1 // SPDX-License-Identifier: GPL-2.0
5 * Portions Copyright (C) 2020 Christoph Hellwig
8 #include <linux/module.h>
9 #include <linux/ctype.h>
11 #include <linux/genhd.h>
12 #include <linux/kdev_t.h>
13 #include <linux/kernel.h>
14 #include <linux/blkdev.h>
15 #include <linux/backing-dev.h>
16 #include <linux/init.h>
17 #include <linux/spinlock.h>
18 #include <linux/proc_fs.h>
19 #include <linux/seq_file.h>
20 #include <linux/slab.h>
21 #include <linux/kmod.h>
22 #include <linux/mutex.h>
23 #include <linux/idr.h>
24 #include <linux/log2.h>
25 #include <linux/pm_runtime.h>
26 #include <linux/badblocks.h>
30 static struct kobject
*block_depr
;
32 DECLARE_RWSEM(bdev_lookup_sem
);
34 /* for extended dynamic devt allocation, currently only one major is used */
35 #define NR_EXT_DEVT (1 << MINORBITS)
36 static DEFINE_IDA(ext_devt_ida
);
38 static void disk_check_events(struct disk_events
*ev
,
39 unsigned int *clearing_ptr
);
40 static void disk_alloc_events(struct gendisk
*disk
);
41 static void disk_add_events(struct gendisk
*disk
);
42 static void disk_del_events(struct gendisk
*disk
);
43 static void disk_release_events(struct gendisk
*disk
);
45 void set_capacity(struct gendisk
*disk
, sector_t sectors
)
47 struct block_device
*bdev
= disk
->part0
;
50 spin_lock_irqsave(&bdev
->bd_size_lock
, flags
);
51 i_size_write(bdev
->bd_inode
, (loff_t
)sectors
<< SECTOR_SHIFT
);
52 spin_unlock_irqrestore(&bdev
->bd_size_lock
, flags
);
54 EXPORT_SYMBOL(set_capacity
);
57 * Set disk capacity and notify if the size is not currently zero and will not
58 * be set to zero. Returns true if a uevent was sent, otherwise false.
60 bool set_capacity_and_notify(struct gendisk
*disk
, sector_t size
)
62 sector_t capacity
= get_capacity(disk
);
63 char *envp
[] = { "RESIZE=1", NULL
};
65 set_capacity(disk
, size
);
68 * Only print a message and send a uevent if the gendisk is user visible
69 * and alive. This avoids spamming the log and udev when setting the
70 * initial capacity during probing.
72 if (size
== capacity
||
73 (disk
->flags
& (GENHD_FL_UP
| GENHD_FL_HIDDEN
)) != GENHD_FL_UP
)
76 pr_info("%s: detected capacity change from %lld to %lld\n",
77 disk
->disk_name
, capacity
, size
);
80 * Historically we did not send a uevent for changes to/from an empty
83 if (!capacity
|| !size
)
85 kobject_uevent_env(&disk_to_dev(disk
)->kobj
, KOBJ_CHANGE
, envp
);
88 EXPORT_SYMBOL_GPL(set_capacity_and_notify
);
91 * Format the device name of the indicated disk into the supplied buffer and
92 * return a pointer to that same buffer for convenience.
94 char *disk_name(struct gendisk
*hd
, int partno
, char *buf
)
97 snprintf(buf
, BDEVNAME_SIZE
, "%s", hd
->disk_name
);
98 else if (isdigit(hd
->disk_name
[strlen(hd
->disk_name
)-1]))
99 snprintf(buf
, BDEVNAME_SIZE
, "%sp%d", hd
->disk_name
, partno
);
101 snprintf(buf
, BDEVNAME_SIZE
, "%s%d", hd
->disk_name
, partno
);
106 const char *bdevname(struct block_device
*bdev
, char *buf
)
108 return disk_name(bdev
->bd_disk
, bdev
->bd_partno
, buf
);
110 EXPORT_SYMBOL(bdevname
);
112 static void part_stat_read_all(struct block_device
*part
,
113 struct disk_stats
*stat
)
117 memset(stat
, 0, sizeof(struct disk_stats
));
118 for_each_possible_cpu(cpu
) {
119 struct disk_stats
*ptr
= per_cpu_ptr(part
->bd_stats
, cpu
);
122 for (group
= 0; group
< NR_STAT_GROUPS
; group
++) {
123 stat
->nsecs
[group
] += ptr
->nsecs
[group
];
124 stat
->sectors
[group
] += ptr
->sectors
[group
];
125 stat
->ios
[group
] += ptr
->ios
[group
];
126 stat
->merges
[group
] += ptr
->merges
[group
];
129 stat
->io_ticks
+= ptr
->io_ticks
;
133 static unsigned int part_in_flight(struct block_device
*part
)
135 unsigned int inflight
= 0;
138 for_each_possible_cpu(cpu
) {
139 inflight
+= part_stat_local_read_cpu(part
, in_flight
[0], cpu
) +
140 part_stat_local_read_cpu(part
, in_flight
[1], cpu
);
142 if ((int)inflight
< 0)
148 static void part_in_flight_rw(struct block_device
*part
,
149 unsigned int inflight
[2])
155 for_each_possible_cpu(cpu
) {
156 inflight
[0] += part_stat_local_read_cpu(part
, in_flight
[0], cpu
);
157 inflight
[1] += part_stat_local_read_cpu(part
, in_flight
[1], cpu
);
159 if ((int)inflight
[0] < 0)
161 if ((int)inflight
[1] < 0)
165 struct block_device
*__disk_get_part(struct gendisk
*disk
, int partno
)
167 struct disk_part_tbl
*ptbl
= rcu_dereference(disk
->part_tbl
);
169 if (unlikely(partno
< 0 || partno
>= ptbl
->len
))
171 return rcu_dereference(ptbl
->part
[partno
]);
175 * disk_part_iter_init - initialize partition iterator
176 * @piter: iterator to initialize
177 * @disk: disk to iterate over
178 * @flags: DISK_PITER_* flags
180 * Initialize @piter so that it iterates over partitions of @disk.
185 void disk_part_iter_init(struct disk_part_iter
*piter
, struct gendisk
*disk
,
188 struct disk_part_tbl
*ptbl
;
191 ptbl
= rcu_dereference(disk
->part_tbl
);
196 if (flags
& DISK_PITER_REVERSE
)
197 piter
->idx
= ptbl
->len
- 1;
198 else if (flags
& (DISK_PITER_INCL_PART0
| DISK_PITER_INCL_EMPTY_PART0
))
203 piter
->flags
= flags
;
207 EXPORT_SYMBOL_GPL(disk_part_iter_init
);
210 * disk_part_iter_next - proceed iterator to the next partition and return it
211 * @piter: iterator of interest
213 * Proceed @piter to the next partition and return it.
218 struct block_device
*disk_part_iter_next(struct disk_part_iter
*piter
)
220 struct disk_part_tbl
*ptbl
;
223 /* put the last partition */
224 disk_part_iter_exit(piter
);
228 ptbl
= rcu_dereference(piter
->disk
->part_tbl
);
230 /* determine iteration parameters */
231 if (piter
->flags
& DISK_PITER_REVERSE
) {
233 if (piter
->flags
& (DISK_PITER_INCL_PART0
|
234 DISK_PITER_INCL_EMPTY_PART0
))
243 /* iterate to the next partition */
244 for (; piter
->idx
!= end
; piter
->idx
+= inc
) {
245 struct block_device
*part
;
247 part
= rcu_dereference(ptbl
->part
[piter
->idx
]);
250 piter
->part
= bdgrab(part
);
253 if (!bdev_nr_sectors(part
) &&
254 !(piter
->flags
& DISK_PITER_INCL_EMPTY
) &&
255 !(piter
->flags
& DISK_PITER_INCL_EMPTY_PART0
&&
270 EXPORT_SYMBOL_GPL(disk_part_iter_next
);
273 * disk_part_iter_exit - finish up partition iteration
274 * @piter: iter of interest
276 * Called when iteration is over. Cleans up @piter.
