1 #include <linux/capability.h>
2 #include <linux/blkdev.h>
3 #include <linux/export.h>
5 #include <linux/blkpg.h>
6 #include <linux/hdreg.h>
7 #include <linux/badblocks.h>
8 #include <linux/backing-dev.h>
10 #include <linux/blktrace_api.h>
12 #include <asm/uaccess.h>
14 static int blkpg_ioctl(struct block_device *bdev, struct blkpg_ioctl_arg __user *arg)
16 struct block_device *bdevp;
18 struct hd_struct *part, *lpart;
19 struct blkpg_ioctl_arg a;
20 struct blkpg_partition p;
21 struct disk_part_iter piter;
22 long long start, length;
25 if (!capable(CAP_SYS_ADMIN))
27 if (copy_from_user(&a, arg, sizeof(struct blkpg_ioctl_arg)))
29 if (copy_from_user(&p, a.data, sizeof(struct blkpg_partition)))
32 if (bdev != bdev->bd_contains)
38 case BLKPG_ADD_PARTITION:
40 length = p.length >> 9;
41 /* check for fit in a hd_struct */
42 if (sizeof(sector_t) == sizeof(long) &&
43 sizeof(long long) > sizeof(long)) {
44 long pstart = start, plength = length;
45 if (pstart != start || plength != length
46 || pstart < 0 || plength < 0 || partno > 65535)
50 mutex_lock(&bdev->bd_mutex);
53 disk_part_iter_init(&piter, disk,
54 DISK_PITER_INCL_EMPTY);
55 while ((part = disk_part_iter_next(&piter))) {
56 if (!(start + length <= part->start_sect ||
57 start >= part->start_sect + part->nr_sects)) {
58 disk_part_iter_exit(&piter);
59 mutex_unlock(&bdev->bd_mutex);
63 disk_part_iter_exit(&piter);
66 part = add_partition(disk, partno, start, length,
67 ADDPART_FLAG_NONE, NULL);
68 mutex_unlock(&bdev->bd_mutex);
69 return PTR_ERR_OR_ZERO(part);
70 case BLKPG_DEL_PARTITION:
71 part = disk_get_part(disk, partno);
75 bdevp = bdget(part_devt(part));
80 mutex_lock(&bdevp->bd_mutex);
81 if (bdevp->bd_openers) {
82 mutex_unlock(&bdevp->bd_mutex);
88 invalidate_bdev(bdevp);
90 mutex_lock_nested(&bdev->bd_mutex, 1);
91 delete_partition(disk, partno);
92 mutex_unlock(&bdev->bd_mutex);
93 mutex_unlock(&bdevp->bd_mutex);
97 case BLKPG_RESIZE_PARTITION:
99 /* new length of partition in bytes */
100 length = p.length >> 9;
101 /* check for fit in a hd_struct */
102 if (sizeof(sector_t) == sizeof(long) &&
103 sizeof(long long) > sizeof(long)) {
104 long pstart = start, plength = length;
105 if (pstart != start || plength != length
106 || pstart < 0 || plength < 0)
109 part = disk_get_part(disk, partno);
112 bdevp = bdget(part_devt(part));
117 mutex_lock(&bdevp->bd_mutex);
118 mutex_lock_nested(&bdev->bd_mutex, 1);
119 if (start != part->start_sect) {
120 mutex_unlock(&bdevp->bd_mutex);
121 mutex_unlock(&bdev->bd_mutex);
127 disk_part_iter_init(&piter, disk,
128 DISK_PITER_INCL_EMPTY);
129 while ((lpart = disk_part_iter_next(&piter))) {
130 if (lpart->partno != partno &&
131 !(start + length <= lpart->start_sect ||
132 start >= lpart->start_sect + lpart->nr_sects)
134 disk_part_iter_exit(&piter);
135 mutex_unlock(&bdevp->bd_mutex);
136 mutex_unlock(&bdev->bd_mutex);
142 disk_part_iter_exit(&piter);
143 part_nr_sects_write(part, (sector_t)length);
144 i_size_write(bdevp->bd_inode, p.length);
145 mutex_unlock(&bdevp->bd_mutex);
146 mutex_unlock(&bdev->bd_mutex);
156 * This is an exported API for the block driver, and will not
157 * acquire bd_mutex. This API should be used in case that
158 * caller has held bd_mutex already.
