5 #include <linux/module.h>
7 #include <linux/genhd.h>
8 #include <linux/kdev_t.h>
9 #include <linux/kernel.h>
10 #include <linux/blkdev.h>
11 #include <linux/backing-dev.h>
12 #include <linux/init.h>
13 #include <linux/spinlock.h>
14 #include <linux/proc_fs.h>
15 #include <linux/seq_file.h>
16 #include <linux/slab.h>
17 #include <linux/kmod.h>
18 #include <linux/kobj_map.h>
19 #include <linux/mutex.h>
20 #include <linux/idr.h>
21 #include <linux/log2.h>
22 #include <linux/pm_runtime.h>
23 #include <linux/badblocks.h>
27 static DEFINE_MUTEX(block_class_lock
);
28 struct kobject
*block_depr
;
30 /* for extended dynamic devt allocation, currently only one major is used */
31 #define NR_EXT_DEVT (1 << MINORBITS)
33 /* For extended devt allocation. ext_devt_lock prevents look up
34 * results from going away underneath its user.
36 static DEFINE_SPINLOCK(ext_devt_lock
);
37 static DEFINE_IDR(ext_devt_idr
);
39 static struct device_type disk_type
;
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 * disk_get_part - get partition
50 * @disk: disk to look partition from
51 * @partno: partition number
53 * Look for partition @partno from @disk. If found, increment
54 * reference count and return it.
60 * Pointer to the found partition on success, NULL if not found.
62 struct hd_struct
*disk_get_part(struct gendisk
*disk
, int partno
)
64 struct hd_struct
*part
= NULL
;
65 struct disk_part_tbl
*ptbl
;
67 if (unlikely(partno
< 0))
72 ptbl
= rcu_dereference(disk
->part_tbl
);
73 if (likely(partno
< ptbl
->len
)) {
74 part
= rcu_dereference(ptbl
->part
[partno
]);
76 get_device(part_to_dev(part
));
83 EXPORT_SYMBOL_GPL(disk_get_part
);
86 * disk_part_iter_init - initialize partition iterator
87 * @piter: iterator to initialize
88 * @disk: disk to iterate over
89 * @flags: DISK_PITER_* flags
91 * Initialize @piter so that it iterates over partitions of @disk.
96 void disk_part_iter_init(struct disk_part_iter
*piter
, struct gendisk
*disk
,
99 struct disk_part_tbl
*ptbl
;
102 ptbl
= rcu_dereference(disk
->part_tbl
);
107 if (flags
& DISK_PITER_REVERSE
)
108 piter
->idx
= ptbl
->len
- 1;
109 else if (flags
& (DISK_PITER_INCL_PART0
| DISK_PITER_INCL_EMPTY_PART0
))
114 piter
->flags
= flags
;
118 EXPORT_SYMBOL_GPL(disk_part_iter_init
);
121 * disk_part_iter_next - proceed iterator to the next partition and return it
122 * @piter: iterator of interest
124 * Proceed @piter to the next partition and return it.
129 struct hd_struct
*disk_part_iter_next(struct disk_part_iter
*piter
)
131 struct disk_part_tbl
*ptbl
;
134 /* put the last partition */
135 disk_put_part(piter
->part
);
140 ptbl
= rcu_dereference(piter
->disk
->part_tbl
);
142 /* determine iteration parameters */
143 if (piter
->flags
& DISK_PITER_REVERSE
) {
145 if (piter
->flags
& (DISK_PITER_INCL_PART0
|
146 DISK_PITER_INCL_EMPTY_PART0
))
155 /* iterate to the next partition */
156 for (; piter
->idx
!= end
; piter
->idx
+= inc
) {
157 struct hd_struct
*part
;
159 part
= rcu_dereference(ptbl
->part
[piter
->idx
]);
162 if (!part_nr_sects_read(part
) &&
163 !(piter
->flags
& DISK_PITER_INCL_EMPTY
) &&
164 !(piter
->flags
& DISK_PITER_INCL_EMPTY_PART0
&&
168 get_device(part_to_dev(part
));
178 EXPORT_SYMBOL_GPL(disk_part_iter_next
);
181 * disk_part_iter_exit - finish up partition iteration
182 * @piter: iter of interest
184 * Called when iteration is over. Cleans up @piter.
189 void disk_part_iter_exit(struct disk_part_iter
*piter
)
191 disk_put_part(piter
->part
);
194 EXPORT_SYMBOL_GPL(disk_part_iter_exit
);
196 static inline int sector_in_part(struct hd_struct
*part
, sector_t sector
)
198 return part
->start_sect
<= sector
&&
199 sector
< part
->start_sect
+ part_nr_sects_read(part
);
203 * disk_map_sector_rcu - map sector to partition
204 * @disk: gendisk of interest
205 * @sector: sector to map
207 * Find out which partition @sector maps to on @disk. This is
208 * primarily used for stats accounting.
211 * RCU read locked. The returned partition pointer is valid only
212 * while preemption is disabled.
215 * Found partition on success, part0 is returned if no partition matches
217 struct hd_struct
*disk_map_sector_rcu(struct gendisk
*disk
, sector_t sector
)
219 struct disk_part_tbl
*ptbl
;
220 struct hd_struct
*part
;
223 ptbl
= rcu_dereference(disk
->part_tbl
);
225 part
= rcu_dereference(ptbl
->last_lookup
);
226 if (part
&& sector_in_part(part
, sector
))
229 for (i
= 1; i
< ptbl
->len
; i
++) {
230 part
= rcu_dereference(ptbl
->part
[i
]);
232 if (part
&& sector_in_part(part
, sector
)) {
233 rcu_assign_pointer(ptbl
->last_lookup
, part
);
239 EXPORT_SYMBOL_GPL(disk_map_sector_rcu
);
242 * Can be deleted altogether. Later.
245 static struct blk_major_name
{
246 struct blk_major_name
*next
;
249 } *major_names
[BLKDEV_MAJOR_HASH_SIZE
];
251 /* index in the above - for now: assume no multimajor ranges */
252 static inline int major_to_index(unsigned major
)
254 return major
% BLKDEV_MAJOR_HASH_SIZE
;
257 #ifdef CONFIG_PROC_FS
258 void blkdev_show(struct seq_file
*seqf
, off_t offset
)
260 struct blk_major_name
*dp
;
262 if (offset
< BLKDEV_MAJOR_HASH_SIZE
) {
263 mutex_lock(&block_class_lock
);
264 for (dp
= major_names
[offset
]; dp
; dp
= dp
->next
)
265 seq_printf(seqf
, "%3d %s\n", dp
->major
, dp
->name
);
266 mutex_unlock(&block_class_lock
);
269 #endif /* CONFIG_PROC_FS */
272 * register_blkdev - register a new block device
274 * @major: the requested major device number [1..255]. If @major=0, try to
275 * allocate any unused major number.
276 * @name: the name of the new block device as a zero terminated string
278 * The @name must be unique within the system.
280 * The return value depends on the @major input parameter.