281 void disk_part_iter_exit(struct disk_part_iter
*piter
)
287 EXPORT_SYMBOL_GPL(disk_part_iter_exit
);
289 static inline int sector_in_part(struct block_device
*part
, sector_t sector
)
291 return part
->bd_start_sect
<= sector
&&
292 sector
< part
->bd_start_sect
+ bdev_nr_sectors(part
);
296 * disk_map_sector_rcu - map sector to partition
297 * @disk: gendisk of interest
298 * @sector: sector to map
300 * Find out which partition @sector maps to on @disk. This is
301 * primarily used for stats accounting.
307 * Found partition on success, part0 is returned if no partition matches
308 * or the matched partition is being deleted.
310 struct block_device
*disk_map_sector_rcu(struct gendisk
*disk
, sector_t sector
)
312 struct disk_part_tbl
*ptbl
;
313 struct block_device
*part
;
317 ptbl
= rcu_dereference(disk
->part_tbl
);
319 part
= rcu_dereference(ptbl
->last_lookup
);
320 if (part
&& sector_in_part(part
, sector
))
323 for (i
= 1; i
< ptbl
->len
; i
++) {
324 part
= rcu_dereference(ptbl
->part
[i
]);
325 if (part
&& sector_in_part(part
, sector
)) {
326 rcu_assign_pointer(ptbl
->last_lookup
, part
);
336 EXPORT_SYMBOL_GPL(disk_map_sector_rcu
);
339 * disk_has_partitions
340 * @disk: gendisk of interest
342 * Walk through the partition table and check if valid partition exists.
348 * True if the gendisk has at least one valid non-zero size partition.
351 bool disk_has_partitions(struct gendisk
*disk
)
353 struct disk_part_tbl
*ptbl
;
358 ptbl
= rcu_dereference(disk
->part_tbl
);
360 /* Iterate partitions skipping the whole device at index 0 */
361 for (i
= 1; i
< ptbl
->len
; i
++) {
362 if (rcu_dereference(ptbl
->part
[i
])) {
372 EXPORT_SYMBOL_GPL(disk_has_partitions
);
375 * Can be deleted altogether. Later.
378 #define BLKDEV_MAJOR_HASH_SIZE 255
379 static struct blk_major_name
{
380 struct blk_major_name
*next
;
383 void (*probe
)(dev_t devt
);
384 } *major_names
[BLKDEV_MAJOR_HASH_SIZE
];
385 static DEFINE_MUTEX(major_names_lock
);
387 /* index in the above - for now: assume no multimajor ranges */
388 static inline int major_to_index(unsigned major
)
390 return major
% BLKDEV_MAJOR_HASH_SIZE
;
393 #ifdef CONFIG_PROC_FS
394 void blkdev_show(struct seq_file
*seqf
, off_t offset
)
396 struct blk_major_name
*dp
;
398 mutex_lock(&major_names_lock
);
399 for (dp
= major_names
[major_to_index(offset
)]; dp
; dp
= dp
->next
)
400 if (dp
->major
== offset
)
401 seq_printf(seqf
, "%3d %s\n", dp
->major
, dp
->name
);
402 mutex_unlock(&major_names_lock
);
404 #endif /* CONFIG_PROC_FS */
407 * __register_blkdev - register a new block device
409 * @major: the requested major device number [1..BLKDEV_MAJOR_MAX-1]. If
410 * @major = 0, try to allocate any unused major number.
411 * @name: the name of the new block device as a zero terminated string
412 * @probe: allback that is called on access to any minor number of @major
414 * The @name must be unique within the system.
416 * The return value depends on the @major input parameter:
418 * - if a major device number was requested in range [1..BLKDEV_MAJOR_MAX-1]
419 * then the function returns zero on success, or a negative error code
420 * - if any unused major number was requested with @major = 0 parameter
421 * then the return value is the allocated major number in range
422 * [1..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise
424 * See Documentation/admin-guide/devices.txt for the list of allocated
427 * Use register_blkdev instead for any new code.
429 int __register_blkdev(unsigned int major
, const char *name
,
430 void (*probe
)(dev_t devt
))
432 struct blk_major_name
**n
, *p
;
435 mutex_lock(&major_names_lock
);
439 for (index
= ARRAY_SIZE(major_names
)-1; index
> 0; index
--) {
440 if (major_names
[index
] == NULL
)
445 printk("%s: failed to get major for %s\n",
454 if (major
>= BLKDEV_MAJOR_MAX
) {
455 pr_err("%s: major requested (%u) is greater than the maximum (%u) for %s\n",
456 __func__
, major
, BLKDEV_MAJOR_MAX
-1, name
);
462 p
= kmalloc(sizeof(struct blk_major_name
), GFP_KERNEL
);
470 strlcpy(p
->name
, name
, sizeof(p
->name
));
472 index
= major_to_index(major
);
474 for (n
= &major_names
[index
]; *n
; n
= &(*n
)->next
) {
475 if ((*n
)->major
== major
)
484 printk("register_blkdev: cannot get major %u for %s\n",
489 mutex_unlock(&major_names_lock
);
492 EXPORT_SYMBOL(__register_blkdev
);
494 void unregister_blkdev(unsigned int major
, const char *name
)
496 struct blk_major_name
**n
;
497 struct blk_major_name
*p
= NULL
;
498 int index
= major_to_index(major
);
500 mutex_lock(&major_names_lock
);
501 for (n
= &major_names
[index
]; *n
; n
= &(*n
)->next
)
502 if ((*n
)->major
== major
)
504 if (!*n
|| strcmp((*n
)->name
, name
)) {
510 mutex_unlock(&major_names_lock
);
514 EXPORT_SYMBOL(unregister_blkdev
);
517 * blk_mangle_minor - scatter minor numbers apart
518 * @minor: minor number to mangle
520 * Scatter consecutively allocated @minor number apart if MANGLE_DEVT
521 * is enabled. Mangling twice gives the original value.
529 static int blk_mangle_minor(int minor
)
531 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
534 for (i
= 0; i
< MINORBITS
/ 2; i
++) {
535 int low
= minor
& (1 << i
);
536 int high
= minor
& (1 << (MINORBITS
- 1 - i
));
537 int distance
= MINORBITS
- 1 - 2 * i
;
539 minor
^= low
| high
; /* clear both bits */
540 low
<<= distance
; /* swap the positions */
542 minor
|= low
| high
; /* and set */
549 * blk_alloc_devt - allocate a dev_t for a block device
550 * @bdev: block device to allocate dev_t for
551 * @devt: out parameter for resulting dev_t
553 * Allocate a dev_t for block device.
556 * 0 on success, allocated dev_t is returned in *@devt. -errno on
562 int blk_alloc_devt(struct block_device
*bdev
, dev_t
*devt
)
564 struct gendisk
*disk
= bdev
->bd_disk
;
567 /* in consecutive minor range? */
568 if (bdev
->bd_partno
< disk
->minors
) {
569 *devt
= MKDEV(disk
->major
, disk
->first_minor
+ bdev
->bd_partno
);
573 idx
= ida_alloc_range(&ext_devt_ida
, 0, NR_EXT_DEVT
, GFP_KERNEL
);
575 return idx
== -ENOSPC
? -EBUSY
: idx
;
577 *devt
= MKDEV(BLOCK_EXT_MAJOR
, blk_mangle_minor(idx
));
582 * blk_free_devt - free a dev_t
583 * @devt: dev_t to free
585 * Free @devt which was allocated using blk_alloc_devt().
