1 // SPDX-License-Identifier: GPL-2.0
3 * Copyright (C) 2007 Oracle. All rights reserved.
6 #include <linux/kernel.h>
8 #include <linux/file.h>
10 #include <linux/fsnotify.h>
11 #include <linux/pagemap.h>
12 #include <linux/highmem.h>
13 #include <linux/time.h>
14 #include <linux/string.h>
15 #include <linux/backing-dev.h>
16 #include <linux/mount.h>
17 #include <linux/namei.h>
18 #include <linux/writeback.h>
19 #include <linux/compat.h>
20 #include <linux/security.h>
21 #include <linux/xattr.h>
23 #include <linux/slab.h>
24 #include <linux/blkdev.h>
25 #include <linux/uuid.h>
26 #include <linux/btrfs.h>
27 #include <linux/uaccess.h>
28 #include <linux/iversion.h>
29 #include <linux/fileattr.h>
30 #include <linux/fsverity.h>
31 #include <linux/sched/xacct.h>
35 #include "transaction.h"
36 #include "btrfs_inode.h"
37 #include "print-tree.h"
41 #include "rcu-string.h"
43 #include "dev-replace.h"
48 #include "compression.h"
49 #include "space-info.h"
50 #include "delalloc-space.h"
51 #include "block-group.h"
55 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
56 * structures are incorrect, as the timespec structure from userspace
57 * is 4 bytes too small. We define these alternatives here to teach
58 * the kernel about the 32-bit struct packing.
60 struct btrfs_ioctl_timespec_32 {
63 } __attribute__ ((__packed__));
65 struct btrfs_ioctl_received_subvol_args_32 {
66 char uuid[BTRFS_UUID_SIZE]; /* in */
67 __u64 stransid; /* in */
68 __u64 rtransid; /* out */
69 struct btrfs_ioctl_timespec_32 stime; /* in */
70 struct btrfs_ioctl_timespec_32 rtime; /* out */
72 __u64 reserved[16]; /* in */
73 } __attribute__ ((__packed__));
75 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
76 struct btrfs_ioctl_received_subvol_args_32)
79 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
80 struct btrfs_ioctl_send_args_32 {
81 __s64 send_fd; /* in */
82 __u64 clone_sources_count; /* in */
83 compat_uptr_t clone_sources; /* in */
84 __u64 parent_root; /* in */
86 __u32 version; /* in */
87 __u8 reserved[28]; /* in */
88 } __attribute__ ((__packed__));
90 #define BTRFS_IOC_SEND_32 _IOW(BTRFS_IOCTL_MAGIC, 38, \
91 struct btrfs_ioctl_send_args_32)
93 struct btrfs_ioctl_encoded_io_args_32 {
95 compat_ulong_t iovcnt;
100 __u64 unencoded_offset;
106 #define BTRFS_IOC_ENCODED_READ_32 _IOR(BTRFS_IOCTL_MAGIC, 64, \
107 struct btrfs_ioctl_encoded_io_args_32)
108 #define BTRFS_IOC_ENCODED_WRITE_32 _IOW(BTRFS_IOCTL_MAGIC, 64, \
109 struct btrfs_ioctl_encoded_io_args_32)
112 /* Mask out flags that are inappropriate for the given type of inode. */
113 static unsigned int btrfs_mask_fsflags_for_type(struct inode *inode,
116 if (S_ISDIR(inode->i_mode))
118 else if (S_ISREG(inode->i_mode))
119 return flags & ~FS_DIRSYNC_FL;
121 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
125 * Export internal inode flags to the format expected by the FS_IOC_GETFLAGS
128 static unsigned int btrfs_inode_flags_to_fsflags(struct btrfs_inode *binode)
130 unsigned int iflags = 0;
131 u32 flags = binode->flags;
132 u32 ro_flags = binode->ro_flags;
134 if (flags & BTRFS_INODE_SYNC)
135 iflags |= FS_SYNC_FL;
136 if (flags & BTRFS_INODE_IMMUTABLE)
137 iflags |= FS_IMMUTABLE_FL;
138 if (flags & BTRFS_INODE_APPEND)
139 iflags |= FS_APPEND_FL;
140 if (flags & BTRFS_INODE_NODUMP)
141 iflags |= FS_NODUMP_FL;
142 if (flags & BTRFS_INODE_NOATIME)
143 iflags |= FS_NOATIME_FL;
144 if (flags & BTRFS_INODE_DIRSYNC)
145 iflags |= FS_DIRSYNC_FL;
146 if (flags & BTRFS_INODE_NODATACOW)
147 iflags |= FS_NOCOW_FL;
148 if (ro_flags & BTRFS_INODE_RO_VERITY)
149 iflags |= FS_VERITY_FL;
151 if (flags & BTRFS_INODE_NOCOMPRESS)
152 iflags |= FS_NOCOMP_FL;
153 else if (flags & BTRFS_INODE_COMPRESS)
154 iflags |= FS_COMPR_FL;
160 * Update inode->i_flags based on the btrfs internal flags.
162 void btrfs_sync_inode_flags_to_i_flags(struct inode *inode)
164 struct btrfs_inode *binode = BTRFS_I(inode);
165 unsigned int new_fl = 0;
167 if (binode->flags & BTRFS_INODE_SYNC)
169 if (binode->flags & BTRFS_INODE_IMMUTABLE)
170 new_fl |= S_IMMUTABLE;
171 if (binode->flags & BTRFS_INODE_APPEND)
173 if (binode->flags & BTRFS_INODE_NOATIME)
175 if (binode->flags & BTRFS_INODE_DIRSYNC)
177 if (binode->ro_flags & BTRFS_INODE_RO_VERITY)
180 set_mask_bits(&inode->i_flags,
181 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC |
186 * Check if @flags are a supported and valid set of FS_*_FL flags and that
187 * the old and new flags are not conflicting
189 static int check_fsflags(unsigned int old_flags, unsigned int flags)
191 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
192 FS_NOATIME_FL | FS_NODUMP_FL | \
193 FS_SYNC_FL | FS_DIRSYNC_FL | \
194 FS_NOCOMP_FL | FS_COMPR_FL |
198 /* COMPR and NOCOMP on new/old are valid */
199 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
202 if ((flags & FS_COMPR_FL) && (flags & FS_NOCOW_FL))
205 /* NOCOW and compression options are mutually exclusive */
206 if ((old_flags & FS_NOCOW_FL) && (flags & (FS_COMPR_FL | FS_NOCOMP_FL)))
208 if ((flags & FS_NOCOW_FL) && (old_flags & (FS_COMPR_FL | FS_NOCOMP_FL)))
214 static int check_fsflags_compatible(struct btrfs_fs_info *fs_info,
217 if (btrfs_is_zoned(fs_info) && (flags & FS_NOCOW_FL))
224 * Set flags/xflags from the internal inode flags. The remaining items of
225 * fsxattr are zeroed.
227 int btrfs_fileattr_get(struct dentry *dentry, struct fileattr *fa)
229 struct btrfs_inode *binode = BTRFS_I(d_inode(dentry));
231 fileattr_fill_flags(fa, btrfs_inode_flags_to_fsflags(binode));
235 int btrfs_fileattr_set(struct user_namespace *mnt_userns,
236 struct dentry *dentry, struct fileattr *fa)
238 struct inode *inode = d_inode(dentry);
239 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
240 struct btrfs_inode *binode = BTRFS_I(inode);
241 struct btrfs_root *root = binode->root;
242 struct btrfs_trans_handle *trans;
243 unsigned int fsflags, old_fsflags;
245 const char *comp = NULL;
248 if (btrfs_root_readonly(root))
251 if (fileattr_has_fsx(fa))
254 fsflags = btrfs_mask_fsflags_for_type(inode, fa->flags);
255 old_fsflags = btrfs_inode_flags_to_fsflags(binode);
256 ret = check_fsflags(old_fsflags, fsflags);
260 ret = check_fsflags_compatible(fs_info, fsflags);
264 binode_flags = binode->flags;
265 if (fsflags & FS_SYNC_FL)
266 binode_flags |= BTRFS_INODE_SYNC;
268 binode_flags &= ~BTRFS_INODE_SYNC;
269 if (fsflags & FS_IMMUTABLE_FL)
270 binode_flags |= BTRFS_INODE_IMMUTABLE;
272 binode_flags &= ~BTRFS_INODE_IMMUTABLE;
273 if (fsflags & FS_APPEND_FL)
274 binode_flags |= BTRFS_INODE_APPEND;
276 binode_flags &= ~BTRFS_INODE_APPEND;
277 if (fsflags & FS_NODUMP_FL)
278 binode_flags |= BTRFS_INODE_NODUMP;
280 binode_flags &= ~BTRFS_INODE_NODUMP;
281 if (fsflags & FS_NOATIME_FL)
282 binode_flags |= BTRFS_INODE_NOATIME;
284 binode_flags &= ~BTRFS_INODE_NOATIME;
286 /* If coming from FS_IOC_FSSETXATTR then skip unconverted flags */
287 if (!fa->flags_valid) {
288 /* 1 item for the inode */
289 trans = btrfs_start_transaction(root, 1);
291 return PTR_ERR(trans);
295 if (fsflags & FS_DIRSYNC_FL)
296 binode_flags |= BTRFS_INODE_DIRSYNC;
298 binode_flags &= ~BTRFS_INODE_DIRSYNC;
299 if (fsflags & FS_NOCOW_FL) {
300 if (S_ISREG(inode->i_mode)) {
302 * It's safe to turn csums off here, no extents exist.
303 * Otherwise we want the flag to reflect the real COW
304 * status of the file and will not set it.
306 if (inode->i_size == 0)
307 binode_flags |= BTRFS_INODE_NODATACOW |
308 BTRFS_INODE_NODATASUM;
310 binode_flags |= BTRFS_INODE_NODATACOW;
314 * Revert back under same assumptions as above
316 if (S_ISREG(inode->i_mode)) {
317 if (inode->i_size == 0)
318 binode_flags &= ~(BTRFS_INODE_NODATACOW |
319 BTRFS_INODE_NODATASUM);
321 binode_flags &= ~BTRFS_INODE_NODATACOW;
326 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
327 * flag may be changed automatically if compression code won't make
330 if (fsflags & FS_NOCOMP_FL) {
331 binode_flags &= ~BTRFS_INODE_COMPRESS;
332 binode_flags |= BTRFS_INODE_NOCOMPRESS;
333 } else if (fsflags & FS_COMPR_FL) {
335 if (IS_SWAPFILE(inode))
338 binode_flags |= BTRFS_INODE_COMPRESS;
339 binode_flags &= ~BTRFS_INODE_NOCOMPRESS;
341 comp = btrfs_compress_type2str(fs_info->compress_type);
342 if (!comp || comp[0] == 0)
343 comp = btrfs_compress_type2str(BTRFS_COMPRESS_ZLIB);
345 binode_flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
352 trans = btrfs_start_transaction(root, 3);
354 return PTR_ERR(trans);
357 ret = btrfs_set_prop(trans, inode, "btrfs.compression", comp,
360 btrfs_abort_transaction(trans, ret);
364 ret = btrfs_set_prop(trans, inode, "btrfs.compression", NULL,
366 if (ret && ret != -ENODATA) {
367 btrfs_abort_transaction(trans, ret);
373 binode->flags = binode_flags;
374 btrfs_sync_inode_flags_to_i_flags(inode);
375 inode_inc_iversion(inode);
376 inode->i_ctime = current_time(inode);
377 ret = btrfs_update_inode(trans, root, BTRFS_I(inode));
380 btrfs_end_transaction(trans);
385 * Start exclusive operation @type, return true on success
387 bool btrfs_exclop_start(struct btrfs_fs_info *fs_info,
388 enum btrfs_exclusive_operation type)
392 spin_lock(&fs_info->super_lock);
393 if (fs_info->exclusive_operation == BTRFS_EXCLOP_NONE) {
394 fs_info->exclusive_operation = type;
397 spin_unlock(&fs_info->super_lock);
403 * Conditionally allow to enter the exclusive operation in case it's compatible
404 * with the running one. This must be paired with btrfs_exclop_start_unlock and
405 * btrfs_exclop_finish.
408 * - the same type is already running
409 * - when trying to add a device and balance has been paused
410 * - not BTRFS_EXCLOP_NONE - this is intentionally incompatible and the caller
411 * must check the condition first that would allow none -> @type
413 bool btrfs_exclop_start_try_lock(struct btrfs_fs_info *fs_info,
414 enum btrfs_exclusive_operation type)
416 spin_lock(&fs_info->super_lock);
417 if (fs_info->exclusive_operation == type ||
418 (fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE_PAUSED &&
419 type == BTRFS_EXCLOP_DEV_ADD))
422 spin_unlock(&fs_info->super_lock);
426 void btrfs_exclop_start_unlock(struct btrfs_fs_info *fs_info)
428 spin_unlock(&fs_info->super_lock);
431 void btrfs_exclop_finish(struct btrfs_fs_info *fs_info)
433 spin_lock(&fs_info->super_lock);
434 WRITE_ONCE(fs_info->exclusive_operation, BTRFS_EXCLOP_NONE);
435 spin_unlock(&fs_info->super_lock);
436 sysfs_notify(&fs_info->fs_devices->fsid_kobj, NULL, "exclusive_operation");
439 void btrfs_exclop_balance(struct btrfs_fs_info *fs_info,
440 enum btrfs_exclusive_operation op)
443 case BTRFS_EXCLOP_BALANCE_PAUSED:
444 spin_lock(&fs_info->super_lock);
445 ASSERT(fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE ||
446 fs_info->exclusive_operation == BTRFS_EXCLOP_DEV_ADD);
447 fs_info->exclusive_operation = BTRFS_EXCLOP_BALANCE_PAUSED;
448 spin_unlock(&fs_info->super_lock);
450 case BTRFS_EXCLOP_BALANCE:
451 spin_lock(&fs_info->super_lock);
452 ASSERT(fs_info->exclusive_operation == BTRFS_EXCLOP_BALANCE_PAUSED);
453 fs_info->exclusive_operation = BTRFS_EXCLOP_BALANCE;
454 spin_unlock(&fs_info->super_lock);
458 "invalid exclop balance operation %d requested", op);
462 static int btrfs_ioctl_getversion(struct inode *inode, int __user *arg)
464 return put_user(inode->i_generation, arg);
467 static noinline int btrfs_ioctl_fitrim(struct btrfs_fs_info *fs_info,
470 struct btrfs_device *device;
471 struct request_queue *q;
472 struct fstrim_range range;
473 u64 minlen = ULLONG_MAX;
477 if (!capable(CAP_SYS_ADMIN))
481 * btrfs_trim_block_group() depends on space cache, which is not
482 * available in zoned filesystem. So, disallow fitrim on a zoned
483 * filesystem for now.
485 if (btrfs_is_zoned(fs_info))
489 * If the fs is mounted with nologreplay, which requires it to be
490 * mounted in RO mode as well, we can not allow discard on free space
491 * inside block groups, because log trees refer to extents that are not
492 * pinned in a block group's free space cache (pinning the extents is
493 * precisely the first phase of replaying a log tree).
495 if (btrfs_test_opt(fs_info, NOLOGREPLAY))
499 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
503 q = bdev_get_queue(device->bdev);
504 if (blk_queue_discard(q)) {
506 minlen = min_t(u64, q->limits.discard_granularity,
514 if (copy_from_user(&range, arg, sizeof(range)))
518 * NOTE: Don't truncate the range using super->total_bytes. Bytenr of
519 * block group is in the logical address space, which can be any
520 * sectorsize aligned bytenr in the range [0, U64_MAX].
522 if (range.len < fs_info->sb->s_blocksize)
525 range.minlen = max(range.minlen, minlen);
526 ret = btrfs_trim_fs(fs_info, &range);
530 if (copy_to_user(arg, &range, sizeof(range)))
536 int __pure btrfs_is_empty_uuid(u8 *uuid)
540 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
547 static noinline int create_subvol(struct user_namespace *mnt_userns,
548 struct inode *dir, struct dentry *dentry,
549 const char *name, int namelen,
550 struct btrfs_qgroup_inherit *inherit)
552 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
553 struct btrfs_trans_handle *trans;
554 struct btrfs_key key;
555 struct btrfs_root_item *root_item;
556 struct btrfs_inode_item *inode_item;
557 struct extent_buffer *leaf;
558 struct btrfs_root *root = BTRFS_I(dir)->root;
559 struct btrfs_root *new_root;
560 struct btrfs_block_rsv block_rsv;
561 struct timespec64 cur_time = current_time(dir);
568 root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
572 ret = btrfs_get_free_objectid(fs_info->tree_root, &objectid);
577 * Don't create subvolume whose level is not zero. Or qgroup will be
578 * screwed up since it assumes subvolume qgroup's level to be 0.
580 if (btrfs_qgroup_level(objectid)) {
585 ret = get_anon_bdev(&anon_dev);
589 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
591 * The same as the snapshot creation, please see the comment
592 * of create_snapshot().
