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_lock prevents look up
32 * results from going away underneath its user.
34 static DEFINE_SPINLOCK(ext_devt_lock
);
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 */
423 idr_preload(GFP_KERNEL
);
425 spin_lock(&ext_devt_lock
);
426 idx
= idr_alloc(&ext_devt_idr
, part
, 0, NR_EXT_DEVT
, GFP_NOWAIT
);
427 spin_unlock(&ext_devt_lock
);
431 return idx
== -ENOSPC
? -EBUSY
: idx
;
433 *devt
= MKDEV(BLOCK_EXT_MAJOR
, blk_mangle_minor(idx
));
438 * blk_free_devt - free a dev_t
439 * @devt: dev_t to free
441 * Free @devt which was allocated using blk_alloc_devt().
446 void blk_free_devt(dev_t devt
)
450 if (devt
== MKDEV(0, 0))
453 if (MAJOR(devt
) == BLOCK_EXT_MAJOR
) {
454 spin_lock(&ext_devt_lock
);
455 idr_remove(&ext_devt_idr
, blk_mangle_minor(MINOR(devt
)));
456 spin_unlock(&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 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
= disk
->driverfs_dev
;
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
);
576 * add_disk - add partitioning information to kernel list
577 * @disk: per-device partitioning information
579 * This function registers the partitioning information in @disk
582 * FIXME: error handling
584 void add_disk(struct gendisk
*disk
)
586 struct backing_dev_info
*bdi
;
590 /* minors == 0 indicates to use ext devt from part0 and should
591 * be accompanied with EXT_DEVT flag. Make sure all
592 * parameters make sense.
594 WARN_ON(disk
->minors
&& !(disk
->major
|| disk
->first_minor
));
595 WARN_ON(!disk
->minors
&& !(disk
->flags
& GENHD_FL_EXT_DEVT
));
597 disk
->flags
|= GENHD_FL_UP
;
599 retval
= blk_alloc_devt(&disk
->part0
, &devt
);
604 disk_to_dev(disk
)->devt
= devt
;
606 /* ->major and ->first_minor aren't supposed to be
607 * dereferenced from here on, but set them just in case.
609 disk
->major
= MAJOR(devt
);
610 disk
->first_minor
= MINOR(devt
);
612 disk_alloc_events(disk
);
614 /* Register BDI before referencing it from bdev */
615 bdi
= &disk
->queue
->backing_dev_info
;
616 bdi_register_dev(bdi
, disk_devt(disk
));
618 blk_register_region(disk_devt(disk
), disk
->minors
, NULL
,
619 exact_match
, exact_lock
, disk
);
621 blk_register_queue(disk
);
624 * Take an extra ref on queue which will be put on disk_release()
625 * so that it sticks around as long as @disk is there.
627 WARN_ON_ONCE(!blk_get_queue(disk
->queue
));
629 retval
= sysfs_create_link(&disk_to_dev(disk
)->kobj
, &bdi
->dev
->kobj
,
633 disk_add_events(disk
);
635 EXPORT_SYMBOL(add_disk
);
637 void del_gendisk(struct gendisk
*disk
)
639 struct disk_part_iter piter
;
640 struct hd_struct
*part
;
642 disk_del_events(disk
);
644 /* invalidate stuff */
645 disk_part_iter_init(&piter
, disk
,
646 DISK_PITER_INCL_EMPTY
| DISK_PITER_REVERSE
);
647 while ((part
= disk_part_iter_next(&piter
))) {
648 invalidate_partition(disk
, part
->partno
);
649 delete_partition(disk
, part
->partno
);
651 disk_part_iter_exit(&piter
);
653 invalidate_partition(disk
, 0);
654 set_capacity(disk
, 0);
655 disk
->flags
&= ~GENHD_FL_UP
;
657 sysfs_remove_link(&disk_to_dev(disk
)->kobj
, "bdi");
658 bdi_unregister(&disk
->queue
->backing_dev_info
);
659 blk_unregister_queue(disk
);
660 blk_unregister_region(disk_devt(disk
), disk
->minors
);
662 part_stat_set_all(&disk
->part0
, 0);
663 disk
->part0
.stamp
= 0;
665 kobject_put(disk
->part0
.holder_dir
);
666 kobject_put(disk
->slave_dir
);
667 disk
->driverfs_dev
= NULL
;
668 if (!sysfs_deprecated
)
669 sysfs_remove_link(block_depr
, dev_name(disk_to_dev(disk
)));
670 pm_runtime_set_memalloc_noio(disk_to_dev(disk
), false);
671 device_del(disk_to_dev(disk
));
673 EXPORT_SYMBOL(del_gendisk
);
676 * get_gendisk - get partitioning information for a given device
677 * @devt: device to get partitioning information for
678 * @partno: returned partition index
680 * This function gets the structure containing partitioning
681 * information for the given device @devt.
