5 #include <linux/module.h>
7 #include <linux/genhd.h>
8 #include <linux/kdev_t.h>
9 #include <linux/kernel.h>
10 #include <linux/blkdev.h>
11 #include <linux/backing-dev.h>
12 #include <linux/init.h>
13 #include <linux/spinlock.h>
14 #include <linux/proc_fs.h>
15 #include <linux/seq_file.h>
16 #include <linux/slab.h>
17 #include <linux/kmod.h>
18 #include <linux/kobj_map.h>
19 #include <linux/mutex.h>
20 #include <linux/idr.h>
21 #include <linux/log2.h>
22 #include <linux/pm_runtime.h>
23 #include <linux/badblocks.h>
27 static DEFINE_MUTEX(block_class_lock
);
28 struct kobject
*block_depr
;
30 /* for extended dynamic devt allocation, currently only one major is used */
31 #define NR_EXT_DEVT (1 << MINORBITS)
33 /* For extended devt allocation. ext_devt_lock prevents look up
34 * results from going away underneath its user.
36 static DEFINE_SPINLOCK(ext_devt_lock
);
37 static DEFINE_IDR(ext_devt_idr
);
39 static const 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
);
48 void part_inc_in_flight(struct request_queue
*q
, struct hd_struct
*part
, int rw
)
53 atomic_inc(&part
->in_flight
[rw
]);
55 atomic_inc(&part_to_disk(part
)->part0
.in_flight
[rw
]);
58 void part_dec_in_flight(struct request_queue
*q
, struct hd_struct
*part
, int rw
)
63 atomic_dec(&part
->in_flight
[rw
]);
65 atomic_dec(&part_to_disk(part
)->part0
.in_flight
[rw
]);
68 void part_in_flight(struct request_queue
*q
, struct hd_struct
*part
,
69 unsigned int inflight
[2])
72 blk_mq_in_flight(q
, part
, inflight
);
76 inflight
[0] = atomic_read(&part
->in_flight
[0]) +
77 atomic_read(&part
->in_flight
[1]);
79 part
= &part_to_disk(part
)->part0
;
80 inflight
[1] = atomic_read(&part
->in_flight
[0]) +
81 atomic_read(&part
->in_flight
[1]);
85 struct hd_struct
*__disk_get_part(struct gendisk
*disk
, int partno
)
87 struct disk_part_tbl
*ptbl
= rcu_dereference(disk
->part_tbl
);
89 if (unlikely(partno
< 0 || partno
>= ptbl
->len
))
91 return rcu_dereference(ptbl
->part
[partno
]);
95 * disk_get_part - get partition
96 * @disk: disk to look partition from
97 * @partno: partition number
99 * Look for partition @partno from @disk. If found, increment
100 * reference count and return it.
106 * Pointer to the found partition on success, NULL if not found.
108 struct hd_struct
*disk_get_part(struct gendisk
*disk
, int partno
)
110 struct hd_struct
*part
;
113 part
= __disk_get_part(disk
, partno
);
115 get_device(part_to_dev(part
));
120 EXPORT_SYMBOL_GPL(disk_get_part
);
123 * disk_part_iter_init - initialize partition iterator
124 * @piter: iterator to initialize
125 * @disk: disk to iterate over
126 * @flags: DISK_PITER_* flags
128 * Initialize @piter so that it iterates over partitions of @disk.
133 void disk_part_iter_init(struct disk_part_iter
*piter
, struct gendisk
*disk
,
136 struct disk_part_tbl
*ptbl
;
139 ptbl
= rcu_dereference(disk
->part_tbl
);
144 if (flags
& DISK_PITER_REVERSE
)
145 piter
->idx
= ptbl
->len
- 1;
146 else if (flags
& (DISK_PITER_INCL_PART0
| DISK_PITER_INCL_EMPTY_PART0
))
151 piter
->flags
= flags
;
155 EXPORT_SYMBOL_GPL(disk_part_iter_init
);
158 * disk_part_iter_next - proceed iterator to the next partition and return it
159 * @piter: iterator of interest
161 * Proceed @piter to the next partition and return it.
166 struct hd_struct
*disk_part_iter_next(struct disk_part_iter
*piter
)
168 struct disk_part_tbl
*ptbl
;
171 /* put the last partition */
172 disk_put_part(piter
->part
);
177 ptbl
= rcu_dereference(piter
->disk
->part_tbl
);
179 /* determine iteration parameters */
180 if (piter
->flags
& DISK_PITER_REVERSE
) {
182 if (piter
->flags
& (DISK_PITER_INCL_PART0
|
183 DISK_PITER_INCL_EMPTY_PART0
))
192 /* iterate to the next partition */
193 for (; piter
->idx
!= end
; piter
->idx
+= inc
) {
194 struct hd_struct
*part
;
196 part
= rcu_dereference(ptbl
->part
[piter
->idx
]);
199 if (!part_nr_sects_read(part
) &&
200 !(piter
->flags
& DISK_PITER_INCL_EMPTY
) &&
201 !(piter
->flags
& DISK_PITER_INCL_EMPTY_PART0
&&
205 get_device(part_to_dev(part
));
215 EXPORT_SYMBOL_GPL(disk_part_iter_next
);
218 * disk_part_iter_exit - finish up partition iteration
219 * @piter: iter of interest
221 * Called when iteration is over. Cleans up @piter.
226 void disk_part_iter_exit(struct disk_part_iter
*piter
)
228 disk_put_part(piter
->part
);
231 EXPORT_SYMBOL_GPL(disk_part_iter_exit
);
233 static inline int sector_in_part(struct hd_struct
*part
, sector_t sector
)
235 return part
->start_sect
<= sector
&&
236 sector
< part
->start_sect
+ part_nr_sects_read(part
);
240 * disk_map_sector_rcu - map sector to partition
241 * @disk: gendisk of interest
242 * @sector: sector to map
244 * Find out which partition @sector maps to on @disk. This is
245 * primarily used for stats accounting.
248 * RCU read locked. The returned partition pointer is valid only
249 * while preemption is disabled.
252 * Found partition on success, part0 is returned if no partition matches
254 struct hd_struct
*disk_map_sector_rcu(struct gendisk
*disk
, sector_t sector
)
256 struct disk_part_tbl
*ptbl
;
257 struct hd_struct
*part
;
260 ptbl
= rcu_dereference(disk
->part_tbl
);
262 part
= rcu_dereference(ptbl
->last_lookup
);
263 if (part
&& sector_in_part(part
, sector
))
266 for (i
= 1; i
< ptbl
->len
; i
++) {
267 part
= rcu_dereference(ptbl
->part
[i
]);
269 if (part
&& sector_in_part(part
, sector
)) {
270 rcu_assign_pointer(ptbl
->last_lookup
, part
);
276 EXPORT_SYMBOL_GPL(disk_map_sector_rcu
);
279 * Can be deleted altogether. Later.
