1 // SPDX-License-Identifier: GPL-2.0
6 #include <linux/module.h>
7 #include <linux/ctype.h>
9 #include <linux/genhd.h>
10 #include <linux/kdev_t.h>
11 #include <linux/kernel.h>
12 #include <linux/blkdev.h>
13 #include <linux/backing-dev.h>
14 #include <linux/init.h>
15 #include <linux/spinlock.h>
16 #include <linux/proc_fs.h>
17 #include <linux/seq_file.h>
18 #include <linux/slab.h>
19 #include <linux/kmod.h>
20 #include <linux/mutex.h>
21 #include <linux/idr.h>
22 #include <linux/log2.h>
23 #include <linux/pm_runtime.h>
24 #include <linux/badblocks.h>
28 static struct kobject
*block_depr
;
30 DECLARE_RWSEM(bdev_lookup_sem
);
32 /* for extended dynamic devt allocation, currently only one major is used */
33 #define NR_EXT_DEVT (1 << MINORBITS)
34 static DEFINE_IDA(ext_devt_ida
);
36 static void disk_check_events(struct disk_events
*ev
,
37 unsigned int *clearing_ptr
);
38 static void disk_alloc_events(struct gendisk
*disk
);
39 static void disk_add_events(struct gendisk
*disk
);
40 static void disk_del_events(struct gendisk
*disk
);
41 static void disk_release_events(struct gendisk
*disk
);
43 void set_capacity(struct gendisk
*disk
, sector_t sectors
)
45 struct block_device
*bdev
= disk
->part0
;
47 spin_lock(&bdev
->bd_size_lock
);
48 i_size_write(bdev
->bd_inode
, (loff_t
)sectors
<< SECTOR_SHIFT
);
49 spin_unlock(&bdev
->bd_size_lock
);
51 EXPORT_SYMBOL(set_capacity
);
54 * Set disk capacity and notify if the size is not currently zero and will not
55 * be set to zero. Returns true if a uevent was sent, otherwise false.
57 bool set_capacity_and_notify(struct gendisk
*disk
, sector_t size
)
59 sector_t capacity
= get_capacity(disk
);
60 char *envp
[] = { "RESIZE=1", NULL
};
62 set_capacity(disk
, size
);
65 * Only print a message and send a uevent if the gendisk is user visible
66 * and alive. This avoids spamming the log and udev when setting the
67 * initial capacity during probing.
69 if (size
== capacity
||
70 (disk
->flags
& (GENHD_FL_UP
| GENHD_FL_HIDDEN
)) != GENHD_FL_UP
)
73 pr_info("%s: detected capacity change from %lld to %lld\n",
74 disk
->disk_name
, size
, capacity
);
77 * Historically we did not send a uevent for changes to/from an empty
80 if (!capacity
|| !size
)
82 kobject_uevent_env(&disk_to_dev(disk
)->kobj
, KOBJ_CHANGE
, envp
);
85 EXPORT_SYMBOL_GPL(set_capacity_and_notify
);
88 * Format the device name of the indicated disk into the supplied buffer and
89 * return a pointer to that same buffer for convenience.
91 char *disk_name(struct gendisk
*hd
, int partno
, char *buf
)
94 snprintf(buf
, BDEVNAME_SIZE
, "%s", hd
->disk_name
);
95 else if (isdigit(hd
->disk_name
[strlen(hd
->disk_name
)-1]))
96 snprintf(buf
, BDEVNAME_SIZE
, "%sp%d", hd
->disk_name
, partno
);
98 snprintf(buf
, BDEVNAME_SIZE
, "%s%d", hd
->disk_name
, partno
);
103 const char *bdevname(struct block_device
*bdev
, char *buf
)
105 return disk_name(bdev
->bd_disk
, bdev
->bd_partno
, buf
);
107 EXPORT_SYMBOL(bdevname
);
109 static void part_stat_read_all(struct block_device
*part
,
110 struct disk_stats
*stat
)
114 memset(stat
, 0, sizeof(struct disk_stats
));
115 for_each_possible_cpu(cpu
) {
116 struct disk_stats
*ptr
= per_cpu_ptr(part
->bd_stats
, cpu
);
119 for (group
= 0; group
< NR_STAT_GROUPS
; group
++) {
120 stat
->nsecs
[group
] += ptr
->nsecs
[group
];
121 stat
->sectors
[group
] += ptr
->sectors
[group
];
122 stat
->ios
[group
] += ptr
->ios
[group
];
123 stat
->merges
[group
] += ptr
->merges
[group
];
126 stat
->io_ticks
+= ptr
->io_ticks
;
130 static unsigned int part_in_flight(struct block_device
*part
)
132 unsigned int inflight
= 0;
135 for_each_possible_cpu(cpu
) {
136 inflight
+= part_stat_local_read_cpu(part
, in_flight
[0], cpu
) +
137 part_stat_local_read_cpu(part
, in_flight
[1], cpu
);
139 if ((int)inflight
< 0)
145 static void part_in_flight_rw(struct block_device
*part
,
146 unsigned int inflight
[2])
152 for_each_possible_cpu(cpu
) {
153 inflight
[0] += part_stat_local_read_cpu(part
, in_flight
[0], cpu
);
154 inflight
[1] += part_stat_local_read_cpu(part
, in_flight
[1], cpu
);
156 if ((int)inflight
[0] < 0)
158 if ((int)inflight
[1] < 0)
162 struct block_device
*__disk_get_part(struct gendisk
*disk
, int partno
)
164 struct disk_part_tbl
*ptbl
= rcu_dereference(disk
->part_tbl
);
166 if (unlikely(partno
< 0 || partno
>= ptbl
->len
))
168 return rcu_dereference(ptbl
->part
[partno
]);
172 * disk_part_iter_init - initialize partition iterator
173 * @piter: iterator to initialize
174 * @disk: disk to iterate over
175 * @flags: DISK_PITER_* flags
177 * Initialize @piter so that it iterates over partitions of @disk.
182 void disk_part_iter_init(struct disk_part_iter
*piter
, struct gendisk
*disk
,
185 struct disk_part_tbl
*ptbl
;
188 ptbl
= rcu_dereference(disk
->part_tbl
);
193 if (flags
& DISK_PITER_REVERSE
)
194 piter
->idx
= ptbl
->len
- 1;
195 else if (flags
& (DISK_PITER_INCL_PART0
| DISK_PITER_INCL_EMPTY_PART0
))
200 piter
->flags
= flags
;
204 EXPORT_SYMBOL_GPL(disk_part_iter_init
);
207 * disk_part_iter_next - proceed iterator to the next partition and return it
208 * @piter: iterator of interest
210 * Proceed @piter to the next partition and return it.
215 struct block_device
*disk_part_iter_next(struct disk_part_iter
*piter
)
217 struct disk_part_tbl
*ptbl
;
220 /* put the last partition */
221 disk_part_iter_exit(piter
);
225 ptbl
= rcu_dereference(piter
->disk
->part_tbl
);
227 /* determine iteration parameters */
228 if (piter
->flags
& DISK_PITER_REVERSE
) {
230 if (piter
->flags
& (DISK_PITER_INCL_PART0
|
231 DISK_PITER_INCL_EMPTY_PART0
))
240 /* iterate to the next partition */
241 for (; piter
->idx
!= end
; piter
->idx
+= inc
) {
242 struct block_device
*part
;
244 part
= rcu_dereference(ptbl
->part
[piter
->idx
]);
247 if (!bdev_nr_sectors(part
) &&
248 !(piter
->flags
& DISK_PITER_INCL_EMPTY
) &&
249 !(piter
->flags
& DISK_PITER_INCL_EMPTY_PART0
&&
253 piter
->part
= bdgrab(part
);
264 EXPORT_SYMBOL_GPL(disk_part_iter_next
);
267 * disk_part_iter_exit - finish up partition iteration
268 * @piter: iter of interest
270 * Called when iteration is over. Cleans up @piter.
