1 // SPDX-License-Identifier: GPL-2.0
3 * Code extracted from drivers/block/genhd.c
4 * Copyright (C) 1991-1998 Linus Torvalds
5 * Re-organised Feb 1998 Russell King
7 * We now have independent partition support from the
8 * block drivers, which allows all the partition code to
9 * be grouped in one location, and it to be mostly self
13 #include <linux/init.h>
14 #include <linux/module.h>
16 #include <linux/slab.h>
17 #include <linux/kmod.h>
18 #include <linux/ctype.h>
19 #include <linux/genhd.h>
20 #include <linux/blktrace_api.h>
22 #include "partitions/check.h"
24 #ifdef CONFIG_BLK_DEV_MD
25 extern void md_autodetect_dev(dev_t dev
);
29 * disk_name() is used by partition check code and the genhd driver.
30 * It formats the devicename of the indicated disk into
31 * the supplied buffer (of size at least 32), and returns
32 * a pointer to that same buffer (for convenience).
35 char *disk_name(struct gendisk
*hd
, int partno
, char *buf
)
38 snprintf(buf
, BDEVNAME_SIZE
, "%s", hd
->disk_name
);
39 else if (isdigit(hd
->disk_name
[strlen(hd
->disk_name
)-1]))
40 snprintf(buf
, BDEVNAME_SIZE
, "%sp%d", hd
->disk_name
, partno
);
42 snprintf(buf
, BDEVNAME_SIZE
, "%s%d", hd
->disk_name
, partno
);
47 const char *bdevname(struct block_device
*bdev
, char *buf
)
49 return disk_name(bdev
->bd_disk
, bdev
->bd_part
->partno
, buf
);
52 EXPORT_SYMBOL(bdevname
);
54 const char *bio_devname(struct bio
*bio
, char *buf
)
56 return disk_name(bio
->bi_disk
, bio
->bi_partno
, buf
);
58 EXPORT_SYMBOL(bio_devname
);
61 * There's very little reason to use this, you should really
62 * have a struct block_device just about everywhere and use
65 const char *__bdevname(dev_t dev
, char *buffer
)
67 scnprintf(buffer
, BDEVNAME_SIZE
, "unknown-block(%u,%u)",
68 MAJOR(dev
), MINOR(dev
));
72 EXPORT_SYMBOL(__bdevname
);
74 static ssize_t
part_partition_show(struct device
*dev
,
75 struct device_attribute
*attr
, char *buf
)
77 struct hd_struct
*p
= dev_to_part(dev
);
79 return sprintf(buf
, "%d\n", p
->partno
);
82 static ssize_t
part_start_show(struct device
*dev
,
83 struct device_attribute
*attr
, char *buf
)
85 struct hd_struct
*p
= dev_to_part(dev
);
87 return sprintf(buf
, "%llu\n",(unsigned long long)p
->start_sect
);
90 ssize_t
part_size_show(struct device
*dev
,
91 struct device_attribute
*attr
, char *buf
)
93 struct hd_struct
*p
= dev_to_part(dev
);
94 return sprintf(buf
, "%llu\n",(unsigned long long)part_nr_sects_read(p
));
97 static ssize_t
part_ro_show(struct device
*dev
,
98 struct device_attribute
*attr
, char *buf
)
100 struct hd_struct
*p
= dev_to_part(dev
);
101 return sprintf(buf
, "%d\n", p
->policy
? 1 : 0);
104 static ssize_t
part_alignment_offset_show(struct device
*dev
,
105 struct device_attribute
*attr
, char *buf
)
107 struct hd_struct
*p
= dev_to_part(dev
);
108 return sprintf(buf
, "%llu\n", (unsigned long long)p
->alignment_offset
);
111 static ssize_t
part_discard_alignment_show(struct device
*dev
,
112 struct device_attribute
*attr
, char *buf
)
114 struct hd_struct
*p
= dev_to_part(dev
);
115 return sprintf(buf
, "%u\n", p
->discard_alignment
);
118 ssize_t
part_stat_show(struct device
*dev
,
119 struct device_attribute
*attr
, char *buf
)
121 struct hd_struct
*p
= dev_to_part(dev
);
122 struct request_queue
*q
= part_to_disk(p
)->queue
;
123 unsigned int inflight
[2];
126 cpu
= part_stat_lock();
127 part_round_stats(q
, cpu
, p
);
129 part_in_flight(q
, p
, inflight
);
131 "%8lu %8lu %8llu %8u "
132 "%8lu %8lu %8llu %8u "
135 part_stat_read(p
