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/init.h>
12 #include <linux/spinlock.h>
13 #include <linux/proc_fs.h>
14 #include <linux/seq_file.h>
15 #include <linux/slab.h>
16 #include <linux/kmod.h>
17 #include <linux/kobj_map.h>
18 #include <linux/mutex.h>
19 #include <linux/idr.h>
20 #include <linux/log2.h>
21 #include <linux/pm_runtime.h>
25 static DEFINE_MUTEX(block_class_lock
);
26 struct kobject
*block_depr
;
28 /* for extended dynamic devt allocation, currently only one major is used */
29 #define NR_EXT_DEVT (1 << MINORBITS)
31 /* For extended devt allocation. ext_devt_mutex prevents look up
32 * results from going away underneath its user.
34 static DEFINE_MUTEX(ext_devt_mutex
);
35 static DEFINE_IDR(ext_devt_idr
);
37 static struct device_type disk_type
;
39 static void disk_check_events(struct disk_events
*ev
,
40 unsigned int *clearing_ptr
);
41 static void disk_alloc_events(struct gendisk
*disk
);
42 static void disk_add_events(struct gendisk
*disk
);
43 static void disk_del_events(struct gendisk
*disk
);
44 static void disk_release_events(struct gendisk
*disk
);
47 * disk_get_part - get partition
48 * @disk: disk to look partition from
49 * @partno: partition number
51 * Look for partition @partno from @disk. If found, increment
52 * reference count and return it.
58 * Pointer to the found partition on success, NULL if not found.
60 struct hd_struct
*disk_get_part(struct gendisk
*disk
, int partno
)
62 struct hd_struct
*part
= NULL
;
63 struct disk_part_tbl
*ptbl
;
65 if (unlikely(partno
< 0))
70 ptbl
= rcu_dereference(disk
->part_tbl
);
71 if (likely(partno
< ptbl
->len
)) {
72 part
= rcu_dereference(ptbl
->part
[partno
]);
74 get_device(part_to_dev(part
));
81 EXPORT_SYMBOL_GPL(disk_get_part
);
84 * disk_part_iter_init - initialize partition iterator
85 * @piter: iterator to initialize
86 * @disk: disk to iterate over
87 * @flags: DISK_PITER_* flags
89 * Initialize @piter so that it iterates over partitions of @disk.
94 void disk_part_iter_init(struct disk_part_iter
*piter
, struct gendisk
*disk
,
97 struct disk_part_tbl
*ptbl
;
100 ptbl
= rcu_dereference(disk
->part_tbl
);
105 if (flags
& DISK_PITER_REVERSE
)
106 piter
->idx
= ptbl
->len
- 1;
107 else if (flags
& (DISK_PITER_INCL_PART0
| DISK_PITER_INCL_EMPTY_PART0
))
112 piter
->flags
= flags
;
116 EXPORT_SYMBOL_GPL(disk_part_iter_init
);
119 * disk_part_iter_next - proceed iterator to the next partition and return it
120 * @piter: iterator of interest
122 * Proceed @piter to the next partition and return it.
127 struct hd_struct
*disk_part_iter_next(struct disk_part_iter
*piter
)
129 struct disk_part_tbl
*ptbl
;
132 /* put the last partition */
133 disk_put_part(piter
->part
);
138 ptbl
= rcu_dereference(piter
->disk
->part_tbl
);
140 /* determine iteration parameters */
141 if (piter
->flags
& DISK_PITER_REVERSE
) {
143 if (piter
->flags
& (DISK_PITER_INCL_PART0
|
144 DISK_PITER_INCL_EMPTY_PART0
))
153 /* iterate to the next partition */
154 for (; piter
->idx
!= end
; piter
->idx
+= inc
) {
155 struct hd_struct
*part
;
157 part
= rcu_dereference(ptbl
->part
[piter
->idx
]);
160 if (!part_nr_sects_read(part
) &&
161 !(piter
->flags
& DISK_PITER_INCL_EMPTY
) &&
162 !(piter
->flags
& DISK_PITER_INCL_EMPTY_PART0
&&
166 get_device(part_to_dev(part
));
176 EXPORT_SYMBOL_GPL(disk_part_iter_next
);
179 * disk_part_iter_exit - finish up partition iteration
180 * @piter: iter of interest
182 * Called when iteration is over. Cleans up @piter.
187 void disk_part_iter_exit(struct disk_part_iter
*piter
)
189 disk_put_part(piter
->part
);
192 EXPORT_SYMBOL_GPL(disk_part_iter_exit
);
194 static inline int sector_in_part(struct hd_struct
*part
, sector_t sector
)
196 return part
->start_sect
<= sector
&&
197 sector
< part
->start_sect
+ part_nr_sects_read(part
);
201 * disk_map_sector_rcu - map sector to partition
202 * @disk: gendisk of interest
203 * @sector: sector to map
205 * Find out which partition @sector maps to on @disk. This is
206 * primarily used for stats accounting.
209 * RCU read locked. The returned partition pointer is valid only
210 * while preemption is disabled.
213 * Found partition on success, part0 is returned if no partition matches
215 struct hd_struct
*disk_map_sector_rcu(struct gendisk
*disk
, sector_t sector
)
217 struct disk_part_tbl
*ptbl
;
218 struct hd_struct
*part
;
221 ptbl
= rcu_dereference(disk
->part_tbl
);
223 part
= rcu_dereference(ptbl
->last_lookup
);
224 if (part
&& sector_in_part(part
, sector
))
227 for (i
= 1; i
< ptbl
->len
; i
++) {
228 part
= rcu_dereference(ptbl
->part
[i
]);
230 if (part
&& sector_in_part(part
, sector
)) {
231 rcu_assign_pointer(ptbl
->last_lookup
, part
);
237 EXPORT_SYMBOL_GPL(disk_map_sector_rcu
);
240 * Can be deleted altogether. Later.
243 static struct blk_major_name
{
244 struct blk_major_name
*next
;
247 } *major_names
[BLKDEV_MAJOR_HASH_SIZE
];
249 /* index in the above - for now: assume no multimajor ranges */
250 static inline int major_to_index(unsigned major
)
252 return major
% BLKDEV_MAJOR_HASH_SIZE
;
255 #ifdef CONFIG_PROC_FS
256 void blkdev_show(struct seq_file
*seqf
, off_t offset
)
258 struct blk_major_name
*dp
;
260 if (offset
< BLKDEV_MAJOR_HASH_SIZE
) {
261 mutex_lock(&block_class_lock
);
262 for (dp
= major_names
[offset
]; dp
; dp
= dp
->next
)
263 seq_printf(seqf
, "%3d %s\n", dp
->major
, dp
->name
);
264 mutex_unlock(&block_class_lock
);
267 #endif /* CONFIG_PROC_FS */
270 * register_blkdev - register a new block device
272 * @major: the requested major device number [1..255]. If @major=0, try to
273 * allocate any unused major number.
