1 // SPDX-License-Identifier: GPL-2.0
5 * Portions Copyright (C) 2020 Christoph Hellwig
8 #include <linux/module.h>
9 #include <linux/ctype.h>
11 #include <linux/genhd.h>
12 #include <linux/kdev_t.h>
13 #include <linux/kernel.h>
14 #include <linux/blkdev.h>
15 #include <linux/backing-dev.h>
16 #include <linux/init.h>
17 #include <linux/spinlock.h>
18 #include <linux/proc_fs.h>
19 #include <linux/seq_file.h>
20 #include <linux/slab.h>
21 #include <linux/kmod.h>
22 #include <linux/mutex.h>
23 #include <linux/idr.h>
24 #include <linux/log2.h>
25 #include <linux/pm_runtime.h>
26 #include <linux/badblocks.h>
30 static struct kobject
*block_depr
;
33 * Unique, monotonically increasing sequential number associated with block
34 * devices instances (i.e. incremented each time a device is attached).
35 * Associating uevents with block devices in userspace is difficult and racy:
36 * the uevent netlink socket is lossy, and on slow and overloaded systems has
37 * a very high latency.
38 * Block devices do not have exclusive owners in userspace, any process can set
39 * one up (e.g. loop devices). Moreover, device names can be reused (e.g. loop0
40 * can be reused again and again).
41 * A userspace process setting up a block device and watching for its events
42 * cannot thus reliably tell whether an event relates to the device it just set
43 * up or another earlier instance with the same name.
44 * This sequential number allows userspace processes to solve this problem, and
45 * uniquely associate an uevent to the lifetime to a device.
47 static atomic64_t diskseq
;
49 /* for extended dynamic devt allocation, currently only one major is used */
50 #define NR_EXT_DEVT (1 << MINORBITS)
51 static DEFINE_IDA(ext_devt_ida
);
53 void set_capacity(struct gendisk
*disk
, sector_t sectors
)
55 struct block_device
*bdev
= disk
->part0
;
57 spin_lock(&bdev
->bd_size_lock
);
58 i_size_write(bdev
->bd_inode
, (loff_t
)sectors
<< SECTOR_SHIFT
);
59 spin_unlock(&bdev
->bd_size_lock
);
61 EXPORT_SYMBOL(set_capacity
);
64 * Set disk capacity and notify if the size is not currently zero and will not
65 * be set to zero. Returns true if a uevent was sent, otherwise false.
67 bool set_capacity_and_notify(struct gendisk
*disk
, sector_t size
)
69 sector_t capacity
= get_capacity(disk
);
70 char *envp
[] = { "RESIZE=1", NULL
};
72 set_capacity(disk
, size
);
75 * Only print a message and send a uevent if the gendisk is user visible
76 * and alive. This avoids spamming the log and udev when setting the
77 * initial capacity during probing.
79 if (size
== capacity
||
81 (disk
->flags
& GENHD_FL_HIDDEN
))
84 pr_info("%s: detected capacity change from %lld to %lld\n",
85 disk
->disk_name
, capacity
, size
);
88 * Historically we did not send a uevent for changes to/from an empty
91 if (!capacity
|| !size
)
93 kobject_uevent_env(&disk_to_dev(disk
)->kobj
, KOBJ_CHANGE
, envp
);
96 EXPORT_SYMBOL_GPL(set_capacity_and_notify
);
99 * Format the device name of the indicated block device into the supplied buffer
100 * and return a pointer to that same buffer for convenience.
102 * Note: do not use this in new code, use the %pg specifier to sprintf and
105 const char *bdevname(struct block_device
*bdev
, char *buf
)
107 struct gendisk
*hd
= bdev
->bd_disk
;
108 int partno
= bdev
->bd_partno
;
111 snprintf(buf
, BDEVNAME_SIZE
, "%s", hd
->disk_name
);
112 else if (isdigit(hd
->disk_name
[strlen(hd
->disk_name
)-1]))
113 snprintf(buf
, BDEVNAME_SIZE
, "%sp%d", hd
->disk_name
, partno
);
115 snprintf(buf
, BDEVNAME_SIZE
, "%s%d", hd
->disk_name
, partno
);
119 EXPORT_SYMBOL(bdevname
);
121 static void part_stat_read_all(struct block_device
*part
,
122 struct disk_stats
*stat
)
126 memset(stat
, 0, sizeof(struct disk_stats
));
127 for_each_possible_cpu(cpu
) {
128 struct disk_stats
*ptr
= per_cpu_ptr(part
->bd_stats
, cpu
);
131 for (group
= 0; group
< NR_STAT_GROUPS
; group
++) {
132 stat
->nsecs
[group
] += ptr
->nsecs
[group
];
133 stat
->sectors
[group
] += ptr
->sectors
[group
];
134 stat
->ios
[group
] += ptr
->ios
[group
];
135 stat
->merges
[group
] += ptr
->merges
[group
];
138 stat
->io_ticks
+= ptr
->io_ticks
;
142 static unsigned int part_in_flight(struct block_device
*part
)
144 unsigned int inflight
= 0;
147 for_each_possible_cpu(cpu
) {
148 inflight
+= part_stat_local_read_cpu(part
, in_flight
[0], cpu
) +
149 part_stat_local_read_cpu(part
, in_flight
[1], cpu
);
151 if ((int)inflight
< 0)
157 static void part_in_flight_rw(struct block_device
*part
,
158 unsigned int inflight
[2])
164 for_each_possible_cpu(cpu
) {
165 inflight
[0] += part_stat_local_read_cpu(part
, in_flight
[0], cpu
);
166 inflight
[1] += part_stat_local_read_cpu(part
, in_flight
[1], cpu
);
168 if ((int)inflight
[0] < 0)
170 if ((int)inflight
[1] < 0)
175 * Can be deleted altogether. Later.