, ios
[READ
]),
136 part_stat_read(p
, merges
[READ
]),
137 (unsigned long long)part_stat_read(p
, sectors
[READ
]),
138 jiffies_to_msecs(part_stat_read(p
, ticks
[READ
])),
139 part_stat_read(p
, ios
[WRITE
]),
140 part_stat_read(p
, merges
[WRITE
]),
141 (unsigned long long)part_stat_read(p
, sectors
[WRITE
]),
142 jiffies_to_msecs(part_stat_read(p
, ticks
[WRITE
])),
144 jiffies_to_msecs(part_stat_read(p
, io_ticks
)),
145 jiffies_to_msecs(part_stat_read(p
, time_in_queue
)));
148 ssize_t
part_inflight_show(struct device
*dev
,
149 struct device_attribute
*attr
, char *buf
)
151 struct hd_struct
*p
= dev_to_part(dev
);
153 return sprintf(buf
, "%8u %8u\n", atomic_read(&p
->in_flight
[0]),
154 atomic_read(&p
->in_flight
[1]));
157 #ifdef CONFIG_FAIL_MAKE_REQUEST
158 ssize_t
part_fail_show(struct device
*dev
,
159 struct device_attribute
*attr
, char *buf
)
161 struct hd_struct
*p
= dev_to_part(dev
);
163 return sprintf(buf
, "%d\n", p
->make_it_fail
);
166 ssize_t
part_fail_store(struct device
*dev
,
167 struct device_attribute
*attr
,
168 const char *buf
, size_t count
)
170 struct hd_struct
*p
= dev_to_part(dev
);
173 if (count
> 0 && sscanf(buf
, "%d", &i
) > 0)
174 p
->make_it_fail
= (i
== 0) ? 0 : 1;
180 static DEVICE_ATTR(partition
, S_IRUGO
, part_partition_show
, NULL
);
181 static DEVICE_ATTR(start
, S_IRUGO
, part_start_show
, NULL
);
182 static DEVICE_ATTR(size
, S_IRUGO
, part_size_show
, NULL
);
183 static DEVICE_ATTR(ro
, S_IRUGO
, part_ro_show
, NULL
);
184 static DEVICE_ATTR(alignment_offset
, S_IRUGO
, part_alignment_offset_show
, NULL
);
185 static DEVICE_ATTR(discard_alignment
, S_IRUGO
, part_discard_alignment_show
,
187 static DEVICE_ATTR(stat
, S_IRUGO
, part_stat_show
, NULL
);
188 static DEVICE_ATTR(inflight
, S_IRUGO
, part_inflight_show
, NULL
);
189 #ifdef CONFIG_FAIL_MAKE_REQUEST
190 static struct device_attribute dev_attr_fail
=
191 __ATTR(make
-it
-fail
, S_IRUGO
|S_IWUSR
, part_fail_show
, part_fail_store
);
194 static struct attribute
*part_attrs
[] = {
195 &dev_attr_partition
.attr
,
196 &dev_attr_start
.attr
,
199 &dev_attr_alignment_offset
.attr
,
200 &dev_attr_discard_alignment
.attr
,
202 &dev_attr_inflight
.attr
,
203 #ifdef CONFIG_FAIL_MAKE_REQUEST
209 static struct attribute_group part_attr_group
= {
213 static const struct attribute_group
*part_attr_groups
[] = {
215 #ifdef CONFIG_BLK_DEV_IO_TRACE
216 &blk_trace_attr_group
,
221 static void part_release(struct device
*dev
)
223 struct hd_struct
*p
= dev_to_part(dev
);
224 blk_free_devt(dev
->devt
);
229 static int part_uevent(struct device
*dev
, struct kobj_uevent_env
*env
)
231 struct hd_struct
*part
= dev_to_part(dev
);
233 add_uevent_var(env
, "PARTN=%u", part
->partno
);
234 if (part
->info
&& part
->info
->volname
[0])
235 add_uevent_var(env
, "PARTNAME=%s", part
->info
->volname
);
239 struct device_type part_type
= {
241 .groups
= part_attr_groups
,
242 .release
= part_release
,
243 .uevent
= part_uevent
,
246 static void delete_partition_rcu_cb(struct rcu_head
*head
)
248 struct hd_struct
*part
= container_of(head
, struct hd_struct
, rcu_head
);
250 part
->start_sect
= 0;
252 part_stat_set_all(part
, 0);
253 put_device(part_to_dev(part
));
256 void __delete_partition(struct percpu_ref
*ref
)
258 struct hd_struct
*part
= container_of(ref
, struct hd_struct
, ref
);
259 call_rcu(&part
->rcu_head
, delete_partition_rcu_cb
);
263 * Must be called either with bd_mutex held, before a disk can be opened or
264 * after all disk users are gone.
