
},
490 .show
= queue_nomerges_show
,
491 .store
= queue_nomerges_store
,
494 static struct queue_sysfs_entry queue_rq_affinity_entry
= {
495 .attr
= {.name
= "rq_affinity", .mode
= S_IRUGO
| S_IWUSR
},
496 .show
= queue_rq_affinity_show
,
497 .store
= queue_rq_affinity_store
,
500 static struct queue_sysfs_entry queue_iostats_entry
= {
501 .attr
= {.name
= "iostats", .mode
= S_IRUGO
| S_IWUSR
},
502 .show
= queue_show_iostats
,
503 .store
= queue_store_iostats
,
506 static struct queue_sysfs_entry queue_random_entry
= {
507 .attr
= {.name
= "add_random", .mode
= S_IRUGO
| S_IWUSR
},
508 .show
= queue_show_random
,
509 .store
= queue_store_random
,
512 static struct queue_sysfs_entry queue_poll_entry
= {
513 .attr
= {.name
= "io_poll", .mode
= S_IRUGO
| S_IWUSR
},
514 .show
= queue_poll_show
,
515 .store
= queue_poll_store
,
518 static struct queue_sysfs_entry queue_wc_entry
= {
519 .attr
= {.name
= "write_cache", .mode
= S_IRUGO
| S_IWUSR
},
520 .show
= queue_wc_show
,
521 .store
= queue_wc_store
,
524 static struct queue_sysfs_entry queue_dax_entry
= {
525 .attr
= {.name
= "dax", .mode
= S_IRUGO
},
526 .show
= queue_dax_show
,
529 static struct attribute
*default_attrs
[] = {
530 &queue_requests_entry
.attr
,
531 &queue_ra_entry
.attr
,
532 &queue_max_hw_sectors_entry
.attr
,
533 &queue_max_sectors_entry
.attr
,
534 &queue_max_segments_entry
.attr
,
535 &queue_max_integrity_segments_entry
.attr
,
536 &queue_max_segment_size_entry
.attr
,
537 &queue_iosched_entry
.attr
,
538 &queue_hw_sector_size_entry
.attr
,
539 &queue_logical_block_size_entry
.attr
,
540 &queue_physical_block_size_entry
.attr
,
541 &queue_io_min_entry
.attr
,
542 &queue_io_opt_entry
.attr
,
543 &queue_discard_granularity_entry
.attr
,
544 &queue_discard_max_entry
.attr
,
545 &queue_discard_max_hw_entry
.attr
,
546 &queue_discard_zeroes_data_entry
.attr
,
547 &queue_write_same_max_entry
.attr
,
548 &queue_nonrot_entry
.attr
,
549 &queue_nomerges_entry
.attr
,
550 &queue_rq_affinity_entry
.attr
,
551 &queue_iostats_entry
.attr
,
552 &queue_random_entry
.attr
,
553 &queue_poll_entry
.attr
,
554 &queue_wc_entry
.attr
,
555 &queue_dax_entry
.attr
,
559 #define to_queue(atr) container_of((atr), struct queue_sysfs_entry, attr)
562 queue_attr_show(struct kobject
*kobj
, struct attribute
*attr
, char *page
)
564 struct queue_sysfs_entry
*entry
= to_queue(attr
);
565 struct request_queue
*q
=
566 container_of(kobj
, struct request_queue
, kobj
);
571 mutex_lock(&q
->sysfs_lock
);
572 if (blk_queue_dying(q
)) {
573 mutex_unlock(&q
->sysfs_lock
);
576 res
= entry
->show(q
, page
);
577 mutex_unlock(&q
->sysfs_lock
);
582 queue_attr_store(struct kobject
*kobj
, struct attribute
*attr
,
583 const char *page
, size_t length
)
585 struct queue_sysfs_entry
*entry
= to_queue(attr
);
586 struct request_queue
*q
;
592 q
= container_of(kobj
, struct request_queue
, kobj
);
593 mutex_lock(&q
->sysfs_lock
);
594 if (blk_queue_dying(q
)) {
595 mutex_unlock(&q
->sysfs_lock
);
598 res
= entry
->store(q
, page
, length
);
599 mutex_unlock(&q
->sysfs_lock
);
603 static void blk_free_queue_rcu(struct rcu_head
*rcu_head
)
605 struct request_queue
*q
= container_of(rcu_head
, struct request_queue
,
607 kmem_cache_free(blk_requestq_cachep
, q
);
611 * blk_release_queue: - release a &struct request_queue when it is no longer needed
612 * @kobj: the kobj belonging to the request queue to be released
615 * blk_release_queue is the pair to blk_init_queue() or
616 * blk_queue_make_request(). It should be called when a request queue is
617 * being released; typically when a block device is being de-registered.