594 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv, 8, false);
598 trans = btrfs_start_transaction(root, 0);
600 ret = PTR_ERR(trans);
601 btrfs_subvolume_release_metadata(root, &block_rsv);
604 trans->block_rsv = &block_rsv;
605 trans->bytes_reserved = block_rsv.size;
607 ret = btrfs_qgroup_inherit(trans, 0, objectid, inherit);
611 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0,
612 BTRFS_NESTING_NORMAL);
618 btrfs_mark_buffer_dirty(leaf);
620 inode_item = &root_item->inode;
621 btrfs_set_stack_inode_generation(inode_item, 1);
622 btrfs_set_stack_inode_size(inode_item, 3);
623 btrfs_set_stack_inode_nlink(inode_item, 1);
624 btrfs_set_stack_inode_nbytes(inode_item,
626 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
628 btrfs_set_root_flags(root_item, 0);
629 btrfs_set_root_limit(root_item, 0);
630 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
632 btrfs_set_root_bytenr(root_item, leaf->start);
633 btrfs_set_root_generation(root_item, trans->transid);
634 btrfs_set_root_level(root_item, 0);
635 btrfs_set_root_refs(root_item, 1);
636 btrfs_set_root_used(root_item, leaf->len);
637 btrfs_set_root_last_snapshot(root_item, 0);
639 btrfs_set_root_generation_v2(root_item,
640 btrfs_root_generation(root_item));
641 generate_random_guid(root_item->uuid);
642 btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
643 btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
644 root_item->ctime = root_item->otime;
645 btrfs_set_root_ctransid(root_item, trans->transid);
646 btrfs_set_root_otransid(root_item, trans->transid);
648 btrfs_tree_unlock(leaf);
650 btrfs_set_root_dirid(root_item, BTRFS_FIRST_FREE_OBJECTID);
652 key.objectid = objectid;
654 key.type = BTRFS_ROOT_ITEM_KEY;
655 ret = btrfs_insert_root(trans, fs_info->tree_root, &key,
659 * Since we don't abort the transaction in this case, free the
660 * tree block so that we don't leak space and leave the
661 * filesystem in an inconsistent state (an extent item in the
662 * extent tree with a backreference for a root that does not
665 btrfs_tree_lock(leaf);
666 btrfs_clean_tree_block(leaf);
667 btrfs_tree_unlock(leaf);
668 btrfs_free_tree_block(trans, objectid, leaf, 0, 1);
669 free_extent_buffer(leaf);
673 free_extent_buffer(leaf);
676 key.offset = (u64)-1;
677 new_root = btrfs_get_new_fs_root(fs_info, objectid, anon_dev);
678 if (IS_ERR(new_root)) {
679 ret = PTR_ERR(new_root);
680 btrfs_abort_transaction(trans, ret);
683 /* anon_dev is owned by new_root now. */
686 ret = btrfs_record_root_in_trans(trans, new_root);
688 btrfs_put_root(new_root);
689 btrfs_abort_transaction(trans, ret);
693 ret = btrfs_create_subvol_root(trans, new_root, root, mnt_userns);
694 btrfs_put_root(new_root);
696 /* We potentially lose an unused inode item here */
697 btrfs_abort_transaction(trans, ret);
702 * insert the directory item
704 ret = btrfs_set_inode_index(BTRFS_I(dir), &index);
706 btrfs_abort_transaction(trans, ret);
710 ret = btrfs_insert_dir_item(trans, name, namelen, BTRFS_I(dir), &key,
711 BTRFS_FT_DIR, index);
713 btrfs_abort_transaction(trans, ret);
717 btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2);
718 ret = btrfs_update_inode(trans, root, BTRFS_I(dir));
720 btrfs_abort_transaction(trans, ret);
724 ret = btrfs_add_root_ref(trans, objectid, root->root_key.objectid,
725 btrfs_ino(BTRFS_I(dir)), index, name, namelen);
727 btrfs_abort_transaction(trans, ret);
731 ret = btrfs_uuid_tree_add(trans, root_item->uuid,
732 BTRFS_UUID_KEY_SUBVOL, objectid);
734 btrfs_abort_transaction(trans, ret);
737 trans->block_rsv = NULL;
738 trans->bytes_reserved = 0;
739 btrfs_subvolume_release_metadata(root, &block_rsv);
742 btrfs_end_transaction(trans);
744 ret = btrfs_commit_transaction(trans);
747 inode = btrfs_lookup_dentry(dir, dentry);
749 return PTR_ERR(inode);
750 d_instantiate(dentry, inode);
754 free_anon_bdev(anon_dev);
760 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
761 struct dentry *dentry, bool readonly,
762 struct btrfs_qgroup_inherit *inherit)
764 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
766 struct btrfs_pending_snapshot *pending_snapshot;
767 struct btrfs_trans_handle *trans;
770 /* We do not support snapshotting right now. */
771 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
773 "extent tree v2 doesn't support snapshotting yet");
777 if (!test_bit(BTRFS_ROOT_SHAREABLE, &root->state))
780 if (atomic_read(&root->nr_swapfiles)) {
782 "cannot snapshot subvolume with active swapfile");
786 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
787 if (!pending_snapshot)
790 ret = get_anon_bdev(&pending_snapshot->anon_dev);
793 pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
795 pending_snapshot->path = btrfs_alloc_path();
796 if (!pending_snapshot->root_item || !pending_snapshot->path) {
801 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
802 BTRFS_BLOCK_RSV_TEMP);
804 * 1 - parent dir inode
807 * 2 - root ref/backref
808 * 1 - root of snapshot
811 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
812 &pending_snapshot->block_rsv, 8,
817 pending_snapshot->dentry = dentry;
818 pending_snapshot->root = root;
819 pending_snapshot->readonly = readonly;
820 pending_snapshot->dir = dir;
821 pending_snapshot->inherit = inherit;
823 trans = btrfs_start_transaction(root, 0);
825 ret = PTR_ERR(trans);
829 trans->pending_snapshot = pending_snapshot;
831 ret = btrfs_commit_transaction(trans);
835 ret = pending_snapshot->error;
839 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
843 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
845 ret = PTR_ERR(inode);
849 d_instantiate(dentry, inode);
851 pending_snapshot->anon_dev = 0;
853 /* Prevent double freeing of anon_dev */
854 if (ret && pending_snapshot->snap)
855 pending_snapshot->snap->anon_dev = 0;
856 btrfs_put_root(pending_snapshot->snap);
857 btrfs_subvolume_release_metadata(root, &pending_snapshot->block_rsv);
859 if (pending_snapshot->anon_dev)
860 free_anon_bdev(pending_snapshot->anon_dev);
861 kfree(pending_snapshot->root_item);
862 btrfs_free_path(pending_snapshot->path);
863 kfree(pending_snapshot);
868 /* copy of may_delete in fs/namei.c()
869 * Check whether we can remove a link victim from directory dir, check
870 * whether the type of victim is right.
871 * 1. We can't do it if dir is read-only (done in permission())
872 * 2. We should have write and exec permissions on dir
873 * 3. We can't remove anything from append-only dir
874 * 4. We can't do anything with immutable dir (done in permission())
875 * 5. If the sticky bit on dir is set we should either
876 * a. be owner of dir, or
877 * b. be owner of victim, or
878 * c. have CAP_FOWNER capability
879 * 6. If the victim is append-only or immutable we can't do anything with
880 * links pointing to it.
881 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
882 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
883 * 9. We can't remove a root or mountpoint.
884 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
885 * nfs_async_unlink().
888 static int btrfs_may_delete(struct user_namespace *mnt_userns,
889 struct inode *dir, struct dentry *victim, int isdir)
893 if (d_really_is_negative(victim))
896 BUG_ON(d_inode(victim->d_parent) != dir);
897 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
899 error = inode_permission(mnt_userns, dir, MAY_WRITE | MAY_EXEC);
904 if (check_sticky(mnt_userns, dir, d_inode(victim)) ||
905 IS_APPEND(d_inode(victim)) || IS_IMMUTABLE(d_inode(victim)) ||
906 IS_SWAPFILE(d_inode(victim)))
909 if (!d_is_dir(victim))
913 } else if (d_is_dir(victim))
917 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
922 /* copy of may_create in fs/namei.c() */
923 static inline int btrfs_may_create(struct user_namespace *mnt_userns,
924 struct inode *dir, struct dentry *child)
926 if (d_really_is_positive(child))
930 if (!fsuidgid_has_mapping(dir->i_sb, mnt_userns))
932 return inode_permission(mnt_userns, dir, MAY_WRITE | MAY_EXEC);
936 * Create a new subvolume below @parent. This is largely modeled after
937 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
938 * inside this filesystem so it's quite a bit simpler.
940 static noinline int btrfs_mksubvol(const struct path *parent,
941 struct user_namespace *mnt_userns,
942 const char *name, int namelen,
943 struct btrfs_root *snap_src,
945 struct btrfs_qgroup_inherit *inherit)
947 struct inode *dir = d_inode(parent->dentry);
948 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
949 struct dentry *dentry;
952 error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
956 dentry = lookup_one(mnt_userns, name, parent->dentry, namelen);
957 error = PTR_ERR(dentry);
961 error = btrfs_may_create(mnt_userns, dir, dentry);
966 * even if this name doesn't exist, we may get hash collisions.
967 * check for them now when we can safely fail
969 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
975 down_read(&fs_info->subvol_sem);
977 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
981 error = create_snapshot(snap_src, dir, dentry, readonly, inherit);
983 error = create_subvol(mnt_userns, dir, dentry, name, namelen, inherit);
986 fsnotify_mkdir(dir, dentry);
988 up_read(&fs_info->subvol_sem);
992 btrfs_inode_unlock(dir, 0);
996 static noinline int btrfs_mksnapshot(const struct path *parent,
997 struct user_namespace *mnt_userns,
998 const char *name, int namelen,
999 struct btrfs_root *root,
1001 struct btrfs_qgroup_inherit *inherit)
1004 bool snapshot_force_cow = false;
1007 * Force new buffered writes to reserve space even when NOCOW is
1008 * possible. This is to avoid later writeback (running dealloc) to
1009 * fallback to COW mode and unexpectedly fail with ENOSPC.
1011 btrfs_drew_read_lock(&root->snapshot_lock);
1013 ret = btrfs_start_delalloc_snapshot(root, false);
1018 * All previous writes have started writeback in NOCOW mode, so now
1019 * we force future writes to fallback to COW mode during snapshot
1022 atomic_inc(&root->snapshot_force_cow);
1023 snapshot_force_cow = true;
1025 btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1);
1027 ret = btrfs_mksubvol(parent, mnt_userns, name, namelen,
1028 root, readonly, inherit);
1030 if (snapshot_force_cow)
1031 atomic_dec(&root->snapshot_force_cow);
1032 btrfs_drew_read_unlock(&root->snapshot_lock);
1037 * Defrag specific helper to get an extent map.
1039 * Differences between this and btrfs_get_extent() are:
1041 * - No extent_map will be added to inode->extent_tree
1042 * To reduce memory usage in the long run.
1044 * - Extra optimization to skip file extents older than @newer_than
1045 * By using btrfs_search_forward() we can skip entire file ranges that
1046 * have extents created in past transactions, because btrfs_search_forward()
1047 * will not visit leaves and nodes with a generation smaller than given
1048 * minimal generation threshold (@newer_than).
1050 * Return valid em if we find a file extent matching the requirement.
1051 * Return NULL if we can not find a file extent matching the requirement.
1053 * Return ERR_PTR() for error.
1055 static struct extent_map *defrag_get_extent(struct btrfs_inode *inode,
1056 u64 start, u64 newer_than)
1058 struct btrfs_root *root = inode->root;
1059 struct btrfs_file_extent_item *fi;
1060 struct btrfs_path path = { 0 };
1061 struct extent_map *em;
1062 struct btrfs_key key;
1063 u64 ino = btrfs_ino(inode);
1066 em = alloc_extent_map();
1073 key.type = BTRFS_EXTENT_DATA_KEY;
1077 ret = btrfs_search_forward(root, &key, &path, newer_than);
1080 /* Can't find anything newer */
1084 ret = btrfs_search_slot(NULL, root, &key, &path, 0, 0);
1088 if (path.slots[0] >= btrfs_header_nritems(path.nodes[0])) {
1090 * If btrfs_search_slot() makes path to point beyond nritems,
1091 * we should not have an empty leaf, as this inode must at
1092 * least have its INODE_ITEM.
1094 ASSERT(btrfs_header_nritems(path.nodes[0]));
1095 path.slots[0] = btrfs_header_nritems(path.nodes[0]) - 1;
1097 btrfs_item_key_to_cpu(path.nodes[0], &key, path.slots[0]);
1098 /* Perfect match, no need to go one slot back */
1099 if (key.objectid == ino && key.type == BTRFS_EXTENT_DATA_KEY &&
1100 key.offset == start)
1103 /* We didn't find a perfect match, needs to go one slot back */
1104 if (path.slots[0] > 0) {
1105 btrfs_item_key_to_cpu(path.nodes[0], &key, path.slots[0]);
1106 if (key.objectid == ino && key.type == BTRFS_EXTENT_DATA_KEY)
1111 /* Iterate through the path to find a file extent covering @start */
1115 if (path.slots[0] >= btrfs_header_nritems(path.nodes[0]))
1118 btrfs_item_key_to_cpu(path.nodes[0], &key, path.slots[0]);
1121 * We may go one slot back to INODE_REF/XATTR item, then
1122 * need to go forward until we reach an EXTENT_DATA.
1123 * But we should still has the correct ino as key.objectid.
1125 if (WARN_ON(key.objectid < ino) || key.type < BTRFS_EXTENT_DATA_KEY)
1128 /* It's beyond our target range, definitely not extent found */
1129 if (key.objectid > ino || key.type > BTRFS_EXTENT_DATA_KEY)
1133 * | |<- File extent ->|
1136 * This means there is a hole between start and key.offset.
1138 if (key.offset > start) {
1140 em->orig_start = start;
1141 em->block_start = EXTENT_MAP_HOLE;
1142 em->len = key.offset - start;
1146 fi = btrfs_item_ptr(path.nodes[0], path.slots[0],
1147 struct btrfs_file_extent_item);
1148 extent_end = btrfs_file_extent_end(&path);
1151 * |<- file extent ->| |
1154 * We haven't reached start, search next slot.
1156 if (extent_end <= start)
1159 /* Now this extent covers @start, convert it to em */
1160 btrfs_extent_item_to_extent_map(inode, &path, fi, false, em);
1163 ret = btrfs_next_item(root, &path);
1169 btrfs_release_path(&path);
1173 btrfs_release_path(&path);
1174 free_extent_map(em);
1178 btrfs_release_path(&path);
1179 free_extent_map(em);
1180 return ERR_PTR(ret);
1183 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start,
1184 u64 newer_than, bool locked)
1186 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1187 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1188 struct extent_map *em;
1189 const u32 sectorsize = BTRFS_I(inode)->root->fs_info->sectorsize;
1192 * hopefully we have this extent in the tree already, try without
1193 * the full extent lock
1195 read_lock(&em_tree->lock);
1196 em = lookup_extent_mapping(em_tree, start, sectorsize);
1197 read_unlock(&em_tree->lock);
1200 * We can get a merged extent, in that case, we need to re-search
1201 * tree to get the original em for defrag.
1203 * If @newer_than is 0 or em::generation < newer_than, we can trust
1204 * this em, as either we don't care about the generation, or the
1205 * merged extent map will be rejected anyway.
1207 if (em && test_bit(EXTENT_FLAG_MERGED, &em->flags) &&
1208 newer_than && em->generation >= newer_than) {
1209 free_extent_map(em);
1214 struct extent_state *cached = NULL;
1215 u64 end = start + sectorsize - 1;
1217 /* get the big lock and read metadata off disk */
1219 lock_extent_bits(io_tree, start, end, &cached);
1220 em = defrag_get_extent(BTRFS_I(inode), start, newer_than);
1222 unlock_extent_cached(io_tree, start, end, &cached);
1231 static u32 get_extent_max_capacity(const struct extent_map *em)
1233 if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
1234 return BTRFS_MAX_COMPRESSED;
1235 return BTRFS_MAX_EXTENT_SIZE;
1238 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em,
1239 u32 extent_thresh, u64 newer_than, bool locked)
1241 struct extent_map *next;
1244 /* this is the last extent */
1245 if (em->start + em->len >= i_size_read(inode))
1249 * Here we need to pass @newer_then when checking the next extent, or
1250 * we will hit a case we mark current extent for defrag, but the next
1251 * one will not be a target.
1252 * This will just cause extra IO without really reducing the fragments.
1254 next = defrag_lookup_extent(inode, em->start + em->len, newer_than, locked);
1255 /* No more em or hole */
1256 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1258 if (test_bit(EXTENT_FLAG_PREALLOC, &next->flags))
1261 * If the next extent is at its max capacity, defragging current extent
1262 * makes no sense, as the total number of extents won't change.
1264 if (next->len >= get_extent_max_capacity(em))
1266 /* Skip older extent */
1267 if (next->generation < newer_than)
1269 /* Also check extent size */
1270 if (next->len >= extent_thresh)
1275 free_extent_map(next);
1280 * Prepare one page to be defragged.
1284 * - Returned page is locked and has been set up properly.
1285 * - No ordered extent exists in the page.
1286 * - The page is uptodate.