266 void delete_partition(struct gendisk
*disk
, int partno
)
268 struct disk_part_tbl
*ptbl
=
269 rcu_dereference_protected(disk
->part_tbl
, 1);
270 struct hd_struct
*part
;
272 if (partno
>= ptbl
->len
)
275 part
= rcu_dereference_protected(ptbl
->part
[partno
], 1);
279 rcu_assign_pointer(ptbl
->part
[partno
], NULL
);
280 rcu_assign_pointer(ptbl
->last_lookup
, NULL
);
281 kobject_put(part
->holder_dir
);
282 device_del(part_to_dev(part
));
284 hd_struct_kill(part
);
287 static ssize_t
whole_disk_show(struct device
*dev
,
288 struct device_attribute
*attr
, char *buf
)
292 static DEVICE_ATTR(whole_disk
, S_IRUSR
| S_IRGRP
| S_IROTH
,
293 whole_disk_show
, NULL
);
296 * Must be called either with bd_mutex held, before a disk can be opened or
297 * after all disk users are gone.
299 struct hd_struct
*add_partition(struct gendisk
*disk
, int partno
,
300 sector_t start
, sector_t len
, int flags
,
301 struct partition_meta_info
*info
)
304 dev_t devt
= MKDEV(0, 0);
305 struct device
*ddev
= disk_to_dev(disk
);
307 struct disk_part_tbl
*ptbl
;
311 err
= disk_expand_part_tbl(disk
, partno
);
314 ptbl
= rcu_dereference_protected(disk
->part_tbl
, 1);
316 if (ptbl
->part
[partno
])
317 return ERR_PTR(-EBUSY
);
319 p
= kzalloc(sizeof(*p
), GFP_KERNEL
);
321 return ERR_PTR(-EBUSY
);
323 if (!init_part_stats(p
)) {
328 seqcount_init(&p
->nr_sects_seq
);
329 pdev
= part_to_dev(p
);
331 p
->start_sect
= start
;
332 p
->alignment_offset
=
333 queue_limit_alignment_offset(&disk
->queue
->limits
, start
);
334 p
->discard_alignment
=
335 queue_limit_discard_alignment(&disk
->queue
->limits
, start
);
338 p
->policy
= get_disk_ro(disk
);
341 struct partition_meta_info
*pinfo
= alloc_part_info(disk
);
346 memcpy(pinfo
, info
, sizeof(*info
));
350 dname
= dev_name(ddev
);
351 if (isdigit(dname
[strlen(dname
) - 1]))
352 dev_set_name(pdev
, "%sp%d", dname
, partno
);
354 dev_set_name(pdev
, "%s%d", dname
, partno
);
356 device_initialize(pdev
);
357 pdev
->class = &block_class
;
358 pdev
->type
= &part_type
;
361 err
= blk_alloc_devt(p
, &devt
);
366 /* delay uevent until 'holders' subdir is created */
367 dev_set_uevent_suppress(pdev
, 1);
368 err
= device_add(pdev
);
373 p
->holder_dir
= kobject_create_and_add("holders", &pdev
->kobj
);
377 dev_set_uevent_suppress(pdev
, 0);
378 if (flags
& ADDPART_FLAG_WHOLEDISK
) {
379 err
= device_create_file(pdev
, &dev_attr_whole_disk
);
384 err
= hd_ref_init(p
);
386 if (flags
& ADDPART_FLAG_WHOLEDISK
)
387 goto out_remove_file
;
391 /* everything is up and running, commence */
392 rcu_assign_pointer(ptbl
->part
[partno
], p
);
394 /* suppress uevent if the disk suppresses it */
395 if (!dev_get_uevent_suppress(ddev
))
396 kobject_uevent(&pdev
->kobj
, KOBJ_ADD
);
407 device_remove_file(pdev