590 void blk_free_devt(dev_t devt
)
592 if (MAJOR(devt
) == BLOCK_EXT_MAJOR
)
593 ida_free(&ext_devt_ida
, blk_mangle_minor(MINOR(devt
)));
596 static char *bdevt_str(dev_t devt
, char *buf
)
598 if (MAJOR(devt
) <= 0xff && MINOR(devt
) <= 0xff) {
599 char tbuf
[BDEVT_SIZE
];
600 snprintf(tbuf
, BDEVT_SIZE
, "%02x%02x", MAJOR(devt
), MINOR(devt
));
601 snprintf(buf
, BDEVT_SIZE
, "%-9s", tbuf
);
603 snprintf(buf
, BDEVT_SIZE
, "%03x:%05x", MAJOR(devt
), MINOR(devt
));
608 static void disk_scan_partitions(struct gendisk
*disk
)
610 struct block_device
*bdev
;
612 if (!get_capacity(disk
) || !disk_part_scan_enabled(disk
))
615 set_bit(GD_NEED_PART_SCAN
, &disk
->state
);
616 bdev
= blkdev_get_by_dev(disk_devt(disk
), FMODE_READ
, NULL
);
618 blkdev_put(bdev
, FMODE_READ
);
621 static void register_disk(struct device
*parent
, struct gendisk
*disk
,
622 const struct attribute_group
**groups
)
624 struct device
*ddev
= disk_to_dev(disk
);
625 struct disk_part_iter piter
;
626 struct block_device
*part
;
629 ddev
->parent
= parent
;
631 dev_set_name(ddev
, "%s", disk
->disk_name
);
633 /* delay uevents, until we scanned partition table */
634 dev_set_uevent_suppress(ddev
, 1);
637 WARN_ON(ddev
->groups
);
638 ddev
->groups
= groups
;
640 if (device_add(ddev
))
642 if (!sysfs_deprecated
) {
643 err
= sysfs_create_link(block_depr
, &ddev
->kobj
,
644 kobject_name(&ddev
->kobj
));
652 * avoid probable deadlock caused by allocating memory with
653 * GFP_KERNEL in runtime_resume callback of its all ancestor
656 pm_runtime_set_memalloc_noio(ddev
, true);
658 disk
->part0
->bd_holder_dir
=
659 kobject_create_and_add("holders", &ddev
->kobj
);
660 disk
->slave_dir
= kobject_create_and_add("slaves", &ddev
->kobj
);
662 if (disk
->flags
& GENHD_FL_HIDDEN
)
665 disk_scan_partitions(disk
);
667 /* announce disk after possible partitions are created */
668 dev_set_uevent_suppress(ddev
, 0);
669 kobject_uevent(&ddev
->kobj
, KOBJ_ADD
);
671 /* announce possible partitions */
672 disk_part_iter_init(&piter
, disk
, 0);
673 while ((part
= disk_part_iter_next(&piter
)))
674 kobject_uevent(bdev_kobj(part
), KOBJ_ADD
);
675 disk_part_iter_exit(&piter
);
677 if (disk
->queue
->backing_dev_info
->dev
) {
678 err
= sysfs_create_link(&ddev
->kobj
,
679 &disk
->queue
->backing_dev_info
->dev
->kobj
,
686 * __device_add_disk - add disk information to kernel list
687 * @parent: parent device for the disk
688 * @disk: per-device partitioning information
689 * @groups: Additional per-device sysfs groups
690 * @register_queue: register the queue if set to true
692 * This function registers the partitioning information in @disk
695 * FIXME: error handling
697 static void __device_add_disk(struct device
*parent
, struct gendisk
*disk
,
698 const struct attribute_group
**groups
,
705 * The disk queue should now be all set with enough information about
706 * the device for the elevator code to pick an adequate default
707 * elevator if one is needed, that is, for devices requesting queue
711 elevator_init_mq(disk
->queue
);
713 /* minors == 0 indicates to use ext devt from part0 and should
714 * be accompanied with EXT_DEVT flag. Make sure all
715 * parameters make sense.
717 WARN_ON(disk
->minors
&& !(disk
->major
|| disk
->first_minor
));
718 WARN_ON(!disk
->minors
&&
719 !(disk
->flags
& (GENHD_FL_EXT_DEVT
| GENHD_FL_HIDDEN
)));
721 disk
->flags
|= GENHD_FL_UP
;
723 retval
= blk_alloc_devt(disk
->part0
, &devt
);
728 disk
->major
= MAJOR(devt
);
729 disk
->first_minor
= MINOR(devt
);
731 disk_alloc_events(disk
);
733 if (disk
->flags
& GENHD_FL_HIDDEN
) {
735 * Don't let hidden disks show up in /proc/partitions,
736 * and don't bother scanning for partitions either.
738 disk
->flags
|= GENHD_FL_SUPPRESS_PARTITION_INFO
;
739 disk
->flags
|= GENHD_FL_NO_PART_SCAN
;
741 struct backing_dev_info
*bdi
= disk
->queue
->backing_dev_info
;
742 struct device
*dev
= disk_to_dev(disk
);
745 /* Register BDI before referencing it from bdev */
747 ret
= bdi_register(bdi
, "%u:%u", MAJOR(devt
), MINOR(devt
));
749 bdi_set_owner(bdi
, dev
);
750 bdev_add(disk
->part0
, devt
);
752 register_disk(parent
, disk
, groups
);
754 blk_register_queue(disk
);
757 * Take an extra ref on queue which will be put on disk_release()
758 * so that it sticks around as long as @disk is there.
760 WARN_ON_ONCE(!blk_get_queue(disk
->queue
));
762 disk_add_events(disk
);
763 blk_integrity_add(disk
);
766 void device_add_disk(struct device
*parent
, struct gendisk
*disk
,
767 const struct attribute_group
**groups
)
770 __device_add_disk(parent
, disk
, groups
, true);
772 EXPORT_SYMBOL(device_add_disk
);
774 void device_add_disk_no_queue_reg(struct device
*parent
, struct gendisk
*disk
)
776 __device_add_disk(parent
, disk
, NULL
, false);
778 EXPORT_SYMBOL(device_add_disk_no_queue_reg
);
780 static void invalidate_partition(struct block_device
*bdev
)
783 __invalidate_device(bdev
, true);
786 * Unhash the bdev inode for this device so that it can't be looked
787 * up any more even if openers still hold references to it.
789 remove_inode_hash(bdev
->bd_inode
);
793 * del_gendisk - remove the gendisk
794 * @disk: the struct gendisk to remove
796 * Removes the gendisk and all its associated resources. This deletes the
797 * partitions associated with the gendisk, and unregisters the associated
800 * This is the counter to the respective __device_add_disk() call.
802 * The final removal of the struct gendisk happens when its refcount reaches 0
803 * with put_disk(), which should be called after del_gendisk(), if
804 * __device_add_disk() was used.
806 * Drivers exist which depend on the release of the gendisk to be synchronous,
807 * it should not be deferred.
811 void del_gendisk(struct gendisk
*disk
)
813 struct disk_part_iter piter
;
814 struct block_device
*part
;
818 if (WARN_ON_ONCE(!disk
->queue
))
821 blk_integrity_del(disk
);
822 disk_del_events(disk
);
825 * Block lookups of the disk until all bdevs are unhashed and the
826 * disk is marked as dead (GENHD_FL_UP cleared).
828 down_write(&bdev_lookup_sem
);
830 /* invalidate stuff */
831 disk_part_iter_init(&piter
, disk
,
832 DISK_PITER_INCL_EMPTY
| DISK_PITER_REVERSE
);
833 while ((part
= disk_part_iter_next(&piter
))) {
834 invalidate_partition(part
);
835 delete_partition(part
);
837 disk_part_iter_exit(&piter
);
839 invalidate_partition(disk
->part0
);
840 set_capacity(disk
, 0);
841 disk
->flags
&= ~GENHD_FL_UP
;
842 up_write(&bdev_lookup_sem
);
844 if (!(disk
->flags
& GENHD_FL_HIDDEN
)) {
845 sysfs_remove_link(&disk_to_dev(disk
)->kobj
, "bdi");
848 * Unregister bdi before releasing device numbers (as they can
849 * get reused and we'd get clashes in sysfs).
