2 * Copyright (C) 2007 Oracle. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
19 #include <linux/kernel.h>
20 #include <linux/bio.h>
21 #include <linux/buffer_head.h>
22 #include <linux/file.h>
24 #include <linux/fsnotify.h>
25 #include <linux/pagemap.h>
26 #include <linux/highmem.h>
27 #include <linux/time.h>
28 #include <linux/init.h>
29 #include <linux/string.h>
30 #include <linux/backing-dev.h>
31 #include <linux/mount.h>
32 #include <linux/mpage.h>
33 #include <linux/namei.h>
34 #include <linux/swap.h>
35 #include <linux/writeback.h>
36 #include <linux/statfs.h>
37 #include <linux/compat.h>
38 #include <linux/bit_spinlock.h>
39 #include <linux/security.h>
40 #include <linux/xattr.h>
41 #include <linux/vmalloc.h>
42 #include <linux/slab.h>
43 #include <linux/blkdev.h>
44 #include <linux/uuid.h>
45 #include <linux/btrfs.h>
46 #include <linux/uaccess.h>
49 #include "transaction.h"
50 #include "btrfs_inode.h"
51 #include "print-tree.h"
54 #include "inode-map.h"
56 #include "rcu-string.h"
58 #include "dev-replace.h"
63 #include "compression.h"
66 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
67 * structures are incorrect, as the timespec structure from userspace
68 * is 4 bytes too small. We define these alternatives here to teach
69 * the kernel about the 32-bit struct packing.
71 struct btrfs_ioctl_timespec_32 {
74 } __attribute__ ((__packed__));
76 struct btrfs_ioctl_received_subvol_args_32 {
77 char uuid[BTRFS_UUID_SIZE]; /* in */
78 __u64 stransid; /* in */
79 __u64 rtransid; /* out */
80 struct btrfs_ioctl_timespec_32 stime; /* in */
81 struct btrfs_ioctl_timespec_32 rtime; /* out */
83 __u64 reserved[16]; /* in */
84 } __attribute__ ((__packed__));
86 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
87 struct btrfs_ioctl_received_subvol_args_32)
91 static int btrfs_clone(struct inode *src, struct inode *inode,
92 u64 off, u64 olen, u64 olen_aligned, u64 destoff,
95 /* Mask out flags that are inappropriate for the given type of inode. */
96 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
100 else if (S_ISREG(mode))
101 return flags & ~FS_DIRSYNC_FL;
103 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
107 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
109 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
111 unsigned int iflags = 0;
113 if (flags & BTRFS_INODE_SYNC)
114 iflags |= FS_SYNC_FL;
115 if (flags & BTRFS_INODE_IMMUTABLE)
116 iflags |= FS_IMMUTABLE_FL;
117 if (flags & BTRFS_INODE_APPEND)
118 iflags |= FS_APPEND_FL;
119 if (flags & BTRFS_INODE_NODUMP)
120 iflags |= FS_NODUMP_FL;
121 if (flags & BTRFS_INODE_NOATIME)
122 iflags |= FS_NOATIME_FL;
123 if (flags & BTRFS_INODE_DIRSYNC)
124 iflags |= FS_DIRSYNC_FL;
125 if (flags & BTRFS_INODE_NODATACOW)
126 iflags |= FS_NOCOW_FL;
128 if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
129 iflags |= FS_COMPR_FL;
130 else if (flags & BTRFS_INODE_NOCOMPRESS)
131 iflags |= FS_NOCOMP_FL;
137 * Update inode->i_flags based on the btrfs internal flags.
139 void btrfs_update_iflags(struct inode *inode)
141 struct btrfs_inode *ip = BTRFS_I(inode);
142 unsigned int new_fl = 0;
144 if (ip->flags & BTRFS_INODE_SYNC)
146 if (ip->flags & BTRFS_INODE_IMMUTABLE)
147 new_fl |= S_IMMUTABLE;
148 if (ip->flags & BTRFS_INODE_APPEND)
150 if (ip->flags & BTRFS_INODE_NOATIME)
152 if (ip->flags & BTRFS_INODE_DIRSYNC)
155 set_mask_bits(&inode->i_flags,
156 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC,
161 * Inherit flags from the parent inode.
163 * Currently only the compression flags and the cow flags are inherited.
165 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
172 flags = BTRFS_I(dir)->flags;
174 if (flags & BTRFS_INODE_NOCOMPRESS) {
175 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
176 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
177 } else if (flags & BTRFS_INODE_COMPRESS) {
178 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
179 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
182 if (flags & BTRFS_INODE_NODATACOW) {
183 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
184 if (S_ISREG(inode->i_mode))
185 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
188 btrfs_update_iflags(inode);
191 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
193 struct btrfs_inode *ip = BTRFS_I(file_inode(file));
194 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
196 if (copy_to_user(arg, &flags, sizeof(flags)))
201 static int check_flags(unsigned int flags)
203 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
204 FS_NOATIME_FL | FS_NODUMP_FL | \
205 FS_SYNC_FL | FS_DIRSYNC_FL | \
206 FS_NOCOMP_FL | FS_COMPR_FL |
210 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
216 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
218 struct inode *inode = file_inode(file);
219 struct btrfs_inode *ip = BTRFS_I(inode);
220 struct btrfs_root *root = ip->root;
221 struct btrfs_trans_handle *trans;
222 unsigned int flags, oldflags;
225 unsigned int i_oldflags;
228 if (!inode_owner_or_capable(inode))
231 if (btrfs_root_readonly(root))
234 if (copy_from_user(&flags, arg, sizeof(flags)))
237 ret = check_flags(flags);
241 ret = mnt_want_write_file(file);
247 ip_oldflags = ip->flags;
248 i_oldflags = inode->i_flags;
249 mode = inode->i_mode;
251 flags = btrfs_mask_flags(inode->i_mode, flags);
252 oldflags = btrfs_flags_to_ioctl(ip->flags);
253 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
254 if (!capable(CAP_LINUX_IMMUTABLE)) {
260 if (flags & FS_SYNC_FL)
261 ip->flags |= BTRFS_INODE_SYNC;
263 ip->flags &= ~BTRFS_INODE_SYNC;
264 if (flags & FS_IMMUTABLE_FL)
265 ip->flags |= BTRFS_INODE_IMMUTABLE;
267 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
268 if (flags & FS_APPEND_FL)
269 ip->flags |= BTRFS_INODE_APPEND;
271 ip->flags &= ~BTRFS_INODE_APPEND;
272 if (flags & FS_NODUMP_FL)
273 ip->flags |= BTRFS_INODE_NODUMP;
275 ip->flags &= ~BTRFS_INODE_NODUMP;
276 if (flags & FS_NOATIME_FL)
277 ip->flags |= BTRFS_INODE_NOATIME;
279 ip->flags &= ~BTRFS_INODE_NOATIME;
280 if (flags & FS_DIRSYNC_FL)
281 ip->flags |= BTRFS_INODE_DIRSYNC;
283 ip->flags &= ~BTRFS_INODE_DIRSYNC;
284 if (flags & FS_NOCOW_FL) {
287 * It's safe to turn csums off here, no extents exist.
288 * Otherwise we want the flag to reflect the real COW
289 * status of the file and will not set it.
291 if (inode->i_size == 0)
292 ip->flags |= BTRFS_INODE_NODATACOW
293 | BTRFS_INODE_NODATASUM;
295 ip->flags |= BTRFS_INODE_NODATACOW;
299 * Revert back under same assuptions as above
302 if (inode->i_size == 0)
303 ip->flags &= ~(BTRFS_INODE_NODATACOW
304 | BTRFS_INODE_NODATASUM);
306 ip->flags &= ~BTRFS_INODE_NODATACOW;
311 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
312 * flag may be changed automatically if compression code won't make
315 if (flags & FS_NOCOMP_FL) {
316 ip->flags &= ~BTRFS_INODE_COMPRESS;
317 ip->flags |= BTRFS_INODE_NOCOMPRESS;
319 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
320 if (ret && ret != -ENODATA)
322 } else if (flags & FS_COMPR_FL) {
325 ip->flags |= BTRFS_INODE_COMPRESS;
326 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
328 if (root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
332 ret = btrfs_set_prop(inode, "btrfs.compression",
333 comp, strlen(comp), 0);
338 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
339 if (ret && ret != -ENODATA)
341 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
344 trans = btrfs_start_transaction(root, 1);
346 ret = PTR_ERR(trans);
350 btrfs_update_iflags(inode);
351 inode_inc_iversion(inode);
352 inode->i_ctime = current_fs_time(inode->i_sb);
353 ret = btrfs_update_inode(trans, root, inode);
355 btrfs_end_transaction(trans, root);
358 ip->flags = ip_oldflags;
359 inode->i_flags = i_oldflags;
364 mnt_drop_write_file(file);
368 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
370 struct inode *inode = file_inode(file);
372 return put_user(inode->i_generation, arg);
375 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
377 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
378 struct btrfs_device *device;
379 struct request_queue *q;
380 struct fstrim_range range;
381 u64 minlen = ULLONG_MAX;
383 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
386 if (!capable(CAP_SYS_ADMIN))
390 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
394 q = bdev_get_queue(device->bdev);
395 if (blk_queue_discard(q)) {
397 minlen = min((u64)q->limits.discard_granularity,
405 if (copy_from_user(&range, arg, sizeof(range)))
407 if (range.start > total_bytes ||
408 range.len < fs_info->sb->s_blocksize)
411 range.len = min(range.len, total_bytes - range.start);
412 range.minlen = max(range.minlen, minlen);
413 ret = btrfs_trim_fs(fs_info->tree_root, &range);
417 if (copy_to_user(arg, &range, sizeof(range)))
423 int btrfs_is_empty_uuid(u8 *uuid)
427 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
434 static noinline int create_subvol(struct inode *dir,
435 struct dentry *dentry,
436 char *name, int namelen,
438 struct btrfs_qgroup_inherit *inherit)
440 struct btrfs_trans_handle *trans;
441 struct btrfs_key key;
442 struct btrfs_root_item root_item;
443 struct btrfs_inode_item *inode_item;
444 struct extent_buffer *leaf;
445 struct btrfs_root *root = BTRFS_I(dir)->root;
446 struct btrfs_root *new_root;
447 struct btrfs_block_rsv block_rsv;
448 struct timespec cur_time = current_fs_time(dir->i_sb);
453 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
458 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
463 * Don't create subvolume whose level is not zero. Or qgroup will be
464 * screwed up since it assume subvolme qgroup's level to be 0.
