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/kobj_map.h>
21 #include <linux/mutex.h>
22 #include <linux/idr.h>
23 #include <linux/log2.h>
24 #include <linux/pm_runtime.h>
25 #include <linux/badblocks.h>
29 static DEFINE_MUTEX(block_class_lock
);
30 static struct kobject
*block_depr
;
32 /* for extended dynamic devt allocation, currently only one major is used */
33 #define NR_EXT_DEVT (1 << MINORBITS)
35 /* For extended devt allocation. ext_devt_lock prevents look up
36 * results from going away underneath its user.
38 static DEFINE_SPINLOCK(ext_devt_lock
);
39 static DEFINE_IDR(ext_devt_idr
);
41 static void disk_check_events(struct disk_events
*ev
,
42 unsigned int *clearing_ptr
);
43 static void disk_alloc_events(struct gendisk
*disk
);
44 static void disk_add_events(struct gendisk
*disk
);
45 static void disk_del_events(struct gendisk
*disk
);
46 static void disk_release_events(struct gendisk
*disk
);
49 * Set disk capacity and notify if the size is not currently
50 * zero and will not be set to zero
52 bool set_capacity_revalidate_and_notify(struct gendisk
*disk
, sector_t size
,
55 sector_t capacity
= get_capacity(disk
);
57 set_capacity(disk
, size
);
59 revalidate_disk_size(disk
, true);
61 if (capacity
!= size
&& capacity
!= 0 && size
!= 0) {
62 char *envp
[] = { "RESIZE=1", NULL
};
64 kobject_uevent_env(&disk_to_dev(disk
)->kobj
, KOBJ_CHANGE
, envp
);
71 EXPORT_SYMBOL_GPL(set_capacity_revalidate_and_notify
);
74 * Format the device name of the indicated disk into the supplied buffer and
75 * return a pointer to that same buffer for convenience.
77 char *disk_name(struct gendisk
*hd
, int partno
, char *buf
)
80 snprintf(buf
, BDEVNAME_SIZE
, "%s", hd
->disk_name
);
81 else if (isdigit(hd
->disk_name
[strlen(hd
->disk_name
)-1]))
82 snprintf(buf
, BDEVNAME_SIZE
, "%sp%d", hd
->disk_name
, partno
);
84 snprintf(buf
, BDEVNAME_SIZE
, "%s%d", hd
->disk_name
, partno
);
89 const char *bdevname(struct block_device
*bdev
, char *buf
)
91 return disk_name(bdev
->bd_disk
, bdev
->bd_partno
, buf
);
93 EXPORT_SYMBOL(bdevname
);
95 static void part_stat_read_all(struct hd_struct
*part
, struct disk_stats
*stat
)
99 memset(stat
, 0, sizeof(struct disk_stats
));
100 for_each_possible_cpu(cpu
) {
101 struct disk_stats
*ptr
= per_cpu_ptr(part
->dkstats
, cpu
);
104 for (group
= 0; group
< NR_STAT_GROUPS
; group
++) {
105 stat
->nsecs
[group
] += ptr
->nsecs
[group
];
106 stat
->sectors
[group
] += ptr
->sectors
[group
];
107 stat
->ios
[group
] += ptr
->ios
[group
];
108 stat
->merges
[group
] += ptr
->merges
[group
];
111 stat
->io_ticks
+= ptr
->io_ticks
;
115 static unsigned int part_in_flight(struct hd_struct
*part
)
117 unsigned int inflight
= 0;
120 for_each_possible_cpu(cpu
) {
121 inflight
+= part_stat_local_read_cpu(part
, in_flight
[0], cpu
) +
122 part_stat_local_read_cpu(part
, in_flight
[1], cpu
);
124 if ((int)inflight
< 0)
130 static void part_in_flight_rw(struct hd_struct
*part
, unsigned int inflight
[2])
136 for_each_possible_cpu(cpu
) {
137 inflight
[0] += part_stat_local_read_cpu(part
, in_flight
[0], cpu
);
138 inflight
[1] += part_stat_local_read_cpu(part
, in_flight
[1], cpu
);
140 if ((int)inflight
[0] < 0)
142 if ((int)inflight
[1] < 0)
146 struct hd_struct
*__disk_get_part(struct gendisk
*disk
, int partno
)
148 struct disk_part_tbl
*ptbl
= rcu_dereference(disk
->part_tbl
);
150 if (unlikely(partno
< 0 || partno
>= ptbl
->len
))
152 return rcu_dereference(ptbl
->part
[partno
]);
156 * disk_get_part - get partition
157 * @disk: disk to look partition from
158 * @partno: partition number
160 * Look for partition @partno from @disk. If found, increment
161 * reference count and return it.
167 * Pointer to the found partition on success, NULL if not found.
169 struct hd_struct
*disk_get_part(struct gendisk
*disk
, int partno
)
171 struct hd_struct
*part
;
174 part
= __disk_get_part(disk
, partno
);
176 get_device(part_to_dev(part
));
183 * disk_part_iter_init - initialize partition iterator
184 * @piter: iterator to initialize
185 * @disk: disk to iterate over
186 * @flags: DISK_PITER_* flags
188 * Initialize @piter so that it iterates over partitions of @disk.
193 void disk_part_iter_init(struct disk_part_iter
*piter
, struct gendisk
*disk
,
196 struct disk_part_tbl
*ptbl
;
199 ptbl
= rcu_dereference(disk
->part_tbl
);
204 if (flags
& DISK_PITER_REVERSE
)
205 piter
->idx
= ptbl
->len
- 1;
206 else if (flags
& (DISK_PITER_INCL_PART0
| DISK_PITER_INCL_EMPTY_PART0
))
211 piter
->flags
= flags
;
215 EXPORT_SYMBOL_GPL(disk_part_iter_init
);
218 * disk_part_iter_next - proceed iterator to the next partition and return it
219 * @piter: iterator of interest
221 * Proceed @piter to the next partition and return it.
226 struct hd_struct
*disk_part_iter_next(struct disk_part_iter
*piter
)
228 struct disk_part_tbl
*ptbl
;
231 /* put the last partition */
232 disk_put_part(piter
->part
);
237 ptbl
= rcu_dereference(piter
->disk
->part_tbl
);
239 /* determine iteration parameters */
240 if (piter
->flags
& DISK_PITER_REVERSE
) {
242 if (piter
->flags
& (DISK_PITER_INCL_PART0
|
243 DISK_PITER_INCL_EMPTY_PART0
))
252 /* iterate to the next partition */
253 for (; piter
->idx
!= end
; piter
->idx
+= inc
) {
254 struct hd_struct
*part
;
256 part
= rcu_dereference(ptbl
->part
[piter
->idx
]);
259 if (!part_nr_sects_read(part
) &&
260 !(piter
->flags
& DISK_PITER_INCL_EMPTY
) &&
261 !(piter
->flags
& DISK_PITER_INCL_EMPTY_PART0
&&
265 get_device(part_to_dev(part
));
275 EXPORT_SYMBOL_GPL(disk_part_iter_next
);
278 * disk_part_iter_exit - finish up partition iteration
279 * @piter: iter of interest
281 * Called when iteration is over. Cleans up @piter.
286 void disk_part_iter_exit(struct disk_part_iter
*piter
)
288 disk_put_part(piter
->part
);
291 EXPORT_SYMBOL_GPL(disk_part_iter_exit
);
293 static inline int sector_in_part(struct hd_struct
*part
, sector_t sector
)
295 return part
->start_sect
<= sector
&&
296 sector
< part
->start_sect
+ part_nr_sects_read(part
);
300 * disk_map_sector_rcu - map sector to partition
301 * @disk: gendisk of interest
302 * @sector: sector to map
304 * Find out which partition @sector maps to on @disk. This is
305 * primarily used for stats accounting.
308 * RCU read locked. The returned partition pointer is always valid
309 * because its refcount is grabbed except for part0, which lifetime
310 * is same with the disk.
313 * Found partition on success, part0 is returned if no partition matches
314 * or the matched partition is being deleted.
316 struct hd_struct
*disk_map_sector_rcu(struct gendisk
*disk
, sector_t sector
)
318 struct disk_part_tbl
*ptbl
;
319 struct hd_struct
*part
;
323 ptbl
= rcu_dereference(disk
->part_tbl
);
325 part
= rcu_dereference(ptbl
->last_lookup
);
326 if (part
&& sector_in_part(part
, sector
) && hd_struct_try_get(part
))
329 for (i
= 1; i
< ptbl
->len
; i
++) {
330 part
= rcu_dereference(ptbl
->part
[i
]);
332 if (part
&& sector_in_part(part
, sector
)) {
334 * only live partition can be cached for lookup,
335 * so use-after-free on cached & deleting partition
338 if (!hd_struct_try_get(part
))
340 rcu_assign_pointer(ptbl
->last_lookup
, part
);
352 * disk_has_partitions
353 * @disk: gendisk of interest
355 * Walk through the partition table and check if valid partition exists.
