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/ctype.h>
21 #include <linux/fs_uuid.h>
22 #include <linux/log2.h>
23 #include <linux/pm_runtime.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 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_bh(&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_bh(&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
);
834 seqf
->private = NULL
;
838 static void *show_partition_start(struct seq_file
*seqf
, loff_t
*pos
)
842 p
= disk_seqf_start(seqf
, pos
);
843 if (!IS_ERR_OR_NULL(p
) && !*pos
)
844 seq_puts(seqf
, "major minor #blocks name\n\n");
848 static int show_partition(struct seq_file
*seqf
, void *v
)
850 struct gendisk
*sgp
= v
;
851 struct disk_part_iter piter
;
852 struct hd_struct
*part
;
853 char buf
[BDEVNAME_SIZE
];
855 /* Don't show non-partitionable removeable devices or empty devices */
856 if (!get_capacity(sgp
) || (!disk_max_parts(sgp
) &&
857 (sgp
->flags
& GENHD_FL_REMOVABLE
)))
859 if (sgp
->flags
& GENHD_FL_SUPPRESS_PARTITION_INFO
)
862 /* show the full disk and all non-0 size partitions of it */
863 disk_part_iter_init(&piter
, sgp
, DISK_PITER_INCL_PART0
);
864 while ((part
= disk_part_iter_next(&piter
)))
865 seq_printf(seqf
, "%4d %7d %10llu %s\n",
866 MAJOR(part_devt(part
)), MINOR(part_devt(part
)),
867 (unsigned long long)part_nr_sects_read(part
) >> 1,
868 disk_name(sgp
, part
->partno
, buf
));
869 disk_part_iter_exit(&piter
);
874 static const struct seq_operations partitions_op
= {
875 .start
= show_partition_start
,
876 .next
= disk_seqf_next
,
877 .stop
= disk_seqf_stop
,
878 .show
= show_partition
881 static int partitions_open(struct inode
*inode
, struct file
*file
)
883 return seq_open(file
, &partitions_op
);
886 static const struct file_operations proc_partitions_operations
= {
887 .open
= partitions_open
,
890 .release
= seq_release
,
895 static struct kobject
*base_probe(dev_t devt
, int *partno
, void *data
)
897 if (request_module("block-major-%d-%d", MAJOR(devt
), MINOR(devt
)) > 0)
898 /* Make old-style 2.4 aliases work */
899 request_module("block-major-%d", MAJOR(devt
));
903 static int __init
genhd_device_init(void)
907 block_class
.dev_kobj
= sysfs_dev_block_kobj
;
908 error
= class_register(&block_class
);
911 bdev_map
= kobj_map_init(base_probe
, &block_class_lock
);
914 register_blkdev(BLOCK_EXT_MAJOR
, "blkext");
916 /* create top-level block dir */
917 if (!sysfs_deprecated
)
918 block_depr
= kobject_create_and_add("block", NULL
);
922 subsys_initcall(genhd_device_init
);
924 static ssize_t
disk_range_show(struct device
*dev
,
925 struct device_attribute
*attr
, char *buf
)
927 struct gendisk
*disk
= dev_to_disk(dev
);
929 return sprintf(buf
, "%d\n", disk
->minors
);
932 static ssize_t
disk_ext_range_show(struct device
*dev
,
933 struct device_attribute
*attr
, char *buf
)
935 struct gendisk
*disk
= dev_to_disk(dev
);
937 return sprintf(buf
, "%d\n", disk_max_parts(disk
));
940 static ssize_t
disk_removable_show(struct device
*dev
,
941 struct device_attribute
*attr
, char *buf
)
943 struct gendisk
*disk
= dev_to_disk(dev
);
945 return sprintf(buf
, "%d\n",
946 (disk
->flags
& GENHD_FL_REMOVABLE
? 1 : 0));
949 static ssize_t
disk_ro_show(struct device
*dev
,
950 struct device_attribute
*attr
, char *buf
)
952 struct gendisk
*disk
= dev_to_disk(dev
);
954 return sprintf(buf
, "%d\n", get_disk_ro(disk
) ? 1 : 0);
957 static ssize_t
disk_capability_show(struct device
*dev
,
958 struct device_attribute
*attr
, char *buf
)
960 struct gendisk
*disk
