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 DECLARE_RWSEM(bdev_lookup_sem
);
34 /* for extended dynamic devt allocation, currently only one major is used */
35 #define NR_EXT_DEVT (1 << MINORBITS)
36 static DEFINE_IDA(ext_devt_ida
);
38 static void disk_check_events(struct disk_events
*ev
,
39 unsigned int *clearing_ptr
);
40 static void disk_alloc_events(struct gendisk
*disk
);
41 static void disk_add_events(struct gendisk
*disk
);
42 static void disk_del_events(struct gendisk
*disk
);
43 static void disk_release_events(struct gendisk
*disk
);
45 void set_capacity(struct gendisk
*disk
, sector_t sectors
)
47 struct block_device
*bdev
= disk
->part0
;
49 spin_lock(&bdev
->bd_size_lock
);
50 i_size_write(bdev
->bd_inode
, (loff_t
)sectors
<< SECTOR_SHIFT
);
51 spin_unlock(&bdev
->bd_size_lock
);
53 EXPORT_SYMBOL(set_capacity
);
56 * Set disk capacity and notify if the size is not currently zero and will not
57 * be set to zero. Returns true if a uevent was sent, otherwise false.
59 bool set_capacity_and_notify(struct gendisk
*disk
, sector_t size
)
61 sector_t capacity
= get_capacity(disk
);
62 char *envp
[] = { "RESIZE=1", NULL
};
64 set_capacity(disk
, size
);
67 * Only print a message and send a uevent if the gendisk is user visible
68 * and alive. This avoids spamming the log and udev when setting the
69 * initial capacity during probing.
71 if (size
== capacity
||
72 (disk
->flags
& (GENHD_FL_UP
| GENHD_FL_HIDDEN
)) != GENHD_FL_UP
)
75 pr_info("%s: detected capacity change from %lld to %lld\n",
76 disk
->disk_name
, capacity
, size
);
79 * Historically we did not send a uevent for changes to/from an empty
82 if (!capacity
|| !size
)
84 kobject_uevent_env(&disk_to_dev(disk
)->kobj
, KOBJ_CHANGE
, envp
);
87 EXPORT_SYMBOL_GPL(set_capacity_and_notify
);
90 * Format the device name of the indicated disk into the supplied buffer and
91 * return a pointer to that same buffer for convenience.
93 char *disk_name(struct gendisk
*hd
, int partno
, char *buf
)
96 snprintf(buf
, BDEVNAME_SIZE
, "%s", hd
->disk_name
);
97 else if (isdigit(hd
->disk_name
[strlen(hd
->disk_name
)-1]))
98 snprintf(buf
, BDEVNAME_SIZE
, "%sp%d", hd
->disk_name
, partno
);
100 snprintf(buf
, BDEVNAME_SIZE
, "%s%d", hd
->disk_name
, partno
);
105 const char *bdevname(struct block_device
*bdev
, char *buf
)
107 return disk_name(bdev
->bd_disk
, bdev
->bd_partno
, buf
);
109 EXPORT_SYMBOL(bdevname
);
111 static void part_stat_read_all(struct block_device
*part
,
112 struct disk_stats
*stat
)
116 memset(stat
, 0, sizeof(struct disk_stats
));
117 for_each_possible_cpu(cpu
) {
118 struct disk_stats
*ptr
= per_cpu_ptr(part
->bd_stats
, cpu
);
121 for (group
= 0; group
< NR_STAT_GROUPS
; group
++) {
122 stat
->nsecs
[group
] += ptr
->nsecs
[group
];
123 stat
->sectors
[group
] += ptr
->sectors
[group
];
124 stat
->ios
[group
] += ptr
->ios
[group
];
125 stat
->merges
[group
] += ptr
->merges
[group
];
128 stat
->io_ticks
+= ptr
->io_ticks
;
132 static unsigned int part_in_flight(struct block_device
*part
)
134 unsigned int inflight
= 0;
137 for_each_possible_cpu(cpu
) {
138 inflight
+= part_stat_local_read_cpu(part
, in_flight
[0], cpu
) +
139 part_stat_local_read_cpu(part
, in_flight
[1], cpu
);
141 if ((int)inflight
< 0)
147 static void part_in_flight_rw(struct block_device
*part
,
148 unsigned int inflight
[2])
154 for_each_possible_cpu(cpu
) {
155 inflight
[0] += part_stat_local_read_cpu(part
, in_flight
[0], cpu
);
156 inflight
[1] += part_stat_local_read_cpu(part
, in_flight
[1], cpu
);
158 if ((int)inflight
[0] < 0)
160 if ((int)inflight
[1] < 0)
165 * disk_part_iter_init - initialize partition iterator
166 * @piter: iterator to initialize
167 * @disk: disk to iterate over
168 * @flags: DISK_PITER_* flags
170 * Initialize @piter so that it iterates over partitions of @disk.
175 void disk_part_iter_init(struct disk_part_iter
*piter
, struct gendisk
*disk
,
180 if (flags
& (DISK_PITER_INCL_PART0
| DISK_PITER_INCL_EMPTY_PART0
))
184 piter
->flags
= flags
;
188 * disk_part_iter_next - proceed iterator to the next partition and return it
189 * @piter: iterator of interest
191 * Proceed @piter to the next partition and return it.
196 struct block_device
*disk_part_iter_next(struct disk_part_iter
*piter
)
198 struct block_device
*part
;
201 /* put the last partition */
202 disk_part_iter_exit(piter
);
205 xa_for_each_start(&piter
->disk
->part_tbl
, idx
, part
, piter
->idx
) {
206 if (!bdev_nr_sectors(part
) &&
207 !(piter
->flags
& DISK_PITER_INCL_EMPTY
) &&
208 !(piter
->flags
& DISK_PITER_INCL_EMPTY_PART0
&&
212 piter
->part
= bdgrab(part
);
215 piter
->idx
= idx
+ 1;
224 * disk_part_iter_exit - finish up partition iteration
225 * @piter: iter of interest
227 * Called when iteration is over. Cleans up @piter.
232 void disk_part_iter_exit(struct disk_part_iter
*piter
)
240 * Can be deleted altogether. Later.
243 #define BLKDEV_MAJOR_HASH_SIZE 255
244 static struct blk_major_name
{
245 struct blk_major_name
*next
;
248 void (*probe
)(dev_t devt
);
249 } *major_names
[BLKDEV_MAJOR_HASH_SIZE
];
250 static DEFINE_MUTEX(major_names_lock
);
252 /* index in the above - for now: assume no multimajor ranges */
253 static inline int major_to_index(unsigned major
)
255 return major
% BLKDEV_MAJOR_HASH_SIZE
;
258 #ifdef CONFIG_PROC_FS
259 void blkdev_show(struct seq_file
*seqf
, off_t offset
)
261 struct blk_major_name
*dp
;
263 mutex_lock(&major_names_lock
);
264 for (dp
= major_names
[major_to_index(offset
)]; dp
; dp
= dp
->next
)
265 if (dp
->major
== offset
)
266 seq_printf(seqf
, "%3d %s\n", dp
->major
, dp
->name
);
267 mutex_unlock(&major_names_lock
);
269 #endif /* CONFIG_PROC_FS */
272 * __register_blkdev - register a new block device
274 * @major: the requested major device number [1..BLKDEV_MAJOR_MAX-1]. If
275 * @major = 0, try to allocate any unused major number.
