1 // SPDX-License-Identifier: GPL-2.0
2 #include <linux/capability.h>
3 #include <linux/compat.h>
4 #include <linux/blkdev.h>
5 #include <linux/export.h>
7 #include <linux/blkpg.h>
8 #include <linux/hdreg.h>
9 #include <linux/backing-dev.h>
11 #include <linux/blktrace_api.h>
13 #include <linux/uaccess.h>
16 static int blkpg_do_ioctl(struct block_device
*bdev
,
17 struct blkpg_partition __user
*upart
, int op
)
19 struct gendisk
*disk
= bdev
->bd_disk
;
20 struct blkpg_partition p
;
21 sector_t start
, length
, capacity
, end
;
23 if (!capable(CAP_SYS_ADMIN
))
25 if (copy_from_user(&p
, upart
, sizeof(struct blkpg_partition
)))
27 if (bdev_is_partition(bdev
))
33 if (op
== BLKPG_DEL_PARTITION
)
34 return bdev_del_partition(disk
, p
.pno
);
36 if (p
.start
< 0 || p
.length
<= 0 || p
.start
+ p
.length
< 0)
38 /* Check that the partition is aligned to the block size */
39 if (!IS_ALIGNED(p
.start
| p
.length
, bdev_logical_block_size(bdev
)))
42 start
= p
.start
>> SECTOR_SHIFT
;
43 length
= p
.length
>> SECTOR_SHIFT
;
44 capacity
= get_capacity(disk
);
46 if (check_add_overflow(start
, length
, &end
))
49 if (start
>= capacity
|| end
> capacity
)
53 case BLKPG_ADD_PARTITION
:
54 return bdev_add_partition(disk
, p
.pno
, start
, length
);
55 case BLKPG_RESIZE_PARTITION
:
56 return bdev_resize_partition(disk
, p
.pno
, start
, length
);
62 static int blkpg_ioctl(struct block_device
*bdev
,
63 struct blkpg_ioctl_arg __user
*arg
)
65 struct blkpg_partition __user
*udata
;
68 if (get_user(op
, &arg
->op
) || get_user(udata
, &arg
->data
))
71 return blkpg_do_ioctl(bdev
, udata
, op
);
75 struct compat_blkpg_ioctl_arg
{
82 static int compat_blkpg_ioctl(struct block_device
*bdev
,
83 struct compat_blkpg_ioctl_arg __user
*arg
)
88 if (get_user(op
, &arg
->op
) || get_user(udata
, &arg
->data
))
91 return blkpg_do_ioctl(bdev
, compat_ptr(udata
), op
);
95 static int blk_ioctl_discard(struct block_device
*bdev
, blk_mode_t mode
,
100 struct inode
*inode
= bdev
->bd_inode
;
103 if (!(mode
& BLK_OPEN_WRITE
))
106 if (!bdev_max_discard_sectors(bdev
))
109 if (copy_from_user(range
, (void __user
*)arg
, sizeof(range
)))
120 if (start
+ len
> bdev_nr_bytes(bdev
))
123 filemap_invalidate_lock(inode
->i_mapping
);
124 err
= truncate_bdev_range(bdev
, mode
, start
, start
+ len
- 1);
127 err
= blkdev_issue_discard(bdev
, start
>> 9, len
>> 9, GFP_KERNEL
);
129 filemap_invalidate_unlock(inode
->i_mapping
);
133 static int blk_ioctl_secure_erase(struct block_device
*bdev
, blk_mode_t mode
,
140 if (!(mode
& BLK_OPEN_WRITE
))
142 if (!bdev_max_secure_erase_sectors(bdev
))
144 if (copy_from_user(range
, argp
, sizeof(range
)))
149 if ((start
& 511) || (len
& 511))
151 if (start
+ len
> bdev_nr_bytes(bdev
))
154 filemap_invalidate_lock(bdev
->bd_inode
->i_mapping
);
155 err
= truncate_bdev_range(bdev
, mode
, start
, start
+ len
- 1);
157 err
= blkdev_issue_secure_erase(bdev
, start
>> 9, len
>> 9,
