1 // SPDX-License-Identifier: GPL-2.0
5 * Portions Copyright (C) 2020 Christoph Hellwig
8 #include <linux/module.h>
9 #include <linux/ctype.h>
11 #include <linux/genhd.h>
12 #include <linux/kdev_t.h>
13 #include <linux/kernel.h>
14 #include <linux/blkdev.h>
15 #include <linux/backing-dev.h>
16 #include <linux/init.h>
17 #include <linux/spinlock.h>
18 #include <linux/proc_fs.h>
19 #include <linux/seq_file.h>
20 #include <linux/slab.h>
21 #include <linux/kmod.h>
22 #include <linux/mutex.h>
23 #include <linux/idr.h>
24 #include <linux/log2.h>
25 #include <linux/pm_runtime.h>
26 #include <linux/badblocks.h>
30 static struct kobject
*block_depr
;
32 /* for extended dynamic devt allocation, currently only one major is used */
33 #define NR_EXT_DEVT (1 << MINORBITS)
34 static DEFINE_IDA(ext_devt_ida
);
36 static void disk_check_events(struct disk_events
*ev
,
37 unsigned int *clearing_ptr
);
38 static void disk_alloc_events(struct gendisk
*disk
);
39 static void disk_add_events(struct gendisk
*disk
);
40 static void disk_del_events(struct gendisk
*disk
);
41 static void disk_release_events(struct gendisk
*disk
);
43 void set_capacity(struct gendisk
*disk
, sector_t sectors
)
45 struct block_device
*bdev
= disk
->part0
;
47 spin_lock(&bdev
->bd_size_lock
);
48 i_size_write(bdev
->bd_inode
, (loff_t
)sectors
<< SECTOR_SHIFT
);
49 spin_unlock(&bdev
->bd_size_lock
);
51 EXPORT_SYMBOL(set_capacity
);
54 * Set disk capacity and notify if the size is not currently zero and will not
55 * be set to zero. Returns true if a uevent was sent, otherwise false.
57 bool set_capacity_and_notify(struct gendisk
*disk
, sector_t size
)
59 sector_t capacity
= get_capacity(disk
);
60 char *envp
[] = { "RESIZE=1", NULL
};
62 set_capacity(disk
, size
);
65 * Only print a message and send a uevent if the gendisk is user visible
66 * and alive. This avoids spamming the log and udev when setting the
67 * initial capacity during probing.
69 if (size
== capacity
||
70 (disk
->flags
& (GENHD_FL_UP
| GENHD_FL_HIDDEN
)) != GENHD_FL_UP
)
73 pr_info("%s: detected capacity change from %lld to %lld\n",
74 disk
->disk_name
, capacity
, size
);
77 * Historically we did not send a uevent for changes to/from an empty
80 if (!capacity
|| !size
)
82 kobject_uevent_env(&disk_to_dev(disk
)->kobj
, KOBJ_CHANGE
, envp
);
85 EXPORT_SYMBOL_GPL(set_capacity_and_notify
);
88 * Format the device name of the indicated disk into the supplied buffer and
89 * return a pointer to that same buffer for convenience.
91 char *disk_name(struct gendisk
*hd
, int partno
, char *buf
)
94 snprintf(buf
, BDEVNAME_SIZE
, "%s", hd
->disk_name
);
95 else if (isdigit(hd
->disk_name
[strlen(hd
->disk_name
)-1]))
96 snprintf(buf
, BDEVNAME_SIZE
, "%sp%d", hd
->disk_name
, partno
);
98 snprintf(buf
, BDEVNAME_SIZE
, "%s%d", hd
->disk_name
, partno
);
103 const char *bdevname(struct block_device
*bdev
, char *buf
)
105 return disk_name(bdev
->bd_disk
, bdev
->bd_partno
, buf
);
107 EXPORT_SYMBOL(bdevname
);
109 static void part_stat_read_all(struct block_device
*part
,
110 struct disk_stats
*stat
)
114 memset(stat
, 0, sizeof(struct disk_stats
));
115 for_each_possible_cpu(cpu
) {
116 struct disk_stats
*ptr
= per_cpu_ptr(part
->bd_stats
, cpu
);
119 for (group
= 0; group
< NR_STAT_GROUPS
; group
++) {
120 stat
->nsecs
[group
] += ptr
->nsecs
[group
];
121 stat
->sectors
[group
] += ptr
->sectors
[group
];
122 stat
->ios
[group
] += ptr
->ios
[group
];
123 stat
->merges
[group
] += ptr
->merges
[group
];
126 stat
->io_ticks
+= ptr
->io_ticks
;
130 static unsigned int part_in_flight(struct block_device
*part
)
132 unsigned int inflight
= 0;
135 for_each_possible_cpu(cpu
) {
136 inflight
+= part_stat_local_read_cpu(part
, in_flight
[0], cpu
) +
137 part_stat_local_read_cpu(part
, in_flight
[1], cpu
);
139 if ((int)inflight
< 0)
145 static void part_in_flight_rw(struct block_device
*part
,
146 unsigned int inflight
[2])
152 for_each_possible_cpu(cpu
) {
153 inflight
[0] += part_stat_local_read_cpu(part
, in_flight
[0], cpu
);
154 inflight
[1] += part_stat_local_read_cpu(part
, in_flight
[1], cpu
);
156 if ((int)inflight
[0] < 0)
158 if ((int)inflight
[1] < 0)
163 * Can be deleted altogether. Later.
166 #define BLKDEV_MAJOR_HASH_SIZE 255
167 static struct blk_major_name
{
168 struct blk_major_name
*next
;
171 void (*probe
)(dev_t devt
);
172 } *major_names
[BLKDEV_MAJOR_HASH_SIZE
];
173 static DEFINE_MUTEX(major_names_lock
);
175 /* index in the above - for now: assume no multimajor ranges */
176 static inline int major_to_index(unsigned major
)
178 return major
% BLKDEV_MAJOR_HASH_SIZE
;
181 #ifdef CONFIG_PROC_FS
182 void blkdev_show(struct seq_file
*seqf
, off_t offset
)
184 struct blk_major_name
*dp
;
186 mutex_lock(&major_names_lock
);
187 for (dp
= major_names
[major_to_index(offset
)]; dp
; dp
= dp
->next
)
188 if (dp
->major
== offset
)
189 seq_printf(seqf
, "%3d %s\n", dp
->major
, dp
->name
);
190 mutex_unlock(&major_names_lock
);
192 #endif /* CONFIG_PROC_FS */
195 * __register_blkdev - register a new block device
197 * @major: the requested major device number [1..BLKDEV_MAJOR_MAX-1]. If
198 * @major = 0, try to allocate any unused major number.
199 * @name: the name of the new block device as a zero terminated string
200 * @probe: allback that is called on access to any minor number of @major
202 * The @name must be unique within the system.
204 * The return value depends on the @major input parameter:
206 * - if a major device number was requested in range [1..BLKDEV_MAJOR_MAX-1]
207 * then the function returns zero on success, or a negative error code
208 * - if any unused major number was requested with @major = 0 parameter
209 * then the return value is the allocated major number in range
210 * [1..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise
212 * See Documentation/admin-guide/devices.txt for the list of allocated
215 * Use register_blkdev instead for any new code.