466 if (btrfs_qgroup_level(objectid))
469 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
471 * The same as the snapshot creation, please see the comment
472 * of create_snapshot().
474 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
475 8, &qgroup_reserved, false);
479 trans = btrfs_start_transaction(root, 0);
481 ret = PTR_ERR(trans);
482 btrfs_subvolume_release_metadata(root, &block_rsv,
486 trans->block_rsv = &block_rsv;
487 trans->bytes_reserved = block_rsv.size;
489 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
493 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
499 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
500 btrfs_set_header_bytenr(leaf, leaf->start);
501 btrfs_set_header_generation(leaf, trans->transid);
502 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
503 btrfs_set_header_owner(leaf, objectid);
505 write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(),
507 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
508 btrfs_header_chunk_tree_uuid(leaf),
510 btrfs_mark_buffer_dirty(leaf);
512 memset(&root_item, 0, sizeof(root_item));
514 inode_item = &root_item.inode;
515 btrfs_set_stack_inode_generation(inode_item, 1);
516 btrfs_set_stack_inode_size(inode_item, 3);
517 btrfs_set_stack_inode_nlink(inode_item, 1);
518 btrfs_set_stack_inode_nbytes(inode_item, root->nodesize);
519 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
521 btrfs_set_root_flags(&root_item, 0);
522 btrfs_set_root_limit(&root_item, 0);
523 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
525 btrfs_set_root_bytenr(&root_item, leaf->start);
526 btrfs_set_root_generation(&root_item, trans->transid);
527 btrfs_set_root_level(&root_item, 0);
528 btrfs_set_root_refs(&root_item, 1);
529 btrfs_set_root_used(&root_item, leaf->len);
530 btrfs_set_root_last_snapshot(&root_item, 0);
532 btrfs_set_root_generation_v2(&root_item,
533 btrfs_root_generation(&root_item));
534 uuid_le_gen(&new_uuid);
535 memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
536 btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec);
537 btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec);
538 root_item.ctime = root_item.otime;
539 btrfs_set_root_ctransid(&root_item, trans->transid);
540 btrfs_set_root_otransid(&root_item, trans->transid);
542 btrfs_tree_unlock(leaf);
543 free_extent_buffer(leaf);
546 btrfs_set_root_dirid(&root_item, new_dirid);
548 key.objectid = objectid;
550 key.type = BTRFS_ROOT_ITEM_KEY;
551 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
556 key.offset = (u64)-1;
557 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
558 if (IS_ERR(new_root)) {
559 ret = PTR_ERR(new_root);
560 btrfs_abort_transaction(trans, root, ret);
564 btrfs_record_root_in_trans(trans, new_root);
566 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
568 /* We potentially lose an unused inode item here */
569 btrfs_abort_transaction(trans, root, ret);
573 mutex_lock(&new_root->objectid_mutex);
574 new_root->highest_objectid = new_dirid;
575 mutex_unlock(&new_root->objectid_mutex);
578 * insert the directory item
580 ret = btrfs_set_inode_index(dir, &index);
582 btrfs_abort_transaction(trans, root, ret);
586 ret = btrfs_insert_dir_item(trans, root,
587 name, namelen, dir, &key,
588 BTRFS_FT_DIR, index);
590 btrfs_abort_transaction(trans, root, ret);
594 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
595 ret = btrfs_update_inode(trans, root, dir);
598 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
599 objectid, root->root_key.objectid,
600 btrfs_ino(dir), index, name, namelen);
603 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
604 root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
607 btrfs_abort_transaction(trans, root, ret);
610 trans->block_rsv = NULL;
611 trans->bytes_reserved = 0;
612 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
615 *async_transid = trans->transid;
616 err = btrfs_commit_transaction_async(trans, root, 1);
618 err = btrfs_commit_transaction(trans, root);
620 err = btrfs_commit_transaction(trans, root);
626 inode = btrfs_lookup_dentry(dir, dentry);
628 return PTR_ERR(inode);
629 d_instantiate(dentry, inode);
634 static void btrfs_wait_for_no_snapshoting_writes(struct btrfs_root *root)
640 prepare_to_wait(&root->subv_writers->wait, &wait,
641 TASK_UNINTERRUPTIBLE);
643 writers = percpu_counter_sum(&root->subv_writers->counter);
647 finish_wait(&root->subv_writers->wait, &wait);
651 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
652 struct dentry *dentry, char *name, int namelen,
653 u64 *async_transid, bool readonly,
654 struct btrfs_qgroup_inherit *inherit)
657 struct btrfs_pending_snapshot *pending_snapshot;
658 struct btrfs_trans_handle *trans;
661 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
664 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
665 if (!pending_snapshot)
668 pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
670 pending_snapshot->path = btrfs_alloc_path();
671 if (!pending_snapshot->root_item || !pending_snapshot->path) {
676 atomic_inc(&root->will_be_snapshoted);
677 smp_mb__after_atomic();
678 btrfs_wait_for_no_snapshoting_writes(root);
680 ret = btrfs_start_delalloc_inodes(root, 0);
684 btrfs_wait_ordered_extents(root, -1);
686 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
687 BTRFS_BLOCK_RSV_TEMP);
689 * 1 - parent dir inode
692 * 2 - root ref/backref
693 * 1 - root of snapshot
696 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
697 &pending_snapshot->block_rsv, 8,
698 &pending_snapshot->qgroup_reserved,
703 pending_snapshot->dentry = dentry;
704 pending_snapshot->root = root;
705 pending_snapshot->readonly = readonly;
706 pending_snapshot->dir = dir;
707 pending_snapshot->inherit = inherit;
709 trans = btrfs_start_transaction(root, 0);
711 ret = PTR_ERR(trans);
715 spin_lock(&root->fs_info->trans_lock);
716 list_add(&pending_snapshot->list,
717 &trans->transaction->pending_snapshots);
718 spin_unlock(&root->fs_info->trans_lock);
720 *async_transid = trans->transid;
721 ret = btrfs_commit_transaction_async(trans,
722 root->fs_info->extent_root, 1);
724 ret = btrfs_commit_transaction(trans, root);
726 ret = btrfs_commit_transaction(trans,
727 root->fs_info->extent_root);
732 ret = pending_snapshot->error;
736 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
740 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
742 ret = PTR_ERR(inode);
746 d_instantiate(dentry, inode);
749 btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
750 &pending_snapshot->block_rsv,
751 pending_snapshot->qgroup_reserved);
753 if (atomic_dec_and_test(&root->will_be_snapshoted))
754 wake_up_atomic_t(&root->will_be_snapshoted);
756 kfree(pending_snapshot->root_item);
757 btrfs_free_path(pending_snapshot->path);
758 kfree(pending_snapshot);
763 /* copy of may_delete in fs/namei.c()
764 * Check whether we can remove a link victim from directory dir, check
765 * whether the type of victim is right.
766 * 1. We can't do it if dir is read-only (done in permission())
767 * 2. We should have write and exec permissions on dir
768 * 3. We can't remove anything from append-only dir
769 * 4. We can't do anything with immutable dir (done in permission())
770 * 5. If the sticky bit on dir is set we should either
771 * a. be owner of dir, or
772 * b. be owner of victim, or
773 * c. have CAP_FOWNER capability
774 * 6. If the victim is append-only or immutable we can't do antyhing with
775 * links pointing to it.
776 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
777 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
778 * 9. We can't remove a root or mountpoint.
779 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
780 * nfs_async_unlink().
783 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
787 if (d_really_is_negative(victim))
790 BUG_ON(d_inode(victim->d_parent) != dir);
791 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
793 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
798 if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) ||
799 IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim)))
802 if (!d_is_dir(victim))
806 } else if (d_is_dir(victim))
810 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
815 /* copy of may_create in fs/namei.c() */
816 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
818 if (d_really_is_positive(child))
822 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
826 * Create a new subvolume below @parent. This is largely modeled after
827 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
828 * inside this filesystem so it's quite a bit simpler.