= dev_to_disk(dev
);
962 return sprintf(buf
, "%x\n", disk
->flags
);
965 static ssize_t
disk_alignment_offset_show(struct device
*dev
,
966 struct device_attribute
*attr
,
969 struct gendisk
*disk
= dev_to_disk(dev
);
971 return sprintf(buf
, "%d\n", queue_alignment_offset(disk
->queue
));
974 static ssize_t
disk_discard_alignment_show(struct device
*dev
,
975 struct device_attribute
*attr
,
978 struct gendisk
*disk
= dev_to_disk(dev
);
980 return sprintf(buf
, "%d\n", queue_discard_alignment(disk
->queue
));
983 static DEVICE_ATTR(range
, S_IRUGO
, disk_range_show
, NULL
);
984 static DEVICE_ATTR(ext_range
, S_IRUGO
, disk_ext_range_show
, NULL
);
985 static DEVICE_ATTR(removable
, S_IRUGO
, disk_removable_show
, NULL
);
986 static DEVICE_ATTR(ro
, S_IRUGO
, disk_ro_show
, NULL
);
987 static DEVICE_ATTR(size
, S_IRUGO
, part_size_show
, NULL
);
988 static DEVICE_ATTR(alignment_offset
, S_IRUGO
, disk_alignment_offset_show
, NULL
);
989 static DEVICE_ATTR(discard_alignment
, S_IRUGO
, disk_discard_alignment_show
,
991 static DEVICE_ATTR(capability
, S_IRUGO
, disk_capability_show
, NULL
);
992 static DEVICE_ATTR(stat
, S_IRUGO
, part_stat_show
, NULL
);
993 static DEVICE_ATTR(inflight
, S_IRUGO
, part_inflight_show
, NULL
);
994 #ifdef CONFIG_FAIL_MAKE_REQUEST
995 static struct device_attribute dev_attr_fail
=
996 __ATTR(make
-it
-fail
, S_IRUGO
|S_IWUSR
, part_fail_show
, part_fail_store
);
998 #ifdef CONFIG_FAIL_IO_TIMEOUT
999 static struct device_attribute dev_attr_fail_timeout
=
1000 __ATTR(io
-timeout
-fail
, S_IRUGO
|S_IWUSR
, part_timeout_show
,
1001 part_timeout_store
);
1004 static struct attribute
*disk_attrs
[] = {
1005 &dev_attr_range
.attr
,
1006 &dev_attr_ext_range
.attr
,
1007 &dev_attr_removable
.attr
,
1009 &dev_attr_size
.attr
,
1010 &dev_attr_alignment_offset
.attr
,
1011 &dev_attr_discard_alignment
.attr
,
1012 &dev_attr_capability
.attr
,
1013 &dev_attr_stat
.attr
,
1014 &dev_attr_inflight
.attr
,
1015 #ifdef CONFIG_FAIL_MAKE_REQUEST
1016 &dev_attr_fail
.attr
,
1018 #ifdef CONFIG_FAIL_IO_TIMEOUT
1019 &dev_attr_fail_timeout
.attr
,
1024 static struct attribute_group disk_attr_group
= {
1025 .attrs
= disk_attrs
,
1028 static const struct attribute_group
*disk_attr_groups
[] = {
1034 * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way
1035 * @disk: disk to replace part_tbl for
1036 * @new_ptbl: new part_tbl to install
1038 * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The
1039 * original ptbl is freed using RCU callback.
1042 * Matching bd_mutx locked.
1044 static void disk_replace_part_tbl(struct gendisk
*disk
,
1045 struct disk_part_tbl
*new_ptbl
)
1047 struct disk_part_tbl
*old_ptbl
= disk
->part_tbl
;
1049 rcu_assign_pointer(disk
->part_tbl
, new_ptbl
);
1052 rcu_assign_pointer(old_ptbl
->last_lookup
, NULL
);
1053 kfree_rcu(old_ptbl
, rcu_head
);
1058 * disk_expand_part_tbl - expand disk->part_tbl
1059 * @disk: disk to expand part_tbl for
1060 * @partno: expand such that this partno can fit in
1062 * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl
1063 * uses RCU to allow unlocked dereferencing for stats and other stuff.
1066 * Matching bd_mutex locked, might sleep.
1069 * 0 on success, -errno on failure.
1071 int disk_expand_part_tbl(struct gendisk
*disk
, int partno
)
1073 struct disk_part_tbl
*old_ptbl
= disk
->part_tbl
;
1074 struct disk_part_tbl
*new_ptbl
;
1075 int len
= old_ptbl
? old_ptbl
->len
: 0;
1080 * check for int overflow, since we can get here from blkpg_ioctl()
1081 * with a user passed 'partno'.