281 * - if a major device number was requested in range [1..255] then the
282 * function returns zero on success, or a negative error code
283 * - if any unused major number was requested with @major=0 parameter
284 * then the return value is the allocated major number in range
285 * [1..255] or a negative error code otherwise
287 int register_blkdev(unsigned int major
, const char *name
)
289 struct blk_major_name
**n
, *p
;
292 mutex_lock(&block_class_lock
);
296 for (index
= ARRAY_SIZE(major_names
)-1; index
> 0; index
--) {
297 if (major_names
[index
] == NULL
)
302 printk("register_blkdev: failed to get major for %s\n",
311 p
= kmalloc(sizeof(struct blk_major_name
), GFP_KERNEL
);
318 strlcpy(p
->name
, name
, sizeof(p
->name
));
320 index
= major_to_index(major
);
322 for (n
= &major_names
[index
]; *n
; n
= &(*n
)->next
) {
323 if ((*n
)->major
== major
)
332 printk("register_blkdev: cannot get major %d for %s\n",
337 mutex_unlock(&block_class_lock
);
341 EXPORT_SYMBOL(register_blkdev
);
343 void unregister_blkdev(unsigned int major
, const char *name
)
345 struct blk_major_name
**n
;
346 struct blk_major_name
*p
= NULL
;
347 int index
= major_to_index(major
);
349 mutex_lock(&block_class_lock
);
350 for (n
= &major_names
[index
]; *n
; n
= &(*n
)->next
)
351 if ((*n
)->major
== major
)
353 if (!*n
|| strcmp((*n
)->name
, name
)) {
359 mutex_unlock(&block_class_lock
);
363 EXPORT_SYMBOL(unregister_blkdev
);
365 static struct kobj_map
*bdev_map
;
368 * blk_mangle_minor - scatter minor numbers apart
369 * @minor: minor number to mangle
371 * Scatter consecutively allocated @minor number apart if MANGLE_DEVT
372 * is enabled. Mangling twice gives the original value.
380 static int blk_mangle_minor(int minor
)
382 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
385 for (i
= 0; i
< MINORBITS
/ 2; i
++) {
386 int low
= minor
& (1 << i
);
387 int high
= minor
& (1 << (MINORBITS
- 1 - i
));
388 int distance
= MINORBITS
- 1 - 2 * i
;
390 minor
^= low
| high
; /* clear both bits */
391 low
<<= distance
; /* swap the positions */
393 minor
|= low
| high
; /* and set */
400 * blk_alloc_devt - allocate a dev_t for a partition
401 * @part: partition to allocate dev_t for
402 * @devt: out parameter for resulting dev_t
404 * Allocate a dev_t for block device.
407 * 0 on success, allocated dev_t is returned in *@devt. -errno on
413 int blk_alloc_devt(struct hd_struct
*part
, dev_t
*devt
)
415 struct gendisk
*disk
= part_to_disk(part
);
418 /* in consecutive minor range? */
419 if (part
->partno
< disk
->minors
) {
420 *devt
= MKDEV(disk
->major
, disk
->first_minor
+ part
->partno
);
424 /* allocate ext devt */
425 idr_preload(GFP_KERNEL
);
427 spin_lock_bh(&ext_devt_lock
);
428 idx
= idr_alloc(&ext_devt_idr
, part
, 0, NR_EXT_DEVT
, GFP_NOWAIT
);
429 spin_unlock_bh(&ext_devt_lock
);
433 return idx
== -ENOSPC
? -EBUSY
: idx
;
435 *devt
= MKDEV(BLOCK_EXT_MAJOR
, blk_mangle_minor(idx
));
440 * blk_free_devt - free a dev_t
441 * @devt: dev_t to free
443 * Free @devt which was allocated using blk_alloc_devt().
448 void blk_free_devt(dev_t devt
)
450 if (devt
== MKDEV(0, 0))
453 if (MAJOR(devt
) == BLOCK_EXT_MAJOR
) {
454 spin_lock_bh(&ext_devt_lock
);
455 idr_remove(&ext_devt_idr
, blk_mangle_minor(MINOR(devt
)));
456 spin_unlock_bh(&ext_devt_lock
);
460 static char *bdevt_str(dev_t devt
, char *buf
)
462 if (MAJOR(devt
) <= 0xff && MINOR(devt
) <= 0xff) {
463 char tbuf
[BDEVT_SIZE
];
464 snprintf(tbuf
, BDEVT_SIZE
, "%02x%02x", MAJOR(devt
), MINOR(devt
));
465 snprintf(buf
, BDEVT_SIZE
, "%-9s", tbuf
);
467 snprintf(buf
, BDEVT_SIZE
, "%03x:%05x", MAJOR(devt
), MINOR(devt
));
473 * Register device numbers dev..(dev+range-1)
474 * range must be nonzero
475 * The hash chain is sorted on range, so that subranges can override.
477 void blk_register_region(dev_t devt
, unsigned long range
, struct module
*module
,
478 struct kobject
*(*probe
)(dev_t
, int *, void *),
479 int (*lock
)(dev_t
, void *), void *data
)
481 kobj_map(bdev_map
, devt
, range
, module
, probe
, lock
, data
);
484 EXPORT_SYMBOL(blk_register_region
);
486 void blk_unregister_region(dev_t devt
, unsigned long range
)
488 kobj_unmap(bdev_map
, devt
, range
);
491 EXPORT_SYMBOL(blk_unregister_region
);
493 static struct kobject
*exact_match(dev_t devt
, int *partno
, void *data
)
495 struct gendisk
*p
= data
;
497 return &disk_to_dev(p
)->kobj
;
500 static int exact_lock(dev_t devt
, void *data
)
502 struct gendisk
*p
= data
;
509 static void register_disk(struct device
*parent
, struct gendisk
*disk
)
511 struct device
*ddev
= disk_to_dev(disk
);
512 struct block_device
*bdev
;
513 struct disk_part_iter piter
;
514 struct hd_struct
*part
;
517 ddev
->parent
= parent
;
519 dev_set_name(ddev
, "%s", disk
->disk_name
);
521 /* delay uevents, until we scanned partition table */
522 dev_set_uevent_suppress(ddev
, 1);
524 if (device_add(ddev
))
526 if (!sysfs_deprecated
) {
527 err
= sysfs_create_link(block_depr
, &ddev
->kobj
,
528 kobject_name(&ddev
->kobj
));
536 * avoid probable deadlock caused by allocating memory with
537 * GFP_KERNEL in runtime_resume callback of its all ancestor
540 pm_runtime_set_memalloc_noio(ddev
, true);
542 disk
->part0
.holder_dir
= kobject_create_and_add("holders", &ddev
->kobj
);
543 disk
->slave_dir
= kobject_create_and_add("slaves", &ddev
->kobj
);
545 /* No minors to use for partitions */
546 if (!disk_part_scan_enabled(disk
))
549 /* No such device (e.g., media were just removed) */
550 if (!get_capacity(disk
))
553 bdev
= bdget_disk(disk
, 0);
557 bdev
->bd_invalidated
= 1;
558 err
= blkdev_get(bdev
, FMODE_READ
, NULL
);
561 blkdev_put(bdev
, FMODE_READ
);
564 /* announce disk after possible partitions are created */
565 dev_set_uevent_suppress(ddev
, 0);
566 kobject_uevent(&ddev
->kobj
, KOBJ_ADD
);
568 /* announce possible partitions */
569 disk_part_iter_init(&piter
, disk
, 0);
570 while ((part
= disk_part_iter_next(&piter
)))
571 kobject_uevent(&part_to_dev(part
)->kobj
, KOBJ_ADD
);
572 disk_part_iter_exit(&piter
);
575 void put_disk_devt(struct disk_devt
*disk_devt
)
577 if (disk_devt
&& atomic_dec_and_test(&disk_devt
->count
))
578 disk_devt
->release(disk_devt
);
580 EXPORT_SYMBOL(put_disk_devt
);
582 void get_disk_devt(struct disk_devt
*disk_devt
)
585 atomic_inc(&disk_devt
->count
);
587 EXPORT_SYMBOL(get_disk_devt
);
590 * device_add_disk - add partitioning information to kernel list
591 * @parent: parent device for the disk
592 * @disk: per-device partitioning information
594 * This function registers the partitioning information in @disk
597 * FIXME: error handling
599 void device_add_disk(struct device
*parent
, struct gendisk
*disk
)
601 struct backing_dev_info
*bdi
;
605 /* minors == 0 indicates to use ext devt from part0 and should
606 * be accompanied with EXT_DEVT flag. Make sure all
607 * parameters make sense.