361 * True if the gendisk has at least one valid non-zero size partition.
364 bool disk_has_partitions(struct gendisk
*disk
)
366 struct disk_part_tbl
*ptbl
;
371 ptbl
= rcu_dereference(disk
->part_tbl
);
373 /* Iterate partitions skipping the whole device at index 0 */
374 for (i
= 1; i
< ptbl
->len
; i
++) {
375 if (rcu_dereference(ptbl
->part
[i
])) {
385 EXPORT_SYMBOL_GPL(disk_has_partitions
);
388 * Can be deleted altogether. Later.
391 #define BLKDEV_MAJOR_HASH_SIZE 255
392 static struct blk_major_name
{
393 struct blk_major_name
*next
;
396 } *major_names
[BLKDEV_MAJOR_HASH_SIZE
];
398 /* index in the above - for now: assume no multimajor ranges */
399 static inline int major_to_index(unsigned major
)
401 return major
% BLKDEV_MAJOR_HASH_SIZE
;
404 #ifdef CONFIG_PROC_FS
405 void blkdev_show(struct seq_file
*seqf
, off_t offset
)
407 struct blk_major_name
*dp
;
409 mutex_lock(&block_class_lock
);
410 for (dp
= major_names
[major_to_index(offset
)]; dp
; dp
= dp
->next
)
411 if (dp
->major
== offset
)
412 seq_printf(seqf
, "%3d %s\n", dp
->major
, dp
->name
);
413 mutex_unlock(&block_class_lock
);
415 #endif /* CONFIG_PROC_FS */
418 * register_blkdev - register a new block device
420 * @major: the requested major device number [1..BLKDEV_MAJOR_MAX-1]. If
421 * @major = 0, try to allocate any unused major number.
422 * @name: the name of the new block device as a zero terminated string
424 * The @name must be unique within the system.
426 * The return value depends on the @major input parameter:
428 * - if a major device number was requested in range [1..BLKDEV_MAJOR_MAX-1]
429 * then the function returns zero on success, or a negative error code
430 * - if any unused major number was requested with @major = 0 parameter
431 * then the return value is the allocated major number in range
432 * [1..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise
434 * See Documentation/admin-guide/devices.txt for the list of allocated
437 int register_blkdev(unsigned int major
, const char *name
)
439 struct blk_major_name
**n
, *p
;
442 mutex_lock(&block_class_lock
);
446 for (index
= ARRAY_SIZE(major_names
)-1; index
> 0; index
--) {
447 if (major_names
[index
] == NULL
)
452 printk("%s: failed to get major for %s\n",
461 if (major
>= BLKDEV_MAJOR_MAX
) {
462 pr_err("%s: major requested (%u) is greater than the maximum (%u) for %s\n",
463 __func__
, major
, BLKDEV_MAJOR_MAX
-1, name
);
469 p
= kmalloc(sizeof(struct blk_major_name
), GFP_KERNEL
);
476 strlcpy(p
->name
, name
, sizeof(p
->name
));
478 index
= major_to_index(major
);
480 for (n
= &major_names
[index
]; *n
; n
= &(*n
)->next
) {
481 if ((*n
)->major
== major
)
490 printk("register_blkdev: cannot get major %u for %s\n",
495 mutex_unlock(&block_class_lock
);
499 EXPORT_SYMBOL(register_blkdev
);
501 void unregister_blkdev(unsigned int major
, const char *name
)
503 struct blk_major_name
**n
;
504 struct blk_major_name
*p
= NULL
;
505 int index
= major_to_index(major
);
507 mutex_lock(&block_class_lock
);
508 for (n
= &major_names
[index
]; *n
; n
= &(*n
)->next
)
509 if ((*n
)->major
== major
)
511 if (!*n
|| strcmp((*n
)->name
, name
)) {
517 mutex_unlock(&block_class_lock
);
521 EXPORT_SYMBOL(unregister_blkdev
);
523 static struct kobj_map
*bdev_map
;
526 * blk_mangle_minor - scatter minor numbers apart
527 * @minor: minor number to mangle
529 * Scatter consecutively allocated @minor number apart if MANGLE_DEVT
530 * is enabled. Mangling twice gives the original value.
538 static int blk_mangle_minor(int minor
)
540 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
543 for (i
= 0; i
< MINORBITS
/ 2; i
++) {
544 int low
= minor
& (1 << i
);
545 int high
= minor
& (1 << (MINORBITS
- 1 - i
));
546 int distance
= MINORBITS
- 1 - 2 * i
;
548 minor
^= low
| high
; /* clear both bits */
549 low
<<= distance
; /* swap the positions */
551 minor
|= low
| high
; /* and set */
558 * blk_alloc_devt - allocate a dev_t for a partition
559 * @part: partition to allocate dev_t for
560 * @devt: out parameter for resulting dev_t
562 * Allocate a dev_t for block device.
565 * 0 on success, allocated dev_t is returned in *@devt. -errno on
571 int blk_alloc_devt(struct hd_struct
*part
, dev_t
*devt
)
573 struct gendisk
*disk
= part_to_disk(part
);
576 /* in consecutive minor range? */
577 if (part
->partno
< disk
->minors
) {
578 *devt
= MKDEV(disk
->major
, disk
->first_minor
+ part
->partno
);
582 /* allocate ext devt */
583 idr_preload(GFP_KERNEL
);
585 spin_lock_bh(&ext_devt_lock
);
586 idx
= idr_alloc(&ext_devt_idr
, part
, 0, NR_EXT_DEVT
, GFP_NOWAIT
);
587 spin_unlock_bh(&ext_devt_lock
);
591 return idx
== -ENOSPC
? -EBUSY
: idx
;
593 *devt
= MKDEV(BLOCK_EXT_MAJOR
, blk_mangle_minor(idx
));
598 * blk_free_devt - free a dev_t
599 * @devt: dev_t to free
601 * Free @devt which was allocated using blk_alloc_devt().
606 void blk_free_devt(dev_t devt
)
608 if (devt
== MKDEV(0, 0))
611 if (MAJOR(devt
) == BLOCK_EXT_MAJOR
) {
612 spin_lock_bh(&ext_devt_lock
);
613 idr_remove(&ext_devt_idr
, blk_mangle_minor(MINOR(devt
)));
614 spin_unlock_bh(&ext_devt_lock
);
619 * We invalidate devt by assigning NULL pointer for devt in idr.
621 void blk_invalidate_devt(dev_t devt
)
623 if (MAJOR(devt
) == BLOCK_EXT_MAJOR
) {
624 spin_lock_bh(&ext_devt_lock
);
625 idr_replace(&ext_devt_idr
, NULL
, blk_mangle_minor(MINOR(devt
)));
626 spin_unlock_bh(&ext_devt_lock
);
630 static char *bdevt_str(dev_t devt
, char *buf
)
632 if (MAJOR(devt
) <= 0xff && MINOR(devt
) <= 0xff) {
633 char tbuf
[BDEVT_SIZE
];
634 snprintf(tbuf
, BDEVT_SIZE
, "%02x%02x", MAJOR(devt
), MINOR(devt
));
635 snprintf(buf
, BDEVT_SIZE
, "%-9s", tbuf
);
637 snprintf(buf
, BDEVT_SIZE
, "%03x:%05x", MAJOR(devt
), MINOR(devt
));
643 * Register device numbers dev..(dev+range-1)
644 * range must be nonzero
645 * The hash chain is sorted on range, so that subranges can override.