1083 target
= partno
+ 1;
1087 /* disk_max_parts() is zero during initialization, ignore if so */
1088 if (disk_max_parts(disk
) && target
> disk_max_parts(disk
))
1094 size
= sizeof(*new_ptbl
) + target
* sizeof(new_ptbl
->part
[0]);
1095 new_ptbl
= kzalloc_node(size
, GFP_KERNEL
, disk
->node_id
);
1099 new_ptbl
->len
= target
;
1101 for (i
= 0; i
< len
; i
++)
1102 rcu_assign_pointer(new_ptbl
->part
[i
], old_ptbl
->part
[i
]);
1104 disk_replace_part_tbl(disk
, new_ptbl
);
1108 static void disk_release(struct device
*dev
)
1110 struct gendisk
*disk
= dev_to_disk(dev
);
1112 blk_free_devt(dev
->devt
);
1113 disk_release_events(disk
);
1114 kfree(disk
->random
);
1115 disk_replace_part_tbl(disk
, NULL
);
1116 free_part_stats(&disk
->part0
);
1117 free_part_info(&disk
->part0
);
1119 blk_put_queue(disk
->queue
);
1123 static int disk_uevent(struct device
*dev
, struct kobj_uevent_env
*env
)
1125 struct gendisk
*disk
= dev_to_disk(dev
);
1126 struct disk_part_iter piter
;
1127 struct hd_struct
*part
;
1130 disk_part_iter_init(&piter
, disk
, 0);
1131 while((part
= disk_part_iter_next(&piter
)))
1133 disk_part_iter_exit(&piter
);
1134 add_uevent_var(env
, "NPARTS=%u", cnt
);
1138 struct class block_class
= {
1142 static char *block_devnode(struct device
*dev
, umode_t
*mode
,
1143 kuid_t
*uid
, kgid_t
*gid
)
1145 struct gendisk
*disk
= dev_to_disk(dev
);
1148 return disk
->devnode(disk
, mode
);
1152 static struct device_type disk_type
= {
1154 .groups
= disk_attr_groups
,
1155 .release
= disk_release
,
1156 .devnode
= block_devnode
,
1157 .uevent
= disk_uevent
,
1160 #ifdef CONFIG_PROC_FS
1162 * aggregate disk stat collector. Uses the same stats that the sysfs
1163 * entries do, above, but makes them available through one seq_file.
1165 * The output looks suspiciously like /proc/partitions with a bunch of
1168 static int diskstats_show(struct seq_file
*seqf
, void *v
)
1170 struct gendisk
*gp
= v
;
1171 struct disk_part_iter piter
;
1172 struct hd_struct
*hd
;
1173 char buf
[BDEVNAME_SIZE
];
1177 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1178 seq_puts(seqf, "major minor name"
1179 " rio rmerge rsect ruse wio wmerge "
1180 "wsect wuse running use aveq"
1184 disk_part_iter_init(&piter
, gp
, DISK_PITER_INCL_EMPTY_PART0
);
1185 while ((hd
= disk_part_iter_next(&piter
))) {
1186 cpu
= part_stat_lock();
1187 part_round_stats(cpu
, hd
);
1189 seq_printf(seqf
, "%4d %7d %s %lu %lu %lu "
1190 "%u %lu %lu %lu %u %u %u %u\n",
1191 MAJOR(part_devt(hd
)), MINOR(part_devt(hd
)),
1192 disk_name(gp
, hd
->partno
, buf
),
1193 part_stat_read(hd
, ios
[READ
]),
1194 part_stat_read(hd
, merges
[READ
]),
1195 part_stat_read(hd
, sectors
[READ
]),
1196 jiffies_to_msecs(part_stat_read(hd
, ticks
[READ
])),
1197 part_stat_read(hd
, ios
[WRITE
]),
1198 part_stat_read(hd
, merges
[WRITE
]),
1199 part_stat_read(hd
, sectors
[WRITE
]),
1200 jiffies_to_msecs(part_stat_read(hd
, ticks
[WRITE
])),
1202 jiffies_to_msecs(part_stat_read(hd
, io_ticks
)),
1203 jiffies_to_msecs(part_stat_read(hd
, time_in_queue
))
1206 disk_part_iter_exit(&piter
);
1211 static const struct seq_operations diskstats_op
= {
1212 .start
= disk_seqf_start
,
1213 .next
= disk_seqf_next
,
1214 .stop
= disk_seqf_stop
,
1215 .show
= diskstats_show
1218 static int diskstats_open(struct inode
*inode
, struct file
*file
)
1220 return seq_open(file
, &diskstats_op
);
1223 static const struct file_operations proc_diskstats_operations
= {
1224 .open
= diskstats_open
,
1226 .llseek
= seq_lseek
,
1227 .release
= seq_release
,
1230 static int __init
proc_genhd_init(void)
1232 proc_create("diskstats", 0, NULL
, &proc_diskstats_operations
);
1233 proc_create("partitions", 0, NULL
, &proc_partitions_operations
);
1236 module_init(proc_genhd_init
);
1237 #endif /* CONFIG_PROC_FS */
1239 dev_t
blk_lookup_devt(const char *name
, int partno
)
1241 dev_t devt
= MKDEV(0, 0);
1242 struct class_dev_iter iter
;
1245 class_dev_iter_init(&iter
, &block_class
, NULL
, &disk_type
);
1246 while ((dev
= class_dev_iter_next(&iter
))) {
1247 struct gendisk
*disk
= dev_to_disk(dev
);
1248 struct hd_struct
*part
;
1250 if (strcmp(dev_name(dev
), name
))
1253 if (partno
< disk
->minors
) {
1254 /* We need to return the right devno, even
1255 * if the partition doesn't exist yet.
