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 long long start
, length
;
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 start
= p
.start
>> SECTOR_SHIFT
;
37 length
= p
.length
>> SECTOR_SHIFT
;
40 case BLKPG_ADD_PARTITION
:
41 /* check if partition is aligned to blocksize */
42 if (p
.start
& (bdev_logical_block_size(bdev
) - 1))
44 return bdev_add_partition(disk
, p
.pno
, start
, length
);
45 case BLKPG_RESIZE_PARTITION
:
46 return bdev_resize_partition(disk
, p
.pno
, start
, length
);
52 static int blkpg_ioctl(struct block_device
*bdev
,
53 struct blkpg_ioctl_arg __user
*arg
)
55 struct blkpg_partition __user
*udata
;
58 if (get_user(op
, &arg
->op
) || get_user(udata
, &arg
->data
))
61 return blkpg_do_ioctl(bdev
, udata
, op
);
65 struct compat_blkpg_ioctl_arg
{
72 static int compat_blkpg_ioctl(struct block_device
*bdev
,
73 struct compat_blkpg_ioctl_arg __user
*arg
)
78 if (get_user(op
, &arg
->op
) || get_user(udata
, &arg
->data
))
81 return blkpg_do_ioctl(bdev
, compat_ptr(udata
), op
);
85 static int blkdev_reread_part(struct block_device
*bdev
, fmode_t mode
)
87 struct block_device
*tmp
;
89 if (!disk_part_scan_enabled(bdev
->bd_disk
) || bdev_is_partition(bdev
))
91 if (!capable(CAP_SYS_ADMIN
))
93 if (bdev
->bd_disk
->open_partitions
)
97 * Reopen the device to revalidate the driver state and force a
101 set_bit(GD_NEED_PART_SCAN
, &bdev
->bd_disk
->state
);
103 tmp
= blkdev_get_by_dev(bdev
->bd_dev
, mode
, NULL
);
106 blkdev_put(tmp
, mode
);
110 static int blk_ioctl_discard(struct block_device
*bdev
, fmode_t mode
,
111 unsigned long arg
, unsigned long flags
)
115 struct request_queue
*q
= bdev_get_queue(bdev
);
116 struct inode
*inode
= bdev
->bd_inode
;
119 if (!(mode
& FMODE_WRITE
))
122 if (!blk_queue_discard(q
))
125 if (copy_from_user(range
, (void __user
*)arg
, sizeof(range
)))
136 if (start
+ len
> i_size_read(bdev
->bd_inode
))
139 filemap_invalidate_lock(inode
->i_mapping
);
140 err
= truncate_bdev_range(bdev
, mode
, start
, start
+ len
- 1);
144 err
= blkdev_issue_discard(bdev
, start
>> 9, len
>> 9,
148 filemap_invalidate_unlock(inode
->i_mapping
);
152 static int blk_ioctl_zeroout(struct block_device
*bdev
, fmode_t mode
,
156 uint64_t start
, end
, len
;
157 struct inode
*inode
= bdev
->bd_inode
;
160 if (!(mode
& FMODE_WRITE
))
163 if (copy_from_user(range
, (void __user
*)arg
, sizeof(range
)))
168 end
= start
+ len
- 1;
174 if (end
>= (uint64_t)i_size_read(bdev
->bd_inode
))
179 /* Invalidate the page cache, including dirty pages */
180 filemap_invalidate_lock(inode
->i_mapping
);
181 err
= truncate_bdev_range(bdev
, mode
, start
, end
);
185 err
= blkdev_issue_zeroout(bdev
, start
>> 9, len
>> 9, GFP_KERNEL
,
186 BLKDEV_ZERO_NOUNMAP
);
189 filemap_invalidate_unlock(inode
->i_mapping
);
193 static int put_ushort(unsigned short __user
*argp
, unsigned short val
)
195 return put_user(val
, argp
);
198 static int put_int(int __user
*argp
, int val
)
200 return put_user(val
, argp
);
203 static int put_uint(unsigned int __user
*argp
, unsigned int val
)
205 return put_user(val
, argp
);
208 static int put_long(long __user
*argp
, long val
)
210 return put_user(val
, argp
);
213 static int put_ulong(unsigned long __user
*argp
, unsigned long val
)
215 return put_user(val
, argp
);
218 static int put_u64(u64 __user
*argp
, u64 val
)
220 return put_user(val
, argp
);
224 static int compat_put_long(compat_long_t __user