159 filemap_invalidate_unlock(bdev
->bd_inode
->i_mapping
);
164 static int blk_ioctl_zeroout(struct block_device
*bdev
, blk_mode_t mode
,
168 uint64_t start
, end
, len
;
169 struct inode
*inode
= bdev
->bd_inode
;
172 if (!(mode
& BLK_OPEN_WRITE
))
175 if (copy_from_user(range
, (void __user
*)arg
, sizeof(range
)))
180 end
= start
+ len
- 1;
186 if (end
>= (uint64_t)bdev_nr_bytes(bdev
))
191 /* Invalidate the page cache, including dirty pages */
192 filemap_invalidate_lock(inode
->i_mapping
);
193 err
= truncate_bdev_range(bdev
, mode
, start
, end
);
197 err
= blkdev_issue_zeroout(bdev
, start
>> 9, len
>> 9, GFP_KERNEL
,
198 BLKDEV_ZERO_NOUNMAP
);
201 filemap_invalidate_unlock(inode
->i_mapping
);
205 static int put_ushort(unsigned short __user
*argp
, unsigned short val
)
207 return put_user(val
, argp
);
210 static int put_int(int __user
*argp
, int val
)
212 return put_user(val
, argp
);
215 static int put_uint(unsigned int __user
*argp
, unsigned int val
)
217 return put_user(val
, argp
);
220 static int put_long(long __user
*argp
, long val
)
222 return put_user(val
, argp
);
225 static int put_ulong(unsigned long __user
*argp
, unsigned long val
)
227 return put_user(val
, argp
);
230 static int put_u64(u64 __user
*argp
, u64 val
)
232 return put_user(val
, argp
);
236 static int compat_put_long(compat_long_t __user
*argp
, long val
)
238 return put_user(val
, argp
);
241 static int compat_put_ulong(compat_ulong_t __user
*argp
, compat_ulong_t val
)
243 return put_user(val
, argp
);
249 * This is the equivalent of compat_ptr_ioctl(), to be used by block
250 * drivers that implement only commands that are completely compatible
251 * between 32-bit and 64-bit user space
253 int blkdev_compat_ptr_ioctl(struct block_device
*bdev
, blk_mode_t mode
,
254 unsigned cmd
, unsigned long arg
)
256 struct gendisk
*disk
= bdev
->bd_disk
;
258 if (disk
->fops
->ioctl
)
259 return disk
->fops
->ioctl(bdev
, mode
, cmd
,
260 (unsigned long)compat_ptr(arg
));
264 EXPORT_SYMBOL(blkdev_compat_ptr_ioctl
);
267 static bool blkdev_pr_allowed(struct block_device
*bdev
, blk_mode_t mode
)
269 /* no sense to make reservations for partitions */
270 if (bdev_is_partition(bdev
))
273 if (capable(CAP_SYS_ADMIN
))
276 * Only allow unprivileged reservations if the file descriptor is open
279 return mode
& BLK_OPEN_WRITE
;
282 static int blkdev_pr_register(struct block_device
*bdev
, blk_mode_t mode
,
283 struct pr_registration __user
*arg
)
285 const struct pr_ops
*ops
= bdev
->bd_disk
->fops
->pr_ops
;
286 struct pr_registration reg
;
288 if (!blkdev_pr_allowed(bdev
, mode
))
290 if (!ops
|| !ops
->pr_register
)
292 if (copy_from_user(®
, arg
, sizeof(reg
)))
295 if (reg
.flags
& ~PR_FL_IGNORE_KEY
)
297 return ops
->pr_register(bdev
, reg
.old_key
, reg
.new_key
, reg
.flags
);
300 static int blkdev_pr_reserve(struct block_device
*bdev
, blk_mode_t mode
,
301 struct pr_reservation __user
*arg
)
303 const struct pr_ops