217 int __register_blkdev(unsigned int major
, const char *name
,
218 void (*probe
)(dev_t devt
))
220 struct blk_major_name
**n
, *p
;
223 mutex_lock(&major_names_lock
);
227 for (index
= ARRAY_SIZE(major_names
)-1; index
> 0; index
--) {
228 if (major_names
[index
] == NULL
)
233 printk("%s: failed to get major for %s\n",
242 if (major
>= BLKDEV_MAJOR_MAX
) {
243 pr_err("%s: major requested (%u) is greater than the maximum (%u) for %s\n",
244 __func__
, major
, BLKDEV_MAJOR_MAX
-1, name
);
250 p
= kmalloc(sizeof(struct blk_major_name
), GFP_KERNEL
);
258 strlcpy(p
->name
, name
, sizeof(p
->name
));
260 index
= major_to_index(major
);
262 for (n
= &major_names
[index
]; *n
; n
= &(*n
)->next
) {
263 if ((*n
)->major
== major
)
272 printk("register_blkdev: cannot get major %u for %s\n",
277 mutex_unlock(&major_names_lock
);
280 EXPORT_SYMBOL(__register_blkdev
);
282 void unregister_blkdev(unsigned int major
, const char *name
)
284 struct blk_major_name
**n
;
285 struct blk_major_name
*p
= NULL
;
286 int index
= major_to_index(major
);
288 mutex_lock(&major_names_lock
);
289 for (n
= &major_names
[index
]; *n
; n
= &(*n
)->next
)
290 if ((*n
)->major
== major
)
292 if (!*n
|| strcmp((*n
)->name
, name
)) {
298 mutex_unlock(&major_names_lock
);
302 EXPORT_SYMBOL(unregister_blkdev
);
305 * blk_mangle_minor - scatter minor numbers apart
306 * @minor: minor number to mangle
308 * Scatter consecutively allocated @minor number apart if MANGLE_DEVT
309 * is enabled. Mangling twice gives the original value.
317 static int blk_mangle_minor(int minor
)
319 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
322 for (i
= 0; i
< MINORBITS
/ 2; i
++) {
323 int low
= minor
& (1 << i
);
324 int high
= minor
& (1 << (MINORBITS
- 1 - i
));
325 int distance
= MINORBITS
- 1 - 2 * i
;
327 minor
^= low
| high
; /* clear both bits */
328 low
<<= distance
; /* swap the positions */
330 minor
|= low
| high
; /* and set */
337 * blk_alloc_devt - allocate a dev_t for a block device
338 * @bdev: block device to allocate dev_t for
339 * @devt: out parameter for resulting dev_t
341 * Allocate a dev_t for block device.
344 * 0 on success, allocated dev_t is returned in *@devt. -errno on
350 int blk_alloc_devt(struct block_device
*bdev
, dev_t
*devt
)
352 struct gendisk
*disk
= bdev
->bd_disk
;
355 /* in consecutive minor range? */
356 if (bdev
->bd_partno
< disk
->minors
) {
357 *devt
= MKDEV(disk
->major
, disk
->first_minor
+ bdev
->bd_partno
);
361 idx
= ida_alloc_range(&ext_devt_ida
, 0, NR_EXT_DEVT
, GFP_KERNEL
);
363 return idx
== -ENOSPC
? -EBUSY
: idx
;
365 *devt
= MKDEV(BLOCK_EXT_MAJOR
, blk_mangle_minor(idx
));
370 * blk_free_devt - free a dev_t
371 * @devt: dev_t to free
373 * Free @devt which was allocated using blk_alloc_devt().
378 void blk_free_devt(dev_t devt
)
380 if (MAJOR(devt
) == BLOCK_EXT_MAJOR
)
381 ida_free(&ext_devt_ida
, blk_mangle_minor(MINOR(devt
)));
384 static char *bdevt_str(dev_t devt
, char *buf
)
386 if (MAJOR(devt
) <= 0xff && MINOR(devt
) <= 0xff) {
387 char tbuf
[BDEVT_SIZE
];
388 snprintf(tbuf
, BDEVT_SIZE
, "%02x%02x", MAJOR(devt
), MINOR(devt
));
389 snprintf(buf
, BDEVT_SIZE
, "%-9s", tbuf
);
391 snprintf(buf
, BDEVT_SIZE
, "%03x:%05x", MAJOR(devt
), MINOR(devt
));
396 void disk_uevent(struct gendisk
*disk
, enum kobject_action action
)
398 struct block_device
*part
;
402 xa_for_each(&disk
->part_tbl
, idx
, part
) {
403 if (bdev_is_partition(part
) && !bdev_nr_sectors(part
))
409 kobject_uevent(bdev_kobj(part
), action
);
415 EXPORT_SYMBOL_GPL(disk_uevent
);
417 static void disk_scan_partitions(struct gendisk
*disk
)
419 struct block_device
*bdev
;
421 if (!get_capacity(disk
) || !disk_part_scan_enabled(disk
))
424 set_bit(GD_NEED_PART_SCAN
, &disk
->state
);
425 bdev
= blkdev_get_by_dev(disk_devt(disk
), FMODE_READ
, NULL
);
427 blkdev_put(bdev
, FMODE_READ
);
430 static void register_disk(struct device
*parent
, struct gendisk
*disk
,
431 const struct attribute_group
**groups
)
433 struct device
*ddev
= disk_to_dev(disk
);
436 ddev
->parent
= parent
;
438 dev_set_name(ddev
, "%s", disk
->disk_name
);
440 /* delay uevents, until we scanned partition table */
441 dev_set_uevent_suppress(ddev
, 1);
444 WARN_ON(ddev
->groups
);
445 ddev
->groups
= groups
;
447 if (device_add(ddev
))
449 if (!sysfs_deprecated
) {
450 err
= sysfs_create_link(block_depr
, &ddev
->kobj
,
451 kobject_name(&ddev
->kobj
));
459 * avoid probable deadlock caused by allocating memory with
460 * GFP_KERNEL in runtime_resume callback of its all ancestor
463 pm_runtime_set_memalloc_noio(ddev
, true);
465 disk
->part0
->bd_holder_dir
=
466 kobject_create_and_add("holders", &ddev
->kobj
);
467 disk
->slave_dir
= kobject_create_and_add("slaves", &ddev
->kobj
);
469 if (disk
->flags
& GENHD_FL_HIDDEN
)
472 disk_scan_partitions(disk
);
474 /* announce the disk and partitions after all partitions are created */
475 dev_set_uevent_suppress(ddev
, 0);
476 disk_uevent(disk
, KOBJ_ADD
);
478 if (disk
->queue
->backing_dev_info
->dev
) {
479 err
= sysfs_create_link(&ddev
->kobj
,
480 &disk
->queue
->backing_dev_info
->dev
->kobj
,
487 * __device_add_disk - add disk information to kernel list
488 * @parent: parent device for the disk
489 * @disk: per-device partitioning information
490 * @groups: Additional per-device sysfs groups
491 * @register_queue: register the queue if set to true
493 * This function registers the partitioning information in @disk
496 * FIXME: error handling
498 static void __device_add_disk(struct device
*parent
, struct gendisk
*disk
,
499 const struct attribute_group
**groups
,
506 * The disk queue should now be all set with enough information about
507 * the device for the elevator code to pick an adequate default
508 * elevator if one is needed, that is, for devices requesting queue
512 elevator_init_mq(disk
->queue
);
514 /* minors == 0 indicates to use ext devt from part0 and should
515 * be accompanied with EXT_DEVT flag. Make sure all
516 * parameters make sense.