830 static noinline int btrfs_mksubvol(struct path *parent,
831 char *name, int namelen,
832 struct btrfs_root *snap_src,
833 u64 *async_transid, bool readonly,
834 struct btrfs_qgroup_inherit *inherit)
836 struct inode *dir = d_inode(parent->dentry);
837 struct dentry *dentry;
840 inode_lock_nested(dir, I_MUTEX_PARENT);
841 // XXX: should've been
842 // mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
843 // if (error == -EINTR)
846 dentry = lookup_one_len(name, parent->dentry, namelen);
847 error = PTR_ERR(dentry);
851 error = btrfs_may_create(dir, dentry);
856 * even if this name doesn't exist, we may get hash collisions.
857 * check for them now when we can safely fail
859 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
865 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
867 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
871 error = create_snapshot(snap_src, dir, dentry, name, namelen,
872 async_transid, readonly, inherit);
874 error = create_subvol(dir, dentry, name, namelen,
875 async_transid, inherit);
878 fsnotify_mkdir(dir, dentry);
880 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
889 * When we're defragging a range, we don't want to kick it off again
890 * if it is really just waiting for delalloc to send it down.
891 * If we find a nice big extent or delalloc range for the bytes in the
892 * file you want to defrag, we return 0 to let you know to skip this
895 static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
897 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
898 struct extent_map *em = NULL;
899 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
902 read_lock(&em_tree->lock);
903 em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE);
904 read_unlock(&em_tree->lock);
907 end = extent_map_end(em);
909 if (end - offset > thresh)
912 /* if we already have a nice delalloc here, just stop */
914 end = count_range_bits(io_tree, &offset, offset + thresh,
915 thresh, EXTENT_DELALLOC, 1);
922 * helper function to walk through a file and find extents
923 * newer than a specific transid, and smaller than thresh.
925 * This is used by the defragging code to find new and small
928 static int find_new_extents(struct btrfs_root *root,
929 struct inode *inode, u64 newer_than,
930 u64 *off, u32 thresh)
932 struct btrfs_path *path;
933 struct btrfs_key min_key;
934 struct extent_buffer *leaf;
935 struct btrfs_file_extent_item *extent;
938 u64 ino = btrfs_ino(inode);
940 path = btrfs_alloc_path();
944 min_key.objectid = ino;
945 min_key.type = BTRFS_EXTENT_DATA_KEY;
946 min_key.offset = *off;
949 ret = btrfs_search_forward(root, &min_key, path, newer_than);
953 if (min_key.objectid != ino)
955 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
958 leaf = path->nodes[0];
959 extent = btrfs_item_ptr(leaf, path->slots[0],
960 struct btrfs_file_extent_item);
962 type = btrfs_file_extent_type(leaf, extent);
963 if (type == BTRFS_FILE_EXTENT_REG &&
964 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
965 check_defrag_in_cache(inode, min_key.offset, thresh)) {
966 *off = min_key.offset;
967 btrfs_free_path(path);
972 if (path->slots[0] < btrfs_header_nritems(leaf)) {
973 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
977 if (min_key.offset == (u64)-1)
981 btrfs_release_path(path);
984 btrfs_free_path(path);
988 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
990 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
991 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
992 struct extent_map *em;
996 * hopefully we have this extent in the tree already, try without
997 * the full extent lock
999 read_lock(&em_tree->lock);
1000 em = lookup_extent_mapping(em_tree, start, len);
1001 read_unlock(&em_tree->lock);
1004 struct extent_state *cached = NULL;
1005 u64 end = start + len - 1;
1007 /* get the big lock and read metadata off disk */
1008 lock_extent_bits(io_tree, start, end, &cached);
1009 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
1010 unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS);
1019 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1021 struct extent_map *next;
1024 /* this is the last extent */
1025 if (em->start + em->len >= i_size_read(inode))
1028 next = defrag_lookup_extent(inode, em->start + em->len);
1029 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1031 else if ((em->block_start + em->block_len == next->block_start) &&
1032 (em->block_len > SZ_128K && next->block_len > SZ_128K))
1035 free_extent_map(next);
1039 static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
1040 u64 *last_len, u64 *skip, u64 *defrag_end,
1043 struct extent_map *em;
1045 bool next_mergeable = true;
1046 bool prev_mergeable = true;
1049 * make sure that once we start defragging an extent, we keep on
1052 if (start < *defrag_end)
1057 em = defrag_lookup_extent(inode, start);
1061 /* this will cover holes, and inline extents */
1062 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1068 prev_mergeable = false;
1070 next_mergeable = defrag_check_next_extent(inode, em);
1072 * we hit a real extent, if it is big or the next extent is not a
1073 * real extent, don't bother defragging it
1075 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1076 (em->len >= thresh || (!next_mergeable && !prev_mergeable)))
1080 * last_len ends up being a counter of how many bytes we've defragged.
1081 * every time we choose not to defrag an extent, we reset *last_len
1082 * so that the next tiny extent will force a defrag.
1084 * The end result of this is that tiny extents before a single big
1085 * extent will force at least part of that big extent to be defragged.
1088 *defrag_end = extent_map_end(em);
1091 *skip = extent_map_end(em);
1095 free_extent_map(em);
1100 * it doesn't do much good to defrag one or two pages
1101 * at a time. This pulls in a nice chunk of pages
1102 * to COW and defrag.
1104 * It also makes sure the delalloc code has enough
1105 * dirty data to avoid making new small extents as part
1108 * It's a good idea to start RA on this range
1109 * before calling this.
1111 static int cluster_pages_for_defrag(struct inode *inode,
1112 struct page **pages,
1113 unsigned long start_index,
1114 unsigned long num_pages)
1116 unsigned long file_end;
1117 u64 isize = i_size_read(inode);
1124 struct btrfs_ordered_extent *ordered;
1125 struct extent_state *cached_state = NULL;
1126 struct extent_io_tree *tree;
1127 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1129 file_end = (isize - 1) >> PAGE_SHIFT;
1130 if (!isize || start_index > file_end)
1133 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1135 ret = btrfs_delalloc_reserve_space(inode,
1136 start_index << PAGE_SHIFT,
1137 page_cnt << PAGE_SHIFT);
1141 tree = &BTRFS_I(inode)->io_tree;
1143 /* step one, lock all the pages */
1144 for (i = 0; i < page_cnt; i++) {
1147 page = find_or_create_page(inode->i_mapping,
1148 start_index + i, mask);
1152 page_start = page_offset(page);
1153 page_end = page_start + PAGE_SIZE - 1;
1155 lock_extent_bits(tree, page_start, page_end,
1157 ordered = btrfs_lookup_ordered_extent(inode,
1159 unlock_extent_cached(tree, page_start, page_end,
1160 &cached_state, GFP_NOFS);
1165 btrfs_start_ordered_extent(inode, ordered, 1);
1166 btrfs_put_ordered_extent(ordered);
1169 * we unlocked the page above, so we need check if
1170 * it was released or not.
1172 if (page->mapping != inode->i_mapping) {
1179 if (!PageUptodate(page)) {
1180 btrfs_readpage(NULL, page);
1182 if (!PageUptodate(page)) {
1190 if (page->mapping != inode->i_mapping) {
1202 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1206 * so now we have a nice long stream of locked
1207 * and up to date pages, lets wait on them
1209 for (i = 0; i < i_done; i++)
1210 wait_on_page_writeback(pages[i]);
1212 page_start = page_offset(pages[0]);
1213 page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE;
1215 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1216 page_start, page_end - 1, &cached_state);
1217 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1218 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1219 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1220 &cached_state, GFP_NOFS);
1222 if (i_done != page_cnt) {
1223 spin_lock(&BTRFS_I(inode)->lock);
1224 BTRFS_I(inode)->outstanding_extents++;
1225 spin_unlock(&BTRFS_I(inode)->lock);
1226 btrfs_delalloc_release_space(inode,
1227 start_index << PAGE_SHIFT,
1228 (page_cnt - i_done) << PAGE_SHIFT);
1232 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1233 &cached_state, GFP_NOFS);
1235 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1236 page_start, page_end - 1, &cached_state,
1239 for (i = 0; i < i_done; i++) {
1240 clear_page_dirty_for_io(pages[i]);
1241 ClearPageChecked(pages[i]);
1242 set_page_extent_mapped(pages[i]);
1243 set_page_dirty(pages[i]);
1244 unlock_page(pages[i]);
1249 for (i = 0; i < i_done; i++) {
1250 unlock_page(pages[i]);
1253 btrfs_delalloc_release_space(inode,
1254 start_index << PAGE_SHIFT,
1255 page_cnt << PAGE_SHIFT);
1260 int btrfs_defrag_file(struct inode *inode, struct file *file,
1261 struct btrfs_ioctl_defrag_range_args *range,
1262 u64 newer_than, unsigned long max_to_defrag)
1264 struct btrfs_root *root = BTRFS_I(inode)->root;
1265 struct file_ra_state *ra = NULL;
1266 unsigned long last_index;
1267 u64 isize = i_size_read(inode);
1271 u64 newer_off = range->start;
1273 unsigned long ra_index = 0;
1275 int defrag_count = 0;
1276 int compress_type = BTRFS_COMPRESS_ZLIB;
1277 u32 extent_thresh = range->extent_thresh;
1278 unsigned long max_cluster = SZ_256K >> PAGE_SHIFT;
1279 unsigned long cluster = max_cluster;
1280 u64 new_align = ~((u64)SZ_128K - 1);
1281 struct page **pages = NULL;
1286 if (range->start >= isize)
1289 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1290 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1292 if (range->compress_type)
1293 compress_type = range->compress_type;
1296 if (extent_thresh == 0)
1297 extent_thresh = SZ_256K;
1300 * if we were not given a file, allocate a readahead
1304 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1307 file_ra_state_init(ra, inode->i_mapping);
1312 pages = kmalloc_array(max_cluster, sizeof(struct page *),
1319 /* find the last page to defrag */
1320 if (range->start + range->len > range->start) {
1321 last_index = min_t(u64, isize - 1,
1322 range->start + range->len - 1) >> PAGE_SHIFT;
1324 last_index = (isize - 1) >> PAGE_SHIFT;
1328 ret = find_new_extents(root, inode, newer_than,
1329 &newer_off, SZ_64K);
1331 range->start = newer_off;
1333 * we always align our defrag to help keep
1334 * the extents in the file evenly spaced
1336 i = (newer_off & new_align) >> PAGE_SHIFT;
1340 i = range->start >> PAGE_SHIFT;
1343 max_to_defrag = last_index - i + 1;
1346 * make writeback starts from i, so the defrag range can be
1347 * written sequentially.