276 * @name: the name of the new block device as a zero terminated string
277 * @probe: allback that is called on access to any minor number of @major
279 * The @name must be unique within the system.
281 * The return value depends on the @major input parameter:
283 * - if a major device number was requested in range [1..BLKDEV_MAJOR_MAX-1]
284 * then the function returns zero on success, or a negative error code
285 * - if any unused major number was requested with @major = 0 parameter
286 * then the return value is the allocated major number in range
287 * [1..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise
289 * See Documentation/admin-guide/devices.txt for the list of allocated
292 * Use register_blkdev instead for any new code.
294 int __register_blkdev(unsigned int major
, const char *name
,
295 void (*probe
)(dev_t devt
))
297 struct blk_major_name
**n
, *p
;
300 mutex_lock(&major_names_lock
);
304 for (index
= ARRAY_SIZE(major_names
)-1; index
> 0; index
--) {
305 if (major_names
[index
] == NULL
)
310 printk("%s: failed to get major for %s\n",
319 if (major
>= BLKDEV_MAJOR_MAX
) {
320 pr_err("%s: major requested (%u) is greater than the maximum (%u) for %s\n",
321 __func__
, major
, BLKDEV_MAJOR_MAX
-1, name
);
327 p
= kmalloc(sizeof(struct blk_major_name
), GFP_KERNEL
);
335 strlcpy(p
->name
, name
, sizeof(p
->name
));
337 index
= major_to_index(major
);
339 for (n
= &major_names
[index
]; *n
; n
= &(*n
)->next
) {
340 if ((*n
)->major
== major
)
349 printk("register_blkdev: cannot get major %u for %s\n",
354 mutex_unlock(&major_names_lock
);
357 EXPORT_SYMBOL(__register_blkdev
);
359 void unregister_blkdev(unsigned int major
, const char *name
)
361 struct blk_major_name
**n
;
362 struct blk_major_name
*p
= NULL
;
363 int index
= major_to_index(major
);
365 mutex_lock(&major_names_lock
);
366 for (n
= &major_names
[index
]; *n
; n
= &(*n
)->next
)
367 if ((*n
)->major
== major
)
369 if (!*n
|| strcmp((*n
)->name
, name
)) {
375 mutex_unlock(&major_names_lock
);
379 EXPORT_SYMBOL(unregister_blkdev
);
382 * blk_mangle_minor - scatter minor numbers apart
383 * @minor: minor number to mangle
385 * Scatter consecutively allocated @minor number apart if MANGLE_DEVT
386 * is enabled. Mangling twice gives the original value.
394 static int blk_mangle_minor(int minor
)
396 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
399 for (i
= 0; i
< MINORBITS
/ 2; i
++) {
400 int low
= minor
& (1 << i
);
401 int high
= minor
& (1 << (MINORBITS
- 1 - i
));
402 int distance
= MINORBITS
- 1 - 2 * i
;
404 minor
^= low
| high
; /* clear both bits */
405 low
<<= distance
; /* swap the positions */
407 minor
|= low
| high
; /* and set */
414 * blk_alloc_devt - allocate a dev_t for a block device
415 * @bdev: block device to allocate dev_t for
416 * @devt: out parameter for resulting dev_t
418 * Allocate a dev_t for block device.
421 * 0 on success, allocated dev_t is returned in *@devt. -errno on
427 int blk_alloc_devt(struct block_device
*bdev
, dev_t
*devt
)
429 struct gendisk
*disk
= bdev
->bd_disk
;
432 /* in consecutive minor range? */
433 if (bdev
->bd_partno
< disk
->minors
) {
434 *devt
= MKDEV(disk
->major
, disk
->first_minor
+ bdev
->bd_partno
);
438 idx
= ida_alloc_range(&ext_devt_ida
, 0, NR_EXT_DEVT
, GFP_KERNEL
);
440 return idx
== -ENOSPC
? -EBUSY
: idx
;
442 *devt
= MKDEV(BLOCK_EXT_MAJOR
, blk_mangle_minor(idx
));
447 * blk_free_devt - free a dev_t
448 * @devt: dev_t to free
450 * Free @devt which was allocated using blk_alloc_devt().
455 void blk_free_devt(dev_t devt
)
457 if (MAJOR(devt
) == BLOCK_EXT_MAJOR
)
458 ida_free(&ext_devt_ida
, blk_mangle_minor(MINOR(devt
)));
461 static char *bdevt_str(dev_t devt
, char *buf
)
463 if (MAJOR(devt
) <= 0xff && MINOR(devt
) <= 0xff) {
464 char tbuf
[BDEVT_SIZE
];
465 snprintf(tbuf
, BDEVT_SIZE
, "%02x%02x", MAJOR(devt
), MINOR(devt
));
466 snprintf(buf
, BDEVT_SIZE
, "%-9s", tbuf
);
468 snprintf(buf
, BDEVT_SIZE
, "%03x:%05x", MAJOR(devt
), MINOR(devt
));
473 void disk_uevent(struct gendisk
*disk
, enum kobject_action action
)
475 struct disk_part_iter piter
;
476 struct block_device
*part
;
478 disk_part_iter_init(&piter
, disk
, DISK_PITER_INCL_EMPTY_PART0
);
479 while ((part
= disk_part_iter_next(&piter
)))
480 kobject_uevent(bdev_kobj(part
), action
);
481 disk_part_iter_exit(&piter
);
483 EXPORT_SYMBOL_GPL(disk_uevent
);
485 static void disk_scan_partitions(struct gendisk
*disk
)
487 struct block_device
*bdev
;
489 if (!get_capacity(disk
) || !disk_part_scan_enabled(disk
))
492 set_bit(GD_NEED_PART_SCAN
, &disk
->state
);
493 bdev
= blkdev_get_by_dev(disk_devt(disk
), FMODE_READ
, NULL
);
495 blkdev_put(bdev
, FMODE_READ
);
498 static void register_disk(struct device
*parent
, struct gendisk
*disk
,
499 const struct attribute_group
**groups
)
501 struct device
*ddev
= disk_to_dev(disk
);
504 ddev
->parent
= parent
;
506 dev_set_name(ddev
, "%s", disk
->disk_name
);
508 /* delay uevents, until we scanned partition table */
509 dev_set_uevent_suppress(ddev
, 1);
512 WARN_ON(ddev
->groups
);
513 ddev
->groups
= groups
;
515 if (device_add(ddev
))
517 if (!sysfs_deprecated
) {
518 err
= sysfs_create_link(block_depr
, &ddev
->kobj
,
519 kobject_name(&ddev
->kobj
));
527 * avoid probable deadlock caused by allocating memory with
528 * GFP_KERNEL in runtime_resume callback of its all ancestor
531 pm_runtime_set_memalloc_noio(ddev
, true);
533 disk
->part0
->bd_holder_dir
=
534 kobject_create_and_add("holders", &ddev
->kobj
);
535 disk
->slave_dir
= kobject_create_and_add("slaves", &ddev
->kobj
);
537 if (disk
->flags
& GENHD_FL_HIDDEN
)
540 disk_scan_partitions(disk
);
542 /* announce the disk and partitions after all partitions are created */
543 dev_set_uevent_suppress(ddev
, 0);
544 disk_uevent(disk
, KOBJ_ADD
);
546 if (disk
->queue
->backing_dev_info
->dev
) {
547 err
= sysfs_create_link(&ddev
->kobj
,
548 &disk
->queue
->backing_dev_info
->dev
->kobj
,
555 * __device_add_disk - add disk information to kernel list
556 * @parent: parent device for the disk
557 * @disk: per-device partitioning information
558 * @groups: Additional per-device sysfs groups
559 * @register_queue: register the queue if set to true
561 * This function registers the partitioning information in @disk
564 * FIXME: error handling
566 static void __device_add_disk(struct device
*parent
, struct gendisk
*disk
,
567 const struct attribute_group
**groups
,
574 * The disk queue should now be all set with enough information about
575 * the device for the elevator code to pick an adequate default
576 * elevator if one is needed, that is, for devices requesting queue
580 elevator_init_mq(disk
->queue
);
582 /* minors == 0 indicates to use ext devt from part0 and should
583 * be accompanied with EXT_DEVT flag. Make sure all
584 * parameters make sense.