851 bdi_unregister(disk
->queue
->backing_dev_info
);
854 blk_unregister_queue(disk
);
856 kobject_put(disk
->part0
->bd_holder_dir
);
857 kobject_put(disk
->slave_dir
);
859 part_stat_set_all(disk
->part0
, 0);
860 disk
->part0
->bd_stamp
= 0;
861 if (!sysfs_deprecated
)
862 sysfs_remove_link(block_depr
, dev_name(disk_to_dev(disk
)));
863 pm_runtime_set_memalloc_noio(disk_to_dev(disk
), false);
864 device_del(disk_to_dev(disk
));
866 EXPORT_SYMBOL(del_gendisk
);
868 /* sysfs access to bad-blocks list. */
869 static ssize_t
disk_badblocks_show(struct device
*dev
,
870 struct device_attribute
*attr
,
873 struct gendisk
*disk
= dev_to_disk(dev
);
876 return sprintf(page
, "\n");
878 return badblocks_show(disk
->bb
, page
, 0);
881 static ssize_t
disk_badblocks_store(struct device
*dev
,
882 struct device_attribute
*attr
,
883 const char *page
, size_t len
)
885 struct gendisk
*disk
= dev_to_disk(dev
);
890 return badblocks_store(disk
->bb
, page
, len
, 0);
893 void blk_request_module(dev_t devt
)
895 unsigned int major
= MAJOR(devt
);
896 struct blk_major_name
**n
;
898 mutex_lock(&major_names_lock
);
899 for (n
= &major_names
[major_to_index(major
)]; *n
; n
= &(*n
)->next
) {
900 if ((*n
)->major
== major
&& (*n
)->probe
) {
902 mutex_unlock(&major_names_lock
);
906 mutex_unlock(&major_names_lock
);
908 if (request_module("block-major-%d-%d", MAJOR(devt
), MINOR(devt
)) > 0)
909 /* Make old-style 2.4 aliases work */
910 request_module("block-major-%d", MAJOR(devt
));
914 * bdget_disk - do bdget() by gendisk and partition number
915 * @disk: gendisk of interest
916 * @partno: partition number
918 * Find partition @partno from @disk, do bdget() on it.
924 * Resulting block_device on success, NULL on failure.
926 struct block_device
*bdget_disk(struct gendisk
*disk
, int partno
)
928 struct block_device
*bdev
= NULL
;
931 bdev
= __disk_get_part(disk
, partno
);
932 if (bdev
&& !bdgrab(bdev
))
940 * print a full list of all partitions - intended for places where the root
941 * filesystem can't be mounted and thus to give the victim some idea of what
944 void __init
printk_all_partitions(void)
946 struct class_dev_iter iter
;
949 class_dev_iter_init(&iter
, &block_class
, NULL
, &disk_type
);
950 while ((dev
= class_dev_iter_next(&iter
))) {
951 struct gendisk
*disk
= dev_to_disk(dev
);
952 struct disk_part_iter piter
;
953 struct block_device
*part
;
954 char name_buf
[BDEVNAME_SIZE
];
955 char devt_buf
[BDEVT_SIZE
];
958 * Don't show empty devices or things that have been
961 if (get_capacity(disk
) == 0 ||
962 (disk
->flags
& GENHD_FL_SUPPRESS_PARTITION_INFO
))
966 * Note, unlike /proc/partitions, I am showing the
967 * numbers in hex - the same format as the root=
970 disk_part_iter_init(&piter
, disk
, DISK_PITER_INCL_PART0
);
971 while ((part
= disk_part_iter_next(&piter
))) {
972 bool is_part0
= part
== disk
->part0
;
974 printk("%s%s %10llu %s %s", is_part0
? "" : " ",
975 bdevt_str(part
->bd_dev
, devt_buf
),
976 bdev_nr_sectors(part
) >> 1,
977 disk_name(disk
, part
->bd_partno
, name_buf
),
979 part
->bd_meta_info
->uuid
: "");
981 if (dev
->parent
&& dev
->parent
->driver
)
982 printk(" driver: %s\n",
983 dev
->parent
->driver
->name
);
985 printk(" (driver?)\n");
989 disk_part_iter_exit(&piter
);
991 class_dev_iter_exit(&iter
);
994 #ifdef CONFIG_PROC_FS
996 static void *disk_seqf_start(struct seq_file
*seqf
, loff_t
*pos
)
999 struct class_dev_iter
*iter
;
1002 iter
= kmalloc(sizeof(*iter
), GFP_KERNEL
);
1004 return ERR_PTR(-ENOMEM
);
1006 seqf
->private = iter
;
1007 class_dev_iter_init(iter
, &block_class
, NULL
, &disk_type
);
1009 dev
= class_dev_iter_next(iter
);
1014 return dev_to_disk(dev
);
1017 static void *disk_seqf_next(struct seq_file
*seqf
, void *v
, loff_t
*pos
)
1022 dev
= class_dev_iter_next(seqf
->private);
1024 return dev_to_disk(dev
);
1029 static void disk_seqf_stop(struct seq_file
*seqf
, void *v
)
1031 struct class_dev_iter
*iter
= seqf
->private;
1033 /* stop is called even after start failed :-( */
1035 class_dev_iter_exit(iter
);
1037 seqf
->private = NULL
;
1041 static void *show_partition_start(struct seq_file
*seqf
, loff_t
*pos
)
1045 p
= disk_seqf_start(seqf
, pos
);
1046 if (!IS_ERR_OR_NULL(p
) && !*pos
)
1047 seq_puts(seqf
, "major minor #blocks name\n\n");
1051 static int show_partition(struct seq_file
*seqf
, void *v
)
1053 struct gendisk
*sgp
= v
;
1054 struct disk_part_iter piter
;
1055 struct block_device
*part
;
1056 char buf
[BDEVNAME_SIZE
];
1058 /* Don't show non-partitionable removeable devices or empty devices */
1059 if (!get_capacity(sgp
) || (!disk_max_parts(sgp
) &&
1060 (sgp
->flags
& GENHD_FL_REMOVABLE
)))
1062 if (sgp
->flags
& GENHD_FL_SUPPRESS_PARTITION_INFO
)
1065 /* show the full disk and all non-0 size partitions of it */
1066 disk_part_iter_init(&piter
, sgp
, DISK_PITER_INCL_PART0
);
1067 while ((part
= disk_part_iter_next(&piter
)))
1068 seq_printf(seqf
, "%4d %7d %10llu %s\n",
1069 MAJOR(part
->bd_dev
), MINOR(part
->bd_dev
),
1070 bdev_nr_sectors(part
) >> 1,
1071 disk_name(sgp
, part
->bd_partno
, buf
));
1072 disk_part_iter_exit(&piter
);
1077 static const struct seq_operations partitions_op
= {
1078 .start
= show_partition_start
,
1079 .next
= disk_seqf_next
,
1080 .stop
= disk_seqf_stop
,
1081 .show
= show_partition
1085 static int __init
genhd_device_init(void)
1089 block_class
.dev_kobj
= sysfs_dev_block_kobj
;
1090 error
= class_register(&block_class
);
1091 if (unlikely(error
))
1095 register_blkdev(BLOCK_EXT_MAJOR
, "blkext");
1097 /* create top-level block dir */
1098 if (!sysfs_deprecated
)
1099 block_depr
= kobject_create_and_add("block", NULL
);
1103 subsys_initcall(genhd_device_init
);
1105 static ssize_t
disk_range_show(struct device
*dev
,
1106 struct device_attribute
*attr
, char *buf
)
1108 struct gendisk
*disk
= dev_to_disk(dev
);
1110 return sprintf(buf
, "%d\n", disk
->minors
);
1113 static ssize_t
disk_ext_range_show(struct device
*dev
,
1114 struct device_attribute
*attr
, char *buf
)
1116 struct gendisk
*disk
= dev_to_disk(dev
);
1118 return sprintf(buf
, "%d\n", disk_max_parts(disk
));
1121 static ssize_t
disk_removable_show(struct device
*dev
,
1122 struct device_attribute
*attr
, char *buf
)
1124 struct gendisk
*disk
= dev_to_disk(dev
);
1126 return sprintf(buf
, "%d\n",
1127 (disk
->flags
& GENHD_FL_REMOVABLE
? 1 : 0));
1130 static ssize_t
disk_hidden_show(struct device
*dev
,
1131 struct device_attribute
*attr
, char *buf
)
1133 struct gendisk