275 void disk_part_iter_exit(struct disk_part_iter
*piter
)
281 EXPORT_SYMBOL_GPL(disk_part_iter_exit
);
283 static inline int sector_in_part(struct block_device
*part
, sector_t sector
)
285 return part
->bd_start_sect
<= sector
&&
286 sector
< part
->bd_start_sect
+ bdev_nr_sectors(part
);
290 * disk_map_sector_rcu - map sector to partition
291 * @disk: gendisk of interest
292 * @sector: sector to map
294 * Find out which partition @sector maps to on @disk. This is
295 * primarily used for stats accounting.
301 * Found partition on success, part0 is returned if no partition matches
302 * or the matched partition is being deleted.
304 struct block_device
*disk_map_sector_rcu(struct gendisk
*disk
, sector_t sector
)
306 struct disk_part_tbl
*ptbl
;
307 struct block_device
*part
;
311 ptbl
= rcu_dereference(disk
->part_tbl
);
313 part
= rcu_dereference(ptbl
->last_lookup
);
314 if (part
&& sector_in_part(part
, sector
))
317 for (i
= 1; i
< ptbl
->len
; i
++) {
318 part
= rcu_dereference(ptbl
->part
[i
]);
319 if (part
&& sector_in_part(part
, sector
)) {
320 rcu_assign_pointer(ptbl
->last_lookup
, part
);
332 * disk_has_partitions
333 * @disk: gendisk of interest
335 * Walk through the partition table and check if valid partition exists.
341 * True if the gendisk has at least one valid non-zero size partition.
344 bool disk_has_partitions(struct gendisk
*disk
)
346 struct disk_part_tbl
*ptbl
;
351 ptbl
= rcu_dereference(disk
->part_tbl
);
353 /* Iterate partitions skipping the whole device at index 0 */
354 for (i
= 1; i
< ptbl
->len
; i
++) {
355 if (rcu_dereference(ptbl
->part
[i
])) {
365 EXPORT_SYMBOL_GPL(disk_has_partitions
);
368 * Can be deleted altogether. Later.
371 #define BLKDEV_MAJOR_HASH_SIZE 255
372 static struct blk_major_name
{
373 struct blk_major_name
*next
;
376 void (*probe
)(dev_t devt
);
377 } *major_names
[BLKDEV_MAJOR_HASH_SIZE
];
378 static DEFINE_MUTEX(major_names_lock
);
380 /* index in the above - for now: assume no multimajor ranges */
381 static inline int major_to_index(unsigned major
)
383 return major
% BLKDEV_MAJOR_HASH_SIZE
;
386 #ifdef CONFIG_PROC_FS
387 void blkdev_show(struct seq_file
*seqf
, off_t offset
)
389 struct blk_major_name
*dp
;
391 mutex_lock(&major_names_lock
);
392 for (dp
= major_names
[major_to_index(offset
)]; dp
; dp
= dp
->next
)
393 if (dp
->major
== offset
)
394 seq_printf(seqf
, "%3d %s\n", dp
->major
, dp
->name
);
395 mutex_unlock(&major_names_lock
);
397 #endif /* CONFIG_PROC_FS */
400 * __register_blkdev - register a new block device
402 * @major: the requested major device number [1..BLKDEV_MAJOR_MAX-1]. If
403 * @major = 0, try to allocate any unused major number.
404 * @name: the name of the new block device as a zero terminated string
405 * @probe: allback that is called on access to any minor number of @major
407 * The @name must be unique within the system.
409 * The return value depends on the @major input parameter:
411 * - if a major device number was requested in range [1..BLKDEV_MAJOR_MAX-1]
412 * then the function returns zero on success, or a negative error code
413 * - if any unused major number was requested with @major = 0 parameter
414 * then the return value is the allocated major number in range
415 * [1..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise
417 * See Documentation/admin-guide/devices.txt for the list of allocated
420 * Use register_blkdev instead for any new code.
422 int __register_blkdev(unsigned int major
, const char *name
,
423 void (*probe
)(dev_t devt
))
425 struct blk_major_name
**n
, *p
;
428 mutex_lock(&major_names_lock
);
432 for (index
= ARRAY_SIZE(major_names
)-1; index
> 0; index
--) {
433 if (major_names
[index
] == NULL
)
438 printk("%s: failed to get major for %s\n",
447 if (major
>= BLKDEV_MAJOR_MAX
) {
448 pr_err("%s: major requested (%u) is greater than the maximum (%u) for %s\n",
449 __func__
, major
, BLKDEV_MAJOR_MAX
-1, name
);
455 p
= kmalloc(sizeof(struct blk_major_name
), GFP_KERNEL
);
463 strlcpy(p
->name
, name
, sizeof(p
->name
));
465 index
= major_to_index(major
);
467 for (n
= &major_names
[index
]; *n
; n
= &(*n
)->next
) {
468 if ((*n
)->major
== major
)
477 printk("register_blkdev: cannot get major %u for %s\n",
482 mutex_unlock(&major_names_lock
);
485 EXPORT_SYMBOL(__register_blkdev
);
487 void unregister_blkdev(unsigned int major
, const char *name
)
489 struct blk_major_name
**n
;
490 struct blk_major_name
*p
= NULL
;
491 int index
= major_to_index(major
);
493 mutex_lock(&major_names_lock
);
494 for (n
= &major_names
[index
]; *n
; n
= &(*n
)->next
)
495 if ((*n
)->major
== major
)
497 if (!*n
|| strcmp((*n
)->name
, name
)) {
503 mutex_unlock(&major_names_lock
);
507 EXPORT_SYMBOL(unregister_blkdev
);
510 * blk_mangle_minor - scatter minor numbers apart
511 * @minor: minor number to mangle
513 * Scatter consecutively allocated @minor number apart if MANGLE_DEVT
514 * is enabled. Mangling twice gives the original value.
522 static int blk_mangle_minor(int minor
)
524 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
527 for (i
= 0; i
< MINORBITS
/ 2; i
++) {
528 int low
= minor
& (1 << i
);
529 int high
= minor
& (1 << (MINORBITS
- 1 - i
));
530 int distance
= MINORBITS
- 1 - 2 * i
;
532 minor
^= low
| high
; /* clear both bits */
533 low
<<= distance
; /* swap the positions */
535 minor
|= low
| high
; /* and set */
542 * blk_alloc_devt - allocate a dev_t for a block device
543 * @bdev: block device to allocate dev_t for
544 * @devt: out parameter for resulting dev_t
546 * Allocate a dev_t for block device.
549 * 0 on success, allocated dev_t is returned in *@devt. -errno on
555 int blk_alloc_devt(struct block_device
*bdev
, dev_t
*devt
)
557 struct gendisk
*disk
= bdev
->bd_disk
;
560 /* in consecutive minor range? */
561 if (bdev
->bd_partno
< disk
->minors
) {
562 *devt
= MKDEV(disk
->major
, disk
->first_minor
+ bdev
->bd_partno
);
566 idx
= ida_alloc_range(&ext_devt_ida
, 0, NR_EXT_DEVT
, GFP_KERNEL
);
568 return idx
== -ENOSPC
? -EBUSY
: idx
;
570 *devt
= MKDEV(BLOCK_EXT_MAJOR
, blk_mangle_minor(idx
));
575 * blk_free_devt - free a dev_t
576 * @devt: dev_t to free
578 * Free @devt which was allocated using blk_alloc_devt().