1257 devt
= MKDEV(MAJOR(dev
->devt
),
1258 MINOR(dev
->devt
) + partno
);
1261 part
= disk_get_part(disk
, partno
);
1263 devt
= part_devt(part
);
1264 disk_put_part(part
);
1267 disk_put_part(part
);
1269 class_dev_iter_exit(&iter
);
1272 EXPORT_SYMBOL(blk_lookup_devt
);
1274 struct gendisk
*alloc_disk(int minors
)
1276 return alloc_disk_node(minors
, NUMA_NO_NODE
);
1278 EXPORT_SYMBOL(alloc_disk
);
1280 struct gendisk
*alloc_disk_node(int minors
, int node_id
)
1282 struct gendisk
*disk
;
1284 disk
= kmalloc_node(sizeof(struct gendisk
),
1285 GFP_KERNEL
| __GFP_ZERO
, node_id
);
1287 if (!init_part_stats(&disk
->part0
)) {
1291 disk
->node_id
= node_id
;
1292 if (disk_expand_part_tbl(disk
, 0)) {
1293 free_part_stats(&disk
->part0
);
1297 disk
->part_tbl
->part
[0] = &disk
->part0
;
1300 * set_capacity() and get_capacity() currently don't use
1301 * seqcounter to read/update the part0->nr_sects. Still init
1302 * the counter as we can read the sectors in IO submission
1303 * patch using seqence counters.
1305 * TODO: Ideally set_capacity() and get_capacity() should be
1306 * converted to make use of bd_mutex and sequence counters.
1308 seqcount_init(&disk
->part0
.nr_sects_seq
);
1309 hd_ref_init(&disk
->part0
);
1311 disk
->minors
= minors
;
1312 rand_initialize_disk(disk
);
1313 disk_to_dev(disk
)->class = &block_class
;
1314 disk_to_dev(disk
)->type
= &disk_type
;
1315 device_initialize(disk_to_dev(disk
));
1319 EXPORT_SYMBOL(alloc_disk_node
);
1321 struct kobject
*get_disk(struct gendisk
*disk
)
1323 struct module
*owner
;
1324 struct kobject
*kobj
;
1328 owner
= disk
->fops
->owner
;
1329 if (owner
&& !try_module_get(owner
))
1331 kobj
= kobject_get(&disk_to_dev(disk
)->kobj
);
1340 EXPORT_SYMBOL(get_disk
);
1342 void put_disk(struct gendisk
*disk
)
1345 kobject_put(&disk_to_dev(disk
)->kobj
);
1348 EXPORT_SYMBOL(put_disk
);
1350 static void set_disk_ro_uevent(struct gendisk
*gd
, int ro
)
1352 char event
[] = "DISK_RO=1";
1353 char *envp
[] = { event
, NULL
};
1357 kobject_uevent_env(&disk_to_dev(gd
)->kobj
, KOBJ_CHANGE
, envp
);
1360 void set_device_ro(struct block_device
*bdev
, int flag
)
1362 bdev
->bd_part
->policy
= flag
;
1365 EXPORT_SYMBOL(set_device_ro
);
1367 void set_disk_ro(struct gendisk
*disk
, int flag
)
1369 struct disk_part_iter piter
;
1370 struct hd_struct
*part
;
1372 if (disk
->part0
.policy
!= flag
) {
1373 set_disk_ro_uevent(disk
, flag
);
1374 disk
->part0
.policy
= flag
;
1377 disk_part_iter_init(&piter
, disk
, DISK_PITER_INCL_EMPTY
);
1378 while ((part
= disk_part_iter_next(&piter
)))
1379 part
->policy
= flag
;
1380 disk_part_iter_exit(&piter
);
1383 EXPORT_SYMBOL(set_disk_ro
);
1385 int bdev_read_only(struct block_device
*bdev
)
1389 return bdev
->bd_part
->policy
;
1392 EXPORT_SYMBOL(bdev_read_only
);
1394 int invalidate_partition(struct gendisk
*disk
, int partno
)
1397 struct block_device
*bdev
= bdget_disk(disk
, partno
);
1400 res
= __invalidate_device(bdev
, true);
1406 EXPORT_SYMBOL(invalidate_partition
);
1408 dev_t
blk_lookup_fs_info(struct fs_info
*seek
)
1410 dev_t devt
= MKDEV(0, 0);
1411 struct class_dev_iter iter
;
1415 class_dev_iter_init(&iter
, &block_class
, NULL
, &disk_type
);
1416 while (best_score
< 3 && (dev
= class_dev_iter_next(&iter
))) {
1417 struct gendisk
*disk
= dev_to_disk(dev
);
1418 struct disk_part_iter piter
