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 || LLONG_MAX - p.length < p.start)
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,
98 unsigned int bs_mask = bdev_logical_block_size(bdev) - 1;
99 uint64_t range[2], start, len, end;
100 struct bio *prev = NULL, *bio;
101 sector_t sector, nr_sects;
102 struct blk_plug plug;
105 if (!(mode & BLK_OPEN_WRITE))
108 if (!bdev_max_discard_sectors(bdev))
110 if (bdev_read_only(bdev))
113 if (copy_from_user(range, (void __user *)arg, sizeof(range)))
121 if ((start | len) & bs_mask)
124 if (check_add_overflow(start, len, &end) ||
125 end > bdev_nr_bytes(bdev))
128 filemap_invalidate_lock(bdev->bd_mapping);
129 err = truncate_bdev_range(bdev, mode, start, start + len - 1);
133 sector = start >> SECTOR_SHIFT;
134 nr_sects = len >> SECTOR_SHIFT;
136 blk_start_plug(&plug);
138 if (fatal_signal_pending(current)) {
140 bio_await_chain(prev);
144 bio = blk_alloc_discard_bio(bdev, §or, &nr_sects,
148 prev = bio_chain_and_submit(prev, bio);
151 err = submit_bio_wait(prev);
152 if (err == -EOPNOTSUPP)
157 blk_finish_plug(&plug);
159 filemap_invalidate_unlock(bdev->bd_mapping);
163 static int blk_ioctl_secure_erase(struct block_device *bdev, blk_mode_t mode,
170 if (!(mode & BLK_OPEN_WRITE))
172 if (!bdev_max_secure_erase_sectors(bdev))
174 if (copy_from_user(range, argp, sizeof(range)))
179 if ((start & 511) || (len & 511))
181 if (start + len > bdev_nr_bytes(bdev))
184 filemap_invalidate_lock(bdev->bd_mapping);
185 err = truncate_bdev_range(bdev, mode, start, start + len - 1);
187 err = blkdev_issue_secure_erase(bdev, start >> 9, len >> 9,
189 filemap_invalidate_unlock(bdev->bd_mapping);
194 static int blk_ioctl_zeroout(struct block_device *bdev, blk_mode_t mode,
198 uint64_t start, end, len;
201 if (!(mode & BLK_OPEN_WRITE))
204 if (copy_from_user(range, (void __user *)arg, sizeof(range)))
209 end = start + len - 1;
215 if (end >= (uint64_t)bdev_nr_bytes(bdev))
220 /* Invalidate the page cache, including dirty pages */
221 filemap_invalidate_lock(bdev->bd_mapping);
222 err = truncate_bdev_range(bdev, mode, start, end);
226 err = blkdev_issue_zeroout(bdev, start >> 9, len >> 9, GFP_KERNEL,
227 BLKDEV_ZERO_NOUNMAP | BLKDEV_ZERO_KILLABLE);
230 filemap_invalidate_unlock(bdev->bd_mapping);
234 static int put_ushort(unsigned short __user *argp, unsigned short val)
236 return put_user(val, argp);
239 static int put_int(int __user *argp, int val)
241 return put_user(val, argp);
244 static int put_uint(unsigned int __user *argp, unsigned int val)
246 return put_user(val, argp);
249 static int put_long(long __user *argp, long val)
251 return put_user(val, argp);
254 static int put_ulong(unsigned long __user *argp, unsigned long val)
256 return put_user(val, argp);
259 static int put_u64(u64 __user *argp, u64 val)
261 return put_user(val, argp);
265 static int compat_put_long(compat_long_t __user *argp, long val)
267 return put_user(val, argp);
270 static int compat_put_ulong(compat_ulong_t __user *argp, compat_ulong_t val)
272 return put_user(val, argp);
278 * This is the equivalent of compat_ptr_ioctl(), to be used by block
279 * drivers that implement only commands that are completely compatible
280 * between 32-bit and 64-bit user space
282 int blkdev_compat_ptr_ioctl(struct block_device *bdev, blk_mode_t mode,
283 unsigned cmd, unsigned long arg)