683 struct gendisk
*get_gendisk(dev_t devt
, int *partno
)
685 struct gendisk
*disk
= NULL
;
687 if (MAJOR(devt
) != BLOCK_EXT_MAJOR
) {
688 struct kobject
*kobj
;
690 kobj
= kobj_lookup(bdev_map
, devt
, partno
);
692 disk
= dev_to_disk(kobj_to_dev(kobj
));
694 struct hd_struct
*part
;
696 spin_lock(&ext_devt_lock
);
697 part
= idr_find(&ext_devt_idr
, blk_mangle_minor(MINOR(devt
)));
698 if (part
&& get_disk(part_to_disk(part
))) {
699 *partno
= part
->partno
;
700 disk
= part_to_disk(part
);
702 spin_unlock(&ext_devt_lock
);
707 EXPORT_SYMBOL(get_gendisk
);
710 * bdget_disk - do bdget() by gendisk and partition number
711 * @disk: gendisk of interest
712 * @partno: partition number
714 * Find partition @partno from @disk, do bdget() on it.
720 * Resulting block_device on success, NULL on failure.
722 struct block_device
*bdget_disk(struct gendisk
*disk
, int partno
)
724 struct hd_struct
*part
;
725 struct block_device
*bdev
= NULL
;
727 part
= disk_get_part(disk
, partno
);
729 bdev
= bdget(part_devt(part
));
734 EXPORT_SYMBOL(bdget_disk
);
737 * print a full list of all partitions - intended for places where the root
738 * filesystem can't be mounted and thus to give the victim some idea of what
741 void __init
printk_all_partitions(void)
743 struct class_dev_iter iter
;
746 class_dev_iter_init(&iter
, &block_class
, NULL
, &disk_type
);
747 while ((dev
= class_dev_iter_next(&iter
))) {
748 struct gendisk
*disk
= dev_to_disk(dev
);
749 struct disk_part_iter piter
;
750 struct hd_struct
*part
;
751 char name_buf
[BDEVNAME_SIZE
];
752 char devt_buf
[BDEVT_SIZE
];
755 * Don't show empty devices or things that have been
758 if (get_capacity(disk
) == 0 ||
759 (disk
->flags
& GENHD_FL_SUPPRESS_PARTITION_INFO
))
763 * Note, unlike /proc/partitions, I am showing the
764 * numbers in hex - the same format as the root=
767 disk_part_iter_init(&piter
, disk
, DISK_PITER_INCL_PART0
);
768 while ((part
= disk_part_iter_next(&piter
))) {
769 bool is_part0
= part
== &disk
->part0
;
771 printk("%s%s %10llu %s %s", is_part0
? "" : " ",
772 bdevt_str(part_devt(part
), devt_buf
),
773 (unsigned long long)part_nr_sects_read(part
) >> 1
774 , disk_name(disk
, part
->partno
, name_buf
),
775 part
->info
? part
->info
->uuid
: "");
777 if (disk
->driverfs_dev
!= NULL
&&
778 disk
->driverfs_dev
->driver
!= NULL
)
779 printk(" driver: %s\n",
780 disk
->driverfs_dev
->driver
->name
);
782 printk(" (driver?)\n");
786 disk_part_iter_exit(&piter
);
788 class_dev_iter_exit(&iter
);
791 #ifdef CONFIG_PROC_FS
793 static void *disk_seqf_start(struct seq_file
*seqf
, loff_t
*pos
)
796 struct class_dev_iter
*iter
;
799 iter
= kmalloc(sizeof(*iter
), GFP_KERNEL
);
801 return ERR_PTR(-ENOMEM
);
803 seqf
->private = iter
;
804 class_dev_iter_init(iter
, &block_class
, NULL
, &disk_type
);
806 dev
= class_dev_iter_next(iter
);
811 return dev_to_disk(dev
);
814 static void *disk_seqf_next(struct seq_file
*seqf
, void *v
, loff_t
*pos
)
819 dev
= class_dev_iter_next(seqf
->private);
821 return dev_to_disk(dev
);
826 static void disk_seqf_stop(struct seq_file
*seqf
, void *v
)
828 struct class_dev_iter
*iter
= seqf
->private;
830 /* stop is called even after start failed :-( */
832 class_dev_iter_exit(iter
);
837 static void *show_partition_start(struct seq_file
*seqf
, loff_t
*pos
)
841 p
= disk_seqf_start(seqf
, pos
);
842 if (!IS_ERR_OR_NULL(p
) && !*pos
)
843 seq_puts(seqf
, "major minor #blocks name\n\n");
847 static int show_partition(struct seq_file
*seqf
, void *v
)
849 struct gendisk
*sgp
= v
;
850 struct disk_part_iter piter
;
851 struct hd_struct
*part
;
852 char buf
[BDEVNAME_SIZE
];
854 /* Don't show non-partitionable removeable devices or empty devices */
855 if (!get_capacity(sgp
) || (!disk_max_parts(sgp
) &&
856 (sgp
->flags
& GENHD_FL_REMOVABLE
)))
858 if (sgp
->flags
& GENHD_FL_SUPPRESS_PARTITION_INFO
)
861 /* show the full disk and all non-0 size partitions of it */
862 disk_part_iter_init(&piter
, sgp
, DISK_PITER_INCL_PART0
);
863 while ((part
= disk_part_iter_next(&piter
)))
864 seq_printf(seqf
, "%4d %7d %10llu %s\n",
865 MAJOR(part_devt(part
)), MINOR(part_devt(part
)),
866 (unsigned long long)part_nr_sects_read(part
) >> 1,
867 disk_name(sgp
, part
->partno
, buf
));
868 disk_part_iter_exit(&piter
);
873 static const struct seq_operations partitions_op
= {
874 .start
= show_partition_start
,
875 .next
= disk_seqf_next
,
876 .stop
= disk_seqf_stop
,