;
1419 struct hd_struct
*part
;
1421 disk_part_iter_init(&piter
, disk
, DISK_PITER_INCL_PART0
);
1423 while (best_score
< 3 && (part
= disk_part_iter_next(&piter
))) {
1424 int score
= part_matches_fs_info(part
, seek
);
1425 if (score
> best_score
) {
1426 devt
= part_devt(part
);
1430 disk_part_iter_exit(&piter
);
1432 class_dev_iter_exit(&iter
);
1435 EXPORT_SYMBOL_GPL(blk_lookup_fs_info
);
1437 /* Caller uses NULL, key to start. For each match found, we return a bdev on
1438 * which we have done blkdev_get, and we do the blkdev_put on block devices
1439 * that are passed to us. When no more matches are found, we return NULL.
1441 struct block_device
*next_bdev_of_type(struct block_device
*last
,
1444 dev_t devt
= MKDEV(0, 0);
1445 struct class_dev_iter iter
;
1447 struct block_device
*next
= NULL
, *bdev
;
1453 class_dev_iter_init(&iter
, &block_class
, NULL
, &disk_type
);
1454 while (!devt
&& (dev
= class_dev_iter_next(&iter
))) {
1455 struct gendisk
*disk
= dev_to_disk(dev
);
1456 struct disk_part_iter piter
;
1457 struct hd_struct
*part
;
1459 disk_part_iter_init(&piter
, disk
, DISK_PITER_INCL_PART0
);
1461 while ((part
= disk_part_iter_next(&piter
))) {
1462 bdev
= bdget(part_devt(part
));
1463 if (last
&& !got_last
) {
1469 if (blkdev_get(bdev
, FMODE_READ
, 0))
1472 if (bdev_matches_key(bdev
, key
)) {
1477 blkdev_put(bdev
, FMODE_READ
);
1479 disk_part_iter_exit(&piter
);
1481 class_dev_iter_exit(&iter
);
1484 blkdev_put(last
, FMODE_READ
);
1487 EXPORT_SYMBOL_GPL(next_bdev_of_type
);
1490 * Disk events - monitor disk events like media change and eject request.
1492 struct disk_events
{
1493 struct list_head node
; /* all disk_event's */
1494 struct gendisk
*disk
; /* the associated disk */
1497 struct mutex block_mutex
; /* protects blocking */
1498 int block
; /* event blocking depth */
1499 unsigned int pending
; /* events already sent out */
1500 unsigned int clearing
; /* events being cleared */
1502 long poll_msecs
; /* interval, -1 for default */
1503 struct delayed_work dwork
;
1506 static const char *disk_events_strs
[] = {
1507 [ilog2(DISK_EVENT_MEDIA_CHANGE
)] = "media_change",
1508 [ilog2(DISK_EVENT_EJECT_REQUEST
)] = "eject_request",
1509 #ifdef CONFIG_MTK_MULTI_PARTITION_MOUNT_ONLY_SUPPORT
1510 [ilog2(DISK_EVENT_MEDIA_DISAPPEAR
)] = "media_disappear",
1514 static char *disk_uevents
[] = {
1515 [ilog2(DISK_EVENT_MEDIA_CHANGE
)] = "DISK_MEDIA_CHANGE=1",
1516 [ilog2(DISK_EVENT_EJECT_REQUEST
)] = "DISK_EJECT_REQUEST=1",
1517 #ifdef CONFIG_MTK_MULTI_PARTITION_MOUNT_ONLY_SUPPORT
1518 [ilog2(DISK_EVENT_MEDIA_DISAPPEAR
)] = "DISK_EVENT_MEDIA_DISAPPEAR=1",
1522 /* list of all disk_events */
1523 static DEFINE_MUTEX(disk_events_mutex
);
1524 static LIST_HEAD(disk_events
);
1526 /* disable in-kernel polling by default */
1527 //ALPS00319570, CL955952 merged back, begin
1528 //static unsigned long disk_events_dfl_poll_msecs = 0; //original
1529 static unsigned long disk_events_dfl_poll_msecs
= 2000;
1530 //ALPS00319570, CL955952 merged back, end
1532 static unsigned long disk_events_poll_jiffies(struct gendisk
*disk
)
1534 struct disk_events
*ev
= disk
->ev
;
1535 long intv_msecs
= 0;
1538 * If device-specific poll interval is set, always use it. If
1539 * the default is being used, poll iff there are events which
1540 * can't be monitored asynchronously.