274 * @name: the name of the new block device as a zero terminated string
276 * The @name must be unique within the system.
278 * The return value depends on the @major input parameter.
279 * - if a major device number was requested in range [1..255] then the
280 * function returns zero on success, or a negative error code
281 * - if any unused major number was requested with @major=0 parameter
282 * then the return value is the allocated major number in range
283 * [1..255] or a negative error code otherwise
285 int register_blkdev(unsigned int major
, const char *name
)
287 struct blk_major_name
**n
, *p
;
290 mutex_lock(&block_class_lock
);
294 for (index
= ARRAY_SIZE(major_names
)-1; index
> 0; index
--) {
295 if (major_names
[index
] == NULL
)
300 printk("register_blkdev: failed to get major for %s\n",
309 p
= kmalloc(sizeof(struct blk_major_name
), GFP_KERNEL
);
316 strlcpy(p
->name
, name
, sizeof(p
->name
));
318 index
= major_to_index(major
);
320 for (n
= &major_names
[index
]; *n
; n
= &(*n
)->next
) {
321 if ((*n
)->major
== major
)
330 printk("register_blkdev: cannot get major %d for %s\n",
335 mutex_unlock(&block_class_lock
);
339 EXPORT_SYMBOL(register_blkdev
);
341 void unregister_blkdev(unsigned int major
, const char *name
)
343 struct blk_major_name
**n
;
344 struct blk_major_name
*p
= NULL
;
345 int index
= major_to_index(major
);
347 mutex_lock(&block_class_lock
);
348 for (n
= &major_names
[index
]; *n
; n
= &(*n
)->next
)
349 if ((*n
)->major
== major
)
351 if (!*n
|| strcmp((*n
)->name
, name
)) {
357 mutex_unlock(&block_class_lock
);
361 EXPORT_SYMBOL(unregister_blkdev
);
363 static struct kobj_map
*bdev_map
;
366 * blk_mangle_minor - scatter minor numbers apart
367 * @minor: minor number to mangle
369 * Scatter consecutively allocated @minor number apart if MANGLE_DEVT
370 * is enabled. Mangling twice gives the original value.
378 static int blk_mangle_minor(int minor
)
380 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
383 for (i
= 0; i
< MINORBITS
/ 2; i
++) {
384 int low
= minor
& (1 << i
);
385 int high
= minor
& (1 << (MINORBITS
- 1 - i
));
386 int distance
= MINORBITS
- 1 - 2 * i
;
388 minor
^= low
| high
; /* clear both bits */
389 low
<<= distance
; /* swap the positions */
391 minor
|= low
| high
; /* and set */
398 * blk_alloc_devt - allocate a dev_t for a partition
399 * @part: partition to allocate dev_t for
400 * @devt: out parameter for resulting dev_t
402 * Allocate a dev_t for block device.
405 * 0 on success, allocated dev_t is returned in *@devt. -errno on
411 int blk_alloc_devt(struct hd_struct
*part
, dev_t
*devt
)
413 struct gendisk
*disk
= part_to_disk(part
);
416 /* in consecutive minor range? */
417 if (part
->partno
< disk
->minors
) {
418 *devt
= MKDEV(disk
->major
, disk
->first_minor
+ part
->partno
);
422 /* allocate ext devt */
424 if (!idr_pre_get(&ext_devt_idr
, GFP_KERNEL
))
426 mutex_lock(&ext_devt_mutex
);
427 rc
= idr_get_new(&ext_devt_idr
, part
, &idx
);
428 if (!rc
&& idx
>= NR_EXT_DEVT
) {
429 idr_remove(&ext_devt_idr
, idx
);
432 mutex_unlock(&ext_devt_mutex
);
433 } while (rc
== -EAGAIN
);
438 *devt
= MKDEV(BLOCK_EXT_MAJOR
, blk_mangle_minor(idx
));
443 * blk_free_devt - free a dev_t
444 * @devt: dev_t to free
446 * Free @devt which was allocated using blk_alloc_devt().
451 void blk_free_devt(dev_t devt
)
455 if (devt
== MKDEV(0, 0))
458 if (MAJOR(devt
) == BLOCK_EXT_MAJOR
) {
459 mutex_lock(&ext_devt_mutex
);
460 idr_remove(&ext_devt_idr
, blk_mangle_minor(MINOR(devt
)));
461 mutex_unlock(&ext_devt_mutex
);
465 static char *bdevt_str(dev_t devt
, char *buf
)
467 if (MAJOR(devt
) <= 0xff && MINOR(devt
) <= 0xff) {
468 char tbuf
[BDEVT_SIZE
];
469 snprintf(tbuf
, BDEVT_SIZE
, "%02x%02x", MAJOR(devt
), MINOR(devt
));
470 snprintf(buf
, BDEVT_SIZE
, "%-9s", tbuf
);
472 snprintf(buf
, BDEVT_SIZE
, "%03x:%05x", MAJOR(devt
), MINOR(devt
));
478 * Register device numbers dev..(dev+range-1)
479 * range must be nonzero
480 * The hash chain is sorted on range, so that subranges can override.
482 void blk_register_region(dev_t devt
, unsigned long range
, struct module
*module
,
483 struct kobject
*(*probe
)(dev_t
, int *, void *),
484 int (*lock
)(dev_t
, void *), void *data
)
486 kobj_map(bdev_map
, devt
, range
, module
, probe
, lock
, data
);
489 EXPORT_SYMBOL(blk_register_region
);
491 void blk_unregister_region(dev_t devt
, unsigned long range
)
493 kobj_unmap(bdev_map
, devt
, range
);
496 EXPORT_SYMBOL(blk_unregister_region
);
498 static struct kobject
*exact_match(dev_t devt
, int *partno
, void *data
)
500 struct gendisk
*p
= data
;
502 return &disk_to_dev(p
)->kobj
;
505 static int exact_lock(dev_t devt
, void *data
)
507 struct gendisk
*p
= data
;
514 static void register_disk(struct gendisk
*disk
)
516 struct device
*ddev
= disk_to_dev(disk
);
517 struct block_device
*bdev
;
518 struct disk_part_iter piter
;
519 struct hd_struct
*part
;
522 ddev
->parent
= disk
->driverfs_dev
;
524 dev_set_name(ddev
, disk
->disk_name
);
526 /* delay uevents, until we scanned partition table */
527 dev_set_uevent_suppress(ddev
, 1);
529 if (device_add(ddev
))
531 if (!sysfs_deprecated
) {
532 err
= sysfs_create_link(block_depr
, &ddev
->kobj
,
533 kobject_name(&ddev
->kobj
));
541 * avoid probable deadlock caused by allocating memory with
542 * GFP_KERNEL in runtime_resume callback of its all ancestor
545 pm_runtime_set_memalloc_noio(ddev
, true);
547 disk
->part0
.holder_dir
= kobject_create_and_add("holders", &ddev
->kobj
);
548 disk
->slave_dir
= kobject_create_and_add("slaves", &ddev
->kobj
);
550 /* No minors to use for partitions */
551 if (!disk_part_scan_enabled(disk
))
554 /* No such device (e.g., media were just removed) */
555 if (!get_capacity(disk
))
558 bdev
= bdget_disk(disk
, 0);
562 bdev
->bd_invalidated
= 1;
563 err
= blkdev_get(bdev
, FMODE_READ
, NULL
);
566 blkdev_put(bdev
, FMODE_READ
);
569 /* announce disk after possible partitions are created */
570 dev_set_uevent_suppress(ddev
, 0);
571 kobject_uevent(&ddev
->kobj
, KOBJ_ADD
);
573 /* announce possible partitions */
574 disk_part_iter_init(&piter
, disk
, 0);
575 while ((part
= disk_part_iter_next(&piter
)))
576 kobject_uevent(&part_to_dev(part
)->kobj
, KOBJ_ADD
);
577 disk_part_iter_exit(&piter
);
581 * add_disk - add partitioning information to kernel list
582 * @disk: per-device partitioning information
584 * This function registers the partitioning information in @disk
587 * FIXME: error handling
589 void add_disk(struct gendisk
*disk
)
591 struct backing_dev_info
*bdi
;
595 /* minors == 0 indicates to use ext devt from part0 and should
596 * be accompanied with EXT_DEVT flag. Make sure all
597 * parameters make sense.