583 void blk_free_devt(dev_t devt
)
585 if (MAJOR(devt
) == BLOCK_EXT_MAJOR
)
586 ida_free(&ext_devt_ida
, blk_mangle_minor(MINOR(devt
)));
589 static char *bdevt_str(dev_t devt
, char *buf
)
591 if (MAJOR(devt
) <= 0xff && MINOR(devt
) <= 0xff) {
592 char tbuf
[BDEVT_SIZE
];
593 snprintf(tbuf
, BDEVT_SIZE
, "%02x%02x", MAJOR(devt
), MINOR(devt
));
594 snprintf(buf
, BDEVT_SIZE
, "%-9s", tbuf
);
596 snprintf(buf
, BDEVT_SIZE
, "%03x:%05x", MAJOR(devt
), MINOR(devt
));
601 static void disk_scan_partitions(struct gendisk
*disk
)
603 struct block_device
*bdev
;
605 if (!get_capacity(disk
) || !disk_part_scan_enabled(disk
))
608 set_bit(GD_NEED_PART_SCAN
, &disk
->state
);
609 bdev
= blkdev_get_by_dev(disk_devt(disk
), FMODE_READ
, NULL
);
611 blkdev_put(bdev
, FMODE_READ
);
614 static void register_disk(struct device
*parent
, struct gendisk
*disk
,
615 const struct attribute_group
**groups
)
617 struct device
*ddev
= disk_to_dev(disk
);
618 struct disk_part_iter piter
;
619 struct block_device
*part
;
622 ddev
->parent
= parent
;
624 dev_set_name(ddev
, "%s", disk
->disk_name
);
626 /* delay uevents, until we scanned partition table */
627 dev_set_uevent_suppress(ddev
, 1);
630 WARN_ON(ddev
->groups
);
631 ddev
->groups
= groups
;
633 if (device_add(ddev
))
635 if (!sysfs_deprecated
) {
636 err
= sysfs_create_link(block_depr
, &ddev
->kobj
,
637 kobject_name(&ddev
->kobj
));
645 * avoid probable deadlock caused by allocating memory with
646 * GFP_KERNEL in runtime_resume callback of its all ancestor
649 pm_runtime_set_memalloc_noio(ddev
, true);
651 disk
->part0
->bd_holder_dir
=
652 kobject_create_and_add("holders", &ddev
->kobj
);
653 disk
->slave_dir
= kobject_create_and_add("slaves", &ddev
->kobj
);
655 if (disk
->flags
& GENHD_FL_HIDDEN
) {
656 dev_set_uevent_suppress(ddev
, 0);
660 disk_scan_partitions(disk
);
662 /* announce disk after possible partitions are created */
663 dev_set_uevent_suppress(ddev
, 0);
664 kobject_uevent(&ddev
->kobj
, KOBJ_ADD
);
666 /* announce possible partitions */
667 disk_part_iter_init(&piter
, disk
, 0);
668 while ((part
= disk_part_iter_next(&piter
)))
669 kobject_uevent(bdev_kobj(part
), KOBJ_ADD
);
670 disk_part_iter_exit(&piter
);
672 if (disk
->queue
->backing_dev_info
->dev
) {
673 err
= sysfs_create_link(&ddev
->kobj
,
674 &disk
->queue
->backing_dev_info
->dev
->kobj
,
681 * __device_add_disk - add disk information to kernel list
682 * @parent: parent device for the disk
683 * @disk: per-device partitioning information
684 * @groups: Additional per-device sysfs groups
685 * @register_queue: register the queue if set to true
687 * This function registers the partitioning information in @disk
690 * FIXME: error handling
692 static void __device_add_disk(struct device
*parent
, struct gendisk
*disk
,
693 const struct attribute_group
**groups
,
700 * The disk queue should now be all set with enough information about
701 * the device for the elevator code to pick an adequate default
702 * elevator if one is needed, that is, for devices requesting queue
706 elevator_init_mq(disk
->queue
);
708 /* minors == 0 indicates to use ext devt from part0 and should
709 * be accompanied with EXT_DEVT flag. Make sure all
710 * parameters make sense.
712 WARN_ON(disk
->minors
&& !(disk
->major
|| disk
->first_minor
));
713 WARN_ON(!disk
->minors
&&
714 !(disk
->flags
& (GENHD_FL_EXT_DEVT
| GENHD_FL_HIDDEN
)));
716 disk
->flags
|= GENHD_FL_UP
;
718 retval
= blk_alloc_devt(disk
->part0
, &devt
);
723 disk
->major
= MAJOR(devt
);
724 disk
->first_minor
= MINOR(devt
);
726 disk_alloc_events(disk
);
728 if (disk
->flags
& GENHD_FL_HIDDEN
) {
730 * Don't let hidden disks show up in /proc/partitions,
731 * and don't bother scanning for partitions either.
733 disk
->flags
|= GENHD_FL_SUPPRESS_PARTITION_INFO
;
734 disk
->flags
|= GENHD_FL_NO_PART_SCAN
;
736 struct backing_dev_info
*bdi
= disk
->queue
->backing_dev_info
;
737 struct device
*dev
= disk_to_dev(disk
);
740 /* Register BDI before referencing it from bdev */
742 ret
= bdi_register(bdi
, "%u:%u", MAJOR(devt
), MINOR(devt
));
744 bdi_set_owner(bdi
, dev
);
745 bdev_add(disk
->part0
, devt
);
747 register_disk(parent
, disk
, groups
);
749 blk_register_queue(disk
);
752 * Take an extra ref on queue which will be put on disk_release()
753 * so that it sticks around as long as @disk is there.
755 WARN_ON_ONCE(!blk_get_queue(disk
->queue
));
757 disk_add_events(disk
);
758 blk_integrity_add(disk
);
761 void device_add_disk(struct device
*parent
, struct gendisk
*disk
,
762 const struct attribute_group
**groups
)
765 __device_add_disk(parent
, disk
, groups
, true);
767 EXPORT_SYMBOL(device_add_disk
);
769 void device_add_disk_no_queue_reg(struct device
*parent
, struct gendisk
*disk
)
771 __device_add_disk(parent
, disk
, NULL
, false);
773 EXPORT_SYMBOL(device_add_disk_no_queue_reg
);
775 static void invalidate_partition(struct block_device
*bdev
)
778 __invalidate_device(bdev
, true);
781 * Unhash the bdev inode for this device so that it can't be looked
782 * up any more even if openers still hold references to it.
784 remove_inode_hash(bdev
->bd_inode
);
788 * del_gendisk - remove the gendisk
789 * @disk: the struct gendisk to remove
791 * Removes the gendisk and all its associated resources. This deletes the
792 * partitions associated with the gendisk, and unregisters the associated
795 * This is the counter to the respective __device_add_disk() call.
797 * The final removal of the struct gendisk happens when its refcount reaches 0
798 * with put_disk(), which should be called after del_gendisk(), if
799 * __device_add_disk() was used.
801 * Drivers exist which depend on the release of the gendisk to be synchronous,
802 * it should not be deferred.
806 void del_gendisk(struct gendisk
*disk
)
808 struct disk_part_iter piter
;
809 struct block_device
*part
;
813 if (WARN_ON_ONCE(!disk
->queue
))
816 blk_integrity_del(disk
);
817 disk_del_events(disk
);
820 * Block lookups of the disk until all bdevs are unhashed and the
821 * disk is marked as dead (GENHD_FL_UP cleared).
823 down_write(&bdev_lookup_sem
);
825 /* invalidate stuff */
826 disk_part_iter_init(&piter
, disk
,
827 DISK_PITER_INCL_EMPTY
| DISK_PITER_REVERSE
);
828 while ((part
= disk_part_iter_next(&piter
))) {
829 invalidate_partition(part
);
830 delete_partition(part
);
832 disk_part_iter_exit(&piter
);
834 invalidate_partition(disk
->part0
);
835 set_capacity(disk
, 0);
836 disk
->flags
&= ~GENHD_FL_UP
;
837 up_write(&bdev_lookup_sem
);
839 if (!(disk
->flags
& GENHD_FL_HIDDEN
)) {
840 sysfs_remove_link(&disk_to_dev(disk
)->kobj
, "bdi");
843 * Unregister bdi before releasing device numbers (as they can
844 * get reused and we'd get clashes in sysfs).