518 WARN_ON(disk
->minors
&& !(disk
->major
|| disk
->first_minor
));
519 WARN_ON(!disk
->minors
&&
520 !(disk
->flags
& (GENHD_FL_EXT_DEVT
| GENHD_FL_HIDDEN
)));
522 disk
->flags
|= GENHD_FL_UP
;
524 retval
= blk_alloc_devt(disk
->part0
, &devt
);
529 disk
->major
= MAJOR(devt
);
530 disk
->first_minor
= MINOR(devt
);
532 disk_alloc_events(disk
);
534 if (disk
->flags
& GENHD_FL_HIDDEN
) {
536 * Don't let hidden disks show up in /proc/partitions,
537 * and don't bother scanning for partitions either.
539 disk
->flags
|= GENHD_FL_SUPPRESS_PARTITION_INFO
;
540 disk
->flags
|= GENHD_FL_NO_PART_SCAN
;
542 struct backing_dev_info
*bdi
= disk
->queue
->backing_dev_info
;
543 struct device
*dev
= disk_to_dev(disk
);
546 /* Register BDI before referencing it from bdev */
548 ret
= bdi_register(bdi
, "%u:%u", MAJOR(devt
), MINOR(devt
));
550 bdi_set_owner(bdi
, dev
);
551 bdev_add(disk
->part0
, devt
);
553 register_disk(parent
, disk
, groups
);
555 blk_register_queue(disk
);
558 * Take an extra ref on queue which will be put on disk_release()
559 * so that it sticks around as long as @disk is there.
561 WARN_ON_ONCE(!blk_get_queue(disk
->queue
));
563 disk_add_events(disk
);
564 blk_integrity_add(disk
);
567 void device_add_disk(struct device
*parent
, struct gendisk
*disk
,
568 const struct attribute_group
**groups
)
571 __device_add_disk(parent
, disk
, groups
, true);
573 EXPORT_SYMBOL(device_add_disk
);
575 void device_add_disk_no_queue_reg(struct device
*parent
, struct gendisk
*disk
)
577 __device_add_disk(parent
, disk
, NULL
, false);
579 EXPORT_SYMBOL(device_add_disk_no_queue_reg
);
582 * del_gendisk - remove the gendisk
583 * @disk: the struct gendisk to remove
585 * Removes the gendisk and all its associated resources. This deletes the
586 * partitions associated with the gendisk, and unregisters the associated
589 * This is the counter to the respective __device_add_disk() call.
591 * The final removal of the struct gendisk happens when its refcount reaches 0
592 * with put_disk(), which should be called after del_gendisk(), if
593 * __device_add_disk() was used.
595 * Drivers exist which depend on the release of the gendisk to be synchronous,
596 * it should not be deferred.
600 void del_gendisk(struct gendisk
*disk
)
604 if (WARN_ON_ONCE(!disk
->queue
))
607 blk_integrity_del(disk
);
608 disk_del_events(disk
);
610 mutex_lock(&disk
->part0
->bd_mutex
);
611 disk
->flags
&= ~GENHD_FL_UP
;
612 blk_drop_partitions(disk
);
613 mutex_unlock(&disk
->part0
->bd_mutex
);
615 fsync_bdev(disk
->part0
);
616 __invalidate_device(disk
->part0
, true);
619 * Unhash the bdev inode for this device so that it can't be looked
620 * up any more even if openers still hold references to it.
622 remove_inode_hash(disk
->part0
->bd_inode
);
624 set_capacity(disk
, 0);
626 if (!(disk
->flags
& GENHD_FL_HIDDEN
)) {
627 sysfs_remove_link(&disk_to_dev(disk
)->kobj
, "bdi");
630 * Unregister bdi before releasing device numbers (as they can
631 * get reused and we'd get clashes in sysfs).
633 bdi_unregister(disk
->queue
->backing_dev_info
);
636 blk_unregister_queue(disk
);
638 kobject_put(disk
->part0
->bd_holder_dir
);
639 kobject_put(disk
->slave_dir
);
641 part_stat_set_all(disk
->part0
, 0);
642 disk
->part0
->bd_stamp
= 0;
643 if (!sysfs_deprecated
)
644 sysfs_remove_link(block_depr
, dev_name(disk_to_dev(disk
)));
645 pm_runtime_set_memalloc_noio(disk_to_dev(disk
), false);
646 device_del(disk_to_dev(disk
));
648 EXPORT_SYMBOL(del_gendisk
);
650 /* sysfs access to bad-blocks list. */
651 static ssize_t
disk_badblocks_show(struct device
*dev
,
652 struct device_attribute
*attr
,
655 struct gendisk
*disk
= dev_to_disk(dev
);
658 return sprintf(page
, "\n");
660 return badblocks_show(disk
->bb
, page
, 0);
663 static ssize_t
disk_badblocks_store(struct device
*dev
,
664 struct device_attribute
*attr
,
665 const char *page
, size_t len
)
667 struct gendisk
*disk
= dev_to_disk(dev
);
672 return badblocks_store(disk
->bb
, page
, len
, 0);
675 void blk_request_module(dev_t devt
)
677 unsigned int major
= MAJOR(devt
);
678 struct blk_major_name
**n
;
680 mutex_lock(&major_names_lock
);
681 for (n
= &major_names
[major_to_index(major
)]; *n
; n
= &(*n
)->next
) {
682 if ((*n
)->major
== major
&& (*n
)->probe
) {
684 mutex_unlock(&major_names_lock
);
688 mutex_unlock(&major_names_lock
);
690 if (request_module("block-major-%d-%d", MAJOR(devt
), MINOR(devt
)) > 0)
691 /* Make old-style 2.4 aliases work */
692 request_module("block-major-%d", MAJOR(devt
));
696 * bdget_disk - do bdget() by gendisk and partition number
697 * @disk: gendisk of interest
698 * @partno: partition number
700 * Find partition @partno from @disk, do bdget() on it.
706 * Resulting block_device on success, NULL on failure.
708 struct block_device
*bdget_disk(struct gendisk
*disk
, int partno
)
710 struct block_device
*bdev
= NULL
;
713 bdev
= xa_load(&disk
->part_tbl
, partno
);
714 if (bdev
&& !bdgrab(bdev
))
722 * print a full list of all partitions - intended for places where the root
723 * filesystem can't be mounted and thus to give the victim some idea of what
726 void __init
printk_all_partitions(void)
728 struct class_dev_iter iter
;
731 class_dev_iter_init(&iter
, &block_class
, NULL
, &disk_type
);
732 while ((dev
= class_dev_iter_next(&iter
))) {
733 struct gendisk
*disk
= dev_to_disk(dev
);
734 struct block_device
*part
;
735 char name_buf
[BDEVNAME_SIZE
];
736 char devt_buf
[BDEVT_SIZE
];
740 * Don't show empty devices or things that have been
743 if (get_capacity(disk
) == 0 ||
744 (disk
->flags
& GENHD_FL_SUPPRESS_PARTITION_INFO
))
748 * Note, unlike /proc/partitions, I am showing the numbers in
749 * hex - the same format as the root= option takes.