*argp
, long val
)
226 return put_user(val
, argp
);
229 static int compat_put_ulong(compat_ulong_t __user
*argp
, compat_ulong_t val
)
231 return put_user(val
, argp
);
237 * This is the equivalent of compat_ptr_ioctl(), to be used by block
238 * drivers that implement only commands that are completely compatible
239 * between 32-bit and 64-bit user space
241 int blkdev_compat_ptr_ioctl(struct block_device
*bdev
, fmode_t mode
,
242 unsigned cmd
, unsigned long arg
)
244 struct gendisk
*disk
= bdev
->bd_disk
;
246 if (disk
->fops
->ioctl
)
247 return disk
->fops
->ioctl(bdev
, mode
, cmd
,
248 (unsigned long)compat_ptr(arg
));
252 EXPORT_SYMBOL(blkdev_compat_ptr_ioctl
);
255 static int blkdev_pr_register(struct block_device
*bdev
,
256 struct pr_registration __user
*arg
)
258 const struct pr_ops
*ops
= bdev
->bd_disk
->fops
->pr_ops
;
259 struct pr_registration reg
;
261 if (!capable(CAP_SYS_ADMIN
))
263 if (!ops
|| !ops
->pr_register
)
265 if (copy_from_user(®
, arg
, sizeof(reg
)))
268 if (reg
.flags
& ~PR_FL_IGNORE_KEY
)
270 return ops
->pr_register(bdev
, reg
.old_key
, reg
.new_key
, reg
.flags
);
273 static int blkdev_pr_reserve(struct block_device
*bdev
,
274 struct pr_reservation __user
*arg
)
276 const struct pr_ops
*ops
= bdev
->bd_disk
->fops
->pr_ops
;
277 struct pr_reservation rsv
;
279 if (!capable(CAP_SYS_ADMIN
))
281 if (!ops
|| !ops
->pr_reserve
)
283 if (copy_from_user(&rsv
, arg
, sizeof(rsv
)))
286 if (rsv
.flags
& ~PR_FL_IGNORE_KEY
)
288 return ops
->pr_reserve(bdev
, rsv
.key
, rsv
.type
, rsv
.flags
);
291 static int blkdev_pr_release(struct block_device
*bdev
,
292 struct pr_reservation __user
*arg
)
294 const struct pr_ops
*ops
= bdev
->bd_disk
->fops
->pr_ops
;
295 struct pr_reservation rsv
;
297 if (!capable(CAP_SYS_ADMIN
))
299 if (!ops
|| !ops
->pr_release
)
301 if (copy_from_user(&rsv
, arg
, sizeof(rsv
)))
306 return ops
->pr_release(bdev
, rsv
.key
, rsv
.type
);
309 static int blkdev_pr_preempt(struct block_device
*bdev
,
310 struct pr_preempt __user
*arg
, bool abort
)
312 const struct pr_ops
*ops
= bdev
->bd_disk
->fops
->pr_ops
;
315 if (!capable(CAP_SYS_ADMIN
))
317 if (!ops
|| !ops
->pr_preempt
)
319 if (copy_from_user(&p
, arg
, sizeof(p
)))
324 return ops
->pr_preempt(bdev
, p
.old_key
, p
.new_key
, p
.type
, abort
);
327 static int blkdev_pr_clear(struct block_device
*bdev
,
328 struct pr_clear __user
*arg
)
330 const struct pr_ops
*ops
= bdev
->bd_disk
->fops
->pr_ops
;
333 if (!capable(CAP_SYS_ADMIN
))
335 if (!ops
|| !ops
->pr_clear
)
337 if (copy_from_user(&c
, arg
, sizeof(c
)))
342 return ops
->pr_clear(bdev
, c
.key
);
345 static int blkdev_flushbuf(struct block_device
*bdev
, fmode_t mode
,
346 unsigned cmd
, unsigned long arg
)
348 if (!capable(CAP_SYS_ADMIN
))
351 invalidate_bdev(bdev
);
355 static int blkdev_roset(struct block_device
*bdev
, fmode_t mode
,
356 unsigned cmd
, unsigned long arg
)
360 if (!capable(CAP_SYS_ADMIN
))
363 if (get_user(n
, (int __user
*)arg
))
365 if (bdev
->bd_disk
->fops
->set_read_only
) {
366 ret
= bdev
->bd_disk
->fops
->set_read_only(bdev
, n
);
370 bdev
->bd_read_only
= n
;
374 static int blkdev_getgeo(struct block_device
*bdev
,
375 struct hd_geometry __user
*argp
)
377 struct gendisk
*disk
= bdev
->bd_disk
;
378 struct hd_geometry geo
;
383 if (!disk
->fops
->getgeo
)
387 * We need to set the startsect first, the driver may
388 * want to override it.