647 void blk_register_region(dev_t devt
, unsigned long range
, struct module
*module
,
648 struct kobject
*(*probe
)(dev_t
, int *, void *),
649 int (*lock
)(dev_t
, void *), void *data
)
651 kobj_map(bdev_map
, devt
, range
, module
, probe
, lock
, data
);
654 EXPORT_SYMBOL(blk_register_region
);
656 void blk_unregister_region(dev_t devt
, unsigned long range
)
658 kobj_unmap(bdev_map
, devt
, range
);
661 EXPORT_SYMBOL(blk_unregister_region
);
663 static struct kobject
*exact_match(dev_t devt
, int *partno
, void *data
)
665 struct gendisk
*p
= data
;
667 return &disk_to_dev(p
)->kobj
;
670 static int exact_lock(dev_t devt
, void *data
)
672 struct gendisk
*p
= data
;
674 if (!get_disk_and_module(p
))
679 static void disk_scan_partitions(struct gendisk
*disk
)
681 struct block_device
*bdev
;
683 if (!get_capacity(disk
) || !disk_part_scan_enabled(disk
))
686 set_bit(GD_NEED_PART_SCAN
, &disk
->state
);
687 bdev
= blkdev_get_by_dev(disk_devt(disk
), FMODE_READ
, NULL
);
689 blkdev_put(bdev
, FMODE_READ
);
692 static void register_disk(struct device
*parent
, struct gendisk
*disk
,
693 const struct attribute_group
**groups
)
695 struct device
*ddev
= disk_to_dev(disk
);
696 struct disk_part_iter piter
;
697 struct hd_struct
*part
;
700 ddev
->parent
= parent
;
702 dev_set_name(ddev
, "%s", disk
->disk_name
);
704 /* delay uevents, until we scanned partition table */
705 dev_set_uevent_suppress(ddev
, 1);
708 WARN_ON(ddev
->groups
);
709 ddev
->groups
= groups
;
711 if (device_add(ddev
))
713 if (!sysfs_deprecated
) {
714 err
= sysfs_create_link(block_depr
, &ddev
->kobj
,
715 kobject_name(&ddev
->kobj
));
723 * avoid probable deadlock caused by allocating memory with
724 * GFP_KERNEL in runtime_resume callback of its all ancestor
727 pm_runtime_set_memalloc_noio(ddev
, true);
729 disk
->part0
.holder_dir
= kobject_create_and_add("holders", &ddev
->kobj
);
730 disk
->slave_dir
= kobject_create_and_add("slaves", &ddev
->kobj
);
732 if (disk
->flags
& GENHD_FL_HIDDEN
) {
733 dev_set_uevent_suppress(ddev
, 0);
737 disk_scan_partitions(disk
);
739 /* announce disk after possible partitions are created */
740 dev_set_uevent_suppress(ddev
, 0);
741 kobject_uevent(&ddev
->kobj
, KOBJ_ADD
);
743 /* announce possible partitions */
744 disk_part_iter_init(&piter
, disk
, 0);
745 while ((part
= disk_part_iter_next(&piter
)))
746 kobject_uevent(&part_to_dev(part
)->kobj
, KOBJ_ADD
);
747 disk_part_iter_exit(&piter
);
749 if (disk
->queue
->backing_dev_info
->dev
) {
750 err
= sysfs_create_link(&ddev
->kobj
,
751 &disk
->queue
->backing_dev_info
->dev
->kobj
,
758 * __device_add_disk - add disk information to kernel list
759 * @parent: parent device for the disk
760 * @disk: per-device partitioning information
761 * @groups: Additional per-device sysfs groups
762 * @register_queue: register the queue if set to true
764 * This function registers the partitioning information in @disk
767 * FIXME: error handling
769 static void __device_add_disk(struct device
*parent
, struct gendisk
*disk
,
770 const struct attribute_group
**groups
,
777 * The disk queue should now be all set with enough information about
778 * the device for the elevator code to pick an adequate default
779 * elevator if one is needed, that is, for devices requesting queue
783 elevator_init_mq(disk
->queue
);
785 /* minors == 0 indicates to use ext devt from part0 and should
786 * be accompanied with EXT_DEVT flag. Make sure all
787 * parameters make sense.
789 WARN_ON(disk
->minors
&& !(disk
->major
|| disk
->first_minor
));
790 WARN_ON(!disk
->minors
&&
791 !(disk
->flags
& (GENHD_FL_EXT_DEVT
| GENHD_FL_HIDDEN
)));
793 disk
->flags
|= GENHD_FL_UP
;
795 retval
= blk_alloc_devt(&disk
->part0
, &devt
);
800 disk
->major
= MAJOR(devt
);
801 disk
->first_minor
= MINOR(devt
);
803 disk_alloc_events(disk
);
805 if (disk
->flags
& GENHD_FL_HIDDEN
) {
807 * Don't let hidden disks show up in /proc/partitions,
808 * and don't bother scanning for partitions either.
810 disk
->flags
|= GENHD_FL_SUPPRESS_PARTITION_INFO
;
811 disk
->flags
|= GENHD_FL_NO_PART_SCAN
;
813 struct backing_dev_info
*bdi
= disk
->queue
->backing_dev_info
;
814 struct device
*dev
= disk_to_dev(disk
);
817 /* Register BDI before referencing it from bdev */
819 ret
= bdi_register(bdi
, "%u:%u", MAJOR(devt
), MINOR(devt
));
821 bdi_set_owner(bdi
, dev
);
822 blk_register_region(disk_devt(disk
), disk
->minors
, NULL
,
823 exact_match
, exact_lock
, disk
);
825 register_disk(parent
, disk
, groups
);
827 blk_register_queue(disk
);
830 * Take an extra ref on queue which will be put on disk_release()
831 * so that it sticks around as long as @disk is there.
833 WARN_ON_ONCE(!blk_get_queue(disk
->queue
));
835 disk_add_events(disk
);
836 blk_integrity_add(disk
);
839 void device_add_disk(struct device
*parent
, struct gendisk
*disk
,
840 const struct attribute_group
**groups
)
843 __device_add_disk(parent
, disk
, groups
, true);
845 EXPORT_SYMBOL(device_add_disk
);
847 void device_add_disk_no_queue_reg(struct device
*parent
, struct gendisk
*disk
)
849 __device_add_disk(parent
, disk
, NULL
, false);
851 EXPORT_SYMBOL(device_add_disk_no_queue_reg
);
853 static void invalidate_partition(struct gendisk
*disk
, int partno
)
855 struct block_device
*bdev
;
857 bdev
= bdget_disk(disk
, partno
);
862 __invalidate_device(bdev
, true);
865 * Unhash the bdev inode for this device so that it gets evicted as soon
866 * as last inode reference is dropped.
868 remove_inode_hash(bdev
->bd_inode
);
873 * del_gendisk - remove the gendisk
874 * @disk: the struct gendisk to remove
876 * Removes the gendisk and all its associated resources. This deletes the
877 * partitions associated with the gendisk, and unregisters the associated
880 * This is the counter to the respective __device_add_disk() call.
882 * The final removal of the struct gendisk happens when its refcount reaches 0
883 * with put_disk(), which should be called after del_gendisk(), if
884 * __device_add_disk() was used.
886 * Drivers exist which depend on the release of the gendisk to be synchronous,
887 * it should not be deferred.
891 void del_gendisk(struct gendisk
*disk
)
893 struct disk_part_iter piter
;
894 struct hd_struct
*part
;
898 blk_integrity_del(disk
);
899 disk_del_events(disk
);
902 * Block lookups of the disk until all bdevs are unhashed and the
903 * disk is marked as dead (GENHD_FL_UP cleared).
905 down_write(&disk
->lookup_sem
);
906 /* invalidate stuff */
907 disk_part_iter_init(&piter
, disk
,
908 DISK_PITER_INCL_EMPTY
| DISK_PITER_REVERSE
);
909 while ((part
= disk_part_iter_next(&piter
))) {
910 invalidate_partition(disk
, part
->partno
);
911 delete_partition(part
);
913 disk_part_iter_exit(&piter
);
915 invalidate_partition(disk
, 0);
916 set_capacity(disk
, 0);
917 disk
->flags
&= ~GENHD_FL_UP
;
918 up_write(&disk
->lookup_sem
);
920 if (!(disk
->flags
& GENHD_FL_HIDDEN
))
921 sysfs_remove_link(&disk_to_dev(disk
)->kobj
, "bdi");
924 * Unregister bdi before releasing device numbers (as they can
925 * get reused and we'd get clashes in sysfs).
927 if (!(disk
->flags
& GENHD_FL_HIDDEN
))
928 bdi_unregister(disk
->queue
->backing_dev_info
);
929 blk_unregister_queue(disk
);
934 if (!(disk
->flags
& GENHD_FL_HIDDEN
))
935 blk_unregister_region(disk_devt(disk
), disk
->minors
);
937 * Remove gendisk pointer from idr so that it cannot be looked up
938 * while RCU period before freeing gendisk is running to prevent
939 * use-after-free issues. Note that the device number stays
940 * "in-use" until we really free the gendisk.
942 blk_invalidate_devt(disk_devt(disk
));
944 kobject_put(disk
->part0
.holder_dir
);
945 kobject_put(disk
->slave_dir
);
947 part_stat_set_all(&disk
->part0
, 0);
948 disk
->part0
.stamp
= 0;
949 if (!sysfs_deprecated
)
950 sysfs_remove_link(block_depr
, dev_name(disk_to_dev(disk
)));
951 pm_runtime_set_memalloc_noio(disk_to_dev(disk
), false);
952 device_del(disk_to_dev(disk
));
954 EXPORT_SYMBOL(del_gendisk
);
956 /* sysfs access to bad-blocks list. */
957 static ssize_t
disk_badblocks_show(struct device
*dev
,
958 struct device_attribute
*attr
,
961 struct gendisk
*disk
= dev_to_disk(dev
);
964 return sprintf(page
, "\n");
966 return badblocks_show(disk
->bb
, page
, 0);
969 static ssize_t
disk_badblocks_store(struct device
*dev
,
970 struct device_attribute
*attr
,
971 const char *page
, size_t len
)
973 struct gendisk
*disk
= dev_to_disk(dev
);
978 return badblocks_store(disk
->bb
, page
, len
, 0);
982 * get_gendisk - get partitioning information for a given device
983 * @devt: device to get partitioning information for
984 * @partno: returned partition index
986 * This function gets the structure containing partitioning
987 * information for the given device @devt.