752 xa_for_each(&disk
->part_tbl
, idx
, part
) {
753 if (!bdev_nr_sectors(part
))
755 printk("%s%s %10llu %s %s",
756 bdev_is_partition(part
) ? " " : "",
757 bdevt_str(part
->bd_dev
, devt_buf
),
758 bdev_nr_sectors(part
) >> 1,
759 disk_name(disk
, part
->bd_partno
, name_buf
),
761 part
->bd_meta_info
->uuid
: "");
762 if (bdev_is_partition(part
))
764 else if (dev
->parent
&& dev
->parent
->driver
)
765 printk(" driver: %s\n",
766 dev
->parent
->driver
->name
);
768 printk(" (driver?)\n");
772 class_dev_iter_exit(&iter
);
775 #ifdef CONFIG_PROC_FS
777 static void *disk_seqf_start(struct seq_file
*seqf
, loff_t
*pos
)
780 struct class_dev_iter
*iter
;
783 iter
= kmalloc(sizeof(*iter
), GFP_KERNEL
);
785 return ERR_PTR(-ENOMEM
);
787 seqf
->private = iter
;
788 class_dev_iter_init(iter
, &block_class
, NULL
, &disk_type
);
790 dev
= class_dev_iter_next(iter
);
795 return dev_to_disk(dev
);
798 static void *disk_seqf_next(struct seq_file
*seqf
, void *v
, loff_t
*pos
)
803 dev
= class_dev_iter_next(seqf
->private);
805 return dev_to_disk(dev
);
810 static void disk_seqf_stop(struct seq_file
*seqf
, void *v
)
812 struct class_dev_iter
*iter
= seqf
->private;
814 /* stop is called even after start failed :-( */
816 class_dev_iter_exit(iter
);
818 seqf
->private = NULL
;
822 static void *show_partition_start(struct seq_file
*seqf
, loff_t
*pos
)
826 p
= disk_seqf_start(seqf
, pos
);
827 if (!IS_ERR_OR_NULL(p
) && !*pos
)
828 seq_puts(seqf
, "major minor #blocks name\n\n");
832 static int show_partition(struct seq_file
*seqf
, void *v
)
834 struct gendisk
*sgp
= v
;
835 struct block_device
*part
;
837 char buf
[BDEVNAME_SIZE
];
839 /* Don't show non-partitionable removeable devices or empty devices */
840 if (!get_capacity(sgp
) || (!disk_max_parts(sgp
) &&
841 (sgp
->flags
& GENHD_FL_REMOVABLE
)))
843 if (sgp
->flags
& GENHD_FL_SUPPRESS_PARTITION_INFO
)
847 xa_for_each(&sgp
->part_tbl
, idx
, part
) {
848 if (!bdev_nr_sectors(part
))
850 seq_printf(seqf
, "%4d %7d %10llu %s\n",
851 MAJOR(part
->bd_dev
), MINOR(part
->bd_dev
),
852 bdev_nr_sectors(part
) >> 1,
853 disk_name(sgp
, part
->bd_partno
, buf
));
859 static const struct seq_operations partitions_op
= {
860 .start
= show_partition_start
,
861 .next
= disk_seqf_next
,
862 .stop
= disk_seqf_stop
,
863 .show
= show_partition
867 static int __init
genhd_device_init(void)
871 block_class
.dev_kobj
= sysfs_dev_block_kobj
;
872 error
= class_register(&block_class
);
877 register_blkdev(BLOCK_EXT_MAJOR
, "blkext");
879 /* create top-level block dir */
880 if (!sysfs_deprecated
)
881 block_depr
= kobject_create_and_add("block", NULL
);
885 subsys_initcall(genhd_device_init
);
887 static ssize_t
disk_range_show(struct device
*dev
,
888 struct device_attribute
*attr
, char *buf
)
890 struct gendisk
*disk
= dev_to_disk(dev
);
892 return sprintf(buf
, "%d\n", disk
->minors
);
895 static ssize_t
disk_ext_range_show(struct device
*dev
,
896 struct device_attribute
*attr
, char *buf
)
898 struct gendisk
*disk
= dev_to_disk(dev
);
900 return sprintf(buf
, "%d\n", disk_max_parts(disk
));
903 static ssize_t
disk_removable_show(struct device
*dev
,
904 struct device_attribute
*attr
, char *buf
)
906 struct gendisk
*disk
= dev_to_disk(dev
);
908 return sprintf(buf
, "%d\n",
909 (disk
->flags
& GENHD_FL_REMOVABLE
? 1 : 0));
912 static ssize_t
disk_hidden_show(struct device
*dev
,
913 struct device_attribute
*attr
, char *buf
)
915 struct gendisk
*disk
= dev_to_disk(dev
);
917 return sprintf(buf
, "%d\n",
918 (disk
->flags
& GENHD_FL_HIDDEN
? 1 : 0));
921 static ssize_t
disk_ro_show(struct device
*dev
,
922 struct device_attribute
*attr
, char *buf
)
924 struct gendisk
*disk
= dev_to_disk(dev
);
926 return sprintf(buf
, "%d\n", get_disk_ro(disk
) ? 1 : 0);
929 ssize_t
part_size_show(struct device
*dev
,
930 struct device_attribute
*attr
, char *buf
)
932 return sprintf(buf
, "%llu\n", bdev_nr_sectors(dev_to_bdev(dev
)));
935 ssize_t
part_stat_show(struct device
*dev
,
936 struct device_attribute
*attr
, char *buf
)
938 struct block_device
*bdev
= dev_to_bdev(dev
);
939 struct request_queue
*q
= bdev
->bd_disk
->queue
;
940 struct disk_stats stat
;
941 unsigned int inflight
;
943 part_stat_read_all(bdev
, &stat
);
945 inflight
= blk_mq_in_flight(q
, bdev
);
947 inflight
= part_in_flight(bdev
);
950 "%8lu %8lu %8llu %8u "
951 "%8lu %8lu %8llu %8u "
953 "%8lu %8lu %8llu %8u "
957 stat
.merges
[STAT_READ
],
958 (unsigned long long)stat
.sectors
[STAT_READ
],
959 (unsigned int)div_u64(stat
.nsecs
[STAT_READ
], NSEC_PER_MSEC
),
960 stat
.ios
[STAT_WRITE
],
961 stat
.merges
[STAT_WRITE
],
962 (unsigned long long)stat
.sectors
[STAT_WRITE
],
963 (unsigned int)div_u64(stat
.nsecs
[STAT_WRITE
], NSEC_PER_MSEC
),
965 jiffies_to_msecs(stat
.io_ticks
),
966 (unsigned int)div_u64(stat
.nsecs
[STAT_READ
] +
967 stat
.nsecs
[STAT_WRITE
] +
968 stat
.nsecs
[STAT_DISCARD
] +
969 stat
.nsecs
[STAT_FLUSH
],
971 stat
.ios
[STAT_DISCARD
],
972 stat
.merges
[STAT_DISCARD
],
973 (unsigned long long)stat
.sectors
[STAT_DISCARD
],
974 (unsigned int)div_u64(stat
.nsecs
[STAT_DISCARD
], NSEC_PER_MSEC
),
975 stat
.ios
[STAT_FLUSH
],