178 #define BLKDEV_MAJOR_HASH_SIZE 255
179 static struct blk_major_name
{
180 struct blk_major_name
*next
;
183 void (*probe
)(dev_t devt
);
184 } *major_names
[BLKDEV_MAJOR_HASH_SIZE
];
185 static DEFINE_MUTEX(major_names_lock
);
187 /* index in the above - for now: assume no multimajor ranges */
188 static inline int major_to_index(unsigned major
)
190 return major
% BLKDEV_MAJOR_HASH_SIZE
;
193 #ifdef CONFIG_PROC_FS
194 void blkdev_show(struct seq_file
*seqf
, off_t offset
)
196 struct blk_major_name
*dp
;
198 mutex_lock(&major_names_lock
);
199 for (dp
= major_names
[major_to_index(offset
)]; dp
; dp
= dp
->next
)
200 if (dp
->major
== offset
)
201 seq_printf(seqf
, "%3d %s\n", dp
->major
, dp
->name
);
202 mutex_unlock(&major_names_lock
);
204 #endif /* CONFIG_PROC_FS */
207 * __register_blkdev - register a new block device
209 * @major: the requested major device number [1..BLKDEV_MAJOR_MAX-1]. If
210 * @major = 0, try to allocate any unused major number.
211 * @name: the name of the new block device as a zero terminated string
212 * @probe: allback that is called on access to any minor number of @major
214 * The @name must be unique within the system.
216 * The return value depends on the @major input parameter:
218 * - if a major device number was requested in range [1..BLKDEV_MAJOR_MAX-1]
219 * then the function returns zero on success, or a negative error code
220 * - if any unused major number was requested with @major = 0 parameter
221 * then the return value is the allocated major number in range
222 * [1..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise
224 * See Documentation/admin-guide/devices.txt for the list of allocated
227 * Use register_blkdev instead for any new code.
229 int __register_blkdev(unsigned int major
, const char *name
,
230 void (*probe
)(dev_t devt
))
232 struct blk_major_name
**n
, *p
;
235 mutex_lock(&major_names_lock
);
239 for (index
= ARRAY_SIZE(major_names
)-1; index
> 0; index
--) {
240 if (major_names
[index
] == NULL
)
245 printk("%s: failed to get major for %s\n",
254 if (major
>= BLKDEV_MAJOR_MAX
) {
255 pr_err("%s: major requested (%u) is greater than the maximum (%u) for %s\n",
256 __func__
, major
, BLKDEV_MAJOR_MAX
-1, name
);
262 p
= kmalloc(sizeof(struct blk_major_name
), GFP_KERNEL
);
270 strlcpy(p
->name
, name
, sizeof(p
->name
));
272 index
= major_to_index(major
);
274 for (n
= &major_names
[index
]; *n
; n
= &(*n
)->next
) {
275 if ((*n
)->major
== major
)
284 printk("register_blkdev: cannot get major %u for %s\n",
289 mutex_unlock(&major_names_lock
);
292 EXPORT_SYMBOL(__register_blkdev
);
294 void unregister_blkdev(unsigned int major
, const char *name
)
296 struct blk_major_name
**n
;
297 struct blk_major_name
*p
= NULL
;
298 int index
= major_to_index(major
);
300 mutex_lock(&major_names_lock
);
301 for (n
= &major_names
[index
]; *n
; n
= &(*n
)->next
)
302 if ((*n
)->major
== major
)
304 if (!*n
|| strcmp((*n
)->name
, name
)) {
310 mutex_unlock(&major_names_lock
);
314 EXPORT_SYMBOL(unregister_blkdev
);
317 * blk_mangle_minor - scatter minor numbers apart
318 * @minor: minor number to mangle
320 * Scatter consecutively allocated @minor number apart if MANGLE_DEVT
321 * is enabled. Mangling twice gives the original value.