1288 * NOTE: Caller should also wait for page writeback after the cluster is
1289 * prepared, here we don't do writeback wait for each page.
1291 static struct page *defrag_prepare_one_page(struct btrfs_inode *inode,
1294 struct address_space *mapping = inode->vfs_inode.i_mapping;
1295 gfp_t mask = btrfs_alloc_write_mask(mapping);
1296 u64 page_start = (u64)index << PAGE_SHIFT;
1297 u64 page_end = page_start + PAGE_SIZE - 1;
1298 struct extent_state *cached_state = NULL;
1303 page = find_or_create_page(mapping, index, mask);
1305 return ERR_PTR(-ENOMEM);
1308 * Since we can defragment files opened read-only, we can encounter
1309 * transparent huge pages here (see CONFIG_READ_ONLY_THP_FOR_FS). We
1310 * can't do I/O using huge pages yet, so return an error for now.
1311 * Filesystem transparent huge pages are typically only used for
1312 * executables that explicitly enable them, so this isn't very
1315 if (PageCompound(page)) {
1318 return ERR_PTR(-ETXTBSY);
1321 ret = set_page_extent_mapped(page);
1325 return ERR_PTR(ret);
1328 /* Wait for any existing ordered extent in the range */
1330 struct btrfs_ordered_extent *ordered;
1332 lock_extent_bits(&inode->io_tree, page_start, page_end, &cached_state);
1333 ordered = btrfs_lookup_ordered_range(inode, page_start, PAGE_SIZE);
1334 unlock_extent_cached(&inode->io_tree, page_start, page_end,
1340 btrfs_start_ordered_extent(ordered, 1);
1341 btrfs_put_ordered_extent(ordered);
1344 * We unlocked the page above, so we need check if it was
1347 if (page->mapping != mapping || !PagePrivate(page)) {
1355 * Now the page range has no ordered extent any more. Read the page to
1358 if (!PageUptodate(page)) {
1359 btrfs_readpage(NULL, page);
1361 if (page->mapping != mapping || !PagePrivate(page)) {
1366 if (!PageUptodate(page)) {
1369 return ERR_PTR(-EIO);
1375 struct defrag_target_range {
1376 struct list_head list;
1382 * Collect all valid target extents.
1384 * @start: file offset to lookup
1385 * @len: length to lookup
1386 * @extent_thresh: file extent size threshold, any extent size >= this value
1388 * @newer_than: only defrag extents newer than this value
1389 * @do_compress: whether the defrag is doing compression
1390 * if true, @extent_thresh will be ignored and all regular
1391 * file extents meeting @newer_than will be targets.
1392 * @locked: if the range has already held extent lock
1393 * @target_list: list of targets file extents
1395 static int defrag_collect_targets(struct btrfs_inode *inode,
1396 u64 start, u64 len, u32 extent_thresh,
1397 u64 newer_than, bool do_compress,
1398 bool locked, struct list_head *target_list,
1399 u64 *last_scanned_ret)
1401 bool last_is_target = false;
1405 while (cur < start + len) {
1406 struct extent_map *em;
1407 struct defrag_target_range *new;
1408 bool next_mergeable = true;
1411 last_is_target = false;
1412 em = defrag_lookup_extent(&inode->vfs_inode, cur,
1413 newer_than, locked);
1417 /* Skip hole/inline/preallocated extents */
1418 if (em->block_start >= EXTENT_MAP_LAST_BYTE ||
1419 test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
1422 /* Skip older extent */
1423 if (em->generation < newer_than)
1426 /* This em is under writeback, no need to defrag */
1427 if (em->generation == (u64)-1)
1431 * Our start offset might be in the middle of an existing extent
1432 * map, so take that into account.
1434 range_len = em->len - (cur - em->start);
1436 * If this range of the extent map is already flagged for delalloc,
1439 * 1) We could deadlock later, when trying to reserve space for
1440 * delalloc, because in case we can't immediately reserve space
1441 * the flusher can start delalloc and wait for the respective
1442 * ordered extents to complete. The deadlock would happen
1443 * because we do the space reservation while holding the range
1444 * locked, and starting writeback, or finishing an ordered
1445 * extent, requires locking the range;
1447 * 2) If there's delalloc there, it means there's dirty pages for
1448 * which writeback has not started yet (we clean the delalloc
1449 * flag when starting writeback and after creating an ordered
1450 * extent). If we mark pages in an adjacent range for defrag,
1451 * then we will have a larger contiguous range for delalloc,
1452 * very likely resulting in a larger extent after writeback is
1453 * triggered (except in a case of free space fragmentation).
1455 if (test_range_bit(&inode->io_tree, cur, cur + range_len - 1,
1456 EXTENT_DELALLOC, 0, NULL))
1460 * For do_compress case, we want to compress all valid file
1461 * extents, thus no @extent_thresh or mergeable check.
1466 /* Skip too large extent */
1467 if (range_len >= extent_thresh)
1471 * Skip extents already at its max capacity, this is mostly for
1472 * compressed extents, which max cap is only 128K.
1474 if (em->len >= get_extent_max_capacity(em))
1477 next_mergeable = defrag_check_next_extent(&inode->vfs_inode, em,
1478 extent_thresh, newer_than, locked);
1479 if (!next_mergeable) {
1480 struct defrag_target_range *last;
1482 /* Empty target list, no way to merge with last entry */
1483 if (list_empty(target_list))
1485 last = list_entry(target_list->prev,
1486 struct defrag_target_range, list);
1487 /* Not mergeable with last entry */
1488 if (last->start + last->len != cur)
1491 /* Mergeable, fall through to add it to @target_list. */
1495 last_is_target = true;
1496 range_len = min(extent_map_end(em), start + len) - cur;
1498 * This one is a good target, check if it can be merged into
1499 * last range of the target list.
1501 if (!list_empty(target_list)) {
1502 struct defrag_target_range *last;
1504 last = list_entry(target_list->prev,
1505 struct defrag_target_range, list);
1506 ASSERT(last->start + last->len <= cur);
1507 if (last->start + last->len == cur) {
1508 /* Mergeable, enlarge the last entry */
1509 last->len += range_len;
1512 /* Fall through to allocate a new entry */
1515 /* Allocate new defrag_target_range */
1516 new = kmalloc(sizeof(*new), GFP_NOFS);
1518 free_extent_map(em);
1523 new->len = range_len;
1524 list_add_tail(&new->list, target_list);
1527 cur = extent_map_end(em);
1528 free_extent_map(em);
1531 struct defrag_target_range *entry;
1532 struct defrag_target_range *tmp;
1534 list_for_each_entry_safe(entry, tmp, target_list, list) {
1535 list_del_init(&entry->list);
1539 if (!ret && last_scanned_ret) {
1541 * If the last extent is not a target, the caller can skip to
1542 * the end of that extent.
1543 * Otherwise, we can only go the end of the specified range.
1545 if (!last_is_target)
1546 *last_scanned_ret = max(cur, *last_scanned_ret);
1548 *last_scanned_ret = max(start + len, *last_scanned_ret);
1553 #define CLUSTER_SIZE (SZ_256K)
1554 static_assert(IS_ALIGNED(CLUSTER_SIZE, PAGE_SIZE));
1557 * Defrag one contiguous target range.
1559 * @inode: target inode
1560 * @target: target range to defrag
1561 * @pages: locked pages covering the defrag range
1562 * @nr_pages: number of locked pages
1564 * Caller should ensure:
1566 * - Pages are prepared
1567 * Pages should be locked, no ordered extent in the pages range,
1570 * - Extent bits are locked
1572 static int defrag_one_locked_target(struct btrfs_inode *inode,
1573 struct defrag_target_range *target,
1574 struct page **pages, int nr_pages,
1575 struct extent_state **cached_state)
1577 struct btrfs_fs_info *fs_info = inode->root->fs_info;
1578 struct extent_changeset *data_reserved = NULL;
1579 const u64 start = target->start;
1580 const u64 len = target->len;
1581 unsigned long last_index = (start + len - 1) >> PAGE_SHIFT;
1582 unsigned long start_index = start >> PAGE_SHIFT;
1583 unsigned long first_index = page_index(pages[0]);
1587 ASSERT(last_index - first_index + 1 <= nr_pages);
1589 ret = btrfs_delalloc_reserve_space(inode, &data_reserved, start, len);
1592 clear_extent_bit(&inode->io_tree, start, start + len - 1,
1593 EXTENT_DELALLOC | EXTENT_DO_ACCOUNTING |
1594 EXTENT_DEFRAG, 0, 0, cached_state);
1595 set_extent_defrag(&inode->io_tree, start, start + len - 1, cached_state);
1597 /* Update the page status */
1598 for (i = start_index - first_index; i <= last_index - first_index; i++) {
1599 ClearPageChecked(pages[i]);
1600 btrfs_page_clamp_set_dirty(fs_info, pages[i], start, len);
1602 btrfs_delalloc_release_extents(inode, len);
1603 extent_changeset_free(data_reserved);
1608 static int defrag_one_range(struct btrfs_inode *inode, u64 start, u32 len,
1609 u32 extent_thresh, u64 newer_than, bool do_compress,
1610 u64 *last_scanned_ret)
1612 struct extent_state *cached_state = NULL;
1613 struct defrag_target_range *entry;
1614 struct defrag_target_range *tmp;
1615 LIST_HEAD(target_list);
1616 struct page **pages;
1617 const u32 sectorsize = inode->root->fs_info->sectorsize;
1618 u64 last_index = (start + len - 1) >> PAGE_SHIFT;
1619 u64 start_index = start >> PAGE_SHIFT;
1620 unsigned int nr_pages = last_index - start_index + 1;
1624 ASSERT(nr_pages <= CLUSTER_SIZE / PAGE_SIZE);
1625 ASSERT(IS_ALIGNED(start, sectorsize) && IS_ALIGNED(len, sectorsize));
1627 pages = kcalloc(nr_pages, sizeof(struct page *), GFP_NOFS);
1631 /* Prepare all pages */
1632 for (i = 0; i < nr_pages; i++) {
1633 pages[i] = defrag_prepare_one_page(inode, start_index + i);
1634 if (IS_ERR(pages[i])) {
1635 ret = PTR_ERR(pages[i]);
1640 for (i = 0; i < nr_pages; i++)
1641 wait_on_page_writeback(pages[i]);
1643 /* Lock the pages range */
1644 lock_extent_bits(&inode->io_tree, start_index << PAGE_SHIFT,
1645 (last_index << PAGE_SHIFT) + PAGE_SIZE - 1,
1648 * Now we have a consistent view about the extent map, re-check
1649 * which range really needs to be defragged.
1651 * And this time we have extent locked already, pass @locked = true
1652 * so that we won't relock the extent range and cause deadlock.
1654 ret = defrag_collect_targets(inode, start, len, extent_thresh,
1655 newer_than, do_compress, true,
1656 &target_list, last_scanned_ret);
1660 list_for_each_entry(entry, &target_list, list) {
1661 ret = defrag_one_locked_target(inode, entry, pages, nr_pages,
1667 list_for_each_entry_safe(entry, tmp, &target_list, list) {
1668 list_del_init(&entry->list);
1672 unlock_extent_cached(&inode->io_tree, start_index << PAGE_SHIFT,
1673 (last_index << PAGE_SHIFT) + PAGE_SIZE - 1,
1676 for (i = 0; i < nr_pages; i++) {
1678 unlock_page(pages[i]);
1686 static int defrag_one_cluster(struct btrfs_inode *inode,
1687 struct file_ra_state *ra,
1688 u64 start, u32 len, u32 extent_thresh,
1689 u64 newer_than, bool do_compress,
1690 unsigned long *sectors_defragged,
1691 unsigned long max_sectors,
1692 u64 *last_scanned_ret)
1694 const u32 sectorsize = inode->root->fs_info->sectorsize;
1695 struct defrag_target_range *entry;
1696 struct defrag_target_range *tmp;
1697 LIST_HEAD(target_list);
1700 ret = defrag_collect_targets(inode, start, len, extent_thresh,
1701 newer_than, do_compress, false,
1702 &target_list, NULL);
1706 list_for_each_entry(entry, &target_list, list) {
1707 u32 range_len = entry->len;
1709 /* Reached or beyond the limit */
1710 if (max_sectors && *sectors_defragged >= max_sectors) {
1716 range_len = min_t(u32, range_len,
1717 (max_sectors - *sectors_defragged) * sectorsize);
1720 * If defrag_one_range() has updated last_scanned_ret,
1721 * our range may already be invalid (e.g. hole punched).
1722 * Skip if our range is before last_scanned_ret, as there is
1723 * no need to defrag the range anymore.
1725 if (entry->start + range_len <= *last_scanned_ret)
1729 page_cache_sync_readahead(inode->vfs_inode.i_mapping,
1730 ra, NULL, entry->start >> PAGE_SHIFT,
1731 ((entry->start + range_len - 1) >> PAGE_SHIFT) -
1732 (entry->start >> PAGE_SHIFT) + 1);
1734 * Here we may not defrag any range if holes are punched before
1735 * we locked the pages.
1736 * But that's fine, it only affects the @sectors_defragged
1739 ret = defrag_one_range(inode, entry->start, range_len,
1740 extent_thresh, newer_than, do_compress,
1744 *sectors_defragged += range_len >>
1745 inode->root->fs_info->sectorsize_bits;
1748 list_for_each_entry_safe(entry, tmp, &target_list, list) {
1749 list_del_init(&entry->list);
1753 *last_scanned_ret = max(*last_scanned_ret, start + len);
1758 * Entry point to file defragmentation.
1760 * @inode: inode to be defragged
1761 * @ra: readahead state (can be NUL)
1762 * @range: defrag options including range and flags
1763 * @newer_than: minimum transid to defrag
1764 * @max_to_defrag: max number of sectors to be defragged, if 0, the whole inode
1765 * will be defragged.
1767 * Return <0 for error.
1768 * Return >=0 for the number of sectors defragged, and range->start will be updated
1769 * to indicate the file offset where next defrag should be started at.
1770 * (Mostly for autodefrag, which sets @max_to_defrag thus we may exit early without
1771 * defragging all the range).
1773 int btrfs_defrag_file(struct inode *inode, struct file_ra_state *ra,
1774 struct btrfs_ioctl_defrag_range_args *range,
1775 u64 newer_than, unsigned long max_to_defrag)
1777 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1778 unsigned long sectors_defragged = 0;
1779 u64 isize = i_size_read(inode);
1782 bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS;
1783 bool ra_allocated = false;
1784 int compress_type = BTRFS_COMPRESS_ZLIB;
1786 u32 extent_thresh = range->extent_thresh;
1787 pgoff_t start_index;
1792 if (range->start >= isize)
1796 if (range->compress_type >= BTRFS_NR_COMPRESS_TYPES)
1798 if (range->compress_type)
1799 compress_type = range->compress_type;
1802 if (extent_thresh == 0)
1803 extent_thresh = SZ_256K;
1805 if (range->start + range->len > range->start) {
1806 /* Got a specific range */
1807 last_byte = min(isize, range->start + range->len);
1809 /* Defrag until file end */
1813 /* Align the range */
1814 cur = round_down(range->start, fs_info->sectorsize);
1815 last_byte = round_up(last_byte, fs_info->sectorsize) - 1;
1818 * If we were not given a ra, allocate a readahead context. As
1819 * readahead is just an optimization, defrag will work without it so
1820 * we don't error out.
1823 ra_allocated = true;
1824 ra = kzalloc(sizeof(*ra), GFP_KERNEL);
1826 file_ra_state_init(ra, inode->i_mapping);
1830 * Make writeback start from the beginning of the range, so that the
1831 * defrag range can be written sequentially.
1833 start_index = cur >> PAGE_SHIFT;
1834 if (start_index < inode->i_mapping->writeback_index)
1835 inode->i_mapping->writeback_index = start_index;
1837 while (cur < last_byte) {
1838 const unsigned long prev_sectors_defragged = sectors_defragged;
1839 u64 last_scanned = cur;
1842 if (btrfs_defrag_cancelled(fs_info)) {
1847 /* We want the cluster end at page boundary when possible */
1848 cluster_end = (((cur >> PAGE_SHIFT) +
1849 (SZ_256K >> PAGE_SHIFT)) << PAGE_SHIFT) - 1;
1850 cluster_end = min(cluster_end, last_byte);
1852 btrfs_inode_lock(inode, 0);
1853 if (IS_SWAPFILE(inode)) {
1855 btrfs_inode_unlock(inode, 0);
1858 if (!(inode->i_sb->s_flags & SB_ACTIVE)) {
1859 btrfs_inode_unlock(inode, 0);
1863 BTRFS_I(inode)->defrag_compress = compress_type;
1864 ret = defrag_one_cluster(BTRFS_I(inode), ra, cur,
1865 cluster_end + 1 - cur, extent_thresh,
1866 newer_than, do_compress, §ors_defragged,
1867 max_to_defrag, &last_scanned);
1869 if (sectors_defragged > prev_sectors_defragged)
1870 balance_dirty_pages_ratelimited(inode->i_mapping);
1872 btrfs_inode_unlock(inode, 0);
1875 cur = max(cluster_end + 1, last_scanned);
1886 * Update range.start for autodefrag, this will indicate where to start
1890 if (sectors_defragged) {
1892 * We have defragged some sectors, for compression case they
1893 * need to be written back immediately.
