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
, size
, capacity
);
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)
164 struct block_device
*__disk_get_part(struct gendisk
*disk
, int partno
)
166 struct disk_part_tbl
*ptbl
= rcu_dereference(disk
->part_tbl
);
168 if (unlikely(partno
< 0 || partno
>= ptbl
->len
))
170 return rcu_dereference(ptbl
->part
[partno
]);
174 * disk_part_iter_init - initialize partition iterator
175 * @piter: iterator to initialize
176 * @disk: disk to iterate over
177 * @flags: DISK_PITER_* flags
179 * Initialize @piter so that it iterates over partitions of @disk.
184 void disk_part_iter_init(struct disk_part_iter
*piter
, struct gendisk
*disk
,
187 struct disk_part_tbl
*ptbl
;
190 ptbl
= rcu_dereference(disk
->part_tbl
);
195 if (flags
& DISK_PITER_REVERSE
)
196 piter
->idx
= ptbl
->len
- 1;
197 else if (flags
& (DISK_PITER_INCL_PART0
| DISK_PITER_INCL_EMPTY_PART0
))
202 piter
->flags
= flags
;
206 EXPORT_SYMBOL_GPL(disk_part_iter_init
);
209 * disk_part_iter_next - proceed iterator to the next partition and return it
210 * @piter: iterator of interest
212 * Proceed @piter to the next partition and return it.
217 struct block_device
*disk_part_iter_next(struct disk_part_iter
*piter
)
219 struct disk_part_tbl
*ptbl
;
222 /* put the last partition */
223 disk_part_iter_exit(piter
);
227 ptbl
= rcu_dereference(piter
->disk
->part_tbl
);
229 /* determine iteration parameters */
230 if (piter
->flags
& DISK_PITER_REVERSE
) {
232 if (piter
->flags
& (DISK_PITER_INCL_PART0
|
233 DISK_PITER_INCL_EMPTY_PART0
))
242 /* iterate to the next partition */
243 for (; piter
->idx
!= end
; piter
->idx
+= inc
) {
244 struct block_device
*part
;
246 part
= rcu_dereference(ptbl
->part
[piter
->idx
]);
249 if (!bdev_nr_sectors(part
) &&
250 !(piter
->flags
& DISK_PITER_INCL_EMPTY
) &&
251 !(piter
->flags
& DISK_PITER_INCL_EMPTY_PART0
&&
255 piter
->part
= bdgrab(part
);
266 EXPORT_SYMBOL_GPL(disk_part_iter_next
);
269 * disk_part_iter_exit - finish up partition iteration
270 * @piter: iter of interest
272 * Called when iteration is over. Cleans up @piter.
277 void disk_part_iter_exit(struct disk_part_iter
*piter
)
283 EXPORT_SYMBOL_GPL(disk_part_iter_exit
);
285 static inline int sector_in_part(struct block_device
*part
, sector_t sector
)
287 return part
->bd_start_sect
<= sector
&&
288 sector
< part
->bd_start_sect
+ bdev_nr_sectors(part
);
292 * disk_map_sector_rcu - map sector to partition
293 * @disk: gendisk of interest
294 * @sector: sector to map
296 * Find out which partition @sector maps to on @disk. This is
297 * primarily used for stats accounting.
303 * Found partition on success, part0 is returned if no partition matches
304 * or the matched partition is being deleted.
306 struct block_device
*disk_map_sector_rcu(struct gendisk
*disk
, sector_t sector
)
308 struct disk_part_tbl
*ptbl
;
309 struct block_device
*part
;
313 ptbl
= rcu_dereference(disk
->part_tbl
);
315 part
= rcu_dereference(ptbl
->last_lookup
);
316 if (part
&& sector_in_part(part
, sector
))
319 for (i
= 1; i
< ptbl
->len
; i
++) {
320 part
= rcu_dereference(ptbl
->part
[i
]);
321 if (part
&& sector_in_part(part
, sector
)) {
322 rcu_assign_pointer(ptbl
->last_lookup
, part
);
334 * disk_has_partitions
335 * @disk: gendisk of interest
337 * Walk through the partition table and check if valid partition exists.
343 * True if the gendisk has at least one valid non-zero size partition.
346 bool disk_has_partitions(struct gendisk
*disk
)
348 struct disk_part_tbl
*ptbl
;
353 ptbl
= rcu_dereference(disk
->part_tbl
);
355 /* Iterate partitions skipping the whole device at index 0 */
356 for (i
= 1; i
< ptbl
->len
; i
++) {
357 if (rcu_dereference(ptbl
->part
[i
])) {
367 EXPORT_SYMBOL_GPL(disk_has_partitions
);
370 * Can be deleted altogether. Later.
373 #define BLKDEV_MAJOR_HASH_SIZE 255
374 static struct blk_major_name
{
375 struct blk_major_name
*next
;
378 void (*probe
)(dev_t devt
);
379 } *major_names
[BLKDEV_MAJOR_HASH_SIZE
];
380 static DEFINE_MUTEX(major_names_lock
);
382 /* index in the above - for now: assume no multimajor ranges */
383 static inline int major_to_index(unsigned major
)
385 return major
% BLKDEV_MAJOR_HASH_SIZE
;
388 #ifdef CONFIG_PROC_FS
389 void blkdev_show(struct seq_file
*seqf
, off_t offset
)
391 struct blk_major_name
*dp
;
393 mutex_lock(&major_names_lock
);
394 for (dp
= major_names
[major_to_index(offset
)]; dp
; dp
= dp
->next
)
395 if (dp
->major
== offset
)
396 seq_printf(seqf
, "%3d %s\n", dp
->major
, dp
->name
);
397 mutex_unlock(&major_names_lock
);
399 #endif /* CONFIG_PROC_FS */
402 * __register_blkdev - register a new block device
404 * @major: the requested major device number [1..BLKDEV_MAJOR_MAX-1]. If
405 * @major = 0, try to allocate any unused major number.
406 * @name: the name of the new block device as a zero terminated string
407 * @probe: allback that is called on access to any minor number of @major
409 * The @name must be unique within the system.
411 * The return value depends on the @major input parameter:
413 * - if a major device number was requested in range [1..BLKDEV_MAJOR_MAX-1]
414 * then the function returns zero on success, or a negative error code
415 * - if any unused major number was requested with @major = 0 parameter
416 * then the return value is the allocated major number in range
417 * [1..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise
419 * See Documentation/admin-guide/devices.txt for the list of allocated
422 * Use register_blkdev instead for any new code.
424 int __register_blkdev(unsigned int major
, const char *name
,
425 void (*probe
)(dev_t devt
))
427 struct blk_major_name
**n
, *p
;
430 mutex_lock(&major_names_lock
);
434 for (index
= ARRAY_SIZE(major_names
)-1; index
> 0; index
--) {
435 if (major_names
[index
] == NULL
)
440 printk("%s: failed to get major for %s\n",
449 if (major
>= BLKDEV_MAJOR_MAX
) {
450 pr_err("%s: major requested (%u) is greater than the maximum (%u) for %s\n",
451 __func__
, major
, BLKDEV_MAJOR_MAX
-1, name
);
457 p
= kmalloc(sizeof(struct blk_major_name
), GFP_KERNEL
);
465 strlcpy(p
->name
, name
, sizeof(p
->name
));
467 index
= major_to_index(major
);
469 for (n
= &major_names
[index
]; *n
; n
= &(*n
)->next
) {
470 if ((*n
)->major
== major
)
479 printk("register_blkdev: cannot get major %u for %s\n",
484 mutex_unlock(&major_names_lock
);
487 EXPORT_SYMBOL(__register_blkdev
);
489 void unregister_blkdev(unsigned int major
, const char *name
)
491 struct blk_major_name
**n
;
492 struct blk_major_name
*p
= NULL
;
493 int index
= major_to_index(major
);
495 mutex_lock(&major_names_lock
);
496 for (n
= &major_names
[index
]; *n
; n
= &(*n
)->next
)
497 if ((*n
)->major
== major
)
499 if (!*n
|| strcmp((*n
)->name
, name
)) {
505 mutex_unlock(&major_names_lock
);
509 EXPORT_SYMBOL(unregister_blkdev
);
512 * blk_mangle_minor - scatter minor numbers apart
513 * @minor: minor number to mangle
515 * Scatter consecutively allocated @minor number apart if MANGLE_DEVT
516 * is enabled. Mangling twice gives the original value.