329 static int blk_mangle_minor(int minor
)
331 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
334 for (i
= 0; i
< MINORBITS
/ 2; i
++) {
335 int low
= minor
& (1 << i
);
336 int high
= minor
& (1 << (MINORBITS
- 1 - i
));
337 int distance
= MINORBITS
- 1 - 2 * i
;
339 minor
^= low
| high
; /* clear both bits */
340 low
<<= distance
; /* swap the positions */
342 minor
|= low
| high
; /* and set */
348 int blk_alloc_ext_minor(void)
352 idx
= ida_alloc_range(&ext_devt_ida
, 0, NR_EXT_DEVT
, GFP_KERNEL
);
358 return blk_mangle_minor(idx
);
361 void blk_free_ext_minor(unsigned int minor
)
363 ida_free(&ext_devt_ida
, blk_mangle_minor(minor
));
366 static char *bdevt_str(dev_t devt
, char *buf
)
368 if (MAJOR(devt
) <= 0xff && MINOR(devt
) <= 0xff) {
369 char tbuf
[BDEVT_SIZE
];
370 snprintf(tbuf
, BDEVT_SIZE
, "%02x%02x", MAJOR(devt
), MINOR(devt
));
371 snprintf(buf
, BDEVT_SIZE
, "%-9s", tbuf
);
373 snprintf(buf
, BDEVT_SIZE
, "%03x:%05x", MAJOR(devt
), MINOR(devt
));
378 void disk_uevent(struct gendisk
*disk
, enum kobject_action action
)
380 struct block_device
*part
;
384 xa_for_each(&disk
->part_tbl
, idx
, part
) {
385 if (bdev_is_partition(part
) && !bdev_nr_sectors(part
))
387 if (!kobject_get_unless_zero(&part
->bd_device
.kobj
))
391 kobject_uevent(bdev_kobj(part
), action
);
392 put_device(&part
->bd_device
);
397 EXPORT_SYMBOL_GPL(disk_uevent
);
399 static void disk_scan_partitions(struct gendisk
*disk
)
401 struct block_device
*bdev
;
403 if (!get_capacity(disk
) || !disk_part_scan_enabled(disk
))
406 set_bit(GD_NEED_PART_SCAN
, &disk
->state
);
407 bdev
= blkdev_get_by_dev(disk_devt(disk
), FMODE_READ
, NULL
);
409 blkdev_put(bdev
, FMODE_READ
);
412 static void register_disk(struct device
*parent
, struct gendisk
*disk
,
413 const struct attribute_group
**groups
)
415 struct device
*ddev
= disk_to_dev(disk
);
418 ddev
->parent
= parent
;
420 dev_set_name(ddev
, "%s", disk
->disk_name
);
422 /* delay uevents, until we scanned partition table */
423 dev_set_uevent_suppress(ddev
, 1);
426 WARN_ON(ddev
->groups
);
427 ddev
->groups
= groups
;
429 if (device_add(ddev
))
431 if (!sysfs_deprecated
) {
432 err
= sysfs_create_link(block_depr
, &ddev
->kobj
,
433 kobject_name(&ddev
->kobj
));
441 * avoid probable deadlock caused by allocating memory with
442 * GFP_KERNEL in runtime_resume callback of its all ancestor
445 pm_runtime_set_memalloc_noio(ddev
, true);
447 disk
->part0
->bd_holder_dir
=
448 kobject_create_and_add("holders", &ddev
->kobj
);
449 disk
->slave_dir
= kobject_create_and_add("slaves", &ddev
->kobj
);
452 * XXX: this is a mess, can't wait for real error handling in add_disk.
453 * Make sure ->slave_dir is NULL if we failed some of the registration
454 * so that the cleanup in bd_unlink_disk_holder works properly.
456 if (bd_register_pending_holders(disk
) < 0) {
457 kobject_put(disk
->slave_dir
);
458 disk
->slave_dir
= NULL
;
461 if (disk
->flags
& GENHD_FL_HIDDEN
)
464 disk_scan_partitions(disk
);
466 /* announce the disk and partitions after all partitions are created */
467 dev_set_uevent_suppress(ddev
, 0);
468 disk_uevent(disk
, KOBJ_ADD
);
470 if (disk
->bdi
->dev
) {
471 err
= sysfs_create_link(&ddev
->kobj
, &disk
->bdi
->dev
->kobj
,
478 * device_add_disk - add disk information to kernel list
479 * @parent: parent device for the disk
480 * @disk: per-device partitioning information
481 * @groups: Additional per-device sysfs groups
483 * This function registers the partitioning information in @disk
486 * FIXME: error handling
489 void device_add_disk(struct device
*parent
, struct gendisk
*disk
,
490 const struct attribute_group
**groups
)
496 * The disk queue should now be all set with enough information about
497 * the device for the elevator code to pick an adequate default
498 * elevator if one is needed, that is, for devices requesting queue
501 elevator_init_mq(disk
->queue
);
504 * If the driver provides an explicit major number it also must provide
505 * the number of minors numbers supported, and those will be used to
507 * Otherwise just allocate the device numbers for both the whole device
508 * and all partitions from the extended dev_t space.
511 WARN_ON(!disk
->minors
);
513 if (disk
->minors
> DISK_MAX_PARTS
) {
514 pr_err("block: can't allocate more than %d partitions\n",
516 disk
->minors
= DISK_MAX_PARTS
;
519 WARN_ON(disk
->minors
);
521 ret
= blk_alloc_ext_minor();
526 disk
->major
= BLOCK_EXT_MAJOR
;
527 disk
->first_minor
= MINOR(ret
);
528 disk
->flags
|= GENHD_FL_EXT_DEVT
;
531 disk_alloc_events(disk
);
533 if (disk
->flags
& GENHD_FL_HIDDEN
) {
535 * Don't let hidden disks show up in /proc/partitions,
536 * and don't bother scanning for partitions either.
538 disk
->flags
|= GENHD_FL_SUPPRESS_PARTITION_INFO
;
539 disk
->flags
|= GENHD_FL_NO_PART_SCAN
;
541 struct device
*dev
= disk_to_dev(disk
);
543 /* Register BDI before referencing it from bdev */
544 dev
->devt
= MKDEV(disk
->major
, disk
->first_minor
);
545 ret
= bdi_register(disk
->bdi
, "%u:%u",
546 disk
->major
, disk
->first_minor
);
548 bdi_set_owner(disk
->bdi
, dev
);
549 bdev_add(disk
->part0
, dev
->devt
);
551 register_disk(parent
, disk
, groups
);
552 blk_register_queue(disk
);
554 disk_add_events(disk
);
555 blk_integrity_add(disk
);
557 EXPORT_SYMBOL(device_add_disk
);
560 * del_gendisk - remove the gendisk
561 * @disk: the struct gendisk to remove
563 * Removes the gendisk and all its associated resources. This deletes the
564 * partitions associated with the gendisk, and unregisters the associated
567 * This is the counter to the respective __device_add_disk() call.