1349 if (i < inode->i_mapping->writeback_index)
1350 inode->i_mapping->writeback_index = i;
1352 while (i <= last_index && defrag_count < max_to_defrag &&
1353 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) {
1355 * make sure we stop running if someone unmounts
1358 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1361 if (btrfs_defrag_cancelled(root->fs_info)) {
1362 btrfs_debug(root->fs_info, "defrag_file cancelled");
1367 if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT,
1368 extent_thresh, &last_len, &skip,
1369 &defrag_end, range->flags &
1370 BTRFS_DEFRAG_RANGE_COMPRESS)) {
1373 * the should_defrag function tells us how much to skip
1374 * bump our counter by the suggested amount
1376 next = DIV_ROUND_UP(skip, PAGE_SIZE);
1377 i = max(i + 1, next);
1382 cluster = (PAGE_ALIGN(defrag_end) >>
1384 cluster = min(cluster, max_cluster);
1386 cluster = max_cluster;
1389 if (i + cluster > ra_index) {
1390 ra_index = max(i, ra_index);
1391 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1393 ra_index += cluster;
1397 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1398 BTRFS_I(inode)->force_compress = compress_type;
1399 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1401 inode_unlock(inode);
1405 defrag_count += ret;
1406 balance_dirty_pages_ratelimited(inode->i_mapping);
1407 inode_unlock(inode);
1410 if (newer_off == (u64)-1)
1416 newer_off = max(newer_off + 1,
1417 (u64)i << PAGE_SHIFT);
1419 ret = find_new_extents(root, inode, newer_than,
1420 &newer_off, SZ_64K);
1422 range->start = newer_off;
1423 i = (newer_off & new_align) >> PAGE_SHIFT;
1430 last_len += ret << PAGE_SHIFT;
1438 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1439 filemap_flush(inode->i_mapping);
1440 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1441 &BTRFS_I(inode)->runtime_flags))
1442 filemap_flush(inode->i_mapping);
1445 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1446 /* the filemap_flush will queue IO into the worker threads, but
1447 * we have to make sure the IO is actually started and that
1448 * ordered extents get created before we return
1450 atomic_inc(&root->fs_info->async_submit_draining);
1451 while (atomic_read(&root->fs_info->nr_async_submits) ||
1452 atomic_read(&root->fs_info->async_delalloc_pages)) {
1453 wait_event(root->fs_info->async_submit_wait,
1454 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1455 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1457 atomic_dec(&root->fs_info->async_submit_draining);
1460 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1461 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1467 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1469 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1470 inode_unlock(inode);
1478 static noinline int btrfs_ioctl_resize(struct file *file,
1484 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
1485 struct btrfs_ioctl_vol_args *vol_args;
1486 struct btrfs_trans_handle *trans;
1487 struct btrfs_device *device = NULL;
1490 char *devstr = NULL;
1494 if (!capable(CAP_SYS_ADMIN))
1497 ret = mnt_want_write_file(file);
1501 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1503 mnt_drop_write_file(file);
1504 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1507 mutex_lock(&root->fs_info->volume_mutex);
1508 vol_args = memdup_user(arg, sizeof(*vol_args));
1509 if (IS_ERR(vol_args)) {
1510 ret = PTR_ERR(vol_args);
1514 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1516 sizestr = vol_args->name;
1517 devstr = strchr(sizestr, ':');
1519 sizestr = devstr + 1;
1521 devstr = vol_args->name;
1522 ret = kstrtoull(devstr, 10, &devid);
1529 btrfs_info(root->fs_info, "resizing devid %llu", devid);
1532 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
1534 btrfs_info(root->fs_info, "resizer unable to find device %llu",
1540 if (!device->writeable) {
1541 btrfs_info(root->fs_info,
1542 "resizer unable to apply on readonly device %llu",
1548 if (!strcmp(sizestr, "max"))
1549 new_size = device->bdev->bd_inode->i_size;
1551 if (sizestr[0] == '-') {
1554 } else if (sizestr[0] == '+') {
1558 new_size = memparse(sizestr, &retptr);
1559 if (*retptr != '\0' || new_size == 0) {
1565 if (device->is_tgtdev_for_dev_replace) {
1570 old_size = btrfs_device_get_total_bytes(device);
1573 if (new_size > old_size) {
1577 new_size = old_size - new_size;
1578 } else if (mod > 0) {
1579 if (new_size > ULLONG_MAX - old_size) {
1583 new_size = old_size + new_size;
1586 if (new_size < SZ_256M) {
1590 if (new_size > device->bdev->bd_inode->i_size) {
1595 new_size = div_u64(new_size, root->sectorsize);
1596 new_size *= root->sectorsize;
1598 btrfs_info_in_rcu(root->fs_info, "new size for %s is %llu",
1599 rcu_str_deref(device->name), new_size);
1601 if (new_size > old_size) {
1602 trans = btrfs_start_transaction(root, 0);
1603 if (IS_ERR(trans)) {
1604 ret = PTR_ERR(trans);
1607 ret = btrfs_grow_device(trans, device, new_size);
1608 btrfs_commit_transaction(trans, root);
1609 } else if (new_size < old_size) {
1610 ret = btrfs_shrink_device(device, new_size);
1611 } /* equal, nothing need to do */
1616 mutex_unlock(&root->fs_info->volume_mutex);
1617 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
1618 mnt_drop_write_file(file);
1622 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1623 char *name, unsigned long fd, int subvol,
1624 u64 *transid, bool readonly,
1625 struct btrfs_qgroup_inherit *inherit)
1630 ret = mnt_want_write_file(file);
1634 namelen = strlen(name);
1635 if (strchr(name, '/')) {
1637 goto out_drop_write;
1640 if (name[0] == '.' &&
1641 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1643 goto out_drop_write;
1647 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1648 NULL, transid, readonly, inherit);
1650 struct fd src = fdget(fd);
1651 struct inode *src_inode;
1654 goto out_drop_write;
1657 src_inode = file_inode(src.file);
1658 if (src_inode->i_sb != file_inode(file)->i_sb) {
1659 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
1660 "Snapshot src from another FS");
1662 } else if (!inode_owner_or_capable(src_inode)) {
1664 * Subvolume creation is not restricted, but snapshots
1665 * are limited to own subvolumes only
1669 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1670 BTRFS_I(src_inode)->root,
1671 transid, readonly, inherit);
1676 mnt_drop_write_file(file);
1681 static noinline int btrfs_ioctl_snap_create(struct file *file,
1682 void __user *arg, int subvol)
1684 struct btrfs_ioctl_vol_args *vol_args;
1687 vol_args = memdup_user(arg, sizeof(*vol_args));
1688 if (IS_ERR(vol_args))
1689 return PTR_ERR(vol_args);
1690 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1692 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1693 vol_args->fd, subvol,
1700 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1701 void __user *arg, int subvol)
1703 struct btrfs_ioctl_vol_args_v2 *vol_args;
1707 bool readonly = false;
1708 struct btrfs_qgroup_inherit *inherit = NULL;
1710 vol_args = memdup_user(arg, sizeof(*vol_args));
1711 if (IS_ERR(vol_args))
1712 return PTR_ERR(vol_args);
1713 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1715 if (vol_args->flags &
1716 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1717 BTRFS_SUBVOL_QGROUP_INHERIT)) {
1722 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1724 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1726 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1727 if (vol_args->size > PAGE_SIZE) {
1731 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1732 if (IS_ERR(inherit)) {
1733 ret = PTR_ERR(inherit);
1738 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1739 vol_args->fd, subvol, ptr,
1744 if (ptr && copy_to_user(arg +
1745 offsetof(struct btrfs_ioctl_vol_args_v2,
1757 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1760 struct inode *inode = file_inode(file);
1761 struct btrfs_root *root = BTRFS_I(inode)->root;
1765 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1768 down_read(&root->fs_info->subvol_sem);
1769 if (btrfs_root_readonly(root))
1770 flags |= BTRFS_SUBVOL_RDONLY;
1771 up_read(&root->fs_info->subvol_sem);