390 memset(&geo
, 0, sizeof(geo
));
391 geo
.start
= get_start_sect(bdev
);
392 ret
= disk
->fops
->getgeo(bdev
, &geo
);
395 if (copy_to_user(argp
, &geo
, sizeof(geo
)))
401 struct compat_hd_geometry
{
403 unsigned char sectors
;
404 unsigned short cylinders
;
408 static int compat_hdio_getgeo(struct block_device
*bdev
,
409 struct compat_hd_geometry __user
*ugeo
)
411 struct gendisk
*disk
= bdev
->bd_disk
;
412 struct hd_geometry geo
;
417 if (!disk
->fops
->getgeo
)
420 memset(&geo
, 0, sizeof(geo
));
422 * We need to set the startsect first, the driver may
423 * want to override it.
425 geo
.start
= get_start_sect(bdev
);
426 ret
= disk
->fops
->getgeo(bdev
, &geo
);
430 ret
= copy_to_user(ugeo
, &geo
, 4);
431 ret
|= put_user(geo
.start
, &ugeo
->start
);
439 /* set the logical block size */
440 static int blkdev_bszset(struct block_device
*bdev
, fmode_t mode
,
445 if (!capable(CAP_SYS_ADMIN
))
449 if (get_user(n
, argp
))
452 if (mode
& FMODE_EXCL
)
453 return set_blocksize(bdev
, n
);
455 if (IS_ERR(blkdev_get_by_dev(bdev
->bd_dev
, mode
| FMODE_EXCL
, &bdev
)))
457 ret
= set_blocksize(bdev
, n
);
458 blkdev_put(bdev
, mode
| FMODE_EXCL
);
464 * Common commands that are handled the same way on native and compat
465 * user space. Note the separate arg/argp parameters that are needed
466 * to deal with the compat_ptr() conversion.
468 static int blkdev_common_ioctl(struct block_device
*bdev
, fmode_t mode
,
469 unsigned cmd
, unsigned long arg
, void __user
*argp
)
471 unsigned int max_sectors
;
475 return blkdev_flushbuf(bdev
, mode
, cmd
, arg
);
477 return blkdev_roset(bdev
, mode
, cmd
, arg
);
479 return blk_ioctl_discard(bdev
, mode
, arg
, 0);
481 return blk_ioctl_discard(bdev
, mode
, arg
,
482 BLKDEV_DISCARD_SECURE
);
484 return blk_ioctl_zeroout(bdev
, mode
, arg
);
486 return put_u64(argp
, bdev
->bd_disk
->diskseq
);
488 return blkdev_report_zones_ioctl(bdev
, mode
, cmd
, arg
);
493 return blkdev_zone_mgmt_ioctl(bdev
, mode
, cmd
, arg
);
495 return put_uint(argp
, bdev_zone_sectors(bdev
));
497 return put_uint(argp
, blkdev_nr_zones(bdev
->bd_disk
));
499 return put_int(argp
, bdev_read_only(bdev
) != 0);
500 case BLKSSZGET
: /* get block device logical block size */
501 return put_int(argp
, bdev_logical_block_size(bdev
));
502 case BLKPBSZGET
: /* get block device physical block size */
503 return put_uint(argp
, bdev_physical_block_size(bdev
));
505 return put_uint(argp
, bdev_io_min(bdev
));
507 return put_uint(argp
, bdev_io_opt(bdev
));
509 return put_int(argp
, bdev_alignment_offset(bdev
));
510 case BLKDISCARDZEROES
:
511 return put_uint(argp
, 0);
513 max_sectors
= min_t(unsigned int, USHRT_MAX
,
514 queue_max_sectors(bdev_get_queue(bdev
)));
515 return put_ushort(argp
, max_sectors
);
517 return put_ushort(argp
, !blk_queue_nonrot(bdev_get_queue(bdev
)));
520 if(!capable(CAP_SYS_ADMIN
))
522 bdev
->bd_disk
->bdi
->ra_pages
= (arg
* 512) / PAGE_SIZE
;
525 return blkdev_reread_part(bdev
, mode
);
528 case BLKTRACETEARDOWN
:
529 return blk_trace_ioctl(bdev
, cmd
, argp
);
530 case IOC_PR_REGISTER
:
531 return blkdev_pr_register(bdev
, argp
);
533 return blkdev_pr_reserve(bdev
, argp
);
535 return blkdev_pr_release(bdev
, argp
);
537 return blkdev_pr_preempt(bdev
, argp
, false);
538 case IOC_PR_PREEMPT_ABORT
:
539 return blkdev_pr_preempt(bdev
, argp
, true);
541 return blkdev_pr_clear(bdev
, argp
);
548 * Always keep this in sync with compat_blkdev_ioctl()
549 * to handle all incompatible commands in both functions.