609 WARN_ON(disk
->minors
&& !(disk
->major
|| disk
->first_minor
));
610 WARN_ON(!disk
->minors
&& !(disk
->flags
& GENHD_FL_EXT_DEVT
));
612 disk
->flags
|= GENHD_FL_UP
;
614 retval
= blk_alloc_devt(&disk
->part0
, &devt
);
619 disk_to_dev(disk
)->devt
= devt
;
621 /* ->major and ->first_minor aren't supposed to be
622 * dereferenced from here on, but set them just in case.
624 disk
->major
= MAJOR(devt
);
625 disk
->first_minor
= MINOR(devt
);
627 disk_alloc_events(disk
);
630 * Take a reference on the devt and assign it to queue since it
631 * must not be reallocated while the bdi is registered
633 disk
->queue
->disk_devt
= disk
->disk_devt
;
634 get_disk_devt(disk
->disk_devt
);
636 /* Register BDI before referencing it from bdev */
637 bdi
= disk
->queue
->backing_dev_info
;
638 bdi_register_owner(bdi
, disk_to_dev(disk
));
640 blk_register_region(disk_devt(disk
), disk
->minors
, NULL
,
641 exact_match
, exact_lock
, disk
);
642 register_disk(parent
, disk
);
643 blk_register_queue(disk
);
646 * Take an extra ref on queue which will be put on disk_release()
647 * so that it sticks around as long as @disk is there.
649 WARN_ON_ONCE(!blk_get_queue(disk
->queue
));
651 retval
= sysfs_create_link(&disk_to_dev(disk
)->kobj
, &bdi
->dev
->kobj
,
655 disk_add_events(disk
);
656 blk_integrity_add(disk
);
658 EXPORT_SYMBOL(device_add_disk
);
660 void del_gendisk(struct gendisk
*disk
)
662 struct disk_part_iter piter
;
663 struct hd_struct
*part
;
665 blk_integrity_del(disk
);
666 disk_del_events(disk
);
668 /* invalidate stuff */
669 disk_part_iter_init(&piter
, disk
,
670 DISK_PITER_INCL_EMPTY
| DISK_PITER_REVERSE
);
671 while ((part
= disk_part_iter_next(&piter
))) {
672 invalidate_partition(disk
, part
->partno
);
673 bdev_unhash_inode(part_devt(part
));
674 delete_partition(disk
, part
->partno
);
676 disk_part_iter_exit(&piter
);
678 invalidate_partition(disk
, 0);
679 bdev_unhash_inode(disk_devt(disk
));
680 set_capacity(disk
, 0);
681 disk
->flags
&= ~GENHD_FL_UP
;
683 sysfs_remove_link(&disk_to_dev(disk
)->kobj
, "bdi");
685 * Unregister bdi before releasing device numbers (as they can get
686 * reused and we'd get clashes in sysfs).
688 bdi_unregister(disk
->queue
->backing_dev_info
);
689 blk_unregister_queue(disk
);
690 blk_unregister_region(disk_devt(disk
), disk
->minors
);
692 part_stat_set_all(&disk
->part0
, 0);
693 disk
->part0
.stamp
= 0;
695 kobject_put(disk
->part0
.holder_dir
);
696 kobject_put(disk
->slave_dir
);
697 if (!sysfs_deprecated
)
698 sysfs_remove_link(block_depr
, dev_name(disk_to_dev(disk
)));
699 pm_runtime_set_memalloc_noio(disk_to_dev(disk
), false);
700 device_del(disk_to_dev(disk
));
702 EXPORT_SYMBOL(del_gendisk
);
704 /* sysfs access to bad-blocks list. */
705 static ssize_t
disk_badblocks_show(struct device
*dev
,
706 struct device_attribute
*attr
,
709 struct gendisk
*disk
= dev_to_disk(dev
);
712 return sprintf(page
, "\n");
714 return badblocks_show(disk
->bb
, page
, 0);
717 static ssize_t
disk_badblocks_store(struct device
*dev
,
718 struct device_attribute
*attr
,
719 const char *page
, size_t len
)
721 struct gendisk
*disk
= dev_to_disk(dev
);
726 return badblocks_store(disk
->bb
, page
, len
, 0);
730 * get_gendisk - get partitioning information for a given device
731 * @devt: device to get partitioning information for
732 * @partno: returned partition index
734 * This function gets the structure containing partitioning
735 * information for the given device @devt.
737 struct gendisk
*get_gendisk(dev_t devt
, int *partno
)
739 struct gendisk
*disk
= NULL
;
741 if (MAJOR(devt
) != BLOCK_EXT_MAJOR
) {
742 struct kobject
*kobj
;
744 kobj
= kobj_lookup(bdev_map
, devt
, partno
);
746 disk
= dev_to_disk(kobj_to_dev(kobj
));
748 struct hd_struct
*part
;
750 spin_lock_bh(&ext_devt_lock
);
751 part
= idr_find(&ext_devt_idr
, blk_mangle_minor(MINOR(devt
)));
752 if (part
&& get_disk(part_to_disk(part
))) {
753 *partno
= part
->partno
;
754 disk
= part_to_disk(part
);
756 spin_unlock_bh(&ext_devt_lock
);
761 EXPORT_SYMBOL(get_gendisk
);
764 * bdget_disk - do bdget() by gendisk and partition number
765 * @disk: gendisk of interest
766 * @partno: partition number
768 * Find partition @partno from @disk, do bdget() on it.
774 * Resulting block_device on success, NULL on failure.