*ops
= bdev
->bd_disk
->fops
->pr_ops
;
304 struct pr_reservation rsv
;
306 if (!blkdev_pr_allowed(bdev
, mode
))
308 if (!ops
|| !ops
->pr_reserve
)
310 if (copy_from_user(&rsv
, arg
, sizeof(rsv
)))
313 if (rsv
.flags
& ~PR_FL_IGNORE_KEY
)
315 return ops
->pr_reserve(bdev
, rsv
.key
, rsv
.type
, rsv
.flags
);
318 static int blkdev_pr_release(struct block_device
*bdev
, blk_mode_t mode
,
319 struct pr_reservation __user
*arg
)
321 const struct pr_ops
*ops
= bdev
->bd_disk
->fops
->pr_ops
;
322 struct pr_reservation rsv
;
324 if (!blkdev_pr_allowed(bdev
, mode
))
326 if (!ops
|| !ops
->pr_release
)
328 if (copy_from_user(&rsv
, arg
, sizeof(rsv
)))
333 return ops
->pr_release(bdev
, rsv
.key
, rsv
.type
);
336 static int blkdev_pr_preempt(struct block_device
*bdev
, blk_mode_t mode
,
337 struct pr_preempt __user
*arg
, bool abort
)
339 const struct pr_ops
*ops
= bdev
->bd_disk
->fops
->pr_ops
;
342 if (!blkdev_pr_allowed(bdev
, mode
))
344 if (!ops
|| !ops
->pr_preempt
)
346 if (copy_from_user(&p
, arg
, sizeof(p
)))
351 return ops
->pr_preempt(bdev
, p
.old_key
, p
.new_key
, p
.type
, abort
);
354 static int blkdev_pr_clear(struct block_device
*bdev
, blk_mode_t mode
,
355 struct pr_clear __user
*arg
)
357 const struct pr_ops
*ops
= bdev
->bd_disk
->fops
->pr_ops
;
360 if (!blkdev_pr_allowed(bdev
, mode
))
362 if (!ops
|| !ops
->pr_clear
)
364 if (copy_from_user(&c
, arg
, sizeof(c
)))
369 return ops
->pr_clear(bdev
, c
.key
);
372 static int blkdev_flushbuf(struct block_device
*bdev
, unsigned cmd
,
375 if (!capable(CAP_SYS_ADMIN
))
378 mutex_lock(&bdev
->bd_holder_lock
);
379 if (bdev
->bd_holder_ops
&& bdev
->bd_holder_ops
->sync
)
380 bdev
->bd_holder_ops
->sync(bdev
);
382 mutex_unlock(&bdev
->bd_holder_lock
);
386 invalidate_bdev(bdev
);
390 static int blkdev_roset(struct block_device
*bdev
, unsigned cmd
,
395 if (!capable(CAP_SYS_ADMIN
))
398 if (get_user(n
, (int __user
*)arg
))
400 if (bdev
->bd_disk
->fops
->set_read_only
) {
401 ret
= bdev
->bd_disk
->fops
->set_read_only(bdev
, n
);
405 bdev
->bd_read_only
= n
;
409 static int blkdev_getgeo(struct block_device
*bdev
,
410 struct hd_geometry __user
*argp
)
412 struct gendisk
*disk
= bdev
->bd_disk
;
413 struct hd_geometry geo
;
418 if (!disk
->fops
->getgeo
)
422 * We need to set the startsect first, the driver may
423 * want to override it.
425 memset(&geo
, 0, sizeof(geo
));
426 geo
.start
= get_start_sect(bdev
);
427 ret
= disk
->fops
->getgeo(bdev
, &geo
);
430 if (copy_to_user(argp
, &geo
, sizeof(geo
)))
436 struct compat_hd_geometry
{
438 unsigned char sectors
;
439 unsigned short cylinders
;
443 static int compat_hdio_getgeo(struct block_device
*bdev
,
444 struct compat_hd_geometry __user
*ugeo
)
446 struct gendisk
*disk
= bdev
->bd_disk
;
447 struct hd_geometry geo
;
452 if (!disk
->fops
->getgeo
)
455 memset(&geo
, 0, sizeof(geo
));
457 * We need to set the startsect first, the driver may
458 * want to override it.