569 * The final removal of the struct gendisk happens when its refcount reaches 0
570 * with put_disk(), which should be called after del_gendisk(), if
571 * __device_add_disk() was used.
573 * Drivers exist which depend on the release of the gendisk to be synchronous,
574 * it should not be deferred.
578 void del_gendisk(struct gendisk
*disk
)
582 if (WARN_ON_ONCE(!disk_live(disk
)))
585 blk_integrity_del(disk
);
586 disk_del_events(disk
);
588 mutex_lock(&disk
->open_mutex
);
589 remove_inode_hash(disk
->part0
->bd_inode
);
590 blk_drop_partitions(disk
);
591 mutex_unlock(&disk
->open_mutex
);
593 fsync_bdev(disk
->part0
);
594 __invalidate_device(disk
->part0
, true);
596 set_capacity(disk
, 0);
598 if (!(disk
->flags
& GENHD_FL_HIDDEN
)) {
599 sysfs_remove_link(&disk_to_dev(disk
)->kobj
, "bdi");
602 * Unregister bdi before releasing device numbers (as they can
603 * get reused and we'd get clashes in sysfs).
605 bdi_unregister(disk
->bdi
);
608 blk_unregister_queue(disk
);
610 kobject_put(disk
->part0
->bd_holder_dir
);
611 kobject_put(disk
->slave_dir
);
613 part_stat_set_all(disk
->part0
, 0);
614 disk
->part0
->bd_stamp
= 0;
615 if (!sysfs_deprecated
)
616 sysfs_remove_link(block_depr
, dev_name(disk_to_dev(disk
)));
617 pm_runtime_set_memalloc_noio(disk_to_dev(disk
), false);
618 device_del(disk_to_dev(disk
));
620 EXPORT_SYMBOL(del_gendisk
);
622 /* sysfs access to bad-blocks list. */
623 static ssize_t
disk_badblocks_show(struct device
*dev
,
624 struct device_attribute
*attr
,
627 struct gendisk
*disk
= dev_to_disk(dev
);
630 return sprintf(page
, "\n");
632 return badblocks_show(disk
->bb
, page
, 0);
635 static ssize_t
disk_badblocks_store(struct device
*dev
,
636 struct device_attribute
*attr
,
637 const char *page
, size_t len
)
639 struct gendisk
*disk
= dev_to_disk(dev
);
644 return badblocks_store(disk
->bb
, page
, len
, 0);
647 void blk_request_module(dev_t devt
)
649 unsigned int major
= MAJOR(devt
);
650 struct blk_major_name
**n
;
652 mutex_lock(&major_names_lock
);
653 for (n
= &major_names
[major_to_index(major
)]; *n
; n
= &(*n
)->next
) {
654 if ((*n
)->major
== major
&& (*n
)->probe
) {
656 mutex_unlock(&major_names_lock
);
660 mutex_unlock(&major_names_lock
);
662 if (request_module("block-major-%d-%d", MAJOR(devt
), MINOR(devt
)) > 0)
663 /* Make old-style 2.4 aliases work */
664 request_module("block-major-%d", MAJOR(devt
));
668 * print a full list of all partitions - intended for places where the root
669 * filesystem can't be mounted and thus to give the victim some idea of what
672 void __init
printk_all_partitions(void)
674 struct class_dev_iter iter
;
677 class_dev_iter_init(&iter
, &block_class
, NULL
, &disk_type
);
678 while ((dev
= class_dev_iter_next(&iter
))) {
679 struct gendisk
*disk
= dev_to_disk(dev
);
680 struct block_device
*part
;
681 char devt_buf
[BDEVT_SIZE
];
685 * Don't show empty devices or things that have been
688 if (get_capacity(disk
) == 0 ||
689 (disk
->flags
& GENHD_FL_SUPPRESS_PARTITION_INFO
))
693 * Note, unlike /proc/partitions, I am showing the numbers in
694 * hex - the same format as the root= option takes.