776 struct block_device
*bdget_disk(struct gendisk
*disk
, int partno
)
778 struct hd_struct
*part
;
779 struct block_device
*bdev
= NULL
;
781 part
= disk_get_part(disk
, partno
);
783 bdev
= bdget(part_devt(part
));
788 EXPORT_SYMBOL(bdget_disk
);
791 * print a full list of all partitions - intended for places where the root
792 * filesystem can't be mounted and thus to give the victim some idea of what
795 void __init
printk_all_partitions(void)
797 struct class_dev_iter iter
;
800 class_dev_iter_init(&iter
, &block_class
, NULL
, &disk_type
);
801 while ((dev
= class_dev_iter_next(&iter
))) {
802 struct gendisk
*disk
= dev_to_disk(dev
);
803 struct disk_part_iter piter
;
804 struct hd_struct
*part
;
805 char name_buf
[BDEVNAME_SIZE
];
806 char devt_buf
[BDEVT_SIZE
];
809 * Don't show empty devices or things that have been
812 if (get_capacity(disk
) == 0 ||
813 (disk
->flags
& GENHD_FL_SUPPRESS_PARTITION_INFO
))
817 * Note, unlike /proc/partitions, I am showing the
818 * numbers in hex - the same format as the root=
821 disk_part_iter_init(&piter
, disk
, DISK_PITER_INCL_PART0
);
822 while ((part
= disk_part_iter_next(&piter
))) {
823 bool is_part0
= part
== &disk
->part0
;
825 printk("%s%s %10llu %s %s", is_part0
? "" : " ",
826 bdevt_str(part_devt(part
), devt_buf
),
827 (unsigned long long)part_nr_sects_read(part
) >> 1
828 , disk_name(disk
, part
->partno
, name_buf
),
829 part
->info
? part
->info
->uuid
: "");
831 if (dev
->parent
&& dev
->parent
->driver
)
832 printk(" driver: %s\n",
833 dev
->parent
->driver
->name
);
835 printk(" (driver?)\n");
839 disk_part_iter_exit(&piter
);
841 class_dev_iter_exit(&iter
);
844 #ifdef CONFIG_PROC_FS
846 static void *disk_seqf_start(struct seq_file
*seqf
, loff_t
*pos
)
849 struct class_dev_iter
*iter
;
852 iter
= kmalloc(sizeof(*iter
), GFP_KERNEL
);
854 return ERR_PTR(-ENOMEM
);
856 seqf
->private = iter
;
857 class_dev_iter_init(iter
, &block_class
, NULL
, &disk_type
);
859 dev
= class_dev_iter_next(iter
);
864 return dev_to_disk(dev
);
867 static void *disk_seqf_next(struct seq_file
*seqf
, void *v
, loff_t
*pos
)
872 dev
= class_dev_iter_next(seqf
->private);
874 return dev_to_disk(dev
);
879 static void disk_seqf_stop(struct seq_file
*seqf
, void *v
)
881 struct class_dev_iter
*iter
= seqf
->private;
883 /* stop is called even after start failed :-( */
885 class_dev_iter_exit(iter
);
887 seqf
->private = NULL
;
891 static void *show_partition_start(struct seq_file
*seqf
, loff_t
*pos
)
895 p
= disk_seqf_start(seqf
, pos
);
896 if (!IS_ERR_OR_NULL(p
) && !*pos
)
897 seq_puts(seqf
, "major minor #blocks name\n\n");
901 static int show_partition(struct seq_file
*seqf
, void *v
)
903 struct gendisk
*sgp
= v
;
904 struct disk_part_iter piter
;
905 struct hd_struct
*part
;
906 char buf
[BDEVNAME_SIZE
];
908 /* Don't show non-partitionable removeable devices or empty devices */
909 if (!get_capacity(sgp
) || (!disk_max_parts(sgp
) &&
910 (sgp
->flags
& GENHD_FL_REMOVABLE
)))
912 if (sgp
->flags
& GENHD_FL_SUPPRESS_PARTITION_INFO
)
915 /* show the full disk and all non-0 size partitions of it */
916 disk_part_iter_init(&piter
, sgp
, DISK_PITER_INCL_PART0
);
917 while ((part
= disk_part_iter_next(&piter
)))
918 seq_printf(seqf
, "%4d %7d %10llu %s\n",
919 MAJOR(part_devt(part
)), MINOR(part_devt(part
)),
920 (unsigned long long)part_nr_sects_read(part
) >> 1,
921 disk_name(sgp
, part
->partno
, buf
));
922 disk_part_iter_exit(&piter
);
927 static const struct seq_operations partitions_op
= {
928 .start
= show_partition_start
,
929 .next
= disk_seqf_next
,
930 .stop
= disk_seqf_stop
,
931 .show
= show_partition
934 static int partitions_open(struct inode
*inode
, struct file
*file
)
936 return seq_open(file
, &partitions_op
);
939 static const struct file_operations proc_partitions_operations
= {
940 .open
= partitions_open
,
943 .release
= seq_release
,
948 static struct kobject
*base_probe(dev_t devt
, int *partno
, void *data
)
950 if (request_module("block-major-%d-%d", MAJOR(devt
), MINOR(devt
)) > 0)
951 /* Make old-style 2.4 aliases work */
952 request_module("block-major-%d", MAJOR(devt
));
956 static int __init
genhd_device_init(void)
960 block_class
.dev_kobj
= sysfs_dev_block_kobj
;
961 error
= class_register(&block_class
);
964 bdev_map
= kobj_map_init(base_probe
, &block_class_lock
);
967 register_blkdev(BLOCK_EXT_MAJOR
, "blkext");
969 /* create top-level block dir */
970 if (!sysfs_deprecated
)
971 block_depr
= kobject_create_and_add("block", NULL
);
975 subsys_initcall(genhd_device_init
);
977 static ssize_t
disk_range_show(struct device
*dev
,
978 struct device_attribute
*attr
, char *buf
)
980 struct gendisk
*disk
= dev_to_disk(dev
);
982 return sprintf(buf
, "%d\n", disk
->minors
);
985 static ssize_t
disk_ext_range_show(struct device
*dev
,
986 struct device_attribute
*attr
, char *buf
)
988 struct gendisk
*disk
= dev_to_disk(dev
);
990 return sprintf(buf
, "%d\n", disk_max_parts(disk
));
993 static ssize_t
disk_removable_show(struct device
*dev
,
994 struct device_attribute
*attr
, char *buf
)
996 struct gendisk
*disk
= dev_to_disk(dev
);
998 return sprintf(buf
, "%d\n",
999 (disk
->flags
& GENHD_FL_REMOVABLE
? 1 : 0));
1002 static ssize_t
disk_ro_show(struct device
*dev
,
1003 struct device_attribute
*attr