586 WARN_ON(disk
->minors
&& !(disk
->major
|| disk
->first_minor
));
587 WARN_ON(!disk
->minors
&&
588 !(disk
->flags
& (GENHD_FL_EXT_DEVT
| GENHD_FL_HIDDEN
)));
590 disk
->flags
|= GENHD_FL_UP
;
592 retval
= blk_alloc_devt(disk
->part0
, &devt
);
597 disk
->major
= MAJOR(devt
);
598 disk
->first_minor
= MINOR(devt
);
600 disk_alloc_events(disk
);
602 if (disk
->flags
& GENHD_FL_HIDDEN
) {
604 * Don't let hidden disks show up in /proc/partitions,
605 * and don't bother scanning for partitions either.
607 disk
->flags
|= GENHD_FL_SUPPRESS_PARTITION_INFO
;
608 disk
->flags
|= GENHD_FL_NO_PART_SCAN
;
610 struct backing_dev_info
*bdi
= disk
->queue
->backing_dev_info
;
611 struct device
*dev
= disk_to_dev(disk
);
614 /* Register BDI before referencing it from bdev */
616 ret
= bdi_register(bdi
, "%u:%u", MAJOR(devt
), MINOR(devt
));
618 bdi_set_owner(bdi
, dev
);
619 bdev_add(disk
->part0
, devt
);
621 register_disk(parent
, disk
, groups
);
623 blk_register_queue(disk
);
626 * Take an extra ref on queue which will be put on disk_release()
627 * so that it sticks around as long as @disk is there.
629 WARN_ON_ONCE(!blk_get_queue(disk
->queue
));
631 disk_add_events(disk
);
632 blk_integrity_add(disk
);
635 void device_add_disk(struct device
*parent
, struct gendisk
*disk
,
636 const struct attribute_group
**groups
)
639 __device_add_disk(parent
, disk
, groups
, true);
641 EXPORT_SYMBOL(device_add_disk
);
643 void device_add_disk_no_queue_reg(struct device
*parent
, struct gendisk
*disk
)
645 __device_add_disk(parent
, disk
, NULL
, false);
647 EXPORT_SYMBOL(device_add_disk_no_queue_reg
);
649 static void invalidate_partition(struct block_device
*bdev
)
652 __invalidate_device(bdev
, true);
655 * Unhash the bdev inode for this device so that it can't be looked
656 * up any more even if openers still hold references to it.
658 remove_inode_hash(bdev
->bd_inode
);
662 * del_gendisk - remove the gendisk
663 * @disk: the struct gendisk to remove
665 * Removes the gendisk and all its associated resources. This deletes the
666 * partitions associated with the gendisk, and unregisters the associated
669 * This is the counter to the respective __device_add_disk() call.
671 * The final removal of the struct gendisk happens when its refcount reaches 0
672 * with put_disk(), which should be called after del_gendisk(), if
673 * __device_add_disk() was used.
675 * Drivers exist which depend on the release of the gendisk to be synchronous,
676 * it should not be deferred.
680 void del_gendisk(struct gendisk
*disk
)
682 struct disk_part_iter piter
;
683 struct block_device
*part
;
687 if (WARN_ON_ONCE(!disk
->queue
))
690 blk_integrity_del(disk
);
691 disk_del_events(disk
);
694 * Block lookups of the disk until all bdevs are unhashed and the
695 * disk is marked as dead (GENHD_FL_UP cleared).
697 down_write(&bdev_lookup_sem
);
699 /* invalidate stuff */
700 disk_part_iter_init(&piter
, disk
, DISK_PITER_INCL_EMPTY
);
701 while ((part
= disk_part_iter_next(&piter
))) {
702 invalidate_partition(part
);
703 delete_partition(part
);
705 disk_part_iter_exit(&piter
);
707 invalidate_partition(disk
->part0
);
708 set_capacity(disk
, 0);
709 disk
->flags
&= ~GENHD_FL_UP
;
710 up_write(&bdev_lookup_sem
);
712 if (!(disk
->flags
& GENHD_FL_HIDDEN
)) {
713 sysfs_remove_link(&disk_to_dev(disk
)->kobj
, "bdi");
716 * Unregister bdi before releasing device numbers (as they can
717 * get reused and we'd get clashes in sysfs).
719 bdi_unregister(disk
->queue
->backing_dev_info
);
722 blk_unregister_queue(disk
);
724 kobject_put(disk
->part0
->bd_holder_dir
);
725 kobject_put(disk
->slave_dir
);
727 part_stat_set_all(disk
->part0
, 0);
728 disk
->part0
->bd_stamp
= 0;
729 if (!sysfs_deprecated
)
730 sysfs_remove_link(block_depr
, dev_name(disk_to_dev(disk
)));
731 pm_runtime_set_memalloc_noio(disk_to_dev(disk
), false);
732 device_del(disk_to_dev(disk
));
734 EXPORT_SYMBOL(del_gendisk
);
736 /* sysfs access to bad-blocks list. */
737 static ssize_t
disk_badblocks_show(struct device
*dev
,
738 struct device_attribute
*attr
,
741 struct gendisk
*disk
= dev_to_disk(dev
);
744 return sprintf(page
, "\n");
746 return badblocks_show(disk
->bb
, page
, 0);
749 static ssize_t
disk_badblocks_store(struct device
*dev
,
750 struct device_attribute
*attr
,
751 const char *page
, size_t len
)
753 struct gendisk
*disk
= dev_to_disk(dev
);
758 return badblocks_store(disk
->bb
, page
, len
, 0);
761 void blk_request_module(dev_t devt
)
763 unsigned int major
= MAJOR(devt
);
764 struct blk_major_name
**n
;
766 mutex_lock(&major_names_lock
);
767 for (n
= &major_names
[major_to_index(major
)]; *n
; n
= &(*n
)->next
) {
768 if ((*n
)->major
== major
&& (*n
)->probe
) {
770 mutex_unlock(&major_names_lock
);
774 mutex_unlock(&major_names_lock
);
776 if (request_module("block-major-%d-%d", MAJOR(devt
), MINOR(devt
)) > 0)
777 /* Make old-style 2.4 aliases work */
778 request_module("block-major-%d", MAJOR(devt
));
782 * bdget_disk - do bdget() by gendisk and partition number
783 * @disk: gendisk of interest
784 * @partno: partition number
786 * Find partition @partno from @disk, do bdget() on it.
792 * Resulting block_device on success, NULL on failure.