1895 if (range->flags & BTRFS_DEFRAG_RANGE_START_IO) {
1896 filemap_flush(inode->i_mapping);
1897 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1898 &BTRFS_I(inode)->runtime_flags))
1899 filemap_flush(inode->i_mapping);
1901 if (range->compress_type == BTRFS_COMPRESS_LZO)
1902 btrfs_set_fs_incompat(fs_info, COMPRESS_LZO);
1903 else if (range->compress_type == BTRFS_COMPRESS_ZSTD)
1904 btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD);
1905 ret = sectors_defragged;
1908 btrfs_inode_lock(inode, 0);
1909 BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE;
1910 btrfs_inode_unlock(inode, 0);
1916 * Try to start exclusive operation @type or cancel it if it's running.
1919 * 0 - normal mode, newly claimed op started
1920 * >0 - normal mode, something else is running,
1921 * return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS to user space
1922 * ECANCELED - cancel mode, successful cancel
1923 * ENOTCONN - cancel mode, operation not running anymore
1925 static int exclop_start_or_cancel_reloc(struct btrfs_fs_info *fs_info,
1926 enum btrfs_exclusive_operation type, bool cancel)
1929 /* Start normal op */
1930 if (!btrfs_exclop_start(fs_info, type))
1931 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1932 /* Exclusive operation is now claimed */
1936 /* Cancel running op */
1937 if (btrfs_exclop_start_try_lock(fs_info, type)) {
1939 * This blocks any exclop finish from setting it to NONE, so we
1940 * request cancellation. Either it runs and we will wait for it,
1941 * or it has finished and no waiting will happen.
1943 atomic_inc(&fs_info->reloc_cancel_req);
1944 btrfs_exclop_start_unlock(fs_info);
1946 if (test_bit(BTRFS_FS_RELOC_RUNNING, &fs_info->flags))
1947 wait_on_bit(&fs_info->flags, BTRFS_FS_RELOC_RUNNING,
1948 TASK_INTERRUPTIBLE);
1953 /* Something else is running or none */
1957 static noinline int btrfs_ioctl_resize(struct file *file,
1960 BTRFS_DEV_LOOKUP_ARGS(args);
1961 struct inode *inode = file_inode(file);
1962 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1966 struct btrfs_root *root = BTRFS_I(inode)->root;
1967 struct btrfs_ioctl_vol_args *vol_args;
1968 struct btrfs_trans_handle *trans;
1969 struct btrfs_device *device = NULL;
1972 char *devstr = NULL;
1977 if (!capable(CAP_SYS_ADMIN))
1980 ret = mnt_want_write_file(file);
1985 * Read the arguments before checking exclusivity to be able to
1986 * distinguish regular resize and cancel
1988 vol_args = memdup_user(arg, sizeof(*vol_args));
1989 if (IS_ERR(vol_args)) {
1990 ret = PTR_ERR(vol_args);
1993 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1994 sizestr = vol_args->name;
1995 cancel = (strcmp("cancel", sizestr) == 0);
1996 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_RESIZE, cancel);
1999 /* Exclusive operation is now claimed */
2001 devstr = strchr(sizestr, ':');
2003 sizestr = devstr + 1;
2005 devstr = vol_args->name;
2006 ret = kstrtoull(devstr, 10, &devid);
2013 btrfs_info(fs_info, "resizing devid %llu", devid);
2017 device = btrfs_find_device(fs_info->fs_devices, &args);
2019 btrfs_info(fs_info, "resizer unable to find device %llu",
2025 if (!test_bit(BTRFS_DEV_STATE_WRITEABLE, &device->dev_state)) {
2027 "resizer unable to apply on readonly device %llu",
2033 if (!strcmp(sizestr, "max"))
2034 new_size = bdev_nr_bytes(device->bdev);
2036 if (sizestr[0] == '-') {
2039 } else if (sizestr[0] == '+') {
2043 new_size = memparse(sizestr, &retptr);
2044 if (*retptr != '\0' || new_size == 0) {
2050 if (test_bit(BTRFS_DEV_STATE_REPLACE_TGT, &device->dev_state)) {
2055 old_size = btrfs_device_get_total_bytes(device);
2058 if (new_size > old_size) {
2062 new_size = old_size - new_size;
2063 } else if (mod > 0) {
2064 if (new_size > ULLONG_MAX - old_size) {
2068 new_size = old_size + new_size;
2071 if (new_size < SZ_256M) {
2075 if (new_size > bdev_nr_bytes(device->bdev)) {
2080 new_size = round_down(new_size, fs_info->sectorsize);
2082 if (new_size > old_size) {
2083 trans = btrfs_start_transaction(root, 0);
2084 if (IS_ERR(trans)) {
2085 ret = PTR_ERR(trans);
2088 ret = btrfs_grow_device(trans, device, new_size);
2089 btrfs_commit_transaction(trans);
2090 } else if (new_size < old_size) {
2091 ret = btrfs_shrink_device(device, new_size);
2092 } /* equal, nothing need to do */
2094 if (ret == 0 && new_size != old_size)
2095 btrfs_info_in_rcu(fs_info,
2096 "resize device %s (devid %llu) from %llu to %llu",
2097 rcu_str_deref(device->name), device->devid,
2098 old_size, new_size);
2100 btrfs_exclop_finish(fs_info);
2104 mnt_drop_write_file(file);
2108 static noinline int __btrfs_ioctl_snap_create(struct file *file,
2109 struct user_namespace *mnt_userns,
2110 const char *name, unsigned long fd, int subvol,
2112 struct btrfs_qgroup_inherit *inherit)
2117 if (!S_ISDIR(file_inode(file)->i_mode))
2120 ret = mnt_want_write_file(file);
2124 namelen = strlen(name);
2125 if (strchr(name, '/')) {
2127 goto out_drop_write;
2130 if (name[0] == '.' &&
2131 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
2133 goto out_drop_write;
2137 ret = btrfs_mksubvol(&file->f_path, mnt_userns, name,
2138 namelen, NULL, readonly, inherit);
2140 struct fd src = fdget(fd);
2141 struct inode *src_inode;
2144 goto out_drop_write;
2147 src_inode = file_inode(src.file);
2148 if (src_inode->i_sb != file_inode(file)->i_sb) {
2149 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
2150 "Snapshot src from another FS");
2152 } else if (!inode_owner_or_capable(mnt_userns, src_inode)) {
2154 * Subvolume creation is not restricted, but snapshots
2155 * are limited to own subvolumes only
2159 ret = btrfs_mksnapshot(&file->f_path, mnt_userns,
2161 BTRFS_I(src_inode)->root,
2167 mnt_drop_write_file(file);
2172 static noinline int btrfs_ioctl_snap_create(struct file *file,
2173 void __user *arg, int subvol)
2175 struct btrfs_ioctl_vol_args *vol_args;
2178 if (!S_ISDIR(file_inode(file)->i_mode))
2181 vol_args = memdup_user(arg, sizeof(*vol_args));
2182 if (IS_ERR(vol_args))
2183 return PTR_ERR(vol_args);
2184 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2186 ret = __btrfs_ioctl_snap_create(file, file_mnt_user_ns(file),
2187 vol_args->name, vol_args->fd, subvol,
2194 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
2195 void __user *arg, int subvol)
2197 struct btrfs_ioctl_vol_args_v2 *vol_args;
2199 bool readonly = false;
2200 struct btrfs_qgroup_inherit *inherit = NULL;
2202 if (!S_ISDIR(file_inode(file)->i_mode))
2205 vol_args = memdup_user(arg, sizeof(*vol_args));
2206 if (IS_ERR(vol_args))
2207 return PTR_ERR(vol_args);
2208 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
2210 if (vol_args->flags & ~BTRFS_SUBVOL_CREATE_ARGS_MASK) {
2215 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
2217 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
2220 if (vol_args->size < sizeof(*inherit) ||
2221 vol_args->size > PAGE_SIZE) {
2225 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
2226 if (IS_ERR(inherit)) {
2227 ret = PTR_ERR(inherit);
2231 if (inherit->num_qgroups > PAGE_SIZE ||
2232 inherit->num_ref_copies > PAGE_SIZE ||
2233 inherit->num_excl_copies > PAGE_SIZE) {
2238 nums = inherit->num_qgroups + 2 * inherit->num_ref_copies +
2239 2 * inherit->num_excl_copies;
2240 if (vol_args->size != struct_size(inherit, qgroups, nums)) {
2246 ret = __btrfs_ioctl_snap_create(file, file_mnt_user_ns(file),
2247 vol_args->name, vol_args->fd, subvol,
2258 static noinline int btrfs_ioctl_subvol_getflags(struct inode *inode,
2261 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2262 struct btrfs_root *root = BTRFS_I(inode)->root;
2266 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID)
2269 down_read(&fs_info->subvol_sem);
2270 if (btrfs_root_readonly(root))
2271 flags |= BTRFS_SUBVOL_RDONLY;
2272 up_read(&fs_info->subvol_sem);
2274 if (copy_to_user(arg, &flags, sizeof(flags)))
2280 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
2283 struct inode *inode = file_inode(file);
2284 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2285 struct btrfs_root *root = BTRFS_I(inode)->root;
2286 struct btrfs_trans_handle *trans;
2291 if (!inode_owner_or_capable(file_mnt_user_ns(file), inode))
2294 ret = mnt_want_write_file(file);
2298 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
2300 goto out_drop_write;
2303 if (copy_from_user(&flags, arg, sizeof(flags))) {
2305 goto out_drop_write;
2308 if (flags & ~BTRFS_SUBVOL_RDONLY) {
2310 goto out_drop_write;
2313 down_write(&fs_info->subvol_sem);
2316 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
2319 root_flags = btrfs_root_flags(&root->root_item);
2320 if (flags & BTRFS_SUBVOL_RDONLY) {
2321 btrfs_set_root_flags(&root->root_item,
2322 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
2325 * Block RO -> RW transition if this subvolume is involved in
2328 spin_lock(&root->root_item_lock);
2329 if (root->send_in_progress == 0) {
2330 btrfs_set_root_flags(&root->root_item,
2331 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
2332 spin_unlock(&root->root_item_lock);
2334 spin_unlock(&root->root_item_lock);
2336 "Attempt to set subvolume %llu read-write during send",
2337 root->root_key.objectid);
2343 trans = btrfs_start_transaction(root, 1);
2344 if (IS_ERR(trans)) {
2345 ret = PTR_ERR(trans);
2349 ret = btrfs_update_root(trans, fs_info->tree_root,
2350 &root->root_key, &root->root_item);
2352 btrfs_end_transaction(trans);
2356 ret = btrfs_commit_transaction(trans);
2360 btrfs_set_root_flags(&root->root_item, root_flags);
2362 up_write(&fs_info->subvol_sem);
2364 mnt_drop_write_file(file);
2369 static noinline int key_in_sk(struct btrfs_key *key,
2370 struct btrfs_ioctl_search_key *sk)
2372 struct btrfs_key test;
2375 test.objectid = sk->min_objectid;
2376 test.type = sk->min_type;
2377 test.offset = sk->min_offset;
2379 ret = btrfs_comp_cpu_keys(key, &test);
2383 test.objectid = sk->max_objectid;
2384 test.type = sk->max_type;
2385 test.offset = sk->max_offset;
2387 ret = btrfs_comp_cpu_keys(key, &test);
2393 static noinline int copy_to_sk(struct btrfs_path *path,
2394 struct btrfs_key *key,
2395 struct btrfs_ioctl_search_key *sk,
2398 unsigned long *sk_offset,
2402 struct extent_buffer *leaf;
2403 struct btrfs_ioctl_search_header sh;
2404 struct btrfs_key test;
2405 unsigned long item_off;
2406 unsigned long item_len;
2412 leaf = path->nodes[0];
2413 slot = path->slots[0];
2414 nritems = btrfs_header_nritems(leaf);
2416 if (btrfs_header_generation(leaf) > sk->max_transid) {
2420 found_transid = btrfs_header_generation(leaf);
2422 for (i = slot; i < nritems; i++) {
2423 item_off = btrfs_item_ptr_offset(leaf, i);
2424 item_len = btrfs_item_size(leaf, i);
2426 btrfs_item_key_to_cpu(leaf, key, i);
2427 if (!key_in_sk(key, sk))
2430 if (sizeof(sh) + item_len > *buf_size) {
2437 * return one empty item back for v1, which does not
2441 *buf_size = sizeof(sh) + item_len;
2446 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
2451 sh.objectid = key->objectid;
2452 sh.offset = key->offset;
2453 sh.type = key->type;
2455 sh.transid = found_transid;
2458 * Copy search result header. If we fault then loop again so we
2459 * can fault in the pages and -EFAULT there if there's a
2460 * problem. Otherwise we'll fault and then copy the buffer in
2461 * properly this next time through
2463 if (copy_to_user_nofault(ubuf + *sk_offset, &sh, sizeof(sh))) {
2468 *sk_offset += sizeof(sh);
2471 char __user *up = ubuf + *sk_offset;
2473 * Copy the item, same behavior as above, but reset the
2474 * * sk_offset so we copy the full thing again.
2476 if (read_extent_buffer_to_user_nofault(leaf, up,
2477 item_off, item_len)) {
2479 *sk_offset -= sizeof(sh);
2483 *sk_offset += item_len;
2487 if (ret) /* -EOVERFLOW from above */
2490 if (*num_found >= sk->nr_items) {
2497 test.objectid = sk->max_objectid;
2498 test.type = sk->max_type;
2499 test.offset = sk->max_offset;
2500 if (btrfs_comp_cpu_keys(key, &test) >= 0)
2502 else if (key->offset < (u64)-1)
2504 else if (key->type < (u8)-1) {
2507 } else if (key->objectid < (u64)-1) {
2515 * 0: all items from this leaf copied, continue with next
2516 * 1: * more items can be copied, but unused buffer is too small
2517 * * all items were found
2518 * Either way, it will stops the loop which iterates to the next
2520 * -EOVERFLOW: item was to large for buffer
2521 * -EFAULT: could not copy extent buffer back to userspace
2526 static noinline int search_ioctl(struct inode *inode,
2527 struct btrfs_ioctl_search_key *sk,
2531 struct btrfs_fs_info *info = btrfs_sb(inode->i_sb);
2532 struct btrfs_root *root;
2533 struct btrfs_key key;
2534 struct btrfs_path *path;
2537 unsigned long sk_offset = 0;
2539 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2540 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2544 path = btrfs_alloc_path();
2548 if (sk->tree_id == 0) {
2549 /* search the root of the inode that was passed */
2550 root = btrfs_grab_root(BTRFS_I(inode)->root);
2552 root = btrfs_get_fs_root(info, sk->tree_id, true);
2554 btrfs_free_path(path);
2555 return PTR_ERR(root);
2559 key.objectid = sk->min_objectid;
2560 key.type = sk->min_type;
2561 key.offset = sk->min_offset;
2565 if (fault_in_writeable(ubuf + sk_offset, *buf_size - sk_offset))
2568 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2574 ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
2575 &sk_offset, &num_found);
2576 btrfs_release_path(path);
2584 sk->nr_items = num_found;
2585 btrfs_put_root(root);
2586 btrfs_free_path(path);
2590 static noinline int btrfs_ioctl_tree_search(struct inode *inode,
2593 struct btrfs_ioctl_search_args __user *uargs;
2594 struct btrfs_ioctl_search_key sk;
2598 if (!capable(CAP_SYS_ADMIN))
2601 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2603 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2606 buf_size = sizeof(uargs->buf);
2608 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2611 * In the origin implementation an overflow is handled by returning a
2612 * search header with a len of zero, so reset ret.
2614 if (ret == -EOVERFLOW)
2617 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2622 static noinline int btrfs_ioctl_tree_search_v2(struct inode *inode,
2625 struct btrfs_ioctl_search_args_v2 __user *uarg;
2626 struct btrfs_ioctl_search_args_v2 args;
2629 const size_t buf_limit = SZ_16M;
2631 if (!capable(CAP_SYS_ADMIN))
2634 /* copy search header and buffer size */
2635 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2636 if (copy_from_user(&args, uarg, sizeof(args)))
2639 buf_size = args.buf_size;
2641 /* limit result size to 16MB */
2642 if (buf_size > buf_limit)
2643 buf_size = buf_limit;
2645 ret = search_ioctl(inode, &args.key, &buf_size,
2646 (char __user *)(&uarg->buf[0]));
2647 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2649 else if (ret == -EOVERFLOW &&
2650 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2657 * Search INODE_REFs to identify path name of 'dirid' directory
2658 * in a 'tree_id' tree. and sets path name to 'name'.