697 xa_for_each(&disk
->part_tbl
, idx
, part
) {
698 if (!bdev_nr_sectors(part
))
700 printk("%s%s %10llu %pg %s",
701 bdev_is_partition(part
) ? " " : "",
702 bdevt_str(part
->bd_dev
, devt_buf
),
703 bdev_nr_sectors(part
) >> 1, part
,
705 part
->bd_meta_info
->uuid
: "");
706 if (bdev_is_partition(part
))
708 else if (dev
->parent
&& dev
->parent
->driver
)
709 printk(" driver: %s\n",
710 dev
->parent
->driver
->name
);
712 printk(" (driver?)\n");
716 class_dev_iter_exit(&iter
);
719 #ifdef CONFIG_PROC_FS
721 static void *disk_seqf_start(struct seq_file
*seqf
, loff_t
*pos
)
724 struct class_dev_iter
*iter
;
727 iter
= kmalloc(sizeof(*iter
), GFP_KERNEL
);
729 return ERR_PTR(-ENOMEM
);
731 seqf
->private = iter
;
732 class_dev_iter_init(iter
, &block_class
, NULL
, &disk_type
);
734 dev
= class_dev_iter_next(iter
);
739 return dev_to_disk(dev
);
742 static void *disk_seqf_next(struct seq_file
*seqf
, void *v
, loff_t
*pos
)
747 dev
= class_dev_iter_next(seqf
->private);
749 return dev_to_disk(dev
);
754 static void disk_seqf_stop(struct seq_file
*seqf
, void *v
)
756 struct class_dev_iter
*iter
= seqf
->private;
758 /* stop is called even after start failed :-( */
760 class_dev_iter_exit(iter
);
762 seqf
->private = NULL
;
766 static void *show_partition_start(struct seq_file
*seqf
, loff_t
*pos
)
770 p
= disk_seqf_start(seqf
, pos
);
771 if (!IS_ERR_OR_NULL(p
) && !*pos
)
772 seq_puts(seqf
, "major minor #blocks name\n\n");
776 static int show_partition(struct seq_file
*seqf
, void *v
)
778 struct gendisk
*sgp
= v
;
779 struct block_device
*part
;
782 /* Don't show non-partitionable removeable devices or empty devices */
783 if (!get_capacity(sgp
) || (!disk_max_parts(sgp
) &&
784 (sgp
->flags
& GENHD_FL_REMOVABLE
)))
786 if (sgp
->flags
& GENHD_FL_SUPPRESS_PARTITION_INFO
)
790 xa_for_each(&sgp
->part_tbl
, idx
, part
) {
791 if (!bdev_nr_sectors(part
))
793 seq_printf(seqf
, "%4d %7d %10llu %pg\n",
794 MAJOR(part
->bd_dev
), MINOR(part
->bd_dev
),
795 bdev_nr_sectors(part
) >> 1, part
);
801 static const struct seq_operations partitions_op
= {
802 .start
= show_partition_start
,
803 .next
= disk_seqf_next
,
804 .stop
= disk_seqf_stop
,
805 .show
= show_partition
809 static int __init
genhd_device_init(void)
813 block_class
.dev_kobj
= sysfs_dev_block_kobj
;
814 error
= class_register(&block_class
);
819 register_blkdev(BLOCK_EXT_MAJOR
, "blkext");
821 /* create top-level block dir */
822 if (!sysfs_deprecated
)
823 block_depr
= kobject_create_and_add("block", NULL
);
827 subsys_initcall(genhd_device_init
);
829 static ssize_t
disk_range_show(struct device
*dev
,
830 struct device_attribute
*attr
, char *buf
)
832 struct gendisk
*disk
= dev_to_disk(dev
);
834 return sprintf(buf
, "%d\n", disk
->minors
);
837 static ssize_t
disk_ext_range_show(struct device
*dev
,
838 struct device_attribute
*attr
, char *buf
)
840 struct gendisk
*disk
= dev_to_disk(dev
);
842 return sprintf(buf
, "%d\n", disk_max_parts(disk
));
845 static ssize_t
disk_removable_show(struct device
*dev
,
846 struct device_attribute
*attr
, char *buf
)
848 struct gendisk
*disk
= dev_to_disk(dev
);
850 return sprintf(buf
, "%d\n",
851 (disk
->flags
& GENHD_FL_REMOVABLE
? 1 : 0));
854 static ssize_t
disk_hidden_show(struct device
*dev
,
855 struct device_attribute
*attr
, char *buf
)
857 struct gendisk
*disk
= dev_to_disk(dev
);
859 return sprintf(buf
, "%d\n",
860 (disk
->flags
& GENHD_FL_HIDDEN
? 1 : 0));
863 static ssize_t
disk_ro_show(struct device
*dev
,
864 struct device_attribute
*attr
, char *buf
)
866 struct gendisk
*disk
= dev_to_disk(dev
);
868 return sprintf(buf
, "%d\n", get_disk_ro(disk
) ? 1 : 0);
871 ssize_t
part_size_show(struct device
*dev
,
872 struct device_attribute
*attr
, char *buf
)
874 return sprintf(buf
, "%llu\n", bdev_nr_sectors(dev_to_bdev(dev
)));
877 ssize_t
part_stat_show(struct device
*dev
,
878 struct device_attribute
*attr
, char *buf
)
880 struct block_device
*bdev
= dev_to_bdev(dev
);
881 struct request_queue
*q
= bdev
->bd_disk
->queue
;
882 struct disk_stats stat
;
883 unsigned int inflight
;
885 part_stat_read_all(bdev
, &stat
);
887 inflight
= blk_mq_in_flight(q
, bdev
);
889 inflight
= part_in_flight(bdev
);
892 "%8lu %8lu %8llu %8u "
893 "%8lu %8lu %8llu %8u "
895 "%8lu %8lu %8llu %8u "
899 stat
.merges
[STAT_READ
],
900 (unsigned long long)stat
.sectors
[STAT_READ
],
901 (unsigned int)div_u64(stat
.nsecs
[STAT_READ
], NSEC_PER_MSEC
),
902 stat
.ios
[STAT_WRITE
],
903 stat
.merges
[STAT_WRITE
],
904 (unsigned long long)stat
.sectors
[STAT_WRITE
],
905 (unsigned int)div_u64(stat
.nsecs
[STAT_WRITE
], NSEC_PER_MSEC
),
907 jiffies_to_msecs(stat
.io_ticks
),