1773 if (copy_to_user(arg, &flags, sizeof(flags)))
1779 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1782 struct inode *inode = file_inode(file);
1783 struct btrfs_root *root = BTRFS_I(inode)->root;
1784 struct btrfs_trans_handle *trans;
1789 if (!inode_owner_or_capable(inode))
1792 ret = mnt_want_write_file(file);
1796 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1798 goto out_drop_write;
1801 if (copy_from_user(&flags, arg, sizeof(flags))) {
1803 goto out_drop_write;
1806 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1808 goto out_drop_write;
1811 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1813 goto out_drop_write;
1816 down_write(&root->fs_info->subvol_sem);
1819 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1822 root_flags = btrfs_root_flags(&root->root_item);
1823 if (flags & BTRFS_SUBVOL_RDONLY) {
1824 btrfs_set_root_flags(&root->root_item,
1825 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1828 * Block RO -> RW transition if this subvolume is involved in
1831 spin_lock(&root->root_item_lock);
1832 if (root->send_in_progress == 0) {
1833 btrfs_set_root_flags(&root->root_item,
1834 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1835 spin_unlock(&root->root_item_lock);
1837 spin_unlock(&root->root_item_lock);
1838 btrfs_warn(root->fs_info,
1839 "Attempt to set subvolume %llu read-write during send",
1840 root->root_key.objectid);
1846 trans = btrfs_start_transaction(root, 1);
1847 if (IS_ERR(trans)) {
1848 ret = PTR_ERR(trans);
1852 ret = btrfs_update_root(trans, root->fs_info->tree_root,
1853 &root->root_key, &root->root_item);
1855 btrfs_commit_transaction(trans, root);
1858 btrfs_set_root_flags(&root->root_item, root_flags);
1860 up_write(&root->fs_info->subvol_sem);
1862 mnt_drop_write_file(file);
1868 * helper to check if the subvolume references other subvolumes
1870 static noinline int may_destroy_subvol(struct btrfs_root *root)
1872 struct btrfs_path *path;
1873 struct btrfs_dir_item *di;
1874 struct btrfs_key key;
1878 path = btrfs_alloc_path();
1882 /* Make sure this root isn't set as the default subvol */
1883 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1884 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1885 dir_id, "default", 7, 0);
1886 if (di && !IS_ERR(di)) {
1887 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1888 if (key.objectid == root->root_key.objectid) {
1890 btrfs_err(root->fs_info, "deleting default subvolume "
1891 "%llu is not allowed", key.objectid);
1894 btrfs_release_path(path);
1897 key.objectid = root->root_key.objectid;
1898 key.type = BTRFS_ROOT_REF_KEY;
1899 key.offset = (u64)-1;
1901 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1908 if (path->slots[0] > 0) {
1910 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1911 if (key.objectid == root->root_key.objectid &&
1912 key.type == BTRFS_ROOT_REF_KEY)
1916 btrfs_free_path(path);
1920 static noinline int key_in_sk(struct btrfs_key *key,
1921 struct btrfs_ioctl_search_key *sk)
1923 struct btrfs_key test;
1926 test.objectid = sk->min_objectid;
1927 test.type = sk->min_type;
1928 test.offset = sk->min_offset;
1930 ret = btrfs_comp_cpu_keys(key, &test);
1934 test.objectid = sk->max_objectid;
1935 test.type = sk->max_type;
1936 test.offset = sk->max_offset;
1938 ret = btrfs_comp_cpu_keys(key, &test);
1944 static noinline int copy_to_sk(struct btrfs_root *root,
1945 struct btrfs_path *path,
1946 struct btrfs_key *key,
1947 struct btrfs_ioctl_search_key *sk,
1950 unsigned long *sk_offset,
1954 struct extent_buffer *leaf;
1955 struct btrfs_ioctl_search_header sh;
1956 struct btrfs_key test;
1957 unsigned long item_off;
1958 unsigned long item_len;
1964 leaf = path->nodes[0];
1965 slot = path->slots[0];
1966 nritems = btrfs_header_nritems(leaf);
1968 if (btrfs_header_generation(leaf) > sk->max_transid) {
1972 found_transid = btrfs_header_generation(leaf);
1974 for (i = slot; i < nritems; i++) {
1975 item_off = btrfs_item_ptr_offset(leaf, i);
1976 item_len = btrfs_item_size_nr(leaf, i);
1978 btrfs_item_key_to_cpu(leaf, key, i);
1979 if (!key_in_sk(key, sk))
1982 if (sizeof(sh) + item_len > *buf_size) {
1989 * return one empty item back for v1, which does not
1993 *buf_size = sizeof(sh) + item_len;
1998 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
2003 sh.objectid = key->objectid;
2004 sh.offset = key->offset;
2005 sh.type = key->type;
2007 sh.transid = found_transid;
2009 /* copy search result header */
2010 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
2015 *sk_offset += sizeof(sh);
2018 char __user *up = ubuf + *sk_offset;
2020 if (read_extent_buffer_to_user(leaf, up,
2021 item_off, item_len)) {
2026 *sk_offset += item_len;
2030 if (ret) /* -EOVERFLOW from above */
2033 if (*num_found >= sk->nr_items) {
2040 test.objectid = sk->max_objectid;
2041 test.type = sk->max_type;
2042 test.offset = sk->max_offset;
2043 if (btrfs_comp_cpu_keys(key, &test) >= 0)
2045 else if (key->offset < (u64)-1)
2047 else if (key->type < (u8)-1) {
2050 } else if (key->objectid < (u64)-1) {
2058 * 0: all items from this leaf copied, continue with next
2059 * 1: * more items can be copied, but unused buffer is too small
2060 * * all items were found
2061 * Either way, it will stops the loop which iterates to the next
2063 * -EOVERFLOW: item was to large for buffer
2064 * -EFAULT: could not copy extent buffer back to userspace
2069 static noinline int search_ioctl(struct inode *inode,
2070 struct btrfs_ioctl_search_key *sk,
2074 struct btrfs_root *root;
2075 struct btrfs_key key;
2076 struct btrfs_path *path;
2077 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
2080 unsigned long sk_offset = 0;
2082 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2083 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2087 path = btrfs_alloc_path();
2091 if (sk->tree_id == 0) {
2092 /* search the root of the inode that was passed */
2093 root = BTRFS_I(inode)->root;
2095 key.objectid = sk->tree_id;
2096 key.type = BTRFS_ROOT_ITEM_KEY;
2097 key.offset = (u64)-1;
2098 root = btrfs_read_fs_root_no_name(info, &key);
2100 btrfs_free_path(path);
2105 key.objectid = sk->min_objectid;
2106 key.type = sk->min_type;
2107 key.offset = sk->min_offset;
2110 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2116 ret = copy_to_sk(root, path, &key, sk, buf_size, ubuf,
2117 &sk_offset, &num_found);
2118 btrfs_release_path(path);
2126 sk->nr_items = num_found;
2127 btrfs_free_path(path);
2131 static noinline int btrfs_ioctl_tree_search(struct file *file,
2134 struct btrfs_ioctl_search_args __user *uargs;
2135 struct btrfs_ioctl_search_key sk;
2136 struct inode *inode;
2140 if (!capable(CAP_SYS_ADMIN))
2143 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2145 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2148 buf_size = sizeof(uargs->buf);
2150 inode = file_inode(file);
2151 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2154 * In the origin implementation an overflow is handled by returning a
2155 * search header with a len of zero, so reset ret.
2157 if (ret == -EOVERFLOW)
2160 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2165 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2168 struct btrfs_ioctl_search_args_v2 __user *uarg;
2169 struct btrfs_ioctl_search_args_v2 args;
2170 struct inode *inode;
2173 const size_t buf_limit = SZ_16M;
2175 if (!capable(CAP_SYS_ADMIN))
2178 /* copy search header and buffer size */
2179 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2180 if (copy_from_user(&args, uarg, sizeof(args)))
2183 buf_size = args.buf_size;
2185 if (buf_size < sizeof(struct btrfs_ioctl_search_header))
2188 /* limit result size to 16MB */
2189 if (buf_size > buf_limit)
2190 buf_size = buf_limit;
2192 inode = file_inode(file);
2193 ret = search_ioctl(inode, &args.key, &buf_size,
2194 (char *)(&uarg->buf[0]));
2195 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2197 else if (ret == -EOVERFLOW &&
2198 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2205 * Search INODE_REFs to identify path name of 'dirid' directory
2206 * in a 'tree_id' tree. and sets path name to 'name'.