877 .show
= show_partition
880 static int partitions_open(struct inode
*inode
, struct file
*file
)
882 return seq_open(file
, &partitions_op
);
885 static const struct file_operations proc_partitions_operations
= {
886 .open
= partitions_open
,
889 .release
= seq_release
,
894 static struct kobject
*base_probe(dev_t devt
, int *partno
, void *data
)
896 if (request_module("block-major-%d-%d", MAJOR(devt
), MINOR(devt
)) > 0)
897 /* Make old-style 2.4 aliases work */
898 request_module("block-major-%d", MAJOR(devt
));
902 static int __init
genhd_device_init(void)
906 block_class
.dev_kobj
= sysfs_dev_block_kobj
;
907 error
= class_register(&block_class
);
910 bdev_map
= kobj_map_init(base_probe
, &block_class_lock
);
913 register_blkdev(BLOCK_EXT_MAJOR
, "blkext");
915 /* create top-level block dir */
916 if (!sysfs_deprecated
)
917 block_depr
= kobject_create_and_add("block", NULL
);
921 subsys_initcall(genhd_device_init
);
923 static ssize_t
disk_range_show(struct device
*dev
,
924 struct device_attribute
*attr
, char *buf
)
926 struct gendisk
*disk
= dev_to_disk(dev
);
928 return sprintf(buf
, "%d\n", disk
->minors
);
931 static ssize_t
disk_ext_range_show(struct device
*dev
,
932 struct device_attribute
*attr
, char *buf
)
934 struct gendisk
*disk
= dev_to_disk(dev
);
936 return sprintf(buf
, "%d\n", disk_max_parts(disk
));
939 static ssize_t
disk_removable_show(struct device
*dev
,
940 struct device_attribute
*attr
, char *buf
)
942 struct gendisk
*disk
= dev_to_disk(dev
);
944 return sprintf(buf
, "%d\n",
945 (disk
->flags
& GENHD_FL_REMOVABLE
? 1 : 0));
948 static ssize_t
disk_ro_show(struct device
*dev
,
949 struct device_attribute
*attr
, char *buf
)
951 struct gendisk
*disk
= dev_to_disk(dev
);
953 return sprintf(buf
, "%d\n", get_disk_ro(disk
) ? 1 : 0);
956 static ssize_t
disk_capability_show(struct device
*dev
,
957 struct device_attribute
*attr
, char *buf
)
959 struct gendisk
*disk
= dev_to_disk(dev
);
961 return sprintf(buf
, "%x\n", disk
->flags
);
964 static ssize_t
disk_alignment_offset_show(struct device
*dev
,
965 struct device_attribute
*attr
,
968 struct gendisk
*disk
= dev_to_disk(dev
);
970 return sprintf(buf
, "%d\n", queue_alignment_offset(disk
->queue
));
973 static ssize_t
disk_discard_alignment_show(struct device
*dev
,
974 struct device_attribute
*attr
,
977 struct gendisk
*disk
= dev_to_disk(dev
);
979 return sprintf(buf
, "%d\n", queue_discard_alignment(disk
->queue
));
982 static DEVICE_ATTR(range
, S_IRUGO
, disk_range_show
, NULL
);
983 static DEVICE_ATTR(ext_range
, S_IRUGO
, disk_ext_range_show
, NULL
);
984 static DEVICE_ATTR(removable
, S_IRUGO
, disk_removable_show
, NULL
);
985 static DEVICE_ATTR(ro
, S_IRUGO
, disk_ro_show
, NULL
);
986 static DEVICE_ATTR(size
, S_IRUGO
, part_size_show
, NULL
);
987 static DEVICE_ATTR(alignment_offset
, S_IRUGO
, disk_alignment_offset_show
, NULL
);
988 static DEVICE_ATTR(discard_alignment
, S_IRUGO
, disk_discard_alignment_show
,
990 static DEVICE_ATTR(capability
, S_IRUGO
, disk_capability_show
, NULL
);
991 static DEVICE_ATTR(stat
, S_IRUGO
, part_stat_show
, NULL
);
992 static DEVICE_ATTR(inflight
, S_IRUGO
, part_inflight_show
, NULL
);
993 #ifdef CONFIG_FAIL_MAKE_REQUEST
994 static struct device_attribute dev_attr_fail
=
995 __ATTR(make
-it
-fail
, S_IRUGO
|S_IWUSR
, part_fail_show
, part_fail_store
);
997 #ifdef CONFIG_FAIL_IO_TIMEOUT
998 static struct device_attribute dev_attr_fail_timeout
=
999 __ATTR(io
-timeout
-fail
, S_IRUGO
|S_IWUSR
, part_timeout_show
,
1000 part_timeout_store
);
1003 static struct attribute
*disk_attrs
[] = {
1004 &dev_attr_range
.attr
,
1005 &dev_attr_ext_range
.attr
,
1006 &dev_attr_removable
.attr
,
1008 &dev_attr_size
.attr
,
1009 &dev_attr_alignment_offset
.attr
,
1010 &dev_attr_discard_alignment
.attr
,
1011 &dev_attr_capability
.attr
,
1012 &dev_attr_stat
.attr
,
1013 &dev_attr_inflight
.attr
,
1014 #ifdef CONFIG_FAIL_MAKE_REQUEST
1015 &dev_attr_fail
.attr
,
1017 #ifdef CONFIG_FAIL_IO_TIMEOUT
1018 &dev_attr_fail_timeout
.attr
,
1023 static struct attribute_group disk_attr_group
= {
1024 .attrs
= disk_attrs
,
1027 static const struct attribute_group
*disk_attr_groups
[] = {
1033 * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way
1034 * @disk: disk to replace part_tbl for
1035 * @new_ptbl: new part_tbl to install
1037 * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The
1038 * original ptbl is freed using RCU callback.