846 bdi_unregister(disk
->queue
->backing_dev_info
);
849 blk_unregister_queue(disk
);
851 kobject_put(disk
->part0
->bd_holder_dir
);
852 kobject_put(disk
->slave_dir
);
854 part_stat_set_all(disk
->part0
, 0);
855 disk
->part0
->bd_stamp
= 0;
856 if (!sysfs_deprecated
)
857 sysfs_remove_link(block_depr
, dev_name(disk_to_dev(disk
)));
858 pm_runtime_set_memalloc_noio(disk_to_dev(disk
), false);
859 device_del(disk_to_dev(disk
));
861 EXPORT_SYMBOL(del_gendisk
);
863 /* sysfs access to bad-blocks list. */
864 static ssize_t
disk_badblocks_show(struct device
*dev
,
865 struct device_attribute
*attr
,
868 struct gendisk
*disk
= dev_to_disk(dev
);
871 return sprintf(page
, "\n");
873 return badblocks_show(disk
->bb
, page
, 0);
876 static ssize_t
disk_badblocks_store(struct device
*dev
,
877 struct device_attribute
*attr
,
878 const char *page
, size_t len
)
880 struct gendisk
*disk
= dev_to_disk(dev
);
885 return badblocks_store(disk
->bb
, page
, len
, 0);
888 void blk_request_module(dev_t devt
)
890 unsigned int major
= MAJOR(devt
);
891 struct blk_major_name
**n
;
893 mutex_lock(&major_names_lock
);
894 for (n
= &major_names
[major_to_index(major
)]; *n
; n
= &(*n
)->next
) {
895 if ((*n
)->major
== major
&& (*n
)->probe
) {
897 mutex_unlock(&major_names_lock
);
901 mutex_unlock(&major_names_lock
);
903 if (request_module("block-major-%d-%d", MAJOR(devt
), MINOR(devt
)) > 0)
904 /* Make old-style 2.4 aliases work */
905 request_module("block-major-%d", MAJOR(devt
));
909 * bdget_disk - do bdget() by gendisk and partition number
910 * @disk: gendisk of interest
911 * @partno: partition number
913 * Find partition @partno from @disk, do bdget() on it.
919 * Resulting block_device on success, NULL on failure.
921 struct block_device
*bdget_disk(struct gendisk
*disk
, int partno
)
923 struct block_device
*bdev
= NULL
;
926 bdev
= __disk_get_part(disk
, partno
);
927 if (bdev
&& !bdgrab(bdev
))
935 * print a full list of all partitions - intended for places where the root
936 * filesystem can't be mounted and thus to give the victim some idea of what
939 void __init
printk_all_partitions(void)
941 struct class_dev_iter iter
;
944 class_dev_iter_init(&iter
, &block_class
, NULL
, &disk_type
);
945 while ((dev
= class_dev_iter_next(&iter
))) {
946 struct gendisk
*disk
= dev_to_disk(dev
);
947 struct disk_part_iter piter
;
948 struct block_device
*part
;
949 char name_buf
[BDEVNAME_SIZE
];
950 char devt_buf
[BDEVT_SIZE
];
953 * Don't show empty devices or things that have been
956 if (get_capacity(disk
) == 0 ||
957 (disk
->flags
& GENHD_FL_SUPPRESS_PARTITION_INFO
))
961 * Note, unlike /proc/partitions, I am showing the
962 * numbers in hex - the same format as the root=
965 disk_part_iter_init(&piter
, disk
, DISK_PITER_INCL_PART0
);
966 while ((part
= disk_part_iter_next(&piter
))) {
967 bool is_part0
= part
== disk
->part0
;
969 printk("%s%s %10llu %s %s", is_part0
? "" : " ",
970 bdevt_str(part
->bd_dev
, devt_buf
),
971 bdev_nr_sectors(part
) >> 1,
972 disk_name(disk
, part
->bd_partno
, name_buf
),
974 part
->bd_meta_info
->uuid
: "");
976 if (dev
->parent
&& dev
->parent
->driver
)
977 printk(" driver: %s\n",
978 dev
->parent
->driver
->name
);
980 printk(" (driver?)\n");
984 disk_part_iter_exit(&piter
);
986 class_dev_iter_exit(&iter
);
989 #ifdef CONFIG_PROC_FS
991 static void *disk_seqf_start(struct seq_file
*seqf
, loff_t
*pos
)
994 struct class_dev_iter
*iter
;
997 iter
= kmalloc(sizeof(*iter
), GFP_KERNEL
);
999 return ERR_PTR(-ENOMEM
);
1001 seqf
->private = iter
;
1002 class_dev_iter_init(iter
, &block_class
, NULL
, &disk_type
);
1004 dev
= class_dev_iter_next(iter
);
1009 return dev_to_disk(dev
);
1012 static void *disk_seqf_next(struct seq_file
*seqf
, void *v
, loff_t
*pos
)
1017 dev
= class_dev_iter_next(seqf
->private);
1019 return dev_to_disk(dev
);
1024 static void disk_seqf_stop(struct seq_file
*seqf
, void *v
)
1026 struct class_dev_iter
*iter
= seqf
->private;
1028 /* stop is called even after start failed :-( */
1030 class_dev_iter_exit(iter
);
1032 seqf
->private = NULL
;
1036 static void *show_partition_start(struct seq_file
*seqf
, loff_t
*pos
)
1040 p
= disk_seqf_start(seqf
, pos
);
1041 if (!IS_ERR_OR_NULL(p
) && !*pos
)
1042 seq_puts(seqf
, "major minor #blocks name\n\n");
1046 static int show_partition(struct seq_file
*seqf
, void *v
)
1048 struct gendisk
*sgp
= v
;
1049 struct disk_part_iter piter
;
1050 struct block_device
*part
;
1051 char buf
[BDEVNAME_SIZE
];
1053 /* Don't show non-partitionable removeable devices or empty devices */
1054 if (!get_capacity(sgp
) || (!disk_max_parts(sgp
) &&
1055 (sgp
->flags
& GENHD_FL_REMOVABLE
)))
1057 if (sgp
->flags
& GENHD_FL_SUPPRESS_PARTITION_INFO
)
1060 /* show the full disk and all non-0 size partitions of it */
1061 disk_part_iter_init(&piter
, sgp
, DISK_PITER_INCL_PART0
);
1062 while ((part
= disk_part_iter_next(&piter
)))
1063 seq_printf(seqf
, "%4d %7d %10llu %s\n",
1064 MAJOR(part
->bd_dev
), MINOR(part
->bd_dev
),
1065 bdev_nr_sectors(part
) >> 1,
1066 disk_name(sgp
, part
->bd_partno
, buf
));
1067 disk_part_iter_exit(&piter
);
1072 static const struct seq_operations partitions_op
= {
1073 .start
= show_partition_start
,
1074 .next
= disk_seqf_next
,
1075 .stop
= disk_seqf_stop
,
1076 .show
= show_partition
1080 static int __init
genhd_device_init(void)
1084 block_class
.dev_kobj
= sysfs_dev_block_kobj
;
1085 error
= class_register(&block_class
);
1086 if (unlikely(error
))
1090 register_blkdev(BLOCK_EXT_MAJOR
, "blkext");
1092 /* create top-level block dir */
1093 if (!sysfs_deprecated
)
1094 block_depr
= kobject_create_and_add("block", NULL
);
1098 subsys_initcall(genhd_device_init
);
1100 static ssize_t
disk_range_show(struct device
*dev
,
1101 struct device_attribute
*attr
, char *buf
)
1103 struct gendisk
*disk
= dev_to_disk(dev
);
1105 return sprintf(buf
, "%d\n", disk
->minors
);
1108 static ssize_t
disk_ext_range_show(struct device
*dev
,
1109 struct device_attribute
*attr
, char *buf
)
1111 struct gendisk
*disk
= dev_to_disk(dev
);
1113 return sprintf(buf
, "%d\n", disk_max_parts(disk
));
1116 static ssize_t
disk_removable_show(struct device
*dev
,
1117 struct device_attribute
*attr
, char *buf
)
1119 struct gendisk
*disk
= dev_to_disk(dev
);
1121 return sprintf(buf
, "%d\n",
1122 (disk
->flags
& GENHD_FL_REMOVABLE
? 1 : 0));
1125 static ssize_t
disk_hidden_show(struct device
*dev
,
1126 struct device_attribute
*attr
, char *buf
)
1128 struct gendisk