282 #define BLKDEV_MAJOR_HASH_SIZE 255
283 static struct blk_major_name
{
284 struct blk_major_name
*next
;
287 } *major_names
[BLKDEV_MAJOR_HASH_SIZE
];
289 /* index in the above - for now: assume no multimajor ranges */
290 static inline int major_to_index(unsigned major
)
292 return major
% BLKDEV_MAJOR_HASH_SIZE
;
295 #ifdef CONFIG_PROC_FS
296 void blkdev_show(struct seq_file
*seqf
, off_t offset
)
298 struct blk_major_name
*dp
;
300 mutex_lock(&block_class_lock
);
301 for (dp
= major_names
[major_to_index(offset
)]; dp
; dp
= dp
->next
)
302 if (dp
->major
== offset
)
303 seq_printf(seqf
, "%3d %s\n", dp
->major
, dp
->name
);
304 mutex_unlock(&block_class_lock
);
306 #endif /* CONFIG_PROC_FS */
309 * register_blkdev - register a new block device
311 * @major: the requested major device number [1..255]. If @major = 0, try to
312 * allocate any unused major number.
313 * @name: the name of the new block device as a zero terminated string
315 * The @name must be unique within the system.
317 * The return value depends on the @major input parameter:
319 * - if a major device number was requested in range [1..255] then the
320 * function returns zero on success, or a negative error code
321 * - if any unused major number was requested with @major = 0 parameter
322 * then the return value is the allocated major number in range
323 * [1..255] or a negative error code otherwise
325 int register_blkdev(unsigned int major
, const char *name
)
327 struct blk_major_name
**n
, *p
;
330 mutex_lock(&block_class_lock
);
334 for (index
= ARRAY_SIZE(major_names
)-1; index
> 0; index
--) {
335 if (major_names
[index
] == NULL
)
340 printk("register_blkdev: failed to get major for %s\n",
349 if (major
>= BLKDEV_MAJOR_MAX
) {
350 pr_err("register_blkdev: major requested (%d) is greater than the maximum (%d) for %s\n",
351 major
, BLKDEV_MAJOR_MAX
, name
);
357 p
= kmalloc(sizeof(struct blk_major_name
), GFP_KERNEL
);
364 strlcpy(p
->name
, name
, sizeof(p
->name
));
366 index
= major_to_index(major
);
368 for (n
= &major_names
[index
]; *n
; n
= &(*n
)->next
) {
369 if ((*n
)->major
== major
)
378 printk("register_blkdev: cannot get major %d for %s\n",
383 mutex_unlock(&block_class_lock
);
387 EXPORT_SYMBOL(register_blkdev
);
389 void unregister_blkdev(unsigned int major
, const char *name
)
391 struct blk_major_name
**n
;
392 struct blk_major_name
*p
= NULL
;
393 int index
= major_to_index(major
);
395 mutex_lock(&block_class_lock
);
396 for (n
= &major_names
[index
]; *n
; n
= &(*n
)->next
)
397 if ((*n
)->major
== major
)
399 if (!*n
|| strcmp((*n
)->name
, name
)) {
405 mutex_unlock(&block_class_lock
);
409 EXPORT_SYMBOL(unregister_blkdev
);
411 static struct kobj_map
*bdev_map
;
414 * blk_mangle_minor - scatter minor numbers apart
415 * @minor: minor number to mangle
417 * Scatter consecutively allocated @minor number apart if MANGLE_DEVT
418 * is enabled. Mangling twice gives the original value.
426 static int blk_mangle_minor(int minor
)
428 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
431 for (i
= 0; i
< MINORBITS
/ 2; i
++) {
432 int low
= minor
& (1 << i
);
433 int high
= minor
& (1 << (MINORBITS
- 1 - i
));
434 int distance
= MINORBITS
- 1 - 2 * i
;
436 minor
^= low
| high
; /* clear both bits */
437 low
<<= distance
; /* swap the positions */
439 minor
|= low
| high
; /* and set */
446 * blk_alloc_devt - allocate a dev_t for a partition
447 * @part: partition to allocate dev_t for
448 * @devt: out parameter for resulting dev_t
450 * Allocate a dev_t for block device.
453 * 0 on success, allocated dev_t is returned in *@devt. -errno on
459 int blk_alloc_devt(struct hd_struct
*part
, dev_t
*devt
)
461 struct gendisk
*disk
= part_to_disk(part
);
464 /* in consecutive minor range? */
465 if (part
->partno
< disk
->minors
) {
466 *devt
= MKDEV(disk
->major
, disk
->first_minor
+ part
->partno
);
470 /* allocate ext devt */
471 idr_preload(GFP_KERNEL
);
473 spin_lock_bh(&ext_devt_lock
);
474 idx
= idr_alloc(&ext_devt_idr
, part
, 0, NR_EXT_DEVT
, GFP_NOWAIT
);
475 spin_unlock_bh(&ext_devt_lock
);
479 return idx
== -ENOSPC
? -EBUSY
: idx
;
481 *devt
= MKDEV(BLOCK_EXT_MAJOR
, blk_mangle_minor(idx
));
486 * blk_free_devt - free a dev_t
487 * @devt: dev_t to free
489 * Free @devt which was allocated using blk_alloc_devt().
494 void blk_free_devt(dev_t devt
)
496 if (devt
== MKDEV(0, 0))
499 if (MAJOR(devt
) == BLOCK_EXT_MAJOR
) {
500 spin_lock_bh(&ext_devt_lock
);
501 idr_remove(&ext_devt_idr
, blk_mangle_minor(MINOR(devt
)));
502 spin_unlock_bh(&ext_devt_lock
);
506 static char *bdevt_str(dev_t devt
, char *buf
)
508 if (MAJOR(devt
) <= 0xff && MINOR(devt
) <= 0xff) {
509 char tbuf
[BDEVT_SIZE
];
510 snprintf(tbuf
, BDEVT_SIZE
, "%02x%02x", MAJOR(devt
), MINOR(devt
));
511 snprintf(buf
, BDEVT_SIZE
, "%-9s", tbuf
);
513 snprintf(buf
, BDEVT_SIZE
, "%03x:%05x", MAJOR(devt
), MINOR(devt
));
519 * Register device numbers dev..(dev+range-1)
520 * range must be nonzero
521 * The hash chain is sorted on range, so that subranges can override.