, char *buf
)
1005 struct gendisk
*disk
= dev_to_disk(dev
);
1007 return sprintf(buf
, "%d\n", get_disk_ro(disk
) ? 1 : 0);
1010 static ssize_t
disk_capability_show(struct device
*dev
,
1011 struct device_attribute
*attr
, char *buf
)
1013 struct gendisk
*disk
= dev_to_disk(dev
);
1015 return sprintf(buf
, "%x\n", disk
->flags
);
1018 static ssize_t
disk_alignment_offset_show(struct device
*dev
,
1019 struct device_attribute
*attr
,
1022 struct gendisk
*disk
= dev_to_disk(dev
);
1024 return sprintf(buf
, "%d\n", queue_alignment_offset(disk
->queue
));
1027 static ssize_t
disk_discard_alignment_show(struct device
*dev
,
1028 struct device_attribute
*attr
,
1031 struct gendisk
*disk
= dev_to_disk(dev
);
1033 return sprintf(buf
, "%d\n", queue_discard_alignment(disk
->queue
));
1036 static DEVICE_ATTR(range
, S_IRUGO
, disk_range_show
, NULL
);
1037 static DEVICE_ATTR(ext_range
, S_IRUGO
, disk_ext_range_show
, NULL
);
1038 static DEVICE_ATTR(removable
, S_IRUGO
, disk_removable_show
, NULL
);
1039 static DEVICE_ATTR(ro
, S_IRUGO
, disk_ro_show
, NULL
);
1040 static DEVICE_ATTR(size
, S_IRUGO
, part_size_show
, NULL
);
1041 static DEVICE_ATTR(alignment_offset
, S_IRUGO
, disk_alignment_offset_show
, NULL
);
1042 static DEVICE_ATTR(discard_alignment
, S_IRUGO
, disk_discard_alignment_show
,
1044 static DEVICE_ATTR(capability
, S_IRUGO
, disk_capability_show
, NULL
);
1045 static DEVICE_ATTR(stat
, S_IRUGO
, part_stat_show
, NULL
);
1046 static DEVICE_ATTR(inflight
, S_IRUGO
, part_inflight_show
, NULL
);
1047 static DEVICE_ATTR(badblocks
, S_IRUGO
| S_IWUSR
, disk_badblocks_show
,
1048 disk_badblocks_store
);
1049 #ifdef CONFIG_FAIL_MAKE_REQUEST
1050 static struct device_attribute dev_attr_fail
=
1051 __ATTR(make
-it
-fail
, S_IRUGO
|S_IWUSR
, part_fail_show
, part_fail_store
);
1053 #ifdef CONFIG_FAIL_IO_TIMEOUT
1054 static struct device_attribute dev_attr_fail_timeout
=
1055 __ATTR(io
-timeout
-fail
, S_IRUGO
|S_IWUSR
, part_timeout_show
,
1056 part_timeout_store
);
1059 static struct attribute
*disk_attrs
[] = {
1060 &dev_attr_range
.attr
,
1061 &dev_attr_ext_range
.attr
,
1062 &dev_attr_removable
.attr
,
1064 &dev_attr_size
.attr
,
1065 &dev_attr_alignment_offset
.attr
,
1066 &dev_attr_discard_alignment
.attr
,
1067 &dev_attr_capability
.attr
,
1068 &dev_attr_stat
.attr
,
1069 &dev_attr_inflight
.attr
,
1070 &dev_attr_badblocks
.attr
,
1071 #ifdef CONFIG_FAIL_MAKE_REQUEST
1072 &dev_attr_fail
.attr
,
1074 #ifdef CONFIG_FAIL_IO_TIMEOUT
1075 &dev_attr_fail_timeout
.attr
,
1080 static struct attribute_group disk_attr_group
= {
1081 .attrs
= disk_attrs
,
1084 static const struct attribute_group
*disk_attr_groups
[] = {
1090 * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way
1091 * @disk: disk to replace part_tbl for
1092 * @new_ptbl: new part_tbl to install
1094 * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The
1095 * original ptbl is freed using RCU callback.
1098 * Matching bd_mutx locked.
1100 static void disk_replace_part_tbl(struct gendisk
*disk
,
1101 struct disk_part_tbl
*new_ptbl
)
1103 struct disk_part_tbl
*old_ptbl
= disk
->part_tbl
;
1105 rcu_assign_pointer(disk
->part_tbl
, new_ptbl
);
1108 rcu_assign_pointer(old_ptbl
->last_lookup
, NULL
);
1109 kfree_rcu(old_ptbl
, rcu_head
);
1114 * disk_expand_part_tbl - expand disk->part_tbl
1115 * @disk: disk to expand part_tbl for
1116 * @partno: expand such that this partno can fit in
1118 * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl
1119 * uses RCU to allow unlocked dereferencing for stats and other stuff.
1122 * Matching bd_mutex locked, might sleep.
1125 * 0 on success, -errno on failure.
1127 int disk_expand_part_tbl(struct gendisk
*disk
, int partno
)
1129 struct disk_part_tbl
*old_ptbl
= disk
->part_tbl
;
1130 struct disk_part_tbl
*new_ptbl
;
1131 int len
= old_ptbl
? old_ptbl
->len
: 0;
1136 * check for int overflow, since we can get here from blkpg_ioctl()
1137 * with a user passed 'partno'.
1139 target
= partno
+ 1;
1143 /* disk_max_parts() is zero during initialization, ignore if so */
1144 if (disk_max_parts(disk
) && target
> disk_max_parts(disk
))
1150 size
= sizeof(*new_ptbl
) + target
* sizeof(new_ptbl
->part
[0]);
1151 new_ptbl
= kzalloc_node(size
, GFP_KERNEL
, disk
->node_id
);
1155 new_ptbl
->len
= target
;
1157 for (i
= 0; i
< len
; i
++)
1158 rcu_assign_pointer(new_ptbl
->part
[i
], old_ptbl
->part
[i
]);
1160 disk_replace_part_tbl(disk
, new_ptbl
);
1164 static void disk_release(struct device
*dev
)
1166 struct gendisk
*disk
= dev_to_disk(dev
);
1168 blk_free_devt(dev
->devt
);
1169 disk_release_events(disk
);
1170 kfree(disk
->random
);
1171 disk_replace_part_tbl(disk
, NULL
);
1172 hd_free_part(&disk
->part0
);
1174 blk_put_queue(disk
->queue
);
1177 struct class block_class
= {
1181 static char *block_devnode(struct device
*dev
, umode_t
*mode
,
1182 kuid_t
*uid
, kgid_t
*gid
)
1184 struct gendisk
*disk
= dev_to_disk(dev
);
1187 return disk
->devnode(disk
, mode
);
1191 static struct device_type disk_type
= {
1193 .groups
= disk_attr_groups
,
1194 .release
= disk_release
,
1195 .devnode
= block_devnode
,
1198 #ifdef CONFIG_PROC_FS
1200 * aggregate disk stat collector. Uses the same stats that the sysfs
1201 * entries do, above, but makes them available through one seq_file.