794 struct block_device
*bdget_disk(struct gendisk
*disk
, int partno
)
796 struct block_device
*bdev
= NULL
;
799 bdev
= xa_load(&disk
->part_tbl
, partno
);
800 if (bdev
&& !bdgrab(bdev
))
808 * print a full list of all partitions - intended for places where the root
809 * filesystem can't be mounted and thus to give the victim some idea of what
812 void __init
printk_all_partitions(void)
814 struct class_dev_iter iter
;
817 class_dev_iter_init(&iter
, &block_class
, NULL
, &disk_type
);
818 while ((dev
= class_dev_iter_next(&iter
))) {
819 struct gendisk
*disk
= dev_to_disk(dev
);
820 struct disk_part_iter piter
;
821 struct block_device
*part
;
822 char name_buf
[BDEVNAME_SIZE
];
823 char devt_buf
[BDEVT_SIZE
];
826 * Don't show empty devices or things that have been
829 if (get_capacity(disk
) == 0 ||
830 (disk
->flags
& GENHD_FL_SUPPRESS_PARTITION_INFO
))
834 * Note, unlike /proc/partitions, I am showing the
835 * numbers in hex - the same format as the root=
838 disk_part_iter_init(&piter
, disk
, DISK_PITER_INCL_PART0
);
839 while ((part
= disk_part_iter_next(&piter
))) {
840 bool is_part0
= part
== disk
->part0
;
842 printk("%s%s %10llu %s %s", is_part0
? "" : " ",
843 bdevt_str(part
->bd_dev
, devt_buf
),
844 bdev_nr_sectors(part
) >> 1,
845 disk_name(disk
, part
->bd_partno
, name_buf
),
847 part
->bd_meta_info
->uuid
: "");
849 if (dev
->parent
&& dev
->parent
->driver
)
850 printk(" driver: %s\n",
851 dev
->parent
->driver
->name
);
853 printk(" (driver?)\n");
857 disk_part_iter_exit(&piter
);
859 class_dev_iter_exit(&iter
);
862 #ifdef CONFIG_PROC_FS
864 static void *disk_seqf_start(struct seq_file
*seqf
, loff_t
*pos
)
867 struct class_dev_iter
*iter
;
870 iter
= kmalloc(sizeof(*iter
), GFP_KERNEL
);
872 return ERR_PTR(-ENOMEM
);
874 seqf
->private = iter
;
875 class_dev_iter_init(iter
, &block_class
, NULL
, &disk_type
);
877 dev
= class_dev_iter_next(iter
);
882 return dev_to_disk(dev
);
885 static void *disk_seqf_next(struct seq_file
*seqf
, void *v
, loff_t
*pos
)
890 dev
= class_dev_iter_next(seqf
->private);
892 return dev_to_disk(dev
);
897 static void disk_seqf_stop(struct seq_file
*seqf
, void *v
)
899 struct class_dev_iter
*iter
= seqf
->private;
901 /* stop is called even after start failed :-( */
903 class_dev_iter_exit(iter
);
905 seqf
->private = NULL
;
909 static void *show_partition_start(struct seq_file
*seqf
, loff_t
*pos
)
913 p
= disk_seqf_start(seqf
, pos
);
914 if (!IS_ERR_OR_NULL(p
) && !*pos
)
915 seq_puts(seqf
, "major minor #blocks name\n\n");
919 static int show_partition(struct seq_file
*seqf
, void *v
)
921 struct gendisk
*sgp
= v
;
922 struct disk_part_iter piter
;
923 struct block_device
*part
;
924 char buf
[BDEVNAME_SIZE
];
926 /* Don't show non-partitionable removeable devices or empty devices */
927 if (!get_capacity(sgp
) || (!disk_max_parts(sgp
) &&
928 (sgp
->flags
& GENHD_FL_REMOVABLE
)))
930 if (sgp
->flags
& GENHD_FL_SUPPRESS_PARTITION_INFO
)
933 /* show the full disk and all non-0 size partitions of it */
934 disk_part_iter_init(&piter
, sgp
, DISK_PITER_INCL_PART0
);
935 while ((part
= disk_part_iter_next(&piter
)))
936 seq_printf(seqf
, "%4d %7d %10llu %s\n",
937 MAJOR(part
->bd_dev
), MINOR(part
->bd_dev
),
938 bdev_nr_sectors(part
) >> 1,
939 disk_name(sgp
, part
->bd_partno
, buf
));
940 disk_part_iter_exit(&piter
);
945 static const struct seq_operations partitions_op
= {
946 .start
= show_partition_start
,
947 .next
= disk_seqf_next
,
948 .stop
= disk_seqf_stop
,
949 .show
= show_partition
953 static int __init
genhd_device_init(void)
957 block_class
.dev_kobj
= sysfs_dev_block_kobj
;
958 error
= class_register(&block_class
);
963 register_blkdev(BLOCK_EXT_MAJOR
, "blkext");
965 /* create top-level block dir */
966 if (!sysfs_deprecated
)
967 block_depr
= kobject_create_and_add("block", NULL
);
971 subsys_initcall(genhd_device_init
);
973 static ssize_t
disk_range_show(struct device
*dev
,
974 struct device_attribute
*attr
, char *buf
)
976 struct gendisk
*disk
= dev_to_disk(dev
);
978 return sprintf(buf
, "%d\n", disk
->minors
);
981 static ssize_t
disk_ext_range_show(struct device
*dev
,
982 struct device_attribute
*attr
, char *buf
)
984 struct gendisk
*disk
= dev_to_disk(dev
);
986 return sprintf(buf
, "%d\n", disk_max_parts(disk
));
989 static ssize_t
disk_removable_show(struct device
*dev
,
990 struct device_attribute
*attr
, char *buf
)
992 struct gendisk
*disk
= dev_to_disk(dev
);
994 return sprintf(buf
, "%d\n",
995 (disk
->flags
& GENHD_FL_REMOVABLE
? 1 : 0));
998 static ssize_t
disk_hidden_show(struct device
*dev
,
999 struct device_attribute
*attr
, char *buf
)
1001 struct gendisk
*disk
= dev_to_disk(dev
);
1003 return sprintf(buf
, "%d\n",
1004 (disk
->flags
& GENHD_FL_HIDDEN
? 1 : 0));
1007 static ssize_t
disk_ro_show(struct device
*dev
,
1008 struct device_attribute
*attr
, char *buf
)
1010 struct gendisk
*disk
= dev_to_disk(dev
);
1012 return sprintf(buf
, "%d\n", get_disk_ro(disk
) ? 1 : 0);
1015 ssize_t
part_size_show(struct device
*dev
,
1016 struct device_attribute
*attr
, char *buf
)
1018 return sprintf(buf
, "%llu\n", bdev_nr_sectors(dev_to_bdev(dev
)));
1021 ssize_t
part_stat_show(struct device
*dev
,
1022 struct device_attribute
*attr
, char *buf
)
1024 struct block_device
*bdev
= dev_to_bdev(dev
);
1025 struct request_queue
*q
= bdev
->bd_disk
->queue
;
1026 struct disk_stats stat
;
1027 unsigned int inflight
;
1029 part_stat_read_all(bdev
, &stat
);
1031 inflight
= blk_mq_in_flight(q
, bdev
);
1033 inflight
= part_in_flight(bdev
);
1036 "%8lu %8lu %8llu %8u "
1037 "%8lu %8lu %8llu %8u "
1039 "%8lu %8lu %8llu %8u "
1042 stat
.ios
[STAT_READ
],
1043 stat
.merges
[STAT_READ
],
1044 (unsigned long long)stat
.sectors
[STAT_READ
],
1045 (unsigned int)div_u64(stat
.nsecs
[STAT_READ