160 int __blkdev_reread_part(struct block_device *bdev)
162 struct gendisk *disk = bdev->bd_disk;
164 if (!disk_part_scan_enabled(disk) || bdev != bdev->bd_contains)
166 if (!capable(CAP_SYS_ADMIN))
169 lockdep_assert_held(&bdev->bd_mutex);
171 return rescan_partitions(disk, bdev);
173 EXPORT_SYMBOL(__blkdev_reread_part);
176 * This is an exported API for the block driver, and will
177 * try to acquire bd_mutex. If bd_mutex has been held already
178 * in current context, please call __blkdev_reread_part().
180 * Make sure the held locks in current context aren't required
181 * in open()/close() handler and I/O path for avoiding ABBA deadlock:
182 * - bd_mutex is held before calling block driver's open/close
184 * - reading partition table may submit I/O to the block device
186 int blkdev_reread_part(struct block_device *bdev)
190 mutex_lock(&bdev->bd_mutex);
191 res = __blkdev_reread_part(bdev);
192 mutex_unlock(&bdev->bd_mutex);
196 EXPORT_SYMBOL(blkdev_reread_part);
198 static int blk_ioctl_discard(struct block_device *bdev, fmode_t mode,
199 unsigned long arg, unsigned long flags)
204 if (!(mode & FMODE_WRITE))
207 if (copy_from_user(range, (void __user *)arg, sizeof(range)))
220 if (start + len > (i_size_read(bdev->bd_inode) >> 9))
222 return blkdev_issue_discard(bdev, start, len, GFP_KERNEL, flags);
225 static int blk_ioctl_zeroout(struct block_device *bdev, fmode_t mode,
231 if (!(mode & FMODE_WRITE))
234 if (copy_from_user(range, (void __user *)arg, sizeof(range)))
247 if (start + len > (i_size_read(bdev->bd_inode) >> 9))
250 return blkdev_issue_zeroout(bdev, start, len, GFP_KERNEL, false);
253 static int put_ushort(unsigned long arg, unsigned short val)
255 return put_user(val, (unsigned short __user *)arg);
258 static int put_int(unsigned long arg, int val)
260 return put_user(val, (int __user *)arg);
263 static int put_uint(unsigned long arg, unsigned int val)
265 return put_user(val, (unsigned int __user *)arg);
268 static int put_long(unsigned long arg, long val)
270 return put_user(val, (long __user *)arg);
273 static int put_ulong(unsigned long arg, unsigned long val)
275 return put_user(val, (unsigned long __user *)arg);
278 static int put_u64(unsigned long arg, u64 val)
280 return put_user(val, (u64 __user *)arg);
283 int __blkdev_driver_ioctl(struct block_device *bdev, fmode_t mode,
284 unsigned cmd, unsigned long arg)
286 struct gendisk *disk = bdev->bd_disk;
288 if (disk->fops->ioctl)
289 return disk->fops->ioctl(bdev, mode, cmd, arg);
294 * For the record: _GPL here is only because somebody decided to slap it
295 * on the previous export. Sheer idiocy, since it wasn't copyrightable
296 * at all and could be open-coded without any exports by anybody who cares.