1203 * The output looks suspiciously like /proc/partitions with a bunch of
1206 static int diskstats_show(struct seq_file
*seqf
, void *v
)
1208 struct gendisk
*gp
= v
;
1209 struct disk_part_iter piter
;
1210 struct hd_struct
*hd
;
1211 char buf
[BDEVNAME_SIZE
];
1215 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1216 seq_puts(seqf, "major minor name"
1217 " rio rmerge rsect ruse wio wmerge "
1218 "wsect wuse running use aveq"
1222 disk_part_iter_init(&piter
, gp
, DISK_PITER_INCL_EMPTY_PART0
);
1223 while ((hd
= disk_part_iter_next(&piter
))) {
1224 cpu
= part_stat_lock();
1225 part_round_stats(cpu
, hd
);
1227 seq_printf(seqf
, "%4d %7d %s %lu %lu %lu "
1228 "%u %lu %lu %lu %u %u %u %u\n",
1229 MAJOR(part_devt(hd
)), MINOR(part_devt(hd
)),
1230 disk_name(gp
, hd
->partno
, buf
),
1231 part_stat_read(hd
, ios
[READ
]),
1232 part_stat_read(hd
, merges
[READ
]),
1233 part_stat_read(hd
, sectors
[READ
]),
1234 jiffies_to_msecs(part_stat_read(hd
, ticks
[READ
])),
1235 part_stat_read(hd
, ios
[WRITE
]),
1236 part_stat_read(hd
, merges
[WRITE
]),
1237 part_stat_read(hd
, sectors
[WRITE
]),
1238 jiffies_to_msecs(part_stat_read(hd
, ticks
[WRITE
])),
1240 jiffies_to_msecs(part_stat_read(hd
, io_ticks
)),
1241 jiffies_to_msecs(part_stat_read(hd
, time_in_queue
))
1244 disk_part_iter_exit(&piter
);
1249 static const struct seq_operations diskstats_op
= {
1250 .start
= disk_seqf_start
,
1251 .next
= disk_seqf_next
,
1252 .stop
= disk_seqf_stop
,
1253 .show
= diskstats_show
1256 static int diskstats_open(struct inode
*inode
, struct file
*file
)
1258 return seq_open(file
, &diskstats_op
);
1261 static const struct file_operations proc_diskstats_operations
= {
1262 .open
= diskstats_open
,
1264 .llseek
= seq_lseek
,
1265 .release
= seq_release
,
1268 static int __init
proc_genhd_init(void)
1270 proc_create("diskstats", 0, NULL
, &proc_diskstats_operations
);
1271 proc_create("partitions", 0, NULL
, &proc_partitions_operations
);
1274 module_init(proc_genhd_init
);
1275 #endif /* CONFIG_PROC_FS */
1277 dev_t
blk_lookup_devt(const char *name
, int partno
)
1279 dev_t devt
= MKDEV(0, 0);
1280 struct class_dev_iter iter
;
1283 class_dev_iter_init(&iter
, &block_class
, NULL
, &disk_type
);
1284 while ((dev
= class_dev_iter_next(&iter
))) {
1285 struct gendisk
*disk
= dev_to_disk(dev
);
1286 struct hd_struct
*part
;
1288 if (strcmp(dev_name(dev
), name
))
1291 if (partno
< disk
->minors
) {
1292 /* We need to return the right devno, even
1293 * if the partition doesn't exist yet.
1295 devt
= MKDEV(MAJOR(dev
->devt
),
1296 MINOR(dev
->devt
) + partno
);
1299 part
= disk_get_part(disk
, partno
);
1301 devt
= part_devt(part
);
1302 disk_put_part(part
);
1305 disk_put_part(part
);
1307 class_dev_iter_exit(&iter
);
1310 EXPORT_SYMBOL(blk_lookup_devt
);
1312 struct gendisk
*alloc_disk(int minors
)
1314 return alloc_disk_node(minors
, NUMA_NO_NODE
);
1316 EXPORT_SYMBOL(alloc_disk
);
1318 struct gendisk
*alloc_disk_node(int minors
, int node_id
)
1320 struct gendisk
*disk
;
1322 disk
= kzalloc_node(sizeof(struct gendisk
), GFP_KERNEL
, node_id
);
1324 if (!init_part_stats(&disk
->part0
)) {
1328 disk
->node_id
= node_id
;
1329 if (disk_expand_part_tbl(disk
, 0)) {
1330 free_part_stats(&disk
->part0
);
1334 disk
->part_tbl
->part
[0] = &disk
->part0
;
1337 * set_capacity() and get_capacity() currently don't use
1338 * seqcounter to read/update the part0->nr_sects. Still init
1339 * the counter as we can read the sectors in IO submission
1340 * patch using seqence counters.
1342 * TODO: Ideally set_capacity() and get_capacity() should be
1343 * converted to make use of bd_mutex and sequence counters.
1345 seqcount_init(&disk
->part0
.nr_sects_seq
);
1346 if (hd_ref_init(&disk
->part0
)) {
1347 hd_free_part(&disk
->part0
);
1352 disk
->minors
= minors
;
1353 rand_initialize_disk(disk
);
1354 disk_to_dev(disk
)->class = &block_class
;
1355 disk_to_dev(disk
)->type
= &disk_type
;
1356 device_initialize(disk_to_dev(disk
));
1360 EXPORT_SYMBOL(alloc_disk_node
);
1362 struct kobject
*get_disk(struct gendisk
*disk
)
1364 struct module
*owner
;
1365 struct kobject
*kobj
;
1369 owner
= disk
->fops
->owner
;
1370 if (owner
&& !try_module_get(owner
))
1372 kobj
= kobject_get(&disk_to_dev(disk
)->kobj
);
1381 EXPORT_SYMBOL(get_disk
);
1383 void put_disk(struct gendisk
*disk
)
1386 kobject_put(&disk_to_dev(disk
)->kobj
);
1389 EXPORT_SYMBOL(put_disk
);
1391 static void set_disk_ro_uevent(struct gendisk
*gd
, int ro
)
1393 char event
[] = "DISK_RO=1";
1394 char *envp
[] = { event
, NULL
};
1398 kobject_uevent_env(&disk_to_dev(gd
)->kobj
, KOBJ_CHANGE
, envp
);
1401 void set_device_ro(struct block_device
*bdev
, int flag
)
1403 bdev
->bd_part
->policy
= flag
;
1406 EXPORT_SYMBOL(set_device_ro
);
1408 void set_disk_ro(struct gendisk
*disk
, int flag
)
1410 struct disk_part_iter piter
;
1411 struct hd_struct
*part
;
1413 if (disk
->part0
.policy
!= flag
) {
1414 set_disk_ro_uevent(disk
, flag
);
1415 disk
->part0
.policy
= flag
;
1418 disk_part_iter_init(&piter
, disk
, DISK_PITER_INCL_EMPTY
);
1419 while ((part
= disk_part_iter_next(&piter
)))
1420 part
->policy
= flag
;
1421 disk_part_iter_exit(&piter
);
1424 EXPORT_SYMBOL(set_disk_ro
);
1426 int bdev_read_only(struct block_device
*bdev
)
1430 return bdev
->bd_part
->policy
;
1433 EXPORT_SYMBOL(bdev_read_only
);
1435 int invalidate_partition(struct gendisk
*disk
, int partno
)
1438 struct block_device
*bdev
= bdget_disk(disk
, partno
);
1441 res
= __invalidate_device(bdev
, true);
1447 EXPORT_SYMBOL(invalidate_partition
);
1450 * Disk events - monitor disk events like media change and eject request.