*disk
= dev_to_disk(dev
);
1135 return sprintf(buf
, "%d\n",
1136 (disk
->flags
& GENHD_FL_HIDDEN
? 1 : 0));
1139 static ssize_t
disk_ro_show(struct device
*dev
,
1140 struct device_attribute
*attr
, char *buf
)
1142 struct gendisk
*disk
= dev_to_disk(dev
);
1144 return sprintf(buf
, "%d\n", get_disk_ro(disk
) ? 1 : 0);
1147 ssize_t
part_size_show(struct device
*dev
,
1148 struct device_attribute
*attr
, char *buf
)
1150 return sprintf(buf
, "%llu\n", bdev_nr_sectors(dev_to_bdev(dev
)));
1153 ssize_t
part_stat_show(struct device
*dev
,
1154 struct device_attribute
*attr
, char *buf
)
1156 struct block_device
*bdev
= dev_to_bdev(dev
);
1157 struct request_queue
*q
= bdev
->bd_disk
->queue
;
1158 struct disk_stats stat
;
1159 unsigned int inflight
;
1161 part_stat_read_all(bdev
, &stat
);
1163 inflight
= blk_mq_in_flight(q
, bdev
);
1165 inflight
= part_in_flight(bdev
);
1168 "%8lu %8lu %8llu %8u "
1169 "%8lu %8lu %8llu %8u "
1171 "%8lu %8lu %8llu %8u "
1174 stat
.ios
[STAT_READ
],
1175 stat
.merges
[STAT_READ
],
1176 (unsigned long long)stat
.sectors
[STAT_READ
],
1177 (unsigned int)div_u64(stat
.nsecs
[STAT_READ
], NSEC_PER_MSEC
),
1178 stat
.ios
[STAT_WRITE
],
1179 stat
.merges
[STAT_WRITE
],
1180 (unsigned long long)stat
.sectors
[STAT_WRITE
],
1181 (unsigned int)div_u64(stat
.nsecs
[STAT_WRITE
], NSEC_PER_MSEC
),
1183 jiffies_to_msecs(stat
.io_ticks
),
1184 (unsigned int)div_u64(stat
.nsecs
[STAT_READ
] +
1185 stat
.nsecs
[STAT_WRITE
] +
1186 stat
.nsecs
[STAT_DISCARD
] +
1187 stat
.nsecs
[STAT_FLUSH
],
1189 stat
.ios
[STAT_DISCARD
],
1190 stat
.merges
[STAT_DISCARD
],
1191 (unsigned long long)stat
.sectors
[STAT_DISCARD
],
1192 (unsigned int)div_u64(stat
.nsecs
[STAT_DISCARD
], NSEC_PER_MSEC
),
1193 stat
.ios
[STAT_FLUSH
],
1194 (unsigned int)div_u64(stat
.nsecs
[STAT_FLUSH
], NSEC_PER_MSEC
));
1197 ssize_t
part_inflight_show(struct device
*dev
, struct device_attribute
*attr
,
1200 struct block_device
*bdev
= dev_to_bdev(dev
);
1201 struct request_queue
*q
= bdev
->bd_disk
->queue
;
1202 unsigned int inflight
[2];
1205 blk_mq_in_flight_rw(q
, bdev
, inflight
);
1207 part_in_flight_rw(bdev
, inflight
);
1209 return sprintf(buf
, "%8u %8u\n", inflight
[0], inflight
[1]);
1212 static ssize_t
disk_capability_show(struct device
*dev
,
1213 struct device_attribute
*attr
, char *buf
)
1215 struct gendisk
*disk
= dev_to_disk(dev
);
1217 return sprintf(buf
, "%x\n", disk
->flags
);
1220 static ssize_t
disk_alignment_offset_show(struct device
*dev
,
1221 struct device_attribute
*attr
,
1224 struct gendisk
*disk
= dev_to_disk(dev
);
1226 return sprintf(buf
, "%d\n", queue_alignment_offset(disk
->queue
));
1229 static ssize_t
disk_discard_alignment_show(struct device
*dev
,
1230 struct device_attribute
*attr
,
1233 struct gendisk
*disk
= dev_to_disk(dev
);
1235 return sprintf(buf
, "%d\n", queue_discard_alignment(disk
->queue
));
1238 static DEVICE_ATTR(range
, 0444, disk_range_show
, NULL
);
1239 static DEVICE_ATTR(ext_range
, 0444, disk_ext_range_show
, NULL
);
1240 static DEVICE_ATTR(removable
, 0444, disk_removable_show
, NULL
);
1241 static DEVICE_ATTR(hidden
, 0444, disk_hidden_show
, NULL
);
1242 static DEVICE_ATTR(ro
, 0444, disk_ro_show
, NULL
);
1243 static DEVICE_ATTR(size
, 0444, part_size_show
, NULL
);
1244 static DEVICE_ATTR(alignment_offset
, 0444, disk_alignment_offset_show
, NULL
);
1245 static DEVICE_ATTR(discard_alignment
, 0444, disk_discard_alignment_show
, NULL
);
1246 static DEVICE_ATTR(capability
, 0444, disk_capability_show
, NULL
);
1247 static DEVICE_ATTR(stat
, 0444, part_stat_show
, NULL
);
1248 static DEVICE_ATTR(inflight
, 0444, part_inflight_show
, NULL
);
1249 static DEVICE_ATTR(badblocks
, 0644, disk_badblocks_show
, disk_badblocks_store
);
1251 #ifdef CONFIG_FAIL_MAKE_REQUEST
1252 ssize_t
part_fail_show(struct device
*dev
,
1253 struct device_attribute
*attr
, char *buf
)
1255 return sprintf(buf
, "%d\n", dev_to_bdev(dev
)->bd_make_it_fail
);
1258 ssize_t
part_fail_store(struct device
*dev
,
1259 struct device_attribute
*attr
,
1260 const char *buf
, size_t count
)
1264 if (count
> 0 && sscanf(buf
, "%d", &i
) > 0)
1265 dev_to_bdev(dev
)->bd_make_it_fail
= i
;
1270 static struct device_attribute dev_attr_fail
=
1271 __ATTR(make
-it
-fail
, 0644, part_fail_show
, part_fail_store
);
1272 #endif /* CONFIG_FAIL_MAKE_REQUEST */
1274 #ifdef CONFIG_FAIL_IO_TIMEOUT
1275 static struct device_attribute dev_attr_fail_timeout
=
1276 __ATTR(io
-timeout
-fail
, 0644, part_timeout_show
, part_timeout_store
);
1279 static struct attribute
*disk_attrs
[] = {
1280 &dev_attr_range
.attr
,
1281 &dev_attr_ext_range
.attr
,
1282 &dev_attr_removable
.attr
,
1283 &dev_attr_hidden
.attr
,
1285 &dev_attr_size
.attr
,
1286 &dev_attr_alignment_offset
.attr
,
1287 &dev_attr_discard_alignment
.attr
,
1288 &dev_attr_capability
.attr
,
1289 &dev_attr_stat
.attr
,
1290 &dev_attr_inflight
.attr
,
1291 &dev_attr_badblocks
.attr
,
1292 #ifdef CONFIG_FAIL_MAKE_REQUEST
1293 &dev_attr_fail
.attr
,
1295 #ifdef CONFIG_FAIL_IO_TIMEOUT
1296 &dev_attr_fail_timeout
.attr
,
1301 static umode_t
disk_visible(struct kobject
*kobj
, struct attribute
*a
, int n
)
1303 struct device
*dev
= container_of(kobj
, typeof(*dev
), kobj
);
1304 struct gendisk
*disk
= dev_to_disk(dev
);
1306 if (a
== &dev_attr_badblocks
.attr
&& !disk
->bb
)
1311 static struct attribute_group disk_attr_group
= {
1312 .attrs
= disk_attrs
,
1313 .is_visible
= disk_visible
,
1316 static const struct attribute_group
*disk_attr_groups
[] = {
1322 * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way
1323 * @disk: disk to replace part_tbl for
1324 * @new_ptbl: new part_tbl to install
1326 * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The
1327 * original ptbl is freed using RCU callback.
1330 * Matching bd_mutex locked or the caller is the only user of @disk.
1332 static void disk_replace_part_tbl(struct gendisk
*disk
,
1333 struct disk_part_tbl
*new_ptbl
)
1335 struct disk_part_tbl
*old_ptbl
=
1336 rcu_dereference_protected(disk
->part_tbl
, 1);
1338 rcu_assign_pointer(disk
->part_tbl
, new_ptbl
);
1341 rcu_assign_pointer(old_ptbl
->last_lookup
, NULL
);
1342 kfree_rcu(old_ptbl
, rcu_head
);
1347 * disk_expand_part_tbl - expand disk->part_tbl
1348 * @disk: disk to expand part_tbl for
1349 * @partno: expand such that this partno can fit in
1351 * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl
1352 * uses RCU to allow unlocked dereferencing for stats and other stuff.