523 void blk_register_region(dev_t devt
, unsigned long range
, struct module
*module
,
524 struct kobject
*(*probe
)(dev_t
, int *, void *),
525 int (*lock
)(dev_t
, void *), void *data
)
527 kobj_map(bdev_map
, devt
, range
, module
, probe
, lock
, data
);
530 EXPORT_SYMBOL(blk_register_region
);
532 void blk_unregister_region(dev_t devt
, unsigned long range
)
534 kobj_unmap(bdev_map
, devt
, range
);
537 EXPORT_SYMBOL(blk_unregister_region
);
539 static struct kobject
*exact_match(dev_t devt
, int *partno
, void *data
)
541 struct gendisk
*p
= data
;
543 return &disk_to_dev(p
)->kobj
;
546 static int exact_lock(dev_t devt
, void *data
)
548 struct gendisk
*p
= data
;
555 static void register_disk(struct device
*parent
, struct gendisk
*disk
)
557 struct device
*ddev
= disk_to_dev(disk
);
558 struct block_device
*bdev
;
559 struct disk_part_iter piter
;
560 struct hd_struct
*part
;
563 ddev
->parent
= parent
;
565 dev_set_name(ddev
, "%s", disk
->disk_name
);
567 /* delay uevents, until we scanned partition table */
568 dev_set_uevent_suppress(ddev
, 1);
570 if (device_add(ddev
))
572 if (!sysfs_deprecated
) {
573 err
= sysfs_create_link(block_depr
, &ddev
->kobj
,
574 kobject_name(&ddev
->kobj
));
582 * avoid probable deadlock caused by allocating memory with
583 * GFP_KERNEL in runtime_resume callback of its all ancestor
586 pm_runtime_set_memalloc_noio(ddev
, true);
588 disk
->part0
.holder_dir
= kobject_create_and_add("holders", &ddev
->kobj
);
589 disk
->slave_dir
= kobject_create_and_add("slaves", &ddev
->kobj
);
591 /* No minors to use for partitions */
592 if (!disk_part_scan_enabled(disk
))
595 /* No such device (e.g., media were just removed) */
596 if (!get_capacity(disk
))
599 bdev
= bdget_disk(disk
, 0);
603 bdev
->bd_invalidated
= 1;
604 err
= blkdev_get(bdev
, FMODE_READ
, NULL
);
607 blkdev_put(bdev
, FMODE_READ
);
610 /* announce disk after possible partitions are created */
611 dev_set_uevent_suppress(ddev
, 0);
612 kobject_uevent(&ddev
->kobj
, KOBJ_ADD
);
614 /* announce possible partitions */
615 disk_part_iter_init(&piter
, disk
, 0);
616 while ((part
= disk_part_iter_next(&piter
)))
617 kobject_uevent(&part_to_dev(part
)->kobj
, KOBJ_ADD
);
618 disk_part_iter_exit(&piter
);
622 * device_add_disk - add partitioning information to kernel list
623 * @parent: parent device for the disk
624 * @disk: per-device partitioning information
626 * This function registers the partitioning information in @disk
629 * FIXME: error handling
631 void device_add_disk(struct device
*parent
, struct gendisk
*disk
)
633 struct backing_dev_info
*bdi
;
637 /* minors == 0 indicates to use ext devt from part0 and should
638 * be accompanied with EXT_DEVT flag. Make sure all
639 * parameters make sense.
641 WARN_ON(disk
->minors
&& !(disk
->major
|| disk
->first_minor
));
642 WARN_ON(!disk
->minors
&& !(disk
->flags
& GENHD_FL_EXT_DEVT
));
644 disk
->flags
|= GENHD_FL_UP
;
646 retval
= blk_alloc_devt(&disk
->part0
, &devt
);
651 disk_to_dev(disk
)->devt
= devt
;
652 disk
->major
= MAJOR(devt
);
653 disk
->first_minor
= MINOR(devt
);
655 disk_alloc_events(disk
);
657 /* Register BDI before referencing it from bdev */
658 bdi
= disk
->queue
->backing_dev_info
;
659 bdi_register_owner(bdi
, disk_to_dev(disk
));
661 blk_register_region(disk_devt(disk
), disk
->minors
, NULL
,
662 exact_match
, exact_lock
, disk
);
663 register_disk(parent
, disk
);
664 blk_register_queue(disk
);
667 * Take an extra ref on queue which will be put on disk_release()
668 * so that it sticks around as long as @disk is there.
670 WARN_ON_ONCE(!blk_get_queue(disk
->queue
));
672 retval
= sysfs_create_link(&disk_to_dev(disk
)->kobj
, &bdi
->dev
->kobj
,
676 disk_add_events(disk
);
677 blk_integrity_add(disk
);
679 EXPORT_SYMBOL(device_add_disk
);
681 void del_gendisk(struct gendisk
*disk
)
683 struct disk_part_iter piter
;
684 struct hd_struct
*part
;
686 blk_integrity_del(disk
);
687 disk_del_events(disk
);
689 /* invalidate stuff */
690 disk_part_iter_init(&piter
, disk
,
691 DISK_PITER_INCL_EMPTY
| DISK_PITER_REVERSE
);
692 while ((part
= disk_part_iter_next(&piter
))) {
693 invalidate_partition(disk
, part
->partno
);
694 bdev_unhash_inode(part_devt(part
));
695 delete_partition(disk
, part
->partno
);
697 disk_part_iter_exit(&piter
);
699 invalidate_partition(disk
, 0);
700 bdev_unhash_inode(disk_devt(disk
));
701 set_capacity(disk
, 0);
702 disk
->flags
&= ~GENHD_FL_UP
;
704 sysfs_remove_link(&disk_to_dev(disk
)->kobj
, "bdi");
707 * Unregister bdi before releasing device numbers (as they can
708 * get reused and we'd get clashes in sysfs).