551 * New commands must be compatible and go into blkdev_common_ioctl
553 int blkdev_ioctl(struct block_device
*bdev
, fmode_t mode
, unsigned cmd
,
558 void __user
*argp
= (void __user
*)arg
;
561 /* These need separate implementations for the data structure */
563 return blkdev_getgeo(bdev
, argp
);
565 return blkpg_ioctl(bdev
, argp
);
567 /* Compat mode returns 32-bit data instead of 'long' */
572 return put_long(argp
,
573 (bdev
->bd_disk
->bdi
->ra_pages
* PAGE_SIZE
) / 512);
575 size
= i_size_read(bdev
->bd_inode
);
576 if ((size
>> 9) > ~0UL)
578 return put_ulong(argp
, size
>> 9);
580 /* The data is compatible, but the command number is different */
581 case BLKBSZGET
: /* get block device soft block size (cf. BLKSSZGET) */
582 return put_int(argp
, block_size(bdev
));
584 return blkdev_bszset(bdev
, mode
, argp
);
586 return put_u64(argp
, i_size_read(bdev
->bd_inode
));
588 /* Incompatible alignment on i386 */
590 return blk_trace_ioctl(bdev
, cmd
, argp
);
595 ret
= blkdev_common_ioctl(bdev
, mode
, cmd
, arg
, argp
);
596 if (ret
!= -ENOIOCTLCMD
)
599 if (!bdev
->bd_disk
->fops
->ioctl
)
601 return bdev
->bd_disk
->fops
->ioctl(bdev
, mode
, cmd
, arg
);
603 EXPORT_SYMBOL_GPL(blkdev_ioctl
); /* for /dev/raw */
607 #define BLKBSZGET_32 _IOR(0x12, 112, int)
608 #define BLKBSZSET_32 _IOW(0x12, 113, int)
609 #define BLKGETSIZE64_32 _IOR(0x12, 114, int)
611 /* Most of the generic ioctls are handled in the normal fallback path.
612 This assumes the blkdev's low level compat_ioctl always returns
613 ENOIOCTLCMD for unknown ioctls. */
614 long compat_blkdev_ioctl(struct file
*file
, unsigned cmd
, unsigned long arg
)
617 void __user
*argp
= compat_ptr(arg
);
618 struct block_device
*bdev
= I_BDEV(file
->f_mapping
->host
);
619 struct gendisk
*disk
= bdev
->bd_disk
;
620 fmode_t mode
= file
->f_mode
;
624 * O_NDELAY can be altered using fcntl(.., F_SETFL, ..), so we have
625 * to updated it before every ioctl.
627 if (file
->f_flags
& O_NDELAY
)
628 mode
|= FMODE_NDELAY
;
630 mode
&= ~FMODE_NDELAY
;
633 /* These need separate implementations for the data structure */
635 return compat_hdio_getgeo(bdev
, argp
);
637 return compat_blkpg_ioctl(bdev
, argp
);
639 /* Compat mode returns 32-bit data instead of 'long' */
644 return compat_put_long(argp
,
645 (bdev
->bd_disk
->bdi
->ra_pages
* PAGE_SIZE
) / 512);
647 size
= i_size_read(bdev
->bd_inode
);
648 if ((size
>> 9) > ~(compat_ulong_t
)0)
650 return compat_put_ulong(argp
, size
>> 9);
652 /* The data is compatible, but the command number is different */
653 case BLKBSZGET_32
: /* get the logical block size (cf. BLKSSZGET) */
654 return put_int(argp
, bdev_logical_block_size(bdev
));
656 return blkdev_bszset(bdev
, mode
, argp
);
657 case BLKGETSIZE64_32
:
658 return put_u64(argp
, i_size_read(bdev
->bd_inode
));
660 /* Incompatible alignment on i386 */
661 case BLKTRACESETUP32
:
662 return blk_trace_ioctl(bdev
, cmd
, argp
);
667 ret
= blkdev_common_ioctl(bdev
, mode
, cmd
, arg
, argp
);
668 if (ret
== -ENOIOCTLCMD
&& disk
->fops
->compat_ioctl
)
669 ret
= disk
->fops
->compat_ioctl(bdev
, mode
, cmd
, arg
);