, &dev_attr_whole_disk
);
409 kobject_put(p
->holder_dir
);
416 static bool disk_unlock_native_capacity(struct gendisk
*disk
)
418 const struct block_device_operations
*bdops
= disk
->fops
;
420 if (bdops
->unlock_native_capacity
&&
421 !(disk
->flags
& GENHD_FL_NATIVE_CAPACITY
)) {
422 printk(KERN_CONT
"enabling native capacity\n");
423 bdops
->unlock_native_capacity(disk
);
424 disk
->flags
|= GENHD_FL_NATIVE_CAPACITY
;
427 printk(KERN_CONT
"truncated\n");
432 static int drop_partitions(struct gendisk
*disk
, struct block_device
*bdev
)
434 struct disk_part_iter piter
;
435 struct hd_struct
*part
;
438 if (bdev
->bd_part_count
|| bdev
->bd_super
)
440 res
= invalidate_partition(disk
, 0);
444 disk_part_iter_init(&piter
, disk
, DISK_PITER_INCL_EMPTY
);
445 while ((part
= disk_part_iter_next(&piter
)))
446 delete_partition(disk
, part
->partno
);
447 disk_part_iter_exit(&piter
);
452 static bool part_zone_aligned(struct gendisk
*disk
,
453 struct block_device
*bdev
,
454 sector_t from
, sector_t size
)
456 unsigned int zone_sectors
= bdev_zone_sectors(bdev
);
459 * If this function is called, then the disk is a zoned block device
460 * (host-aware or host-managed). This can be detected even if the
461 * zoned block device support is disabled (CONFIG_BLK_DEV_ZONED not
462 * set). In this case, however, only host-aware devices will be seen
463 * as a block device is not created for host-managed devices. Without
464 * zoned block device support, host-aware drives can still be used as
465 * regular block devices (no zone operation) and their zone size will
466 * be reported as 0. Allow this case.
472 * Check partition start and size alignement. If the drive has a
473 * smaller last runt zone, ignore it and allow the partition to
474 * use it. Check the zone size too: it should be a power of 2 number
477 if (WARN_ON_ONCE(!is_power_of_2(zone_sectors
))) {
480 div_u64_rem(from
, zone_sectors
, &rem
);
483 if ((from
+ size
) < get_capacity(disk
)) {
484 div_u64_rem(size
, zone_sectors
, &rem
);
491 if (from
& (zone_sectors
- 1))
493 if ((from
+ size
) < get_capacity(disk
) &&
494 (size
& (zone_sectors
- 1)))
502 int rescan_partitions(struct gendisk
*disk
, struct block_device
*bdev
)
504 struct parsed_partitions
*state
= NULL
;
505 struct hd_struct
*part
;
508 if (state
&& !IS_ERR(state
)) {
509 free_partitions(state
);
513 res
= drop_partitions(disk
, bdev
);
517 if (disk
->fops
->revalidate_disk
)
518 disk
->fops
->revalidate_disk(disk
);
519 check_disk_size_change(disk
, bdev
);
520 bdev
->bd_invalidated
= 0;
521 if (!get_capacity(disk
) || !(state
= check_partition(disk
, bdev
)))
525 * I/O error reading the partition table. If any
526 * partition code tried to read beyond EOD, retry
527 * after unlocking native capacity.