], NSEC_PER_MSEC
),
1046 stat
.ios
[STAT_WRITE
],
1047 stat
.merges
[STAT_WRITE
],
1048 (unsigned long long)stat
.sectors
[STAT_WRITE
],
1049 (unsigned int)div_u64(stat
.nsecs
[STAT_WRITE
], NSEC_PER_MSEC
),
1051 jiffies_to_msecs(stat
.io_ticks
),
1052 (unsigned int)div_u64(stat
.nsecs
[STAT_READ
] +
1053 stat
.nsecs
[STAT_WRITE
] +
1054 stat
.nsecs
[STAT_DISCARD
] +
1055 stat
.nsecs
[STAT_FLUSH
],
1057 stat
.ios
[STAT_DISCARD
],
1058 stat
.merges
[STAT_DISCARD
],
1059 (unsigned long long)stat
.sectors
[STAT_DISCARD
],
1060 (unsigned int)div_u64(stat
.nsecs
[STAT_DISCARD
], NSEC_PER_MSEC
),
1061 stat
.ios
[STAT_FLUSH
],
1062 (unsigned int)div_u64(stat
.nsecs
[STAT_FLUSH
], NSEC_PER_MSEC
));
1065 ssize_t
part_inflight_show(struct device
*dev
, struct device_attribute
*attr
,
1068 struct block_device
*bdev
= dev_to_bdev(dev
);
1069 struct request_queue
*q
= bdev
->bd_disk
->queue
;
1070 unsigned int inflight
[2];
1073 blk_mq_in_flight_rw(q
, bdev
, inflight
);
1075 part_in_flight_rw(bdev
, inflight
);
1077 return sprintf(buf
, "%8u %8u\n", inflight
[0], inflight
[1]);
1080 static ssize_t
disk_capability_show(struct device
*dev
,
1081 struct device_attribute
*attr
, char *buf
)
1083 struct gendisk
*disk
= dev_to_disk(dev
);
1085 return sprintf(buf
, "%x\n", disk
->flags
);
1088 static ssize_t
disk_alignment_offset_show(struct device
*dev
,
1089 struct device_attribute
*attr
,
1092 struct gendisk
*disk
= dev_to_disk(dev
);
1094 return sprintf(buf
, "%d\n", queue_alignment_offset(disk
->queue
));
1097 static ssize_t
disk_discard_alignment_show(struct device
*dev
,
1098 struct device_attribute
*attr
,
1101 struct gendisk
*disk
= dev_to_disk(dev
);
1103 return sprintf(buf
, "%d\n", queue_discard_alignment(disk
->queue
));
1106 static DEVICE_ATTR(range
, 0444, disk_range_show
, NULL
);
1107 static DEVICE_ATTR(ext_range
, 0444, disk_ext_range_show
, NULL
);
1108 static DEVICE_ATTR(removable
, 0444, disk_removable_show
, NULL
);
1109 static DEVICE_ATTR(hidden
, 0444, disk_hidden_show
, NULL
);
1110 static DEVICE_ATTR(ro
, 0444, disk_ro_show
, NULL
);
1111 static DEVICE_ATTR(size
, 0444, part_size_show
, NULL
);
1112 static DEVICE_ATTR(alignment_offset
, 0444, disk_alignment_offset_show
, NULL
);
1113 static DEVICE_ATTR(discard_alignment
, 0444, disk_discard_alignment_show
, NULL
);
1114 static DEVICE_ATTR(capability
, 0444, disk_capability_show
, NULL
);
1115 static DEVICE_ATTR(stat
, 0444, part_stat_show
, NULL
);
1116 static DEVICE_ATTR(inflight
, 0444, part_inflight_show
, NULL
);
1117 static DEVICE_ATTR(badblocks
, 0644, disk_badblocks_show
, disk_badblocks_store
);
1119 #ifdef CONFIG_FAIL_MAKE_REQUEST
1120 ssize_t
part_fail_show(struct device
*dev
,
1121 struct device_attribute
*attr
, char *buf
)
1123 return sprintf(buf
, "%d\n", dev_to_bdev(dev
)->bd_make_it_fail
);
1126 ssize_t
part_fail_store(struct device
*dev
,
1127 struct device_attribute
*attr
,
1128 const char *buf
, size_t count
)
1132 if (count
> 0 && sscanf(buf
, "%d", &i
) > 0)
1133 dev_to_bdev(dev
)->bd_make_it_fail
= i
;
1138 static struct device_attribute dev_attr_fail
=
1139 __ATTR(make
-it
-fail
, 0644, part_fail_show
, part_fail_store
);
1140 #endif /* CONFIG_FAIL_MAKE_REQUEST */
1142 #ifdef CONFIG_FAIL_IO_TIMEOUT
1143 static struct device_attribute dev_attr_fail_timeout
=
1144 __ATTR(io
-timeout
-fail
, 0644, part_timeout_show
, part_timeout_store
);
1147 static struct attribute
*disk_attrs
[] = {
1148 &dev_attr_range
.attr
,
1149 &dev_attr_ext_range
.attr
,
1150 &dev_attr_removable
.attr
,
1151 &dev_attr_hidden
.attr
,
1153 &dev_attr_size
.attr
,
1154 &dev_attr_alignment_offset
.attr
,
1155 &dev_attr_discard_alignment
.attr
,
1156 &dev_attr_capability
.attr
,
1157 &dev_attr_stat
.attr
,
1158 &dev_attr_inflight
.attr
,
1159 &dev_attr_badblocks
.attr
,
1160 #ifdef CONFIG_FAIL_MAKE_REQUEST
1161 &dev_attr_fail
.attr
,
1163 #ifdef CONFIG_FAIL_IO_TIMEOUT
1164 &dev_attr_fail_timeout
.attr
,
1169 static umode_t
disk_visible(struct kobject
*kobj
, struct attribute
*a
, int n
)
1171 struct device
*dev
= container_of(kobj
, typeof(*dev
), kobj
);
1172 struct gendisk
*disk
= dev_to_disk(dev
);
1174 if (a
== &dev_attr_badblocks
.attr
&& !disk
->bb
)
1179 static struct attribute_group disk_attr_group
= {
1180 .attrs
= disk_attrs
,
1181 .is_visible
= disk_visible
,
1184 static const struct attribute_group
*disk_attr_groups
[] = {
1190 * disk_release - releases all allocated resources of the gendisk
1191 * @dev: the device representing this disk
1193 * This function releases all allocated resources of the gendisk.
1195 * Drivers which used __device_add_disk() have a gendisk with a request_queue
1196 * assigned. Since the request_queue sits on top of the gendisk for these
1197 * drivers we also call blk_put_queue() for them, and we expect the
1198 * request_queue refcount to reach 0 at this point, and so the request_queue
1199 * will also be freed prior to the disk.
1201 * Context: can sleep
1203 static void disk_release(struct device
*dev
)
1205 struct gendisk
*disk
= dev_to_disk(dev
);
1209 blk_free_devt(dev
->devt
);
1210 disk_release_events(disk
);
1211 kfree(disk
->random
);
1212 xa_destroy(&disk
->part_tbl
);
1215 blk_put_queue(disk
->queue
);
1218 struct class block_class
= {
1222 static char *block_devnode(struct device
*dev
, umode_t
*mode
,
1223 kuid_t
*uid
, kgid_t
*gid
)
1225 struct gendisk
*disk
= dev_to_disk(dev
);
1227 if (disk
->fops
->devnode
)
1228 return disk
->fops
->devnode(disk
, mode
);
1232 const struct device_type disk_type
= {
1234 .groups
= disk_attr_groups
,
1235 .release
= disk_release
,
1236 .devnode
= block_devnode
,
1239 #ifdef CONFIG_PROC_FS
1241 * aggregate disk stat collector. Uses the same stats that the sysfs
1242 * entries do, above, but makes them available through one seq_file.