599 WARN_ON(disk
->minors
&& !(disk
->major
|| disk
->first_minor
));
600 WARN_ON(!disk
->minors
&& !(disk
->flags
& GENHD_FL_EXT_DEVT
));
602 disk
->flags
|= GENHD_FL_UP
;
604 retval
= blk_alloc_devt(&disk
->part0
, &devt
);
609 disk_to_dev(disk
)->devt
= devt
;
611 /* ->major and ->first_minor aren't supposed to be
612 * dereferenced from here on, but set them just in case.
614 disk
->major
= MAJOR(devt
);
615 disk
->first_minor
= MINOR(devt
);
617 disk_alloc_events(disk
);
619 /* Register BDI before referencing it from bdev */
620 bdi
= &disk
->queue
->backing_dev_info
;
621 bdi_register_dev(bdi
, disk_devt(disk
));
623 blk_register_region(disk_devt(disk
), disk
->minors
, NULL
,
624 exact_match
, exact_lock
, disk
);
626 blk_register_queue(disk
);
629 * Take an extra ref on queue which will be put on disk_release()
630 * so that it sticks around as long as @disk is there.
632 WARN_ON_ONCE(!blk_get_queue(disk
->queue
));
634 retval
= sysfs_create_link(&disk_to_dev(disk
)->kobj
, &bdi
->dev
->kobj
,
638 disk_add_events(disk
);
640 EXPORT_SYMBOL(add_disk
);
642 void del_gendisk(struct gendisk
*disk
)
644 struct disk_part_iter piter
;
645 struct hd_struct
*part
;
647 disk_del_events(disk
);
649 /* invalidate stuff */
650 disk_part_iter_init(&piter
, disk
,
651 DISK_PITER_INCL_EMPTY
| DISK_PITER_REVERSE
);
652 while ((part
= disk_part_iter_next(&piter
))) {
653 invalidate_partition(disk
, part
->partno
);
654 delete_partition(disk
, part
->partno
);
656 disk_part_iter_exit(&piter
);
658 invalidate_partition(disk
, 0);
659 set_capacity(disk
, 0);
660 disk
->flags
&= ~GENHD_FL_UP
;
662 sysfs_remove_link(&disk_to_dev(disk
)->kobj
, "bdi");
663 bdi_unregister(&disk
->queue
->backing_dev_info
);
664 blk_unregister_queue(disk
);
665 blk_unregister_region(disk_devt(disk
), disk
->minors
);
667 part_stat_set_all(&disk
->part0
, 0);
668 disk
->part0
.stamp
= 0;
670 kobject_put(disk
->part0
.holder_dir
);
671 kobject_put(disk
->slave_dir
);
672 disk
->driverfs_dev
= NULL
;
673 if (!sysfs_deprecated
)
674 sysfs_remove_link(block_depr
, dev_name(disk_to_dev(disk
)));
675 pm_runtime_set_memalloc_noio(disk_to_dev(disk
), false);
676 device_del(disk_to_dev(disk
));
677 blk_free_devt(disk_to_dev(disk
)->devt
);
679 EXPORT_SYMBOL(del_gendisk
);
682 * get_gendisk - get partitioning information for a given device
683 * @devt: device to get partitioning information for
684 * @partno: returned partition index
686 * This function gets the structure containing partitioning
687 * information for the given device @devt.
689 struct gendisk
*get_gendisk(dev_t devt
, int *partno
)
691 struct gendisk
*disk
= NULL
;
693 if (MAJOR(devt
) != BLOCK_EXT_MAJOR
) {
694 struct kobject
*kobj
;
696 kobj
= kobj_lookup(bdev_map
, devt
, partno
);
698 disk
= dev_to_disk(kobj_to_dev(kobj
));
700 struct hd_struct
*part
;
702 mutex_lock(&ext_devt_mutex
);
703 part
= idr_find(&ext_devt_idr
, blk_mangle_minor(MINOR(devt
)));
704 if (part
&& get_disk(part_to_disk(part
))) {
705 *partno
= part
->partno
;
706 disk
= part_to_disk(part
);
708 mutex_unlock(&ext_devt_mutex
);
713 EXPORT_SYMBOL(get_gendisk
);
716 * bdget_disk - do bdget() by gendisk and partition number
717 * @disk: gendisk of interest
718 * @partno: partition number
720 * Find partition @partno from @disk, do bdget() on it.
726 * Resulting block_device on success, NULL on failure.