1041 * Matching bd_mutx locked.
1043 static void disk_replace_part_tbl(struct gendisk
*disk
,
1044 struct disk_part_tbl
*new_ptbl
)
1046 struct disk_part_tbl
*old_ptbl
= disk
->part_tbl
;
1048 rcu_assign_pointer(disk
->part_tbl
, new_ptbl
);
1051 rcu_assign_pointer(old_ptbl
->last_lookup
, NULL
);
1052 kfree_rcu(old_ptbl
, rcu_head
);
1057 * disk_expand_part_tbl - expand disk->part_tbl
1058 * @disk: disk to expand part_tbl for
1059 * @partno: expand such that this partno can fit in
1061 * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl
1062 * uses RCU to allow unlocked dereferencing for stats and other stuff.
1065 * Matching bd_mutex locked, might sleep.
1068 * 0 on success, -errno on failure.
1070 int disk_expand_part_tbl(struct gendisk
*disk
, int partno
)
1072 struct disk_part_tbl
*old_ptbl
= disk
->part_tbl
;
1073 struct disk_part_tbl
*new_ptbl
;
1074 int len
= old_ptbl
? old_ptbl
->len
: 0;
1075 int target
= partno
+ 1;
1079 /* disk_max_parts() is zero during initialization, ignore if so */
1080 if (disk_max_parts(disk
) && target
> disk_max_parts(disk
))
1086 size
= sizeof(*new_ptbl
) + target
* sizeof(new_ptbl
->part
[0]);
1087 new_ptbl
= kzalloc_node(size
, GFP_KERNEL
, disk
->node_id
);
1091 new_ptbl
->len
= target
;
1093 for (i
= 0; i
< len
; i
++)
1094 rcu_assign_pointer(new_ptbl
->part
[i
], old_ptbl
->part
[i
]);
1096 disk_replace_part_tbl(disk
, new_ptbl
);
1100 static void disk_release(struct device
*dev
)
1102 struct gendisk
*disk
= dev_to_disk(dev
);
1104 blk_free_devt(dev
->devt
);
1105 disk_release_events(disk
);
1106 kfree(disk
->random
);
1107 disk_replace_part_tbl(disk
, NULL
);
1108 free_part_stats(&disk
->part0
);
1109 free_part_info(&disk
->part0
);
1111 blk_put_queue(disk
->queue
);
1114 struct class block_class
= {
1118 static char *block_devnode(struct device
*dev
, umode_t
*mode
,
1119 kuid_t
*uid
, kgid_t
*gid
)
1121 struct gendisk
*disk
= dev_to_disk(dev
);
1124 return disk
->devnode(disk
, mode
);
1128 static struct device_type disk_type
= {
1130 .groups
= disk_attr_groups
,
1131 .release
= disk_release
,
1132 .devnode
= block_devnode
,
1135 #ifdef CONFIG_PROC_FS
1137 * aggregate disk stat collector. Uses the same stats that the sysfs
1138 * entries do, above, but makes them available through one seq_file.