285 struct gendisk *disk = bdev->bd_disk;
287 if (disk->fops->ioctl)
288 return disk->fops->ioctl(bdev, mode, cmd,
289 (unsigned long)compat_ptr(arg));
293 EXPORT_SYMBOL(blkdev_compat_ptr_ioctl);
296 static bool blkdev_pr_allowed(struct block_device *bdev, blk_mode_t mode)
298 /* no sense to make reservations for partitions */
299 if (bdev_is_partition(bdev))
302 if (capable(CAP_SYS_ADMIN))
305 * Only allow unprivileged reservations if the file descriptor is open
308 return mode & BLK_OPEN_WRITE;
311 static int blkdev_pr_register(struct block_device *bdev, blk_mode_t mode,
312 struct pr_registration __user *arg)
314 const struct pr_ops *ops = bdev->bd_disk->fops->pr_ops;
315 struct pr_registration reg;
317 if (!blkdev_pr_allowed(bdev, mode))
319 if (!ops || !ops->pr_register)
321 if (copy_from_user(®, arg, sizeof(reg)))
324 if (reg.flags & ~PR_FL_IGNORE_KEY)
326 return ops->pr_register(bdev, reg.old_key, reg.new_key, reg.flags);
329 static int blkdev_pr_reserve(struct block_device *bdev, blk_mode_t mode,
330 struct pr_reservation __user *arg)
332 const struct pr_ops *ops = bdev->bd_disk->fops->pr_ops;
333 struct pr_reservation rsv;
335 if (!blkdev_pr_allowed(bdev, mode))
337 if (!ops || !ops->pr_reserve)
339 if (copy_from_user(&rsv, arg, sizeof(rsv)))
342 if (rsv.flags & ~PR_FL_IGNORE_KEY)
344 return ops->pr_reserve(bdev, rsv.key, rsv.type, rsv.flags);
347 static int blkdev_pr_release(struct block_device *bdev, blk_mode_t mode,
348 struct pr_reservation __user *arg)
350 const struct pr_ops *ops = bdev->bd_disk->fops->pr_ops;
351 struct pr_reservation rsv;
353 if (!blkdev_pr_allowed(bdev, mode))
355 if (!ops || !ops->pr_release)
357 if (copy_from_user(&rsv, arg, sizeof(rsv)))
362 return ops->pr_release(bdev, rsv.key, rsv.type);
365 static int blkdev_pr_preempt(struct block_device *bdev, blk_mode_t mode,
366 struct pr_preempt __user *arg, bool abort)
368 const struct pr_ops *ops = bdev->bd_disk->fops->pr_ops;
371 if (!blkdev_pr_allowed(bdev, mode))
373 if (!ops || !ops->pr_preempt)
375 if (copy_from_user(&p, arg, sizeof(p)))
380 return ops->pr_preempt(bdev, p.old_key, p.new_key, p.type, abort);
383 static int blkdev_pr_clear(struct block_device *bdev, blk_mode_t mode,
384 struct pr_clear __user *arg)
386 const struct pr_ops *ops = bdev->bd_disk->fops->pr_ops;
389 if (!blkdev_pr_allowed(bdev, mode))
391 if (!ops || !ops->pr_clear)
393 if (copy_from_user(&c, arg, sizeof(c)))
398 return ops->pr_clear(bdev, c.key);
401 static int blkdev_flushbuf(struct block_device *bdev, unsigned cmd,
404 if (!capable(CAP_SYS_ADMIN))
407 mutex_lock(&bdev->bd_holder_lock);
408 if (bdev->bd_holder_ops && bdev->bd_holder_ops->sync)
409 bdev->bd_holder_ops->sync(bdev);
411 mutex_unlock(&bdev->bd_holder_lock);
415 invalidate_bdev(bdev);
419 static int blkdev_roset(struct block_device *bdev, unsigned cmd,
424 if (!capable(CAP_SYS_ADMIN))
427 if (get_user(n, (int __user *)arg))
429 if (bdev->bd_disk->fops->set_read_only) {
430 ret = bdev->bd_disk->fops->set_read_only(bdev, n);
435 bdev_set_flag(bdev, BD_READ_ONLY);
437 bdev_clear_flag(bdev, BD_READ_ONLY);
441 static int blkdev_getgeo(struct block_device *bdev,
442 struct hd_geometry __user *argp)
444 struct gendisk *disk = bdev->bd_disk;
445 struct hd_geometry geo;
450 if (!disk->fops->getgeo)
454 * We need to set the startsect first, the driver may
455 * want to override it.