991 struct gendisk
*get_gendisk(dev_t devt
, int *partno
)
993 struct gendisk
*disk
= NULL
;
997 if (MAJOR(devt
) != BLOCK_EXT_MAJOR
) {
998 struct kobject
*kobj
;
1000 kobj
= kobj_lookup(bdev_map
, devt
, partno
);
1002 disk
= dev_to_disk(kobj_to_dev(kobj
));
1004 struct hd_struct
*part
;
1006 spin_lock_bh(&ext_devt_lock
);
1007 part
= idr_find(&ext_devt_idr
, blk_mangle_minor(MINOR(devt
)));
1008 if (part
&& get_disk_and_module(part_to_disk(part
))) {
1009 *partno
= part
->partno
;
1010 disk
= part_to_disk(part
);
1012 spin_unlock_bh(&ext_devt_lock
);
1019 * Synchronize with del_gendisk() to not return disk that is being
1022 down_read(&disk
->lookup_sem
);
1023 if (unlikely((disk
->flags
& GENHD_FL_HIDDEN
) ||
1024 !(disk
->flags
& GENHD_FL_UP
))) {
1025 up_read(&disk
->lookup_sem
);
1026 put_disk_and_module(disk
);
1029 up_read(&disk
->lookup_sem
);
1035 * bdget_disk - do bdget() by gendisk and partition number
1036 * @disk: gendisk of interest
1037 * @partno: partition number
1039 * Find partition @partno from @disk, do bdget() on it.
1045 * Resulting block_device on success, NULL on failure.
1047 struct block_device
*bdget_disk(struct gendisk
*disk
, int partno
)
1049 struct hd_struct
*part
;
1050 struct block_device
*bdev
= NULL
;
1052 part
= disk_get_part(disk
, partno
);
1054 bdev
= bdget_part(part
);
1055 disk_put_part(part
);
1059 EXPORT_SYMBOL(bdget_disk
);
1062 * print a full list of all partitions - intended for places where the root
1063 * filesystem can't be mounted and thus to give the victim some idea of what
1066 void __init
printk_all_partitions(void)
1068 struct class_dev_iter iter
;
1071 class_dev_iter_init(&iter
, &block_class
, NULL
, &disk_type
);
1072 while ((dev
= class_dev_iter_next(&iter
))) {
1073 struct gendisk
*disk
= dev_to_disk(dev
);
1074 struct disk_part_iter piter
;
1075 struct hd_struct
*part
;
1076 char name_buf
[BDEVNAME_SIZE
];
1077 char devt_buf
[BDEVT_SIZE
];
1080 * Don't show empty devices or things that have been
1083 if (get_capacity(disk
) == 0 ||
1084 (disk
->flags
& GENHD_FL_SUPPRESS_PARTITION_INFO
))
1088 * Note, unlike /proc/partitions, I am showing the
1089 * numbers in hex - the same format as the root=
1092 disk_part_iter_init(&piter
, disk
, DISK_PITER_INCL_PART0
);
1093 while ((part
= disk_part_iter_next(&piter
))) {
1094 bool is_part0
= part
== &disk
->part0
;
1096 printk("%s%s %10llu %s %s", is_part0
? "" : " ",
1097 bdevt_str(part_devt(part
), devt_buf
),
1098 (unsigned long long)part_nr_sects_read(part
) >> 1
1099 , disk_name(disk
, part
->partno
, name_buf
),
1100 part
->info
? part
->info
->uuid
: "");
1102 if (dev
->parent
&& dev
->parent
->driver
)
1103 printk(" driver: %s\n",
1104 dev
->parent
->driver
->name
);
1106 printk(" (driver?)\n");
1110 disk_part_iter_exit(&piter
);
1112 class_dev_iter_exit(&iter
);
1115 #ifdef CONFIG_PROC_FS
1117 static void *disk_seqf_start(struct seq_file
*seqf
, loff_t
*pos
)
1120 struct class_dev_iter
*iter
;
1123 iter
= kmalloc(sizeof(*iter
), GFP_KERNEL
);
1125 return ERR_PTR(-ENOMEM
);
1127 seqf
->private = iter
;
1128 class_dev_iter_init(iter
, &block_class
, NULL
, &disk_type
);
1130 dev
= class_dev_iter_next(iter
);
1135 return dev_to_disk(dev
);
1138 static void *disk_seqf_next(struct seq_file
*seqf
, void *v
, loff_t
*pos
)
1143 dev
= class_dev_iter_next(seqf
->private);
1145 return dev_to_disk(dev
);
1150 static void disk_seqf_stop(struct seq_file
*seqf
, void *v
)
1152 struct class_dev_iter
*iter
= seqf
->private;
1154 /* stop is called even after start failed :-( */
1156 class_dev_iter_exit(iter
);
1158 seqf
->private = NULL
;
1162 static void *show_partition_start(struct seq_file
*seqf
, loff_t
*pos
)
1166 p
= disk_seqf_start(seqf
, pos
);
1167 if (!IS_ERR_OR_NULL(p
) && !*pos
)
1168 seq_puts(seqf
, "major minor #blocks name\n\n");
1172 static int show_partition(struct seq_file
*seqf
, void *v
)
1174 struct gendisk
*sgp
= v
;
1175 struct disk_part_iter piter
;
1176 struct hd_struct
*part
;
1177 char buf
[BDEVNAME_SIZE
];
1179 /* Don't show non-partitionable removeable devices or empty devices */
1180 if (!get_capacity(sgp
) || (!disk_max_parts(sgp
) &&
1181 (sgp
->flags
& GENHD_FL_REMOVABLE
)))
1183 if (sgp
->flags
& GENHD_FL_SUPPRESS_PARTITION_INFO
)
1186 /* show the full disk and all non-0 size partitions of it */
1187 disk_part_iter_init(&piter
, sgp
, DISK_PITER_INCL_PART0
);
1188 while ((part
= disk_part_iter_next(&piter
)))
1189 seq_printf(seqf
, "%4d %7d %10llu %s\n",
1190 MAJOR(part_devt(part
)), MINOR(part_devt(part
)),
1191 (unsigned long long)part_nr_sects_read(part
) >> 1,
1192 disk_name(sgp
, part
->partno
, buf
));
1193 disk_part_iter_exit(&piter
);
1198 static const struct seq_operations partitions_op
= {
1199 .start
= show_partition_start
,
1200 .next
= disk_seqf_next
,
1201 .stop
= disk_seqf_stop
,
1202 .show
= show_partition
1207 static struct kobject
*base_probe(dev_t devt
, int *partno
, void *data
)
1209 if (request_module("block-major-%d-%d", MAJOR(devt
), MINOR(devt
)) > 0)
1210 /* Make old-style 2.4 aliases work */
1211 request_module("block-major-%d", MAJOR(devt
));
1215 static int __init
genhd_device_init(void)
1219 block_class
.dev_kobj
= sysfs_dev_block_kobj
;
1220 error
= class_register(&block_class
);
1221 if (unlikely(error
))
1223 bdev_map
= kobj_map_init(base_probe
, &block_class_lock
);
1226 register_blkdev(BLOCK_EXT_MAJOR
, "blkext");
1228 /* create top-level block dir */
1229 if (!sysfs_deprecated
)
1230 block_depr
= kobject_create_and_add("block", NULL
);
1234 subsys_initcall(genhd_device_init
);
1236 static ssize_t
disk_range_show(struct device
*dev
,
1237 struct device_attribute
*attr
, char *buf
)
1239 struct gendisk
*disk
= dev_to_disk(dev
);
1241 return sprintf(buf
, "%d\n", disk
->minors
);
1244 static ssize_t
disk_ext_range_show(struct device
*dev
,
1245 struct device_attribute
*attr
, char *buf
)
1247 struct gendisk
*disk
= dev_to_disk(dev
);
1249 return sprintf(buf
, "%d\n", disk_max_parts(disk
));
1252 static ssize_t
disk_removable_show(struct device
*dev
,
1253 struct device_attribute
*attr
, char *buf
)
1255 struct gendisk
*disk
= dev_to_disk(dev
);
1257 return sprintf(buf
, "%d\n",
1258 (disk
->flags
& GENHD_FL_REMOVABLE
? 1 : 0));
1261 static ssize_t
disk_hidden_show(struct device
*dev
,
1262 struct device_attribute
*attr
, char *buf
)
1264 struct gendisk
*disk
= dev_to_disk(dev
);
1266 return sprintf(buf
, "%d\n",
1267 (disk
->flags
& GENHD_FL_HIDDEN
? 1 : 0));
1270 static ssize_t
disk_ro_show(struct device
*dev
,