*disk
= dev_to_disk(dev
);
1130 return sprintf(buf
, "%d\n",
1131 (disk
->flags
& GENHD_FL_HIDDEN
? 1 : 0));
1134 static ssize_t
disk_ro_show(struct device
*dev
,
1135 struct device_attribute
*attr
, char *buf
)
1137 struct gendisk
*disk
= dev_to_disk(dev
);
1139 return sprintf(buf
, "%d\n", get_disk_ro(disk
) ? 1 : 0);
1142 ssize_t
part_size_show(struct device
*dev
,
1143 struct device_attribute
*attr
, char *buf
)
1145 return sprintf(buf
, "%llu\n", bdev_nr_sectors(dev_to_bdev(dev
)));
1148 ssize_t
part_stat_show(struct device
*dev
,
1149 struct device_attribute
*attr
, char *buf
)
1151 struct block_device
*bdev
= dev_to_bdev(dev
);
1152 struct request_queue
*q
= bdev
->bd_disk
->queue
;
1153 struct disk_stats stat
;
1154 unsigned int inflight
;
1156 part_stat_read_all(bdev
, &stat
);
1158 inflight
= blk_mq_in_flight(q
, bdev
);
1160 inflight
= part_in_flight(bdev
);
1163 "%8lu %8lu %8llu %8u "
1164 "%8lu %8lu %8llu %8u "
1166 "%8lu %8lu %8llu %8u "
1169 stat
.ios
[STAT_READ
],
1170 stat
.merges
[STAT_READ
],
1171 (unsigned long long)stat
.sectors
[STAT_READ
],
1172 (unsigned int)div_u64(stat
.nsecs
[STAT_READ
], NSEC_PER_MSEC
),
1173 stat
.ios
[STAT_WRITE
],
1174 stat
.merges
[STAT_WRITE
],
1175 (unsigned long long)stat
.sectors
[STAT_WRITE
],
1176 (unsigned int)div_u64(stat
.nsecs
[STAT_WRITE
], NSEC_PER_MSEC
),
1178 jiffies_to_msecs(stat
.io_ticks
),
1179 (unsigned int)div_u64(stat
.nsecs
[STAT_READ
] +
1180 stat
.nsecs
[STAT_WRITE
] +
1181 stat
.nsecs
[STAT_DISCARD
] +
1182 stat
.nsecs
[STAT_FLUSH
],
1184 stat
.ios
[STAT_DISCARD
],
1185 stat
.merges
[STAT_DISCARD
],
1186 (unsigned long long)stat
.sectors
[STAT_DISCARD
],
1187 (unsigned int)div_u64(stat
.nsecs
[STAT_DISCARD
], NSEC_PER_MSEC
),
1188 stat
.ios
[STAT_FLUSH
],
1189 (unsigned int)div_u64(stat
.nsecs
[STAT_FLUSH
], NSEC_PER_MSEC
));
1192 ssize_t
part_inflight_show(struct device
*dev
, struct device_attribute
*attr
,
1195 struct block_device
*bdev
= dev_to_bdev(dev
);
1196 struct request_queue
*q
= bdev
->bd_disk
->queue
;
1197 unsigned int inflight
[2];
1200 blk_mq_in_flight_rw(q
, bdev
, inflight
);
1202 part_in_flight_rw(bdev
, inflight
);
1204 return sprintf(buf
, "%8u %8u\n", inflight
[0], inflight
[1]);
1207 static ssize_t
disk_capability_show(struct device
*dev
,
1208 struct device_attribute
*attr
, char *buf
)
1210 struct gendisk
*disk
= dev_to_disk(dev
);
1212 return sprintf(buf
, "%x\n", disk
->flags
);
1215 static ssize_t
disk_alignment_offset_show(struct device
*dev
,
1216 struct device_attribute
*attr
,
1219 struct gendisk
*disk
= dev_to_disk(dev
);
1221 return sprintf(buf
, "%d\n", queue_alignment_offset(disk
->queue
));
1224 static ssize_t
disk_discard_alignment_show(struct device
*dev
,
1225 struct device_attribute
*attr
,
1228 struct gendisk
*disk
= dev_to_disk(dev
);
1230 return sprintf(buf
, "%d\n", queue_discard_alignment(disk
->queue
));
1233 static DEVICE_ATTR(range
, 0444, disk_range_show
, NULL
);
1234 static DEVICE_ATTR(ext_range
, 0444, disk_ext_range_show
, NULL
);
1235 static DEVICE_ATTR(removable
, 0444, disk_removable_show
, NULL
);
1236 static DEVICE_ATTR(hidden
, 0444, disk_hidden_show
, NULL
);
1237 static DEVICE_ATTR(ro
, 0444, disk_ro_show
, NULL
);
1238 static DEVICE_ATTR(size
, 0444, part_size_show
, NULL
);
1239 static DEVICE_ATTR(alignment_offset
, 0444, disk_alignment_offset_show
, NULL
);
1240 static DEVICE_ATTR(discard_alignment
, 0444, disk_discard_alignment_show
, NULL
);
1241 static DEVICE_ATTR(capability
, 0444, disk_capability_show
, NULL
);
1242 static DEVICE_ATTR(stat
, 0444, part_stat_show
, NULL
);
1243 static DEVICE_ATTR(inflight
, 0444, part_inflight_show
, NULL
);
1244 static DEVICE_ATTR(badblocks
, 0644, disk_badblocks_show
, disk_badblocks_store
);
1246 #ifdef CONFIG_FAIL_MAKE_REQUEST
1247 ssize_t
part_fail_show(struct device
*dev
,
1248 struct device_attribute
*attr
, char *buf
)
1250 return sprintf(buf
, "%d\n", dev_to_bdev(dev
)->bd_make_it_fail
);
1253 ssize_t
part_fail_store(struct device
*dev
,
1254 struct device_attribute
*attr
,
1255 const char *buf
, size_t count
)
1259 if (count
> 0 && sscanf(buf
, "%d", &i
) > 0)
1260 dev_to_bdev(dev
)->bd_make_it_fail
= i
;
1265 static struct device_attribute dev_attr_fail
=
1266 __ATTR(make
-it
-fail
, 0644, part_fail_show
, part_fail_store
);
1267 #endif /* CONFIG_FAIL_MAKE_REQUEST */
1269 #ifdef CONFIG_FAIL_IO_TIMEOUT
1270 static struct device_attribute dev_attr_fail_timeout
=
1271 __ATTR(io
-timeout
-fail
, 0644, part_timeout_show
, part_timeout_store
);
1274 static struct attribute
*disk_attrs
[] = {
1275 &dev_attr_range
.attr
,
1276 &dev_attr_ext_range
.attr
,
1277 &dev_attr_removable
.attr
,
1278 &dev_attr_hidden
.attr
,
1280 &dev_attr_size
.attr
,
1281 &dev_attr_alignment_offset
.attr
,
1282 &dev_attr_discard_alignment
.attr
,
1283 &dev_attr_capability
.attr
,
1284 &dev_attr_stat
.attr
,
1285 &dev_attr_inflight
.attr
,
1286 &dev_attr_badblocks
.attr
,
1287 #ifdef CONFIG_FAIL_MAKE_REQUEST
1288 &dev_attr_fail
.attr
,
1290 #ifdef CONFIG_FAIL_IO_TIMEOUT
1291 &dev_attr_fail_timeout
.attr
,
1296 static umode_t
disk_visible(struct kobject
*kobj
, struct attribute
*a
, int n
)
1298 struct device
*dev
= container_of(kobj
, typeof(*dev
), kobj
);
1299 struct gendisk
*disk
= dev_to_disk(dev
);
1301 if (a
== &dev_attr_badblocks
.attr
&& !disk
->bb
)
1306 static struct attribute_group disk_attr_group
= {
1307 .attrs
= disk_attrs
,
1308 .is_visible
= disk_visible
,
1311 static const struct attribute_group
*disk_attr_groups
[] = {
1317 * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way
1318 * @disk: disk to replace part_tbl for
1319 * @new_ptbl: new part_tbl to install
1321 * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The
1322 * original ptbl is freed using RCU callback.
1325 * Matching bd_mutex locked or the caller is the only user of @disk.
1327 static void disk_replace_part_tbl(struct gendisk
*disk
,
1328 struct disk_part_tbl
*new_ptbl
)
1330 struct disk_part_tbl
*old_ptbl
=
1331 rcu_dereference_protected(disk
->part_tbl
, 1);
1333 rcu_assign_pointer(disk
->part_tbl
, new_ptbl
);
1336 rcu_assign_pointer(old_ptbl
->last_lookup
, NULL
);
1337 kfree_rcu(old_ptbl
, rcu_head
);
1342 * disk_expand_part_tbl - expand disk->part_tbl
1343 * @disk: disk to expand part_tbl for
1344 * @partno: expand such that this partno can fit in
1346 * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl
1347 * uses RCU to allow unlocked dereferencing for stats and other stuff.