710 bdi_unregister(disk
->queue
->backing_dev_info
);
711 blk_unregister_queue(disk
);
715 blk_unregister_region(disk_devt(disk
), disk
->minors
);
717 part_stat_set_all(&disk
->part0
, 0);
718 disk
->part0
.stamp
= 0;
720 kobject_put(disk
->part0
.holder_dir
);
721 kobject_put(disk
->slave_dir
);
722 if (!sysfs_deprecated
)
723 sysfs_remove_link(block_depr
, dev_name(disk_to_dev(disk
)));
724 pm_runtime_set_memalloc_noio(disk_to_dev(disk
), false);
725 device_del(disk_to_dev(disk
));
727 EXPORT_SYMBOL(del_gendisk
);
729 /* sysfs access to bad-blocks list. */
730 static ssize_t
disk_badblocks_show(struct device
*dev
,
731 struct device_attribute
*attr
,
734 struct gendisk
*disk
= dev_to_disk(dev
);
737 return sprintf(page
, "\n");
739 return badblocks_show(disk
->bb
, page
, 0);
742 static ssize_t
disk_badblocks_store(struct device
*dev
,
743 struct device_attribute
*attr
,
744 const char *page
, size_t len
)
746 struct gendisk
*disk
= dev_to_disk(dev
);
751 return badblocks_store(disk
->bb
, page
, len
, 0);
755 * get_gendisk - get partitioning information for a given device
756 * @devt: device to get partitioning information for
757 * @partno: returned partition index
759 * This function gets the structure containing partitioning
760 * information for the given device @devt.
762 struct gendisk
*get_gendisk(dev_t devt
, int *partno
)
764 struct gendisk
*disk
= NULL
;
766 if (MAJOR(devt
) != BLOCK_EXT_MAJOR
) {
767 struct kobject
*kobj
;
769 kobj
= kobj_lookup(bdev_map
, devt
, partno
);
771 disk
= dev_to_disk(kobj_to_dev(kobj
));
773 struct hd_struct
*part
;
775 spin_lock_bh(&ext_devt_lock
);
776 part
= idr_find(&ext_devt_idr
, blk_mangle_minor(MINOR(devt
)));
777 if (part
&& get_disk(part_to_disk(part
))) {
778 *partno
= part
->partno
;
779 disk
= part_to_disk(part
);
781 spin_unlock_bh(&ext_devt_lock
);
786 EXPORT_SYMBOL(get_gendisk
);
789 * bdget_disk - do bdget() by gendisk and partition number
790 * @disk: gendisk of interest
791 * @partno: partition number
793 * Find partition @partno from @disk, do bdget() on it.
799 * Resulting block_device on success, NULL on failure.
801 struct block_device
*bdget_disk(struct gendisk
*disk
, int partno
)
803 struct hd_struct
*part
;
804 struct block_device
*bdev
= NULL
;
806 part
= disk_get_part(disk
, partno
);
808 bdev
= bdget(part_devt(part
));
813 EXPORT_SYMBOL(bdget_disk
);
816 * print a full list of all partitions - intended for places where the root
817 * filesystem can't be mounted and thus to give the victim some idea of what
820 void __init
printk_all_partitions(void)
822 struct class_dev_iter iter
;
825 class_dev_iter_init(&iter
, &block_class
, NULL
, &disk_type
);
826 while ((dev
= class_dev_iter_next(&iter
))) {
827 struct gendisk
*disk
= dev_to_disk(dev
);
828 struct disk_part_iter piter
;
829 struct hd_struct
*part
;
830 char name_buf
[BDEVNAME_SIZE
];
831 char devt_buf
[BDEVT_SIZE
];
834 * Don't show empty devices or things that have been
837 if (get_capacity(disk
) == 0 ||
838 (disk
->flags
& GENHD_FL_SUPPRESS_PARTITION_INFO
))
842 * Note, unlike /proc/partitions, I am showing the
843 * numbers in hex - the same format as the root=
846 disk_part_iter_init(&piter
, disk
, DISK_PITER_INCL_PART0
);
847 while ((part
= disk_part_iter_next(&piter
))) {
848 bool is_part0
= part
== &disk
->part0
;
850 printk("%s%s %10llu %s %s", is_part0
? "" : " ",
851 bdevt_str(part_devt(part
), devt_buf
),
852 (unsigned long long)part_nr_sects_read(part
) >> 1
853 , disk_name(disk
, part
->partno
, name_buf
),
854 part
->info
? part
->info
->uuid
: "");
856 if (dev
->parent
&& dev
->parent
->driver
)
857 printk(" driver: %s\n",
858 dev
->parent
->driver
->name
);
860 printk(" (driver?)\n");
864 disk_part_iter_exit(&piter
);
866 class_dev_iter_exit(&iter
);
869 #ifdef CONFIG_PROC_FS
871 static void *disk_seqf_start(struct seq_file
*seqf
, loff_t
*pos
)
874 struct class_dev_iter
*iter
;
877 iter
= kmalloc(sizeof(*iter
), GFP_KERNEL
);
879 return ERR_PTR(-ENOMEM
);
881 seqf
->private = iter
;
882 class_dev_iter_init(iter
, &block_class
, NULL
, &disk_type
);
884 dev
= class_dev_iter_next(iter
);
889 return dev_to_disk(dev
);
892 static void *disk_seqf_next(struct seq_file
*seqf
, void *v
, loff_t
*pos
)
897 dev
= class_dev_iter_next(seqf
->private);
899 return dev_to_disk(dev
);
904 static void disk_seqf_stop(struct seq_file
*seqf
, void *v
)
906 struct class_dev_iter
*iter
= seqf
->private;
908 /* stop is called even after start failed :-( */
910 class_dev_iter_exit(iter
);
912 seqf
->private = NULL
;
916 static void *show_partition_start(struct seq_file
*seqf
, loff_t
*pos
)
920 p
= disk_seqf_start(seqf
, pos
);
921 if (!IS_ERR_OR_NULL(p
) && !*pos
)
922 seq_puts(seqf
, "major minor #blocks name\n\n");
926 static int show_partition(struct seq_file
*seqf
, void *v
)
928 struct gendisk
*sgp
= v
;
929 struct disk_part_iter piter
;
930 struct hd_struct
*part
;
931 char buf
[BDEVNAME_SIZE
];
933 /* Don't show non-partitionable removeable devices or empty devices */
934 if (!get_capacity(sgp
) || (!disk_max_parts(sgp
) &&
935 (sgp
->flags
& GENHD_FL_REMOVABLE
)))
937 if (sgp
->flags
& GENHD_FL_SUPPRESS_PARTITION_INFO
)
940 /* show the full disk and all non-0 size partitions of it */
941 disk_part_iter_init(&piter
, sgp
, DISK_PITER_INCL_PART0
);
942 while ((part