1140 * The output looks suspiciously like /proc/partitions with a bunch of
1143 static int diskstats_show(struct seq_file
*seqf
, void *v
)
1145 struct gendisk
*gp
= v
;
1146 struct disk_part_iter piter
;
1147 struct hd_struct
*hd
;
1148 char buf
[BDEVNAME_SIZE
];
1152 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1153 seq_puts(seqf, "major minor name"
1154 " rio rmerge rsect ruse wio wmerge "
1155 "wsect wuse running use aveq"
1159 disk_part_iter_init(&piter
, gp
, DISK_PITER_INCL_EMPTY_PART0
);
1160 while ((hd
= disk_part_iter_next(&piter
))) {
1161 cpu
= part_stat_lock();
1162 part_round_stats(cpu
, hd
);
1164 seq_printf(seqf
, "%4d %7d %s %lu %lu %lu "
1165 "%u %lu %lu %lu %u %u %u %u\n",
1166 MAJOR(part_devt(hd
)), MINOR(part_devt(hd
)),
1167 disk_name(gp
, hd
->partno
, buf
),
1168 part_stat_read(hd
, ios
[READ
]),
1169 part_stat_read(hd
, merges
[READ
]),
1170 part_stat_read(hd
, sectors
[READ
]),
1171 jiffies_to_msecs(part_stat_read(hd
, ticks
[READ
])),
1172 part_stat_read(hd
, ios
[WRITE
]),
1173 part_stat_read(hd
, merges
[WRITE
]),
1174 part_stat_read(hd
, sectors
[WRITE
]),
1175 jiffies_to_msecs(part_stat_read(hd
, ticks
[WRITE
])),
1177 jiffies_to_msecs(part_stat_read(hd
, io_ticks
)),
1178 jiffies_to_msecs(part_stat_read(hd
, time_in_queue
))
1181 disk_part_iter_exit(&piter
);
1186 static const struct seq_operations diskstats_op
= {
1187 .start
= disk_seqf_start
,
1188 .next
= disk_seqf_next
,
1189 .stop
= disk_seqf_stop
,
1190 .show
= diskstats_show
1193 static int diskstats_open(struct inode
*inode
, struct file
*file
)
1195 return seq_open(file
, &diskstats_op
);
1198 static const struct file_operations proc_diskstats_operations
= {
1199 .open
= diskstats_open
,
1201 .llseek
= seq_lseek
,
1202 .release
= seq_release
,
1205 static int __init
proc_genhd_init(void)
1207 proc_create("diskstats", 0, NULL
, &proc_diskstats_operations
);
1208 proc_create("partitions", 0, NULL
, &proc_partitions_operations
);
1211 module_init(proc_genhd_init
);
1212 #endif /* CONFIG_PROC_FS */
1214 dev_t
blk_lookup_devt(const char *name
, int partno
)
1216 dev_t devt
= MKDEV(0, 0);
1217 struct class_dev_iter iter
;
1220 class_dev_iter_init(&iter
, &block_class
, NULL
, &disk_type
);
1221 while ((dev
= class_dev_iter_next(&iter
))) {
1222 struct gendisk
*disk
= dev_to_disk(dev
);
1223 struct hd_struct
*part
;
1225 if (strcmp(dev_name(dev
), name
))
1228 if (partno
< disk
->minors
) {
1229 /* We need to return the right devno, even
1230 * if the partition doesn't exist yet.
1232 devt
= MKDEV(MAJOR(dev
->devt
),
1233 MINOR(dev
->devt
) + partno
);
1236 part
= disk_get_part(disk
, partno
);
1238 devt
= part_devt(part
);
1239 disk_put_part(part
);
1242 disk_put_part(part
);
1244 class_dev_iter_exit(&iter
);
1247 EXPORT_SYMBOL(blk_lookup_devt
);
1249 struct gendisk
*alloc_disk(int minors
)
1251 return alloc_disk_node(minors
, NUMA_NO_NODE
);
1253 EXPORT_SYMBOL(alloc_disk
);
1255 struct gendisk
*alloc_disk_node(int minors
, int node_id
)
1257 struct gendisk
*disk
;
1259 disk
= kzalloc_node(sizeof(struct gendisk
), GFP_KERNEL
, node_id
);
1261 if (!init_part_stats(&disk
->part0
)) {
1265 disk
->node_id
= node_id
;
1266 if (disk_expand_part_tbl(disk
, 0)) {
1267 free_part_stats(&disk
->part0
);
1271 disk
->part_tbl
->part
[0] = &disk
->part0
;
1274 * set_capacity() and get_capacity() currently don't use
1275 * seqcounter to read/update the part0->nr_sects. Still init
1276 * the counter as we can read the sectors in IO submission
1277 * patch using seqence counters.
1279 * TODO: Ideally set_capacity() and get_capacity() should be
1280 * converted to make use of bd_mutex and sequence counters.
1282 seqcount_init(&disk
->part0
.nr_sects_seq
);
1283 hd_ref_init(&disk
->part0
);
1285 disk
->minors
= minors
;
1286 rand_initialize_disk(disk
);
1287 disk_to_dev(disk
)->class = &block_class
;
1288 disk_to_dev(disk
)->type
= &disk_type
;
1289 device_initialize(disk_to_dev(disk
));
1293 EXPORT_SYMBOL(alloc_disk_node
);
1295 struct kobject
*get_disk(struct gendisk
*disk
)
1297 struct module
*owner
;
1298 struct kobject
*kobj
;
1302 owner
= disk
->fops
->owner
;
1303 if (owner
&& !try_module_get(owner
))
1305 kobj
= kobject_get(&disk_to_dev(disk
)->kobj
);
1314 EXPORT_SYMBOL(get_disk
);
1316 void put_disk(struct gendisk
*disk
)
1319 kobject_put(&disk_to_dev(disk
)->kobj
);
1322 EXPORT_SYMBOL(put_disk
);
1324 static void set_disk_ro_uevent(struct gendisk
*gd
, int ro
)
1326 char event
[] = "DISK_RO=1";
1327 char *envp
[] = { event
, NULL
};
1331 kobject_uevent_env(&disk_to_dev(gd
)->kobj
, KOBJ_CHANGE
, envp
);
1334 void set_device_ro(struct block_device
*bdev
, int flag
)
1336 bdev
->bd_part
->policy
= flag
;
1339 EXPORT_SYMBOL(set_device_ro
);
1341 void set_disk_ro(struct gendisk
*disk
, int flag
)
1343 struct disk_part_iter piter
;
1344 struct hd_struct
*part
;
1346 if (disk
->part0
.policy
!= flag
) {
1347 set_disk_ro_uevent(disk
, flag
);
1348 disk
->part0
.policy
= flag
;
1351 disk_part_iter_init(&piter
, disk
, DISK_PITER_INCL_EMPTY
);
1352 while ((part
= disk_part_iter_next(&piter
)))
1353 part
->policy
= flag
;
1354 disk_part_iter_exit(&piter
);
1357 EXPORT_SYMBOL(set_disk_ro
);
1359 int bdev_read_only(struct block_device
*bdev
)
1363 return bdev
->bd_part
->policy
;
1366 EXPORT_SYMBOL(bdev_read_only
);
1368 int invalidate_partition(struct gendisk
*disk
, int partno
)
1371 struct block_device
*bdev
= bdget_disk(disk
, partno
);
1374 res
= __invalidate_device(bdev
, true);
1380 EXPORT_SYMBOL(invalidate_partition
);
1383 * Disk events - monitor disk events like media change and eject request.