457 memset(&geo, 0, sizeof(geo));
458 geo.start = get_start_sect(bdev);
459 ret = disk->fops->getgeo(bdev, &geo);
462 if (copy_to_user(argp, &geo, sizeof(geo)))
468 struct compat_hd_geometry {
470 unsigned char sectors;
471 unsigned short cylinders;
475 static int compat_hdio_getgeo(struct block_device *bdev,
476 struct compat_hd_geometry __user *ugeo)
478 struct gendisk *disk = bdev->bd_disk;
479 struct hd_geometry geo;
484 if (!disk->fops->getgeo)
487 memset(&geo, 0, sizeof(geo));
489 * We need to set the startsect first, the driver may
490 * want to override it.
492 geo.start = get_start_sect(bdev);
493 ret = disk->fops->getgeo(bdev, &geo);
497 ret = copy_to_user(ugeo, &geo, 4);
498 ret |= put_user(geo.start, &ugeo->start);
506 /* set the logical block size */
507 static int blkdev_bszset(struct file *file, blk_mode_t mode,
510 // this one might be file_inode(file)->i_rdev - a rare valid
511 // use of file_inode() for those.
512 dev_t dev = I_BDEV(file->f_mapping->host)->bd_dev;
513 struct file *excl_file;
516 if (!capable(CAP_SYS_ADMIN))
520 if (get_user(n, argp))
523 if (mode & BLK_OPEN_EXCL)
524 return set_blocksize(file, n);
526 excl_file = bdev_file_open_by_dev(dev, mode, &dev, NULL);
527 if (IS_ERR(excl_file))
529 ret = set_blocksize(excl_file, n);
535 * Common commands that are handled the same way on native and compat
536 * user space. Note the separate arg/argp parameters that are needed
537 * to deal with the compat_ptr() conversion.
539 static int blkdev_common_ioctl(struct block_device *bdev, blk_mode_t mode,
540 unsigned int cmd, unsigned long arg,
543 unsigned int max_sectors;
547 return blkdev_flushbuf(bdev, cmd, arg);
549 return blkdev_roset(bdev, cmd, arg);
551 return blk_ioctl_discard(bdev, mode, arg);
553 return blk_ioctl_secure_erase(bdev, mode, argp);
555 return blk_ioctl_zeroout(bdev, mode, arg);
557 return put_u64(argp, bdev->bd_disk->diskseq);
559 return blkdev_report_zones_ioctl(bdev, cmd, arg);
564 return blkdev_zone_mgmt_ioctl(bdev, mode, cmd, arg);
566 return put_uint(argp, bdev_zone_sectors(bdev));
568 return put_uint(argp, bdev_nr_zones(bdev));
570 return put_int(argp, bdev_read_only(bdev) != 0);
571 case BLKSSZGET: /* get block device logical block size */
572 return put_int(argp, bdev_logical_block_size(bdev));
573 case BLKPBSZGET: /* get block device physical block size */
574 return put_uint(argp, bdev_physical_block_size(bdev));
576 return put_uint(argp, bdev_io_min(bdev));
578 return put_uint(argp, bdev_io_opt(bdev));
580 return put_int(argp, bdev_alignment_offset(bdev));
581 case BLKDISCARDZEROES:
582 return put_uint(argp, 0);
584 max_sectors = min_t(unsigned int, USHRT_MAX,
585 queue_max_sectors(bdev_get_queue(bdev)));
586 return put_ushort(argp, max_sectors);
588 return put_ushort(argp, !bdev_nonrot(bdev));
591 if(!capable(CAP_SYS_ADMIN))
593 bdev->bd_disk->bdi->ra_pages = (arg * 512) / PAGE_SIZE;
596 if (!capable(CAP_SYS_ADMIN))
598 if (bdev_is_partition(bdev))
600 return disk_scan_partitions(bdev->bd_disk,
601 mode | BLK_OPEN_STRICT_SCAN);
604 case BLKTRACETEARDOWN:
605 return blk_trace_ioctl(bdev, cmd, argp);
606 case IOC_PR_REGISTER:
607 return blkdev_pr_register(bdev, mode, argp);
609 return blkdev_pr_reserve(bdev, mode, argp);
611 return blkdev_pr_release(bdev, mode, argp);
613 return blkdev_pr_preempt(bdev, mode, argp, false);
614 case IOC_PR_PREEMPT_ABORT:
615 return blkdev_pr_preempt(bdev, mode, argp, true);
617 return blkdev_pr_clear(bdev, mode, argp);
624 * Always keep this in sync with compat_blkdev_ioctl()
625 * to handle all incompatible commands in both functions.