= disk_part_iter_next(&piter
)))
943 seq_printf(seqf
, "%4d %7d %10llu %s\n",
944 MAJOR(part_devt(part
)), MINOR(part_devt(part
)),
945 (unsigned long long)part_nr_sects_read(part
) >> 1,
946 disk_name(sgp
, part
->partno
, buf
));
947 disk_part_iter_exit(&piter
);
952 static const struct seq_operations partitions_op
= {
953 .start
= show_partition_start
,
954 .next
= disk_seqf_next
,
955 .stop
= disk_seqf_stop
,
956 .show
= show_partition
959 static int partitions_open(struct inode
*inode
, struct file
*file
)
961 return seq_open(file
, &partitions_op
);
964 static const struct file_operations proc_partitions_operations
= {
965 .open
= partitions_open
,
968 .release
= seq_release
,
973 static struct kobject
*base_probe(dev_t devt
, int *partno
, void *data
)
975 if (request_module("block-major-%d-%d", MAJOR(devt
), MINOR(devt
)) > 0)
976 /* Make old-style 2.4 aliases work */
977 request_module("block-major-%d", MAJOR(devt
));
981 static int __init
genhd_device_init(void)
985 block_class
.dev_kobj
= sysfs_dev_block_kobj
;
986 error
= class_register(&block_class
);
989 bdev_map
= kobj_map_init(base_probe
, &block_class_lock
);
992 register_blkdev(BLOCK_EXT_MAJOR
, "blkext");
994 /* create top-level block dir */
995 if (!sysfs_deprecated
)
996 block_depr
= kobject_create_and_add("block", NULL
);
1000 subsys_initcall(genhd_device_init
);
1002 static ssize_t
disk_range_show(struct device
*dev
,
1003 struct device_attribute
*attr
, char *buf
)
1005 struct gendisk
*disk
= dev_to_disk(dev
);
1007 return sprintf(buf
, "%d\n", disk
->minors
);
1010 static ssize_t
disk_ext_range_show(struct device
*dev
,
1011 struct device_attribute
*attr
, char *buf
)
1013 struct gendisk
*disk
= dev_to_disk(dev
);
1015 return sprintf(buf
, "%d\n", disk_max_parts(disk
));
1018 static ssize_t
disk_removable_show(struct device
*dev
,
1019 struct device_attribute
*attr
, char *buf
)
1021 struct gendisk
*disk
= dev_to_disk(dev
);
1023 return sprintf(buf
, "%d\n",
1024 (disk
->flags
& GENHD_FL_REMOVABLE
? 1 : 0));
1027 static ssize_t
disk_ro_show(struct device
*dev
,
1028 struct device_attribute
*attr
, char *buf
)
1030 struct gendisk
*disk
= dev_to_disk(dev
);
1032 return sprintf(buf
, "%d\n", get_disk_ro(disk
) ? 1 : 0);
1035 static ssize_t
disk_capability_show(struct device
*dev
,
1036 struct device_attribute
*attr
, char *buf
)
1038 struct gendisk
*disk
= dev_to_disk(dev
);
1040 return sprintf(buf
, "%x\n", disk
->flags
);
1043 static ssize_t
disk_alignment_offset_show(struct device
*dev
,
1044 struct device_attribute
*attr
,
1047 struct gendisk
*disk
= dev_to_disk(dev
);
1049 return sprintf(buf
, "%d\n", queue_alignment_offset(disk
->queue
));
1052 static ssize_t
disk_discard_alignment_show(struct device
*dev
,
1053 struct device_attribute
*attr
,
1056 struct gendisk
*disk
= dev_to_disk(dev
);
1058 return sprintf(buf
, "%d\n", queue_discard_alignment(disk
->queue
));
1061 static DEVICE_ATTR(range
, S_IRUGO
, disk_range_show
, NULL
);
1062 static DEVICE_ATTR(ext_range
, S_IRUGO
, disk_ext_range_show
, NULL
);
1063 static DEVICE_ATTR(removable
, S_IRUGO
, disk_removable_show
, NULL
);
1064 static DEVICE_ATTR(ro
, S_IRUGO
, disk_ro_show
, NULL
);
1065 static DEVICE_ATTR(size
, S_IRUGO
, part_size_show
, NULL
);
1066 static DEVICE_ATTR(alignment_offset
, S_IRUGO
, disk_alignment_offset_show
, NULL
);
1067 static DEVICE_ATTR(discard_alignment
, S_IRUGO
, disk_discard_alignment_show
,
1069 static DEVICE_ATTR(capability
, S_IRUGO
, disk_capability_show
, NULL
);
1070 static DEVICE_ATTR(stat
, S_IRUGO
, part_stat_show
, NULL
);
1071 static DEVICE_ATTR(inflight
, S_IRUGO
, part_inflight_show
, NULL
);
1072 static DEVICE_ATTR(badblocks
, S_IRUGO
| S_IWUSR
, disk_badblocks_show
,
1073 disk_badblocks_store
);
1074 #ifdef CONFIG_FAIL_MAKE_REQUEST
1075 static struct device_attribute dev_attr_fail
=
1076 __ATTR(make
-it
-fail
, S_IRUGO
|S_IWUSR
, part_fail_show
, part_fail_store
);
1078 #ifdef CONFIG_FAIL_IO_TIMEOUT
1079 static struct device_attribute dev_attr_fail_timeout
=
1080 __ATTR(io
-timeout
-fail
, S_IRUGO
|S_IWUSR
, part_timeout_show
,
1081 part_timeout_store
);
1084 static struct attribute
*disk_attrs
[] = {
1085 &dev_attr_range
.attr
,
1086 &dev_attr_ext_range
.attr
,
1087 &dev_attr_removable
.attr
,
1089 &dev_attr_size
.attr
,
1090 &dev_attr_alignment_offset
.attr
,
1091 &dev_attr_discard_alignment
.attr
,
1092 &dev_attr_capability
.attr
,
1093 &dev_attr_stat
.attr
,
1094 &dev_attr_inflight
.attr
,
1095 &dev_attr_badblocks
.attr
,
1096 #ifdef CONFIG_FAIL_MAKE_REQUEST
1097 &dev_attr_fail
.attr
,
1099 #ifdef CONFIG_FAIL_IO_TIMEOUT
1100 &dev_attr_fail_timeout
.attr
,
1105 static umode_t
disk_visible(struct kobject
*kobj
, struct attribute
*a
, int n
)
1107 struct device
*dev
= container_of(kobj
, typeof(*dev
), kobj
);
1108 struct gendisk
*disk
= dev_to_disk(dev
);
1110 if (a
== &dev_attr_badblocks
.attr
&& !disk
->bb
)
1115 static struct attribute_group disk_attr_group
= {
1116 .attrs
= disk_attrs
,
1117 .is_visible
= disk_visible
,
1120 static const struct attribute_group
*disk_attr_groups
[] = {
1126 * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way
1127 * @disk: disk to replace part_tbl for
1128 * @new_ptbl: new part_tbl to install
1130 * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The
1131 * original ptbl is freed using RCU callback.