1452 struct disk_events
{
1453 struct list_head node
; /* all disk_event's */
1454 struct gendisk
*disk
; /* the associated disk */
1457 struct mutex block_mutex
; /* protects blocking */
1458 int block
; /* event blocking depth */
1459 unsigned int pending
; /* events already sent out */
1460 unsigned int clearing
; /* events being cleared */
1462 long poll_msecs
; /* interval, -1 for default */
1463 struct delayed_work dwork
;
1466 static const char *disk_events_strs
[] = {
1467 [ilog2(DISK_EVENT_MEDIA_CHANGE
)] = "media_change",
1468 [ilog2(DISK_EVENT_EJECT_REQUEST
)] = "eject_request",
1471 static char *disk_uevents
[] = {
1472 [ilog2(DISK_EVENT_MEDIA_CHANGE
)] = "DISK_MEDIA_CHANGE=1",
1473 [ilog2(DISK_EVENT_EJECT_REQUEST
)] = "DISK_EJECT_REQUEST=1",
1476 /* list of all disk_events */
1477 static DEFINE_MUTEX(disk_events_mutex
);
1478 static LIST_HEAD(disk_events
);
1480 /* disable in-kernel polling by default */
1481 static unsigned long disk_events_dfl_poll_msecs
;
1483 static unsigned long disk_events_poll_jiffies(struct gendisk
*disk
)
1485 struct disk_events
*ev
= disk
->ev
;
1486 long intv_msecs
= 0;
1489 * If device-specific poll interval is set, always use it. If
1490 * the default is being used, poll iff there are events which
1491 * can't be monitored asynchronously.
1493 if (ev
->poll_msecs
>= 0)
1494 intv_msecs
= ev
->poll_msecs
;
1495 else if (disk
->events
& ~disk
->async_events
)
1496 intv_msecs
= disk_events_dfl_poll_msecs
;
1498 return msecs_to_jiffies(intv_msecs
);
1502 * disk_block_events - block and flush disk event checking
1503 * @disk: disk to block events for
1505 * On return from this function, it is guaranteed that event checking
1506 * isn't in progress and won't happen until unblocked by
1507 * disk_unblock_events(). Events blocking is counted and the actual
1508 * unblocking happens after the matching number of unblocks are done.
1510 * Note that this intentionally does not block event checking from
1511 * disk_clear_events().
1516 void disk_block_events(struct gendisk
*disk
)
1518 struct disk_events
*ev
= disk
->ev
;
1519 unsigned long flags
;
1526 * Outer mutex ensures that the first blocker completes canceling
1527 * the event work before further blockers are allowed to finish.
1529 mutex_lock(&ev
->block_mutex
);
1531 spin_lock_irqsave(&ev
->lock
, flags
);
1532 cancel
= !ev
->block
++;
1533 spin_unlock_irqrestore(&ev
->lock
, flags
);
1536 cancel_delayed_work_sync(&disk
->ev
->dwork
);
1538 mutex_unlock(&ev
->block_mutex
);
1541 static void __disk_unblock_events(struct gendisk
*disk
, bool check_now
)
1543 struct disk_events
*ev
= disk
->ev
;
1545 unsigned long flags
;
1547 spin_lock_irqsave(&ev
->lock
, flags
);
1549 if (WARN_ON_ONCE(ev
->block
<= 0))
1555 intv
= disk_events_poll_jiffies(disk
);
1557 queue_delayed_work(system_freezable_power_efficient_wq
,
1560 queue_delayed_work(system_freezable_power_efficient_wq
,
1563 spin_unlock_irqrestore(&ev
->lock
, flags
);
1567 * disk_unblock_events - unblock disk event checking
1568 * @disk: disk to unblock events for
1570 * Undo disk_block_events(). When the block count reaches zero, it
1571 * starts events polling if configured.
1574 * Don't care. Safe to call from irq context.
1576 void disk_unblock_events(struct gendisk
*disk
)
1579 __disk_unblock_events(disk
, false);
1583 * disk_flush_events - schedule immediate event checking and flushing
1584 * @disk: disk to check and flush events for
1585 * @mask: events to flush
1587 * Schedule immediate event checking on @disk if not blocked. Events in
1588 * @mask are scheduled to be cleared from the driver. Note that this
1589 * doesn't clear the events from @disk->ev.
1592 * If @mask is non-zero must be called with bdev->bd_mutex held.
1594 void disk_flush_events(struct gendisk
*disk
, unsigned int mask
)
1596 struct disk_events
*ev
= disk
->ev
;
1601 spin_lock_irq(&ev
->lock
);
1602 ev
->clearing
|= mask
;
1604 mod_delayed_work(system_freezable_power_efficient_wq
,
1606 spin_unlock_irq(&ev
->lock
);
1610 * disk_clear_events - synchronously check, clear and return pending events
1611 * @disk: disk to fetch and clear events from
1612 * @mask: mask of events to be fetched and cleared
1614 * Disk events are synchronously checked and pending events in @mask
1615 * are cleared and returned. This ignores the block count.
1620 unsigned int disk_clear_events(struct gendisk
*disk
, unsigned int mask
)
1622 const struct block_device_operations
*bdops
= disk
->fops
;
1623 struct disk_events
*ev
= disk
->ev
;
1624 unsigned int pending
;
1625 unsigned int clearing
= mask
;
1628 /* for drivers still using the old ->media_changed method */
1629 if ((mask
& DISK_EVENT_MEDIA_CHANGE
) &&
1630 bdops
->media_changed
&& bdops
->media_changed(disk
))
1631 return DISK_EVENT_MEDIA_CHANGE
;
1635 disk_block_events(disk
);
1638 * store the union of mask and ev->clearing on the stack so that the
1639 * race with disk_flush_events does not cause ambiguity (ev->clearing
1640 * can still be modified even if events are blocked).
1642 spin_lock_irq(&ev
->lock
);
1643 clearing
|= ev
->clearing
;
1645 spin_unlock_irq(&ev
->lock
);
1647 disk_check_events(ev
, &clearing
);
1649 * if ev->clearing is not 0, the disk_flush_events got called in the
1650 * middle of this function, so we want to run the workfn without delay.
1652 __disk_unblock_events(disk
, ev
->clearing
? true : false);
1654 /* then, fetch and clear pending events */
1655 spin_lock_irq(&ev
->lock
);
1656 pending
= ev
->pending
& mask
;
1657 ev
->pending
&= ~mask
;
1658 spin_unlock_irq(&ev
->lock
);
1659 WARN_ON_ONCE(clearing
& mask
);
1665 * Separate this part out so that a different pointer for clearing_ptr can be
1666 * passed in for disk_clear_events.