298 EXPORT_SYMBOL_GPL(__blkdev_driver_ioctl);
300 static int blkdev_pr_register(struct block_device *bdev,
301 struct pr_registration __user *arg)
303 const struct pr_ops *ops = bdev->bd_disk->fops->pr_ops;
304 struct pr_registration reg;
306 if (!capable(CAP_SYS_ADMIN))
308 if (!ops || !ops->pr_register)
310 if (copy_from_user(®, arg, sizeof(reg)))
313 if (reg.flags & ~PR_FL_IGNORE_KEY)
315 return ops->pr_register(bdev, reg.old_key, reg.new_key, reg.flags);
318 static int blkdev_pr_reserve(struct block_device *bdev,
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 (!capable(CAP_SYS_ADMIN))
326 if (!ops || !ops->pr_reserve)
328 if (copy_from_user(&rsv, arg, sizeof(rsv)))
331 if (rsv.flags & ~PR_FL_IGNORE_KEY)
333 return ops->pr_reserve(bdev, rsv.key, rsv.type, rsv.flags);
336 static int blkdev_pr_release(struct block_device *bdev,
337 struct pr_reservation __user *arg)
339 const struct pr_ops *ops = bdev->bd_disk->fops->pr_ops;
340 struct pr_reservation rsv;
342 if (!capable(CAP_SYS_ADMIN))
344 if (!ops || !ops->pr_release)
346 if (copy_from_user(&rsv, arg, sizeof(rsv)))
351 return ops->pr_release(bdev, rsv.key, rsv.type);
354 static int blkdev_pr_preempt(struct block_device *bdev,
355 struct pr_preempt __user *arg, bool abort)
357 const struct pr_ops *ops = bdev->bd_disk->fops->pr_ops;
360 if (!capable(CAP_SYS_ADMIN))
362 if (!ops || !ops->pr_preempt)
364 if (copy_from_user(&p, arg, sizeof(p)))
369 return ops->pr_preempt(bdev, p.old_key, p.new_key, p.type, abort);
372 static int blkdev_pr_clear(struct block_device *bdev,
373 struct pr_clear __user *arg)
375 const struct pr_ops *ops = bdev->bd_disk->fops->pr_ops;
378 if (!capable(CAP_SYS_ADMIN))
380 if (!ops || !ops->pr_clear)
382 if (copy_from_user(&c, arg, sizeof(c)))
387 return ops->pr_clear(bdev, c.key);
391 * Is it an unrecognized ioctl? The correct returns are either
392 * ENOTTY (final) or ENOIOCTLCMD ("I don't know this one, try a
393 * fallback"). ENOIOCTLCMD gets turned into ENOTTY by the ioctl
394 * code before returning.
396 * Confused drivers sometimes return EINVAL, which is wrong. It
397 * means "I understood the ioctl command, but the parameters to
400 * We should aim to just fix the broken drivers, the EINVAL case
403 static inline int is_unrecognized_ioctl(int ret)
405 return ret == -EINVAL ||
411 bool blkdev_dax_capable(struct block_device *bdev)
413 struct gendisk *disk = bdev->bd_disk;
415 if (!disk->fops->direct_access)
419 * If the partition is not aligned on a page boundary, we can't
422 if ((bdev->bd_part->start_sect % (PAGE_SIZE / 512))
423 || (bdev->bd_part->nr_sects % (PAGE_SIZE / 512)))
427 * If the device has known bad blocks, force all I/O through the
428 * driver / page cache.
430 * TODO: support finer grained dax error handling
432 if (disk->bb && disk->bb->count)
438 static int blkdev_daxset(struct block_device *bdev, unsigned long argp)
443 if (!capable(CAP_SYS_ADMIN))
446 if (get_user(arg, (int __user *)(argp)))
449 if (arg == !!(bdev->bd_inode->i_flags & S_DAX))
455 if (arg && !blkdev_dax_capable(bdev))
458 mutex_lock(&bdev->bd_inode->i_mutex);
459 if (bdev->bd_map_count == 0)
460 inode_set_flags(bdev->bd_inode, arg, S_DAX);
463 mutex_unlock(&bdev->bd_inode->i_mutex);
467 static int blkdev_daxset(struct block_device *bdev, int arg)
475 static int blkdev_flushbuf(struct block_device *bdev, fmode_t mode,
476 unsigned cmd, unsigned long arg)
480 if (!capable(CAP_SYS_ADMIN))
483 ret = __blkdev_driver_ioctl(bdev, mode, cmd, arg);
484 if (!is_unrecognized_ioctl(ret))
488 invalidate_bdev(bdev);
492 static int blkdev_roset(struct block_device *bdev, fmode_t mode,
493 unsigned cmd, unsigned long arg)
497 ret = __blkdev_driver_ioctl(bdev, mode, cmd, arg);
498 if (!is_unrecognized_ioctl(ret))
500 if (!capable(CAP_SYS_ADMIN))
502 if (get_user(n, (int __user *)arg))
504 set_device_ro(bdev, n);
508 static int blkdev_getgeo(struct block_device *bdev,
509 struct hd_geometry __user *argp)
511 struct gendisk *disk = bdev->bd_disk;
512 struct hd_geometry geo;
517 if (!disk->fops->getgeo)
521 * We need to set the startsect first, the driver may
522 * want to override it.