1134 * Matching bd_mutex locked or the caller is the only user of @disk.
1136 static void disk_replace_part_tbl(struct gendisk
*disk
,
1137 struct disk_part_tbl
*new_ptbl
)
1139 struct disk_part_tbl
*old_ptbl
=
1140 rcu_dereference_protected(disk
->part_tbl
, 1);
1142 rcu_assign_pointer(disk
->part_tbl
, new_ptbl
);
1145 rcu_assign_pointer(old_ptbl
->last_lookup
, NULL
);
1146 kfree_rcu(old_ptbl
, rcu_head
);
1151 * disk_expand_part_tbl - expand disk->part_tbl
1152 * @disk: disk to expand part_tbl for
1153 * @partno: expand such that this partno can fit in
1155 * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl
1156 * uses RCU to allow unlocked dereferencing for stats and other stuff.
1159 * Matching bd_mutex locked or the caller is the only user of @disk.
1163 * 0 on success, -errno on failure.
1165 int disk_expand_part_tbl(struct gendisk
*disk
, int partno
)
1167 struct disk_part_tbl
*old_ptbl
=
1168 rcu_dereference_protected(disk
->part_tbl
, 1);
1169 struct disk_part_tbl
*new_ptbl
;
1170 int len
= old_ptbl
? old_ptbl
->len
: 0;
1175 * check for int overflow, since we can get here from blkpg_ioctl()
1176 * with a user passed 'partno'.
1178 target
= partno
+ 1;
1182 /* disk_max_parts() is zero during initialization, ignore if so */
1183 if (disk_max_parts(disk
) && target
> disk_max_parts(disk
))
1189 size
= sizeof(*new_ptbl
) + target
* sizeof(new_ptbl
->part
[0]);
1190 new_ptbl
= kzalloc_node(size
, GFP_KERNEL
, disk
->node_id
);
1194 new_ptbl
->len
= target
;
1196 for (i
= 0; i
< len
; i
++)
1197 rcu_assign_pointer(new_ptbl
->part
[i
], old_ptbl
->part
[i
]);
1199 disk_replace_part_tbl(disk
, new_ptbl
);
1203 static void disk_release(struct device
*dev
)
1205 struct gendisk
*disk
= dev_to_disk(dev
);
1207 blk_free_devt(dev
->devt
);
1208 disk_release_events(disk
);
1209 kfree(disk
->random
);
1210 disk_replace_part_tbl(disk
, NULL
);
1211 hd_free_part(&disk
->part0
);
1213 blk_put_queue(disk
->queue
);
1216 struct class block_class
= {
1220 static char *block_devnode(struct device
*dev
, umode_t
*mode
,
1221 kuid_t
*uid
, kgid_t
*gid
)
1223 struct gendisk
*disk
= dev_to_disk(dev
);
1226 return disk
->devnode(disk
, mode
);
1230 static const struct device_type disk_type
= {
1232 .groups
= disk_attr_groups
,
1233 .release
= disk_release
,
1234 .devnode
= block_devnode
,
1237 #ifdef CONFIG_PROC_FS
1239 * aggregate disk stat collector. Uses the same stats that the sysfs
1240 * entries do, above, but makes them available through one seq_file.
1242 * The output looks suspiciously like /proc/partitions with a bunch of
1245 static int diskstats_show(struct seq_file
*seqf
, void *v
)
1247 struct gendisk
*gp
= v
;
1248 struct disk_part_iter piter
;
1249 struct hd_struct
*hd
;
1250 char buf
[BDEVNAME_SIZE
];
1251 unsigned int inflight
[2];
1255 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1256 seq_puts(seqf, "major minor name"
1257 " rio rmerge rsect ruse wio wmerge "
1258 "wsect wuse running use aveq"
1262 disk_part_iter_init(&piter
, gp
, DISK_PITER_INCL_EMPTY_PART0
);
1263 while ((hd
= disk_part_iter_next(&piter
))) {
1264 cpu
= part_stat_lock();
1265 part_round_stats(gp
->queue
, cpu
, hd
);
1267 part_in_flight(gp
->queue
, hd
, inflight
);
1268 seq_printf(seqf
, "%4d %7d %s %lu %lu %lu "
1269 "%u %lu %lu %lu %u %u %u %u\n",
1270 MAJOR(part_devt(hd
)), MINOR(part_devt(hd
)),
1271 disk_name(gp
, hd
->partno
, buf
),
1272 part_stat_read(hd
, ios
[READ
]),
1273 part_stat_read(hd
, merges
[READ
]),
1274 part_stat_read(hd
, sectors
[READ
]),
1275 jiffies_to_msecs(part_stat_read(hd
, ticks
[READ
])),
1276 part_stat_read(hd
, ios
[WRITE
]),
1277 part_stat_read(hd
, merges
[WRITE
]),
1278 part_stat_read(hd
, sectors
[WRITE
]),
1279 jiffies_to_msecs(part_stat_read(hd
, ticks
[WRITE
])),
1281 jiffies_to_msecs(part_stat_read(hd
, io_ticks
)),
1282 jiffies_to_msecs(part_stat_read(hd
, time_in_queue
))
1285 disk_part_iter_exit(&piter
);
1290 static const struct seq_operations diskstats_op
= {
1291 .start
= disk_seqf_start
,
1292 .next
= disk_seqf_next
,
1293 .stop
= disk_seqf_stop
,
1294 .show
= diskstats_show
1297 static int diskstats_open(struct inode
*inode
, struct file
*file
)
1299 return seq_open(file
, &diskstats_op
);
1302 static const struct file_operations proc_diskstats_operations
= {
1303 .open
= diskstats_open
,
1305 .llseek
= seq_lseek
,
1306 .release
= seq_release
,
1309 static int __init
proc_genhd_init(void)
1311 proc_create("diskstats", 0, NULL
, &proc_diskstats_operations
);
1312 proc_create("partitions", 0, NULL
, &proc_partitions_operations
);
1315 module_init(proc_genhd_init
);
1316 #endif /* CONFIG_PROC_FS */
1318 dev_t
blk_lookup_devt(const char *name
, int partno
)
1320 dev_t devt
= MKDEV(0, 0);
1321 struct class_dev_iter iter
;
1324 class_dev_iter_init(&iter
, &block_class
, NULL
, &disk_type
);
1325 while ((dev
= class_dev_iter_next(&iter
))) {
1326 struct gendisk
*disk
= dev_to_disk(dev
);
1327 struct hd_struct
*part
;
1329 if (strcmp(dev_name(dev
), name
))
1332 if (partno
< disk
->minors
) {
1333 /* We need to return the right devno, even
1334 * if the partition doesn't exist yet.