1355 * Matching bd_mutex locked or the caller is the only user of @disk.
1359 * 0 on success, -errno on failure.
1361 int disk_expand_part_tbl(struct gendisk
*disk
, int partno
)
1363 struct disk_part_tbl
*old_ptbl
=
1364 rcu_dereference_protected(disk
->part_tbl
, 1);
1365 struct disk_part_tbl
*new_ptbl
;
1366 int len
= old_ptbl
? old_ptbl
->len
: 0;
1370 * check for int overflow, since we can get here from blkpg_ioctl()
1371 * with a user passed 'partno'.
1373 target
= partno
+ 1;
1377 /* disk_max_parts() is zero during initialization, ignore if so */
1378 if (disk_max_parts(disk
) && target
> disk_max_parts(disk
))
1384 new_ptbl
= kzalloc_node(struct_size(new_ptbl
, part
, target
), GFP_KERNEL
,
1389 new_ptbl
->len
= target
;
1391 for (i
= 0; i
< len
; i
++)
1392 rcu_assign_pointer(new_ptbl
->part
[i
], old_ptbl
->part
[i
]);
1394 disk_replace_part_tbl(disk
, new_ptbl
);
1399 * disk_release - releases all allocated resources of the gendisk
1400 * @dev: the device representing this disk
1402 * This function releases all allocated resources of the gendisk.
1404 * Drivers which used __device_add_disk() have a gendisk with a request_queue
1405 * assigned. Since the request_queue sits on top of the gendisk for these
1406 * drivers we also call blk_put_queue() for them, and we expect the
1407 * request_queue refcount to reach 0 at this point, and so the request_queue
1408 * will also be freed prior to the disk.
1410 * Context: can sleep
1412 static void disk_release(struct device
*dev
)
1414 struct gendisk
*disk
= dev_to_disk(dev
);
1418 blk_free_devt(dev
->devt
);
1419 disk_release_events(disk
);
1420 kfree(disk
->random
);
1421 disk_replace_part_tbl(disk
, NULL
);
1424 blk_put_queue(disk
->queue
);
1427 struct class block_class
= {
1431 static char *block_devnode(struct device
*dev
, umode_t
*mode
,
1432 kuid_t
*uid
, kgid_t
*gid
)
1434 struct gendisk
*disk
= dev_to_disk(dev
);
1436 if (disk
->fops
->devnode
)
1437 return disk
->fops
->devnode(disk
, mode
);
1441 const struct device_type disk_type
= {
1443 .groups
= disk_attr_groups
,
1444 .release
= disk_release
,
1445 .devnode
= block_devnode
,
1448 #ifdef CONFIG_PROC_FS
1450 * aggregate disk stat collector. Uses the same stats that the sysfs
1451 * entries do, above, but makes them available through one seq_file.
1453 * The output looks suspiciously like /proc/partitions with a bunch of
1456 static int diskstats_show(struct seq_file
*seqf
, void *v
)
1458 struct gendisk
*gp
= v
;
1459 struct disk_part_iter piter
;
1460 struct block_device
*hd
;
1461 char buf
[BDEVNAME_SIZE
];
1462 unsigned int inflight
;
1463 struct disk_stats stat
;
1466 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1467 seq_puts(seqf, "major minor name"
1468 " rio rmerge rsect ruse wio wmerge "
1469 "wsect wuse running use aveq"
1473 disk_part_iter_init(&piter
, gp
, DISK_PITER_INCL_EMPTY_PART0
);
1474 while ((hd
= disk_part_iter_next(&piter
))) {
1475 part_stat_read_all(hd
, &stat
);
1476 if (queue_is_mq(gp
->queue
))
1477 inflight
= blk_mq_in_flight(gp
->queue
, hd
);
1479 inflight
= part_in_flight(hd
);
1481 seq_printf(seqf
, "%4d %7d %s "
1488 MAJOR(hd
->bd_dev
), MINOR(hd
->bd_dev
),
1489 disk_name(gp
, hd
->bd_partno
, buf
),
1490 stat
.ios
[STAT_READ
],
1491 stat
.merges
[STAT_READ
],
1492 stat
.sectors
[STAT_READ
],
1493 (unsigned int)div_u64(stat
.nsecs
[STAT_READ
],
1495 stat
.ios
[STAT_WRITE
],
1496 stat
.merges
[STAT_WRITE
],
1497 stat
.sectors
[STAT_WRITE
],
1498 (unsigned int)div_u64(stat
.nsecs
[STAT_WRITE
],
1501 jiffies_to_msecs(stat
.io_ticks
),
1502 (unsigned int)div_u64(stat
.nsecs
[STAT_READ
] +
1503 stat
.nsecs
[STAT_WRITE
] +
1504 stat
.nsecs
[STAT_DISCARD
] +
1505 stat
.nsecs
[STAT_FLUSH
],
1507 stat
.ios
[STAT_DISCARD
],
1508 stat
.merges
[STAT_DISCARD
],
1509 stat
.sectors
[STAT_DISCARD
],
1510 (unsigned int)div_u64(stat
.nsecs
[STAT_DISCARD
],
1512 stat
.ios
[STAT_FLUSH
],
1513 (unsigned int)div_u64(stat
.nsecs
[STAT_FLUSH
],
1517 disk_part_iter_exit(&piter
);
1522 static const struct seq_operations diskstats_op
= {
1523 .start
= disk_seqf_start
,
1524 .next
= disk_seqf_next
,
1525 .stop
= disk_seqf_stop
,
1526 .show
= diskstats_show
1529 static int __init
proc_genhd_init(void)
1531 proc_create_seq("diskstats", 0, NULL
, &diskstats_op
);
1532 proc_create_seq("partitions", 0, NULL
, &partitions_op
);
1535 module_init(proc_genhd_init
);
1536 #endif /* CONFIG_PROC_FS */
1538 dev_t
blk_lookup_devt(const char *name
, int partno
)
1540 dev_t devt
= MKDEV(0, 0);
1541 struct class_dev_iter iter
;
1544 class_dev_iter_init(&iter
, &block_class
, NULL
, &disk_type
);
1545 while ((dev
= class_dev_iter_next(&iter
))) {
1546 struct gendisk
*disk
= dev_to_disk(dev
);
1547 struct block_device
*part
;
1549 if (strcmp(dev_name(dev
), name
))
1552 if (partno
< disk
->minors
) {
1553 /* We need to return the right devno, even
1554 * if the partition doesn't exist yet.
1556 devt
= MKDEV(MAJOR(dev
->devt
),
1557 MINOR(dev
->devt
) + partno
);
1560 part
= bdget_disk(disk
, partno
);
1562 devt
= part
->bd_dev
;
1567 class_dev_iter_exit(&iter
);
1571 struct gendisk
*__alloc_disk_node(int minors
, int node_id
)
1573 struct gendisk
*disk
;
1574 struct disk_part_tbl
*ptbl
;
1576 if (minors
> DISK_MAX_PARTS
) {
1578 "block: can't allocate more than %d partitions\n",
1580 minors
= DISK_MAX_PARTS
;
1583 disk
= kzalloc_node(sizeof(struct gendisk
), GFP_KERNEL
, node_id
);
1587 disk
->part0
= bdev_alloc(disk
, 0);
1591 disk
->node_id
= node_id
;
1592 if (disk_expand_part_tbl(disk
, 0))
1595 ptbl
= rcu_dereference_protected(disk
->part_tbl
, 1);
1596 rcu_assign_pointer(ptbl
->part
[0], disk
->part0
);
1598 disk
->minors
= minors
;
1599 rand_initialize_disk(disk
);
1600 disk_to_dev(disk
)->class = &block_class
;
1601 disk_to_dev(disk
)->type
= &disk_type
;
1602 device_initialize(disk_to_dev(disk
));
1611 EXPORT_SYMBOL(__alloc_disk_node
);
1614 * put_disk - decrements the gendisk refcount
1615 * @disk: the struct gendisk to decrement the refcount for
1617 * This decrements the refcount for the struct gendisk. When this reaches 0
1618 * we'll have disk_release() called.