524 static int blk_mangle_minor(int minor
)
526 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
529 for (i
= 0; i
< MINORBITS
/ 2; i
++) {
530 int low
= minor
& (1 << i
);
531 int high
= minor
& (1 << (MINORBITS
- 1 - i
));
532 int distance
= MINORBITS
- 1 - 2 * i
;
534 minor
^= low
| high
; /* clear both bits */
535 low
<<= distance
; /* swap the positions */
537 minor
|= low
| high
; /* and set */
544 * blk_alloc_devt - allocate a dev_t for a block device
545 * @bdev: block device to allocate dev_t for
546 * @devt: out parameter for resulting dev_t
548 * Allocate a dev_t for block device.
551 * 0 on success, allocated dev_t is returned in *@devt. -errno on
557 int blk_alloc_devt(struct block_device
*bdev
, dev_t
*devt
)
559 struct gendisk
*disk
= bdev
->bd_disk
;
562 /* in consecutive minor range? */
563 if (bdev
->bd_partno
< disk
->minors
) {
564 *devt
= MKDEV(disk
->major
, disk
->first_minor
+ bdev
->bd_partno
);
568 idx
= ida_alloc_range(&ext_devt_ida
, 0, NR_EXT_DEVT
, GFP_KERNEL
);
570 return idx
== -ENOSPC
? -EBUSY
: idx
;
572 *devt
= MKDEV(BLOCK_EXT_MAJOR
, blk_mangle_minor(idx
));
577 * blk_free_devt - free a dev_t
578 * @devt: dev_t to free
580 * Free @devt which was allocated using blk_alloc_devt().
585 void blk_free_devt(dev_t devt
)
587 if (MAJOR(devt
) == BLOCK_EXT_MAJOR
)
588 ida_free(&ext_devt_ida
, blk_mangle_minor(MINOR(devt
)));
591 static char *bdevt_str(dev_t devt
, char *buf
)
593 if (MAJOR(devt
) <= 0xff && MINOR(devt
) <= 0xff) {
594 char tbuf
[BDEVT_SIZE
];
595 snprintf(tbuf
, BDEVT_SIZE
, "%02x%02x", MAJOR(devt
), MINOR(devt
));
596 snprintf(buf
, BDEVT_SIZE
, "%-9s", tbuf
);
598 snprintf(buf
, BDEVT_SIZE
, "%03x:%05x", MAJOR(devt
), MINOR(devt
));
603 static void disk_scan_partitions(struct gendisk
*disk
)
605 struct block_device
*bdev
;
607 if (!get_capacity(disk
) || !disk_part_scan_enabled(disk
))
610 set_bit(GD_NEED_PART_SCAN
, &disk
->state
);
611 bdev
= blkdev_get_by_dev(disk_devt(disk
), FMODE_READ
, NULL
);
613 blkdev_put(bdev
, FMODE_READ
);
616 static void register_disk(struct device
*parent
, struct gendisk
*disk
,
617 const struct attribute_group
**groups
)
619 struct device
*ddev
= disk_to_dev(disk
);
620 struct disk_part_iter piter
;
621 struct block_device
*part
;
624 ddev
->parent
= parent
;
626 dev_set_name(ddev
, "%s", disk
->disk_name
);
628 /* delay uevents, until we scanned partition table */
629 dev_set_uevent_suppress(ddev
, 1);
632 WARN_ON(ddev
->groups
);
633 ddev
->groups
= groups
;
635 if (device_add(ddev
))
637 if (!sysfs_deprecated
) {
638 err
= sysfs_create_link(block_depr
, &ddev
->kobj
,
639 kobject_name(&ddev
->kobj
));
647 * avoid probable deadlock caused by allocating memory with
648 * GFP_KERNEL in runtime_resume callback of its all ancestor
651 pm_runtime_set_memalloc_noio(ddev
, true);
653 disk
->part0
->bd_holder_dir
=
654 kobject_create_and_add("holders", &ddev
->kobj
);
655 disk
->slave_dir
= kobject_create_and_add("slaves", &ddev
->kobj
);
657 if (disk
->flags
& GENHD_FL_HIDDEN
) {
658 dev_set_uevent_suppress(ddev
, 0);
662 disk_scan_partitions(disk
);
664 /* announce disk after possible partitions are created */
665 dev_set_uevent_suppress(ddev
, 0);
666 kobject_uevent(&ddev
->kobj
, KOBJ_ADD
);
668 /* announce possible partitions */
669 disk_part_iter_init(&piter
, disk
, 0);
670 while ((part
= disk_part_iter_next(&piter
)))
671 kobject_uevent(bdev_kobj(part
), KOBJ_ADD
);
672 disk_part_iter_exit(&piter
);
674 if (disk
->queue
->backing_dev_info
->dev
) {
675 err
= sysfs_create_link(&ddev
->kobj
,
676 &disk
->queue
->backing_dev_info
->dev
->kobj
,
683 * __device_add_disk - add disk information to kernel list
684 * @parent: parent device for the disk
685 * @disk: per-device partitioning information
686 * @groups: Additional per-device sysfs groups
687 * @register_queue: register the queue if set to true
689 * This function registers the partitioning information in @disk
692 * FIXME: error handling
694 static void __device_add_disk(struct device
*parent
, struct gendisk
*disk
,
695 const struct attribute_group
**groups
,
702 * The disk queue should now be all set with enough information about
703 * the device for the elevator code to pick an adequate default
704 * elevator if one is needed, that is, for devices requesting queue
708 elevator_init_mq(disk
->queue
);
710 /* minors == 0 indicates to use ext devt from part0 and should
711 * be accompanied with EXT_DEVT flag. Make sure all
712 * parameters make sense.
714 WARN_ON(disk
->minors
&& !(disk
->major
|| disk
->first_minor
));
715 WARN_ON(!disk
->minors
&&
716 !(disk
->flags
& (GENHD_FL_EXT_DEVT
| GENHD_FL_HIDDEN
)));
718 disk
->flags
|= GENHD_FL_UP
;
720 retval
= blk_alloc_devt(disk
->part0
, &devt
);
725 disk
->major
= MAJOR(devt
);
726 disk
->first_minor
= MINOR(devt
);
728 disk_alloc_events(disk
);
730 if (disk
->flags
& GENHD_FL_HIDDEN
) {
732 * Don't let hidden disks show up in /proc/partitions,
733 * and don't bother scanning for partitions either.