976 (unsigned int)div_u64(stat
.nsecs
[STAT_FLUSH
], NSEC_PER_MSEC
));
979 ssize_t
part_inflight_show(struct device
*dev
, struct device_attribute
*attr
,
982 struct block_device
*bdev
= dev_to_bdev(dev
);
983 struct request_queue
*q
= bdev
->bd_disk
->queue
;
984 unsigned int inflight
[2];
987 blk_mq_in_flight_rw(q
, bdev
, inflight
);
989 part_in_flight_rw(bdev
, inflight
);
991 return sprintf(buf
, "%8u %8u\n", inflight
[0], inflight
[1]);
994 static ssize_t
disk_capability_show(struct device
*dev
,
995 struct device_attribute
*attr
, char *buf
)
997 struct gendisk
*disk
= dev_to_disk(dev
);
999 return sprintf(buf
, "%x\n", disk
->flags
);
1002 static ssize_t
disk_alignment_offset_show(struct device
*dev
,
1003 struct device_attribute
*attr
,
1006 struct gendisk
*disk
= dev_to_disk(dev
);
1008 return sprintf(buf
, "%d\n", queue_alignment_offset(disk
->queue
));
1011 static ssize_t
disk_discard_alignment_show(struct device
*dev
,
1012 struct device_attribute
*attr
,
1015 struct gendisk
*disk
= dev_to_disk(dev
);
1017 return sprintf(buf
, "%d\n", queue_discard_alignment(disk
->queue
));
1020 static DEVICE_ATTR(range
, 0444, disk_range_show
, NULL
);
1021 static DEVICE_ATTR(ext_range
, 0444, disk_ext_range_show
, NULL
);
1022 static DEVICE_ATTR(removable
, 0444, disk_removable_show
, NULL
);
1023 static DEVICE_ATTR(hidden
, 0444, disk_hidden_show
, NULL
);
1024 static DEVICE_ATTR(ro
, 0444, disk_ro_show
, NULL
);
1025 static DEVICE_ATTR(size
, 0444, part_size_show
, NULL
);
1026 static DEVICE_ATTR(alignment_offset
, 0444, disk_alignment_offset_show
, NULL
);
1027 static DEVICE_ATTR(discard_alignment
, 0444, disk_discard_alignment_show
, NULL
);
1028 static DEVICE_ATTR(capability
, 0444, disk_capability_show
, NULL
);
1029 static DEVICE_ATTR(stat
, 0444, part_stat_show
, NULL
);
1030 static DEVICE_ATTR(inflight
, 0444, part_inflight_show
, NULL
);
1031 static DEVICE_ATTR(badblocks
, 0644, disk_badblocks_show
, disk_badblocks_store
);
1033 #ifdef CONFIG_FAIL_MAKE_REQUEST
1034 ssize_t
part_fail_show(struct device
*dev
,
1035 struct device_attribute
*attr
, char *buf
)
1037 return sprintf(buf
, "%d\n", dev_to_bdev(dev
)->bd_make_it_fail
);
1040 ssize_t
part_fail_store(struct device
*dev
,
1041 struct device_attribute
*attr
,
1042 const char *buf
, size_t count
)
1046 if (count
> 0 && sscanf(buf
, "%d", &i
) > 0)
1047 dev_to_bdev(dev
)->bd_make_it_fail
= i
;
1052 static struct device_attribute dev_attr_fail
=
1053 __ATTR(make
-it
-fail
, 0644, part_fail_show
, part_fail_store
);
1054 #endif /* CONFIG_FAIL_MAKE_REQUEST */
1056 #ifdef CONFIG_FAIL_IO_TIMEOUT
1057 static struct device_attribute dev_attr_fail_timeout
=
1058 __ATTR(io
-timeout
-fail
, 0644, part_timeout_show
, part_timeout_store
);
1061 static struct attribute
*disk_attrs
[] = {
1062 &dev_attr_range
.attr
,
1063 &dev_attr_ext_range
.attr
,
1064 &dev_attr_removable
.attr
,
1065 &dev_attr_hidden
.attr
,
1067 &dev_attr_size
.attr
,
1068 &dev_attr_alignment_offset
.attr
,
1069 &dev_attr_discard_alignment
.attr
,
1070 &dev_attr_capability
.attr
,
1071 &dev_attr_stat
.attr
,
1072 &dev_attr_inflight
.attr
,
1073 &dev_attr_badblocks
.attr
,
1074 #ifdef CONFIG_FAIL_MAKE_REQUEST
1075 &dev_attr_fail
.attr
,
1077 #ifdef CONFIG_FAIL_IO_TIMEOUT
1078 &dev_attr_fail_timeout
.attr
,
1083 static umode_t
disk_visible(struct kobject
*kobj
, struct attribute
*a
, int n
)
1085 struct device
*dev
= container_of(kobj
, typeof(*dev
), kobj
);
1086 struct gendisk
*disk
= dev_to_disk(dev
);
1088 if (a
== &dev_attr_badblocks
.attr
&& !disk
->bb
)
1093 static struct attribute_group disk_attr_group
= {
1094 .attrs
= disk_attrs
,
1095 .is_visible
= disk_visible
,
1098 static const struct attribute_group
*disk_attr_groups
[] = {
1104 * disk_release - releases all allocated resources of the gendisk
1105 * @dev: the device representing this disk
1107 * This function releases all allocated resources of the gendisk.
1109 * Drivers which used __device_add_disk() have a gendisk with a request_queue
1110 * assigned. Since the request_queue sits on top of the gendisk for these
1111 * drivers we also call blk_put_queue() for them, and we expect the
1112 * request_queue refcount to reach 0 at this point, and so the request_queue
1113 * will also be freed prior to the disk.
1115 * Context: can sleep
1117 static void disk_release(struct device
*dev
)
1119 struct gendisk
*disk
= dev_to_disk(dev
);
1123 blk_free_devt(dev
->devt
);
1124 disk_release_events(disk
);
1125 kfree(disk
->random
);
1126 xa_destroy(&disk
->part_tbl
);
1129 blk_put_queue(disk
->queue
);
1132 struct class block_class
= {
1136 static char *block_devnode(struct device
*dev
, umode_t
*mode
,
1137 kuid_t
*uid
, kgid_t
*gid
)
1139 struct gendisk
*disk
= dev_to_disk(dev
);
1141 if (disk
->fops
->devnode
)
1142 return disk
->fops
->devnode(disk
, mode
);
1146 const struct device_type disk_type
= {
1148 .groups
= disk_attr_groups
,
1149 .release
= disk_release
,
1150 .devnode
= block_devnode
,
1153 #ifdef CONFIG_PROC_FS
1155 * aggregate disk stat collector. Uses the same stats that the sysfs
1156 * entries do, above, but makes them available through one seq_file.