908 (unsigned int)div_u64(stat
.nsecs
[STAT_READ
] +
909 stat
.nsecs
[STAT_WRITE
] +
910 stat
.nsecs
[STAT_DISCARD
] +
911 stat
.nsecs
[STAT_FLUSH
],
913 stat
.ios
[STAT_DISCARD
],
914 stat
.merges
[STAT_DISCARD
],
915 (unsigned long long)stat
.sectors
[STAT_DISCARD
],
916 (unsigned int)div_u64(stat
.nsecs
[STAT_DISCARD
], NSEC_PER_MSEC
),
917 stat
.ios
[STAT_FLUSH
],
918 (unsigned int)div_u64(stat
.nsecs
[STAT_FLUSH
], NSEC_PER_MSEC
));
921 ssize_t
part_inflight_show(struct device
*dev
, struct device_attribute
*attr
,
924 struct block_device
*bdev
= dev_to_bdev(dev
);
925 struct request_queue
*q
= bdev
->bd_disk
->queue
;
926 unsigned int inflight
[2];
929 blk_mq_in_flight_rw(q
, bdev
, inflight
);
931 part_in_flight_rw(bdev
, inflight
);
933 return sprintf(buf
, "%8u %8u\n", inflight
[0], inflight
[1]);
936 static ssize_t
disk_capability_show(struct device
*dev
,
937 struct device_attribute
*attr
, char *buf
)
939 struct gendisk
*disk
= dev_to_disk(dev
);
941 return sprintf(buf
, "%x\n", disk
->flags
);
944 static ssize_t
disk_alignment_offset_show(struct device
*dev
,
945 struct device_attribute
*attr
,
948 struct gendisk
*disk
= dev_to_disk(dev
);
950 return sprintf(buf
, "%d\n", queue_alignment_offset(disk
->queue
));
953 static ssize_t
disk_discard_alignment_show(struct device
*dev
,
954 struct device_attribute
*attr
,
957 struct gendisk
*disk
= dev_to_disk(dev
);
959 return sprintf(buf
, "%d\n", queue_discard_alignment(disk
->queue
));
962 static ssize_t
diskseq_show(struct device
*dev
,
963 struct device_attribute
*attr
, char *buf
)
965 struct gendisk
*disk
= dev_to_disk(dev
);
967 return sprintf(buf
, "%llu\n", disk
->diskseq
);
970 static DEVICE_ATTR(range
, 0444, disk_range_show
, NULL
);
971 static DEVICE_ATTR(ext_range
, 0444, disk_ext_range_show
, NULL
);
972 static DEVICE_ATTR(removable
, 0444, disk_removable_show
, NULL
);
973 static DEVICE_ATTR(hidden
, 0444, disk_hidden_show
, NULL
);
974 static DEVICE_ATTR(ro
, 0444, disk_ro_show
, NULL
);
975 static DEVICE_ATTR(size
, 0444, part_size_show
, NULL
);
976 static DEVICE_ATTR(alignment_offset
, 0444, disk_alignment_offset_show
, NULL
);
977 static DEVICE_ATTR(discard_alignment
, 0444, disk_discard_alignment_show
, NULL
);
978 static DEVICE_ATTR(capability
, 0444, disk_capability_show
, NULL
);
979 static DEVICE_ATTR(stat
, 0444, part_stat_show
, NULL
);
980 static DEVICE_ATTR(inflight
, 0444, part_inflight_show
, NULL
);
981 static DEVICE_ATTR(badblocks
, 0644, disk_badblocks_show
, disk_badblocks_store
);
982 static DEVICE_ATTR(diskseq
, 0444, diskseq_show
, NULL
);
984 #ifdef CONFIG_FAIL_MAKE_REQUEST
985 ssize_t
part_fail_show(struct device
*dev
,
986 struct device_attribute
*attr
, char *buf
)
988 return sprintf(buf
, "%d\n", dev_to_bdev(dev
)->bd_make_it_fail
);
991 ssize_t
part_fail_store(struct device
*dev
,
992 struct device_attribute
*attr
,
993 const char *buf
, size_t count
)
997 if (count
> 0 && sscanf(buf
, "%d", &i
) > 0)
998 dev_to_bdev(dev
)->bd_make_it_fail
= i
;
1003 static struct device_attribute dev_attr_fail
=
1004 __ATTR(make
-it
-fail
, 0644, part_fail_show
, part_fail_store
);
1005 #endif /* CONFIG_FAIL_MAKE_REQUEST */
1007 #ifdef CONFIG_FAIL_IO_TIMEOUT
1008 static struct device_attribute dev_attr_fail_timeout
=
1009 __ATTR(io
-timeout
-fail
, 0644, part_timeout_show
, part_timeout_store
);
1012 static struct attribute
*disk_attrs
[] = {
1013 &dev_attr_range
.attr
,
1014 &dev_attr_ext_range
.attr
,
1015 &dev_attr_removable
.attr
,
1016 &dev_attr_hidden
.attr
,
1018 &dev_attr_size
.attr
,
1019 &dev_attr_alignment_offset
.attr
,
1020 &dev_attr_discard_alignment
.attr
,
1021 &dev_attr_capability
.attr
,
1022 &dev_attr_stat
.attr
,
1023 &dev_attr_inflight
.attr
,
1024 &dev_attr_badblocks
.attr
,
1025 &dev_attr_events
.attr
,
1026 &dev_attr_events_async
.attr
,
1027 &dev_attr_events_poll_msecs
.attr
,
1028 &dev_attr_diskseq
.attr
,
1029 #ifdef CONFIG_FAIL_MAKE_REQUEST
1030 &dev_attr_fail
.attr
,
1032 #ifdef CONFIG_FAIL_IO_TIMEOUT
1033 &dev_attr_fail_timeout
.attr
,
1038 static umode_t
disk_visible(struct kobject
*kobj
, struct attribute
*a
, int n
)
1040 struct device
*dev
= container_of(kobj
, typeof(*dev
), kobj
);
1041 struct gendisk
*disk
= dev_to_disk(dev
);
1043 if (a
== &dev_attr_badblocks
.attr
&& !disk
->bb
)
1048 static struct attribute_group disk_attr_group
= {
1049 .attrs
= disk_attrs
,
1050 .is_visible
= disk_visible
,
1053 static const struct attribute_group
*disk_attr_groups
[] = {
1059 * disk_release - releases all allocated resources of the gendisk
1060 * @dev: the device representing this disk
1062 * This function releases all allocated resources of the gendisk.