728 struct block_device
*bdget_disk(struct gendisk
*disk
, int partno
)
730 struct hd_struct
*part
;
731 struct block_device
*bdev
= NULL
;
733 part
= disk_get_part(disk
, partno
);
735 bdev
= bdget(part_devt(part
));
740 EXPORT_SYMBOL(bdget_disk
);
743 * print a full list of all partitions - intended for places where the root
744 * filesystem can't be mounted and thus to give the victim some idea of what
747 void __init
printk_all_partitions(void)
749 struct class_dev_iter iter
;
752 class_dev_iter_init(&iter
, &block_class
, NULL
, &disk_type
);
753 while ((dev
= class_dev_iter_next(&iter
))) {
754 struct gendisk
*disk
= dev_to_disk(dev
);
755 struct disk_part_iter piter
;
756 struct hd_struct
*part
;
757 char name_buf
[BDEVNAME_SIZE
];
758 char devt_buf
[BDEVT_SIZE
];
761 * Don't show empty devices or things that have been
764 if (get_capacity(disk
) == 0 ||
765 (disk
->flags
& GENHD_FL_SUPPRESS_PARTITION_INFO
))
769 * Note, unlike /proc/partitions, I am showing the
770 * numbers in hex - the same format as the root=
773 disk_part_iter_init(&piter
, disk
, DISK_PITER_INCL_PART0
);
774 while ((part
= disk_part_iter_next(&piter
))) {
775 bool is_part0
= part
== &disk
->part0
;
777 printk("%s%s %10llu %s %s", is_part0
? "" : " ",
778 bdevt_str(part_devt(part
), devt_buf
),
779 (unsigned long long)part_nr_sects_read(part
) >> 1
780 , disk_name(disk
, part
->partno
, name_buf
),
781 part
->info
? part
->info
->uuid
: "");
783 if (disk
->driverfs_dev
!= NULL
&&
784 disk
->driverfs_dev
->driver
!= NULL
)
785 printk(" driver: %s\n",
786 disk
->driverfs_dev
->driver
->name
);
788 printk(" (driver?)\n");
792 disk_part_iter_exit(&piter
);
794 class_dev_iter_exit(&iter
);
797 #ifdef CONFIG_PROC_FS
799 static void *disk_seqf_start(struct seq_file
*seqf
, loff_t
*pos
)
802 struct class_dev_iter
*iter
;
805 iter
= kmalloc(sizeof(*iter
), GFP_KERNEL
);
807 return ERR_PTR(-ENOMEM
);
809 seqf
->private = iter
;
810 class_dev_iter_init(iter
, &block_class
, NULL
, &disk_type
);
812 dev
= class_dev_iter_next(iter
);
817 return dev_to_disk(dev
);
820 static void *disk_seqf_next(struct seq_file
*seqf
, void *v
, loff_t
*pos
)
825 dev
= class_dev_iter_next(seqf
->private);
827 return dev_to_disk(dev
);
832 static void disk_seqf_stop(struct seq_file
*seqf
, void *v
)
834 struct class_dev_iter
*iter
= seqf
->private;
836 /* stop is called even after start failed :-( */
838 class_dev_iter_exit(iter
);
843 static void *show_partition_start(struct seq_file
*seqf
, loff_t
*pos
)
847 p
= disk_seqf_start(seqf
, pos
);
848 if (!IS_ERR_OR_NULL(p
) && !*pos
)
849 seq_puts(seqf
, "major minor #blocks name\n\n");
853 static int show_partition(struct seq_file
*seqf
, void *v
)
855 struct gendisk
*sgp
= v
;
856 struct disk_part_iter piter
;
857 struct hd_struct
*part
;
858 char buf
[BDEVNAME_SIZE
];
860 /* Don't show non-partitionable removeable devices or empty devices */
861 if (!get_capacity(sgp
) || (!disk_max_parts(sgp
) &&
862 (sgp
->flags
& GENHD_FL_REMOVABLE
)))
864 if (sgp
->flags
& GENHD_FL_SUPPRESS_PARTITION_INFO
)
867 /* show the full disk and all non-0 size partitions of it */
868 disk_part_iter_init(&piter
, sgp
, DISK_PITER_INCL_PART0
);
869 while ((part
= disk_part_iter_next(&piter
)))
870 seq_printf(seqf
, "%4d %7d %10llu %s\n",
871 MAJOR(part_devt(part
)), MINOR(part_devt(part
)),
872 (unsigned long long)part_nr_sects_read(part
) >> 1,
873 disk_name(sgp
, part
->partno
, buf
));
874 disk_part_iter_exit(&piter
);
879 static const struct seq_operations partitions_op
= {
880 .start
= show_partition_start
,
881 .next
= disk_seqf_next
,
882 .stop
= disk_seqf_stop
,
883 .show
= show_partition
886 static int partitions_open(struct inode
*inode
, struct file
*file
)
888 return seq_open(file
, &partitions_op
);
891 static const struct file_operations proc_partitions_operations
= {
892 .open
= partitions_open
,
895 .release
= seq_release
,
900 static struct kobject
*base_probe(dev_t devt
, int *partno
, void *data
)
902 if (request_module("block-major-%d-%d", MAJOR(devt
), MINOR(devt
)) > 0)
903 /* Make old-style 2.4 aliases work */
904 request_module("block-major-%d", MAJOR(devt
));
908 static int __init
genhd_device_init(void)
912 block_class
.dev_kobj
= sysfs_dev_block_kobj
;
913 error
= class_register(&block_class
);
916 bdev_map
= kobj_map_init(base_probe
, &block_class_lock
);
919 register_blkdev(BLOCK_EXT_MAJOR
, "blkext");
921 /* create top-level block dir */
922 if (!sysfs_deprecated
)
923 block_depr
= kobject_create_and_add("block", NULL
);
927 subsys_initcall(genhd_device_init
);
929 static ssize_t
disk_range_show(struct device
*dev
,
930 struct device_attribute
*attr
, char *buf
)
932 struct gendisk
*disk
= dev_to_disk(dev
);
934 return sprintf(buf
, "%d\n", disk
->minors
);
937 static ssize_t
disk_ext_range_show(struct device
*dev
,
938 struct device_attribute
*attr
, char *buf
)
940 struct gendisk
*disk
= dev_to_disk(dev
);
942 return sprintf(buf
, "%d\n", disk_max_parts(disk
));
945 static ssize_t
disk_removable_show(struct device
*dev
,
946 struct device_attribute
*attr
, char *buf
)
948 struct gendisk
*disk
= dev_to_disk(dev
);
950 return sprintf(buf