1244 * The output looks suspiciously like /proc/partitions with a bunch of
1247 static int diskstats_show(struct seq_file
*seqf
, void *v
)
1249 struct gendisk
*gp
= v
;
1250 struct disk_part_iter piter
;
1251 struct block_device
*hd
;
1252 char buf
[BDEVNAME_SIZE
];
1253 unsigned int inflight
;
1254 struct disk_stats stat
;
1257 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1258 seq_puts(seqf, "major minor name"
1259 " rio rmerge rsect ruse wio wmerge "
1260 "wsect wuse running use aveq"
1264 disk_part_iter_init(&piter
, gp
, DISK_PITER_INCL_EMPTY_PART0
);
1265 while ((hd
= disk_part_iter_next(&piter
))) {
1266 part_stat_read_all(hd
, &stat
);
1267 if (queue_is_mq(gp
->queue
))
1268 inflight
= blk_mq_in_flight(gp
->queue
, hd
);
1270 inflight
= part_in_flight(hd
);
1272 seq_printf(seqf
, "%4d %7d %s "
1279 MAJOR(hd
->bd_dev
), MINOR(hd
->bd_dev
),
1280 disk_name(gp
, hd
->bd_partno
, buf
),
1281 stat
.ios
[STAT_READ
],
1282 stat
.merges
[STAT_READ
],
1283 stat
.sectors
[STAT_READ
],
1284 (unsigned int)div_u64(stat
.nsecs
[STAT_READ
],
1286 stat
.ios
[STAT_WRITE
],
1287 stat
.merges
[STAT_WRITE
],
1288 stat
.sectors
[STAT_WRITE
],
1289 (unsigned int)div_u64(stat
.nsecs
[STAT_WRITE
],
1292 jiffies_to_msecs(stat
.io_ticks
),
1293 (unsigned int)div_u64(stat
.nsecs
[STAT_READ
] +
1294 stat
.nsecs
[STAT_WRITE
] +
1295 stat
.nsecs
[STAT_DISCARD
] +
1296 stat
.nsecs
[STAT_FLUSH
],
1298 stat
.ios
[STAT_DISCARD
],
1299 stat
.merges
[STAT_DISCARD
],
1300 stat
.sectors
[STAT_DISCARD
],
1301 (unsigned int)div_u64(stat
.nsecs
[STAT_DISCARD
],
1303 stat
.ios
[STAT_FLUSH
],
1304 (unsigned int)div_u64(stat
.nsecs
[STAT_FLUSH
],
1308 disk_part_iter_exit(&piter
);
1313 static const struct seq_operations diskstats_op
= {
1314 .start
= disk_seqf_start
,
1315 .next
= disk_seqf_next
,
1316 .stop
= disk_seqf_stop
,
1317 .show
= diskstats_show
1320 static int __init
proc_genhd_init(void)
1322 proc_create_seq("diskstats", 0, NULL
, &diskstats_op
);
1323 proc_create_seq("partitions", 0, NULL
, &partitions_op
);
1326 module_init(proc_genhd_init
);
1327 #endif /* CONFIG_PROC_FS */
1329 dev_t
blk_lookup_devt(const char *name
, int partno
)
1331 dev_t devt
= MKDEV(0, 0);
1332 struct class_dev_iter iter
;
1335 class_dev_iter_init(&iter
, &block_class
, NULL
, &disk_type
);
1336 while ((dev
= class_dev_iter_next(&iter
))) {
1337 struct gendisk
*disk
= dev_to_disk(dev
);
1338 struct block_device
*part
;
1340 if (strcmp(dev_name(dev
), name
))
1343 if (partno
< disk
->minors
) {
1344 /* We need to return the right devno, even
1345 * if the partition doesn't exist yet.
1347 devt
= MKDEV(MAJOR(dev
->devt
),
1348 MINOR(dev
->devt
) + partno
);
1351 part
= bdget_disk(disk
, partno
);
1353 devt
= part
->bd_dev
;
1358 class_dev_iter_exit(&iter
);
1362 struct gendisk
*__alloc_disk_node(int minors
, int node_id
)
1364 struct gendisk
*disk
;
1366 if (minors
> DISK_MAX_PARTS
) {
1368 "block: can't allocate more than %d partitions\n",
1370 minors
= DISK_MAX_PARTS
;
1373 disk
= kzalloc_node(sizeof(struct gendisk
), GFP_KERNEL
, node_id
);
1377 disk
->part0
= bdev_alloc(disk
, 0);
1381 disk
->node_id
= node_id
;
1382 xa_init(&disk
->part_tbl
);
1383 if (xa_insert(&disk
->part_tbl
, 0, disk
->part0
, GFP_KERNEL
))
1384 goto out_destroy_part_tbl
;
1386 disk
->minors
= minors
;
1387 rand_initialize_disk(disk
);
1388 disk_to_dev(disk
)->class = &block_class
;
1389 disk_to_dev(disk
)->type
= &disk_type
;
1390 device_initialize(disk_to_dev(disk
));
1393 out_destroy_part_tbl
:
1394 xa_destroy(&disk
->part_tbl
);
1400 EXPORT_SYMBOL(__alloc_disk_node
);
1403 * put_disk - decrements the gendisk refcount
1404 * @disk: the struct gendisk to decrement the refcount for
1406 * This decrements the refcount for the struct gendisk. When this reaches 0
1407 * we'll have disk_release() called.
1409 * Context: Any context, but the last reference must not be dropped from
1412 void put_disk(struct gendisk
*disk
)
1415 put_device(disk_to_dev(disk
));
1417 EXPORT_SYMBOL(put_disk
);
1419 static void set_disk_ro_uevent(struct gendisk
*gd
, int ro
)
1421 char event
[] = "DISK_RO=1";
1422 char *envp
[] = { event
, NULL
};
1426 kobject_uevent_env(&disk_to_dev(gd
)->kobj
, KOBJ_CHANGE
, envp
);
1430 * set_disk_ro - set a gendisk read-only
1431 * @disk: gendisk to operate on
1432 * @read_only: %true to set the disk read-only, %false set the disk read/write
1434 * This function is used to indicate whether a given disk device should have its
1435 * read-only flag set. set_disk_ro() is typically used by device drivers to
1436 * indicate whether the underlying physical device is write-protected.
1438 void set_disk_ro(struct gendisk
*disk
, bool read_only
)
1441 if (test_and_set_bit(GD_READ_ONLY
, &disk
->state
))
1444 if (!test_and_clear_bit(GD_READ_ONLY
, &disk
->state
))
1447 set_disk_ro_uevent(disk
, read_only
);
1449 EXPORT_SYMBOL(set_disk_ro
);
1451 int bdev_read_only(struct block_device
*bdev
)
1453 return bdev
->bd_read_only
|| get_disk_ro(bdev
->bd_disk
);
1455 EXPORT_SYMBOL(bdev_read_only
);
1458 * Disk events - monitor disk events like media change and eject request.