1350 * Matching bd_mutex locked or the caller is the only user of @disk.
1354 * 0 on success, -errno on failure.
1356 int disk_expand_part_tbl(struct gendisk
*disk
, int partno
)
1358 struct disk_part_tbl
*old_ptbl
=
1359 rcu_dereference_protected(disk
->part_tbl
, 1);
1360 struct disk_part_tbl
*new_ptbl
;
1361 int len
= old_ptbl
? old_ptbl
->len
: 0;
1365 * check for int overflow, since we can get here from blkpg_ioctl()
1366 * with a user passed 'partno'.
1368 target
= partno
+ 1;
1372 /* disk_max_parts() is zero during initialization, ignore if so */
1373 if (disk_max_parts(disk
) && target
> disk_max_parts(disk
))
1379 new_ptbl
= kzalloc_node(struct_size(new_ptbl
, part
, target
), GFP_KERNEL
,
1384 new_ptbl
->len
= target
;
1386 for (i
= 0; i
< len
; i
++)
1387 rcu_assign_pointer(new_ptbl
->part
[i
], old_ptbl
->part
[i
]);
1389 disk_replace_part_tbl(disk
, new_ptbl
);
1394 * disk_release - releases all allocated resources of the gendisk
1395 * @dev: the device representing this disk
1397 * This function releases all allocated resources of the gendisk.
1399 * Drivers which used __device_add_disk() have a gendisk with a request_queue
1400 * assigned. Since the request_queue sits on top of the gendisk for these
1401 * drivers we also call blk_put_queue() for them, and we expect the
1402 * request_queue refcount to reach 0 at this point, and so the request_queue
1403 * will also be freed prior to the disk.
1405 * Context: can sleep
1407 static void disk_release(struct device
*dev
)
1409 struct gendisk
*disk
= dev_to_disk(dev
);
1413 blk_free_devt(dev
->devt
);
1414 disk_release_events(disk
);
1415 kfree(disk
->random
);
1416 disk_replace_part_tbl(disk
, NULL
);
1419 blk_put_queue(disk
->queue
);
1422 struct class block_class
= {
1426 static char *block_devnode(struct device
*dev
, umode_t
*mode
,
1427 kuid_t
*uid
, kgid_t
*gid
)
1429 struct gendisk
*disk
= dev_to_disk(dev
);
1431 if (disk
->fops
->devnode
)
1432 return disk
->fops
->devnode(disk
, mode
);
1436 const struct device_type disk_type
= {
1438 .groups
= disk_attr_groups
,
1439 .release
= disk_release
,
1440 .devnode
= block_devnode
,
1443 #ifdef CONFIG_PROC_FS
1445 * aggregate disk stat collector. Uses the same stats that the sysfs
1446 * entries do, above, but makes them available through one seq_file.
1448 * The output looks suspiciously like /proc/partitions with a bunch of
1451 static int diskstats_show(struct seq_file
*seqf
, void *v
)
1453 struct gendisk
*gp
= v
;
1454 struct disk_part_iter piter
;
1455 struct block_device
*hd
;
1456 char buf
[BDEVNAME_SIZE
];
1457 unsigned int inflight
;
1458 struct disk_stats stat
;
1461 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1462 seq_puts(seqf, "major minor name"
1463 " rio rmerge rsect ruse wio wmerge "
1464 "wsect wuse running use aveq"
1468 disk_part_iter_init(&piter
, gp
, DISK_PITER_INCL_EMPTY_PART0
);
1469 while ((hd
= disk_part_iter_next(&piter
))) {
1470 part_stat_read_all(hd
, &stat
);
1471 if (queue_is_mq(gp
->queue
))
1472 inflight
= blk_mq_in_flight(gp
->queue
, hd
);
1474 inflight
= part_in_flight(hd
);
1476 seq_printf(seqf
, "%4d %7d %s "
1483 MAJOR(hd
->bd_dev
), MINOR(hd
->bd_dev
),
1484 disk_name(gp
, hd
->bd_partno
, buf
),
1485 stat
.ios
[STAT_READ
],
1486 stat
.merges
[STAT_READ
],
1487 stat
.sectors
[STAT_READ
],
1488 (unsigned int)div_u64(stat
.nsecs
[STAT_READ
],
1490 stat
.ios
[STAT_WRITE
],
1491 stat
.merges
[STAT_WRITE
],
1492 stat
.sectors
[STAT_WRITE
],
1493 (unsigned int)div_u64(stat
.nsecs
[STAT_WRITE
],
1496 jiffies_to_msecs(stat
.io_ticks
),
1497 (unsigned int)div_u64(stat
.nsecs
[STAT_READ
] +
1498 stat
.nsecs
[STAT_WRITE
] +
1499 stat
.nsecs
[STAT_DISCARD
] +
1500 stat
.nsecs
[STAT_FLUSH
],
1502 stat
.ios
[STAT_DISCARD
],
1503 stat
.merges
[STAT_DISCARD
],
1504 stat
.sectors
[STAT_DISCARD
],
1505 (unsigned int)div_u64(stat
.nsecs
[STAT_DISCARD
],
1507 stat
.ios
[STAT_FLUSH
],
1508 (unsigned int)div_u64(stat
.nsecs
[STAT_FLUSH
],
1512 disk_part_iter_exit(&piter
);
1517 static const struct seq_operations diskstats_op
= {
1518 .start
= disk_seqf_start
,
1519 .next
= disk_seqf_next
,
1520 .stop
= disk_seqf_stop
,
1521 .show
= diskstats_show
1524 static int __init
proc_genhd_init(void)
1526 proc_create_seq("diskstats", 0, NULL
, &diskstats_op
);
1527 proc_create_seq("partitions", 0, NULL
, &partitions_op
);
1530 module_init(proc_genhd_init
);
1531 #endif /* CONFIG_PROC_FS */
1533 dev_t
blk_lookup_devt(const char *name
, int partno
)
1535 dev_t devt
= MKDEV(0, 0);
1536 struct class_dev_iter iter
;
1539 class_dev_iter_init(&iter
, &block_class
, NULL
, &disk_type
);
1540 while ((dev
= class_dev_iter_next(&iter
))) {
1541 struct gendisk
*disk
= dev_to_disk(dev
);
1542 struct block_device
*part
;
1544 if (strcmp(dev_name(dev
), name
))
1547 if (partno
< disk
->minors
) {
1548 /* We need to return the right devno, even
1549 * if the partition doesn't exist yet.
1551 devt
= MKDEV(MAJOR(dev
->devt
),
1552 MINOR(dev
->devt
) + partno
);
1555 part
= bdget_disk(disk
, partno
);
1557 devt
= part
->bd_dev
;
1562 class_dev_iter_exit(&iter
);
1566 struct gendisk
*__alloc_disk_node(int minors
, int node_id
)
1568 struct gendisk
*disk
;
1569 struct disk_part_tbl
*ptbl
;
1571 if (minors
> DISK_MAX_PARTS
) {
1573 "block: can't allocate more than %d partitions\n",
1575 minors
= DISK_MAX_PARTS
;
1578 disk
= kzalloc_node(sizeof(struct gendisk
), GFP_KERNEL
, node_id
);
1582 disk
->part0
= bdev_alloc(disk
, 0);
1586 disk
->node_id
= node_id
;
1587 if (disk_expand_part_tbl(disk
, 0))
1590 ptbl
= rcu_dereference_protected(disk
->part_tbl
, 1);
1591 rcu_assign_pointer(ptbl
->part
[0], disk
->part0
);
1593 disk
->minors
= minors
;
1594 rand_initialize_disk(disk
);
1595 disk_to_dev(disk
)->class = &block_class
;
1596 disk_to_dev(disk
)->type
= &disk_type
;
1597 device_initialize(disk_to_dev(disk
));
1606 EXPORT_SYMBOL(__alloc_disk_node
);
1609 * put_disk - decrements the gendisk refcount
1610 * @disk: the struct gendisk to decrement the refcount for
1612 * This decrements the refcount for the struct gendisk. When this reaches 0
1613 * we'll have disk_release() called.