1542 if (ev
->poll_msecs
>= 0)
1543 intv_msecs
= ev
->poll_msecs
;
1544 else if (disk
->events
& ~disk
->async_events
)
1545 intv_msecs
= disk_events_dfl_poll_msecs
;
1547 return msecs_to_jiffies(intv_msecs
);
1551 * disk_block_events - block and flush disk event checking
1552 * @disk: disk to block events for
1554 * On return from this function, it is guaranteed that event checking
1555 * isn't in progress and won't happen until unblocked by
1556 * disk_unblock_events(). Events blocking is counted and the actual
1557 * unblocking happens after the matching number of unblocks are done.
1559 * Note that this intentionally does not block event checking from
1560 * disk_clear_events().
1565 void disk_block_events(struct gendisk
*disk
)
1567 struct disk_events
*ev
= disk
->ev
;
1568 unsigned long flags
;
1575 * Outer mutex ensures that the first blocker completes canceling
1576 * the event work before further blockers are allowed to finish.
1578 mutex_lock(&ev
->block_mutex
);
1580 spin_lock_irqsave(&ev
->lock
, flags
);
1581 cancel
= !ev
->block
++;
1582 spin_unlock_irqrestore(&ev
->lock
, flags
);
1585 cancel_delayed_work_sync(&disk
->ev
->dwork
);
1587 mutex_unlock(&ev
->block_mutex
);
1590 static void __disk_unblock_events(struct gendisk
*disk
, bool check_now
)
1592 struct disk_events
*ev
= disk
->ev
;
1594 unsigned long flags
;
1596 spin_lock_irqsave(&ev
->lock
, flags
);
1598 if (WARN_ON_ONCE(ev
->block
<= 0))
1605 * Not exactly a latency critical operation, set poll timer
1606 * slack to 25% and kick event check.
1608 intv
= disk_events_poll_jiffies(disk
);
1609 set_timer_slack(&ev
->dwork
.timer
, intv
/ 4);
1611 queue_delayed_work(system_freezable_wq
, &ev
->dwork
, 0);
1613 queue_delayed_work(system_freezable_wq
, &ev
->dwork
, intv
);
1615 spin_unlock_irqrestore(&ev
->lock
, flags
);
1619 * disk_unblock_events - unblock disk event checking
1620 * @disk: disk to unblock events for
1622 * Undo disk_block_events(). When the block count reaches zero, it
1623 * starts events polling if configured.
1626 * Don't care. Safe to call from irq context.
1628 void disk_unblock_events(struct gendisk
*disk
)
1631 __disk_unblock_events(disk
, false);
1635 * disk_flush_events - schedule immediate event checking and flushing
1636 * @disk: disk to check and flush events for
1637 * @mask: events to flush
1639 * Schedule immediate event checking on @disk if not blocked. Events in
1640 * @mask are scheduled to be cleared from the driver. Note that this
1641 * doesn't clear the events from @disk->ev.
1644 * If @mask is non-zero must be called with bdev->bd_mutex held.
1646 void disk_flush_events(struct gendisk
*disk
, unsigned int mask
)
1648 struct disk_events
*ev
= disk
->ev
;
1653 spin_lock_irq(&ev
->lock
);
1654 ev
->clearing
|= mask
;
1656 mod_delayed_work(system_freezable_wq
, &ev
->dwork
, 0);
1657 spin_unlock_irq(&ev
->lock
);
1661 * disk_clear_events - synchronously check, clear and return pending events
1662 * @disk: disk to fetch and clear events from
1663 * @mask: mask of events to be fetched and clearted
1665 * Disk events are synchronously checked and pending events in @mask
1666 * are cleared and returned. This ignores the block count.