1271 struct device_attribute
*attr
, char *buf
)
1273 struct gendisk
*disk
= dev_to_disk(dev
);
1275 return sprintf(buf
, "%d\n", get_disk_ro(disk
) ? 1 : 0);
1278 ssize_t
part_size_show(struct device
*dev
,
1279 struct device_attribute
*attr
, char *buf
)
1281 struct hd_struct
*p
= dev_to_part(dev
);
1283 return sprintf(buf
, "%llu\n",
1284 (unsigned long long)part_nr_sects_read(p
));
1287 ssize_t
part_stat_show(struct device
*dev
,
1288 struct device_attribute
*attr
, char *buf
)
1290 struct hd_struct
*p
= dev_to_part(dev
);
1291 struct request_queue
*q
= part_to_disk(p
)->queue
;
1292 struct disk_stats stat
;
1293 unsigned int inflight
;
1295 part_stat_read_all(p
, &stat
);
1297 inflight
= blk_mq_in_flight(q
, p
);
1299 inflight
= part_in_flight(p
);
1302 "%8lu %8lu %8llu %8u "
1303 "%8lu %8lu %8llu %8u "
1305 "%8lu %8lu %8llu %8u "
1308 stat
.ios
[STAT_READ
],
1309 stat
.merges
[STAT_READ
],
1310 (unsigned long long)stat
.sectors
[STAT_READ
],
1311 (unsigned int)div_u64(stat
.nsecs
[STAT_READ
], NSEC_PER_MSEC
),
1312 stat
.ios
[STAT_WRITE
],
1313 stat
.merges
[STAT_WRITE
],
1314 (unsigned long long)stat
.sectors
[STAT_WRITE
],
1315 (unsigned int)div_u64(stat
.nsecs
[STAT_WRITE
], NSEC_PER_MSEC
),
1317 jiffies_to_msecs(stat
.io_ticks
),
1318 (unsigned int)div_u64(stat
.nsecs
[STAT_READ
] +
1319 stat
.nsecs
[STAT_WRITE
] +
1320 stat
.nsecs
[STAT_DISCARD
] +
1321 stat
.nsecs
[STAT_FLUSH
],
1323 stat
.ios
[STAT_DISCARD
],
1324 stat
.merges
[STAT_DISCARD
],
1325 (unsigned long long)stat
.sectors
[STAT_DISCARD
],
1326 (unsigned int)div_u64(stat
.nsecs
[STAT_DISCARD
], NSEC_PER_MSEC
),
1327 stat
.ios
[STAT_FLUSH
],
1328 (unsigned int)div_u64(stat
.nsecs
[STAT_FLUSH
], NSEC_PER_MSEC
));
1331 ssize_t
part_inflight_show(struct device
*dev
, struct device_attribute
*attr
,
1334 struct hd_struct
*p
= dev_to_part(dev
);
1335 struct request_queue
*q
= part_to_disk(p
)->queue
;
1336 unsigned int inflight
[2];
1339 blk_mq_in_flight_rw(q
, p
, inflight
);
1341 part_in_flight_rw(p
, inflight
);
1343 return sprintf(buf
, "%8u %8u\n", inflight
[0], inflight
[1]);
1346 static ssize_t
disk_capability_show(struct device
*dev
,
1347 struct device_attribute
*attr
, char *buf
)
1349 struct gendisk
*disk
= dev_to_disk(dev
);
1351 return sprintf(buf
, "%x\n", disk
->flags
);
1354 static ssize_t
disk_alignment_offset_show(struct device
*dev
,
1355 struct device_attribute
*attr
,
1358 struct gendisk
*disk
= dev_to_disk(dev
);
1360 return sprintf(buf
, "%d\n", queue_alignment_offset(disk
->queue
));
1363 static ssize_t
disk_discard_alignment_show(struct device
*dev
,
1364 struct device_attribute
*attr
,
1367 struct gendisk
*disk
= dev_to_disk(dev
);
1369 return sprintf(buf
, "%d\n", queue_discard_alignment(disk
->queue
));
1372 static DEVICE_ATTR(range
, 0444, disk_range_show
, NULL
);
1373 static DEVICE_ATTR(ext_range
, 0444, disk_ext_range_show
, NULL
);
1374 static DEVICE_ATTR(removable
, 0444, disk_removable_show
, NULL
);
1375 static DEVICE_ATTR(hidden
, 0444, disk_hidden_show
, NULL
);
1376 static DEVICE_ATTR(ro
, 0444, disk_ro_show
, NULL
);
1377 static DEVICE_ATTR(size
, 0444, part_size_show
, NULL
);
1378 static DEVICE_ATTR(alignment_offset
, 0444, disk_alignment_offset_show
, NULL
);
1379 static DEVICE_ATTR(discard_alignment
, 0444, disk_discard_alignment_show
, NULL
);
1380 static DEVICE_ATTR(capability
, 0444, disk_capability_show
, NULL
);
1381 static DEVICE_ATTR(stat
, 0444, part_stat_show
, NULL
);
1382 static DEVICE_ATTR(inflight
, 0444, part_inflight_show
, NULL
);
1383 static DEVICE_ATTR(badblocks
, 0644, disk_badblocks_show
, disk_badblocks_store
);
1385 #ifdef CONFIG_FAIL_MAKE_REQUEST
1386 ssize_t
part_fail_show(struct device
*dev
,
1387 struct device_attribute
*attr
, char *buf
)
1389 struct hd_struct
*p
= dev_to_part(dev
);
1391 return sprintf(buf
, "%d\n", p
->make_it_fail
);
1394 ssize_t
part_fail_store(struct device
*dev
,
1395 struct device_attribute
*attr
,
1396 const char *buf
, size_t count
)
1398 struct hd_struct
*p
= dev_to_part(dev
);
1401 if (count
> 0 && sscanf(buf
, "%d", &i
) > 0)
1402 p
->make_it_fail
= (i
== 0) ? 0 : 1;
1407 static struct device_attribute dev_attr_fail
=
1408 __ATTR(make
-it
-fail
, 0644, part_fail_show
, part_fail_store
);
1409 #endif /* CONFIG_FAIL_MAKE_REQUEST */
1411 #ifdef CONFIG_FAIL_IO_TIMEOUT
1412 static struct device_attribute dev_attr_fail_timeout
=
1413 __ATTR(io
-timeout
-fail
, 0644, part_timeout_show
, part_timeout_store
);
1416 static struct attribute
*disk_attrs
[] = {
1417 &dev_attr_range
.attr
,
1418 &dev_attr_ext_range
.attr
,
1419 &dev_attr_removable
.attr
,
1420 &dev_attr_hidden
.attr
,
1422 &dev_attr_size
.attr
,
1423 &dev_attr_alignment_offset
.attr
,
1424 &dev_attr_discard_alignment
.attr
,
1425 &dev_attr_capability
.attr
,
1426 &dev_attr_stat
.attr
,
1427 &dev_attr_inflight
.attr
,
1428 &dev_attr_badblocks
.attr
,
1429 #ifdef CONFIG_FAIL_MAKE_REQUEST
1430 &dev_attr_fail
.attr
,
1432 #ifdef CONFIG_FAIL_IO_TIMEOUT
1433 &dev_attr_fail_timeout
.attr
,
1438 static umode_t
disk_visible(struct kobject
*kobj
, struct attribute
*a
, int n
)
1440 struct device
*dev
= container_of(kobj
, typeof(*dev
), kobj
);
1441 struct gendisk
*disk
= dev_to_disk(dev
);
1443 if (a
== &dev_attr_badblocks
.attr
&& !disk
->bb
)
1448 static struct attribute_group disk_attr_group
= {
1449 .attrs
= disk_attrs
,
1450 .is_visible
= disk_visible
,
1453 static const struct attribute_group
*disk_attr_groups
[] = {
1459 * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way
1460 * @disk: disk to replace part_tbl for
1461 * @new_ptbl: new part_tbl to install
1463 * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The
1464 * original ptbl is freed using RCU callback.
1467 * Matching bd_mutex locked or the caller is the only user of @disk.
1469 static void disk_replace_part_tbl(struct gendisk
*disk
,
1470 struct disk_part_tbl
*new_ptbl
)
1472 struct disk_part_tbl
*old_ptbl
=
1473 rcu_dereference_protected(disk
->part_tbl
, 1);
1475 rcu_assign_pointer(disk
->part_tbl
, new_ptbl
);
1478 rcu_assign_pointer(old_ptbl
->last_lookup
, NULL
);
1479 kfree_rcu(old_ptbl
, rcu_head
);
1484 * disk_expand_part_tbl - expand disk->part_tbl
1485 * @disk: disk to expand part_tbl for
1486 * @partno: expand such that this partno can fit in
1488 * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl
1489 * uses RCU to allow unlocked dereferencing for stats and other stuff.
1492 * Matching bd_mutex locked or the caller is the only user of @disk.
1496 * 0 on success, -errno on failure.