1385 struct disk_events
{
1386 struct list_head node
; /* all disk_event's */
1387 struct gendisk
*disk
; /* the associated disk */
1390 struct mutex block_mutex
; /* protects blocking */
1391 int block
; /* event blocking depth */
1392 unsigned int pending
; /* events already sent out */
1393 unsigned int clearing
; /* events being cleared */
1395 long poll_msecs
; /* interval, -1 for default */
1396 struct delayed_work dwork
;
1399 static const char *disk_events_strs
[] = {
1400 [ilog2(DISK_EVENT_MEDIA_CHANGE
)] = "media_change",
1401 [ilog2(DISK_EVENT_EJECT_REQUEST
)] = "eject_request",
1404 static char *disk_uevents
[] = {
1405 [ilog2(DISK_EVENT_MEDIA_CHANGE
)] = "DISK_MEDIA_CHANGE=1",
1406 [ilog2(DISK_EVENT_EJECT_REQUEST
)] = "DISK_EJECT_REQUEST=1",
1409 /* list of all disk_events */
1410 static DEFINE_MUTEX(disk_events_mutex
);
1411 static LIST_HEAD(disk_events
);
1413 /* disable in-kernel polling by default */
1414 static unsigned long disk_events_dfl_poll_msecs
= 0;
1416 static unsigned long disk_events_poll_jiffies(struct gendisk
*disk
)
1418 struct disk_events
*ev
= disk
->ev
;
1419 long intv_msecs
= 0;
1422 * If device-specific poll interval is set, always use it. If
1423 * the default is being used, poll iff there are events which
1424 * can't be monitored asynchronously.
1426 if (ev
->poll_msecs
>= 0)
1427 intv_msecs
= ev
->poll_msecs
;
1428 else if (disk
->events
& ~disk
->async_events
)
1429 intv_msecs
= disk_events_dfl_poll_msecs
;
1431 return msecs_to_jiffies(intv_msecs
);
1435 * disk_block_events - block and flush disk event checking
1436 * @disk: disk to block events for
1438 * On return from this function, it is guaranteed that event checking
1439 * isn't in progress and won't happen until unblocked by
1440 * disk_unblock_events(). Events blocking is counted and the actual
1441 * unblocking happens after the matching number of unblocks are done.
1443 * Note that this intentionally does not block event checking from
1444 * disk_clear_events().
1449 void disk_block_events(struct gendisk
*disk
)
1451 struct disk_events
*ev
= disk
->ev
;
1452 unsigned long flags
;
1459 * Outer mutex ensures that the first blocker completes canceling
1460 * the event work before further blockers are allowed to finish.
1462 mutex_lock(&ev
->block_mutex
);
1464 spin_lock_irqsave(&ev
->lock
, flags
);
1465 cancel
= !ev
->block
++;
1466 spin_unlock_irqrestore(&ev
->lock
, flags
);
1469 cancel_delayed_work_sync(&disk
->ev
->dwork
);
1471 mutex_unlock(&ev
->block_mutex
);
1474 static void __disk_unblock_events(struct gendisk
*disk
, bool check_now
)
1476 struct disk_events
*ev
= disk
->ev
;
1478 unsigned long flags
;
1480 spin_lock_irqsave(&ev
->lock
, flags
);
1482 if (WARN_ON_ONCE(ev
->block
<= 0))
1489 * Not exactly a latency critical operation, set poll timer
1490 * slack to 25% and kick event check.
1492 intv
= disk_events_poll_jiffies(disk
);
1493 set_timer_slack(&ev
->dwork
.timer
, intv
/ 4);
1495 queue_delayed_work(system_freezable_power_efficient_wq
,
1498 queue_delayed_work(system_freezable_power_efficient_wq
,
1501 spin_unlock_irqrestore(&ev
->lock
, flags
);
1505 * disk_unblock_events - unblock disk event checking
1506 * @disk: disk to unblock events for
1508 * Undo disk_block_events(). When the block count reaches zero, it
1509 * starts events polling if configured.
1512 * Don't care. Safe to call from irq context.