1671 unsigned int disk_clear_events(struct gendisk
*disk
, unsigned int mask
)
1673 const struct block_device_operations
*bdops
= disk
->fops
;
1674 struct disk_events
*ev
= disk
->ev
;
1675 unsigned int pending
;
1676 unsigned int clearing
= mask
;
1679 /* for drivers still using the old ->media_changed method */
1680 if ((mask
& DISK_EVENT_MEDIA_CHANGE
) &&
1681 bdops
->media_changed
&& bdops
->media_changed(disk
))
1682 return DISK_EVENT_MEDIA_CHANGE
;
1686 disk_block_events(disk
);
1689 * store the union of mask and ev->clearing on the stack so that the
1690 * race with disk_flush_events does not cause ambiguity (ev->clearing
1691 * can still be modified even if events are blocked).
1693 spin_lock_irq(&ev
->lock
);
1694 clearing
|= ev
->clearing
;
1696 spin_unlock_irq(&ev
->lock
);
1698 disk_check_events(ev
, &clearing
);
1700 * if ev->clearing is not 0, the disk_flush_events got called in the
1701 * middle of this function, so we want to run the workfn without delay.
1703 __disk_unblock_events(disk
, ev
->clearing
? true : false);
1705 /* then, fetch and clear pending events */
1706 spin_lock_irq(&ev
->lock
);
1707 pending
= ev
->pending
& mask
;
1708 ev
->pending
&= ~mask
;
1709 spin_unlock_irq(&ev
->lock
);
1710 WARN_ON_ONCE(clearing
& mask
);
1716 * Separate this part out so that a different pointer for clearing_ptr can be
1717 * passed in for disk_clear_events.
1719 static void disk_events_workfn(struct work_struct
*work
)
1721 struct delayed_work
*dwork
= to_delayed_work(work
);
1722 struct disk_events
*ev
= container_of(dwork
, struct disk_events
, dwork
);
1724 disk_check_events(ev
, &ev
->clearing
);
1727 static void disk_check_events(struct disk_events
*ev
,
1728 unsigned int *clearing_ptr
)
1730 struct gendisk
*disk
= ev
->disk
;
1731 char *envp
[ARRAY_SIZE(disk_uevents
) + 1] = { };
1732 unsigned int clearing
= *clearing_ptr
;
1733 unsigned int events
;
1735 int nr_events
= 0, i
;
1738 events
= disk
->fops
->check_events(disk
, clearing
);
1740 /* accumulate pending events and schedule next poll if necessary */
1741 spin_lock_irq(&ev
->lock
);
1743 events
&= ~ev
->pending
;
1744 ev
->pending
|= events
;
1745 *clearing_ptr
&= ~clearing
;
1747 intv
= disk_events_poll_jiffies(disk
);
1748 if (!ev
->block
&& intv
)
1749 queue_delayed_work(system_freezable_wq
, &ev
->dwork
, intv
);
1751 spin_unlock_irq(&ev
->lock
);
1754 * Tell userland about new events. Only the events listed in
1755 * @disk->events are reported. Unlisted events are processed the
1756 * same internally but never get reported to userland.
1758 for (i
= 0; i
< ARRAY_SIZE(disk_uevents
); i
++)
1759 if (events
& disk
->events
& (1 << i
))
1760 envp
[nr_events
++] = disk_uevents
[i
];
1763 kobject_uevent_env(&disk_to_dev(disk
)->kobj
, KOBJ_CHANGE
, envp
);
1767 * A disk events enabled device has the following sysfs nodes under
1768 * its /sys/block/X/ directory.