1336 devt
= MKDEV(MAJOR(dev
->devt
),
1337 MINOR(dev
->devt
) + partno
);
1340 part
= disk_get_part(disk
, partno
);
1342 devt
= part_devt(part
);
1343 disk_put_part(part
);
1346 disk_put_part(part
);
1348 class_dev_iter_exit(&iter
);
1351 EXPORT_SYMBOL(blk_lookup_devt
);
1353 struct gendisk
*alloc_disk(int minors
)
1355 return alloc_disk_node(minors
, NUMA_NO_NODE
);
1357 EXPORT_SYMBOL(alloc_disk
);
1359 struct gendisk
*alloc_disk_node(int minors
, int node_id
)
1361 struct gendisk
*disk
;
1362 struct disk_part_tbl
*ptbl
;
1364 if (minors
> DISK_MAX_PARTS
) {
1366 "block: can't allocated more than %d partitions\n",
1368 minors
= DISK_MAX_PARTS
;
1371 disk
= kzalloc_node(sizeof(struct gendisk
), GFP_KERNEL
, node_id
);
1373 if (!init_part_stats(&disk
->part0
)) {
1377 disk
->node_id
= node_id
;
1378 if (disk_expand_part_tbl(disk
, 0)) {
1379 free_part_stats(&disk
->part0
);
1383 ptbl
= rcu_dereference_protected(disk
->part_tbl
, 1);
1384 rcu_assign_pointer(ptbl
->part
[0], &disk
->part0
);
1387 * set_capacity() and get_capacity() currently don't use
1388 * seqcounter to read/update the part0->nr_sects. Still init
1389 * the counter as we can read the sectors in IO submission
1390 * patch using seqence counters.
1392 * TODO: Ideally set_capacity() and get_capacity() should be
1393 * converted to make use of bd_mutex and sequence counters.
1395 seqcount_init(&disk
->part0
.nr_sects_seq
);
1396 if (hd_ref_init(&disk
->part0
)) {
1397 hd_free_part(&disk
->part0
);
1402 disk
->minors
= minors
;
1403 rand_initialize_disk(disk
);
1404 disk_to_dev(disk
)->class = &block_class
;
1405 disk_to_dev(disk
)->type
= &disk_type
;
1406 device_initialize(disk_to_dev(disk
));
1410 EXPORT_SYMBOL(alloc_disk_node
);
1412 struct kobject
*get_disk(struct gendisk
*disk
)
1414 struct module
*owner
;
1415 struct kobject
*kobj
;
1419 owner
= disk
->fops
->owner
;
1420 if (owner
&& !try_module_get(owner
))
1422 kobj
= kobject_get_unless_zero(&disk_to_dev(disk
)->kobj
);
1431 EXPORT_SYMBOL(get_disk
);
1433 void put_disk(struct gendisk
*disk
)
1436 kobject_put(&disk_to_dev(disk
)->kobj
);
1439 EXPORT_SYMBOL(put_disk
);
1441 static void set_disk_ro_uevent(struct gendisk
*gd
, int ro
)
1443 char event
[] = "DISK_RO=1";
1444 char *envp
[] = { event
, NULL
};
1448 kobject_uevent_env(&disk_to_dev(gd
)->kobj
, KOBJ_CHANGE
, envp
);
1451 void set_device_ro(struct block_device
*bdev
, int flag
)
1453 bdev
->bd_part
->policy
= flag
;
1456 EXPORT_SYMBOL(set_device_ro
);
1458 void set_disk_ro(struct gendisk
*disk
, int flag
)
1460 struct disk_part_iter piter
;
1461 struct hd_struct
*part
;
1463 if (disk
->part0
.policy
!= flag
) {
1464 set_disk_ro_uevent(disk
, flag
);
1465 disk
->part0
.policy
= flag
;
1468 disk_part_iter_init(&piter
, disk
, DISK_PITER_INCL_EMPTY
);
1469 while ((part
= disk_part_iter_next(&piter
)))
1470 part
->policy
= flag
;
1471 disk_part_iter_exit(&piter
);
1474 EXPORT_SYMBOL(set_disk_ro
);
1476 int bdev_read_only(struct block_device
*bdev
)
1480 return bdev
->bd_part
->policy
;
1483 EXPORT_SYMBOL(bdev_read_only
);
1485 int invalidate_partition(struct gendisk
*disk
, int partno
)
1488 struct block_device
*bdev
= bdget_disk(disk
, partno
);
1491 res
= __invalidate_device(bdev
, true);
1497 EXPORT_SYMBOL(invalidate_partition
);
1500 * Disk events - monitor disk events like media change and eject request.
1502 struct disk_events
{
1503 struct list_head node
; /* all disk_event's */
1504 struct gendisk
*disk
; /* the associated disk */
1507 struct mutex block_mutex
; /* protects blocking */
1508 int block
; /* event blocking depth */
1509 unsigned int pending
; /* events already sent out */
1510 unsigned int clearing
; /* events being cleared */
1512 long poll_msecs
; /* interval, -1 for default */
1513 struct delayed_work dwork
;
1516 static const char *disk_events_strs
[] = {
1517 [ilog2(DISK_EVENT_MEDIA_CHANGE
)] = "media_change",
1518 [ilog2(DISK_EVENT_EJECT_REQUEST
)] = "eject_request",
1521 static char *disk_uevents
[] = {
1522 [ilog2(DISK_EVENT_MEDIA_CHANGE
)] = "DISK_MEDIA_CHANGE=1",
1523 [ilog2(DISK_EVENT_EJECT_REQUEST
)] = "DISK_EJECT_REQUEST=1",
1526 /* list of all disk_events */
1527 static DEFINE_MUTEX(disk_events_mutex
);
1528 static LIST_HEAD(disk_events
);
1530 /* disable in-kernel polling by default */
1531 static unsigned long disk_events_dfl_poll_msecs
;
1533 static unsigned long disk_events_poll_jiffies(struct gendisk
*disk
)
1535 struct disk_events
*ev
= disk
->ev
;
1536 long intv_msecs
= 0;
1539 * If device-specific poll interval is set, always use it. If
1540 * the default is being used, poll iff there are events which
1541 * can't be monitored asynchronously.