735 disk
->flags
|= GENHD_FL_SUPPRESS_PARTITION_INFO
;
736 disk
->flags
|= GENHD_FL_NO_PART_SCAN
;
738 struct backing_dev_info
*bdi
= disk
->queue
->backing_dev_info
;
739 struct device
*dev
= disk_to_dev(disk
);
742 /* Register BDI before referencing it from bdev */
744 ret
= bdi_register(bdi
, "%u:%u", MAJOR(devt
), MINOR(devt
));
746 bdi_set_owner(bdi
, dev
);
747 bdev_add(disk
->part0
, devt
);
749 register_disk(parent
, disk
, groups
);
751 blk_register_queue(disk
);
754 * Take an extra ref on queue which will be put on disk_release()
755 * so that it sticks around as long as @disk is there.
757 WARN_ON_ONCE(!blk_get_queue(disk
->queue
));
759 disk_add_events(disk
);
760 blk_integrity_add(disk
);
763 void device_add_disk(struct device
*parent
, struct gendisk
*disk
,
764 const struct attribute_group
**groups
)
767 __device_add_disk(parent
, disk
, groups
, true);
769 EXPORT_SYMBOL(device_add_disk
);
771 void device_add_disk_no_queue_reg(struct device
*parent
, struct gendisk
*disk
)
773 __device_add_disk(parent
, disk
, NULL
, false);
775 EXPORT_SYMBOL(device_add_disk_no_queue_reg
);
777 static void invalidate_partition(struct block_device
*bdev
)
780 __invalidate_device(bdev
, true);
783 * Unhash the bdev inode for this device so that it can't be looked
784 * up any more even if openers still hold references to it.
786 remove_inode_hash(bdev
->bd_inode
);
790 * del_gendisk - remove the gendisk
791 * @disk: the struct gendisk to remove
793 * Removes the gendisk and all its associated resources. This deletes the
794 * partitions associated with the gendisk, and unregisters the associated
797 * This is the counter to the respective __device_add_disk() call.
799 * The final removal of the struct gendisk happens when its refcount reaches 0
800 * with put_disk(), which should be called after del_gendisk(), if
801 * __device_add_disk() was used.
803 * Drivers exist which depend on the release of the gendisk to be synchronous,
804 * it should not be deferred.
808 void del_gendisk(struct gendisk
*disk
)
810 struct disk_part_iter piter
;
811 struct block_device
*part
;
815 if (WARN_ON_ONCE(!disk
->queue
))
818 blk_integrity_del(disk
);
819 disk_del_events(disk
);
822 * Block lookups of the disk until all bdevs are unhashed and the
823 * disk is marked as dead (GENHD_FL_UP cleared).
825 down_write(&bdev_lookup_sem
);
827 /* invalidate stuff */
828 disk_part_iter_init(&piter
, disk
,
829 DISK_PITER_INCL_EMPTY
| DISK_PITER_REVERSE
);
830 while ((part
= disk_part_iter_next(&piter
))) {
831 invalidate_partition(part
);
832 delete_partition(part
);
834 disk_part_iter_exit(&piter
);
836 invalidate_partition(disk
->part0
);
837 set_capacity(disk
, 0);
838 disk
->flags
&= ~GENHD_FL_UP
;
839 up_write(&bdev_lookup_sem
);
841 if (!(disk
->flags
& GENHD_FL_HIDDEN
)) {
842 sysfs_remove_link(&disk_to_dev(disk
)->kobj
, "bdi");
845 * Unregister bdi before releasing device numbers (as they can
846 * get reused and we'd get clashes in sysfs).
848 bdi_unregister(disk
->queue
->backing_dev_info
);
851 blk_unregister_queue(disk
);
853 kobject_put(disk
->part0
->bd_holder_dir
);
854 kobject_put(disk
->slave_dir
);
856 part_stat_set_all(disk
->part0
, 0);
857 disk
->part0
->bd_stamp
= 0;
858 if (!sysfs_deprecated
)
859 sysfs_remove_link(block_depr
, dev_name(disk_to_dev(disk
)));
860 pm_runtime_set_memalloc_noio(disk_to_dev(disk
), false);
861 device_del(disk_to_dev(disk
));
863 EXPORT_SYMBOL(del_gendisk
);
865 /* sysfs access to bad-blocks list. */
866 static ssize_t
disk_badblocks_show(struct device
*dev
,
867 struct device_attribute
*attr
,
870 struct gendisk
*disk
= dev_to_disk(dev
);
873 return sprintf(page
, "\n");
875 return badblocks_show(disk
->bb
, page
, 0);
878 static ssize_t
disk_badblocks_store(struct device
*dev
,
879 struct device_attribute
*attr
,
880 const char *page
, size_t len
)
882 struct gendisk
*disk
= dev_to_disk(dev
);
887 return badblocks_store(disk
->bb
, page
, len
, 0);
890 void blk_request_module(dev_t devt
)
892 unsigned int major
= MAJOR(devt
);
893 struct blk_major_name
**n
;
895 mutex_lock(&major_names_lock
);
896 for (n
= &major_names
[major_to_index(major
)]; *n
; n
= &(*n
)->next
) {
897 if ((*n
)->major
== major
&& (*n
)->probe
) {
899 mutex_unlock(&major_names_lock
);
903 mutex_unlock(&major_names_lock
);
905 if (request_module("block-major-%d-%d", MAJOR(devt
), MINOR(devt
)) > 0)
906 /* Make old-style 2.4 aliases work */
907 request_module("block-major-%d", MAJOR(devt
));
911 * bdget_disk - do bdget() by gendisk and partition number
912 * @disk: gendisk of interest
913 * @partno: partition number
915 * Find partition @partno from @disk, do bdget() on it.
921 * Resulting block_device on success, NULL on failure.