1615 * Context: Any context, but the last reference must not be dropped from
1618 void put_disk(struct gendisk
*disk
)
1621 put_device(disk_to_dev(disk
));
1623 EXPORT_SYMBOL(put_disk
);
1625 static void set_disk_ro_uevent(struct gendisk
*gd
, int ro
)
1627 char event
[] = "DISK_RO=1";
1628 char *envp
[] = { event
, NULL
};
1632 kobject_uevent_env(&disk_to_dev(gd
)->kobj
, KOBJ_CHANGE
, envp
);
1635 void set_disk_ro(struct gendisk
*disk
, int flag
)
1637 struct disk_part_iter piter
;
1638 struct block_device
*part
;
1640 if (disk
->part0
->bd_read_only
!= flag
) {
1641 set_disk_ro_uevent(disk
, flag
);
1642 disk
->part0
->bd_read_only
= flag
;
1645 disk_part_iter_init(&piter
, disk
, DISK_PITER_INCL_EMPTY
);
1646 while ((part
= disk_part_iter_next(&piter
)))
1647 part
->bd_read_only
= flag
;
1648 disk_part_iter_exit(&piter
);
1651 EXPORT_SYMBOL(set_disk_ro
);
1653 int bdev_read_only(struct block_device
*bdev
)
1657 return bdev
->bd_read_only
;
1660 EXPORT_SYMBOL(bdev_read_only
);
1663 * Disk events - monitor disk events like media change and eject request.
1665 struct disk_events
{
1666 struct list_head node
; /* all disk_event's */
1667 struct gendisk
*disk
; /* the associated disk */
1670 struct mutex block_mutex
; /* protects blocking */
1671 int block
; /* event blocking depth */
1672 unsigned int pending
; /* events already sent out */
1673 unsigned int clearing
; /* events being cleared */
1675 long poll_msecs
; /* interval, -1 for default */
1676 struct delayed_work dwork
;
1679 static const char *disk_events_strs
[] = {
1680 [ilog2(DISK_EVENT_MEDIA_CHANGE
)] = "media_change",
1681 [ilog2(DISK_EVENT_EJECT_REQUEST
)] = "eject_request",
1684 static char *disk_uevents
[] = {
1685 [ilog2(DISK_EVENT_MEDIA_CHANGE
)] = "DISK_MEDIA_CHANGE=1",
1686 [ilog2(DISK_EVENT_EJECT_REQUEST
)] = "DISK_EJECT_REQUEST=1",
1689 /* list of all disk_events */
1690 static DEFINE_MUTEX(disk_events_mutex
);
1691 static LIST_HEAD(disk_events
);
1693 /* disable in-kernel polling by default */
1694 static unsigned long disk_events_dfl_poll_msecs
;
1696 static unsigned long disk_events_poll_jiffies(struct gendisk
*disk
)
1698 struct disk_events
*ev
= disk
->ev
;
1699 long intv_msecs
= 0;
1702 * If device-specific poll interval is set, always use it. If
1703 * the default is being used, poll if the POLL flag is set.
1705 if (ev
->poll_msecs
>= 0)
1706 intv_msecs
= ev
->poll_msecs
;
1707 else if (disk
->event_flags
& DISK_EVENT_FLAG_POLL
)
1708 intv_msecs
= disk_events_dfl_poll_msecs
;
1710 return msecs_to_jiffies(intv_msecs
);
1714 * disk_block_events - block and flush disk event checking
1715 * @disk: disk to block events for
1717 * On return from this function, it is guaranteed that event checking
1718 * isn't in progress and won't happen until unblocked by
1719 * disk_unblock_events(). Events blocking is counted and the actual
1720 * unblocking happens after the matching number of unblocks are done.
1722 * Note that this intentionally does not block event checking from
1723 * disk_clear_events().
1728 void disk_block_events(struct gendisk
*disk
)
1730 struct disk_events
*ev
= disk
->ev
;
1731 unsigned long flags
;
1738 * Outer mutex ensures that the first blocker completes canceling
1739 * the event work before further blockers are allowed to finish.
1741 mutex_lock(&ev
->block_mutex
);
1743 spin_lock_irqsave(&ev
->lock
, flags
);
1744 cancel
= !ev
->block
++;
1745 spin_unlock_irqrestore(&ev
->lock
, flags
);
1748 cancel_delayed_work_sync(&disk
->ev
->dwork
);
1750 mutex_unlock(&ev
->block_mutex
);
1753 static void __disk_unblock_events(struct gendisk
*disk
, bool check_now
)
1755 struct disk_events
*ev
= disk
->ev
;
1757 unsigned long flags
;
1759 spin_lock_irqsave(&ev
->lock
, flags
);
1761 if (WARN_ON_ONCE(ev
->block
<= 0))
1767 intv
= disk_events_poll_jiffies(disk
);
1769 queue_delayed_work(system_freezable_power_efficient_wq
,
1772 queue_delayed_work(system_freezable_power_efficient_wq
,
1775 spin_unlock_irqrestore(&ev
->lock
, flags
);
1779 * disk_unblock_events - unblock disk event checking
1780 * @disk: disk to unblock events for
1782 * Undo disk_block_events(). When the block count reaches zero, it
1783 * starts events polling if configured.
1786 * Don't care. Safe to call from irq context.
1788 void disk_unblock_events(struct gendisk
*disk
)
1791 __disk_unblock_events(disk
, false);
1795 * disk_flush_events - schedule immediate event checking and flushing
1796 * @disk: disk to check and flush events for
1797 * @mask: events to flush
1799 * Schedule immediate event checking on @disk if not blocked. Events in
1800 * @mask are scheduled to be cleared from the driver. Note that this
1801 * doesn't clear the events from @disk->ev.
1804 * If @mask is non-zero must be called with bdev->bd_mutex held.
1806 void disk_flush_events(struct gendisk
*disk
, unsigned int mask
)
1808 struct disk_events
*ev
= disk
->ev
;
1813 spin_lock_irq(&ev
->lock
);
1814 ev
->clearing
|= mask
;
1816 mod_delayed_work(system_freezable_power_efficient_wq
,
1818 spin_unlock_irq(&ev
->lock
);
1822 * disk_clear_events - synchronously check, clear and return pending events
1823 * @disk: disk to fetch and clear events from
1824 * @mask: mask of events to be fetched and cleared
1826 * Disk events are synchronously checked and pending events in @mask
1827 * are cleared and returned. This ignores the block count.
1832 static unsigned int disk_clear_events(struct gendisk
*disk
, unsigned int mask
)
1834 struct disk_events
*ev
= disk
->ev
;
1835 unsigned int pending
;
1836 unsigned int clearing
= mask
;
1841 disk_block_events(disk
);
1844 * store the union of mask and ev->clearing on the stack so that the
1845 * race with disk_flush_events does not cause ambiguity (ev->clearing
1846 * can still be modified even if events are blocked).
1848 spin_lock_irq(&ev
->lock
);
1849 clearing
|= ev
->clearing
;
1851 spin_unlock_irq(&ev
->lock
);
1853 disk_check_events(ev
, &clearing
);
1855 * if ev->clearing is not 0, the disk_flush_events got called in the
1856 * middle of this function, so we want to run the workfn without delay.
1858 __disk_unblock_events(disk
, ev
->clearing
? true : false);
1860 /* then, fetch and clear pending events */
1861 spin_lock_irq(&ev
->lock
);
1862 pending
= ev
->pending
& mask
;
1863 ev
->pending
&= ~mask
;
1864 spin_unlock_irq(&ev
->lock
);
1865 WARN_ON_ONCE(clearing
& mask
);
1871 * bdev_check_media_change - check if a removable media has been changed
1872 * @bdev: block device to check
1874 * Check whether a removable media has been changed, and attempt to free all
1875 * dentries and inodes and invalidates all block device page cache entries in
1878 * Returns %true if the block device changed, or %false if not.