2208 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2209 u64 tree_id, u64 dirid, char *name)
2211 struct btrfs_root *root;
2212 struct btrfs_key key;
2218 struct btrfs_inode_ref *iref;
2219 struct extent_buffer *l;
2220 struct btrfs_path *path;
2222 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2227 path = btrfs_alloc_path();
2231 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
2233 key.objectid = tree_id;
2234 key.type = BTRFS_ROOT_ITEM_KEY;
2235 key.offset = (u64)-1;
2236 root = btrfs_read_fs_root_no_name(info, &key);
2238 btrfs_err(info, "could not find root %llu", tree_id);
2243 key.objectid = dirid;
2244 key.type = BTRFS_INODE_REF_KEY;
2245 key.offset = (u64)-1;
2248 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2252 ret = btrfs_previous_item(root, path, dirid,
2253 BTRFS_INODE_REF_KEY);
2263 slot = path->slots[0];
2264 btrfs_item_key_to_cpu(l, &key, slot);
2266 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2267 len = btrfs_inode_ref_name_len(l, iref);
2269 total_len += len + 1;
2271 ret = -ENAMETOOLONG;
2276 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2278 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2281 btrfs_release_path(path);
2282 key.objectid = key.offset;
2283 key.offset = (u64)-1;
2284 dirid = key.objectid;
2286 memmove(name, ptr, total_len);
2287 name[total_len] = '\0';
2290 btrfs_free_path(path);
2294 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2297 struct btrfs_ioctl_ino_lookup_args *args;
2298 struct inode *inode;
2301 args = memdup_user(argp, sizeof(*args));
2303 return PTR_ERR(args);
2305 inode = file_inode(file);
2308 * Unprivileged query to obtain the containing subvolume root id. The
2309 * path is reset so it's consistent with btrfs_search_path_in_tree.
2311 if (args->treeid == 0)
2312 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2314 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2319 if (!capable(CAP_SYS_ADMIN)) {
2324 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2325 args->treeid, args->objectid,
2329 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2336 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2339 struct dentry *parent = file->f_path.dentry;
2340 struct dentry *dentry;
2341 struct inode *dir = d_inode(parent);
2342 struct inode *inode;
2343 struct btrfs_root *root = BTRFS_I(dir)->root;
2344 struct btrfs_root *dest = NULL;
2345 struct btrfs_ioctl_vol_args *vol_args;
2346 struct btrfs_trans_handle *trans;
2347 struct btrfs_block_rsv block_rsv;
2349 u64 qgroup_reserved;
2354 vol_args = memdup_user(arg, sizeof(*vol_args));
2355 if (IS_ERR(vol_args))
2356 return PTR_ERR(vol_args);
2358 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2359 namelen = strlen(vol_args->name);
2360 if (strchr(vol_args->name, '/') ||
2361 strncmp(vol_args->name, "..", namelen) == 0) {
2366 err = mnt_want_write_file(file);
2371 inode_lock_nested(dir, I_MUTEX_PARENT);
2372 // XXX: should've been
2373 // err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2374 // if (err == -EINTR)
2375 // goto out_drop_write;
2376 dentry = lookup_one_len(vol_args->name, parent, namelen);
2377 if (IS_ERR(dentry)) {
2378 err = PTR_ERR(dentry);
2379 goto out_unlock_dir;
2382 if (d_really_is_negative(dentry)) {
2387 inode = d_inode(dentry);
2388 dest = BTRFS_I(inode)->root;
2389 if (!capable(CAP_SYS_ADMIN)) {
2391 * Regular user. Only allow this with a special mount
2392 * option, when the user has write+exec access to the
2393 * subvol root, and when rmdir(2) would have been
2396 * Note that this is _not_ check that the subvol is
2397 * empty or doesn't contain data that we wouldn't
2398 * otherwise be able to delete.
2400 * Users who want to delete empty subvols should try
2404 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2408 * Do not allow deletion if the parent dir is the same
2409 * as the dir to be deleted. That means the ioctl
2410 * must be called on the dentry referencing the root
2411 * of the subvol, not a random directory contained
2418 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2423 /* check if subvolume may be deleted by a user */
2424 err = btrfs_may_delete(dir, dentry, 1);
2428 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2436 * Don't allow to delete a subvolume with send in progress. This is
2437 * inside the i_mutex so the error handling that has to drop the bit
2438 * again is not run concurrently.
2440 spin_lock(&dest->root_item_lock);
2441 root_flags = btrfs_root_flags(&dest->root_item);
2442 if (dest->send_in_progress == 0) {
2443 btrfs_set_root_flags(&dest->root_item,
2444 root_flags | BTRFS_ROOT_SUBVOL_DEAD);
2445 spin_unlock(&dest->root_item_lock);
2447 spin_unlock(&dest->root_item_lock);
2448 btrfs_warn(root->fs_info,
2449 "Attempt to delete subvolume %llu during send",
2450 dest->root_key.objectid);
2452 goto out_unlock_inode;
2455 down_write(&root->fs_info->subvol_sem);
2457 err = may_destroy_subvol(dest);
2461 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2463 * One for dir inode, two for dir entries, two for root
2466 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2467 5, &qgroup_reserved, true);
2471 trans = btrfs_start_transaction(root, 0);
2472 if (IS_ERR(trans)) {
2473 err = PTR_ERR(trans);
2476 trans->block_rsv = &block_rsv;
2477 trans->bytes_reserved = block_rsv.size;
2479 btrfs_record_snapshot_destroy(trans, dir);
2481 ret = btrfs_unlink_subvol(trans, root, dir,
2482 dest->root_key.objectid,
2483 dentry->d_name.name,
2484 dentry->d_name.len);
2487 btrfs_abort_transaction(trans, root, ret);
2491 btrfs_record_root_in_trans(trans, dest);
2493 memset(&dest->root_item.drop_progress, 0,
2494 sizeof(dest->root_item.drop_progress));
2495 dest->root_item.drop_level = 0;
2496 btrfs_set_root_refs(&dest->root_item, 0);
2498 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
2499 ret = btrfs_insert_orphan_item(trans,
2500 root->fs_info->tree_root,
2501 dest->root_key.objectid);
2503 btrfs_abort_transaction(trans, root, ret);
2509 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2510 dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
2511 dest->root_key.objectid);
2512 if (ret && ret != -ENOENT) {
2513 btrfs_abort_transaction(trans, root, ret);
2517 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2518 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2519 dest->root_item.received_uuid,
2520 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2521 dest->root_key.objectid);
2522 if (ret && ret != -ENOENT) {
2523 btrfs_abort_transaction(trans, root, ret);
2530 trans->block_rsv = NULL;
2531 trans->bytes_reserved = 0;
2532 ret = btrfs_end_transaction(trans, root);
2535 inode->i_flags |= S_DEAD;
2537 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
2539 up_write(&root->fs_info->subvol_sem);
2541 spin_lock(&dest->root_item_lock);
2542 root_flags = btrfs_root_flags(&dest->root_item);
2543 btrfs_set_root_flags(&dest->root_item,
2544 root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
2545 spin_unlock(&dest->root_item_lock);
2548 inode_unlock(inode);
2550 d_invalidate(dentry);
2551 btrfs_invalidate_inodes(dest);
2553 ASSERT(dest->send_in_progress == 0);
2556 if (dest->ino_cache_inode) {
2557 iput(dest->ino_cache_inode);
2558 dest->ino_cache_inode = NULL;
2566 mnt_drop_write_file(file);
2572 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2574 struct inode *inode = file_inode(file);
2575 struct btrfs_root *root = BTRFS_I(inode)->root;
2576 struct btrfs_ioctl_defrag_range_args *range;
2579 ret = mnt_want_write_file(file);
2583 if (btrfs_root_readonly(root)) {
2588 switch (inode->i_mode & S_IFMT) {
2590 if (!capable(CAP_SYS_ADMIN)) {
2594 ret = btrfs_defrag_root(root);
2597 ret = btrfs_defrag_root(root->fs_info->extent_root);
2600 if (!(file->f_mode & FMODE_WRITE)) {
2605 range = kzalloc(sizeof(*range), GFP_KERNEL);
2612 if (copy_from_user(range, argp,
2618 /* compression requires us to start the IO */
2619 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2620 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2621 range->extent_thresh = (u32)-1;
2624 /* the rest are all set to zero by kzalloc */
2625 range->len = (u64)-1;
2627 ret = btrfs_defrag_file(file_inode(file), file,
2637 mnt_drop_write_file(file);
2641 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2643 struct btrfs_ioctl_vol_args *vol_args;
2646 if (!capable(CAP_SYS_ADMIN))
2649 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2651 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2654 mutex_lock(&root->fs_info->volume_mutex);
2655 vol_args = memdup_user(arg, sizeof(*vol_args));
2656 if (IS_ERR(vol_args)) {
2657 ret = PTR_ERR(vol_args);
2661 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2662 ret = btrfs_init_new_device(root, vol_args->name);
2665 btrfs_info(root->fs_info, "disk added %s",vol_args->name);
2669 mutex_unlock(&root->fs_info->volume_mutex);
2670 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2674 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2676 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2677 struct btrfs_ioctl_vol_args *vol_args;
2680 if (!capable(CAP_SYS_ADMIN))
2683 ret = mnt_want_write_file(file);
2687 vol_args = memdup_user(arg, sizeof(*vol_args));
2688 if (IS_ERR(vol_args)) {
2689 ret = PTR_ERR(vol_args);