1460 struct disk_events
{
1461 struct list_head node
; /* all disk_event's */
1462 struct gendisk
*disk
; /* the associated disk */
1465 struct mutex block_mutex
; /* protects blocking */
1466 int block
; /* event blocking depth */
1467 unsigned int pending
; /* events already sent out */
1468 unsigned int clearing
; /* events being cleared */
1470 long poll_msecs
; /* interval, -1 for default */
1471 struct delayed_work dwork
;
1474 static const char *disk_events_strs
[] = {
1475 [ilog2(DISK_EVENT_MEDIA_CHANGE
)] = "media_change",
1476 [ilog2(DISK_EVENT_EJECT_REQUEST
)] = "eject_request",
1479 static char *disk_uevents
[] = {
1480 [ilog2(DISK_EVENT_MEDIA_CHANGE
)] = "DISK_MEDIA_CHANGE=1",
1481 [ilog2(DISK_EVENT_EJECT_REQUEST
)] = "DISK_EJECT_REQUEST=1",
1484 /* list of all disk_events */
1485 static DEFINE_MUTEX(disk_events_mutex
);
1486 static LIST_HEAD(disk_events
);
1488 /* disable in-kernel polling by default */
1489 static unsigned long disk_events_dfl_poll_msecs
;
1491 static unsigned long disk_events_poll_jiffies(struct gendisk
*disk
)
1493 struct disk_events
*ev
= disk
->ev
;
1494 long intv_msecs
= 0;
1497 * If device-specific poll interval is set, always use it. If
1498 * the default is being used, poll if the POLL flag is set.
1500 if (ev
->poll_msecs
>= 0)
1501 intv_msecs
= ev
->poll_msecs
;
1502 else if (disk
->event_flags
& DISK_EVENT_FLAG_POLL
)
1503 intv_msecs
= disk_events_dfl_poll_msecs
;
1505 return msecs_to_jiffies(intv_msecs
);
1509 * disk_block_events - block and flush disk event checking
1510 * @disk: disk to block events for
1512 * On return from this function, it is guaranteed that event checking
1513 * isn't in progress and won't happen until unblocked by
1514 * disk_unblock_events(). Events blocking is counted and the actual
1515 * unblocking happens after the matching number of unblocks are done.
1517 * Note that this intentionally does not block event checking from
1518 * disk_clear_events().
1523 void disk_block_events(struct gendisk
*disk
)
1525 struct disk_events
*ev
= disk
->ev
;
1526 unsigned long flags
;
1533 * Outer mutex ensures that the first blocker completes canceling
1534 * the event work before further blockers are allowed to finish.
1536 mutex_lock(&ev
->block_mutex
);
1538 spin_lock_irqsave(&ev
->lock
, flags
);
1539 cancel
= !ev
->block
++;
1540 spin_unlock_irqrestore(&ev
->lock
, flags
);
1543 cancel_delayed_work_sync(&disk
->ev
->dwork
);
1545 mutex_unlock(&ev
->block_mutex
);
1548 static void __disk_unblock_events(struct gendisk
*disk
, bool check_now
)
1550 struct disk_events
*ev
= disk
->ev
;
1552 unsigned long flags
;
1554 spin_lock_irqsave(&ev
->lock
, flags
);
1556 if (WARN_ON_ONCE(ev
->block
<= 0))
1562 intv
= disk_events_poll_jiffies(disk
);
1564 queue_delayed_work(system_freezable_power_efficient_wq
,
1567 queue_delayed_work(system_freezable_power_efficient_wq
,
1570 spin_unlock_irqrestore(&ev
->lock
, flags
);
1574 * disk_unblock_events - unblock disk event checking
1575 * @disk: disk to unblock events for
1577 * Undo disk_block_events(). When the block count reaches zero, it
1578 * starts events polling if configured.
1581 * Don't care. Safe to call from irq context.
1583 void disk_unblock_events(struct gendisk
*disk
)
1586 __disk_unblock_events(disk
, false);
1590 * disk_flush_events - schedule immediate event checking and flushing
1591 * @disk: disk to check and flush events for
1592 * @mask: events to flush
1594 * Schedule immediate event checking on @disk if not blocked. Events in
1595 * @mask are scheduled to be cleared from the driver. Note that this
1596 * doesn't clear the events from @disk->ev.
1599 * If @mask is non-zero must be called with bdev->bd_mutex held.
1601 void disk_flush_events(struct gendisk
*disk
, unsigned int mask
)
1603 struct disk_events
*ev
= disk
->ev
;
1608 spin_lock_irq(&ev
->lock
);
1609 ev
->clearing
|= mask
;
1611 mod_delayed_work(system_freezable_power_efficient_wq
,
1613 spin_unlock_irq(&ev
->lock
);
1617 * disk_clear_events - synchronously check, clear and return pending events
1618 * @disk: disk to fetch and clear events from
1619 * @mask: mask of events to be fetched and cleared
1621 * Disk events are synchronously checked and pending events in @mask
1622 * are cleared and returned. This ignores the block count.
1627 static unsigned int disk_clear_events(struct gendisk
*disk
, unsigned int mask
)
1629 struct disk_events
*ev
= disk
->ev
;
1630 unsigned int pending
;
1631 unsigned int clearing
= mask
;
1636 disk_block_events(disk
);
1639 * store the union of mask and ev->clearing on the stack so that the
1640 * race with disk_flush_events does not cause ambiguity (ev->clearing
1641 * can still be modified even if events are blocked).
1643 spin_lock_irq(&ev
->lock
);
1644 clearing
|= ev
->clearing
;
1646 spin_unlock_irq(&ev
->lock
);
1648 disk_check_events(ev
, &clearing
);
1650 * if ev->clearing is not 0, the disk_flush_events got called in the
1651 * middle of this function, so we want to run the workfn without delay.
1653 __disk_unblock_events(disk
, ev
->clearing
? true : false);
1655 /* then, fetch and clear pending events */
1656 spin_lock_irq(&ev
->lock
);
1657 pending
= ev
->pending
& mask
;
1658 ev
->pending
&= ~mask
;
1659 spin_unlock_irq(&ev
->lock
);
1660 WARN_ON_ONCE(clearing
& mask
);
1666 * bdev_check_media_change - check if a removable media has been changed
1667 * @bdev: block device to check
1669 * Check whether a removable media has been changed, and attempt to free all
1670 * dentries and inodes and invalidates all block device page cache entries in
1673 * Returns %true if the block device changed, or %false if not.
1675 bool bdev_check_media_change(struct block_device
*bdev
)
1677 unsigned int events
;
1679 events
= disk_clear_events(bdev
->bd_disk
, DISK_EVENT_MEDIA_CHANGE
|
1680 DISK_EVENT_EJECT_REQUEST
);
1681 if (!(events
& DISK_EVENT_MEDIA_CHANGE
))
1684 if (__invalidate_device(bdev
, true))
1685 pr_warn("VFS: busy inodes on changed media %s\n",
1686 bdev
->bd_disk
->disk_name
);
1687 set_bit(GD_NEED_PART_SCAN
, &bdev
->bd_disk
->state
);
1690 EXPORT_SYMBOL(bdev_check_media_change
);
1693 * Separate this part out so that a different pointer for clearing_ptr can be
1694 * passed in for disk_clear_events.