2660 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2661 u64 tree_id, u64 dirid, char *name)
2663 struct btrfs_root *root;
2664 struct btrfs_key key;
2670 struct btrfs_inode_ref *iref;
2671 struct extent_buffer *l;
2672 struct btrfs_path *path;
2674 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2679 path = btrfs_alloc_path();
2683 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1];
2685 root = btrfs_get_fs_root(info, tree_id, true);
2687 ret = PTR_ERR(root);
2692 key.objectid = dirid;
2693 key.type = BTRFS_INODE_REF_KEY;
2694 key.offset = (u64)-1;
2697 ret = btrfs_search_backwards(root, &key, path);
2706 slot = path->slots[0];
2708 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2709 len = btrfs_inode_ref_name_len(l, iref);
2711 total_len += len + 1;
2713 ret = -ENAMETOOLONG;
2718 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2720 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2723 btrfs_release_path(path);
2724 key.objectid = key.offset;
2725 key.offset = (u64)-1;
2726 dirid = key.objectid;
2728 memmove(name, ptr, total_len);
2729 name[total_len] = '\0';
2732 btrfs_put_root(root);
2733 btrfs_free_path(path);
2737 static int btrfs_search_path_in_tree_user(struct user_namespace *mnt_userns,
2738 struct inode *inode,
2739 struct btrfs_ioctl_ino_lookup_user_args *args)
2741 struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
2742 struct super_block *sb = inode->i_sb;
2743 struct btrfs_key upper_limit = BTRFS_I(inode)->location;
2744 u64 treeid = BTRFS_I(inode)->root->root_key.objectid;
2745 u64 dirid = args->dirid;
2746 unsigned long item_off;
2747 unsigned long item_len;
2748 struct btrfs_inode_ref *iref;
2749 struct btrfs_root_ref *rref;
2750 struct btrfs_root *root = NULL;
2751 struct btrfs_path *path;
2752 struct btrfs_key key, key2;
2753 struct extent_buffer *leaf;
2754 struct inode *temp_inode;
2761 path = btrfs_alloc_path();
2766 * If the bottom subvolume does not exist directly under upper_limit,
2767 * construct the path in from the bottom up.
2769 if (dirid != upper_limit.objectid) {
2770 ptr = &args->path[BTRFS_INO_LOOKUP_USER_PATH_MAX - 1];
2772 root = btrfs_get_fs_root(fs_info, treeid, true);
2774 ret = PTR_ERR(root);
2778 key.objectid = dirid;
2779 key.type = BTRFS_INODE_REF_KEY;
2780 key.offset = (u64)-1;
2782 ret = btrfs_search_backwards(root, &key, path);
2790 leaf = path->nodes[0];
2791 slot = path->slots[0];
2793 iref = btrfs_item_ptr(leaf, slot, struct btrfs_inode_ref);
2794 len = btrfs_inode_ref_name_len(leaf, iref);
2796 total_len += len + 1;
2797 if (ptr < args->path) {
2798 ret = -ENAMETOOLONG;
2803 read_extent_buffer(leaf, ptr,
2804 (unsigned long)(iref + 1), len);
2806 /* Check the read+exec permission of this directory */
2807 ret = btrfs_previous_item(root, path, dirid,
2808 BTRFS_INODE_ITEM_KEY);
2811 } else if (ret > 0) {
2816 leaf = path->nodes[0];
2817 slot = path->slots[0];
2818 btrfs_item_key_to_cpu(leaf, &key2, slot);
2819 if (key2.objectid != dirid) {
2824 temp_inode = btrfs_iget(sb, key2.objectid, root);
2825 if (IS_ERR(temp_inode)) {
2826 ret = PTR_ERR(temp_inode);
2829 ret = inode_permission(mnt_userns, temp_inode,
2830 MAY_READ | MAY_EXEC);
2837 if (key.offset == upper_limit.objectid)
2839 if (key.objectid == BTRFS_FIRST_FREE_OBJECTID) {
2844 btrfs_release_path(path);
2845 key.objectid = key.offset;
2846 key.offset = (u64)-1;
2847 dirid = key.objectid;
2850 memmove(args->path, ptr, total_len);
2851 args->path[total_len] = '\0';
2852 btrfs_put_root(root);
2854 btrfs_release_path(path);
2857 /* Get the bottom subvolume's name from ROOT_REF */
2858 key.objectid = treeid;
2859 key.type = BTRFS_ROOT_REF_KEY;
2860 key.offset = args->treeid;
2861 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
2864 } else if (ret > 0) {
2869 leaf = path->nodes[0];
2870 slot = path->slots[0];
2871 btrfs_item_key_to_cpu(leaf, &key, slot);
2873 item_off = btrfs_item_ptr_offset(leaf, slot);
2874 item_len = btrfs_item_size(leaf, slot);
2875 /* Check if dirid in ROOT_REF corresponds to passed dirid */
2876 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
2877 if (args->dirid != btrfs_root_ref_dirid(leaf, rref)) {
2882 /* Copy subvolume's name */
2883 item_off += sizeof(struct btrfs_root_ref);
2884 item_len -= sizeof(struct btrfs_root_ref);
2885 read_extent_buffer(leaf, args->name, item_off, item_len);
2886 args->name[item_len] = 0;
2889 btrfs_put_root(root);
2891 btrfs_free_path(path);
2895 static noinline int btrfs_ioctl_ino_lookup(struct btrfs_root *root,
2898 struct btrfs_ioctl_ino_lookup_args *args;
2901 args = memdup_user(argp, sizeof(*args));
2903 return PTR_ERR(args);
2906 * Unprivileged query to obtain the containing subvolume root id. The
2907 * path is reset so it's consistent with btrfs_search_path_in_tree.
2909 if (args->treeid == 0)
2910 args->treeid = root->root_key.objectid;
2912 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2917 if (!capable(CAP_SYS_ADMIN)) {
2922 ret = btrfs_search_path_in_tree(root->fs_info,
2923 args->treeid, args->objectid,
2927 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2935 * Version of ino_lookup ioctl (unprivileged)
2937 * The main differences from ino_lookup ioctl are:
2939 * 1. Read + Exec permission will be checked using inode_permission() during
2940 * path construction. -EACCES will be returned in case of failure.
2941 * 2. Path construction will be stopped at the inode number which corresponds
2942 * to the fd with which this ioctl is called. If constructed path does not
2943 * exist under fd's inode, -EACCES will be returned.
2944 * 3. The name of bottom subvolume is also searched and filled.
2946 static int btrfs_ioctl_ino_lookup_user(struct file *file, void __user *argp)
2948 struct btrfs_ioctl_ino_lookup_user_args *args;
2949 struct inode *inode;
2952 args = memdup_user(argp, sizeof(*args));
2954 return PTR_ERR(args);
2956 inode = file_inode(file);
2958 if (args->dirid == BTRFS_FIRST_FREE_OBJECTID &&
2959 BTRFS_I(inode)->location.objectid != BTRFS_FIRST_FREE_OBJECTID) {
2961 * The subvolume does not exist under fd with which this is
2968 ret = btrfs_search_path_in_tree_user(file_mnt_user_ns(file), inode, args);
2970 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2977 /* Get the subvolume information in BTRFS_ROOT_ITEM and BTRFS_ROOT_BACKREF */
2978 static int btrfs_ioctl_get_subvol_info(struct inode *inode, void __user *argp)
2980 struct btrfs_ioctl_get_subvol_info_args *subvol_info;
2981 struct btrfs_fs_info *fs_info;
2982 struct btrfs_root *root;
2983 struct btrfs_path *path;
2984 struct btrfs_key key;
2985 struct btrfs_root_item *root_item;
2986 struct btrfs_root_ref *rref;
2987 struct extent_buffer *leaf;
2988 unsigned long item_off;
2989 unsigned long item_len;
2993 path = btrfs_alloc_path();
2997 subvol_info = kzalloc(sizeof(*subvol_info), GFP_KERNEL);
2999 btrfs_free_path(path);
3003 fs_info = BTRFS_I(inode)->root->fs_info;
3005 /* Get root_item of inode's subvolume */
3006 key.objectid = BTRFS_I(inode)->root->root_key.objectid;
3007 root = btrfs_get_fs_root(fs_info, key.objectid, true);
3009 ret = PTR_ERR(root);
3012 root_item = &root->root_item;
3014 subvol_info->treeid = key.objectid;
3016 subvol_info->generation = btrfs_root_generation(root_item);
3017 subvol_info->flags = btrfs_root_flags(root_item);
3019 memcpy(subvol_info->uuid, root_item->uuid, BTRFS_UUID_SIZE);
3020 memcpy(subvol_info->parent_uuid, root_item->parent_uuid,
3022 memcpy(subvol_info->received_uuid, root_item->received_uuid,
3025 subvol_info->ctransid = btrfs_root_ctransid(root_item);
3026 subvol_info->ctime.sec = btrfs_stack_timespec_sec(&root_item->ctime);
3027 subvol_info->ctime.nsec = btrfs_stack_timespec_nsec(&root_item->ctime);
3029 subvol_info->otransid = btrfs_root_otransid(root_item);
3030 subvol_info->otime.sec = btrfs_stack_timespec_sec(&root_item->otime);
3031 subvol_info->otime.nsec = btrfs_stack_timespec_nsec(&root_item->otime);
3033 subvol_info->stransid = btrfs_root_stransid(root_item);
3034 subvol_info->stime.sec = btrfs_stack_timespec_sec(&root_item->stime);
3035 subvol_info->stime.nsec = btrfs_stack_timespec_nsec(&root_item->stime);
3037 subvol_info->rtransid = btrfs_root_rtransid(root_item);
3038 subvol_info->rtime.sec = btrfs_stack_timespec_sec(&root_item->rtime);
3039 subvol_info->rtime.nsec = btrfs_stack_timespec_nsec(&root_item->rtime);
3041 if (key.objectid != BTRFS_FS_TREE_OBJECTID) {
3042 /* Search root tree for ROOT_BACKREF of this subvolume */
3043 key.type = BTRFS_ROOT_BACKREF_KEY;
3045 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
3048 } else if (path->slots[0] >=
3049 btrfs_header_nritems(path->nodes[0])) {
3050 ret = btrfs_next_leaf(fs_info->tree_root, path);
3053 } else if (ret > 0) {
3059 leaf = path->nodes[0];
3060 slot = path->slots[0];
3061 btrfs_item_key_to_cpu(leaf, &key, slot);
3062 if (key.objectid == subvol_info->treeid &&
3063 key.type == BTRFS_ROOT_BACKREF_KEY) {
3064 subvol_info->parent_id = key.offset;
3066 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
3067 subvol_info->dirid = btrfs_root_ref_dirid(leaf, rref);
3069 item_off = btrfs_item_ptr_offset(leaf, slot)
3070 + sizeof(struct btrfs_root_ref);
3071 item_len = btrfs_item_size(leaf, slot)
3072 - sizeof(struct btrfs_root_ref);
3073 read_extent_buffer(leaf, subvol_info->name,
3074 item_off, item_len);
3081 if (copy_to_user(argp, subvol_info, sizeof(*subvol_info)))
3085 btrfs_put_root(root);
3087 btrfs_free_path(path);
3093 * Return ROOT_REF information of the subvolume containing this inode
3094 * except the subvolume name.
3096 static int btrfs_ioctl_get_subvol_rootref(struct btrfs_root *root,
3099 struct btrfs_ioctl_get_subvol_rootref_args *rootrefs;
3100 struct btrfs_root_ref *rref;
3101 struct btrfs_path *path;
3102 struct btrfs_key key;
3103 struct extent_buffer *leaf;
3109 path = btrfs_alloc_path();
3113 rootrefs = memdup_user(argp, sizeof(*rootrefs));
3114 if (IS_ERR(rootrefs)) {
3115 btrfs_free_path(path);
3116 return PTR_ERR(rootrefs);
3119 objectid = root->root_key.objectid;
3120 key.objectid = objectid;
3121 key.type = BTRFS_ROOT_REF_KEY;
3122 key.offset = rootrefs->min_treeid;
3125 root = root->fs_info->tree_root;
3126 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3129 } else if (path->slots[0] >=
3130 btrfs_header_nritems(path->nodes[0])) {
3131 ret = btrfs_next_leaf(root, path);
3134 } else if (ret > 0) {
3140 leaf = path->nodes[0];
3141 slot = path->slots[0];
3143 btrfs_item_key_to_cpu(leaf, &key, slot);
3144 if (key.objectid != objectid || key.type != BTRFS_ROOT_REF_KEY) {
3149 if (found == BTRFS_MAX_ROOTREF_BUFFER_NUM) {
3154 rref = btrfs_item_ptr(leaf, slot, struct btrfs_root_ref);
3155 rootrefs->rootref[found].treeid = key.offset;
3156 rootrefs->rootref[found].dirid =
3157 btrfs_root_ref_dirid(leaf, rref);
3160 ret = btrfs_next_item(root, path);
3163 } else if (ret > 0) {
3170 if (!ret || ret == -EOVERFLOW) {
3171 rootrefs->num_items = found;
3172 /* update min_treeid for next search */
3174 rootrefs->min_treeid =
3175 rootrefs->rootref[found - 1].treeid + 1;
3176 if (copy_to_user(argp, rootrefs, sizeof(*rootrefs)))
3181 btrfs_free_path(path);
3186 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
3190 struct dentry *parent = file->f_path.dentry;
3191 struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb);
3192 struct dentry *dentry;
3193 struct inode *dir = d_inode(parent);
3194 struct inode *inode;
3195 struct btrfs_root *root = BTRFS_I(dir)->root;
3196 struct btrfs_root *dest = NULL;
3197 struct btrfs_ioctl_vol_args *vol_args = NULL;
3198 struct btrfs_ioctl_vol_args_v2 *vol_args2 = NULL;
3199 struct user_namespace *mnt_userns = file_mnt_user_ns(file);
3200 char *subvol_name, *subvol_name_ptr = NULL;
3203 bool destroy_parent = false;
3205 /* We don't support snapshots with extent tree v2 yet. */
3206 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
3208 "extent tree v2 doesn't support snapshot deletion yet");
3213 vol_args2 = memdup_user(arg, sizeof(*vol_args2));
3214 if (IS_ERR(vol_args2))
3215 return PTR_ERR(vol_args2);
3217 if (vol_args2->flags & ~BTRFS_SUBVOL_DELETE_ARGS_MASK) {
3223 * If SPEC_BY_ID is not set, we are looking for the subvolume by
3224 * name, same as v1 currently does.
3226 if (!(vol_args2->flags & BTRFS_SUBVOL_SPEC_BY_ID)) {
3227 vol_args2->name[BTRFS_SUBVOL_NAME_MAX] = 0;
3228 subvol_name = vol_args2->name;
3230 err = mnt_want_write_file(file);
3234 struct inode *old_dir;
3236 if (vol_args2->subvolid < BTRFS_FIRST_FREE_OBJECTID) {
3241 err = mnt_want_write_file(file);
3245 dentry = btrfs_get_dentry(fs_info->sb,
3246 BTRFS_FIRST_FREE_OBJECTID,
3247 vol_args2->subvolid, 0, 0);
3248 if (IS_ERR(dentry)) {
3249 err = PTR_ERR(dentry);
3250 goto out_drop_write;
3254 * Change the default parent since the subvolume being
3255 * deleted can be outside of the current mount point.
3257 parent = btrfs_get_parent(dentry);
3260 * At this point dentry->d_name can point to '/' if the
3261 * subvolume we want to destroy is outsite of the
3262 * current mount point, so we need to release the
3263 * current dentry and execute the lookup to return a new
3264 * one with ->d_name pointing to the
3265 * <mount point>/subvol_name.
3268 if (IS_ERR(parent)) {
3269 err = PTR_ERR(parent);
3270 goto out_drop_write;
3273 dir = d_inode(parent);
3276 * If v2 was used with SPEC_BY_ID, a new parent was
3277 * allocated since the subvolume can be outside of the
3278 * current mount point. Later on we need to release this
3279 * new parent dentry.
3281 destroy_parent = true;
3284 * On idmapped mounts, deletion via subvolid is
3285 * restricted to subvolumes that are immediate
3286 * ancestors of the inode referenced by the file
3287 * descriptor in the ioctl. Otherwise the idmapping
3288 * could potentially be abused to delete subvolumes
3289 * anywhere in the filesystem the user wouldn't be able
3290 * to delete without an idmapped mount.
3292 if (old_dir != dir && mnt_userns != &init_user_ns) {
3297 subvol_name_ptr = btrfs_get_subvol_name_from_objectid(
3298 fs_info, vol_args2->subvolid);
3299 if (IS_ERR(subvol_name_ptr)) {
3300 err = PTR_ERR(subvol_name_ptr);
3303 /* subvol_name_ptr is already nul terminated */
3304 subvol_name = (char *)kbasename(subvol_name_ptr);
3307 vol_args = memdup_user(arg, sizeof(*vol_args));
3308 if (IS_ERR(vol_args))
3309 return PTR_ERR(vol_args);
3311 vol_args->name[BTRFS_PATH_NAME_MAX] = 0;
3312 subvol_name = vol_args->name;
3314 err = mnt_want_write_file(file);
3319 subvol_namelen = strlen(subvol_name);
3321 if (strchr(subvol_name, '/') ||
3322 strncmp(subvol_name, "..", subvol_namelen) == 0) {
3324 goto free_subvol_name;
3327 if (!S_ISDIR(dir->i_mode)) {
3329 goto free_subvol_name;
3332 err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
3334 goto free_subvol_name;
3335 dentry = lookup_one(mnt_userns, subvol_name, parent, subvol_namelen);
3336 if (IS_ERR(dentry)) {
3337 err = PTR_ERR(dentry);
3338 goto out_unlock_dir;
3341 if (d_really_is_negative(dentry)) {
3346 inode = d_inode(dentry);
3347 dest = BTRFS_I(inode)->root;
3348 if (!capable(CAP_SYS_ADMIN)) {
3350 * Regular user. Only allow this with a special mount
3351 * option, when the user has write+exec access to the
3352 * subvol root, and when rmdir(2) would have been
3355 * Note that this is _not_ check that the subvol is
3356 * empty or doesn't contain data that we wouldn't
3357 * otherwise be able to delete.