1514 void disk_unblock_events(struct gendisk
*disk
)
1517 __disk_unblock_events(disk
, false);
1521 * disk_flush_events - schedule immediate event checking and flushing
1522 * @disk: disk to check and flush events for
1523 * @mask: events to flush
1525 * Schedule immediate event checking on @disk if not blocked. Events in
1526 * @mask are scheduled to be cleared from the driver. Note that this
1527 * doesn't clear the events from @disk->ev.
1530 * If @mask is non-zero must be called with bdev->bd_mutex held.
1532 void disk_flush_events(struct gendisk
*disk
, unsigned int mask
)
1534 struct disk_events
*ev
= disk
->ev
;
1539 spin_lock_irq(&ev
->lock
);
1540 ev
->clearing
|= mask
;
1542 mod_delayed_work(system_freezable_power_efficient_wq
,
1544 spin_unlock_irq(&ev
->lock
);
1548 * disk_clear_events - synchronously check, clear and return pending events
1549 * @disk: disk to fetch and clear events from
1550 * @mask: mask of events to be fetched and clearted
1552 * Disk events are synchronously checked and pending events in @mask
1553 * are cleared and returned. This ignores the block count.
1558 unsigned int disk_clear_events(struct gendisk
*disk
, unsigned int mask
)
1560 const struct block_device_operations
*bdops
= disk
->fops
;
1561 struct disk_events
*ev
= disk
->ev
;
1562 unsigned int pending
;
1563 unsigned int clearing
= mask
;
1566 /* for drivers still using the old ->media_changed method */
1567 if ((mask
& DISK_EVENT_MEDIA_CHANGE
) &&
1568 bdops
->media_changed
&& bdops
->media_changed(disk
))
1569 return DISK_EVENT_MEDIA_CHANGE
;
1573 disk_block_events(disk
);
1576 * store the union of mask and ev->clearing on the stack so that the
1577 * race with disk_flush_events does not cause ambiguity (ev->clearing
1578 * can still be modified even if events are blocked).
1580 spin_lock_irq(&ev
->lock
);
1581 clearing
|= ev
->clearing
;
1583 spin_unlock_irq(&ev
->lock
);
1585 disk_check_events(ev
, &clearing
);
1587 * if ev->clearing is not 0, the disk_flush_events got called in the
1588 * middle of this function, so we want to run the workfn without delay.
1590 __disk_unblock_events(disk
, ev
->clearing
? true : false);
1592 /* then, fetch and clear pending events */
1593 spin_lock_irq(&ev
->lock
);
1594 pending
= ev
->pending
& mask
;
1595 ev
->pending
&= ~mask
;
1596 spin_unlock_irq(&ev
->lock
);
1597 WARN_ON_ONCE(clearing
& mask
);
1603 * Separate this part out so that a different pointer for clearing_ptr can be
1604 * passed in for disk_clear_events.
1606 static void disk_events_workfn(struct work_struct
*work
)
1608 struct delayed_work
*dwork
= to_delayed_work(work
);
1609 struct disk_events
*ev
= container_of(dwork
, struct disk_events
, dwork
);
1611 disk_check_events(ev
, &ev
->clearing
);
1614 static void disk_check_events(struct disk_events
*ev
,
1615 unsigned int *clearing_ptr
)
1617 struct gendisk
*disk
= ev
->disk
;
1618 char *envp
[ARRAY_SIZE(disk_uevents
) + 1] = { };
1619 unsigned int clearing
= *clearing_ptr
;
1620 unsigned int events
;
1622 int nr_events
= 0, i
;
1625 events
= disk
->fops
->check_events(disk
, clearing
);
1627 /* accumulate pending events and schedule next poll if necessary */
1628 spin_lock_irq(&ev
->lock
);
1630 events
&= ~ev
->pending
;
1631 ev
->pending
|= events
;
1632 *clearing_ptr
&= ~clearing
;
1634 intv
= disk_events_poll_jiffies(disk
);
1635 if (!ev
->block
&& intv
)
1636 queue_delayed_work(system_freezable_power_efficient_wq
,
1639 spin_unlock_irq(&ev
->lock
);
1642 * Tell userland about new events. Only the events listed in
1643 * @disk->events are reported. Unlisted events are processed the
1644 * same internally but never get reported to userland.
1646 for (i
= 0; i
< ARRAY_SIZE(disk_uevents
); i
++)
1647 if (events
& disk
->events
& (1 << i
))
1648 envp
[nr_events
++] = disk_uevents
[i
];
1651 kobject_uevent_env(&disk_to_dev(disk
)->kobj
, KOBJ_CHANGE
, envp
);
1655 * A disk events enabled device has the following sysfs nodes under
1656 * its /sys/block/X/ directory.