1668 static void disk_events_workfn(struct work_struct
*work
)
1670 struct delayed_work
*dwork
= to_delayed_work(work
);
1671 struct disk_events
*ev
= container_of(dwork
, struct disk_events
, dwork
);
1673 disk_check_events(ev
, &ev
->clearing
);
1676 static void disk_check_events(struct disk_events
*ev
,
1677 unsigned int *clearing_ptr
)
1679 struct gendisk
*disk
= ev
->disk
;
1680 char *envp
[ARRAY_SIZE(disk_uevents
) + 1] = { };
1681 unsigned int clearing
= *clearing_ptr
;
1682 unsigned int events
;
1684 int nr_events
= 0, i
;
1687 events
= disk
->fops
->check_events(disk
, clearing
);
1689 /* accumulate pending events and schedule next poll if necessary */
1690 spin_lock_irq(&ev
->lock
);
1692 events
&= ~ev
->pending
;
1693 ev
->pending
|= events
;
1694 *clearing_ptr
&= ~clearing
;
1696 intv
= disk_events_poll_jiffies(disk
);
1697 if (!ev
->block
&& intv
)
1698 queue_delayed_work(system_freezable_power_efficient_wq
,
1701 spin_unlock_irq(&ev
->lock
);
1704 * Tell userland about new events. Only the events listed in
1705 * @disk->events are reported. Unlisted events are processed the
1706 * same internally but never get reported to userland.
1708 for (i
= 0; i
< ARRAY_SIZE(disk_uevents
); i
++)
1709 if (events
& disk
->events
& (1 << i
))
1710 envp
[nr_events
++] = disk_uevents
[i
];
1713 kobject_uevent_env(&disk_to_dev(disk
)->kobj
, KOBJ_CHANGE
, envp
);
1717 * A disk events enabled device has the following sysfs nodes under
1718 * its /sys/block/X/ directory.
1720 * events : list of all supported events
1721 * events_async : list of events which can be detected w/o polling
1722 * events_poll_msecs : polling interval, 0: disable, -1: system default
1724 static ssize_t
__disk_events_show(unsigned int events
, char *buf
)
1726 const char *delim
= "";
1730 for (i
= 0; i
< ARRAY_SIZE(disk_events_strs
); i
++)
1731 if (events
& (1 << i
)) {
1732 pos
+= sprintf(buf
+ pos
, "%s%s",
1733 delim
, disk_events_strs
[i
]);
1737 pos
+= sprintf(buf
+ pos
, "\n");
1741 static ssize_t
disk_events_show(struct device
*dev
,
1742 struct device_attribute
*attr
, char *buf
)
1744 struct gendisk
*disk
= dev_to_disk(dev
);
1746 return __disk_events_show(disk
->events
, buf
);
1749 static ssize_t
disk_events_async_show(struct device
*dev
,
1750 struct device_attribute
*attr
, char *buf
)
1752 struct gendisk
*disk
= dev_to_disk(dev
);
1754 return __disk_events_show(disk
->async_events
, buf
);
1757 static ssize_t
disk_events_poll_msecs_show(struct device
*dev
,
1758 struct device_attribute
*attr
,
1761 struct gendisk
*disk
= dev_to_disk(dev
);
1763 return sprintf(buf
, "%ld\n", disk
->ev
->poll_msecs
);
1766 static ssize_t
disk_events_poll_msecs_store(struct device
*dev
,
1767 struct device_attribute
*attr
,
1768 const char *buf
, size_t count
)
1770 struct gendisk
*disk
= dev_to_disk(dev
);
1773 if (!count
|| !sscanf(buf
, "%ld", &intv
))
1776 if (intv
< 0 && intv
!= -1)
1779 disk_block_events(disk
);
1780 disk
->ev
->poll_msecs
= intv
;
1781 __disk_unblock_events(disk
, true);
1786 static const DEVICE_ATTR(events
, S_IRUGO
, disk_events_show
, NULL
);
1787 static const DEVICE_ATTR(events_async
, S_IRUGO
, disk_events_async_show
, NULL
);
1788 static const DEVICE_ATTR(events_poll_msecs
, S_IRUGO
|S_IWUSR
,
1789 disk_events_poll_msecs_show
,
1790 disk_events_poll_msecs_store
);
1792 static const struct attribute
*disk_events_attrs
[] = {
1793 &dev_attr_events
.attr
,
1794 &dev_attr_events_async
.attr
,
1795 &dev_attr_events_poll_msecs
.attr
,
1800 * The default polling interval can be specified by the kernel
1801 * parameter block.events_dfl_poll_msecs which defaults to 0
1802 * (disable). This can also be modified runtime by writing to
1803 * /sys/module/block/events_dfl_poll_msecs.
1805 static int disk_events_set_dfl_poll_msecs(const char *val
,
1806 const struct kernel_param
*kp
)
1808 struct disk_events
*ev
;
1811 ret
= param_set_ulong(val
, kp
);
1815 mutex_lock(&disk_events_mutex
);
1817 list_for_each_entry(ev
, &disk_events
, node
)
1818 disk_flush_events(ev
->disk
, 0);
1820 mutex_unlock(&disk_events_mutex
);
1825 static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops
= {
1826 .set
= disk_events_set_dfl_poll_msecs
,
1827 .get
= param_get_ulong
,
1830 #undef MODULE_PARAM_PREFIX
1831 #define MODULE_PARAM_PREFIX "block."
1833 module_param_cb(events_dfl_poll_msecs
, &disk_events_dfl_poll_msecs_param_ops
,
1834 &disk_events_dfl_poll_msecs
, 0644);
1837 * disk_{alloc|add|del|release}_events - initialize and destroy disk_events.
1839 static void disk_alloc_events(struct gendisk
*disk
)
1841 struct disk_events
*ev
;
1843 if (!disk
->fops
->check_events
)
1846 ev
= kzalloc(sizeof(*ev
), GFP_KERNEL
);
1848 pr_warn("%s: failed to initialize events\n", disk
->disk_name
);
1852 INIT_LIST_HEAD(&ev
->node
);
1854 spin_lock_init(&ev
->lock
);
1855 mutex_init(&ev
->block_mutex
);
1857 ev
->poll_msecs
= -1;
1858 INIT_DELAYED_WORK(&ev
->dwork
, disk_events_workfn
);
1863 static void disk_add_events(struct gendisk
*disk
)
1868 /* FIXME: error handling */
1869 if (sysfs_create_files(&disk_to_dev(disk
)->kobj
, disk_events_attrs
) < 0)
1870 pr_warn("%s: failed to create sysfs files for events\n",
1873 mutex_lock(&disk_events_mutex
);
1874 list_add_tail(&disk
->ev
->node
, &disk_events
);
1875 mutex_unlock(&disk_events_mutex
);
1878 * Block count is initialized to 1 and the following initial
1879 * unblock kicks it into action.
1881 __disk_unblock_events(disk
, true);
1884 static void disk_del_events(struct gendisk
*disk
)
1889 disk_block_events(disk
);
1891 mutex_lock(&disk_events_mutex
);
1892 list_del_init(&disk
->ev
->node
);
1893 mutex_unlock(&disk_events_mutex
);
1895 sysfs_remove_files(&disk_to_dev(disk
)->kobj
, disk_events_attrs
);
1898 static void disk_release_events(struct gendisk
*disk
)
1900 /* the block count should be 1 from disk_del_events() */
1901 WARN_ON_ONCE(disk
->ev
&& disk
->ev
->block
!= 1);