923 struct block_device
*bdget_disk(struct gendisk
*disk
, int partno
)
925 struct block_device
*bdev
= NULL
;
928 bdev
= __disk_get_part(disk
, partno
);
929 if (bdev
&& !bdgrab(bdev
))
937 * print a full list of all partitions - intended for places where the root
938 * filesystem can't be mounted and thus to give the victim some idea of what
941 void __init
printk_all_partitions(void)
943 struct class_dev_iter iter
;
946 class_dev_iter_init(&iter
, &block_class
, NULL
, &disk_type
);
947 while ((dev
= class_dev_iter_next(&iter
))) {
948 struct gendisk
*disk
= dev_to_disk(dev
);
949 struct disk_part_iter piter
;
950 struct block_device
*part
;
951 char name_buf
[BDEVNAME_SIZE
];
952 char devt_buf
[BDEVT_SIZE
];
955 * Don't show empty devices or things that have been
958 if (get_capacity(disk
) == 0 ||
959 (disk
->flags
& GENHD_FL_SUPPRESS_PARTITION_INFO
))
963 * Note, unlike /proc/partitions, I am showing the
964 * numbers in hex - the same format as the root=
967 disk_part_iter_init(&piter
, disk
, DISK_PITER_INCL_PART0
);
968 while ((part
= disk_part_iter_next(&piter
))) {
969 bool is_part0
= part
== disk
->part0
;
971 printk("%s%s %10llu %s %s", is_part0
? "" : " ",
972 bdevt_str(part
->bd_dev
, devt_buf
),
973 bdev_nr_sectors(part
) >> 1,
974 disk_name(disk
, part
->bd_partno
, name_buf
),
976 part
->bd_meta_info
->uuid
: "");
978 if (dev
->parent
&& dev
->parent
->driver
)
979 printk(" driver: %s\n",
980 dev
->parent
->driver
->name
);
982 printk(" (driver?)\n");
986 disk_part_iter_exit(&piter
);
988 class_dev_iter_exit(&iter
);
991 #ifdef CONFIG_PROC_FS
993 static void *disk_seqf_start(struct seq_file
*seqf
, loff_t
*pos
)
996 struct class_dev_iter
*iter
;
999 iter
= kmalloc(sizeof(*iter
), GFP_KERNEL
);
1001 return ERR_PTR(-ENOMEM
);
1003 seqf
->private = iter
;
1004 class_dev_iter_init(iter
, &block_class
, NULL
, &disk_type
);
1006 dev
= class_dev_iter_next(iter
);
1011 return dev_to_disk(dev
);
1014 static void *disk_seqf_next(struct seq_file
*seqf
, void *v
, loff_t
*pos
)
1019 dev
= class_dev_iter_next(seqf
->private);
1021 return dev_to_disk(dev
);
1026 static void disk_seqf_stop(struct seq_file
*seqf
, void *v
)
1028 struct class_dev_iter
*iter
= seqf
->private;
1030 /* stop is called even after start failed :-( */
1032 class_dev_iter_exit(iter
);
1034 seqf
->private = NULL
;
1038 static void *show_partition_start(struct seq_file
*seqf
, loff_t
*pos
)
1042 p
= disk_seqf_start(seqf
, pos
);
1043 if (!IS_ERR_OR_NULL(p
) && !*pos
)
1044 seq_puts(seqf
, "major minor #blocks name\n\n");
1048 static int show_partition(struct seq_file
*seqf
, void *v
)
1050 struct gendisk
*sgp
= v
;
1051 struct disk_part_iter piter
;
1052 struct block_device
*part
;
1053 char buf
[BDEVNAME_SIZE
];
1055 /* Don't show non-partitionable removeable devices or empty devices */
1056 if (!get_capacity(sgp
) || (!disk_max_parts(sgp
) &&
1057 (sgp
->flags
& GENHD_FL_REMOVABLE
)))
1059 if (sgp
->flags
& GENHD_FL_SUPPRESS_PARTITION_INFO
)
1062 /* show the full disk and all non-0 size partitions of it */
1063 disk_part_iter_init(&piter
, sgp
, DISK_PITER_INCL_PART0
);
1064 while ((part
= disk_part_iter_next(&piter
)))
1065 seq_printf(seqf
, "%4d %7d %10llu %s\n",
1066 MAJOR(part
->bd_dev
), MINOR(part
->bd_dev
),
1067 bdev_nr_sectors(part
) >> 1,
1068 disk_name(sgp
, part
->bd_partno
, buf
));
1069 disk_part_iter_exit(&piter
);
1074 static const struct seq_operations partitions_op
= {
1075 .start
= show_partition_start
,
1076 .next
= disk_seqf_next
,
1077 .stop
= disk_seqf_stop
,
1078 .show
= show_partition
1082 static int __init
genhd_device_init(void)
1086 block_class
.dev_kobj
= sysfs_dev_block_kobj
;
1087 error
= class_register(&block_class
);
1088 if (unlikely(error
))
1092 register_blkdev(BLOCK_EXT_MAJOR
, "blkext");
1094 /* create top-level block dir */
1095 if (!sysfs_deprecated
)
1096 block_depr
= kobject_create_and_add("block", NULL
);
1100 subsys_initcall(genhd_device_init
);
1102 static ssize_t
disk_range_show(struct device
*dev
,
1103 struct device_attribute
*attr
, char *buf
)
1105 struct gendisk
*disk
= dev_to_disk(dev
);
1107 return sprintf(buf
, "%d\n", disk
->minors
);
1110 static ssize_t
disk_ext_range_show(struct device
*dev
,
1111 struct device_attribute
*attr
, char *buf
)
1113 struct gendisk
*disk
= dev_to_disk(dev
);
1115 return sprintf(buf
, "%d\n", disk_max_parts(disk
));
1118 static ssize_t
disk_removable_show(struct device
*dev
,
1119 struct device_attribute
*attr
, char *buf
)
1121 struct gendisk
*disk
= dev_to_disk(dev
);
1123 return sprintf(buf
, "%d\n",
1124 (disk
->flags
& GENHD_FL_REMOVABLE
? 1 : 0));
1127 static ssize_t
disk_hidden_show(struct device
*dev
,
1128 struct device_attribute
*attr
, char *buf
)
1130 struct gendisk
*disk
= dev_to_disk(dev
);
1132 return sprintf(buf
, "%d\n",
1133 (disk
->flags
& GENHD_FL_HIDDEN
? 1 : 0));
1136 static ssize_t
disk_ro_show(struct device
*dev
,
1137 struct device_attribute
*attr
, char *buf
)
1139 struct gendisk
*disk
= dev_to_disk(dev
);
1141 return sprintf(buf
, "%d\n", get_disk_ro(disk
) ? 1 : 0);
1144 ssize_t
part_size_show(struct device
*dev
,
1145 struct device_attribute
*attr
, char *buf
)
1147 return sprintf(buf
, "%llu\n", bdev_nr_sectors(dev_to_bdev(dev
)));
1150 ssize_t
part_stat_show(struct device
*dev
,
1151 struct device_attribute
*attr
, char *buf
)
1153 struct block_device
*bdev
= dev_to_bdev(dev
);
1154 struct request_queue
*q
= bdev
->bd_disk
->queue
;
1155 struct disk_stats stat
;
1156 unsigned int inflight
;
1158 part_stat_read_all(bdev
, &stat
);
1160 inflight
= blk_mq_in_flight(q
, bdev
);
1162 inflight
= part_in_flight(bdev
);
1165 "%8lu %8lu %8llu %8u "
1166 "%8lu %8lu %8llu %8u "
1168 "%8lu %8lu %8llu %8u "
1171 stat
.ios
[STAT_READ
],
1172 stat
.merges
[STAT_READ
],
1173 (unsigned long long)stat
.sectors
[STAT_READ
],
1174 (unsigned int)div_u64(stat
.nsecs
[STAT_READ
], NSEC_PER_MSEC
),
1175 stat
.ios
[STAT_WRITE
],
1176 stat
.merges
[STAT_WRITE
],
1177 (unsigned long long)stat
.sectors
[STAT_WRITE
],
1178 (unsigned int)div_u64(stat
.nsecs
[STAT_WRITE
], NSEC_PER_MSEC
),
1180 jiffies_to_msecs(stat
.io_ticks
),
1181 (unsigned int)div_u64(stat
.nsecs
[STAT_READ
] +
1182 stat
.nsecs
[STAT_WRITE
] +
1183 stat
.nsecs
[STAT_DISCARD
] +
1184 stat
.nsecs
[STAT_FLUSH
],
1186 stat
.ios
[STAT_DISCARD
],
1187 stat
.merges