1498 int disk_expand_part_tbl(struct gendisk
*disk
, int partno
)
1500 struct disk_part_tbl
*old_ptbl
=
1501 rcu_dereference_protected(disk
->part_tbl
, 1);
1502 struct disk_part_tbl
*new_ptbl
;
1503 int len
= old_ptbl
? old_ptbl
->len
: 0;
1507 * check for int overflow, since we can get here from blkpg_ioctl()
1508 * with a user passed 'partno'.
1510 target
= partno
+ 1;
1514 /* disk_max_parts() is zero during initialization, ignore if so */
1515 if (disk_max_parts(disk
) && target
> disk_max_parts(disk
))
1521 new_ptbl
= kzalloc_node(struct_size(new_ptbl
, part
, target
), GFP_KERNEL
,
1526 new_ptbl
->len
= target
;
1528 for (i
= 0; i
< len
; i
++)
1529 rcu_assign_pointer(new_ptbl
->part
[i
], old_ptbl
->part
[i
]);
1531 disk_replace_part_tbl(disk
, new_ptbl
);
1536 * disk_release - releases all allocated resources of the gendisk
1537 * @dev: the device representing this disk
1539 * This function releases all allocated resources of the gendisk.
1541 * The struct gendisk refcount is incremented with get_gendisk() or
1542 * get_disk_and_module(), and its refcount is decremented with
1543 * put_disk_and_module() or put_disk(). Once the refcount reaches 0 this
1544 * function is called.
1546 * Drivers which used __device_add_disk() have a gendisk with a request_queue
1547 * assigned. Since the request_queue sits on top of the gendisk for these
1548 * drivers we also call blk_put_queue() for them, and we expect the
1549 * request_queue refcount to reach 0 at this point, and so the request_queue
1550 * will also be freed prior to the disk.
1552 * Context: can sleep
1554 static void disk_release(struct device
*dev
)
1556 struct gendisk
*disk
= dev_to_disk(dev
);
1560 blk_free_devt(dev
->devt
);
1561 disk_release_events(disk
);
1562 kfree(disk
->random
);
1563 disk_replace_part_tbl(disk
, NULL
);
1564 hd_free_part(&disk
->part0
);
1566 blk_put_queue(disk
->queue
);
1569 struct class block_class
= {
1573 static char *block_devnode(struct device
*dev
, umode_t
*mode
,
1574 kuid_t
*uid
, kgid_t
*gid
)
1576 struct gendisk
*disk
= dev_to_disk(dev
);
1578 if (disk
->fops
->devnode
)
1579 return disk
->fops
->devnode(disk
, mode
);
1583 const struct device_type disk_type
= {
1585 .groups
= disk_attr_groups
,
1586 .release
= disk_release
,
1587 .devnode
= block_devnode
,
1590 #ifdef CONFIG_PROC_FS
1592 * aggregate disk stat collector. Uses the same stats that the sysfs
1593 * entries do, above, but makes them available through one seq_file.
1595 * The output looks suspiciously like /proc/partitions with a bunch of
1598 static int diskstats_show(struct seq_file
*seqf
, void *v
)
1600 struct gendisk
*gp
= v
;
1601 struct disk_part_iter piter
;
1602 struct hd_struct
*hd
;
1603 char buf
[BDEVNAME_SIZE
];
1604 unsigned int inflight
;
1605 struct disk_stats stat
;
1608 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1609 seq_puts(seqf, "major minor name"
1610 " rio rmerge rsect ruse wio wmerge "
1611 "wsect wuse running use aveq"
1615 disk_part_iter_init(&piter
, gp
, DISK_PITER_INCL_EMPTY_PART0
);
1616 while ((hd
= disk_part_iter_next(&piter
))) {
1617 part_stat_read_all(hd
, &stat
);
1618 if (queue_is_mq(gp
->queue
))
1619 inflight
= blk_mq_in_flight(gp
->queue
, hd
);
1621 inflight
= part_in_flight(hd
);
1623 seq_printf(seqf
, "%4d %7d %s "
1630 MAJOR(part_devt(hd
)), MINOR(part_devt(hd
)),
1631 disk_name(gp
, hd
->partno
, buf
),
1632 stat
.ios
[STAT_READ
],
1633 stat
.merges
[STAT_READ
],
1634 stat
.sectors
[STAT_READ
],
1635 (unsigned int)div_u64(stat
.nsecs
[STAT_READ
],
1637 stat
.ios
[STAT_WRITE
],
1638 stat
.merges
[STAT_WRITE
],
1639 stat
.sectors
[STAT_WRITE
],
1640 (unsigned int)div_u64(stat
.nsecs
[STAT_WRITE
],
1643 jiffies_to_msecs(stat
.io_ticks
),
1644 (unsigned int)div_u64(stat
.nsecs
[STAT_READ
] +
1645 stat
.nsecs
[STAT_WRITE
] +
1646 stat
.nsecs
[STAT_DISCARD
] +
1647 stat
.nsecs
[STAT_FLUSH
],
1649 stat
.ios
[STAT_DISCARD
],
1650 stat
.merges
[STAT_DISCARD
],
1651 stat
.sectors
[STAT_DISCARD
],
1652 (unsigned int)div_u64(stat
.nsecs
[STAT_DISCARD
],
1654 stat
.ios
[STAT_FLUSH
],
1655 (unsigned int)div_u64(stat
.nsecs
[STAT_FLUSH
],
1659 disk_part_iter_exit(&piter
);
1664 static const struct seq_operations diskstats_op
= {
1665 .start
= disk_seqf_start
,
1666 .next
= disk_seqf_next
,
1667 .stop
= disk_seqf_stop
,
1668 .show
= diskstats_show
1671 static int __init
proc_genhd_init(void)
1673 proc_create_seq("diskstats", 0, NULL
, &diskstats_op
);
1674 proc_create_seq("partitions", 0, NULL
, &partitions_op
);
1677 module_init(proc_genhd_init
);
1678 #endif /* CONFIG_PROC_FS */
1680 dev_t
blk_lookup_devt(const char *name
, int partno
)
1682 dev_t devt
= MKDEV(0, 0);
1683 struct class_dev_iter iter
;
1686 class_dev_iter_init(&iter
, &block_class
, NULL
, &disk_type
);
1687 while ((dev
= class_dev_iter_next(&iter
))) {
1688 struct gendisk
*disk
= dev_to_disk(dev
);
1689 struct hd_struct
*part
;
1691 if (strcmp(dev_name(dev
), name
))
1694 if (partno
< disk
->minors
) {
1695 /* We need to return the right devno, even
1696 * if the partition doesn't exist yet.
1698 devt
= MKDEV(MAJOR(dev
->devt
),
1699 MINOR(dev
->devt
) + partno
);
1702 part
= disk_get_part(disk
, partno
);
1704 devt
= part_devt(part
);
1705 disk_put_part(part
);
1708 disk_put_part(part
);
1710 class_dev_iter_exit(&iter
);
1714 struct gendisk
*__alloc_disk_node(int minors
, int node_id
)
1716 struct gendisk
*disk
;
1717 struct disk_part_tbl
*ptbl
;
1719 if (minors
> DISK_MAX_PARTS
) {
1721 "block: can't allocate more than %d partitions\n",
1723 minors
= DISK_MAX_PARTS
;
1726 disk
= kzalloc_node(sizeof(struct gendisk
), GFP_KERNEL
, node_id
);
1730 disk
->part0
.dkstats
= alloc_percpu(struct disk_stats
);
1731 if (!disk
->part0
.dkstats
)
1734 init_rwsem(&disk
->lookup_sem
);
1735 disk
->node_id
= node_id
;
1736 if (disk_expand_part_tbl(disk
, 0)) {
1737 free_percpu(disk
->part0
.dkstats
);
1741 ptbl
= rcu_dereference_protected(disk
->part_tbl
, 1);
1742 rcu_assign_pointer(ptbl
->part
[0], &disk
->part0
);
1745 * set_capacity() and get_capacity() currently don't use
1746 * seqcounter to read/update the part0->nr_sects. Still init
1747 * the counter as we can read the sectors in IO submission
1748 * patch using seqence counters.
1750 * TODO: Ideally set_capacity() and get_capacity() should be
1751 * converted to make use of bd_mutex and sequence counters.
1753 hd_sects_seq_init(&disk
->part0
);
1754 if (hd_ref_init(&disk
->part0
))
1755 goto out_free_part0
;
1757 disk
->minors
= minors
;
1758 rand_initialize_disk(disk
);
1759 disk_to_dev(disk
)->class = &block_class
;
1760 disk_to_dev(disk
)->type
= &disk_type
;
1761 device_initialize(disk_to_dev(disk
));
1765 hd_free_part(&disk
->part0
);
1770 EXPORT_SYMBOL(__alloc_disk_node
);
1773 * get_disk_and_module - increments the gendisk and gendisk fops module refcount
1774 * @disk: the struct gendisk to increment the refcount for
1776 * This increments the refcount for the struct gendisk, and the gendisk's
1777 * fops module owner.