529 if (PTR_ERR(state
) == -ENOSPC
) {
530 printk(KERN_WARNING
"%s: partition table beyond EOD, ",
532 if (disk_unlock_native_capacity(disk
))
538 * If any partition code tried to read beyond EOD, try
539 * unlocking native capacity even if partition table is
540 * successfully read as we could be missing some partitions.
542 if (state
->access_beyond_eod
) {
544 "%s: partition table partially beyond EOD, ",
546 if (disk_unlock_native_capacity(disk
))
550 /* tell userspace that the media / partition table may have changed */
551 kobject_uevent(&disk_to_dev(disk
)->kobj
, KOBJ_CHANGE
);
553 /* Detect the highest partition number and preallocate
554 * disk->part_tbl. This is an optimization and not strictly
557 for (p
= 1, highest
= 0; p
< state
->limit
; p
++)
558 if (state
->parts
[p
].size
)
561 disk_expand_part_tbl(disk
, highest
);
564 for (p
= 1; p
< state
->limit
; p
++) {
567 size
= state
->parts
[p
].size
;
571 from
= state
->parts
[p
].from
;
572 if (from
>= get_capacity(disk
)) {
574 "%s: p%d start %llu is beyond EOD, ",
575 disk
->disk_name
, p
, (unsigned long long) from
);
576 if (disk_unlock_native_capacity(disk
))
581 if (from
+ size
> get_capacity(disk
)) {
583 "%s: p%d size %llu extends beyond EOD, ",
584 disk
->disk_name
, p
, (unsigned long long) size
);
586 if (disk_unlock_native_capacity(disk
)) {
587 /* free state and restart */
591 * we can not ignore partitions of broken tables
592 * created by for example camera firmware, but
593 * we limit them to the end of the disk to avoid
594 * creating invalid block devices
596 size
= get_capacity(disk
) - from
;
601 * On a zoned block device, partitions should be aligned on the
602 * device zone size (i.e. zone boundary crossing not allowed).
603 * Otherwise, resetting the write pointer of the last zone of
604 * one partition may impact the following partition.
606 if (bdev_is_zoned(bdev
) &&
607 !part_zone_aligned(disk
, bdev
, from
, size
)) {
609 "%s: p%d start %llu+%llu is not zone aligned\n",
610 disk
->disk_name
, p
, (unsigned long long) from
,
611 (unsigned long long) size
);
615 part
= add_partition(disk
, p
, from
, size
,
616 state
->parts
[p
].flags
,
617 &state
->parts
[p
].info
);
619 printk(KERN_ERR
" %s: p%d could not be added: %ld\n",
620 disk
->disk_name
, p
, -PTR_ERR(part
));
623 #ifdef CONFIG_BLK_DEV_MD
624 if (state
->parts
[p
].flags
& ADDPART_FLAG_RAID
)
625 md_autodetect_dev(part_to_dev(part
)->devt
);
628 free_partitions(state
);
632 int invalidate_partitions(struct gendisk
*disk
, struct block_device
*bdev
)
636 if (!bdev
->bd_invalidated
)
639 res
= drop_partitions(disk
, bdev
);
643 set_capacity(disk
, 0);
644 check_disk_size_change(disk
, bdev
);
645 bdev
->bd_invalidated
= 0;
646 /* tell userspace that the media / partition table may have changed */
647 kobject_uevent(&disk_to_dev(disk
)->kobj
, KOBJ_CHANGE
);
652 unsigned char *read_dev_sector(struct block_device
*bdev
, sector_t n
, Sector
*p
)
654 struct address_space
*mapping
= bdev
->bd_inode
->i_mapping
;
657 page
= read_mapping_page(mapping
, (pgoff_t
)(n
>> (PAGE_SHIFT
-9)), NULL
);
662 return (unsigned char *)page_address(page
) + ((n
& ((1 << (PAGE_SHIFT
- 9)) - 1)) << 9);
670 EXPORT_SYMBOL(read_dev_sector
);