3359 * Users who want to delete empty subvols should try
3363 if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
3367 * Do not allow deletion if the parent dir is the same
3368 * as the dir to be deleted. That means the ioctl
3369 * must be called on the dentry referencing the root
3370 * of the subvol, not a random directory contained
3377 err = inode_permission(mnt_userns, inode, MAY_WRITE | MAY_EXEC);
3382 /* check if subvolume may be deleted by a user */
3383 err = btrfs_may_delete(mnt_userns, dir, dentry, 1);
3387 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
3392 btrfs_inode_lock(inode, 0);
3393 err = btrfs_delete_subvolume(dir, dentry);
3394 btrfs_inode_unlock(inode, 0);
3396 d_delete_notify(dir, dentry);
3401 btrfs_inode_unlock(dir, 0);
3403 kfree(subvol_name_ptr);
3408 mnt_drop_write_file(file);
3415 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
3417 struct inode *inode = file_inode(file);
3418 struct btrfs_root *root = BTRFS_I(inode)->root;
3419 struct btrfs_ioctl_defrag_range_args range = {0};
3422 ret = mnt_want_write_file(file);
3426 if (btrfs_root_readonly(root)) {
3431 switch (inode->i_mode & S_IFMT) {
3433 if (!capable(CAP_SYS_ADMIN)) {
3437 ret = btrfs_defrag_root(root);
3441 * Note that this does not check the file descriptor for write
3442 * access. This prevents defragmenting executables that are
3443 * running and allows defrag on files open in read-only mode.
3445 if (!capable(CAP_SYS_ADMIN) &&
3446 inode_permission(&init_user_ns, inode, MAY_WRITE)) {
3452 if (copy_from_user(&range, argp, sizeof(range))) {
3456 /* compression requires us to start the IO */
3457 if ((range.flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
3458 range.flags |= BTRFS_DEFRAG_RANGE_START_IO;
3459 range.extent_thresh = (u32)-1;
3462 /* the rest are all set to zero by kzalloc */
3463 range.len = (u64)-1;
3465 ret = btrfs_defrag_file(file_inode(file), &file->f_ra,
3466 &range, BTRFS_OLDEST_GENERATION, 0);
3474 mnt_drop_write_file(file);
3478 static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg)
3480 struct btrfs_ioctl_vol_args *vol_args;
3481 bool restore_op = false;
3484 if (!capable(CAP_SYS_ADMIN))
3487 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
3488 btrfs_err(fs_info, "device add not supported on extent tree v2 yet");
3492 if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_ADD)) {
3493 if (!btrfs_exclop_start_try_lock(fs_info, BTRFS_EXCLOP_DEV_ADD))
3494 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3497 * We can do the device add because we have a paused balanced,
3498 * change the exclusive op type and remember we should bring
3499 * back the paused balance
3501 fs_info->exclusive_operation = BTRFS_EXCLOP_DEV_ADD;
3502 btrfs_exclop_start_unlock(fs_info);
3506 vol_args = memdup_user(arg, sizeof(*vol_args));
3507 if (IS_ERR(vol_args)) {
3508 ret = PTR_ERR(vol_args);
3512 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
3513 ret = btrfs_init_new_device(fs_info, vol_args->name);
3516 btrfs_info(fs_info, "disk added %s", vol_args->name);
3521 btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE_PAUSED);
3523 btrfs_exclop_finish(fs_info);
3527 static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
3529 BTRFS_DEV_LOOKUP_ARGS(args);
3530 struct inode *inode = file_inode(file);
3531 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3532 struct btrfs_ioctl_vol_args_v2 *vol_args;
3533 struct block_device *bdev = NULL;
3536 bool cancel = false;
3538 if (!capable(CAP_SYS_ADMIN))
3541 vol_args = memdup_user(arg, sizeof(*vol_args));
3542 if (IS_ERR(vol_args))
3543 return PTR_ERR(vol_args);
3545 if (vol_args->flags & ~BTRFS_DEVICE_REMOVE_ARGS_MASK) {
3550 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
3551 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
3552 args.devid = vol_args->devid;
3553 } else if (!strcmp("cancel", vol_args->name)) {
3556 ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name);
3561 ret = mnt_want_write_file(file);
3565 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
3570 /* Exclusive operation is now claimed */
3571 ret = btrfs_rm_device(fs_info, &args, &bdev, &mode);
3573 btrfs_exclop_finish(fs_info);
3576 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
3577 btrfs_info(fs_info, "device deleted: id %llu",
3580 btrfs_info(fs_info, "device deleted: %s",
3584 mnt_drop_write_file(file);
3586 blkdev_put(bdev, mode);
3588 btrfs_put_dev_args_from_path(&args);
3593 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
3595 BTRFS_DEV_LOOKUP_ARGS(args);
3596 struct inode *inode = file_inode(file);
3597 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3598 struct btrfs_ioctl_vol_args *vol_args;
3599 struct block_device *bdev = NULL;
3602 bool cancel = false;
3604 if (!capable(CAP_SYS_ADMIN))
3607 vol_args = memdup_user(arg, sizeof(*vol_args));
3608 if (IS_ERR(vol_args))
3609 return PTR_ERR(vol_args);
3611 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
3612 if (!strcmp("cancel", vol_args->name)) {
3615 ret = btrfs_get_dev_args_from_path(fs_info, &args, vol_args->name);
3620 ret = mnt_want_write_file(file);
3624 ret = exclop_start_or_cancel_reloc(fs_info, BTRFS_EXCLOP_DEV_REMOVE,
3627 ret = btrfs_rm_device(fs_info, &args, &bdev, &mode);
3629 btrfs_info(fs_info, "disk deleted %s", vol_args->name);
3630 btrfs_exclop_finish(fs_info);
3633 mnt_drop_write_file(file);
3635 blkdev_put(bdev, mode);
3637 btrfs_put_dev_args_from_path(&args);
3642 static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
3645 struct btrfs_ioctl_fs_info_args *fi_args;
3646 struct btrfs_device *device;
3647 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
3651 fi_args = memdup_user(arg, sizeof(*fi_args));
3652 if (IS_ERR(fi_args))
3653 return PTR_ERR(fi_args);
3655 flags_in = fi_args->flags;
3656 memset(fi_args, 0, sizeof(*fi_args));
3659 fi_args->num_devices = fs_devices->num_devices;
3661 list_for_each_entry_rcu(device, &fs_devices->devices, dev_list) {
3662 if (device->devid > fi_args->max_id)
3663 fi_args->max_id = device->devid;
3667 memcpy(&fi_args->fsid, fs_devices->fsid, sizeof(fi_args->fsid));
3668 fi_args->nodesize = fs_info->nodesize;
3669 fi_args->sectorsize = fs_info->sectorsize;
3670 fi_args->clone_alignment = fs_info->sectorsize;
3672 if (flags_in & BTRFS_FS_INFO_FLAG_CSUM_INFO) {
3673 fi_args->csum_type = btrfs_super_csum_type(fs_info->super_copy);
3674 fi_args->csum_size = btrfs_super_csum_size(fs_info->super_copy);
3675 fi_args->flags |= BTRFS_FS_INFO_FLAG_CSUM_INFO;
3678 if (flags_in & BTRFS_FS_INFO_FLAG_GENERATION) {
3679 fi_args->generation = fs_info->generation;
3680 fi_args->flags |= BTRFS_FS_INFO_FLAG_GENERATION;
3683 if (flags_in & BTRFS_FS_INFO_FLAG_METADATA_UUID) {
3684 memcpy(&fi_args->metadata_uuid, fs_devices->metadata_uuid,
3685 sizeof(fi_args->metadata_uuid));
3686 fi_args->flags |= BTRFS_FS_INFO_FLAG_METADATA_UUID;
3689 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
3696 static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
3699 BTRFS_DEV_LOOKUP_ARGS(args);
3700 struct btrfs_ioctl_dev_info_args *di_args;
3701 struct btrfs_device *dev;
3704 di_args = memdup_user(arg, sizeof(*di_args));
3705 if (IS_ERR(di_args))
3706 return PTR_ERR(di_args);
3708 args.devid = di_args->devid;
3709 if (!btrfs_is_empty_uuid(di_args->uuid))
3710 args.uuid = di_args->uuid;
3713 dev = btrfs_find_device(fs_info->fs_devices, &args);
3719 di_args->devid = dev->devid;
3720 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
3721 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
3722 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
3724 strncpy(di_args->path, rcu_str_deref(dev->name),
3725 sizeof(di_args->path) - 1);
3726 di_args->path[sizeof(di_args->path) - 1] = 0;
3728 di_args->path[0] = '\0';
3733 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
3740 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3742 struct inode *inode = file_inode(file);
3743 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3744 struct btrfs_root *root = BTRFS_I(inode)->root;
3745 struct btrfs_root *new_root;
3746 struct btrfs_dir_item *di;
3747 struct btrfs_trans_handle *trans;
3748 struct btrfs_path *path = NULL;
3749 struct btrfs_disk_key disk_key;
3754 if (!capable(CAP_SYS_ADMIN))
3757 ret = mnt_want_write_file(file);
3761 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3767 objectid = BTRFS_FS_TREE_OBJECTID;
3769 new_root = btrfs_get_fs_root(fs_info, objectid, true);
3770 if (IS_ERR(new_root)) {
3771 ret = PTR_ERR(new_root);
3774 if (!is_fstree(new_root->root_key.objectid)) {
3779 path = btrfs_alloc_path();
3785 trans = btrfs_start_transaction(root, 1);
3786 if (IS_ERR(trans)) {
3787 ret = PTR_ERR(trans);
3791 dir_id = btrfs_super_root_dir(fs_info->super_copy);
3792 di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
3793 dir_id, "default", 7, 1);
3794 if (IS_ERR_OR_NULL(di)) {
3795 btrfs_release_path(path);
3796 btrfs_end_transaction(trans);
3798 "Umm, you don't have the default diritem, this isn't going to work");
3803 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3804 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3805 btrfs_mark_buffer_dirty(path->nodes[0]);
3806 btrfs_release_path(path);
3808 btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
3809 btrfs_end_transaction(trans);
3811 btrfs_put_root(new_root);
3812 btrfs_free_path(path);
3814 mnt_drop_write_file(file);
3818 static void get_block_group_info(struct list_head *groups_list,
3819 struct btrfs_ioctl_space_info *space)
3821 struct btrfs_block_group *block_group;
3823 space->total_bytes = 0;
3824 space->used_bytes = 0;
3826 list_for_each_entry(block_group, groups_list, list) {
3827 space->flags = block_group->flags;
3828 space->total_bytes += block_group->length;
3829 space->used_bytes += block_group->used;
3833 static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
3836 struct btrfs_ioctl_space_args space_args;
3837 struct btrfs_ioctl_space_info space;
3838 struct btrfs_ioctl_space_info *dest;
3839 struct btrfs_ioctl_space_info *dest_orig;
3840 struct btrfs_ioctl_space_info __user *user_dest;
3841 struct btrfs_space_info *info;
3842 static const u64 types[] = {
3843 BTRFS_BLOCK_GROUP_DATA,
3844 BTRFS_BLOCK_GROUP_SYSTEM,
3845 BTRFS_BLOCK_GROUP_METADATA,
3846 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA
3854 if (copy_from_user(&space_args,
3855 (struct btrfs_ioctl_space_args __user *)arg,
3856 sizeof(space_args)))
3859 for (i = 0; i < num_types; i++) {
3860 struct btrfs_space_info *tmp;
3863 list_for_each_entry(tmp, &fs_info->space_info, list) {
3864 if (tmp->flags == types[i]) {
3873 down_read(&info->groups_sem);
3874 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3875 if (!list_empty(&info->block_groups[c]))
3878 up_read(&info->groups_sem);
3882 * Global block reserve, exported as a space_info
3886 /* space_slots == 0 means they are asking for a count */
3887 if (space_args.space_slots == 0) {
3888 space_args.total_spaces = slot_count;
3892 slot_count = min_t(u64, space_args.space_slots, slot_count);
3894 alloc_size = sizeof(*dest) * slot_count;
3896 /* we generally have at most 6 or so space infos, one for each raid
3897 * level. So, a whole page should be more than enough for everyone
3899 if (alloc_size > PAGE_SIZE)
3902 space_args.total_spaces = 0;
3903 dest = kmalloc(alloc_size, GFP_KERNEL);
3908 /* now we have a buffer to copy into */
3909 for (i = 0; i < num_types; i++) {
3910 struct btrfs_space_info *tmp;
3916 list_for_each_entry(tmp, &fs_info->space_info, list) {
3917 if (tmp->flags == types[i]) {
3925 down_read(&info->groups_sem);
3926 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3927 if (!list_empty(&info->block_groups[c])) {
3928 get_block_group_info(&info->block_groups[c],
3930 memcpy(dest, &space, sizeof(space));
3932 space_args.total_spaces++;
3938 up_read(&info->groups_sem);
3942 * Add global block reserve
3945 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
3947 spin_lock(&block_rsv->lock);
3948 space.total_bytes = block_rsv->size;
3949 space.used_bytes = block_rsv->size - block_rsv->reserved;
3950 spin_unlock(&block_rsv->lock);
3951 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3952 memcpy(dest, &space, sizeof(space));
3953 space_args.total_spaces++;
3956 user_dest = (struct btrfs_ioctl_space_info __user *)
3957 (arg + sizeof(struct btrfs_ioctl_space_args));
3959 if (copy_to_user(user_dest, dest_orig, alloc_size))
3964 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3970 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3973 struct btrfs_trans_handle *trans;
3976 trans = btrfs_attach_transaction_barrier(root);
3977 if (IS_ERR(trans)) {
3978 if (PTR_ERR(trans) != -ENOENT)
3979 return PTR_ERR(trans);
3981 /* No running transaction, don't bother */
3982 transid = root->fs_info->last_trans_committed;
3985 transid = trans->transid;
3986 btrfs_commit_transaction_async(trans);
3989 if (copy_to_user(argp, &transid, sizeof(transid)))
3994 static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
4000 if (copy_from_user(&transid, argp, sizeof(transid)))
4003 transid = 0; /* current trans */
4005 return btrfs_wait_for_commit(fs_info, transid);
4008 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
4010 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
4011 struct btrfs_ioctl_scrub_args *sa;
4014 if (!capable(CAP_SYS_ADMIN))
4017 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
4018 btrfs_err(fs_info, "scrub is not supported on extent tree v2 yet");
4022 sa = memdup_user(arg, sizeof(*sa));
4026 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4027 ret = mnt_want_write_file(file);
4032 ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
4033 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4037 * Copy scrub args to user space even if btrfs_scrub_dev() returned an
4038 * error. This is important as it allows user space to know how much
4039 * progress scrub has done. For example, if scrub is canceled we get
4040 * -ECANCELED from btrfs_scrub_dev() and return that error back to user
4041 * space. Later user space can inspect the progress from the structure
4042 * btrfs_ioctl_scrub_args and resume scrub from where it left off
4043 * previously (btrfs-progs does this).
4044 * If we fail to copy the btrfs_ioctl_scrub_args structure to user space
4045 * then return -EFAULT to signal the structure was not copied or it may
4046 * be corrupt and unreliable due to a partial copy.