, "%d\n",
951 (disk
->flags
& GENHD_FL_REMOVABLE
? 1 : 0));
954 static ssize_t
disk_ro_show(struct device
*dev
,
955 struct device_attribute
*attr
, char *buf
)
957 struct gendisk
*disk
= dev_to_disk(dev
);
959 return sprintf(buf
, "%d\n", get_disk_ro(disk
) ? 1 : 0);
962 static ssize_t
disk_capability_show(struct device
*dev
,
963 struct device_attribute
*attr
, char *buf
)
965 struct gendisk
*disk
= dev_to_disk(dev
);
967 return sprintf(buf
, "%x\n", disk
->flags
);
970 static ssize_t
disk_alignment_offset_show(struct device
*dev
,
971 struct device_attribute
*attr
,
974 struct gendisk
*disk
= dev_to_disk(dev
);
976 return sprintf(buf
, "%d\n", queue_alignment_offset(disk
->queue
));
979 static ssize_t
disk_discard_alignment_show(struct device
*dev
,
980 struct device_attribute
*attr
,
983 struct gendisk
*disk
= dev_to_disk(dev
);
985 return sprintf(buf
, "%d\n", queue_discard_alignment(disk
->queue
));
988 static DEVICE_ATTR(range
, S_IRUGO
, disk_range_show
, NULL
);
989 static DEVICE_ATTR(ext_range
, S_IRUGO
, disk_ext_range_show
, NULL
);
990 static DEVICE_ATTR(removable
, S_IRUGO
, disk_removable_show
, NULL
);
991 static DEVICE_ATTR(ro
, S_IRUGO
, disk_ro_show
, NULL
);
992 static DEVICE_ATTR(size
, S_IRUGO
, part_size_show
, NULL
);
993 static DEVICE_ATTR(alignment_offset
, S_IRUGO
, disk_alignment_offset_show
, NULL
);
994 static DEVICE_ATTR(discard_alignment
, S_IRUGO
, disk_discard_alignment_show
,
996 static DEVICE_ATTR(capability
, S_IRUGO
, disk_capability_show
, NULL
);
997 static DEVICE_ATTR(stat
, S_IRUGO
, part_stat_show
, NULL
);
998 static DEVICE_ATTR(inflight
, S_IRUGO
, part_inflight_show
, NULL
);
999 #ifdef CONFIG_FAIL_MAKE_REQUEST
1000 static struct device_attribute dev_attr_fail
=
1001 __ATTR(make
-it
-fail
, S_IRUGO
|S_IWUSR
, part_fail_show
, part_fail_store
);
1003 #ifdef CONFIG_FAIL_IO_TIMEOUT
1004 static struct device_attribute dev_attr_fail_timeout
=
1005 __ATTR(io
-timeout
-fail
, S_IRUGO
|S_IWUSR
, part_timeout_show
,
1006 part_timeout_store
);
1009 static struct attribute
*disk_attrs
[] = {
1010 &dev_attr_range
.attr
,
1011 &dev_attr_ext_range
.attr
,
1012 &dev_attr_removable
.attr
,
1014 &dev_attr_size
.attr
,
1015 &dev_attr_alignment_offset
.attr
,
1016 &dev_attr_discard_alignment
.attr
,
1017 &dev_attr_capability
.attr
,
1018 &dev_attr_stat
.attr
,
1019 &dev_attr_inflight
.attr
,
1020 #ifdef CONFIG_FAIL_MAKE_REQUEST
1021 &dev_attr_fail
.attr
,
1023 #ifdef CONFIG_FAIL_IO_TIMEOUT
1024 &dev_attr_fail_timeout
.attr
,
1029 static struct attribute_group disk_attr_group
= {
1030 .attrs
= disk_attrs
,
1033 static const struct attribute_group
*disk_attr_groups
[] = {
1039 * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way
1040 * @disk: disk to replace part_tbl for
1041 * @new_ptbl: new part_tbl to install
1043 * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The
1044 * original ptbl is freed using RCU callback.
1047 * Matching bd_mutx locked.
1049 static void disk_replace_part_tbl(struct gendisk
*disk
,
1050 struct disk_part_tbl
*new_ptbl
)
1052 struct disk_part_tbl
*old_ptbl
= disk
->part_tbl
;
1054 rcu_assign_pointer(disk
->part_tbl
, new_ptbl
);
1057 rcu_assign_pointer(old_ptbl
->last_lookup
, NULL
);
1058 kfree_rcu(old_ptbl
, rcu_head
);
1063 * disk_expand_part_tbl - expand disk->part_tbl
1064 * @disk: disk to expand part_tbl for
1065 * @partno: expand such that this partno can fit in
1067 * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl
1068 * uses RCU to allow unlocked dereferencing for stats and other stuff.
1071 * Matching bd_mutex locked, might sleep.
1074 * 0 on success, -errno on failure.
1076 int disk_expand_part_tbl(struct gendisk
*disk
, int partno
)
1078 struct disk_part_tbl
*old_ptbl
= disk
->part_tbl
;
1079 struct disk_part_tbl
*new_ptbl
;
1080 int len
= old_ptbl
? old_ptbl
->len
: 0;
1081 int target
= partno
+ 1;
1085 /* disk_max_parts() is zero during initialization, ignore if so */
1086 if (disk_max_parts(disk
) && target
> disk_max_parts(disk
))
1092 size
= sizeof(*new_ptbl
) + target
* sizeof(new_ptbl
->part
[0]);
1093 new_ptbl
= kzalloc_node(size
, GFP_KERNEL
, disk
->node_id
);
1097 new_ptbl
->len
= target
;
1099 for (i
= 0; i
< len
; i
++)
1100 rcu_assign_pointer(new_ptbl
->part
[i
], old_ptbl
->part
[i
]);
1102 disk_replace_part_tbl(disk
, new_ptbl
);
1106 static void disk_release(struct device
*dev
)
1108 struct gendisk
*disk
= dev_to_disk(dev
);
1110 disk_release_events(disk
);
1111 kfree(disk
->random
);
1112 disk_replace_part_tbl(disk
, NULL
);
1113 free_part_stats(&disk
->part0
);
1114 free_part_info(&disk
->part0
);
1116 blk_put_queue(disk
->queue
);
1119 struct class block_class
= {
1123 static char *block_devnode(struct device
*dev
, umode_t
*mode
)
1125 struct gendisk
*disk
= dev_to_disk(dev
);
1128 return disk
->devnode(disk
, mode
);
1132 static struct device_type disk_type
= {
1134 .groups
= disk_attr_groups
,
1135 .release
= disk_release
,
1136 .devnode
= block_devnode
,
1139 #ifdef CONFIG_PROC_FS
1141 * aggregate disk stat collector. Uses the same stats that the sysfs
1142 * entries do, above, but makes them available through one seq_file.