1158 * The output looks suspiciously like /proc/partitions with a bunch of
1161 static int diskstats_show(struct seq_file
*seqf
, void *v
)
1163 struct gendisk
*gp
= v
;
1164 struct block_device
*hd
;
1165 char buf
[BDEVNAME_SIZE
];
1166 unsigned int inflight
;
1167 struct disk_stats stat
;
1171 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1172 seq_puts(seqf, "major minor name"
1173 " rio rmerge rsect ruse wio wmerge "
1174 "wsect wuse running use aveq"
1179 xa_for_each(&gp
->part_tbl
, idx
, hd
) {
1180 if (bdev_is_partition(hd
) && !bdev_nr_sectors(hd
))
1182 part_stat_read_all(hd
, &stat
);
1183 if (queue_is_mq(gp
->queue
))
1184 inflight
= blk_mq_in_flight(gp
->queue
, hd
);
1186 inflight
= part_in_flight(hd
);
1188 seq_printf(seqf
, "%4d %7d %s "
1195 MAJOR(hd
->bd_dev
), MINOR(hd
->bd_dev
),
1196 disk_name(gp
, hd
->bd_partno
, buf
),
1197 stat
.ios
[STAT_READ
],
1198 stat
.merges
[STAT_READ
],
1199 stat
.sectors
[STAT_READ
],
1200 (unsigned int)div_u64(stat
.nsecs
[STAT_READ
],
1202 stat
.ios
[STAT_WRITE
],
1203 stat
.merges
[STAT_WRITE
],
1204 stat
.sectors
[STAT_WRITE
],
1205 (unsigned int)div_u64(stat
.nsecs
[STAT_WRITE
],
1208 jiffies_to_msecs(stat
.io_ticks
),
1209 (unsigned int)div_u64(stat
.nsecs
[STAT_READ
] +
1210 stat
.nsecs
[STAT_WRITE
] +
1211 stat
.nsecs
[STAT_DISCARD
] +
1212 stat
.nsecs
[STAT_FLUSH
],
1214 stat
.ios
[STAT_DISCARD
],
1215 stat
.merges
[STAT_DISCARD
],
1216 stat
.sectors
[STAT_DISCARD
],
1217 (unsigned int)div_u64(stat
.nsecs
[STAT_DISCARD
],
1219 stat
.ios
[STAT_FLUSH
],
1220 (unsigned int)div_u64(stat
.nsecs
[STAT_FLUSH
],
1229 static const struct seq_operations diskstats_op
= {
1230 .start
= disk_seqf_start
,
1231 .next
= disk_seqf_next
,
1232 .stop
= disk_seqf_stop
,
1233 .show
= diskstats_show
1236 static int __init
proc_genhd_init(void)
1238 proc_create_seq("diskstats", 0, NULL
, &diskstats_op
);
1239 proc_create_seq("partitions", 0, NULL
, &partitions_op
);
1242 module_init(proc_genhd_init
);
1243 #endif /* CONFIG_PROC_FS */
1245 dev_t
blk_lookup_devt(const char *name
, int partno
)
1247 dev_t devt
= MKDEV(0, 0);
1248 struct class_dev_iter iter
;
1251 class_dev_iter_init(&iter
, &block_class
, NULL
, &disk_type
);
1252 while ((dev
= class_dev_iter_next(&iter
))) {
1253 struct gendisk
*disk
= dev_to_disk(dev
);
1254 struct block_device
*part
;
1256 if (strcmp(dev_name(dev
), name
))
1259 if (partno
< disk
->minors
) {
1260 /* We need to return the right devno, even
1261 * if the partition doesn't exist yet.
1263 devt
= MKDEV(MAJOR(dev
->devt
),
1264 MINOR(dev
->devt
) + partno
);
1267 part
= bdget_disk(disk
, partno
);
1269 devt
= part
->bd_dev
;
1274 class_dev_iter_exit(&iter
);
1278 struct gendisk
*__alloc_disk_node(int minors
, int node_id
)
1280 struct gendisk
*disk
;
1282 if (minors
> DISK_MAX_PARTS
) {
1284 "block: can't allocate more than %d partitions\n",
1286 minors
= DISK_MAX_PARTS
;
1289 disk
= kzalloc_node(sizeof(struct gendisk
), GFP_KERNEL
, node_id
);
1293 disk
->part0
= bdev_alloc(disk
, 0);
1297 disk
->node_id
= node_id
;
1298 xa_init(&disk
->part_tbl
);
1299 if (xa_insert(&disk
->part_tbl
, 0, disk
->part0
, GFP_KERNEL
))
1300 goto out_destroy_part_tbl
;
1302 disk
->minors
= minors
;
1303 rand_initialize_disk(disk
);
1304 disk_to_dev(disk
)->class = &block_class
;
1305 disk_to_dev(disk
)->type
= &disk_type
;
1306 device_initialize(disk_to_dev(disk
));
1309 out_destroy_part_tbl
:
1310 xa_destroy(&disk
->part_tbl
);
1316 EXPORT_SYMBOL(__alloc_disk_node
);
1319 * put_disk - decrements the gendisk refcount
1320 * @disk: the struct gendisk to decrement the refcount for
1322 * This decrements the refcount for the struct gendisk. When this reaches 0
1323 * we'll have disk_release() called.
1325 * Context: Any context, but the last reference must not be dropped from
1328 void put_disk(struct gendisk
*disk
)
1331 put_device(disk_to_dev(disk
));
1333 EXPORT_SYMBOL(put_disk
);
1335 static void set_disk_ro_uevent(struct gendisk
*gd
, int ro
)
1337 char event
[] = "DISK_RO=1";
1338 char *envp
[] = { event
, NULL
};
1342 kobject_uevent_env(&disk_to_dev(gd
)->kobj
, KOBJ_CHANGE
, envp
);
1346 * set_disk_ro - set a gendisk read-only
1347 * @disk: gendisk to operate on
1348 * @read_only: %true to set the disk read-only, %false set the disk read/write
1350 * This function is used to indicate whether a given disk device should have its
1351 * read-only flag set. set_disk_ro() is typically used by device drivers to
1352 * indicate whether the underlying physical device is write-protected.
1354 void set_disk_ro(struct gendisk
*disk
, bool read_only
)
1357 if (test_and_set_bit(GD_READ_ONLY
, &disk
->state
))
1360 if (!test_and_clear_bit(GD_READ_ONLY
, &disk
->state
))
1363 set_disk_ro_uevent(disk
, read_only
);
1365 EXPORT_SYMBOL(set_disk_ro
);
1367 int bdev_read_only(struct block_device
*bdev
)
1369 return bdev
->bd_read_only
|| get_disk_ro(bdev
->bd_disk
);
1371 EXPORT_SYMBOL(bdev_read_only
);
1374 * Disk events - monitor disk events like media change and eject request.