4048 if (copy_to_user(arg, sa, sizeof(*sa)))
4051 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4052 mnt_drop_write_file(file);
4058 static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
4060 if (!capable(CAP_SYS_ADMIN))
4063 return btrfs_scrub_cancel(fs_info);
4066 static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
4069 struct btrfs_ioctl_scrub_args *sa;
4072 if (!capable(CAP_SYS_ADMIN))
4075 sa = memdup_user(arg, sizeof(*sa));
4079 ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
4081 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
4088 static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
4091 struct btrfs_ioctl_get_dev_stats *sa;
4094 sa = memdup_user(arg, sizeof(*sa));
4098 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4103 ret = btrfs_get_dev_stats(fs_info, sa);
4105 if (ret == 0 && copy_to_user(arg, sa, sizeof(*sa)))
4112 static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
4115 struct btrfs_ioctl_dev_replace_args *p;
4118 if (!capable(CAP_SYS_ADMIN))
4121 if (btrfs_fs_incompat(fs_info, EXTENT_TREE_V2)) {
4122 btrfs_err(fs_info, "device replace not supported on extent tree v2 yet");
4126 p = memdup_user(arg, sizeof(*p));
4131 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
4132 if (sb_rdonly(fs_info->sb)) {
4136 if (!btrfs_exclop_start(fs_info, BTRFS_EXCLOP_DEV_REPLACE)) {
4137 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4139 ret = btrfs_dev_replace_by_ioctl(fs_info, p);
4140 btrfs_exclop_finish(fs_info);
4143 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4144 btrfs_dev_replace_status(fs_info, p);
4147 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4148 p->result = btrfs_dev_replace_cancel(fs_info);
4156 if ((ret == 0 || ret == -ECANCELED) && copy_to_user(arg, p, sizeof(*p)))
4163 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4169 struct btrfs_ioctl_ino_path_args *ipa = NULL;
4170 struct inode_fs_paths *ipath = NULL;
4171 struct btrfs_path *path;
4173 if (!capable(CAP_DAC_READ_SEARCH))
4176 path = btrfs_alloc_path();
4182 ipa = memdup_user(arg, sizeof(*ipa));
4189 size = min_t(u32, ipa->size, 4096);
4190 ipath = init_ipath(size, root, path);
4191 if (IS_ERR(ipath)) {
4192 ret = PTR_ERR(ipath);
4197 ret = paths_from_inode(ipa->inum, ipath);
4201 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
4202 rel_ptr = ipath->fspath->val[i] -
4203 (u64)(unsigned long)ipath->fspath->val;
4204 ipath->fspath->val[i] = rel_ptr;
4207 ret = copy_to_user((void __user *)(unsigned long)ipa->fspath,
4208 ipath->fspath, size);
4215 btrfs_free_path(path);
4222 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4224 struct btrfs_data_container *inodes = ctx;
4225 const size_t c = 3 * sizeof(u64);
4227 if (inodes->bytes_left >= c) {
4228 inodes->bytes_left -= c;
4229 inodes->val[inodes->elem_cnt] = inum;
4230 inodes->val[inodes->elem_cnt + 1] = offset;
4231 inodes->val[inodes->elem_cnt + 2] = root;
4232 inodes->elem_cnt += 3;
4234 inodes->bytes_missing += c - inodes->bytes_left;
4235 inodes->bytes_left = 0;
4236 inodes->elem_missed += 3;
4242 static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
4243 void __user *arg, int version)
4247 struct btrfs_ioctl_logical_ino_args *loi;
4248 struct btrfs_data_container *inodes = NULL;
4249 struct btrfs_path *path = NULL;
4252 if (!capable(CAP_SYS_ADMIN))
4255 loi = memdup_user(arg, sizeof(*loi));
4257 return PTR_ERR(loi);
4260 ignore_offset = false;
4261 size = min_t(u32, loi->size, SZ_64K);
4263 /* All reserved bits must be 0 for now */
4264 if (memchr_inv(loi->reserved, 0, sizeof(loi->reserved))) {
4268 /* Only accept flags we have defined so far */
4269 if (loi->flags & ~(BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET)) {
4273 ignore_offset = loi->flags & BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET;
4274 size = min_t(u32, loi->size, SZ_16M);
4277 path = btrfs_alloc_path();
4283 inodes = init_data_container(size);
4284 if (IS_ERR(inodes)) {
4285 ret = PTR_ERR(inodes);
4290 ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
4291 build_ino_list, inodes, ignore_offset);
4297 ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes,
4303 btrfs_free_path(path);
4311 void btrfs_update_ioctl_balance_args(struct btrfs_fs_info *fs_info,
4312 struct btrfs_ioctl_balance_args *bargs)
4314 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4316 bargs->flags = bctl->flags;
4318 if (test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags))
4319 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4320 if (atomic_read(&fs_info->balance_pause_req))
4321 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
4322 if (atomic_read(&fs_info->balance_cancel_req))
4323 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
4325 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4326 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4327 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4329 spin_lock(&fs_info->balance_lock);
4330 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4331 spin_unlock(&fs_info->balance_lock);
4334 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4336 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4337 struct btrfs_fs_info *fs_info = root->fs_info;
4338 struct btrfs_ioctl_balance_args *bargs;
4339 struct btrfs_balance_control *bctl;
4340 bool need_unlock; /* for mut. excl. ops lock */
4345 "IOC_BALANCE ioctl (v1) is deprecated and will be removed in kernel 5.18");
4347 if (!capable(CAP_SYS_ADMIN))
4350 ret = mnt_want_write_file(file);
4355 if (btrfs_exclop_start(fs_info, BTRFS_EXCLOP_BALANCE)) {
4356 mutex_lock(&fs_info->balance_mutex);
4362 * mut. excl. ops lock is locked. Three possibilities:
4363 * (1) some other op is running
4364 * (2) balance is running
4365 * (3) balance is paused -- special case (think resume)
4367 mutex_lock(&fs_info->balance_mutex);
4368 if (fs_info->balance_ctl) {
4369 /* this is either (2) or (3) */
4370 if (!test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
4371 mutex_unlock(&fs_info->balance_mutex);
4373 * Lock released to allow other waiters to continue,
4374 * we'll reexamine the status again.
4376 mutex_lock(&fs_info->balance_mutex);
4378 if (fs_info->balance_ctl &&
4379 !test_bit(BTRFS_FS_BALANCE_RUNNING, &fs_info->flags)) {
4381 need_unlock = false;
4385 mutex_unlock(&fs_info->balance_mutex);
4389 mutex_unlock(&fs_info->balance_mutex);
4395 mutex_unlock(&fs_info->balance_mutex);
4396 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4403 bargs = memdup_user(arg, sizeof(*bargs));
4404 if (IS_ERR(bargs)) {
4405 ret = PTR_ERR(bargs);
4409 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4410 if (!fs_info->balance_ctl) {
4415 bctl = fs_info->balance_ctl;
4416 spin_lock(&fs_info->balance_lock);
4417 bctl->flags |= BTRFS_BALANCE_RESUME;
4418 spin_unlock(&fs_info->balance_lock);
4419 btrfs_exclop_balance(fs_info, BTRFS_EXCLOP_BALANCE);
4427 if (fs_info->balance_ctl) {
4432 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
4439 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4440 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4441 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4443 bctl->flags = bargs->flags;
4445 /* balance everything - no filters */
4446 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4449 if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
4456 * Ownership of bctl and exclusive operation goes to btrfs_balance.
4457 * bctl is freed in reset_balance_state, or, if restriper was paused
4458 * all the way until unmount, in free_fs_info. The flag should be
4459 * cleared after reset_balance_state.
4461 need_unlock = false;
4463 ret = btrfs_balance(fs_info, bctl, bargs);
4466 if ((ret == 0 || ret == -ECANCELED) && arg) {
4467 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4476 mutex_unlock(&fs_info->balance_mutex);
4478 btrfs_exclop_finish(fs_info);
4480 mnt_drop_write_file(file);
4484 static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
4486 if (!capable(CAP_SYS_ADMIN))
4490 case BTRFS_BALANCE_CTL_PAUSE:
4491 return btrfs_pause_balance(fs_info);
4492 case BTRFS_BALANCE_CTL_CANCEL:
4493 return btrfs_cancel_balance(fs_info);
4499 static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
4502 struct btrfs_ioctl_balance_args *bargs;
4505 if (!capable(CAP_SYS_ADMIN))
4508 mutex_lock(&fs_info->balance_mutex);
4509 if (!fs_info->balance_ctl) {
4514 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
4520 btrfs_update_ioctl_balance_args(fs_info, bargs);
4522 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4527 mutex_unlock(&fs_info->balance_mutex);
4531 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4533 struct inode *inode = file_inode(file);
4534 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4535 struct btrfs_ioctl_quota_ctl_args *sa;
4538 if (!capable(CAP_SYS_ADMIN))
4541 ret = mnt_want_write_file(file);
4545 sa = memdup_user(arg, sizeof(*sa));
4551 down_write(&fs_info->subvol_sem);
4554 case BTRFS_QUOTA_CTL_ENABLE:
4555 ret = btrfs_quota_enable(fs_info);
4557 case BTRFS_QUOTA_CTL_DISABLE:
4558 ret = btrfs_quota_disable(fs_info);
4566 up_write(&fs_info->subvol_sem);
4568 mnt_drop_write_file(file);
4572 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4574 struct inode *inode = file_inode(file);
4575 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4576 struct btrfs_root *root = BTRFS_I(inode)->root;
4577 struct btrfs_ioctl_qgroup_assign_args *sa;
4578 struct btrfs_trans_handle *trans;
4582 if (!capable(CAP_SYS_ADMIN))
4585 ret = mnt_want_write_file(file);
4589 sa = memdup_user(arg, sizeof(*sa));
4595 trans = btrfs_join_transaction(root);
4596 if (IS_ERR(trans)) {
4597 ret = PTR_ERR(trans);
4602 ret = btrfs_add_qgroup_relation(trans, sa->src, sa->dst);
4604 ret = btrfs_del_qgroup_relation(trans, sa->src, sa->dst);
4607 /* update qgroup status and info */
4608 err = btrfs_run_qgroups(trans);
4610 btrfs_handle_fs_error(fs_info, err,
4611 "failed to update qgroup status and info");
4612 err = btrfs_end_transaction(trans);
4619 mnt_drop_write_file(file);
4623 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4625 struct inode *inode = file_inode(file);
4626 struct btrfs_root *root = BTRFS_I(inode)->root;
4627 struct btrfs_ioctl_qgroup_create_args *sa;
4628 struct btrfs_trans_handle *trans;
4632 if (!capable(CAP_SYS_ADMIN))
4635 ret = mnt_want_write_file(file);
4639 sa = memdup_user(arg, sizeof(*sa));
4645 if (!sa->qgroupid) {
4650 trans = btrfs_join_transaction(root);
4651 if (IS_ERR(trans)) {
4652 ret = PTR_ERR(trans);
4657 ret = btrfs_create_qgroup(trans, sa->qgroupid);
4659 ret = btrfs_remove_qgroup(trans, sa->qgroupid);
4662 err = btrfs_end_transaction(trans);
4669 mnt_drop_write_file(file);
4673 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4675 struct inode *inode = file_inode(file);
4676 struct btrfs_root *root = BTRFS_I(inode)->root;
4677 struct btrfs_ioctl_qgroup_limit_args *sa;
4678 struct btrfs_trans_handle *trans;
4683 if (!capable(CAP_SYS_ADMIN))
4686 ret = mnt_want_write_file(file);
4690 sa = memdup_user(arg, sizeof(*sa));
4696 trans = btrfs_join_transaction(root);
4697 if (IS_ERR(trans)) {
4698 ret = PTR_ERR(trans);
4702 qgroupid = sa->qgroupid;
4704 /* take the current subvol as qgroup */
4705 qgroupid = root->root_key.objectid;
4708 ret = btrfs_limit_qgroup(trans, qgroupid, &sa->lim);
4710 err = btrfs_end_transaction(trans);
4717 mnt_drop_write_file(file);
4721 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4723 struct inode *inode = file_inode(file);
4724 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4725 struct btrfs_ioctl_quota_rescan_args *qsa;
4728 if (!capable(CAP_SYS_ADMIN))
4731 ret = mnt_want_write_file(file);
4735 qsa = memdup_user(arg, sizeof(*qsa));
4746 ret = btrfs_qgroup_rescan(fs_info);
4751 mnt_drop_write_file(file);
4755 static long btrfs_ioctl_quota_rescan_status(struct btrfs_fs_info *fs_info,
4758 struct btrfs_ioctl_quota_rescan_args qsa = {0};
4760 if (!capable(CAP_SYS_ADMIN))
4763 if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4765 qsa.progress = fs_info->qgroup_rescan_progress.objectid;
4768 if (copy_to_user(arg, &qsa, sizeof(qsa)))
4774 static long btrfs_ioctl_quota_rescan_wait(struct btrfs_fs_info *fs_info,
4777 if (!capable(CAP_SYS_ADMIN))
4780 return btrfs_qgroup_wait_for_completion(fs_info, true);
4783 static long _btrfs_ioctl_set_received_subvol(struct file *file,
4784 struct user_namespace *mnt_userns,
4785 struct btrfs_ioctl_received_subvol_args *sa)
4787 struct inode *inode = file_inode(file);
4788 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4789 struct btrfs_root *root = BTRFS_I(inode)->root;
4790 struct btrfs_root_item *root_item = &root->root_item;
4791 struct btrfs_trans_handle *trans;
4792 struct timespec64 ct = current_time(inode);
4794 int received_uuid_changed;
4796 if (!inode_owner_or_capable(mnt_userns, inode))
4799 ret = mnt_want_write_file(file);
4803 down_write(&fs_info->subvol_sem);
4805 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
4810 if (btrfs_root_readonly(root)) {
4817 * 2 - uuid items (received uuid + subvol uuid)
4819 trans = btrfs_start_transaction(root, 3);
4820 if (IS_ERR(trans)) {
4821 ret = PTR_ERR(trans);
4826 sa->rtransid = trans->transid;
4827 sa->rtime.sec = ct.tv_sec;
4828 sa->rtime.nsec = ct.tv_nsec;
4830 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4832 if (received_uuid_changed &&
4833 !btrfs_is_empty_uuid(root_item->received_uuid)) {
4834 ret = btrfs_uuid_tree_remove(trans, root_item->received_uuid,
4835 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4836 root->root_key.objectid);
4837 if (ret && ret != -ENOENT) {
4838 btrfs_abort_transaction(trans, ret);
4839 btrfs_end_transaction(trans);
4843 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4844 btrfs_set_root_stransid(root_item, sa->stransid);
4845 btrfs_set_root_rtransid(root_item, sa->rtransid);
4846 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4847 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4848 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4849 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4851 ret = btrfs_update_root(trans, fs_info->tree_root,
4852 &root->root_key, &root->root_item);
4854 btrfs_end_transaction(trans);
4857 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4858 ret = btrfs_uuid_tree_add(trans, sa->uuid,
4859 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4860 root->root_key.objectid);
4861 if (ret < 0 && ret != -EEXIST) {
4862 btrfs_abort_transaction(trans, ret);
4863 btrfs_end_transaction(trans);
4867 ret = btrfs_commit_transaction(trans);
4869 up_write(&fs_info->subvol_sem);
4870 mnt_drop_write_file(file);
4875 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4878 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4879 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4882 args32 = memdup_user(arg, sizeof(*args32));
4884 return PTR_ERR(args32);
4886 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
4892 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4893 args64->stransid = args32->stransid;
4894 args64->rtransid = args32->rtransid;
4895 args64->stime.sec = args32->stime.sec;
4896 args64->stime.nsec = args32->stime.nsec;
4897 args64->rtime.sec = args32->rtime.sec;
4898 args64->rtime.nsec = args32->rtime.nsec;
4899 args64->flags = args32->flags;
4901 ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_user_ns(file), args64);
4905 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4906 args32->stransid = args64->stransid;
4907 args32->rtransid = args64->rtransid;
4908 args32->stime.sec = args64->stime.sec;
4909 args32->stime.nsec = args64->stime.nsec;
4910 args32->rtime.sec = args64->rtime.sec;
4911 args32->rtime.nsec = args64->rtime.nsec;
4912 args32->flags = args64->flags;
4914 ret = copy_to_user(arg, args32, sizeof(*args32));
4925 static long btrfs_ioctl_set_received_subvol(struct file *file,
4928 struct btrfs_ioctl_received_subvol_args *sa = NULL;