1144 * The output looks suspiciously like /proc/partitions with a bunch of
1147 static int diskstats_show(struct seq_file
*seqf
, void *v
)
1149 struct gendisk
*gp
= v
;
1150 struct disk_part_iter piter
;
1151 struct hd_struct
*hd
;
1152 char buf
[BDEVNAME_SIZE
];
1156 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1157 seq_puts(seqf, "major minor name"
1158 " rio rmerge rsect ruse wio wmerge "
1159 "wsect wuse running use aveq"
1163 disk_part_iter_init(&piter
, gp
, DISK_PITER_INCL_EMPTY_PART0
);
1164 while ((hd
= disk_part_iter_next(&piter
))) {
1165 cpu
= part_stat_lock();
1166 part_round_stats(cpu
, hd
);
1168 seq_printf(seqf
, "%4d %7d %s %lu %lu %lu "
1169 "%u %lu %lu %lu %u %u %u %u\n",
1170 MAJOR(part_devt(hd
)), MINOR(part_devt(hd
)),
1171 disk_name(gp
, hd
->partno
, buf
),
1172 part_stat_read(hd
, ios
[READ
]),
1173 part_stat_read(hd
, merges
[READ
]),
1174 part_stat_read(hd
, sectors
[READ
]),
1175 jiffies_to_msecs(part_stat_read(hd
, ticks
[READ
])),
1176 part_stat_read(hd
, ios
[WRITE
]),
1177 part_stat_read(hd
, merges
[WRITE
]),
1178 part_stat_read(hd
, sectors
[WRITE
]),
1179 jiffies_to_msecs(part_stat_read(hd
, ticks
[WRITE
])),
1181 jiffies_to_msecs(part_stat_read(hd
, io_ticks
)),
1182 jiffies_to_msecs(part_stat_read(hd
, time_in_queue
))
1185 disk_part_iter_exit(&piter
);
1190 static const struct seq_operations diskstats_op
= {
1191 .start
= disk_seqf_start
,
1192 .next
= disk_seqf_next
,
1193 .stop
= disk_seqf_stop
,
1194 .show
= diskstats_show
1197 static int diskstats_open(struct inode
*inode
, struct file
*file
)
1199 return seq_open(file
, &diskstats_op
);
1202 static const struct file_operations proc_diskstats_operations
= {
1203 .open
= diskstats_open
,
1205 .llseek
= seq_lseek
,
1206 .release
= seq_release
,
1209 static int __init
proc_genhd_init(void)
1211 proc_create("diskstats", 0, NULL
, &proc_diskstats_operations
);
1212 proc_create("partitions", 0, NULL
, &proc_partitions_operations
);
1215 module_init(proc_genhd_init
);
1216 #endif /* CONFIG_PROC_FS */
1218 dev_t
blk_lookup_devt(const char *name
, int partno
)
1220 dev_t devt
= MKDEV(0, 0);
1221 struct class_dev_iter iter
;
1224 class_dev_iter_init(&iter
, &block_class
, NULL
, &disk_type
);
1225 while ((dev
= class_dev_iter_next(&iter
))) {
1226 struct gendisk
*disk
= dev_to_disk(dev
);
1227 struct hd_struct
*part
;
1229 if (strcmp(dev_name(dev
), name
))
1232 if (partno
< disk
->minors
) {
1233 /* We need to return the right devno, even
1234 * if the partition doesn't exist yet.
1236 devt
= MKDEV(MAJOR(dev
->devt
),
1237 MINOR(dev
->devt
) + partno
);
1240 part
= disk_get_part(disk
, partno
);
1242 devt
= part_devt(part
);
1243 disk_put_part(part
);
1246 disk_put_part(part
);
1248 class_dev_iter_exit(&iter
);
1251 EXPORT_SYMBOL(blk_lookup_devt
);
1253 struct gendisk
*alloc_disk(int minors
)
1255 return alloc_disk_node(minors
, NUMA_NO_NODE
);
1257 EXPORT_SYMBOL(alloc_disk
);
1259 struct gendisk
*alloc_disk_node(int minors
, int node_id
)
1261 struct gendisk
*disk
;
1263 disk
= kmalloc_node(sizeof(struct gendisk
),
1264 GFP_KERNEL
| __GFP_ZERO
, node_id
);
1266 if (!init_part_stats(&disk
->part0
)) {
1270 disk
->node_id
= node_id
;
1271 if (disk_expand_part_tbl(disk
, 0)) {
1272 free_part_stats(&disk
->part0
);
1276 disk
->part_tbl
->part
[0] = &disk
->part0
;
1279 * set_capacity() and get_capacity() currently don't use
1280 * seqcounter to read/update the part0->nr_sects. Still init
1281 * the counter as we can read the sectors in IO submission
1282 * patch using seqence counters.
1284 * TODO: Ideally set_capacity() and get_capacity() should be
1285 * converted to make use of bd_mutex and sequence counters.
1287 seqcount_init(&disk
->part0
.nr_sects_seq
);
1288 hd_ref_init(&disk
->part0
);
1290 disk
->minors
= minors
;
1291 rand_initialize_disk(disk
);
1292 disk_to_dev(disk
)->class = &block_class
;
1293 disk_to_dev(disk
)->type
= &disk_type
;
1294 device_initialize(disk_to_dev(disk
));
1298 EXPORT_SYMBOL(alloc_disk_node
);
1300 struct kobject
*get_disk(struct gendisk
*disk
)
1302 struct module
*owner
;
1303 struct kobject
*kobj
;
1307 owner
= disk
->fops
->owner
;
1308 if (owner
&& !try_module_get(owner
))
1310 kobj
= kobject_get(&disk_to_dev(disk
)->kobj
);
1319 EXPORT_SYMBOL(get_disk
);
1321 void put_disk(struct gendisk
*disk
)
1324 kobject_put(&disk_to_dev(disk
)->kobj
);
1327 EXPORT_SYMBOL(put_disk
);
1329 static void set_disk_ro_uevent(struct gendisk
*gd
, int ro
)
1331 char event
[] = "DISK_RO=1";
1332 char *envp
[] = { event
, NULL
};
1336 kobject_uevent_env(&disk_to_dev(gd
)->kobj
, KOBJ_CHANGE
, envp
);
1339 void set_device_ro(struct block_device
*bdev
, int flag
)
1341 bdev
->bd_part
->policy
= flag
;
1344 EXPORT_SYMBOL(set_device_ro
);
1346 void set_disk_ro(struct gendisk
*disk
, int flag
)
1348 struct disk_part_iter piter
;
1349 struct hd_struct
*part
;
1351 if (disk
->part0
.policy
!= flag
) {
1352 set_disk_ro_uevent(disk
, flag
);
1353 disk
->part0
.policy
= flag
;
1356 disk_part_iter_init(&piter
, disk
, DISK_PITER_INCL_EMPTY
);
1357 while ((part
= disk_part_iter_next(&piter
)))
1358 part
->policy
= flag
;
1359 disk_part_iter_exit(&piter
);
1362 EXPORT_SYMBOL(set_disk_ro
);
1364 int bdev_read_only(struct block_device
*bdev
)
1368 return bdev
->bd_part
->policy
;
1371 EXPORT_SYMBOL(bdev_read_only
);
1373 int invalidate_partition(struct gendisk
*disk
, int partno
)
1376 struct block_device
*bdev
= bdget_disk(disk
, partno
);
1379 res
= __invalidate_device(bdev
, true);
1385 EXPORT_SYMBOL(invalidate_partition
);
1388 * Disk events - monitor disk events like media change and eject request.