1543 if (ev
->poll_msecs
>= 0)
1544 intv_msecs
= ev
->poll_msecs
;
1545 else if (disk
->events
& ~disk
->async_events
)
1546 intv_msecs
= disk_events_dfl_poll_msecs
;
1548 return msecs_to_jiffies(intv_msecs
);
1552 * disk_block_events - block and flush disk event checking
1553 * @disk: disk to block events for
1555 * On return from this function, it is guaranteed that event checking
1556 * isn't in progress and won't happen until unblocked by
1557 * disk_unblock_events(). Events blocking is counted and the actual
1558 * unblocking happens after the matching number of unblocks are done.
1560 * Note that this intentionally does not block event checking from
1561 * disk_clear_events().
1566 void disk_block_events(struct gendisk
*disk
)
1568 struct disk_events
*ev
= disk
->ev
;
1569 unsigned long flags
;
1576 * Outer mutex ensures that the first blocker completes canceling
1577 * the event work before further blockers are allowed to finish.
1579 mutex_lock(&ev
->block_mutex
);
1581 spin_lock_irqsave(&ev
->lock
, flags
);
1582 cancel
= !ev
->block
++;
1583 spin_unlock_irqrestore(&ev
->lock
, flags
);
1586 cancel_delayed_work_sync(&disk
->ev
->dwork
);
1588 mutex_unlock(&ev
->block_mutex
);
1591 static void __disk_unblock_events(struct gendisk
*disk
, bool check_now
)
1593 struct disk_events
*ev
= disk
->ev
;
1595 unsigned long flags
;
1597 spin_lock_irqsave(&ev
->lock
, flags
);
1599 if (WARN_ON_ONCE(ev
->block
<= 0))
1605 intv
= disk_events_poll_jiffies(disk
);
1607 queue_delayed_work(system_freezable_power_efficient_wq
,
1610 queue_delayed_work(system_freezable_power_efficient_wq
,
1613 spin_unlock_irqrestore(&ev
->lock
, flags
);
1617 * disk_unblock_events - unblock disk event checking
1618 * @disk: disk to unblock events for
1620 * Undo disk_block_events(). When the block count reaches zero, it
1621 * starts events polling if configured.
1624 * Don't care. Safe to call from irq context.
1626 void disk_unblock_events(struct gendisk
*disk
)
1629 __disk_unblock_events(disk
, false);
1633 * disk_flush_events - schedule immediate event checking and flushing
1634 * @disk: disk to check and flush events for
1635 * @mask: events to flush
1637 * Schedule immediate event checking on @disk if not blocked. Events in
1638 * @mask are scheduled to be cleared from the driver. Note that this
1639 * doesn't clear the events from @disk->ev.
1642 * If @mask is non-zero must be called with bdev->bd_mutex held.
1644 void disk_flush_events(struct gendisk
*disk
, unsigned int mask
)
1646 struct disk_events
*ev
= disk
->ev
;
1651 spin_lock_irq(&ev
->lock
);
1652 ev
->clearing
|= mask
;
1654 mod_delayed_work(system_freezable_power_efficient_wq
,
1656 spin_unlock_irq(&ev
->lock
);
1660 * disk_clear_events - synchronously check, clear and return pending events
1661 * @disk: disk to fetch and clear events from
1662 * @mask: mask of events to be fetched and cleared
1664 * Disk events are synchronously checked and pending events in @mask
1665 * are cleared and returned. This ignores the block count.
1670 unsigned int disk_clear_events(struct gendisk
*disk
, unsigned int mask
)
1672 const struct block_device_operations
*bdops
= disk
->fops
;
1673 struct disk_events
*ev
= disk
->ev
;
1674 unsigned int pending
;
1675 unsigned int clearing
= mask
;
1678 /* for drivers still using the old ->media_changed method */
1679 if ((mask
& DISK_EVENT_MEDIA_CHANGE
) &&
1680 bdops
->media_changed
&& bdops
->media_changed(disk
))
1681 return DISK_EVENT_MEDIA_CHANGE
;
1685 disk_block_events(disk
);
1688 * store the union of mask and ev->clearing on the stack so that the
1689 * race with disk_flush_events does not cause ambiguity (ev->clearing
1690 * can still be modified even if events are blocked).
1692 spin_lock_irq(&ev
->lock
);
1693 clearing
|= ev
->clearing
;
1695 spin_unlock_irq(&ev
->lock
);
1697 disk_check_events(ev
, &clearing
);
1699 * if ev->clearing is not 0, the disk_flush_events got called in the
1700 * middle of this function, so we want to run the workfn without delay.
1702 __disk_unblock_events(disk
, ev
->clearing
? true : false);
1704 /* then, fetch and clear pending events */
1705 spin_lock_irq(&ev
->lock
);
1706 pending
= ev
->pending
& mask
;
1707 ev
->pending
&= ~mask
;
1708 spin_unlock_irq(&ev
->lock
);
1709 WARN_ON_ONCE(clearing
& mask
);
1715 * Separate this part out so that a different pointer for clearing_ptr can be
1716 * passed in for disk_clear_events.
1718 static void disk_events_workfn(struct work_struct
*work
)
1720 struct delayed_work
*dwork
= to_delayed_work(work
);
1721 struct disk_events
*ev
= container_of(dwork
, struct disk_events
, dwork
);
1723 disk_check_events(ev
, &ev
->clearing
);
1726 static void disk_check_events(struct disk_events
*ev
,
1727 unsigned int *clearing_ptr
)
1729 struct gendisk
*disk
= ev
->disk
;
1730 char *envp
[ARRAY_SIZE(disk_uevents
) + 1] = { };
1731 unsigned int clearing
= *clearing_ptr
;
1732 unsigned int events
;
1734 int nr_events
= 0, i
;
1737 events
= disk
->fops
->check_events(disk
, clearing
);
1739 /* accumulate pending events and schedule next poll if necessary */
1740 spin_lock_irq(&ev
->lock
);
1742 events
&= ~ev
->pending
;
1743 ev
->pending
|= events
;
1744 *clearing_ptr
&= ~clearing
;
1746 intv
= disk_events_poll_jiffies(disk
);
1747 if (!ev
->block
&& intv
)
1748 queue_delayed_work(system_freezable_power_efficient_wq
,
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);