618 * Currently, its primary task it to free all the &struct request
619 * structures that were allocated to the queue and the queue itself.
622 * The low level driver must have finished any outstanding requests first
623 * via blk_cleanup_queue().
625 static void blk_release_queue(struct kobject
*kobj
)
627 struct request_queue
*q
=
628 container_of(kobj
, struct request_queue
, kobj
);
630 bdi_exit(&q
->backing_dev_info
);
634 spin_lock_irq(q
->queue_lock
);
636 spin_unlock_irq(q
->queue_lock
);
637 elevator_exit(q
->elevator
);
640 blk_exit_rl(&q
->root_rl
);
643 __blk_queue_free_tags(q
);
646 blk_free_flush_queue(q
->fq
);
650 blk_trace_shutdown(q
);
653 bioset_free(q
->bio_split
);
655 ida_simple_remove(&blk_queue_ida
, q
->id
);
656 call_rcu(&q
->rcu_head
, blk_free_queue_rcu
);
659 static const struct sysfs_ops queue_sysfs_ops
= {
660 .show
= queue_attr_show
,
661 .store
= queue_attr_store
,
664 struct kobj_type blk_queue_ktype
= {
665 .sysfs_ops
= &queue_sysfs_ops
,
666 .default_attrs
= default_attrs
,
667 .release
= blk_release_queue
,
670 int blk_register_queue(struct gendisk
*disk
)
673 struct device
*dev
= disk_to_dev(disk
);
674 struct request_queue
*q
= disk
->queue
;
680 * SCSI probing may synchronously create and destroy a lot of
681 * request_queues for non-existent devices. Shutting down a fully
682 * functional queue takes measureable wallclock time as RCU grace
683 * periods are involved. To avoid excessive latency in these
684 * cases, a request_queue starts out in a degraded mode which is
685 * faster to shut down and is made fully functional here as
686 * request_queues for non-existent devices never get registered.
688 if (!blk_queue_init_done(q
)) {
689 queue_flag_set_unlocked(QUEUE_FLAG_INIT_DONE
, q
);
690 percpu_ref_switch_to_percpu(&q
->q_usage_counter
);
691 blk_queue_bypass_end(q
);
694 ret
= blk_trace_init_sysfs(dev
);
698 ret
= kobject_add(&q
->kobj
, kobject_get(&dev
->kobj
), "%s", "queue");
700 blk_trace_remove_sysfs(dev
);
704 kobject_uevent(&q
->kobj
, KOBJ_ADD
);
707 blk_mq_register_dev(dev
, q
);
712 ret
= elv_register_queue(q
);
714 kobject_uevent(&q
->kobj
, KOBJ_REMOVE
);
715 kobject_del(&q
->kobj
);
716 blk_trace_remove_sysfs(dev
);
717 kobject_put(&dev
->kobj
);
724 void blk_unregister_queue(struct gendisk
*disk
)
726 struct request_queue
*q
= disk
->queue
;
732 blk_mq_unregister_dev(disk_to_dev(disk
), q
);
735 elv_unregister_queue(q
);
737 kobject_uevent(&q
->kobj
, KOBJ_REMOVE
);
738 kobject_del(&q
->kobj
);
739 blk_trace_remove_sysfs(disk_to_dev(disk
));
740 kobject_put(&disk_to_dev(disk
)->kobj
);