1620 * Context: Any context, but the last reference must not be dropped from
1623 void put_disk(struct gendisk
*disk
)
1626 put_device(disk_to_dev(disk
));
1628 EXPORT_SYMBOL(put_disk
);
1630 static void set_disk_ro_uevent(struct gendisk
*gd
, int ro
)
1632 char event
[] = "DISK_RO=1";
1633 char *envp
[] = { event
, NULL
};
1637 kobject_uevent_env(&disk_to_dev(gd
)->kobj
, KOBJ_CHANGE
, envp
);
1640 void set_disk_ro(struct gendisk
*disk
, int flag
)
1642 struct disk_part_iter piter
;
1643 struct block_device
*part
;
1645 if (disk
->part0
->bd_read_only
!= flag
) {
1646 set_disk_ro_uevent(disk
, flag
);
1647 disk
->part0
->bd_read_only
= flag
;
1650 disk_part_iter_init(&piter
, disk
, DISK_PITER_INCL_EMPTY
);
1651 while ((part
= disk_part_iter_next(&piter
)))
1652 part
->bd_read_only
= flag
;
1653 disk_part_iter_exit(&piter
);
1656 EXPORT_SYMBOL(set_disk_ro
);
1658 int bdev_read_only(struct block_device
*bdev
)
1662 return bdev
->bd_read_only
;
1665 EXPORT_SYMBOL(bdev_read_only
);
1668 * Disk events - monitor disk events like media change and eject request.
1670 struct disk_events
{
1671 struct list_head node
; /* all disk_event's */
1672 struct gendisk
*disk
; /* the associated disk */
1675 struct mutex block_mutex
; /* protects blocking */
1676 int block
; /* event blocking depth */
1677 unsigned int pending
; /* events already sent out */
1678 unsigned int clearing
; /* events being cleared */
1680 long poll_msecs
; /* interval, -1 for default */
1681 struct delayed_work dwork
;
1684 static const char *disk_events_strs
[] = {
1685 [ilog2(DISK_EVENT_MEDIA_CHANGE
)] = "media_change",
1686 [ilog2(DISK_EVENT_EJECT_REQUEST
)] = "eject_request",
1689 static char *disk_uevents
[] = {
1690 [ilog2(DISK_EVENT_MEDIA_CHANGE
)] = "DISK_MEDIA_CHANGE=1",
1691 [ilog2(DISK_EVENT_EJECT_REQUEST
)] = "DISK_EJECT_REQUEST=1",
1694 /* list of all disk_events */
1695 static DEFINE_MUTEX(disk_events_mutex
);
1696 static LIST_HEAD(disk_events
);
1698 /* disable in-kernel polling by default */
1699 static unsigned long disk_events_dfl_poll_msecs
;
1701 static unsigned long disk_events_poll_jiffies(struct gendisk
*disk
)
1703 struct disk_events
*ev
= disk
->ev
;
1704 long intv_msecs
= 0;
1707 * If device-specific poll interval is set, always use it. If
1708 * the default is being used, poll if the POLL flag is set.
1710 if (ev
->poll_msecs
>= 0)
1711 intv_msecs
= ev
->poll_msecs
;
1712 else if (disk
->event_flags
& DISK_EVENT_FLAG_POLL
)
1713 intv_msecs
= disk_events_dfl_poll_msecs
;
1715 return msecs_to_jiffies(intv_msecs
);
1719 * disk_block_events - block and flush disk event checking
1720 * @disk: disk to block events for
1722 * On return from this function, it is guaranteed that event checking
1723 * isn't in progress and won't happen until unblocked by
1724 * disk_unblock_events(). Events blocking is counted and the actual
1725 * unblocking happens after the matching number of unblocks are done.
1727 * Note that this intentionally does not block event checking from
1728 * disk_clear_events().
1733 void disk_block_events(struct gendisk
*disk
)
1735 struct disk_events
*ev
= disk
->ev
;
1736 unsigned long flags
;
1743 * Outer mutex ensures that the first blocker completes canceling
1744 * the event work before further blockers are allowed to finish.
1746 mutex_lock(&ev
->block_mutex
);
1748 spin_lock_irqsave(&ev
->lock
, flags
);
1749 cancel
= !ev
->block
++;
1750 spin_unlock_irqrestore(&ev
->lock
, flags
);
1753 cancel_delayed_work_sync(&disk
->ev
->dwork
);
1755 mutex_unlock(&ev
->block_mutex
);
1758 static void __disk_unblock_events(struct gendisk
*disk
, bool check_now
)
1760 struct disk_events
*ev
= disk
->ev
;
1762 unsigned long flags
;
1764 spin_lock_irqsave(&ev
->lock
, flags
);
1766 if (WARN_ON_ONCE(ev
->block
<= 0))
1772 intv
= disk_events_poll_jiffies(disk
);
1774 queue_delayed_work(system_freezable_power_efficient_wq
,
1777 queue_delayed_work(system_freezable_power_efficient_wq
,
1780 spin_unlock_irqrestore(&ev
->lock
, flags
);
1784 * disk_unblock_events - unblock disk event checking
1785 * @disk: disk to unblock events for
1787 * Undo disk_block_events(). When the block count reaches zero, it
1788 * starts events polling if configured.
1791 * Don't care. Safe to call from irq context.
1793 void disk_unblock_events(struct gendisk
*disk
)
1796 __disk_unblock_events(disk
, false);
1800 * disk_flush_events - schedule immediate event checking and flushing
1801 * @disk: disk to check and flush events for
1802 * @mask: events to flush
1804 * Schedule immediate event checking on @disk if not blocked. Events in
1805 * @mask are scheduled to be cleared from the driver. Note that this
1806 * doesn't clear the events from @disk->ev.
1809 * If @mask is non-zero must be called with bdev->bd_mutex held.
1811 void disk_flush_events(struct gendisk
*disk
, unsigned int mask
)
1813 struct disk_events
*ev
= disk
->ev
;
1818 spin_lock_irq(&ev
->lock
);
1819 ev
->clearing
|= mask
;
1821 mod_delayed_work(system_freezable_power_efficient_wq
,
1823 spin_unlock_irq(&ev
->lock
);
1827 * disk_clear_events - synchronously check, clear and return pending events
1828 * @disk: disk to fetch and clear events from
1829 * @mask: mask of events to be fetched and cleared
1831 * Disk events are synchronously checked and pending events in @mask
1832 * are cleared and returned. This ignores the block count.
1837 static unsigned int disk_clear_events(struct gendisk
*disk
, unsigned int mask
)
1839 struct disk_events
*ev
= disk
->ev
;
1840 unsigned int pending
;
1841 unsigned int clearing
= mask
;
1846 disk_block_events(disk
);
1849 * store the union of mask and ev->clearing on the stack so that the
1850 * race with disk_flush_events does not cause ambiguity (ev->clearing
1851 * can still be modified even if events are blocked).
1853 spin_lock_irq(&ev
->lock
);
1854 clearing
|= ev
->clearing
;
1856 spin_unlock_irq(&ev
->lock
);
1858 disk_check_events(ev
, &clearing
);
1860 * if ev->clearing is not 0, the disk_flush_events got called in the
1861 * middle of this function, so we want to run the workfn without delay.
1863 __disk_unblock_events(disk
, ev
->clearing
? true : false);
1865 /* then, fetch and clear pending events */
1866 spin_lock_irq(&ev
->lock
);
1867 pending
= ev
->pending
& mask
;
1868 ev
->pending
&= ~mask
;
1869 spin_unlock_irq(&ev
->lock
);
1870 WARN_ON_ONCE(clearing
& mask
);
1876 * bdev_check_media_change - check if a removable media has been changed
1877 * @bdev: block device to check
1879 * Check whether a removable media has been changed, and attempt to free all
1880 * dentries and inodes and invalidates all block device page cache entries in
1883 * Returns %true if the block device changed, or %false if not.