460 geo
.start
= get_start_sect(bdev
);
461 ret
= disk
->fops
->getgeo(bdev
, &geo
);
465 ret
= copy_to_user(ugeo
, &geo
, 4);
466 ret
|= put_user(geo
.start
, &ugeo
->start
);
474 /* set the logical block size */
475 static int blkdev_bszset(struct block_device
*bdev
, blk_mode_t mode
,
481 if (!capable(CAP_SYS_ADMIN
))
485 if (get_user(n
, argp
))
488 if (mode
& BLK_OPEN_EXCL
)
489 return set_blocksize(bdev
, n
);
491 file
= bdev_file_open_by_dev(bdev
->bd_dev
, mode
, &bdev
, NULL
);
494 ret
= set_blocksize(bdev
, n
);
500 * Common commands that are handled the same way on native and compat
501 * user space. Note the separate arg/argp parameters that are needed
502 * to deal with the compat_ptr() conversion.
504 static int blkdev_common_ioctl(struct block_device
*bdev
, blk_mode_t mode
,
505 unsigned int cmd
, unsigned long arg
,
508 unsigned int max_sectors
;
512 return blkdev_flushbuf(bdev
, cmd
, arg
);
514 return blkdev_roset(bdev
, cmd
, arg
);
516 return blk_ioctl_discard(bdev
, mode
, arg
);
518 return blk_ioctl_secure_erase(bdev
, mode
, argp
);
520 return blk_ioctl_zeroout(bdev
, mode
, arg
);
522 return put_u64(argp
, bdev
->bd_disk
->diskseq
);
524 return blkdev_report_zones_ioctl(bdev
, cmd
, arg
);
529 return blkdev_zone_mgmt_ioctl(bdev
, mode
, cmd
, arg
);
531 return put_uint(argp
, bdev_zone_sectors(bdev
));
533 return put_uint(argp
, bdev_nr_zones(bdev
));
535 return put_int(argp
, bdev_read_only(bdev
) != 0);
536 case BLKSSZGET
: /* get block device logical block size */
537 return put_int(argp
, bdev_logical_block_size(bdev
));
538 case BLKPBSZGET
: /* get block device physical block size */
539 return put_uint(argp
, bdev_physical_block_size(bdev
));
541 return put_uint(argp
, bdev_io_min(bdev
));
543 return put_uint(argp
, bdev_io_opt(bdev
));
545 return put_int(argp
, bdev_alignment_offset(bdev
));
546 case BLKDISCARDZEROES
:
547 return put_uint(argp
, 0);
549 max_sectors
= min_t(unsigned int, USHRT_MAX
,
550 queue_max_sectors(bdev_get_queue(bdev
)));
551 return put_ushort(argp
, max_sectors
);
553 return put_ushort(argp
, !bdev_nonrot(bdev
));
556 if(!capable(CAP_SYS_ADMIN
))
558 bdev
->bd_disk
->bdi
->ra_pages
= (arg
* 512) / PAGE_SIZE
;
561 if (!capable(CAP_SYS_ADMIN
))
563 if (bdev_is_partition(bdev
))
565 return disk_scan_partitions(bdev
->bd_disk
, mode
);
568 case BLKTRACETEARDOWN
:
569 return blk_trace_ioctl(bdev
, cmd
, argp
);
570 case IOC_PR_REGISTER
:
571 return blkdev_pr_register(bdev
, mode
, argp
);
573 return blkdev_pr_reserve(bdev
, mode
, argp
);
575 return blkdev_pr_release(bdev
, mode
, argp
);
577 return blkdev_pr_preempt(bdev
, mode
, argp
, false);
578 case IOC_PR_PREEMPT_ABORT
:
579 return blkdev_pr_preempt(bdev
, mode
, argp
, true);
581 return blkdev_pr_clear(bdev
, mode
, argp
);
588 * Always keep this in sync with compat_blkdev_ioctl()
589 * to handle all incompatible commands in both functions.