627 * New commands must be compatible and go into blkdev_common_ioctl
629 long blkdev_ioctl(struct file *file, unsigned cmd, unsigned long arg)
631 struct block_device *bdev = I_BDEV(file->f_mapping->host);
632 void __user *argp = (void __user *)arg;
633 blk_mode_t mode = file_to_blk_mode(file);
637 /* These need separate implementations for the data structure */
639 return blkdev_getgeo(bdev, argp);
641 return blkpg_ioctl(bdev, argp);
643 /* Compat mode returns 32-bit data instead of 'long' */
648 return put_long(argp,
649 (bdev->bd_disk->bdi->ra_pages * PAGE_SIZE) / 512);
651 if (bdev_nr_sectors(bdev) > ~0UL)
653 return put_ulong(argp, bdev_nr_sectors(bdev));
655 /* The data is compatible, but the command number is different */
656 case BLKBSZGET: /* get block device soft block size (cf. BLKSSZGET) */
657 return put_int(argp, block_size(bdev));
659 return blkdev_bszset(file, mode, argp);
661 return put_u64(argp, bdev_nr_bytes(bdev));
663 /* Incompatible alignment on i386 */
665 return blk_trace_ioctl(bdev, cmd, argp);
670 ret = blkdev_common_ioctl(bdev, mode, cmd, arg, argp);
671 if (ret != -ENOIOCTLCMD)
674 if (!bdev->bd_disk->fops->ioctl)
676 return bdev->bd_disk->fops->ioctl(bdev, mode, cmd, arg);
681 #define BLKBSZGET_32 _IOR(0x12, 112, int)
682 #define BLKBSZSET_32 _IOW(0x12, 113, int)
683 #define BLKGETSIZE64_32 _IOR(0x12, 114, int)
685 /* Most of the generic ioctls are handled in the normal fallback path.
686 This assumes the blkdev's low level compat_ioctl always returns
687 ENOIOCTLCMD for unknown ioctls. */
688 long compat_blkdev_ioctl(struct file *file, unsigned cmd, unsigned long arg)
691 void __user *argp = compat_ptr(arg);
692 struct block_device *bdev = I_BDEV(file->f_mapping->host);
693 struct gendisk *disk = bdev->bd_disk;
694 blk_mode_t mode = file_to_blk_mode(file);
697 /* These need separate implementations for the data structure */
699 return compat_hdio_getgeo(bdev, argp);
701 return compat_blkpg_ioctl(bdev, argp);
703 /* Compat mode returns 32-bit data instead of 'long' */
708 return compat_put_long(argp,
709 (bdev->bd_disk->bdi->ra_pages * PAGE_SIZE) / 512);
711 if (bdev_nr_sectors(bdev) > ~(compat_ulong_t)0)
713 return compat_put_ulong(argp, bdev_nr_sectors(bdev));
715 /* The data is compatible, but the command number is different */
716 case BLKBSZGET_32: /* get the logical block size (cf. BLKSSZGET) */
717 return put_int(argp, bdev_logical_block_size(bdev));
719 return blkdev_bszset(file, mode, argp);
720 case BLKGETSIZE64_32:
721 return put_u64(argp, bdev_nr_bytes(bdev));
723 /* Incompatible alignment on i386 */
724 case BLKTRACESETUP32:
725 return blk_trace_ioctl(bdev, cmd, argp);
730 ret = blkdev_common_ioctl(bdev, mode, cmd, arg, argp);
731 if (ret == -ENOIOCTLCMD && disk->fops->compat_ioctl)
732 ret = disk->fops->compat_ioctl(bdev, mode, cmd, arg);