1885 bool bdev_check_media_change(struct block_device
*bdev
)
1887 unsigned int events
;
1889 events
= disk_clear_events(bdev
->bd_disk
, DISK_EVENT_MEDIA_CHANGE
|
1890 DISK_EVENT_EJECT_REQUEST
);
1891 if (!(events
& DISK_EVENT_MEDIA_CHANGE
))
1894 if (__invalidate_device(bdev
, true))
1895 pr_warn("VFS: busy inodes on changed media %s\n",
1896 bdev
->bd_disk
->disk_name
);
1897 set_bit(GD_NEED_PART_SCAN
, &bdev
->bd_disk
->state
);
1900 EXPORT_SYMBOL(bdev_check_media_change
);
1903 * Separate this part out so that a different pointer for clearing_ptr can be
1904 * passed in for disk_clear_events.
1906 static void disk_events_workfn(struct work_struct
*work
)
1908 struct delayed_work
*dwork
= to_delayed_work(work
);
1909 struct disk_events
*ev
= container_of(dwork
, struct disk_events
, dwork
);
1911 disk_check_events(ev
, &ev
->clearing
);
1914 static void disk_check_events(struct disk_events
*ev
,
1915 unsigned int *clearing_ptr
)
1917 struct gendisk
*disk
= ev
->disk
;
1918 char *envp
[ARRAY_SIZE(disk_uevents
) + 1] = { };
1919 unsigned int clearing
= *clearing_ptr
;
1920 unsigned int events
;
1922 int nr_events
= 0, i
;
1925 events
= disk
->fops
->check_events(disk
, clearing
);
1927 /* accumulate pending events and schedule next poll if necessary */
1928 spin_lock_irq(&ev
->lock
);
1930 events
&= ~ev
->pending
;
1931 ev
->pending
|= events
;
1932 *clearing_ptr
&= ~clearing
;
1934 intv
= disk_events_poll_jiffies(disk
);
1935 if (!ev
->block
&& intv
)
1936 queue_delayed_work(system_freezable_power_efficient_wq
,
1939 spin_unlock_irq(&ev
->lock
);
1942 * Tell userland about new events. Only the events listed in
1943 * @disk->events are reported, and only if DISK_EVENT_FLAG_UEVENT
1944 * is set. Otherwise, events are processed internally but never
1945 * get reported to userland.
1947 for (i
= 0; i
< ARRAY_SIZE(disk_uevents
); i
++)
1948 if ((events
& disk
->events
& (1 << i
)) &&
1949 (disk
->event_flags
& DISK_EVENT_FLAG_UEVENT
))
1950 envp
[nr_events
++] = disk_uevents
[i
];
1953 kobject_uevent_env(&disk_to_dev(disk
)->kobj
, KOBJ_CHANGE
, envp
);
1957 * A disk events enabled device has the following sysfs nodes under
1958 * its /sys/block/X/ directory.
1960 * events : list of all supported events
1961 * events_async : list of events which can be detected w/o polling
1962 * (always empty, only for backwards compatibility)
1963 * events_poll_msecs : polling interval, 0: disable, -1: system default
1965 static ssize_t
__disk_events_show(unsigned int events
, char *buf
)
1967 const char *delim
= "";
1971 for (i
= 0; i
< ARRAY_SIZE(disk_events_strs
); i
++)
1972 if (events
& (1 << i
)) {
1973 pos
+= sprintf(buf
+ pos
, "%s%s",
1974 delim
, disk_events_strs
[i
]);
1978 pos
+= sprintf(buf
+ pos
, "\n");
1982 static ssize_t
disk_events_show(struct device
*dev
,
1983 struct device_attribute
*attr
, char *buf
)
1985 struct gendisk
*disk
= dev_to_disk(dev
);
1987 if (!(disk
->event_flags
& DISK_EVENT_FLAG_UEVENT
))
1990 return __disk_events_show(disk
->events
, buf
);
1993 static ssize_t
disk_events_async_show(struct device
*dev
,
1994 struct device_attribute
*attr
, char *buf
)
1999 static ssize_t
disk_events_poll_msecs_show(struct device
*dev
,
2000 struct device_attribute
*attr
,
2003 struct gendisk
*disk
= dev_to_disk(dev
);
2006 return sprintf(buf
, "-1\n");
2008 return sprintf(buf
, "%ld\n", disk
->ev
->poll_msecs
);
2011 static ssize_t
disk_events_poll_msecs_store(struct device
*dev
,
2012 struct device_attribute
*attr
,
2013 const char *buf
, size_t count
)
2015 struct gendisk
*disk
= dev_to_disk(dev
);
2018 if (!count
|| !sscanf(buf
, "%ld", &intv
))
2021 if (intv
< 0 && intv
!= -1)
2027 disk_block_events(disk
);
2028 disk
->ev
->poll_msecs
= intv
;
2029 __disk_unblock_events(disk
, true);
2034 static const DEVICE_ATTR(events
, 0444, disk_events_show
, NULL
);
2035 static const DEVICE_ATTR(events_async
, 0444, disk_events_async_show
, NULL
);
2036 static const DEVICE_ATTR(events_poll_msecs
, 0644,
2037 disk_events_poll_msecs_show
,
2038 disk_events_poll_msecs_store
);
2040 static const struct attribute
*disk_events_attrs
[] = {
2041 &dev_attr_events
.attr
,
2042 &dev_attr_events_async
.attr
,
2043 &dev_attr_events_poll_msecs
.attr
,
2048 * The default polling interval can be specified by the kernel
2049 * parameter block.events_dfl_poll_msecs which defaults to 0
2050 * (disable). This can also be modified runtime by writing to
2051 * /sys/module/block/parameters/events_dfl_poll_msecs.
2053 static int disk_events_set_dfl_poll_msecs(const char *val
,
2054 const struct kernel_param
*kp
)
2056 struct disk_events
*ev
;
2059 ret
= param_set_ulong(val
, kp
);
2063 mutex_lock(&disk_events_mutex
);
2065 list_for_each_entry(ev
, &disk_events
, node
)
2066 disk_flush_events(ev
->disk
, 0);
2068 mutex_unlock(&disk_events_mutex
);
2073 static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops
= {
2074 .set
= disk_events_set_dfl_poll_msecs
,
2075 .get
= param_get_ulong
,
2078 #undef MODULE_PARAM_PREFIX
2079 #define MODULE_PARAM_PREFIX "block."
2081 module_param_cb(events_dfl_poll_msecs
, &disk_events_dfl_poll_msecs_param_ops
,
2082 &disk_events_dfl_poll_msecs
, 0644);
2085 * disk_{alloc|add|del|release}_events - initialize and destroy disk_events.
2087 static void disk_alloc_events(struct gendisk
*disk
)
2089 struct disk_events
*ev
;
2091 if (!disk
->fops
->check_events
|| !disk
->events
)
2094 ev
= kzalloc(sizeof(*ev
), GFP_KERNEL
);
2096 pr_warn("%s: failed to initialize events\n", disk
->disk_name
);
2100 INIT_LIST_HEAD(&ev
->node
);
2102 spin_lock_init(&ev
->lock
);
2103 mutex_init(&ev
->block_mutex
);
2105 ev
->poll_msecs
= -1;
2106 INIT_DELAYED_WORK(&ev
->dwork
, disk_events_workfn
);
2111 static void disk_add_events(struct gendisk
*disk
)
2113 /* FIXME: error handling */
2114 if (sysfs_create_files(&disk_to_dev(disk
)->kobj
, disk_events_attrs
) < 0)
2115 pr_warn("%s: failed to create sysfs files for events\n",
2121 mutex_lock(&disk_events_mutex
);
2122 list_add_tail(&disk
->ev
->node
, &disk_events
);
2123 mutex_unlock(&disk_events_mutex
);
2126 * Block count is initialized to 1 and the following initial
2127 * unblock kicks it into action.
2129 __disk_unblock_events(disk
, true);
2132 static void disk_del_events(struct gendisk
*disk
)
2135 disk_block_events(disk
);
2137 mutex_lock(&disk_events_mutex
);
2138 list_del_init(&disk
->ev
->node
);
2139 mutex_unlock(&disk_events_mutex
);
2142 sysfs_remove_files(&disk_to_dev(disk
)->kobj
, disk_events_attrs
);
2145 static void disk_release_events(struct gendisk
*disk
)
2147 /* the block count should be 1 from disk_del_events() */
2148 WARN_ON_ONCE(disk
->ev
&& disk
->ev
->block
!= 1);