1779 * Context: Any context.
1781 struct kobject
*get_disk_and_module(struct gendisk
*disk
)
1783 struct module
*owner
;
1784 struct kobject
*kobj
;
1788 owner
= disk
->fops
->owner
;
1789 if (owner
&& !try_module_get(owner
))
1791 kobj
= kobject_get_unless_zero(&disk_to_dev(disk
)->kobj
);
1799 EXPORT_SYMBOL(get_disk_and_module
);
1802 * put_disk - decrements the gendisk refcount
1803 * @disk: the struct gendisk to decrement the refcount for
1805 * This decrements the refcount for the struct gendisk. When this reaches 0
1806 * we'll have disk_release() called.
1808 * Context: Any context, but the last reference must not be dropped from
1811 void put_disk(struct gendisk
*disk
)
1814 kobject_put(&disk_to_dev(disk
)->kobj
);
1816 EXPORT_SYMBOL(put_disk
);
1819 * put_disk_and_module - decrements the module and gendisk refcount
1820 * @disk: the struct gendisk to decrement the refcount for
1822 * This is a counterpart of get_disk_and_module() and thus also of
1825 * Context: Any context, but the last reference must not be dropped from
1828 void put_disk_and_module(struct gendisk
*disk
)
1831 struct module
*owner
= disk
->fops
->owner
;
1837 EXPORT_SYMBOL(put_disk_and_module
);
1839 static void set_disk_ro_uevent(struct gendisk
*gd
, int ro
)
1841 char event
[] = "DISK_RO=1";
1842 char *envp
[] = { event
, NULL
};
1846 kobject_uevent_env(&disk_to_dev(gd
)->kobj
, KOBJ_CHANGE
, envp
);
1849 void set_device_ro(struct block_device
*bdev
, int flag
)
1851 bdev
->bd_part
->policy
= flag
;
1854 EXPORT_SYMBOL(set_device_ro
);
1856 void set_disk_ro(struct gendisk
*disk
, int flag
)
1858 struct disk_part_iter piter
;
1859 struct hd_struct
*part
;
1861 if (disk
->part0
.policy
!= flag
) {
1862 set_disk_ro_uevent(disk
, flag
);
1863 disk
->part0
.policy
= flag
;
1866 disk_part_iter_init(&piter
, disk
, DISK_PITER_INCL_EMPTY
);
1867 while ((part
= disk_part_iter_next(&piter
)))
1868 part
->policy
= flag
;
1869 disk_part_iter_exit(&piter
);
1872 EXPORT_SYMBOL(set_disk_ro
);
1874 int bdev_read_only(struct block_device
*bdev
)
1878 return bdev
->bd_part
->policy
;
1881 EXPORT_SYMBOL(bdev_read_only
);
1884 * Disk events - monitor disk events like media change and eject request.
1886 struct disk_events
{
1887 struct list_head node
; /* all disk_event's */
1888 struct gendisk
*disk
; /* the associated disk */
1891 struct mutex block_mutex
; /* protects blocking */
1892 int block
; /* event blocking depth */
1893 unsigned int pending
; /* events already sent out */
1894 unsigned int clearing
; /* events being cleared */
1896 long poll_msecs
; /* interval, -1 for default */
1897 struct delayed_work dwork
;
1900 static const char *disk_events_strs
[] = {
1901 [ilog2(DISK_EVENT_MEDIA_CHANGE
)] = "media_change",
1902 [ilog2(DISK_EVENT_EJECT_REQUEST
)] = "eject_request",
1905 static char *disk_uevents
[] = {
1906 [ilog2(DISK_EVENT_MEDIA_CHANGE
)] = "DISK_MEDIA_CHANGE=1",
1907 [ilog2(DISK_EVENT_EJECT_REQUEST
)] = "DISK_EJECT_REQUEST=1",
1910 /* list of all disk_events */
1911 static DEFINE_MUTEX(disk_events_mutex
);
1912 static LIST_HEAD(disk_events
);
1914 /* disable in-kernel polling by default */
1915 static unsigned long disk_events_dfl_poll_msecs
;
1917 static unsigned long disk_events_poll_jiffies(struct gendisk
*disk
)
1919 struct disk_events
*ev
= disk
->ev
;
1920 long intv_msecs
= 0;
1923 * If device-specific poll interval is set, always use it. If
1924 * the default is being used, poll if the POLL flag is set.
1926 if (ev
->poll_msecs
>= 0)
1927 intv_msecs
= ev
->poll_msecs
;
1928 else if (disk
->event_flags
& DISK_EVENT_FLAG_POLL
)
1929 intv_msecs
= disk_events_dfl_poll_msecs
;
1931 return msecs_to_jiffies(intv_msecs
);
1935 * disk_block_events - block and flush disk event checking
1936 * @disk: disk to block events for
1938 * On return from this function, it is guaranteed that event checking
1939 * isn't in progress and won't happen until unblocked by
1940 * disk_unblock_events(). Events blocking is counted and the actual
1941 * unblocking happens after the matching number of unblocks are done.
1943 * Note that this intentionally does not block event checking from
1944 * disk_clear_events().
1949 void disk_block_events(struct gendisk
*disk
)
1951 struct disk_events
*ev
= disk
->ev
;
1952 unsigned long flags
;
1959 * Outer mutex ensures that the first blocker completes canceling
1960 * the event work before further blockers are allowed to finish.
1962 mutex_lock(&ev
->block_mutex
);
1964 spin_lock_irqsave(&ev
->lock
, flags
);
1965 cancel
= !ev
->block
++;
1966 spin_unlock_irqrestore(&ev
->lock
, flags
);
1969 cancel_delayed_work_sync(&disk
->ev
->dwork
);
1971 mutex_unlock(&ev
->block_mutex
);
1974 static void __disk_unblock_events(struct gendisk
*disk
, bool check_now
)
1976 struct disk_events
*ev
= disk
->ev
;
1978 unsigned long flags
;
1980 spin_lock_irqsave(&ev
->lock
, flags
);
1982 if (WARN_ON_ONCE(ev
->block
<= 0))
1988 intv
= disk_events_poll_jiffies(disk
);
1990 queue_delayed_work(system_freezable_power_efficient_wq
,
1993 queue_delayed_work(system_freezable_power_efficient_wq
,
1996 spin_unlock_irqrestore(&ev
->lock
, flags
);
2000 * disk_unblock_events - unblock disk event checking
2001 * @disk: disk to unblock events for
2003 * Undo disk_block_events(). When the block count reaches zero, it
2004 * starts events polling if configured.
2007 * Don't care. Safe to call from irq context.
2009 void disk_unblock_events(struct gendisk
*disk
)
2012 __disk_unblock_events(disk
, false);
2016 * disk_flush_events - schedule immediate event checking and flushing
2017 * @disk: disk to check and flush events for
2018 * @mask: events to flush
2020 * Schedule immediate event checking on @disk if not blocked. Events in
2021 * @mask are scheduled to be cleared from the driver. Note that this
2022 * doesn't clear the events from @disk->ev.
2025 * If @mask is non-zero must be called with bdev->bd_mutex held.
2027 void disk_flush_events(struct gendisk
*disk
, unsigned int mask
)
2029 struct disk_events
*ev
= disk
->ev
;
2034 spin_lock_irq(&ev
->lock
);
2035 ev
->clearing
|= mask
;
2037 mod_delayed_work(system_freezable_power_efficient_wq
,
2039 spin_unlock_irq(&ev
->lock
);
2043 * disk_clear_events - synchronously check, clear and return pending events
2044 * @disk: disk to fetch and clear events from
2045 * @mask: mask of events to be fetched and cleared
2047 * Disk events are synchronously checked and pending events in @mask
2048 * are cleared and returned. This ignores the block count.
2053 static unsigned int disk_clear_events(struct gendisk
*disk
, unsigned int mask
)
2055 struct disk_events
*ev
= disk
->ev
;
2056 unsigned int pending
;
2057 unsigned int clearing
= mask
;
2062 disk_block_events(disk
);
2065 * store the union of mask and ev->clearing on the stack so that the
2066 * race with disk_flush_events does not cause ambiguity (ev->clearing
2067 * can still be modified even if events are blocked).
2069 spin_lock_irq(&ev
->lock
);
2070 clearing
|= ev
->clearing
;
2072 spin_unlock_irq(&ev
->lock
);
2074 disk_check_events(ev
, &clearing
);
2076 * if ev->clearing is not 0, the disk_flush_events got called in the
2077 * middle of this function, so we want to run the workfn without delay.