1376 struct disk_events
{
1377 struct list_head node
; /* all disk_event's */
1378 struct gendisk
*disk
; /* the associated disk */
1381 struct mutex block_mutex
; /* protects blocking */
1382 int block
; /* event blocking depth */
1383 unsigned int pending
; /* events already sent out */
1384 unsigned int clearing
; /* events being cleared */
1386 long poll_msecs
; /* interval, -1 for default */
1387 struct delayed_work dwork
;
1390 static const char *disk_events_strs
[] = {
1391 [ilog2(DISK_EVENT_MEDIA_CHANGE
)] = "media_change",
1392 [ilog2(DISK_EVENT_EJECT_REQUEST
)] = "eject_request",
1395 static char *disk_uevents
[] = {
1396 [ilog2(DISK_EVENT_MEDIA_CHANGE
)] = "DISK_MEDIA_CHANGE=1",
1397 [ilog2(DISK_EVENT_EJECT_REQUEST
)] = "DISK_EJECT_REQUEST=1",
1400 /* list of all disk_events */
1401 static DEFINE_MUTEX(disk_events_mutex
);
1402 static LIST_HEAD(disk_events
);
1404 /* disable in-kernel polling by default */
1405 static unsigned long disk_events_dfl_poll_msecs
;
1407 static unsigned long disk_events_poll_jiffies(struct gendisk
*disk
)
1409 struct disk_events
*ev
= disk
->ev
;
1410 long intv_msecs
= 0;
1413 * If device-specific poll interval is set, always use it. If
1414 * the default is being used, poll if the POLL flag is set.
1416 if (ev
->poll_msecs
>= 0)
1417 intv_msecs
= ev
->poll_msecs
;
1418 else if (disk
->event_flags
& DISK_EVENT_FLAG_POLL
)
1419 intv_msecs
= disk_events_dfl_poll_msecs
;
1421 return msecs_to_jiffies(intv_msecs
);
1425 * disk_block_events - block and flush disk event checking
1426 * @disk: disk to block events for
1428 * On return from this function, it is guaranteed that event checking
1429 * isn't in progress and won't happen until unblocked by
1430 * disk_unblock_events(). Events blocking is counted and the actual
1431 * unblocking happens after the matching number of unblocks are done.
1433 * Note that this intentionally does not block event checking from
1434 * disk_clear_events().
1439 void disk_block_events(struct gendisk
*disk
)
1441 struct disk_events
*ev
= disk
->ev
;
1442 unsigned long flags
;
1449 * Outer mutex ensures that the first blocker completes canceling
1450 * the event work before further blockers are allowed to finish.
1452 mutex_lock(&ev
->block_mutex
);
1454 spin_lock_irqsave(&ev
->lock
, flags
);
1455 cancel
= !ev
->block
++;
1456 spin_unlock_irqrestore(&ev
->lock
, flags
);
1459 cancel_delayed_work_sync(&disk
->ev
->dwork
);
1461 mutex_unlock(&ev
->block_mutex
);
1464 static void __disk_unblock_events(struct gendisk
*disk
, bool check_now
)
1466 struct disk_events
*ev
= disk
->ev
;
1468 unsigned long flags
;
1470 spin_lock_irqsave(&ev
->lock
, flags
);
1472 if (WARN_ON_ONCE(ev
->block
<= 0))
1478 intv
= disk_events_poll_jiffies(disk
);
1480 queue_delayed_work(system_freezable_power_efficient_wq
,
1483 queue_delayed_work(system_freezable_power_efficient_wq
,
1486 spin_unlock_irqrestore(&ev
->lock
, flags
);
1490 * disk_unblock_events - unblock disk event checking
1491 * @disk: disk to unblock events for
1493 * Undo disk_block_events(). When the block count reaches zero, it
1494 * starts events polling if configured.
1497 * Don't care. Safe to call from irq context.
1499 void disk_unblock_events(struct gendisk
*disk
)
1502 __disk_unblock_events(disk
, false);
1506 * disk_flush_events - schedule immediate event checking and flushing
1507 * @disk: disk to check and flush events for
1508 * @mask: events to flush
1510 * Schedule immediate event checking on @disk if not blocked. Events in
1511 * @mask are scheduled to be cleared from the driver. Note that this
1512 * doesn't clear the events from @disk->ev.
1515 * If @mask is non-zero must be called with bdev->bd_mutex held.
1517 void disk_flush_events(struct gendisk
*disk
, unsigned int mask
)
1519 struct disk_events
*ev
= disk
->ev
;
1524 spin_lock_irq(&ev
->lock
);
1525 ev
->clearing
|= mask
;
1527 mod_delayed_work(system_freezable_power_efficient_wq
,
1529 spin_unlock_irq(&ev
->lock
);
1533 * disk_clear_events - synchronously check, clear and return pending events
1534 * @disk: disk to fetch and clear events from
1535 * @mask: mask of events to be fetched and cleared
1537 * Disk events are synchronously checked and pending events in @mask
1538 * are cleared and returned. This ignores the block count.
1543 static unsigned int disk_clear_events(struct gendisk
*disk
, unsigned int mask
)
1545 struct disk_events
*ev
= disk
->ev
;
1546 unsigned int pending
;
1547 unsigned int clearing
= mask
;
1552 disk_block_events(disk
);
1555 * store the union of mask and ev->clearing on the stack so that the
1556 * race with disk_flush_events does not cause ambiguity (ev->clearing
1557 * can still be modified even if events are blocked).
1559 spin_lock_irq(&ev
->lock
);
1560 clearing
|= ev
->clearing
;
1562 spin_unlock_irq(&ev
->lock
);
1564 disk_check_events(ev
, &clearing
);
1566 * if ev->clearing is not 0, the disk_flush_events got called in the
1567 * middle of this function, so we want to run the workfn without delay.
1569 __disk_unblock_events(disk
, ev
->clearing
? true : false);
1571 /* then, fetch and clear pending events */
1572 spin_lock_irq(&ev
->lock
);
1573 pending
= ev
->pending
& mask
;
1574 ev
->pending
&= ~mask
;
1575 spin_unlock_irq(&ev
->lock
);
1576 WARN_ON_ONCE(clearing
& mask
);
1582 * bdev_check_media_change - check if a removable media has been changed
1583 * @bdev: block device to check
1585 * Check whether a removable media has been changed, and attempt to free all
1586 * dentries and inodes and invalidates all block device page cache entries in
1589 * Returns %true if the block device changed, or %false if not.
1591 bool bdev_check_media_change(struct block_device
*bdev
)
1593 unsigned int events
;
1595 events
= disk_clear_events(bdev
->bd_disk
, DISK_EVENT_MEDIA_CHANGE
|
1596 DISK_EVENT_EJECT_REQUEST
);
1597 if (!(events
& DISK_EVENT_MEDIA_CHANGE
))
1600 if (__invalidate_device(bdev
, true))
1601 pr_warn("VFS: busy inodes on changed media %s\n",
1602 bdev
->bd_disk
->disk_name
);
1603 set_bit(GD_NEED_PART_SCAN
, &bdev
->bd_disk
->state
);
1606 EXPORT_SYMBOL(bdev_check_media_change
);
1609 * Separate this part out so that a different pointer for clearing_ptr can be
1610 * passed in for disk_clear_events.