1770 * events : list of all supported events
1771 * events_async : list of events which can be detected w/o polling
1772 * events_poll_msecs : polling interval, 0: disable, -1: system default
1774 static ssize_t
__disk_events_show(unsigned int events
, char *buf
)
1776 const char *delim
= "";
1780 for (i
= 0; i
< ARRAY_SIZE(disk_events_strs
); i
++)
1781 if (events
& (1 << i
)) {
1782 pos
+= sprintf(buf
+ pos
, "%s%s",
1783 delim
, disk_events_strs
[i
]);
1787 pos
+= sprintf(buf
+ pos
, "\n");
1791 static ssize_t
disk_events_show(struct device
*dev
,
1792 struct device_attribute
*attr
, char *buf
)
1794 struct gendisk
*disk
= dev_to_disk(dev
);
1796 return __disk_events_show(disk
->events
, buf
);
1799 static ssize_t
disk_events_async_show(struct device
*dev
,
1800 struct device_attribute
*attr
, char *buf
)
1802 struct gendisk
*disk
= dev_to_disk(dev
);
1804 return __disk_events_show(disk
->async_events
, buf
);
1807 static ssize_t
disk_events_poll_msecs_show(struct device
*dev
,
1808 struct device_attribute
*attr
,
1811 struct gendisk
*disk
= dev_to_disk(dev
);
1813 return sprintf(buf
, "%ld\n", disk
->ev
->poll_msecs
);
1816 static ssize_t
disk_events_poll_msecs_store(struct device
*dev
,
1817 struct device_attribute
*attr
,
1818 const char *buf
, size_t count
)
1820 struct gendisk
*disk
= dev_to_disk(dev
);
1823 if (!count
|| !sscanf(buf
, "%ld", &intv
))
1826 if (intv
< 0 && intv
!= -1)
1829 disk_block_events(disk
);
1830 disk
->ev
->poll_msecs
= intv
;
1831 __disk_unblock_events(disk
, true);
1836 static const DEVICE_ATTR(events
, S_IRUGO
, disk_events_show
, NULL
);
1837 static const DEVICE_ATTR(events_async
, S_IRUGO
, disk_events_async_show
, NULL
);
1838 static const DEVICE_ATTR(events_poll_msecs
, S_IRUGO
|S_IWUSR
,
1839 disk_events_poll_msecs_show
,
1840 disk_events_poll_msecs_store
);
1842 static const struct attribute
*disk_events_attrs
[] = {
1843 &dev_attr_events
.attr
,
1844 &dev_attr_events_async
.attr
,
1845 &dev_attr_events_poll_msecs
.attr
,
1850 * The default polling interval can be specified by the kernel
1851 * parameter block.events_dfl_poll_msecs which defaults to 0
1852 * (disable). This can also be modified runtime by writing to
1853 * /sys/module/block/events_dfl_poll_msecs.
1855 static int disk_events_set_dfl_poll_msecs(const char *val
,
1856 const struct kernel_param
*kp
)
1858 struct disk_events
*ev
;
1861 ret
= param_set_ulong(val
, kp
);
1865 mutex_lock(&disk_events_mutex
);
1867 list_for_each_entry(ev
, &disk_events
, node
)
1868 disk_flush_events(ev
->disk
, 0);
1870 mutex_unlock(&disk_events_mutex
);
1875 static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops
= {
1876 .set
= disk_events_set_dfl_poll_msecs
,
1877 .get
= param_get_ulong
,
1880 #undef MODULE_PARAM_PREFIX
1881 #define MODULE_PARAM_PREFIX "block."
1883 module_param_cb(events_dfl_poll_msecs
, &disk_events_dfl_poll_msecs_param_ops
,
1884 &disk_events_dfl_poll_msecs
, 0644);
1887 * disk_{alloc|add|del|release}_events - initialize and destroy disk_events.
1889 static void disk_alloc_events(struct gendisk
*disk
)
1891 struct disk_events
*ev
;
1893 if (!disk
->fops
->check_events
)
1896 ev
= kzalloc(sizeof(*ev
), GFP_KERNEL
);
1898 pr_warn("%s: failed to initialize events\n", disk
->disk_name
);
1902 INIT_LIST_HEAD(&ev
->node
);
1904 spin_lock_init(&ev
->lock
);
1905 mutex_init(&ev
->block_mutex
);
1907 ev
->poll_msecs
= -1;
1908 INIT_DELAYED_WORK(&ev
->dwork
, disk_events_workfn
);
1913 static void disk_add_events(struct gendisk
*disk
)
1918 /* FIXME: error handling */
1919 if (sysfs_create_files(&disk_to_dev(disk
)->kobj
, disk_events_attrs
) < 0)
1920 pr_warn("%s: failed to create sysfs files for events\n",
1923 mutex_lock(&disk_events_mutex
);
1924 list_add_tail(&disk
->ev
->node
, &disk_events
);
1925 mutex_unlock(&disk_events_mutex
);
1928 * Block count is initialized to 1 and the following initial
1929 * unblock kicks it into action.
1931 __disk_unblock_events(disk
, true);
1934 static void disk_del_events(struct gendisk
*disk
)
1939 disk_block_events(disk
);
1941 mutex_lock(&disk_events_mutex
);
1942 list_del_init(&disk
->ev
->node
);
1943 mutex_unlock(&disk_events_mutex
);
1945 sysfs_remove_files(&disk_to_dev(disk
)->kobj
, disk_events_attrs
);
1948 static void disk_release_events(struct gendisk
*disk
)
1950 /* the block count should be 1 from disk_del_events() */
1951 WARN_ON_ONCE(disk
->ev
&& disk
->ev
->block
!= 1);