2693 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2695 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2697 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2701 mutex_lock(&root->fs_info->volume_mutex);
2702 ret = btrfs_rm_device(root, vol_args->name);
2703 mutex_unlock(&root->fs_info->volume_mutex);
2704 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2707 btrfs_info(root->fs_info, "disk deleted %s",vol_args->name);
2712 mnt_drop_write_file(file);
2716 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2718 struct btrfs_ioctl_fs_info_args *fi_args;
2719 struct btrfs_device *device;
2720 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2723 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2727 mutex_lock(&fs_devices->device_list_mutex);
2728 fi_args->num_devices = fs_devices->num_devices;
2729 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2731 list_for_each_entry(device, &fs_devices->devices, dev_list) {
2732 if (device->devid > fi_args->max_id)
2733 fi_args->max_id = device->devid;
2735 mutex_unlock(&fs_devices->device_list_mutex);
2737 fi_args->nodesize = root->fs_info->super_copy->nodesize;
2738 fi_args->sectorsize = root->fs_info->super_copy->sectorsize;
2739 fi_args->clone_alignment = root->fs_info->super_copy->sectorsize;
2741 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2748 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2750 struct btrfs_ioctl_dev_info_args *di_args;
2751 struct btrfs_device *dev;
2752 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2754 char *s_uuid = NULL;
2756 di_args = memdup_user(arg, sizeof(*di_args));
2757 if (IS_ERR(di_args))
2758 return PTR_ERR(di_args);
2760 if (!btrfs_is_empty_uuid(di_args->uuid))
2761 s_uuid = di_args->uuid;
2763 mutex_lock(&fs_devices->device_list_mutex);
2764 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
2771 di_args->devid = dev->devid;
2772 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2773 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
2774 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2776 struct rcu_string *name;
2779 name = rcu_dereference(dev->name);
2780 strncpy(di_args->path, name->str, sizeof(di_args->path));
2782 di_args->path[sizeof(di_args->path) - 1] = 0;
2784 di_args->path[0] = '\0';
2788 mutex_unlock(&fs_devices->device_list_mutex);
2789 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2796 static struct page *extent_same_get_page(struct inode *inode, pgoff_t index)
2800 page = grab_cache_page(inode->i_mapping, index);
2802 return ERR_PTR(-ENOMEM);
2804 if (!PageUptodate(page)) {
2807 ret = btrfs_readpage(NULL, page);
2809 return ERR_PTR(ret);
2811 if (!PageUptodate(page)) {
2814 return ERR_PTR(-EIO);
2816 if (page->mapping != inode->i_mapping) {
2819 return ERR_PTR(-EAGAIN);
2826 static int gather_extent_pages(struct inode *inode, struct page **pages,
2827 int num_pages, u64 off)
2830 pgoff_t index = off >> PAGE_SHIFT;
2832 for (i = 0; i < num_pages; i++) {
2834 pages[i] = extent_same_get_page(inode, index + i);
2835 if (IS_ERR(pages[i])) {
2836 int err = PTR_ERR(pages[i]);
2847 static int lock_extent_range(struct inode *inode, u64 off, u64 len,
2848 bool retry_range_locking)
2851 * Do any pending delalloc/csum calculations on inode, one way or
2852 * another, and lock file content.
2853 * The locking order is:
2856 * 2) range in the inode's io tree
2859 struct btrfs_ordered_extent *ordered;
2860 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2861 ordered = btrfs_lookup_first_ordered_extent(inode,
2864 ordered->file_offset + ordered->len <= off ||
2865 ordered->file_offset >= off + len) &&
2866 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2867 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2869 btrfs_put_ordered_extent(ordered);
2872 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2874 btrfs_put_ordered_extent(ordered);
2875 if (!retry_range_locking)
2877 btrfs_wait_ordered_range(inode, off, len);
2882 static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2)
2884 inode_unlock(inode1);
2885 inode_unlock(inode2);
2888 static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2)
2890 if (inode1 < inode2)
2891 swap(inode1, inode2);
2893 inode_lock_nested(inode1, I_MUTEX_PARENT);
2894 inode_lock_nested(inode2, I_MUTEX_CHILD);
2897 static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1,
2898 struct inode *inode2, u64 loff2, u64 len)
2900 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2901 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2904 static int btrfs_double_extent_lock(struct inode *inode1, u64 loff1,
2905 struct inode *inode2, u64 loff2, u64 len,
2906 bool retry_range_locking)
2910 if (inode1 < inode2) {
2911 swap(inode1, inode2);
2914 ret = lock_extent_range(inode1, loff1, len, retry_range_locking);
2917 ret = lock_extent_range(inode2, loff2, len, retry_range_locking);
2919 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1,
2926 struct page **src_pages;
2927 struct page **dst_pages;
2930 static void btrfs_cmp_data_free(struct cmp_pages *cmp)
2935 for (i = 0; i < cmp->num_pages; i++) {
2936 pg = cmp->src_pages[i];
2941 pg = cmp->dst_pages[i];
2947 kfree(cmp->src_pages);
2948 kfree(cmp->dst_pages);
2951 static int btrfs_cmp_data_prepare(struct inode *src, u64 loff,
2952 struct inode *dst, u64 dst_loff,
2953 u64 len, struct cmp_pages *cmp)
2956 int num_pages = PAGE_ALIGN(len) >> PAGE_SHIFT;
2957 struct page **src_pgarr, **dst_pgarr;
2960 * We must gather up all the pages before we initiate our
2961 * extent locking. We use an array for the page pointers. Size
2962 * of the array is bounded by len, which is in turn bounded by
2963 * BTRFS_MAX_DEDUPE_LEN.
2965 src_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL);
2966 dst_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL);
2967 if (!src_pgarr || !dst_pgarr) {
2972 cmp->num_pages = num_pages;
2973 cmp->src_pages = src_pgarr;
2974 cmp->dst_pages = dst_pgarr;
2976 ret = gather_extent_pages(src, cmp->src_pages, cmp->num_pages, loff);
2980 ret = gather_extent_pages(dst, cmp->dst_pages, cmp->num_pages, dst_loff);
2984 btrfs_cmp_data_free(cmp);
2988 static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
2989 u64 dst_loff, u64 len, struct cmp_pages *cmp)
2993 struct page *src_page, *dst_page;
2994 unsigned int cmp_len = PAGE_SIZE;
2995 void *addr, *dst_addr;
2999 if (len < PAGE_SIZE)
3002 BUG_ON(i >= cmp->num_pages);
3004 src_page = cmp->src_pages[i];
3005 dst_page = cmp->dst_pages[i];
3006 ASSERT(PageLocked(src_page));
3007 ASSERT(PageLocked(dst_page));
3009 addr = kmap_atomic(src_page);
3010 dst_addr = kmap_atomic(dst_page);
3012 flush_dcache_page(src_page);
3013 flush_dcache_page(dst_page);
3015 if (memcmp(addr, dst_addr, cmp_len))
3018 kunmap_atomic(addr);
3019 kunmap_atomic(dst_addr);
3031 static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen,
3035 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
3037 if (off + olen > inode->i_size || off + olen < off)
3040 /* if we extend to eof, continue to block boundary */
3041 if (off + len == inode->i_size)
3042 *plen = len = ALIGN(inode->i_size, bs) - off;
3044 /* Check that we are block aligned - btrfs_clone() requires this */
3045 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
3051 static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen,
3052 struct inode *dst, u64 dst_loff)
3056 struct cmp_pages cmp;
3058 u64 same_lock_start = 0;
3059 u64 same_lock_len = 0;
3070 ret = extent_same_check_offsets(src, loff, &len, olen);
3073 ret = extent_same_check_offsets(src, dst_loff, &len, olen);
3078 * Single inode case wants the same checks, except we
3079 * don't want our length pushed out past i_size as
3080 * comparing that data range makes no sense.
3082 * extent_same_check_offsets() will do this for an
3083 * unaligned length at i_size, so catch it here and
3084 * reject the request.
3086 * This effectively means we require aligned extents
3087 * for the single-inode case, whereas the other cases
3088 * allow an unaligned length so long as it ends at
3096 /* Check for overlapping ranges */
3097 if (dst_loff + len > loff && dst_loff < loff + len) {
3102 same_lock_start = min_t(u64, loff, dst_loff);
3103 same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start;
3105 btrfs_double_inode_lock(src, dst);
3107 ret = extent_same_check_offsets(src, loff, &len, olen);
3111 ret = extent_same_check_offsets(dst, dst_loff, &len, olen);
3116 /* don't make the dst file partly checksummed */
3117 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3118 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
3124 ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, &cmp);
3129 ret = lock_extent_range(src, same_lock_start, same_lock_len,
3132 ret = btrfs_double_extent_lock(src, loff, dst, dst_loff, len,
3135 * If one of the inodes has dirty pages in the respective range or
3136 * ordered extents, we need to flush dellaloc and wait for all ordered
3137 * extents in the range. We must unlock the pages and the ranges in the
3138 * io trees to avoid deadlocks when flushing delalloc (requires locking
3139 * pages) and when waiting for ordered extents to complete (they require
3142 if (ret == -EAGAIN) {
3144 * Ranges in the io trees already unlocked. Now unlock all
3145 * pages before waiting for all IO to complete.