[STAT_DISCARD
],
1188 (unsigned long long)stat
.sectors
[STAT_DISCARD
],
1189 (unsigned int)div_u64(stat
.nsecs
[STAT_DISCARD
], NSEC_PER_MSEC
),
1190 stat
.ios
[STAT_FLUSH
],
1191 (unsigned int)div_u64(stat
.nsecs
[STAT_FLUSH
], NSEC_PER_MSEC
));
1194 ssize_t
part_inflight_show(struct device
*dev
, struct device_attribute
*attr
,
1197 struct block_device
*bdev
= dev_to_bdev(dev
);
1198 struct request_queue
*q
= bdev
->bd_disk
->queue
;
1199 unsigned int inflight
[2];
1202 blk_mq_in_flight_rw(q
, bdev
, inflight
);
1204 part_in_flight_rw(bdev
, inflight
);
1206 return sprintf(buf
, "%8u %8u\n", inflight
[0], inflight
[1]);
1209 static ssize_t
disk_capability_show(struct device
*dev
,
1210 struct device_attribute
*attr
, char *buf
)
1212 struct gendisk
*disk
= dev_to_disk(dev
);
1214 return sprintf(buf
, "%x\n", disk
->flags
);
1217 static ssize_t
disk_alignment_offset_show(struct device
*dev
,
1218 struct device_attribute
*attr
,
1221 struct gendisk
*disk
= dev_to_disk(dev
);
1223 return sprintf(buf
, "%d\n", queue_alignment_offset(disk
->queue
));
1226 static ssize_t
disk_discard_alignment_show(struct device
*dev
,
1227 struct device_attribute
*attr
,
1230 struct gendisk
*disk
= dev_to_disk(dev
);
1232 return sprintf(buf
, "%d\n", queue_discard_alignment(disk
->queue
));
1235 static DEVICE_ATTR(range
, 0444, disk_range_show
, NULL
);
1236 static DEVICE_ATTR(ext_range
, 0444, disk_ext_range_show
, NULL
);
1237 static DEVICE_ATTR(removable
, 0444, disk_removable_show
, NULL
);
1238 static DEVICE_ATTR(hidden
, 0444, disk_hidden_show
, NULL
);
1239 static DEVICE_ATTR(ro
, 0444, disk_ro_show
, NULL
);
1240 static DEVICE_ATTR(size
, 0444, part_size_show
, NULL
);
1241 static DEVICE_ATTR(alignment_offset
, 0444, disk_alignment_offset_show
, NULL
);
1242 static DEVICE_ATTR(discard_alignment
, 0444, disk_discard_alignment_show
, NULL
);
1243 static DEVICE_ATTR(capability
, 0444, disk_capability_show
, NULL
);
1244 static DEVICE_ATTR(stat
, 0444, part_stat_show
, NULL
);
1245 static DEVICE_ATTR(inflight
, 0444, part_inflight_show
, NULL
);
1246 static DEVICE_ATTR(badblocks
, 0644, disk_badblocks_show
, disk_badblocks_store
);
1248 #ifdef CONFIG_FAIL_MAKE_REQUEST
1249 ssize_t
part_fail_show(struct device
*dev
,
1250 struct device_attribute
*attr
, char *buf
)
1252 return sprintf(buf
, "%d\n", dev_to_bdev(dev
)->bd_make_it_fail
);
1255 ssize_t
part_fail_store(struct device
*dev
,
1256 struct device_attribute
*attr
,
1257 const char *buf
, size_t count
)
1261 if (count
> 0 && sscanf(buf
, "%d", &i
) > 0)
1262 dev_to_bdev(dev
)->bd_make_it_fail
= i
;
1267 static struct device_attribute dev_attr_fail
=
1268 __ATTR(make
-it
-fail
, 0644, part_fail_show
, part_fail_store
);
1269 #endif /* CONFIG_FAIL_MAKE_REQUEST */
1271 #ifdef CONFIG_FAIL_IO_TIMEOUT
1272 static struct device_attribute dev_attr_fail_timeout
=
1273 __ATTR(io
-timeout
-fail
, 0644, part_timeout_show
, part_timeout_store
);
1276 static struct attribute
*disk_attrs
[] = {
1277 &dev_attr_range
.attr
,
1278 &dev_attr_ext_range
.attr
,
1279 &dev_attr_removable
.attr
,
1280 &dev_attr_hidden
.attr
,
1282 &dev_attr_size
.attr
,
1283 &dev_attr_alignment_offset
.attr
,
1284 &dev_attr_discard_alignment
.attr
,
1285 &dev_attr_capability
.attr
,
1286 &dev_attr_stat
.attr
,
1287 &dev_attr_inflight
.attr
,
1288 &dev_attr_badblocks
.attr
,
1289 #ifdef CONFIG_FAIL_MAKE_REQUEST
1290 &dev_attr_fail
.attr
,
1292 #ifdef CONFIG_FAIL_IO_TIMEOUT
1293 &dev_attr_fail_timeout
.attr
,
1298 static umode_t
disk_visible(struct kobject
*kobj
, struct attribute
*a
, int n
)
1300 struct device
*dev
= container_of(kobj
, typeof(*dev
), kobj
);
1301 struct gendisk
*disk
= dev_to_disk(dev
);
1303 if (a
== &dev_attr_badblocks
.attr
&& !disk
->bb
)
1308 static struct attribute_group disk_attr_group
= {
1309 .attrs
= disk_attrs
,
1310 .is_visible
= disk_visible
,
1313 static const struct attribute_group
*disk_attr_groups
[] = {
1319 * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way
1320 * @disk: disk to replace part_tbl for
1321 * @new_ptbl: new part_tbl to install
1323 * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The
1324 * original ptbl is freed using RCU callback.
1327 * Matching bd_mutex locked or the caller is the only user of @disk.
1329 static void disk_replace_part_tbl(struct gendisk
*disk
,
1330 struct disk_part_tbl
*new_ptbl
)
1332 struct disk_part_tbl
*old_ptbl
=
1333 rcu_dereference_protected(disk
->part_tbl
, 1);
1335 rcu_assign_pointer(disk
->part_tbl
, new_ptbl
);
1338 rcu_assign_pointer(old_ptbl
->last_lookup
, NULL
);
1339 kfree_rcu(old_ptbl
, rcu_head
);
1344 * disk_expand_part_tbl - expand disk->part_tbl
1345 * @disk: disk to expand part_tbl for
1346 * @partno: expand such that this partno can fit in
1348 * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl
1349 * uses RCU to allow unlocked dereferencing for stats and other stuff.
1352 * Matching bd_mutex locked or the caller is the only user of @disk.
1356 * 0 on success, -errno on failure.
1358 int disk_expand_part_tbl(struct gendisk
*disk
, int partno
)
1360 struct disk_part_tbl
*old_ptbl
=
1361 rcu_dereference_protected(disk
->part_tbl
, 1);
1362 struct disk_part_tbl
*new_ptbl
;
1363 int len
= old_ptbl
? old_ptbl
->len
: 0;
1367 * check for int overflow, since we can get here from blkpg_ioctl()
1368 * with a user passed 'partno'.
1370 target
= partno
+ 1;
1374 /* disk_max_parts() is zero during initialization, ignore if so */
1375 if (disk_max_parts(disk
) && target
> disk_max_parts(disk
))
1381 new_ptbl
= kzalloc_node(struct_size(new_ptbl
, part
, target
), GFP_KERNEL
,
1386 new_ptbl
->len
= target
;
1388 for (i
= 0; i
< len
; i
++)
1389 rcu_assign_pointer(new_ptbl
->part
[i
], old_ptbl
->part
[i
]);
1391 disk_replace_part_tbl(disk
, new_ptbl
);
1396 * disk_release - releases all allocated resources of the gendisk
1397 * @dev: the device representing this disk
1399 * This function releases all allocated resources of the gendisk.
1401 * Drivers which used __device_add_disk() have a gendisk with a request_queue
1402 * assigned. Since the request_queue sits on top of the gendisk for these
1403 * drivers we also call blk_put_queue() for them, and we expect the
1404 * request_queue refcount to reach 0 at this point, and so the request_queue
1405 * will also be freed prior to the disk.