1064 * Drivers which used __device_add_disk() have a gendisk with a request_queue
1065 * assigned. Since the request_queue sits on top of the gendisk for these
1066 * drivers we also call blk_put_queue() for them, and we expect the
1067 * request_queue refcount to reach 0 at this point, and so the request_queue
1068 * will also be freed prior to the disk.
1070 * Context: can sleep
1072 static void disk_release(struct device
*dev
)
1074 struct gendisk
*disk
= dev_to_disk(dev
);
1078 disk_release_events(disk
);
1079 kfree(disk
->random
);
1080 xa_destroy(&disk
->part_tbl
);
1081 disk
->queue
->disk
= NULL
;
1082 blk_put_queue(disk
->queue
);
1083 iput(disk
->part0
->bd_inode
); /* frees the disk */
1086 static int block_uevent(struct device
*dev
, struct kobj_uevent_env
*env
)
1088 struct gendisk
*disk
= dev_to_disk(dev
);
1090 return add_uevent_var(env
, "DISKSEQ=%llu", disk
->diskseq
);
1093 struct class block_class
= {
1095 .dev_uevent
= block_uevent
,
1098 static char *block_devnode(struct device
*dev
, umode_t
*mode
,
1099 kuid_t
*uid
, kgid_t
*gid
)
1101 struct gendisk
*disk
= dev_to_disk(dev
);
1103 if (disk
->fops
->devnode
)
1104 return disk
->fops
->devnode(disk
, mode
);
1108 const struct device_type disk_type
= {
1110 .groups
= disk_attr_groups
,
1111 .release
= disk_release
,
1112 .devnode
= block_devnode
,
1115 #ifdef CONFIG_PROC_FS
1117 * aggregate disk stat collector. Uses the same stats that the sysfs
1118 * entries do, above, but makes them available through one seq_file.
1120 * The output looks suspiciously like /proc/partitions with a bunch of
1123 static int diskstats_show(struct seq_file
*seqf
, void *v
)
1125 struct gendisk
*gp
= v
;
1126 struct block_device
*hd
;
1127 unsigned int inflight
;
1128 struct disk_stats stat
;
1132 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
1133 seq_puts(seqf, "major minor name"
1134 " rio rmerge rsect ruse wio wmerge "
1135 "wsect wuse running use aveq"
1140 xa_for_each(&gp
->part_tbl
, idx
, hd
) {
1141 if (bdev_is_partition(hd
) && !bdev_nr_sectors(hd
))
1143 part_stat_read_all(hd
, &stat
);
1144 if (queue_is_mq(gp
->queue
))
1145 inflight
= blk_mq_in_flight(gp
->queue
, hd
);
1147 inflight
= part_in_flight(hd
);
1149 seq_printf(seqf
, "%4d %7d %pg "
1156 MAJOR(hd
->bd_dev
), MINOR(hd
->bd_dev
), hd
,
1157 stat
.ios
[STAT_READ
],
1158 stat
.merges
[STAT_READ
],
1159 stat
.sectors
[STAT_READ
],
1160 (unsigned int)div_u64(stat
.nsecs
[STAT_READ
],
1162 stat
.ios
[STAT_WRITE
],
1163 stat
.merges
[STAT_WRITE
],
1164 stat
.sectors
[STAT_WRITE
],
1165 (unsigned int)div_u64(stat
.nsecs
[STAT_WRITE
],
1168 jiffies_to_msecs(stat
.io_ticks
),
1169 (unsigned int)div_u64(stat
.nsecs
[STAT_READ
] +
1170 stat
.nsecs
[STAT_WRITE
] +
1171 stat
.nsecs
[STAT_DISCARD
] +
1172 stat
.nsecs
[STAT_FLUSH
],
1174 stat
.ios
[STAT_DISCARD
],
1175 stat
.merges
[STAT_DISCARD
],
1176 stat
.sectors
[STAT_DISCARD
],
1177 (unsigned int)div_u64(stat
.nsecs
[STAT_DISCARD
],
1179 stat
.ios
[STAT_FLUSH
],
1180 (unsigned int)div_u64(stat
.nsecs
[STAT_FLUSH
],
1189 static const struct seq_operations diskstats_op
= {
1190 .start
= disk_seqf_start
,
1191 .next
= disk_seqf_next
,
1192 .stop
= disk_seqf_stop
,
1193 .show
= diskstats_show
1196 static int __init
proc_genhd_init(void)
1198 proc_create_seq("diskstats", 0, NULL
, &diskstats_op
);
1199 proc_create_seq("partitions", 0, NULL
, &partitions_op
);
1202 module_init(proc_genhd_init
);
1203 #endif /* CONFIG_PROC_FS */
1205 dev_t
part_devt(struct gendisk
*disk
, u8 partno
)
1207 struct block_device
*part
;
1211 part
= xa_load(&disk
->part_tbl
, partno
);
1213 devt
= part
->bd_dev
;
1219 dev_t
blk_lookup_devt(const char *name
, int partno
)
1221 dev_t devt
= MKDEV(0, 0);
1222 struct class_dev_iter iter
;
1225 class_dev_iter_init(&iter
, &block_class
, NULL
, &disk_type
);
1226 while ((dev
= class_dev_iter_next(&iter
))) {
1227 struct gendisk
*disk
= dev_to_disk(dev
);
1229 if (strcmp(dev_name(dev
), name
))
1232 if (partno
< disk
->minors
) {
1233 /* We need to return the right devno, even
1234 * if the partition doesn't exist yet.