4931 sa = memdup_user(arg, sizeof(*sa));
4935 ret = _btrfs_ioctl_set_received_subvol(file, file_mnt_user_ns(file), sa);
4940 ret = copy_to_user(arg, sa, sizeof(*sa));
4949 static int btrfs_ioctl_get_fslabel(struct btrfs_fs_info *fs_info,
4954 char label[BTRFS_LABEL_SIZE];
4956 spin_lock(&fs_info->super_lock);
4957 memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4958 spin_unlock(&fs_info->super_lock);
4960 len = strnlen(label, BTRFS_LABEL_SIZE);
4962 if (len == BTRFS_LABEL_SIZE) {
4964 "label is too long, return the first %zu bytes",
4968 ret = copy_to_user(arg, label, len);
4970 return ret ? -EFAULT : 0;
4973 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4975 struct inode *inode = file_inode(file);
4976 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4977 struct btrfs_root *root = BTRFS_I(inode)->root;
4978 struct btrfs_super_block *super_block = fs_info->super_copy;
4979 struct btrfs_trans_handle *trans;
4980 char label[BTRFS_LABEL_SIZE];
4983 if (!capable(CAP_SYS_ADMIN))
4986 if (copy_from_user(label, arg, sizeof(label)))
4989 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
4991 "unable to set label with more than %d bytes",
4992 BTRFS_LABEL_SIZE - 1);
4996 ret = mnt_want_write_file(file);
5000 trans = btrfs_start_transaction(root, 0);
5001 if (IS_ERR(trans)) {
5002 ret = PTR_ERR(trans);
5006 spin_lock(&fs_info->super_lock);
5007 strcpy(super_block->label, label);
5008 spin_unlock(&fs_info->super_lock);
5009 ret = btrfs_commit_transaction(trans);
5012 mnt_drop_write_file(file);
5016 #define INIT_FEATURE_FLAGS(suffix) \
5017 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5018 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5019 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5021 int btrfs_ioctl_get_supported_features(void __user *arg)
5023 static const struct btrfs_ioctl_feature_flags features[3] = {
5024 INIT_FEATURE_FLAGS(SUPP),
5025 INIT_FEATURE_FLAGS(SAFE_SET),
5026 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5029 if (copy_to_user(arg, &features, sizeof(features)))
5035 static int btrfs_ioctl_get_features(struct btrfs_fs_info *fs_info,
5038 struct btrfs_super_block *super_block = fs_info->super_copy;
5039 struct btrfs_ioctl_feature_flags features;
5041 features.compat_flags = btrfs_super_compat_flags(super_block);
5042 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5043 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5045 if (copy_to_user(arg, &features, sizeof(features)))
5051 static int check_feature_bits(struct btrfs_fs_info *fs_info,
5052 enum btrfs_feature_set set,
5053 u64 change_mask, u64 flags, u64 supported_flags,
5054 u64 safe_set, u64 safe_clear)
5056 const char *type = btrfs_feature_set_name(set);
5058 u64 disallowed, unsupported;
5059 u64 set_mask = flags & change_mask;
5060 u64 clear_mask = ~flags & change_mask;
5062 unsupported = set_mask & ~supported_flags;
5064 names = btrfs_printable_features(set, unsupported);
5067 "this kernel does not support the %s feature bit%s",
5068 names, strchr(names, ',') ? "s" : "");
5072 "this kernel does not support %s bits 0x%llx",
5077 disallowed = set_mask & ~safe_set;
5079 names = btrfs_printable_features(set, disallowed);
5082 "can't set the %s feature bit%s while mounted",
5083 names, strchr(names, ',') ? "s" : "");
5087 "can't set %s bits 0x%llx while mounted",
5092 disallowed = clear_mask & ~safe_clear;
5094 names = btrfs_printable_features(set, disallowed);
5097 "can't clear the %s feature bit%s while mounted",
5098 names, strchr(names, ',') ? "s" : "");
5102 "can't clear %s bits 0x%llx while mounted",
5110 #define check_feature(fs_info, change_mask, flags, mask_base) \
5111 check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \
5112 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5113 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5114 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5116 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5118 struct inode *inode = file_inode(file);
5119 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5120 struct btrfs_root *root = BTRFS_I(inode)->root;
5121 struct btrfs_super_block *super_block = fs_info->super_copy;
5122 struct btrfs_ioctl_feature_flags flags[2];
5123 struct btrfs_trans_handle *trans;
5127 if (!capable(CAP_SYS_ADMIN))
5130 if (copy_from_user(flags, arg, sizeof(flags)))
5134 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5135 !flags[0].incompat_flags)
5138 ret = check_feature(fs_info, flags[0].compat_flags,
5139 flags[1].compat_flags, COMPAT);
5143 ret = check_feature(fs_info, flags[0].compat_ro_flags,
5144 flags[1].compat_ro_flags, COMPAT_RO);
5148 ret = check_feature(fs_info, flags[0].incompat_flags,
5149 flags[1].incompat_flags, INCOMPAT);
5153 ret = mnt_want_write_file(file);
5157 trans = btrfs_start_transaction(root, 0);
5158 if (IS_ERR(trans)) {
5159 ret = PTR_ERR(trans);
5160 goto out_drop_write;
5163 spin_lock(&fs_info->super_lock);
5164 newflags = btrfs_super_compat_flags(super_block);
5165 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5166 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5167 btrfs_set_super_compat_flags(super_block, newflags);
5169 newflags = btrfs_super_compat_ro_flags(super_block);
5170 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5171 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5172 btrfs_set_super_compat_ro_flags(super_block, newflags);
5174 newflags = btrfs_super_incompat_flags(super_block);
5175 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5176 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5177 btrfs_set_super_incompat_flags(super_block, newflags);
5178 spin_unlock(&fs_info->super_lock);
5180 ret = btrfs_commit_transaction(trans);
5182 mnt_drop_write_file(file);
5187 static int _btrfs_ioctl_send(struct inode *inode, void __user *argp, bool compat)
5189 struct btrfs_ioctl_send_args *arg;
5193 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5194 struct btrfs_ioctl_send_args_32 args32;
5196 ret = copy_from_user(&args32, argp, sizeof(args32));
5199 arg = kzalloc(sizeof(*arg), GFP_KERNEL);
5202 arg->send_fd = args32.send_fd;
5203 arg->clone_sources_count = args32.clone_sources_count;
5204 arg->clone_sources = compat_ptr(args32.clone_sources);
5205 arg->parent_root = args32.parent_root;
5206 arg->flags = args32.flags;
5207 memcpy(arg->reserved, args32.reserved,
5208 sizeof(args32.reserved));
5213 arg = memdup_user(argp, sizeof(*arg));
5215 return PTR_ERR(arg);
5217 ret = btrfs_ioctl_send(inode, arg);
5222 static int btrfs_ioctl_encoded_read(struct file *file, void __user *argp,
5225 struct btrfs_ioctl_encoded_io_args args = { 0 };
5226 size_t copy_end_kernel = offsetofend(struct btrfs_ioctl_encoded_io_args,
5229 struct iovec iovstack[UIO_FASTIOV];
5230 struct iovec *iov = iovstack;
5231 struct iov_iter iter;
5236 if (!capable(CAP_SYS_ADMIN)) {
5242 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5243 struct btrfs_ioctl_encoded_io_args_32 args32;
5245 copy_end = offsetofend(struct btrfs_ioctl_encoded_io_args_32,
5247 if (copy_from_user(&args32, argp, copy_end)) {
5251 args.iov = compat_ptr(args32.iov);
5252 args.iovcnt = args32.iovcnt;
5253 args.offset = args32.offset;
5254 args.flags = args32.flags;
5259 copy_end = copy_end_kernel;
5260 if (copy_from_user(&args, argp, copy_end)) {
5265 if (args.flags != 0) {
5270 ret = import_iovec(READ, args.iov, args.iovcnt, ARRAY_SIZE(iovstack),
5275 if (iov_iter_count(&iter) == 0) {
5280 ret = rw_verify_area(READ, file, &pos, args.len);
5284 init_sync_kiocb(&kiocb, file);
5287 ret = btrfs_encoded_read(&kiocb, &iter, &args);
5289 fsnotify_access(file);
5290 if (copy_to_user(argp + copy_end,
5291 (char *)&args + copy_end_kernel,
5292 sizeof(args) - copy_end_kernel))
5300 add_rchar(current, ret);
5305 static int btrfs_ioctl_encoded_write(struct file *file, void __user *argp, bool compat)
5307 struct btrfs_ioctl_encoded_io_args args;
5308 struct iovec iovstack[UIO_FASTIOV];
5309 struct iovec *iov = iovstack;
5310 struct iov_iter iter;
5315 if (!capable(CAP_SYS_ADMIN)) {
5320 if (!(file->f_mode & FMODE_WRITE)) {
5326 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5327 struct btrfs_ioctl_encoded_io_args_32 args32;
5329 if (copy_from_user(&args32, argp, sizeof(args32))) {
5333 args.iov = compat_ptr(args32.iov);
5334 args.iovcnt = args32.iovcnt;
5335 args.offset = args32.offset;
5336 args.flags = args32.flags;
5337 args.len = args32.len;
5338 args.unencoded_len = args32.unencoded_len;
5339 args.unencoded_offset = args32.unencoded_offset;
5340 args.compression = args32.compression;
5341 args.encryption = args32.encryption;
5342 memcpy(args.reserved, args32.reserved, sizeof(args.reserved));
5347 if (copy_from_user(&args, argp, sizeof(args))) {
5354 if (args.flags != 0)
5356 if (memchr_inv(args.reserved, 0, sizeof(args.reserved)))
5358 if (args.compression == BTRFS_ENCODED_IO_COMPRESSION_NONE &&
5359 args.encryption == BTRFS_ENCODED_IO_ENCRYPTION_NONE)
5361 if (args.compression >= BTRFS_ENCODED_IO_COMPRESSION_TYPES ||
5362 args.encryption >= BTRFS_ENCODED_IO_ENCRYPTION_TYPES)
5364 if (args.unencoded_offset > args.unencoded_len)
5366 if (args.len > args.unencoded_len - args.unencoded_offset)
5369 ret = import_iovec(WRITE, args.iov, args.iovcnt, ARRAY_SIZE(iovstack),
5374 file_start_write(file);
5376 if (iov_iter_count(&iter) == 0) {
5381 ret = rw_verify_area(WRITE, file, &pos, args.len);
5385 init_sync_kiocb(&kiocb, file);
5386 ret = kiocb_set_rw_flags(&kiocb, 0);
5391 ret = btrfs_do_write_iter(&kiocb, &iter, &args);
5393 fsnotify_modify(file);
5396 file_end_write(file);
5400 add_wchar(current, ret);
5405 long btrfs_ioctl(struct file *file, unsigned int
5406 cmd, unsigned long arg)
5408 struct inode *inode = file_inode(file);
5409 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5410 struct btrfs_root *root = BTRFS_I(inode)->root;
5411 void __user *argp = (void __user *)arg;
5414 case FS_IOC_GETVERSION:
5415 return btrfs_ioctl_getversion(inode, argp);
5416 case FS_IOC_GETFSLABEL:
5417 return btrfs_ioctl_get_fslabel(fs_info, argp);
5418 case FS_IOC_SETFSLABEL:
5419 return btrfs_ioctl_set_fslabel(file, argp);
5421 return btrfs_ioctl_fitrim(fs_info, argp);
5422 case BTRFS_IOC_SNAP_CREATE:
5423 return btrfs_ioctl_snap_create(file, argp, 0);
5424 case BTRFS_IOC_SNAP_CREATE_V2:
5425 return btrfs_ioctl_snap_create_v2(file, argp, 0);
5426 case BTRFS_IOC_SUBVOL_CREATE:
5427 return btrfs_ioctl_snap_create(file, argp, 1);
5428 case BTRFS_IOC_SUBVOL_CREATE_V2:
5429 return btrfs_ioctl_snap_create_v2(file, argp, 1);
5430 case BTRFS_IOC_SNAP_DESTROY:
5431 return btrfs_ioctl_snap_destroy(file, argp, false);
5432 case BTRFS_IOC_SNAP_DESTROY_V2:
5433 return btrfs_ioctl_snap_destroy(file, argp, true);
5434 case BTRFS_IOC_SUBVOL_GETFLAGS:
5435 return btrfs_ioctl_subvol_getflags(inode, argp);
5436 case BTRFS_IOC_SUBVOL_SETFLAGS:
5437 return btrfs_ioctl_subvol_setflags(file, argp);
5438 case BTRFS_IOC_DEFAULT_SUBVOL:
5439 return btrfs_ioctl_default_subvol(file, argp);
5440 case BTRFS_IOC_DEFRAG:
5441 return btrfs_ioctl_defrag(file, NULL);
5442 case BTRFS_IOC_DEFRAG_RANGE:
5443 return btrfs_ioctl_defrag(file, argp);
5444 case BTRFS_IOC_RESIZE:
5445 return btrfs_ioctl_resize(file, argp);
5446 case BTRFS_IOC_ADD_DEV:
5447 return btrfs_ioctl_add_dev(fs_info, argp);
5448 case BTRFS_IOC_RM_DEV:
5449 return btrfs_ioctl_rm_dev(file, argp);
5450 case BTRFS_IOC_RM_DEV_V2:
5451 return btrfs_ioctl_rm_dev_v2(file, argp);
5452 case BTRFS_IOC_FS_INFO:
5453 return btrfs_ioctl_fs_info(fs_info, argp);
5454 case BTRFS_IOC_DEV_INFO:
5455 return btrfs_ioctl_dev_info(fs_info, argp);
5456 case BTRFS_IOC_TREE_SEARCH:
5457 return btrfs_ioctl_tree_search(inode, argp);
5458 case BTRFS_IOC_TREE_SEARCH_V2:
5459 return btrfs_ioctl_tree_search_v2(inode, argp);
5460 case BTRFS_IOC_INO_LOOKUP:
5461 return btrfs_ioctl_ino_lookup(root, argp);
5462 case BTRFS_IOC_INO_PATHS:
5463 return btrfs_ioctl_ino_to_path(root, argp);
5464 case BTRFS_IOC_LOGICAL_INO:
5465 return btrfs_ioctl_logical_to_ino(fs_info, argp, 1);
5466 case BTRFS_IOC_LOGICAL_INO_V2:
5467 return btrfs_ioctl_logical_to_ino(fs_info, argp, 2);
5468 case BTRFS_IOC_SPACE_INFO:
5469 return btrfs_ioctl_space_info(fs_info, argp);
5470 case BTRFS_IOC_SYNC: {
5473 ret = btrfs_start_delalloc_roots(fs_info, LONG_MAX, false);
5476 ret = btrfs_sync_fs(inode->i_sb, 1);
5478 * The transaction thread may want to do more work,
5479 * namely it pokes the cleaner kthread that will start
5480 * processing uncleaned subvols.
5482 wake_up_process(fs_info->transaction_kthread);
5485 case BTRFS_IOC_START_SYNC:
5486 return btrfs_ioctl_start_sync(root, argp);
5487 case BTRFS_IOC_WAIT_SYNC:
5488 return btrfs_ioctl_wait_sync(fs_info, argp);
5489 case BTRFS_IOC_SCRUB:
5490 return btrfs_ioctl_scrub(file, argp);
5491 case BTRFS_IOC_SCRUB_CANCEL:
5492 return btrfs_ioctl_scrub_cancel(fs_info);
5493 case BTRFS_IOC_SCRUB_PROGRESS:
5494 return btrfs_ioctl_scrub_progress(fs_info, argp);
5495 case BTRFS_IOC_BALANCE_V2:
5496 return btrfs_ioctl_balance(file, argp);
5497 case BTRFS_IOC_BALANCE_CTL:
5498 return btrfs_ioctl_balance_ctl(fs_info, arg);
5499 case BTRFS_IOC_BALANCE_PROGRESS:
5500 return btrfs_ioctl_balance_progress(fs_info, argp);
5501 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5502 return btrfs_ioctl_set_received_subvol(file, argp);
5504 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5505 return btrfs_ioctl_set_received_subvol_32(file, argp);
5507 case BTRFS_IOC_SEND:
5508 return _btrfs_ioctl_send(inode, argp, false);
5509 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5510 case BTRFS_IOC_SEND_32:
5511 return _btrfs_ioctl_send(inode, argp, true);
5513 case BTRFS_IOC_GET_DEV_STATS:
5514 return btrfs_ioctl_get_dev_stats(fs_info, argp);
5515 case BTRFS_IOC_QUOTA_CTL:
5516 return btrfs_ioctl_quota_ctl(file, argp);
5517 case BTRFS_IOC_QGROUP_ASSIGN:
5518 return btrfs_ioctl_qgroup_assign(file, argp);
5519 case BTRFS_IOC_QGROUP_CREATE:
5520 return btrfs_ioctl_qgroup_create(file, argp);
5521 case BTRFS_IOC_QGROUP_LIMIT:
5522 return btrfs_ioctl_qgroup_limit(file, argp);
5523 case BTRFS_IOC_QUOTA_RESCAN:
5524 return btrfs_ioctl_quota_rescan(file, argp);
5525 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5526 return btrfs_ioctl_quota_rescan_status(fs_info, argp);
5527 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5528 return btrfs_ioctl_quota_rescan_wait(fs_info, argp);
5529 case BTRFS_IOC_DEV_REPLACE:
5530 return btrfs_ioctl_dev_replace(fs_info, argp);
5531 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5532 return btrfs_ioctl_get_supported_features(argp);
5533 case BTRFS_IOC_GET_FEATURES:
5534 return btrfs_ioctl_get_features(fs_info, argp);
5535 case BTRFS_IOC_SET_FEATURES:
5536 return btrfs_ioctl_set_features(file, argp);
5537 case BTRFS_IOC_GET_SUBVOL_INFO:
5538 return btrfs_ioctl_get_subvol_info(inode, argp);
5539 case BTRFS_IOC_GET_SUBVOL_ROOTREF:
5540 return btrfs_ioctl_get_subvol_rootref(root, argp);
5541 case BTRFS_IOC_INO_LOOKUP_USER:
5542 return btrfs_ioctl_ino_lookup_user(file, argp);
5543 case FS_IOC_ENABLE_VERITY:
5544 return fsverity_ioctl_enable(file, (const void __user *)argp);
5545 case FS_IOC_MEASURE_VERITY:
5546 return fsverity_ioctl_measure(file, argp);
5547 case BTRFS_IOC_ENCODED_READ:
5548 return btrfs_ioctl_encoded_read(file, argp, false);
5549 case BTRFS_IOC_ENCODED_WRITE:
5550 return btrfs_ioctl_encoded_write(file, argp, false);
5551 #if defined(CONFIG_64BIT) && defined(CONFIG_COMPAT)
5552 case BTRFS_IOC_ENCODED_READ_32:
5553 return btrfs_ioctl_encoded_read(file, argp, true);
5554 case BTRFS_IOC_ENCODED_WRITE_32:
5555 return btrfs_ioctl_encoded_write(file, argp, true);
5562 #ifdef CONFIG_COMPAT
5563 long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
5566 * These all access 32-bit values anyway so no further
5567 * handling is necessary.
5570 case FS_IOC32_GETVERSION:
5571 cmd = FS_IOC_GETVERSION;
5575 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));