1612 static void disk_events_workfn(struct work_struct
*work
)
1614 struct delayed_work
*dwork
= to_delayed_work(work
);
1615 struct disk_events
*ev
= container_of(dwork
, struct disk_events
, dwork
);
1617 disk_check_events(ev
, &ev
->clearing
);
1620 static void disk_check_events(struct disk_events
*ev
,
1621 unsigned int *clearing_ptr
)
1623 struct gendisk
*disk
= ev
->disk
;
1624 char *envp
[ARRAY_SIZE(disk_uevents
) + 1] = { };
1625 unsigned int clearing
= *clearing_ptr
;
1626 unsigned int events
;
1628 int nr_events
= 0, i
;
1631 events
= disk
->fops
->check_events(disk
, clearing
);
1633 /* accumulate pending events and schedule next poll if necessary */
1634 spin_lock_irq(&ev
->lock
);
1636 events
&= ~ev
->pending
;
1637 ev
->pending
|= events
;
1638 *clearing_ptr
&= ~clearing
;
1640 intv
= disk_events_poll_jiffies(disk
);
1641 if (!ev
->block
&& intv
)
1642 queue_delayed_work(system_freezable_power_efficient_wq
,
1645 spin_unlock_irq(&ev
->lock
);
1648 * Tell userland about new events. Only the events listed in
1649 * @disk->events are reported, and only if DISK_EVENT_FLAG_UEVENT
1650 * is set. Otherwise, events are processed internally but never
1651 * get reported to userland.
1653 for (i
= 0; i
< ARRAY_SIZE(disk_uevents
); i
++)
1654 if ((events
& disk
->events
& (1 << i
)) &&
1655 (disk
->event_flags
& DISK_EVENT_FLAG_UEVENT
))
1656 envp
[nr_events
++] = disk_uevents
[i
];
1659 kobject_uevent_env(&disk_to_dev(disk
)->kobj
, KOBJ_CHANGE
, envp
);
1663 * A disk events enabled device has the following sysfs nodes under
1664 * its /sys/block/X/ directory.
1666 * events : list of all supported events
1667 * events_async : list of events which can be detected w/o polling
1668 * (always empty, only for backwards compatibility)
1669 * events_poll_msecs : polling interval, 0: disable, -1: system default
1671 static ssize_t
__disk_events_show(unsigned int events
, char *buf
)
1673 const char *delim
= "";
1677 for (i
= 0; i
< ARRAY_SIZE(disk_events_strs
); i
++)
1678 if (events
& (1 << i
)) {
1679 pos
+= sprintf(buf
+ pos
, "%s%s",
1680 delim
, disk_events_strs
[i
]);
1684 pos
+= sprintf(buf
+ pos
, "\n");
1688 static ssize_t
disk_events_show(struct device
*dev
,
1689 struct device_attribute
*attr
, char *buf
)
1691 struct gendisk
*disk
= dev_to_disk(dev
);
1693 if (!(disk
->event_flags
& DISK_EVENT_FLAG_UEVENT
))
1696 return __disk_events_show(disk
->events
, buf
);
1699 static ssize_t
disk_events_async_show(struct device
*dev
,
1700 struct device_attribute
*attr
, char *buf
)
1705 static ssize_t
disk_events_poll_msecs_show(struct device
*dev
,
1706 struct device_attribute
*attr
,
1709 struct gendisk
*disk
= dev_to_disk(dev
);
1712 return sprintf(buf
, "-1\n");
1714 return sprintf(buf
, "%ld\n", disk
->ev
->poll_msecs
);
1717 static ssize_t
disk_events_poll_msecs_store(struct device
*dev
,
1718 struct device_attribute
*attr
,
1719 const char *buf
, size_t count
)
1721 struct gendisk
*disk
= dev_to_disk(dev
);
1724 if (!count
|| !sscanf(buf
, "%ld", &intv
))
1727 if (intv
< 0 && intv
!= -1)
1733 disk_block_events(disk
);
1734 disk
->ev
->poll_msecs
= intv
;
1735 __disk_unblock_events(disk
, true);
1740 static const DEVICE_ATTR(events
, 0444, disk_events_show
, NULL
);
1741 static const DEVICE_ATTR(events_async
, 0444, disk_events_async_show
, NULL
);
1742 static const DEVICE_ATTR(events_poll_msecs
, 0644,
1743 disk_events_poll_msecs_show
,
1744 disk_events_poll_msecs_store
);
1746 static const struct attribute
*disk_events_attrs
[] = {
1747 &dev_attr_events
.attr
,
1748 &dev_attr_events_async
.attr
,
1749 &dev_attr_events_poll_msecs
.attr
,
1754 * The default polling interval can be specified by the kernel
1755 * parameter block.events_dfl_poll_msecs which defaults to 0
1756 * (disable). This can also be modified runtime by writing to
1757 * /sys/module/block/parameters/events_dfl_poll_msecs.
1759 static int disk_events_set_dfl_poll_msecs(const char *val
,
1760 const struct kernel_param
*kp
)
1762 struct disk_events
*ev
;
1765 ret
= param_set_ulong(val
, kp
);
1769 mutex_lock(&disk_events_mutex
);
1771 list_for_each_entry(ev
, &disk_events
, node
)
1772 disk_flush_events(ev
->disk
, 0);
1774 mutex_unlock(&disk_events_mutex
);
1779 static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops
= {
1780 .set
= disk_events_set_dfl_poll_msecs
,
1781 .get
= param_get_ulong
,
1784 #undef MODULE_PARAM_PREFIX
1785 #define MODULE_PARAM_PREFIX "block."
1787 module_param_cb(events_dfl_poll_msecs
, &disk_events_dfl_poll_msecs_param_ops
,
1788 &disk_events_dfl_poll_msecs
, 0644);
1791 * disk_{alloc|add|del|release}_events - initialize and destroy disk_events.
1793 static void disk_alloc_events(struct gendisk
*disk
)
1795 struct disk_events
*ev
;
1797 if (!disk
->fops
->check_events
|| !disk
->events
)
1800 ev
= kzalloc(sizeof(*ev
), GFP_KERNEL
);
1802 pr_warn("%s: failed to initialize events\n", disk
->disk_name
);
1806 INIT_LIST_HEAD(&ev
->node
);
1808 spin_lock_init(&ev
->lock
);
1809 mutex_init(&ev
->block_mutex
);
1811 ev
->poll_msecs
= -1;
1812 INIT_DELAYED_WORK(&ev
->dwork
, disk_events_workfn
);
1817 static void disk_add_events(struct gendisk
*disk
)
1819 /* FIXME: error handling */
1820 if (sysfs_create_files(&disk_to_dev(disk
)->kobj
, disk_events_attrs
) < 0)
1821 pr_warn("%s: failed to create sysfs files for events\n",
1827 mutex_lock(&disk_events_mutex
);
1828 list_add_tail(&disk
->ev
->node
, &disk_events
);
1829 mutex_unlock(&disk_events_mutex
);
1832 * Block count is initialized to 1 and the following initial
1833 * unblock kicks it into action.
1835 __disk_unblock_events(disk
, true);
1838 static void disk_del_events(struct gendisk
*disk
)
1841 disk_block_events(disk
);
1843 mutex_lock(&disk_events_mutex
);
1844 list_del_init(&disk
->ev
->node
);
1845 mutex_unlock(&disk_events_mutex
);
1848 sysfs_remove_files(&disk_to_dev(disk
)->kobj
, disk_events_attrs
);
1851 static void disk_release_events(struct gendisk
*disk
)
1853 /* the block count should be 1 from disk_del_events() */
1854 WARN_ON_ONCE(disk
->ev
&& disk
->ev
->block
!= 1);