1658 * events : list of all supported events
1659 * events_async : list of events which can be detected w/o polling
1660 * events_poll_msecs : polling interval, 0: disable, -1: system default
1662 static ssize_t
__disk_events_show(unsigned int events
, char *buf
)
1664 const char *delim
= "";
1668 for (i
= 0; i
< ARRAY_SIZE(disk_events_strs
); i
++)
1669 if (events
& (1 << i
)) {
1670 pos
+= sprintf(buf
+ pos
, "%s%s",
1671 delim
, disk_events_strs
[i
]);
1675 pos
+= sprintf(buf
+ pos
, "\n");
1679 static ssize_t
disk_events_show(struct device
*dev
,
1680 struct device_attribute
*attr
, char *buf
)
1682 struct gendisk
*disk
= dev_to_disk(dev
);
1684 return __disk_events_show(disk
->events
, buf
);
1687 static ssize_t
disk_events_async_show(struct device
*dev
,
1688 struct device_attribute
*attr
, char *buf
)
1690 struct gendisk
*disk
= dev_to_disk(dev
);
1692 return __disk_events_show(disk
->async_events
, buf
);
1695 static ssize_t
disk_events_poll_msecs_show(struct device
*dev
,
1696 struct device_attribute
*attr
,
1699 struct gendisk
*disk
= dev_to_disk(dev
);
1701 return sprintf(buf
, "%ld\n", disk
->ev
->poll_msecs
);
1704 static ssize_t
disk_events_poll_msecs_store(struct device
*dev
,
1705 struct device_attribute
*attr
,
1706 const char *buf
, size_t count
)
1708 struct gendisk
*disk
= dev_to_disk(dev
);
1711 if (!count
|| !sscanf(buf
, "%ld", &intv
))
1714 if (intv
< 0 && intv
!= -1)
1717 disk_block_events(disk
);
1718 disk
->ev
->poll_msecs
= intv
;
1719 __disk_unblock_events(disk
, true);
1724 static const DEVICE_ATTR(events
, S_IRUGO
, disk_events_show
, NULL
);
1725 static const DEVICE_ATTR(events_async
, S_IRUGO
, disk_events_async_show
, NULL
);
1726 static const DEVICE_ATTR(events_poll_msecs
, S_IRUGO
|S_IWUSR
,
1727 disk_events_poll_msecs_show
,
1728 disk_events_poll_msecs_store
);
1730 static const struct attribute
*disk_events_attrs
[] = {
1731 &dev_attr_events
.attr
,
1732 &dev_attr_events_async
.attr
,
1733 &dev_attr_events_poll_msecs
.attr
,
1738 * The default polling interval can be specified by the kernel
1739 * parameter block.events_dfl_poll_msecs which defaults to 0
1740 * (disable). This can also be modified runtime by writing to
1741 * /sys/module/block/events_dfl_poll_msecs.
1743 static int disk_events_set_dfl_poll_msecs(const char *val
,
1744 const struct kernel_param
*kp
)
1746 struct disk_events
*ev
;
1749 ret
= param_set_ulong(val
, kp
);
1753 mutex_lock(&disk_events_mutex
);
1755 list_for_each_entry(ev
, &disk_events
, node
)
1756 disk_flush_events(ev
->disk
, 0);
1758 mutex_unlock(&disk_events_mutex
);
1763 static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops
= {
1764 .set
= disk_events_set_dfl_poll_msecs
,
1765 .get
= param_get_ulong
,
1768 #undef MODULE_PARAM_PREFIX
1769 #define MODULE_PARAM_PREFIX "block."
1771 module_param_cb(events_dfl_poll_msecs
, &disk_events_dfl_poll_msecs_param_ops
,
1772 &disk_events_dfl_poll_msecs
, 0644);
1775 * disk_{alloc|add|del|release}_events - initialize and destroy disk_events.
1777 static void disk_alloc_events(struct gendisk
*disk
)
1779 struct disk_events
*ev
;
1781 if (!disk
->fops
->check_events
)
1784 ev
= kzalloc(sizeof(*ev
), GFP_KERNEL
);
1786 pr_warn("%s: failed to initialize events\n", disk
->disk_name
);
1790 INIT_LIST_HEAD(&ev
->node
);
1792 spin_lock_init(&ev
->lock
);
1793 mutex_init(&ev
->block_mutex
);
1795 ev
->poll_msecs
= -1;
1796 INIT_DELAYED_WORK(&ev
->dwork
, disk_events_workfn
);
1801 static void disk_add_events(struct gendisk
*disk
)
1806 /* FIXME: error handling */
1807 if (sysfs_create_files(&disk_to_dev(disk
)->kobj
, disk_events_attrs
) < 0)
1808 pr_warn("%s: failed to create sysfs files for events\n",
1811 mutex_lock(&disk_events_mutex
);
1812 list_add_tail(&disk
->ev
->node
, &disk_events
);
1813 mutex_unlock(&disk_events_mutex
);
1816 * Block count is initialized to 1 and the following initial
1817 * unblock kicks it into action.
1819 __disk_unblock_events(disk
, true);
1822 static void disk_del_events(struct gendisk
*disk
)
1827 disk_block_events(disk
);
1829 mutex_lock(&disk_events_mutex
);
1830 list_del_init(&disk
->ev
->node
);
1831 mutex_unlock(&disk_events_mutex
);
1833 sysfs_remove_files(&disk_to_dev(disk
)->kobj
, disk_events_attrs
);
1836 static void disk_release_events(struct gendisk
*disk
)
1838 /* the block count should be 1 from disk_del_events() */
1839 WARN_ON_ONCE(disk
->ev
&& disk
->ev
->block
!= 1);