1390 struct disk_events
{
1391 struct list_head node
; /* all disk_event's */
1392 struct gendisk
*disk
; /* the associated disk */
1395 struct mutex block_mutex
; /* protects blocking */
1396 int block
; /* event blocking depth */
1397 unsigned int pending
; /* events already sent out */
1398 unsigned int clearing
; /* events being cleared */
1400 long poll_msecs
; /* interval, -1 for default */
1401 struct delayed_work dwork
;
1404 static const char *disk_events_strs
[] = {
1405 [ilog2(DISK_EVENT_MEDIA_CHANGE
)] = "media_change",
1406 [ilog2(DISK_EVENT_EJECT_REQUEST
)] = "eject_request",
1409 static char *disk_uevents
[] = {
1410 [ilog2(DISK_EVENT_MEDIA_CHANGE
)] = "DISK_MEDIA_CHANGE=1",
1411 [ilog2(DISK_EVENT_EJECT_REQUEST
)] = "DISK_EJECT_REQUEST=1",
1414 /* list of all disk_events */
1415 static DEFINE_MUTEX(disk_events_mutex
);
1416 static LIST_HEAD(disk_events
);
1418 /* disable in-kernel polling by default */
1419 static unsigned long disk_events_dfl_poll_msecs
= 0;
1421 static unsigned long disk_events_poll_jiffies(struct gendisk
*disk
)
1423 struct disk_events
*ev
= disk
->ev
;
1424 long intv_msecs
= 0;
1427 * If device-specific poll interval is set, always use it. If
1428 * the default is being used, poll iff there are events which
1429 * can't be monitored asynchronously.
1431 if (ev
->poll_msecs
>= 0)
1432 intv_msecs
= ev
->poll_msecs
;
1433 else if (disk
->events
& ~disk
->async_events
)
1434 intv_msecs
= disk_events_dfl_poll_msecs
;
1436 return msecs_to_jiffies(intv_msecs
);
1440 * disk_block_events - block and flush disk event checking
1441 * @disk: disk to block events for
1443 * On return from this function, it is guaranteed that event checking
1444 * isn't in progress and won't happen until unblocked by
1445 * disk_unblock_events(). Events blocking is counted and the actual
1446 * unblocking happens after the matching number of unblocks are done.
1448 * Note that this intentionally does not block event checking from
1449 * disk_clear_events().
1454 void disk_block_events(struct gendisk
*disk
)
1456 struct disk_events
*ev
= disk
->ev
;
1457 unsigned long flags
;
1464 * Outer mutex ensures that the first blocker completes canceling
1465 * the event work before further blockers are allowed to finish.
1467 mutex_lock(&ev
->block_mutex
);
1469 spin_lock_irqsave(&ev
->lock
, flags
);
1470 cancel
= !ev
->block
++;
1471 spin_unlock_irqrestore(&ev
->lock
, flags
);
1474 cancel_delayed_work_sync(&disk
->ev
->dwork
);
1476 mutex_unlock(&ev
->block_mutex
);
1479 static void __disk_unblock_events(struct gendisk
*disk
, bool check_now
)
1481 struct disk_events
*ev
= disk
->ev
;
1483 unsigned long flags
;
1485 spin_lock_irqsave(&ev
->lock
, flags
);
1487 if (WARN_ON_ONCE(ev
->block
<= 0))
1494 * Not exactly a latency critical operation, set poll timer
1495 * slack to 25% and kick event check.
1497 intv
= disk_events_poll_jiffies(disk
);
1498 set_timer_slack(&ev
->dwork
.timer
, intv
/ 4);
1500 queue_delayed_work(system_freezable_wq
, &ev
->dwork
, 0);
1502 queue_delayed_work(system_freezable_wq
, &ev
->dwork
, intv
);
1504 spin_unlock_irqrestore(&ev
->lock
, flags
);
1508 * disk_unblock_events - unblock disk event checking
1509 * @disk: disk to unblock events for
1511 * Undo disk_block_events(). When the block count reaches zero, it
1512 * starts events polling if configured.
1515 * Don't care. Safe to call from irq context.
1517 void disk_unblock_events(struct gendisk
*disk
)
1520 __disk_unblock_events(disk
, false);
1524 * disk_flush_events - schedule immediate event checking and flushing
1525 * @disk: disk to check and flush events for
1526 * @mask: events to flush
1528 * Schedule immediate event checking on @disk if not blocked. Events in
1529 * @mask are scheduled to be cleared from the driver. Note that this
1530 * doesn't clear the events from @disk->ev.
1533 * If @mask is non-zero must be called with bdev->bd_mutex held.
1535 void disk_flush_events(struct gendisk
*disk
, unsigned int mask
)
1537 struct disk_events
*ev
= disk
->ev
;
1542 spin_lock_irq(&ev
->lock
);
1543 ev
->clearing
|= mask
;
1545 mod_delayed_work(system_freezable_wq
, &ev
->dwork
, 0);
1546 spin_unlock_irq(&ev
->lock
);
1550 * disk_clear_events - synchronously check, clear and return pending events
1551 * @disk: disk to fetch and clear events from
1552 * @mask: mask of events to be fetched and clearted
1554 * Disk events are synchronously checked and pending events in @mask
1555 * are cleared and returned. This ignores the block count.
1560 unsigned int disk_clear_events(struct gendisk
*disk
, unsigned int mask
)
1562 const struct block_device_operations
*bdops
= disk
->fops
;
1563 struct disk_events
*ev
= disk
->ev
;
1564 unsigned int pending
;
1565 unsigned int clearing
= mask
;
1568 /* for drivers still using the old ->media_changed method */
1569 if ((mask
& DISK_EVENT_MEDIA_CHANGE
) &&
1570 bdops
->media_changed
&& bdops
->media_changed(disk
))
1571 return DISK_EVENT_MEDIA_CHANGE
;
1575 disk_block_events(disk
);
1578 * store the union of mask and ev->clearing on the stack so that the
1579 * race with disk_flush_events does not cause ambiguity (ev->clearing
1580 * can still be modified even if events are blocked).
1582 spin_lock_irq(&ev
->lock
);
1583 clearing
|= ev
->clearing
;
1585 spin_unlock_irq(&ev
->lock
);
1587 disk_check_events(ev
, &clearing
);
1589 * if ev->clearing is not 0, the disk_flush_events got called in the
1590 * middle of this function, so we want to run the workfn without delay.
1592 __disk_unblock_events(disk
, ev
->clearing
? true : false);
1594 /* then, fetch and clear pending events */
1595 spin_lock_irq(&ev
->lock
);
1596 pending
= ev
->pending
& mask
;
1597 ev
->pending
&= ~mask
;
1598 spin_unlock_irq(&ev
->lock
);
1599 WARN_ON_ONCE(clearing
& mask
);
1605 * Separate this part out so that a different pointer for clearing_ptr can be
1606 * passed in for disk_clear_events.