2079 __disk_unblock_events(disk
, ev
->clearing
? true : false);
2081 /* then, fetch and clear pending events */
2082 spin_lock_irq(&ev
->lock
);
2083 pending
= ev
->pending
& mask
;
2084 ev
->pending
&= ~mask
;
2085 spin_unlock_irq(&ev
->lock
);
2086 WARN_ON_ONCE(clearing
& mask
);
2092 * bdev_check_media_change - check if a removable media has been changed
2093 * @bdev: block device to check
2095 * Check whether a removable media has been changed, and attempt to free all
2096 * dentries and inodes and invalidates all block device page cache entries in
2099 * Returns %true if the block device changed, or %false if not.
2101 bool bdev_check_media_change(struct block_device
*bdev
)
2103 unsigned int events
;
2105 events
= disk_clear_events(bdev
->bd_disk
, DISK_EVENT_MEDIA_CHANGE
|
2106 DISK_EVENT_EJECT_REQUEST
);
2107 if (!(events
& DISK_EVENT_MEDIA_CHANGE
))
2110 if (__invalidate_device(bdev
, true))
2111 pr_warn("VFS: busy inodes on changed media %s\n",
2112 bdev
->bd_disk
->disk_name
);
2113 set_bit(GD_NEED_PART_SCAN
, &bdev
->bd_disk
->state
);
2116 EXPORT_SYMBOL(bdev_check_media_change
);
2119 * Separate this part out so that a different pointer for clearing_ptr can be
2120 * passed in for disk_clear_events.
2122 static void disk_events_workfn(struct work_struct
*work
)
2124 struct delayed_work
*dwork
= to_delayed_work(work
);
2125 struct disk_events
*ev
= container_of(dwork
, struct disk_events
, dwork
);
2127 disk_check_events(ev
, &ev
->clearing
);
2130 static void disk_check_events(struct disk_events
*ev
,
2131 unsigned int *clearing_ptr
)
2133 struct gendisk
*disk
= ev
->disk
;
2134 char *envp
[ARRAY_SIZE(disk_uevents
) + 1] = { };
2135 unsigned int clearing
= *clearing_ptr
;
2136 unsigned int events
;
2138 int nr_events
= 0, i
;
2141 events
= disk
->fops
->check_events(disk
, clearing
);
2143 /* accumulate pending events and schedule next poll if necessary */
2144 spin_lock_irq(&ev
->lock
);
2146 events
&= ~ev
->pending
;
2147 ev
->pending
|= events
;
2148 *clearing_ptr
&= ~clearing
;
2150 intv
= disk_events_poll_jiffies(disk
);
2151 if (!ev
->block
&& intv
)
2152 queue_delayed_work(system_freezable_power_efficient_wq
,
2155 spin_unlock_irq(&ev
->lock
);
2158 * Tell userland about new events. Only the events listed in
2159 * @disk->events are reported, and only if DISK_EVENT_FLAG_UEVENT
2160 * is set. Otherwise, events are processed internally but never
2161 * get reported to userland.
2163 for (i
= 0; i
< ARRAY_SIZE(disk_uevents
); i
++)
2164 if ((events
& disk
->events
& (1 << i
)) &&
2165 (disk
->event_flags
& DISK_EVENT_FLAG_UEVENT
))
2166 envp
[nr_events
++] = disk_uevents
[i
];
2169 kobject_uevent_env(&disk_to_dev(disk
)->kobj
, KOBJ_CHANGE
, envp
);
2173 * A disk events enabled device has the following sysfs nodes under
2174 * its /sys/block/X/ directory.
2176 * events : list of all supported events
2177 * events_async : list of events which can be detected w/o polling
2178 * (always empty, only for backwards compatibility)
2179 * events_poll_msecs : polling interval, 0: disable, -1: system default
2181 static ssize_t
__disk_events_show(unsigned int events
, char *buf
)
2183 const char *delim
= "";
2187 for (i
= 0; i
< ARRAY_SIZE(disk_events_strs
); i
++)
2188 if (events
& (1 << i
)) {
2189 pos
+= sprintf(buf
+ pos
, "%s%s",
2190 delim
, disk_events_strs
[i
]);
2194 pos
+= sprintf(buf
+ pos
, "\n");
2198 static ssize_t
disk_events_show(struct device
*dev
,
2199 struct device_attribute
*attr
, char *buf
)
2201 struct gendisk
*disk
= dev_to_disk(dev
);
2203 if (!(disk
->event_flags
& DISK_EVENT_FLAG_UEVENT
))
2206 return __disk_events_show(disk
->events
, buf
);
2209 static ssize_t
disk_events_async_show(struct device
*dev
,
2210 struct device_attribute
*attr
, char *buf
)
2215 static ssize_t
disk_events_poll_msecs_show(struct device
*dev
,
2216 struct device_attribute
*attr
,
2219 struct gendisk
*disk
= dev_to_disk(dev
);
2222 return sprintf(buf
, "-1\n");
2224 return sprintf(buf
, "%ld\n", disk
->ev
->poll_msecs
);
2227 static ssize_t
disk_events_poll_msecs_store(struct device
*dev
,
2228 struct device_attribute
*attr
,
2229 const char *buf
, size_t count
)
2231 struct gendisk
*disk
= dev_to_disk(dev
);
2234 if (!count
|| !sscanf(buf
, "%ld", &intv
))
2237 if (intv
< 0 && intv
!= -1)
2243 disk_block_events(disk
);
2244 disk
->ev
->poll_msecs
= intv
;
2245 __disk_unblock_events(disk
, true);
2250 static const DEVICE_ATTR(events
, 0444, disk_events_show
, NULL
);
2251 static const DEVICE_ATTR(events_async
, 0444, disk_events_async_show
, NULL
);
2252 static const DEVICE_ATTR(events_poll_msecs
, 0644,
2253 disk_events_poll_msecs_show
,
2254 disk_events_poll_msecs_store
);
2256 static const struct attribute
*disk_events_attrs
[] = {
2257 &dev_attr_events
.attr
,
2258 &dev_attr_events_async
.attr
,
2259 &dev_attr_events_poll_msecs
.attr
,
2264 * The default polling interval can be specified by the kernel
2265 * parameter block.events_dfl_poll_msecs which defaults to 0
2266 * (disable). This can also be modified runtime by writing to
2267 * /sys/module/block/parameters/events_dfl_poll_msecs.
2269 static int disk_events_set_dfl_poll_msecs(const char *val
,
2270 const struct kernel_param
*kp
)
2272 struct disk_events
*ev
;
2275 ret
= param_set_ulong(val
, kp
);
2279 mutex_lock(&disk_events_mutex
);
2281 list_for_each_entry(ev
, &disk_events
, node
)
2282 disk_flush_events(ev
->disk
, 0);
2284 mutex_unlock(&disk_events_mutex
);
2289 static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops
= {
2290 .set
= disk_events_set_dfl_poll_msecs
,
2291 .get
= param_get_ulong
,
2294 #undef MODULE_PARAM_PREFIX
2295 #define MODULE_PARAM_PREFIX "block."
2297 module_param_cb(events_dfl_poll_msecs
, &disk_events_dfl_poll_msecs_param_ops
,
2298 &disk_events_dfl_poll_msecs
, 0644);
2301 * disk_{alloc|add|del|release}_events - initialize and destroy disk_events.
2303 static void disk_alloc_events(struct gendisk
*disk
)
2305 struct disk_events
*ev
;
2307 if (!disk
->fops
->check_events
|| !disk
->events
)
2310 ev
= kzalloc(sizeof(*ev
), GFP_KERNEL
);
2312 pr_warn("%s: failed to initialize events\n", disk
->disk_name
);
2316 INIT_LIST_HEAD(&ev
->node
);
2318 spin_lock_init(&ev
->lock
);
2319 mutex_init(&ev
->block_mutex
);
2321 ev
->poll_msecs
= -1;
2322 INIT_DELAYED_WORK(&ev
->dwork
, disk_events_workfn
);
2327 static void disk_add_events(struct gendisk
*disk
)
2329 /* FIXME: error handling */
2330 if (sysfs_create_files(&disk_to_dev(disk
)->kobj
, disk_events_attrs
) < 0)
2331 pr_warn("%s: failed to create sysfs files for events\n",
2337 mutex_lock(&disk_events_mutex
);
2338 list_add_tail(&disk
->ev
->node
, &disk_events
);
2339 mutex_unlock(&disk_events_mutex
);
2342 * Block count is initialized to 1 and the following initial
2343 * unblock kicks it into action.
2345 __disk_unblock_events(disk
, true);
2348 static void disk_del_events(struct gendisk
*disk
)
2351 disk_block_events(disk
);
2353 mutex_lock(&disk_events_mutex
);
2354 list_del_init(&disk
->ev
->node
);
2355 mutex_unlock(&disk_events_mutex
);
2358 sysfs_remove_files(&disk_to_dev(disk
)->kobj
, disk_events_attrs
);
2361 static void disk_release_events(struct gendisk
*disk
)
2363 /* the block count should be 1 from disk_del_events() */
2364 WARN_ON_ONCE(disk
->ev
&& disk
->ev
->block
!= 1);