1236 devt
= MKDEV(MAJOR(dev
->devt
),
1237 MINOR(dev
->devt
) + partno
);
1239 devt
= part_devt(disk
, partno
);
1244 class_dev_iter_exit(&iter
);
1248 struct gendisk
*__alloc_disk_node(struct request_queue
*q
, int node_id
,
1249 struct lock_class_key
*lkclass
)
1251 struct gendisk
*disk
;
1253 if (!blk_get_queue(q
))
1256 disk
= kzalloc_node(sizeof(struct gendisk
), GFP_KERNEL
, node_id
);
1260 disk
->bdi
= bdi_alloc(node_id
);
1264 disk
->part0
= bdev_alloc(disk
, 0);
1268 disk
->node_id
= node_id
;
1269 mutex_init(&disk
->open_mutex
);
1270 xa_init(&disk
->part_tbl
);
1271 if (xa_insert(&disk
->part_tbl
, 0, disk
->part0
, GFP_KERNEL
))
1272 goto out_destroy_part_tbl
;
1274 rand_initialize_disk(disk
);
1275 disk_to_dev(disk
)->class = &block_class
;
1276 disk_to_dev(disk
)->type
= &disk_type
;
1277 device_initialize(disk_to_dev(disk
));
1281 lockdep_init_map(&disk
->lockdep_map
, "(bio completion)", lkclass
, 0);
1282 #ifdef CONFIG_BLOCK_HOLDER_DEPRECATED
1283 INIT_LIST_HEAD(&disk
->slave_bdevs
);
1287 out_destroy_part_tbl
:
1288 xa_destroy(&disk
->part_tbl
);
1289 iput(disk
->part0
->bd_inode
);
1298 EXPORT_SYMBOL(__alloc_disk_node
);
1300 struct gendisk
*__blk_alloc_disk(int node
, struct lock_class_key
*lkclass
)
1302 struct request_queue
*q
;
1303 struct gendisk
*disk
;
1305 q
= blk_alloc_queue(node
);
1309 disk
= __alloc_disk_node(q
, node
, lkclass
);
1311 blk_cleanup_queue(q
);
1316 EXPORT_SYMBOL(__blk_alloc_disk
);
1319 * put_disk - decrements the gendisk refcount
1320 * @disk: the struct gendisk to decrement the refcount for
1322 * This decrements the refcount for the struct gendisk. When this reaches 0
1323 * we'll have disk_release() called.
1325 * Context: Any context, but the last reference must not be dropped from
1328 void put_disk(struct gendisk
*disk
)
1331 put_device(disk_to_dev(disk
));
1333 EXPORT_SYMBOL(put_disk
);
1336 * blk_cleanup_disk - shutdown a gendisk allocated by blk_alloc_disk
1337 * @disk: gendisk to shutdown
1339 * Mark the queue hanging off @disk DYING, drain all pending requests, then mark
1340 * the queue DEAD, destroy and put it and the gendisk structure.
1342 * Context: can sleep
1344 void blk_cleanup_disk(struct gendisk
*disk
)
1346 blk_cleanup_queue(disk
->queue
);
1349 EXPORT_SYMBOL(blk_cleanup_disk
);
1351 static void set_disk_ro_uevent(struct gendisk
*gd
, int ro
)
1353 char event
[] = "DISK_RO=1";
1354 char *envp
[] = { event
, NULL
};
1358 kobject_uevent_env(&disk_to_dev(gd
)->kobj
, KOBJ_CHANGE
, envp
);
1362 * set_disk_ro - set a gendisk read-only
1363 * @disk: gendisk to operate on
1364 * @read_only: %true to set the disk read-only, %false set the disk read/write
1366 * This function is used to indicate whether a given disk device should have its
1367 * read-only flag set. set_disk_ro() is typically used by device drivers to
1368 * indicate whether the underlying physical device is write-protected.
1370 void set_disk_ro(struct gendisk
*disk
, bool read_only
)
1373 if (test_and_set_bit(GD_READ_ONLY
, &disk
->state
))
1376 if (!test_and_clear_bit(GD_READ_ONLY
, &disk
->state
))
1379 set_disk_ro_uevent(disk
, read_only
);
1381 EXPORT_SYMBOL(set_disk_ro
);
1383 int bdev_read_only(struct block_device
*bdev
)
1385 return bdev
->bd_read_only
|| get_disk_ro(bdev
->bd_disk
);
1387 EXPORT_SYMBOL(bdev_read_only
);
1389 void inc_diskseq(struct gendisk
*disk
)
1391 disk
->diskseq
= atomic64_inc_return(&diskseq
);