1608 static void disk_events_workfn(struct work_struct
*work
)
1610 struct delayed_work
*dwork
= to_delayed_work(work
);
1611 struct disk_events
*ev
= container_of(dwork
, struct disk_events
, dwork
);
1613 disk_check_events(ev
, &ev
->clearing
);
1616 static void disk_check_events(struct disk_events
*ev
,
1617 unsigned int *clearing_ptr
)
1619 struct gendisk
*disk
= ev
->disk
;
1620 char *envp
[ARRAY_SIZE(disk_uevents
) + 1] = { };
1621 unsigned int clearing
= *clearing_ptr
;
1622 unsigned int events
;
1624 int nr_events
= 0, i
;
1627 events
= disk
->fops
->check_events(disk
, clearing
);
1629 /* accumulate pending events and schedule next poll if necessary */
1630 spin_lock_irq(&ev
->lock
);
1632 events
&= ~ev
->pending
;
1633 ev
->pending
|= events
;
1634 *clearing_ptr
&= ~clearing
;
1636 intv
= disk_events_poll_jiffies(disk
);
1637 if (!ev
->block
&& intv
)
1638 queue_delayed_work(system_freezable_wq
, &ev
->dwork
, intv
);
1640 spin_unlock_irq(&ev
->lock
);
1643 * Tell userland about new events. Only the events listed in
1644 * @disk->events are reported. Unlisted events are processed the
1645 * same internally but never get reported to userland.
1647 for (i
= 0; i
< ARRAY_SIZE(disk_uevents
); i
++)
1648 if (events
& disk
->events
& (1 << i
))
1649 envp
[nr_events
++] = disk_uevents
[i
];
1652 kobject_uevent_env(&disk_to_dev(disk
)->kobj
, KOBJ_CHANGE
, envp
);
1656 * A disk events enabled device has the following sysfs nodes under
1657 * its /sys/block/X/ directory.
1659 * events : list of all supported events
1660 * events_async : list of events which can be detected w/o polling
1661 * events_poll_msecs : polling interval, 0: disable, -1: system default
1663 static ssize_t
__disk_events_show(unsigned int events
, char *buf
)
1665 const char *delim
= "";
1669 for (i
= 0; i
< ARRAY_SIZE(disk_events_strs
); i
++)
1670 if (events
& (1 << i
)) {
1671 pos
+= sprintf(buf
+ pos
, "%s%s",
1672 delim
, disk_events_strs
[i
]);
1676 pos
+= sprintf(buf
+ pos
, "\n");
1680 static ssize_t
disk_events_show(struct device
*dev
,
1681 struct device_attribute
*attr
, char *buf
)
1683 struct gendisk
*disk
= dev_to_disk(dev
);
1685 return __disk_events_show(disk
->events
, buf
);
1688 static ssize_t
disk_events_async_show(struct device
*dev
,
1689 struct device_attribute
*attr
, char *buf
)
1691 struct gendisk
*disk
= dev_to_disk(dev
);
1693 return __disk_events_show(disk
->async_events
, buf
);
1696 static ssize_t
disk_events_poll_msecs_show(struct device
*dev
,
1697 struct device_attribute
*attr
,
1700 struct gendisk
*disk
= dev_to_disk(dev
);
1702 return sprintf(buf
, "%ld\n", disk
->ev
->poll_msecs
);
1705 static ssize_t
disk_events_poll_msecs_store(struct device
*dev
,
1706 struct device_attribute
*attr
,
1707 const char *buf
, size_t count
)
1709 struct gendisk
*disk
= dev_to_disk(dev
);
1712 if (!count
|| !sscanf(buf
, "%ld", &intv
))
1715 if (intv
< 0 && intv
!= -1)
1718 disk_block_events(disk
);
1719 disk
->ev
->poll_msecs
= intv
;
1720 __disk_unblock_events(disk
, true);
1725 static const DEVICE_ATTR(events
, S_IRUGO
, disk_events_show
, NULL
);
1726 static const DEVICE_ATTR(events_async
, S_IRUGO
, disk_events_async_show
, NULL
);
1727 static const DEVICE_ATTR(events_poll_msecs
, S_IRUGO
|S_IWUSR
,
1728 disk_events_poll_msecs_show
,
1729 disk_events_poll_msecs_store
);
1731 static const struct attribute
*disk_events_attrs
[] = {
1732 &dev_attr_events
.attr
,
1733 &dev_attr_events_async
.attr
,
1734 &dev_attr_events_poll_msecs
.attr
,
1739 * The default polling interval can be specified by the kernel
1740 * parameter block.events_dfl_poll_msecs which defaults to 0
1741 * (disable). This can also be modified runtime by writing to
1742 * /sys/module/block/events_dfl_poll_msecs.
1744 static int disk_events_set_dfl_poll_msecs(const char *val
,
1745 const struct kernel_param
*kp
)
1747 struct disk_events
*ev
;
1750 ret
= param_set_ulong(val
, kp
);
1754 mutex_lock(&disk_events_mutex
);
1756 list_for_each_entry(ev
, &disk_events
, node
)
1757 disk_flush_events(ev
->disk
, 0);
1759 mutex_unlock(&disk_events_mutex
);
1764 static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops
= {
1765 .set
= disk_events_set_dfl_poll_msecs
,
1766 .get
= param_get_ulong
,
1769 #undef MODULE_PARAM_PREFIX
1770 #define MODULE_PARAM_PREFIX "block."
1772 module_param_cb(events_dfl_poll_msecs
, &disk_events_dfl_poll_msecs_param_ops
,
1773 &disk_events_dfl_poll_msecs
, 0644);
1776 * disk_{alloc|add|del|release}_events - initialize and destroy disk_events.
1778 static void disk_alloc_events(struct gendisk
*disk
)
1780 struct disk_events
*ev
;
1782 if (!disk
->fops
->check_events
)
1785 ev
= kzalloc(sizeof(*ev
), GFP_KERNEL
);
1787 pr_warn("%s: failed to initialize events\n", disk
->disk_name
);
1791 INIT_LIST_HEAD(&ev
->node
);
1793 spin_lock_init(&ev
->lock
);
1794 mutex_init(&ev
->block_mutex
);
1796 ev
->poll_msecs
= -1;
1797 INIT_DELAYED_WORK(&ev
->dwork
, disk_events_workfn
);
1802 static void disk_add_events(struct gendisk
*disk
)
1807 /* FIXME: error handling */
1808 if (sysfs_create_files(&disk_to_dev(disk
)->kobj
, disk_events_attrs
) < 0)
1809 pr_warn("%s: failed to create sysfs files for events\n",
1812 mutex_lock(&disk_events_mutex
);
1813 list_add_tail(&disk
->ev
->node
, &disk_events
);
1814 mutex_unlock(&disk_events_mutex
);
1817 * Block count is initialized to 1 and the following initial
1818 * unblock kicks it into action.
1820 __disk_unblock_events(disk
, true);
1823 static void disk_del_events(struct gendisk
*disk
)
1828 disk_block_events(disk
);
1830 mutex_lock(&disk_events_mutex
);
1831 list_del_init(&disk
->ev
->node
);
1832 mutex_unlock(&disk_events_mutex
);
1834 sysfs_remove_files(&disk_to_dev(disk
)->kobj
, disk_events_attrs
);
1837 static void disk_release_events(struct gendisk
*disk
)
1839 /* the block count should be 1 from disk_del_events() */
1840 WARN_ON_ONCE(disk
->ev
&& disk
->ev
->block
!= 1);