1407 * Context: can sleep
1409 static void disk_release(struct device
*dev
)
1411 struct gendisk
*disk
= dev_to_disk(dev
);
1415 blk_free_devt(dev
->devt
);
1416 disk_release_events(disk
);
1417 kfree(disk
->random
);
1418 disk_replace_part_tbl(disk
, NULL
);
1421 blk_put_queue(disk
->queue
);
1424 struct class block_class
= {
1428 static char *block_devnode(struct device
*dev
, umode_t
*mode
,
1429 kuid_t
*uid
, kgid_t
*gid
)
1431 struct gendisk
*disk
= dev_to_disk(dev
);
1433 if (disk
->fops
->devnode
)
1434 return disk
->fops
->devnode(disk
, mode
);
1438 const struct device_type disk_type
= {
1440 .groups
= disk_attr_groups
,
1441 .release
= disk_release
,
1442 .devnode
= block_devnode
,
1445 #ifdef CONFIG_PROC_FS
1447 * aggregate disk stat collector. Uses the same stats that the sysfs
1448 * entries do, above, but makes them available through one seq_file.
1450 * The output looks suspiciously like /proc/partitions with a bunch of
1453 static int diskstats_show(struct seq_file
*seqf
, void *v
)
1455 struct gendisk
*gp
= v
;
1456 struct disk_part_iter piter
;
1457 struct block_device
*hd
;
1458 char buf
[BDEVNAME_SIZE
];
1459 unsigned int inflight
;
1460 struct disk_stats stat
;
1463 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1464 seq_puts(seqf, "major minor name"
1465 " rio rmerge rsect ruse wio wmerge "
1466 "wsect wuse running use aveq"
1470 disk_part_iter_init(&piter
, gp
, DISK_PITER_INCL_EMPTY_PART0
);
1471 while ((hd
= disk_part_iter_next(&piter
))) {
1472 part_stat_read_all(hd
, &stat
);
1473 if (queue_is_mq(gp
->queue
))
1474 inflight
= blk_mq_in_flight(gp
->queue
, hd
);
1476 inflight
= part_in_flight(hd
);
1478 seq_printf(seqf
, "%4d %7d %s "
1485 MAJOR(hd
->bd_dev
), MINOR(hd
->bd_dev
),
1486 disk_name(gp
, hd
->bd_partno
, buf
),
1487 stat
.ios
[STAT_READ
],
1488 stat
.merges
[STAT_READ
],
1489 stat
.sectors
[STAT_READ
],
1490 (unsigned int)div_u64(stat
.nsecs
[STAT_READ
],
1492 stat
.ios
[STAT_WRITE
],
1493 stat
.merges
[STAT_WRITE
],
1494 stat
.sectors
[STAT_WRITE
],
1495 (unsigned int)div_u64(stat
.nsecs
[STAT_WRITE
],
1498 jiffies_to_msecs(stat
.io_ticks
),
1499 (unsigned int)div_u64(stat
.nsecs
[STAT_READ
] +
1500 stat
.nsecs
[STAT_WRITE
] +
1501 stat
.nsecs
[STAT_DISCARD
] +
1502 stat
.nsecs
[STAT_FLUSH
],
1504 stat
.ios
[STAT_DISCARD
],
1505 stat
.merges
[STAT_DISCARD
],
1506 stat
.sectors
[STAT_DISCARD
],
1507 (unsigned int)div_u64(stat
.nsecs
[STAT_DISCARD
],
1509 stat
.ios
[STAT_FLUSH
],
1510 (unsigned int)div_u64(stat
.nsecs
[STAT_FLUSH
],
1514 disk_part_iter_exit(&piter
);
1519 static const struct seq_operations diskstats_op
= {
1520 .start
= disk_seqf_start
,
1521 .next
= disk_seqf_next
,
1522 .stop
= disk_seqf_stop
,
1523 .show
= diskstats_show
1526 static int __init
proc_genhd_init(void)
1528 proc_create_seq("diskstats", 0, NULL
, &diskstats_op
);
1529 proc_create_seq("partitions", 0, NULL
, &partitions_op
);
1532 module_init(proc_genhd_init
);
1533 #endif /* CONFIG_PROC_FS */
1535 dev_t
blk_lookup_devt(const char *name
, int partno
)
1537 dev_t devt
= MKDEV(0, 0);
1538 struct class_dev_iter iter
;
1541 class_dev_iter_init(&iter
, &block_class
, NULL
, &disk_type
);
1542 while ((dev
= class_dev_iter_next(&iter
))) {
1543 struct gendisk
*disk
= dev_to_disk(dev
);
1544 struct block_device
*part
;
1546 if (strcmp(dev_name(dev
), name
))
1549 if (partno
< disk
->minors
) {
1550 /* We need to return the right devno, even
1551 * if the partition doesn't exist yet.
1553 devt
= MKDEV(MAJOR(dev
->devt
),
1554 MINOR(dev
->devt
) + partno
);
1557 part
= bdget_disk(disk
, partno
);
1559 devt
= part
->bd_dev
;
1564 class_dev_iter_exit(&iter
);
1568 struct gendisk
*__alloc_disk_node(int minors
, int node_id
)
1570 struct gendisk
*disk
;
1571 struct disk_part_tbl
*ptbl
;
1573 if (minors
> DISK_MAX_PARTS
) {
1575 "block: can't allocate more than %d partitions\n",
1577 minors
= DISK_MAX_PARTS
;
1580 disk
= kzalloc_node(sizeof(struct gendisk
), GFP_KERNEL
, node_id
);
1584 disk
->part0
= bdev_alloc(disk
, 0);
1588 disk
->node_id
= node_id
;
1589 if (disk_expand_part_tbl(disk
, 0))
1592 ptbl
= rcu_dereference_protected(disk
->part_tbl
, 1);
1593 rcu_assign_pointer(ptbl
->part
[0], disk
->part0
);
1595 disk
->minors
= minors
;
1596 rand_initialize_disk(disk
);
1597 disk_to_dev(disk
)->class = &block_class
;
1598 disk_to_dev(disk
)->type
= &disk_type
;
1599 device_initialize(disk_to_dev(disk
));
1608 EXPORT_SYMBOL(__alloc_disk_node
);
1611 * put_disk - decrements the gendisk refcount
1612 * @disk: the struct gendisk to decrement the refcount for
1614 * This decrements the refcount for the struct gendisk. When this reaches 0
1615 * we'll have disk_release() called.
1617 * Context: Any context, but the last reference must not be dropped from
1620 void put_disk(struct gendisk
*disk
)
1623 put_device(disk_to_dev(disk
));
1625 EXPORT_SYMBOL(put_disk
);
1627 static void set_disk_ro_uevent(struct gendisk
*gd
, int ro
)
1629 char event
[] = "DISK_RO=1";
1630 char *envp
[] = { event
, NULL
};
1634 kobject_uevent_env(&disk_to_dev(gd
)->kobj
, KOBJ_CHANGE
, envp
);
1637 void set_disk_ro(struct gendisk
*disk
, int flag
)
1639 struct disk_part_iter piter
;
1640 struct block_device
*part
;
1642 if (disk
->part0
->bd_read_only
!= flag
) {
1643 set_disk_ro_uevent(disk
, flag
);
1644 disk
->part0
->bd_read_only
= flag
;
1647 disk_part_iter_init(&piter
, disk
, DISK_PITER_INCL_EMPTY
);
1648 while ((part
= disk_part_iter_next(&piter
)))
1649 part
->bd_read_only
= flag
;
1650 disk_part_iter_exit(&piter
);
1653 EXPORT_SYMBOL(set_disk_ro
);
1655 int bdev_read_only(struct block_device
*bdev
)
1659 return bdev
->bd_read_only
;
1662 EXPORT_SYMBOL(bdev_read_only
);
1665 * Disk events - monitor disk events like media change and eject request.