591 * New commands must be compatible and go into blkdev_common_ioctl
593 long blkdev_ioctl(struct file
*file
, unsigned cmd
, unsigned long arg
)
595 struct block_device
*bdev
= I_BDEV(file
->f_mapping
->host
);
596 void __user
*argp
= (void __user
*)arg
;
597 blk_mode_t mode
= file_to_blk_mode(file
);
601 /* These need separate implementations for the data structure */
603 return blkdev_getgeo(bdev
, argp
);
605 return blkpg_ioctl(bdev
, argp
);
607 /* Compat mode returns 32-bit data instead of 'long' */
612 return put_long(argp
,
613 (bdev
->bd_disk
->bdi
->ra_pages
* PAGE_SIZE
) / 512);
615 if (bdev_nr_sectors(bdev
) > ~0UL)
617 return put_ulong(argp
, bdev_nr_sectors(bdev
));
619 /* The data is compatible, but the command number is different */
620 case BLKBSZGET
: /* get block device soft block size (cf. BLKSSZGET) */
621 return put_int(argp
, block_size(bdev
));
623 return blkdev_bszset(bdev
, mode
, argp
);
625 return put_u64(argp
, bdev_nr_bytes(bdev
));
627 /* Incompatible alignment on i386 */
629 return blk_trace_ioctl(bdev
, cmd
, argp
);
634 ret
= blkdev_common_ioctl(bdev
, mode
, cmd
, arg
, argp
);
635 if (ret
!= -ENOIOCTLCMD
)
638 if (!bdev
->bd_disk
->fops
->ioctl
)
640 return bdev
->bd_disk
->fops
->ioctl(bdev
, mode
, cmd
, arg
);
645 #define BLKBSZGET_32 _IOR(0x12, 112, int)
646 #define BLKBSZSET_32 _IOW(0x12, 113, int)
647 #define BLKGETSIZE64_32 _IOR(0x12, 114, int)
649 /* Most of the generic ioctls are handled in the normal fallback path.
650 This assumes the blkdev's low level compat_ioctl always returns
651 ENOIOCTLCMD for unknown ioctls. */
652 long compat_blkdev_ioctl(struct file
*file
, unsigned cmd
, unsigned long arg
)
655 void __user
*argp
= compat_ptr(arg
);
656 struct block_device
*bdev
= I_BDEV(file
->f_mapping
->host
);
657 struct gendisk
*disk
= bdev
->bd_disk
;
658 blk_mode_t mode
= file_to_blk_mode(file
);
661 /* These need separate implementations for the data structure */
663 return compat_hdio_getgeo(bdev
, argp
);
665 return compat_blkpg_ioctl(bdev
, argp
);
667 /* Compat mode returns 32-bit data instead of 'long' */
672 return compat_put_long(argp
,
673 (bdev
->bd_disk
->bdi
->ra_pages
* PAGE_SIZE
) / 512);
675 if (bdev_nr_sectors(bdev
) > ~(compat_ulong_t
)0)
677 return compat_put_ulong(argp
, bdev_nr_sectors(bdev
));
679 /* The data is compatible, but the command number is different */
680 case BLKBSZGET_32
: /* get the logical block size (cf. BLKSSZGET) */
681 return put_int(argp
, bdev_logical_block_size(bdev
));
683 return blkdev_bszset(bdev
, mode
, argp
);
684 case BLKGETSIZE64_32
:
685 return put_u64(argp
, bdev_nr_bytes(bdev
));
687 /* Incompatible alignment on i386 */
688 case BLKTRACESETUP32
:
689 return blk_trace_ioctl(bdev
, cmd
, argp
);
694 ret
= blkdev_common_ioctl(bdev
, mode
, cmd
, arg
, argp
);
695 if (ret
== -ENOIOCTLCMD
&& disk
->fops
->compat_ioctl
)
696 ret
= disk
->fops
->compat_ioctl(bdev
, mode
, cmd
, arg
);