1880 bool bdev_check_media_change(struct block_device
*bdev
)
1882 unsigned int events
;
1884 events
= disk_clear_events(bdev
->bd_disk
, DISK_EVENT_MEDIA_CHANGE
|
1885 DISK_EVENT_EJECT_REQUEST
);
1886 if (!(events
& DISK_EVENT_MEDIA_CHANGE
))
1889 if (__invalidate_device(bdev
, true))
1890 pr_warn("VFS: busy inodes on changed media %s\n",
1891 bdev
->bd_disk
->disk_name
);
1892 set_bit(GD_NEED_PART_SCAN
, &bdev
->bd_disk
->state
);
1895 EXPORT_SYMBOL(bdev_check_media_change
);
1898 * Separate this part out so that a different pointer for clearing_ptr can be
1899 * passed in for disk_clear_events.
1901 static void disk_events_workfn(struct work_struct
*work
)
1903 struct delayed_work
*dwork
= to_delayed_work(work
);
1904 struct disk_events
*ev
= container_of(dwork
, struct disk_events
, dwork
);
1906 disk_check_events(ev
, &ev
->clearing
);
1909 static void disk_check_events(struct disk_events
*ev
,
1910 unsigned int *clearing_ptr
)
1912 struct gendisk
*disk
= ev
->disk
;
1913 char *envp
[ARRAY_SIZE(disk_uevents
) + 1] = { };
1914 unsigned int clearing
= *clearing_ptr
;
1915 unsigned int events
;
1917 int nr_events
= 0, i
;
1920 events
= disk
->fops
->check_events(disk
, clearing
);
1922 /* accumulate pending events and schedule next poll if necessary */
1923 spin_lock_irq(&ev
->lock
);
1925 events
&= ~ev
->pending
;
1926 ev
->pending
|= events
;
1927 *clearing_ptr
&= ~clearing
;
1929 intv
= disk_events_poll_jiffies(disk
);
1930 if (!ev
->block
&& intv
)
1931 queue_delayed_work(system_freezable_power_efficient_wq
,
1934 spin_unlock_irq(&ev
->lock
);
1937 * Tell userland about new events. Only the events listed in
1938 * @disk->events are reported, and only if DISK_EVENT_FLAG_UEVENT
1939 * is set. Otherwise, events are processed internally but never
1940 * get reported to userland.
1942 for (i
= 0; i
< ARRAY_SIZE(disk_uevents
); i
++)
1943 if ((events
& disk
->events
& (1 << i
)) &&
1944 (disk
->event_flags
& DISK_EVENT_FLAG_UEVENT
))
1945 envp
[nr_events
++] = disk_uevents
[i
];
1948 kobject_uevent_env(&disk_to_dev(disk
)->kobj
, KOBJ_CHANGE
, envp
);
1952 * A disk events enabled device has the following sysfs nodes under
1953 * its /sys/block/X/ directory.
1955 * events : list of all supported events
1956 * events_async : list of events which can be detected w/o polling
1957 * (always empty, only for backwards compatibility)
1958 * events_poll_msecs : polling interval, 0: disable, -1: system default
1960 static ssize_t
__disk_events_show(unsigned int events
, char *buf
)
1962 const char *delim
= "";
1966 for (i
= 0; i
< ARRAY_SIZE(disk_events_strs
); i
++)
1967 if (events
& (1 << i
)) {
1968 pos
+= sprintf(buf
+ pos
, "%s%s",
1969 delim
, disk_events_strs
[i
]);
1973 pos
+= sprintf(buf
+ pos
, "\n");
1977 static ssize_t
disk_events_show(struct device
*dev
,
1978 struct device_attribute
*attr
, char *buf
)
1980 struct gendisk
*disk
= dev_to_disk(dev
);
1982 if (!(disk
->event_flags
& DISK_EVENT_FLAG_UEVENT
))
1985 return __disk_events_show(disk
->events
, buf
);
1988 static ssize_t
disk_events_async_show(struct device
*dev
,
1989 struct device_attribute
*attr
, char *buf
)
1994 static ssize_t
disk_events_poll_msecs_show(struct device
*dev
,
1995 struct device_attribute
*attr
,
1998 struct gendisk
*disk
= dev_to_disk(dev
);
2001 return sprintf(buf
, "-1\n");
2003 return sprintf(buf
, "%ld\n", disk
->ev
->poll_msecs
);
2006 static ssize_t
disk_events_poll_msecs_store(struct device
*dev
,
2007 struct device_attribute
*attr
,
2008 const char *buf
, size_t count
)
2010 struct gendisk
*disk
= dev_to_disk(dev
);
2013 if (!count
|| !sscanf(buf
, "%ld", &intv
))
2016 if (intv
< 0 && intv
!= -1)
2022 disk_block_events(disk
);
2023 disk
->ev
->poll_msecs
= intv
;
2024 __disk_unblock_events(disk
, true);
2029 static const DEVICE_ATTR(events
, 0444, disk_events_show
, NULL
);
2030 static const DEVICE_ATTR(events_async
, 0444, disk_events_async_show
, NULL
);
2031 static const DEVICE_ATTR(events_poll_msecs
, 0644,
2032 disk_events_poll_msecs_show
,
2033 disk_events_poll_msecs_store
);
2035 static const struct attribute
*disk_events_attrs
[] = {
2036 &dev_attr_events
.attr
,
2037 &dev_attr_events_async
.attr
,
2038 &dev_attr_events_poll_msecs
.attr
,
2043 * The default polling interval can be specified by the kernel
2044 * parameter block.events_dfl_poll_msecs which defaults to 0
2045 * (disable). This can also be modified runtime by writing to
2046 * /sys/module/block/parameters/events_dfl_poll_msecs.
2048 static int disk_events_set_dfl_poll_msecs(const char *val
,
2049 const struct kernel_param
*kp
)
2051 struct disk_events
*ev
;
2054 ret
= param_set_ulong(val
, kp
);
2058 mutex_lock(&disk_events_mutex
);
2060 list_for_each_entry(ev
, &disk_events
, node
)
2061 disk_flush_events(ev
->disk
, 0);
2063 mutex_unlock(&disk_events_mutex
);
2068 static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops
= {
2069 .set
= disk_events_set_dfl_poll_msecs
,
2070 .get
= param_get_ulong
,
2073 #undef MODULE_PARAM_PREFIX
2074 #define MODULE_PARAM_PREFIX "block."
2076 module_param_cb(events_dfl_poll_msecs
, &disk_events_dfl_poll_msecs_param_ops
,
2077 &disk_events_dfl_poll_msecs
, 0644);
2080 * disk_{alloc|add|del|release}_events - initialize and destroy disk_events.
2082 static void disk_alloc_events(struct gendisk
*disk
)
2084 struct disk_events
*ev
;
2086 if (!disk
->fops
->check_events
|| !disk
->events
)
2089 ev
= kzalloc(sizeof(*ev
), GFP_KERNEL
);
2091 pr_warn("%s: failed to initialize events\n", disk
->disk_name
);
2095 INIT_LIST_HEAD(&ev
->node
);
2097 spin_lock_init(&ev
->lock
);
2098 mutex_init(&ev
->block_mutex
);
2100 ev
->poll_msecs
= -1;
2101 INIT_DELAYED_WORK(&ev
->dwork
, disk_events_workfn
);
2106 static void disk_add_events(struct gendisk
*disk
)
2108 /* FIXME: error handling */
2109 if (sysfs_create_files(&disk_to_dev(disk
)->kobj
, disk_events_attrs
) < 0)
2110 pr_warn("%s: failed to create sysfs files for events\n",
2116 mutex_lock(&disk_events_mutex
);
2117 list_add_tail(&disk
->ev
->node
, &disk_events
);
2118 mutex_unlock(&disk_events_mutex
);
2121 * Block count is initialized to 1 and the following initial
2122 * unblock kicks it into action.
2124 __disk_unblock_events(disk
, true);
2127 static void disk_del_events(struct gendisk
*disk
)
2130 disk_block_events(disk
);
2132 mutex_lock(&disk_events_mutex
);
2133 list_del_init(&disk
->ev
->node
);
2134 mutex_unlock(&disk_events_mutex
);
2137 sysfs_remove_files(&disk_to_dev(disk
)->kobj
, disk_events_attrs
);
2140 static void disk_release_events(struct gendisk
*disk
)
2142 /* the block count should be 1 from disk_del_events() */
2143 WARN_ON_ONCE(disk
->ev
&& disk
->ev
->block
!= 1);