1667 struct disk_events
{
1668 struct list_head node
; /* all disk_event's */
1669 struct gendisk
*disk
; /* the associated disk */
1672 struct mutex block_mutex
; /* protects blocking */
1673 int block
; /* event blocking depth */
1674 unsigned int pending
; /* events already sent out */
1675 unsigned int clearing
; /* events being cleared */
1677 long poll_msecs
; /* interval, -1 for default */
1678 struct delayed_work dwork
;
1681 static const char *disk_events_strs
[] = {
1682 [ilog2(DISK_EVENT_MEDIA_CHANGE
)] = "media_change",
1683 [ilog2(DISK_EVENT_EJECT_REQUEST
)] = "eject_request",
1686 static char *disk_uevents
[] = {
1687 [ilog2(DISK_EVENT_MEDIA_CHANGE
)] = "DISK_MEDIA_CHANGE=1",
1688 [ilog2(DISK_EVENT_EJECT_REQUEST
)] = "DISK_EJECT_REQUEST=1",
1691 /* list of all disk_events */
1692 static DEFINE_MUTEX(disk_events_mutex
);
1693 static LIST_HEAD(disk_events
);
1695 /* disable in-kernel polling by default */
1696 static unsigned long disk_events_dfl_poll_msecs
;
1698 static unsigned long disk_events_poll_jiffies(struct gendisk
*disk
)
1700 struct disk_events
*ev
= disk
->ev
;
1701 long intv_msecs
= 0;
1704 * If device-specific poll interval is set, always use it. If
1705 * the default is being used, poll if the POLL flag is set.
1707 if (ev
->poll_msecs
>= 0)
1708 intv_msecs
= ev
->poll_msecs
;
1709 else if (disk
->event_flags
& DISK_EVENT_FLAG_POLL
)
1710 intv_msecs
= disk_events_dfl_poll_msecs
;
1712 return msecs_to_jiffies(intv_msecs
);
1716 * disk_block_events - block and flush disk event checking
1717 * @disk: disk to block events for
1719 * On return from this function, it is guaranteed that event checking
1720 * isn't in progress and won't happen until unblocked by
1721 * disk_unblock_events(). Events blocking is counted and the actual
1722 * unblocking happens after the matching number of unblocks are done.
1724 * Note that this intentionally does not block event checking from
1725 * disk_clear_events().
1730 void disk_block_events(struct gendisk
*disk
)
1732 struct disk_events
*ev
= disk
->ev
;
1733 unsigned long flags
;
1740 * Outer mutex ensures that the first blocker completes canceling
1741 * the event work before further blockers are allowed to finish.
1743 mutex_lock(&ev
->block_mutex
);
1745 spin_lock_irqsave(&ev
->lock
, flags
);
1746 cancel
= !ev
->block
++;
1747 spin_unlock_irqrestore(&ev
->lock
, flags
);
1750 cancel_delayed_work_sync(&disk
->ev
->dwork
);
1752 mutex_unlock(&ev
->block_mutex
);
1755 static void __disk_unblock_events(struct gendisk
*disk
, bool check_now
)
1757 struct disk_events
*ev
= disk
->ev
;
1759 unsigned long flags
;
1761 spin_lock_irqsave(&ev
->lock
, flags
);
1763 if (WARN_ON_ONCE(ev
->block
<= 0))
1769 intv
= disk_events_poll_jiffies(disk
);
1771 queue_delayed_work(system_freezable_power_efficient_wq
,
1774 queue_delayed_work(system_freezable_power_efficient_wq
,
1777 spin_unlock_irqrestore(&ev
->lock
, flags
);
1781 * disk_unblock_events - unblock disk event checking
1782 * @disk: disk to unblock events for
1784 * Undo disk_block_events(). When the block count reaches zero, it
1785 * starts events polling if configured.
1788 * Don't care. Safe to call from irq context.
1790 void disk_unblock_events(struct gendisk
*disk
)
1793 __disk_unblock_events(disk
, false);
1797 * disk_flush_events - schedule immediate event checking and flushing
1798 * @disk: disk to check and flush events for
1799 * @mask: events to flush
1801 * Schedule immediate event checking on @disk if not blocked. Events in
1802 * @mask are scheduled to be cleared from the driver. Note that this
1803 * doesn't clear the events from @disk->ev.
1806 * If @mask is non-zero must be called with bdev->bd_mutex held.
1808 void disk_flush_events(struct gendisk
*disk
, unsigned int mask
)
1810 struct disk_events
*ev
= disk
->ev
;
1815 spin_lock_irq(&ev
->lock
);
1816 ev
->clearing
|= mask
;
1818 mod_delayed_work(system_freezable_power_efficient_wq
,
1820 spin_unlock_irq(&ev
->lock
);
1824 * disk_clear_events - synchronously check, clear and return pending events
1825 * @disk: disk to fetch and clear events from
1826 * @mask: mask of events to be fetched and cleared
1828 * Disk events are synchronously checked and pending events in @mask
1829 * are cleared and returned. This ignores the block count.
1834 static unsigned int disk_clear_events(struct gendisk
*disk
, unsigned int mask
)
1836 struct disk_events
*ev
= disk
->ev
;
1837 unsigned int pending
;
1838 unsigned int clearing
= mask
;
1843 disk_block_events(disk
);
1846 * store the union of mask and ev->clearing on the stack so that the
1847 * race with disk_flush_events does not cause ambiguity (ev->clearing
1848 * can still be modified even if events are blocked).
1850 spin_lock_irq(&ev
->lock
);
1851 clearing
|= ev
->clearing
;
1853 spin_unlock_irq(&ev
->lock
);
1855 disk_check_events(ev
, &clearing
);
1857 * if ev->clearing is not 0, the disk_flush_events got called in the
1858 * middle of this function, so we want to run the workfn without delay.
1860 __disk_unblock_events(disk
, ev
->clearing
? true : false);
1862 /* then, fetch and clear pending events */
1863 spin_lock_irq(&ev
->lock
);
1864 pending
= ev
->pending
& mask
;
1865 ev
->pending
&= ~mask
;
1866 spin_unlock_irq(&ev
->lock
);
1867 WARN_ON_ONCE(clearing
& mask
);
1873 * bdev_check_media_change - check if a removable media has been changed
1874 * @bdev: block device to check
1876 * Check whether a removable media has been changed, and attempt to free all
1877 * dentries and inodes and invalidates all block device page cache entries in
1880 * Returns %true if the block device changed, or %false if not.
1882 bool bdev_check_media_change(struct block_device
*bdev
)
1884 unsigned int events
;
1886 events
= disk_clear_events(bdev
->bd_disk
, DISK_EVENT_MEDIA_CHANGE
|
1887 DISK_EVENT_EJECT_REQUEST
);
1888 if (!(events
& DISK_EVENT_MEDIA_CHANGE
))
1891 if (__invalidate_device(bdev
, true))
1892 pr_warn("VFS: busy inodes on changed media %s\n",
1893 bdev
->bd_disk
->disk_name
);
1894 set_bit(GD_NEED_PART_SCAN
, &bdev
->bd_disk
->state
);
1897 EXPORT_SYMBOL(bdev_check_media_change
);
1900 * Separate this part out so that a different pointer for clearing_ptr can be
1901 * passed in for disk_clear_events.