1696 static void disk_events_workfn(struct work_struct
*work
)
1698 struct delayed_work
*dwork
= to_delayed_work(work
);
1699 struct disk_events
*ev
= container_of(dwork
, struct disk_events
, dwork
);
1701 disk_check_events(ev
, &ev
->clearing
);
1704 static void disk_check_events(struct disk_events
*ev
,
1705 unsigned int *clearing_ptr
)
1707 struct gendisk
*disk
= ev
->disk
;
1708 char *envp
[ARRAY_SIZE(disk_uevents
) + 1] = { };
1709 unsigned int clearing
= *clearing_ptr
;
1710 unsigned int events
;
1712 int nr_events
= 0, i
;
1715 events
= disk
->fops
->check_events(disk
, clearing
);
1717 /* accumulate pending events and schedule next poll if necessary */
1718 spin_lock_irq(&ev
->lock
);
1720 events
&= ~ev
->pending
;
1721 ev
->pending
|= events
;
1722 *clearing_ptr
&= ~clearing
;
1724 intv
= disk_events_poll_jiffies(disk
);
1725 if (!ev
->block
&& intv
)
1726 queue_delayed_work(system_freezable_power_efficient_wq
,
1729 spin_unlock_irq(&ev
->lock
);
1732 * Tell userland about new events. Only the events listed in
1733 * @disk->events are reported, and only if DISK_EVENT_FLAG_UEVENT
1734 * is set. Otherwise, events are processed internally but never
1735 * get reported to userland.
1737 for (i
= 0; i
< ARRAY_SIZE(disk_uevents
); i
++)
1738 if ((events
& disk
->events
& (1 << i
)) &&
1739 (disk
->event_flags
& DISK_EVENT_FLAG_UEVENT
))
1740 envp
[nr_events
++] = disk_uevents
[i
];
1743 kobject_uevent_env(&disk_to_dev(disk
)->kobj
, KOBJ_CHANGE
, envp
);
1747 * A disk events enabled device has the following sysfs nodes under
1748 * its /sys/block/X/ directory.
1750 * events : list of all supported events
1751 * events_async : list of events which can be detected w/o polling
1752 * (always empty, only for backwards compatibility)
1753 * events_poll_msecs : polling interval, 0: disable, -1: system default
1755 static ssize_t
__disk_events_show(unsigned int events
, char *buf
)
1757 const char *delim
= "";
1761 for (i
= 0; i
< ARRAY_SIZE(disk_events_strs
); i
++)
1762 if (events
& (1 << i
)) {
1763 pos
+= sprintf(buf
+ pos
, "%s%s",
1764 delim
, disk_events_strs
[i
]);
1768 pos
+= sprintf(buf
+ pos
, "\n");
1772 static ssize_t
disk_events_show(struct device
*dev
,
1773 struct device_attribute
*attr
, char *buf
)
1775 struct gendisk
*disk
= dev_to_disk(dev
);
1777 if (!(disk
->event_flags
& DISK_EVENT_FLAG_UEVENT
))
1780 return __disk_events_show(disk
->events
, buf
);
1783 static ssize_t
disk_events_async_show(struct device
*dev
,
1784 struct device_attribute
*attr
, char *buf
)
1789 static ssize_t
disk_events_poll_msecs_show(struct device
*dev
,
1790 struct device_attribute
*attr
,
1793 struct gendisk
*disk
= dev_to_disk(dev
);
1796 return sprintf(buf
, "-1\n");
1798 return sprintf(buf
, "%ld\n", disk
->ev
->poll_msecs
);
1801 static ssize_t
disk_events_poll_msecs_store(struct device
*dev
,
1802 struct device_attribute
*attr
,
1803 const char *buf
, size_t count
)
1805 struct gendisk
*disk
= dev_to_disk(dev
);
1808 if (!count
|| !sscanf(buf
, "%ld", &intv
))
1811 if (intv
< 0 && intv
!= -1)
1817 disk_block_events(disk
);
1818 disk
->ev
->poll_msecs
= intv
;
1819 __disk_unblock_events(disk
, true);
1824 static const DEVICE_ATTR(events
, 0444, disk_events_show
, NULL
);
1825 static const DEVICE_ATTR(events_async
, 0444, disk_events_async_show
, NULL
);
1826 static const DEVICE_ATTR(events_poll_msecs
, 0644,
1827 disk_events_poll_msecs_show
,
1828 disk_events_poll_msecs_store
);
1830 static const struct attribute
*disk_events_attrs
[] = {
1831 &dev_attr_events
.attr
,
1832 &dev_attr_events_async
.attr
,
1833 &dev_attr_events_poll_msecs
.attr
,
1838 * The default polling interval can be specified by the kernel
1839 * parameter block.events_dfl_poll_msecs which defaults to 0
1840 * (disable). This can also be modified runtime by writing to
1841 * /sys/module/block/parameters/events_dfl_poll_msecs.
1843 static int disk_events_set_dfl_poll_msecs(const char *val
,
1844 const struct kernel_param
*kp
)
1846 struct disk_events
*ev
;
1849 ret
= param_set_ulong(val
, kp
);
1853 mutex_lock(&disk_events_mutex
);
1855 list_for_each_entry(ev
, &disk_events
, node
)
1856 disk_flush_events(ev
->disk
, 0);
1858 mutex_unlock(&disk_events_mutex
);
1863 static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops
= {
1864 .set
= disk_events_set_dfl_poll_msecs
,
1865 .get
= param_get_ulong
,
1868 #undef MODULE_PARAM_PREFIX
1869 #define MODULE_PARAM_PREFIX "block."
1871 module_param_cb(events_dfl_poll_msecs
, &disk_events_dfl_poll_msecs_param_ops
,
1872 &disk_events_dfl_poll_msecs
, 0644);
1875 * disk_{alloc|add|del|release}_events - initialize and destroy disk_events.
1877 static void disk_alloc_events(struct gendisk
*disk
)
1879 struct disk_events
*ev
;
1881 if (!disk
->fops
->check_events
|| !disk
->events
)
1884 ev
= kzalloc(sizeof(*ev
), GFP_KERNEL
);
1886 pr_warn("%s: failed to initialize events\n", disk
->disk_name
);
1890 INIT_LIST_HEAD(&ev
->node
);
1892 spin_lock_init(&ev
->lock
);
1893 mutex_init(&ev
->block_mutex
);
1895 ev
->poll_msecs
= -1;
1896 INIT_DELAYED_WORK(&ev
->dwork
, disk_events_workfn
);
1901 static void disk_add_events(struct gendisk
*disk
)
1903 /* FIXME: error handling */
1904 if (sysfs_create_files(&disk_to_dev(disk
)->kobj
, disk_events_attrs
) < 0)
1905 pr_warn("%s: failed to create sysfs files for events\n",
1911 mutex_lock(&disk_events_mutex
);
1912 list_add_tail(&disk
->ev
->node
, &disk_events
);
1913 mutex_unlock(&disk_events_mutex
);
1916 * Block count is initialized to 1 and the following initial
1917 * unblock kicks it into action.
1919 __disk_unblock_events(disk
, true);
1922 static void disk_del_events(struct gendisk
*disk
)
1925 disk_block_events(disk
);
1927 mutex_lock(&disk_events_mutex
);
1928 list_del_init(&disk
->ev
->node
);
1929 mutex_unlock(&disk_events_mutex
);
1932 sysfs_remove_files(&disk_to_dev(disk
)->kobj
, disk_events_attrs
);
1935 static void disk_release_events(struct gendisk
*disk
)
1937 /* the block count should be 1 from disk_del_events() */
1938 WARN_ON_ONCE(disk
->ev
&& disk
->ev
->block
!= 1);