524 memset(&geo, 0, sizeof(geo));
525 geo.start = get_start_sect(bdev);
526 ret = disk->fops->getgeo(bdev, &geo);
529 if (copy_to_user(argp, &geo, sizeof(geo)))
534 /* set the logical block size */
535 static int blkdev_bszset(struct block_device *bdev, fmode_t mode,
540 if (!capable(CAP_SYS_ADMIN))
544 if (get_user(n, argp))
547 if (!(mode & FMODE_EXCL)) {
549 if (blkdev_get(bdev, mode | FMODE_EXCL, &bdev) < 0)
553 ret = set_blocksize(bdev, n);
554 if (!(mode & FMODE_EXCL))
555 blkdev_put(bdev, mode | FMODE_EXCL);
560 * always keep this in sync with compat_blkdev_ioctl()
562 int blkdev_ioctl(struct block_device *bdev, fmode_t mode, unsigned cmd,
565 struct backing_dev_info *bdi;
566 void __user *argp = (void __user *)arg;
568 unsigned int max_sectors;
572 return blkdev_flushbuf(bdev, mode, cmd, arg);
574 return blkdev_roset(bdev, mode, cmd, arg);
576 return blk_ioctl_discard(bdev, mode, arg, 0);
578 return blk_ioctl_discard(bdev, mode, arg,
579 BLKDEV_DISCARD_SECURE);
581 return blk_ioctl_zeroout(bdev, mode, arg);
583 return blkdev_getgeo(bdev, argp);
588 bdi = blk_get_backing_dev_info(bdev);
589 return put_long(arg, (bdi->ra_pages * PAGE_CACHE_SIZE) / 512);
591 return put_int(arg, bdev_read_only(bdev) != 0);
592 case BLKBSZGET: /* get block device soft block size (cf. BLKSSZGET) */
593 return put_int(arg, block_size(bdev));
594 case BLKSSZGET: /* get block device logical block size */
595 return put_int(arg, bdev_logical_block_size(bdev));
596 case BLKPBSZGET: /* get block device physical block size */
597 return put_uint(arg, bdev_physical_block_size(bdev));
599 return put_uint(arg, bdev_io_min(bdev));
601 return put_uint(arg, bdev_io_opt(bdev));
603 return put_int(arg, bdev_alignment_offset(bdev));
604 case BLKDISCARDZEROES:
605 return put_uint(arg, bdev_discard_zeroes_data(bdev));
607 max_sectors = min_t(unsigned int, USHRT_MAX,
608 queue_max_sectors(bdev_get_queue(bdev)));
609 return put_ushort(arg, max_sectors);
611 return put_ushort(arg, !blk_queue_nonrot(bdev_get_queue(bdev)));
614 if(!capable(CAP_SYS_ADMIN))
616 bdi = blk_get_backing_dev_info(bdev);
617 bdi->ra_pages = (arg * 512) / PAGE_CACHE_SIZE;
620 return blkdev_bszset(bdev, mode, argp);
622 return blkpg_ioctl(bdev, argp);
624 return blkdev_reread_part(bdev);
626 size = i_size_read(bdev->bd_inode);
627 if ((size >> 9) > ~0UL)
629 return put_ulong(arg, size >> 9);
631 return put_u64(arg, i_size_read(bdev->bd_inode));
635 case BLKTRACETEARDOWN:
636 return blk_trace_ioctl(bdev, cmd, argp);
638 return blkdev_daxset(bdev, arg);
640 return put_int(arg, !!(bdev->bd_inode->i_flags & S_DAX));
642 case IOC_PR_REGISTER:
643 return blkdev_pr_register(bdev, argp);
645 return blkdev_pr_reserve(bdev, argp);
647 return blkdev_pr_release(bdev, argp);
649 return blkdev_pr_preempt(bdev, argp, false);
650 case IOC_PR_PREEMPT_ABORT:
651 return blkdev_pr_preempt(bdev, argp, true);
653 return blkdev_pr_clear(bdev, argp);
655 return __blkdev_driver_ioctl(bdev, mode, cmd, arg);
658 EXPORT_SYMBOL_GPL(blkdev_ioctl);