1903 static void disk_events_workfn(struct work_struct
*work
)
1905 struct delayed_work
*dwork
= to_delayed_work(work
);
1906 struct disk_events
*ev
= container_of(dwork
, struct disk_events
, dwork
);
1908 disk_check_events(ev
, &ev
->clearing
);
1911 static void disk_check_events(struct disk_events
*ev
,
1912 unsigned int *clearing_ptr
)
1914 struct gendisk
*disk
= ev
->disk
;
1915 char *envp
[ARRAY_SIZE(disk_uevents
) + 1] = { };
1916 unsigned int clearing
= *clearing_ptr
;
1917 unsigned int events
;
1919 int nr_events
= 0, i
;
1922 events
= disk
->fops
->check_events(disk
, clearing
);
1924 /* accumulate pending events and schedule next poll if necessary */
1925 spin_lock_irq(&ev
->lock
);
1927 events
&= ~ev
->pending
;
1928 ev
->pending
|= events
;
1929 *clearing_ptr
&= ~clearing
;
1931 intv
= disk_events_poll_jiffies(disk
);
1932 if (!ev
->block
&& intv
)
1933 queue_delayed_work(system_freezable_power_efficient_wq
,
1936 spin_unlock_irq(&ev
->lock
);
1939 * Tell userland about new events. Only the events listed in
1940 * @disk->events are reported, and only if DISK_EVENT_FLAG_UEVENT
1941 * is set. Otherwise, events are processed internally but never
1942 * get reported to userland.
1944 for (i
= 0; i
< ARRAY_SIZE(disk_uevents
); i
++)
1945 if ((events
& disk
->events
& (1 << i
)) &&
1946 (disk
->event_flags
& DISK_EVENT_FLAG_UEVENT
))
1947 envp
[nr_events
++] = disk_uevents
[i
];
1950 kobject_uevent_env(&disk_to_dev(disk
)->kobj
, KOBJ_CHANGE
, envp
);
1954 * A disk events enabled device has the following sysfs nodes under
1955 * its /sys/block/X/ directory.
1957 * events : list of all supported events
1958 * events_async : list of events which can be detected w/o polling
1959 * (always empty, only for backwards compatibility)
1960 * events_poll_msecs : polling interval, 0: disable, -1: system default
1962 static ssize_t
__disk_events_show(unsigned int events
, char *buf
)
1964 const char *delim
= "";
1968 for (i
= 0; i
< ARRAY_SIZE(disk_events_strs
); i
++)
1969 if (events
& (1 << i
)) {
1970 pos
+= sprintf(buf
+ pos
, "%s%s",
1971 delim
, disk_events_strs
[i
]);
1975 pos
+= sprintf(buf
+ pos
, "\n");
1979 static ssize_t
disk_events_show(struct device
*dev
,
1980 struct device_attribute
*attr
, char *buf
)
1982 struct gendisk
*disk
= dev_to_disk(dev
);
1984 if (!(disk
->event_flags
& DISK_EVENT_FLAG_UEVENT
))
1987 return __disk_events_show(disk
->events
, buf
);
1990 static ssize_t
disk_events_async_show(struct device
*dev
,
1991 struct device_attribute
*attr
, char *buf
)
1996 static ssize_t
disk_events_poll_msecs_show(struct device
*dev
,
1997 struct device_attribute
*attr
,
2000 struct gendisk
*disk
= dev_to_disk(dev
);
2003 return sprintf(buf
, "-1\n");
2005 return sprintf(buf
, "%ld\n", disk
->ev
->poll_msecs
);
2008 static ssize_t
disk_events_poll_msecs_store(struct device
*dev
,
2009 struct device_attribute
*attr
,
2010 const char *buf
, size_t count
)
2012 struct gendisk
*disk
= dev_to_disk(dev
);
2015 if (!count
|| !sscanf(buf
, "%ld", &intv
))
2018 if (intv
< 0 && intv
!= -1)
2024 disk_block_events(disk
);
2025 disk
->ev
->poll_msecs
= intv
;
2026 __disk_unblock_events(disk
, true);
2031 static const DEVICE_ATTR(events
, 0444, disk_events_show
, NULL
);
2032 static const DEVICE_ATTR(events_async
, 0444, disk_events_async_show
, NULL
);
2033 static const DEVICE_ATTR(events_poll_msecs
, 0644,
2034 disk_events_poll_msecs_show
,
2035 disk_events_poll_msecs_store
);
2037 static const struct attribute
*disk_events_attrs
[] = {
2038 &dev_attr_events
.attr
,
2039 &dev_attr_events_async
.attr
,
2040 &dev_attr_events_poll_msecs
.attr
,
2045 * The default polling interval can be specified by the kernel
2046 * parameter block.events_dfl_poll_msecs which defaults to 0
2047 * (disable). This can also be modified runtime by writing to
2048 * /sys/module/block/parameters/events_dfl_poll_msecs.
2050 static int disk_events_set_dfl_poll_msecs(const char *val
,
2051 const struct kernel_param
*kp
)
2053 struct disk_events
*ev
;
2056 ret
= param_set_ulong(val
, kp
);
2060 mutex_lock(&disk_events_mutex
);
2062 list_for_each_entry(ev
, &disk_events
, node
)
2063 disk_flush_events(ev
->disk
, 0);
2065 mutex_unlock(&disk_events_mutex
);
2070 static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops
= {
2071 .set
= disk_events_set_dfl_poll_msecs
,
2072 .get
= param_get_ulong
,
2075 #undef MODULE_PARAM_PREFIX
2076 #define MODULE_PARAM_PREFIX "block."
2078 module_param_cb(events_dfl_poll_msecs
, &disk_events_dfl_poll_msecs_param_ops
,
2079 &disk_events_dfl_poll_msecs
, 0644);
2082 * disk_{alloc|add|del|release}_events - initialize and destroy disk_events.
2084 static void disk_alloc_events(struct gendisk
*disk
)
2086 struct disk_events
*ev
;
2088 if (!disk
->fops
->check_events
|| !disk
->events
)
2091 ev
= kzalloc(sizeof(*ev
), GFP_KERNEL
);
2093 pr_warn("%s: failed to initialize events\n", disk
->disk_name
);
2097 INIT_LIST_HEAD(&ev
->node
);
2099 spin_lock_init(&ev
->lock
);
2100 mutex_init(&ev
->block_mutex
);
2102 ev
->poll_msecs
= -1;
2103 INIT_DELAYED_WORK(&ev
->dwork
, disk_events_workfn
);
2108 static void disk_add_events(struct gendisk
*disk
)
2110 /* FIXME: error handling */
2111 if (sysfs_create_files(&disk_to_dev(disk
)->kobj
, disk_events_attrs
) < 0)
2112 pr_warn("%s: failed to create sysfs files for events\n",
2118 mutex_lock(&disk_events_mutex
);
2119 list_add_tail(&disk
->ev
->node
, &disk_events
);
2120 mutex_unlock(&disk_events_mutex
);
2123 * Block count is initialized to 1 and the following initial
2124 * unblock kicks it into action.
2126 __disk_unblock_events(disk
, true);
2129 static void disk_del_events(struct gendisk
*disk
)
2132 disk_block_events(disk
);
2134 mutex_lock(&disk_events_mutex
);
2135 list_del_init(&disk
->ev
->node
);
2136 mutex_unlock(&disk_events_mutex
);
2139 sysfs_remove_files(&disk_to_dev(disk
)->kobj
, disk_events_attrs
);
2142 static void disk_release_events(struct gendisk
*disk
)
2144 /* the block count should be 1 from disk_del_events() */
2145 WARN_ON_ONCE(disk
->ev
&& disk
->ev
->block
!= 1);