3147 btrfs_cmp_data_free(&cmp);
3149 btrfs_wait_ordered_range(src, same_lock_start,
3152 btrfs_wait_ordered_range(src, loff, len);
3153 btrfs_wait_ordered_range(dst, dst_loff, len);
3159 /* ranges in the io trees already unlocked */
3160 btrfs_cmp_data_free(&cmp);
3164 /* pass original length for comparison so we stay within i_size */
3165 ret = btrfs_cmp_data(src, loff, dst, dst_loff, olen, &cmp);
3167 ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1);
3170 unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start,
3171 same_lock_start + same_lock_len - 1);
3173 btrfs_double_extent_unlock(src, loff, dst, dst_loff, len);
3175 btrfs_cmp_data_free(&cmp);
3180 btrfs_double_inode_unlock(src, dst);
3185 #define BTRFS_MAX_DEDUPE_LEN SZ_16M
3187 ssize_t btrfs_dedupe_file_range(struct file *src_file, u64 loff, u64 olen,
3188 struct file *dst_file, u64 dst_loff)
3190 struct inode *src = file_inode(src_file);
3191 struct inode *dst = file_inode(dst_file);
3192 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
3195 if (olen > BTRFS_MAX_DEDUPE_LEN)
3196 olen = BTRFS_MAX_DEDUPE_LEN;
3198 if (WARN_ON_ONCE(bs < PAGE_SIZE)) {
3200 * Btrfs does not support blocksize < page_size. As a
3201 * result, btrfs_cmp_data() won't correctly handle
3202 * this situation without an update.
3207 res = btrfs_extent_same(src, loff, olen, dst, dst_loff);
3213 static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3214 struct inode *inode,
3220 struct btrfs_root *root = BTRFS_I(inode)->root;
3223 inode_inc_iversion(inode);
3224 if (!no_time_update)
3225 inode->i_mtime = inode->i_ctime = current_fs_time(inode->i_sb);
3227 * We round up to the block size at eof when determining which
3228 * extents to clone above, but shouldn't round up the file size.
3230 if (endoff > destoff + olen)
3231 endoff = destoff + olen;
3232 if (endoff > inode->i_size)
3233 btrfs_i_size_write(inode, endoff);
3235 ret = btrfs_update_inode(trans, root, inode);
3237 btrfs_abort_transaction(trans, root, ret);
3238 btrfs_end_transaction(trans, root);
3241 ret = btrfs_end_transaction(trans, root);
3246 static void clone_update_extent_map(struct inode *inode,
3247 const struct btrfs_trans_handle *trans,
3248 const struct btrfs_path *path,
3249 const u64 hole_offset,
3252 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
3253 struct extent_map *em;
3256 em = alloc_extent_map();
3258 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3259 &BTRFS_I(inode)->runtime_flags);
3264 struct btrfs_file_extent_item *fi;
3266 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3267 struct btrfs_file_extent_item);
3268 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3269 em->generation = -1;
3270 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3271 BTRFS_FILE_EXTENT_INLINE)
3272 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3273 &BTRFS_I(inode)->runtime_flags);
3275 em->start = hole_offset;
3277 em->ram_bytes = em->len;
3278 em->orig_start = hole_offset;
3279 em->block_start = EXTENT_MAP_HOLE;
3281 em->orig_block_len = 0;
3282 em->compress_type = BTRFS_COMPRESS_NONE;
3283 em->generation = trans->transid;
3287 write_lock(&em_tree->lock);
3288 ret = add_extent_mapping(em_tree, em, 1);
3289 write_unlock(&em_tree->lock);
3290 if (ret != -EEXIST) {
3291 free_extent_map(em);
3294 btrfs_drop_extent_cache(inode, em->start,
3295 em->start + em->len - 1, 0);
3299 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3300 &BTRFS_I(inode)->runtime_flags);
3304 * Make sure we do not end up inserting an inline extent into a file that has
3305 * already other (non-inline) extents. If a file has an inline extent it can
3306 * not have any other extents and the (single) inline extent must start at the
3307 * file offset 0. Failing to respect these rules will lead to file corruption,
3308 * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc
3310 * We can have extents that have been already written to disk or we can have
3311 * dirty ranges still in delalloc, in which case the extent maps and items are
3312 * created only when we run delalloc, and the delalloc ranges might fall outside
3313 * the range we are currently locking in the inode's io tree. So we check the
3314 * inode's i_size because of that (i_size updates are done while holding the
3315 * i_mutex, which we are holding here).
3316 * We also check to see if the inode has a size not greater than "datal" but has
3317 * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are
3318 * protected against such concurrent fallocate calls by the i_mutex).
3320 * If the file has no extents but a size greater than datal, do not allow the
3321 * copy because we would need turn the inline extent into a non-inline one (even
3322 * with NO_HOLES enabled). If we find our destination inode only has one inline
3323 * extent, just overwrite it with the source inline extent if its size is less
3324 * than the source extent's size, or we could copy the source inline extent's
3325 * data into the destination inode's inline extent if the later is greater then
3328 static int clone_copy_inline_extent(struct inode *src,
3330 struct btrfs_trans_handle *trans,
3331 struct btrfs_path *path,
3332 struct btrfs_key *new_key,
3333 const u64 drop_start,
3339 struct btrfs_root *root = BTRFS_I(dst)->root;
3340 const u64 aligned_end = ALIGN(new_key->offset + datal,
3343 struct btrfs_key key;
3345 if (new_key->offset > 0)
3348 key.objectid = btrfs_ino(dst);
3349 key.type = BTRFS_EXTENT_DATA_KEY;
3351 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3354 } else if (ret > 0) {
3355 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
3356 ret = btrfs_next_leaf(root, path);
3360 goto copy_inline_extent;
3362 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
3363 if (key.objectid == btrfs_ino(dst) &&
3364 key.type == BTRFS_EXTENT_DATA_KEY) {
3365 ASSERT(key.offset > 0);
3368 } else if (i_size_read(dst) <= datal) {
3369 struct btrfs_file_extent_item *ei;
3373 * If the file size is <= datal, make sure there are no other
3374 * extents following (can happen do to an fallocate call with
3375 * the flag FALLOC_FL_KEEP_SIZE).
3377 ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
3378 struct btrfs_file_extent_item);
3380 * If it's an inline extent, it can not have other extents
3383 if (btrfs_file_extent_type(path->nodes[0], ei) ==
3384 BTRFS_FILE_EXTENT_INLINE)
3385 goto copy_inline_extent;
3387 ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei);
3388 if (ext_len > aligned_end)
3391 ret = btrfs_next_item(root, path);
3394 } else if (ret == 0) {
3395 btrfs_item_key_to_cpu(path->nodes[0], &key,
3397 if (key.objectid == btrfs_ino(dst) &&
3398 key.type == BTRFS_EXTENT_DATA_KEY)
3405 * We have no extent items, or we have an extent at offset 0 which may
3406 * or may not be inlined. All these cases are dealt the same way.
3408 if (i_size_read(dst) > datal) {
3410 * If the destination inode has an inline extent...
3411 * This would require copying the data from the source inline
3412 * extent into the beginning of the destination's inline extent.
3413 * But this is really complex, both extents can be compressed
3414 * or just one of them, which would require decompressing and
3415 * re-compressing data (which could increase the new compressed
3416 * size, not allowing the compressed data to fit anymore in an
3418 * So just don't support this case for now (it should be rare,
3419 * we are not really saving space when cloning inline extents).
3424 btrfs_release_path(path);
3425 ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1);
3428 ret = btrfs_insert_empty_item(trans, root, path, new_key, size);
3433 const u32 start = btrfs_file_extent_calc_inline_size(0);
3435 memmove(inline_data + start, inline_data + start + skip, datal);
3438 write_extent_buffer(path->nodes[0], inline_data,
3439 btrfs_item_ptr_offset(path->nodes[0],
3442 inode_add_bytes(dst, datal);
3448 * btrfs_clone() - clone a range from inode file to another
3450 * @src: Inode to clone from
3451 * @inode: Inode to clone to
3452 * @off: Offset within source to start clone from
3453 * @olen: Original length, passed by user, of range to clone
3454 * @olen_aligned: Block-aligned value of olen
3455 * @destoff: Offset within @inode to start clone
3456 * @no_time_update: Whether to update mtime/ctime on the target inode
3458 static int btrfs_clone(struct inode *src, struct inode *inode,
3459 const u64 off, const u64 olen, const u64 olen_aligned,
3460 const u64 destoff, int no_time_update)
3462 struct btrfs_root *root = BTRFS_I(inode)->root;
3463 struct btrfs_path *path = NULL;
3464 struct extent_buffer *leaf;
3465 struct btrfs_trans_handle *trans;
3467 struct btrfs_key key;
3471 const u64 len = olen_aligned;
3472 u64 last_dest_end = destoff;
3475 buf = vmalloc(root->nodesize);
3479 path = btrfs_alloc_path();
3485 path->reada = READA_FORWARD;
3487 key.objectid = btrfs_ino(src);
3488 key.type = BTRFS_EXTENT_DATA_KEY;
3492 u64 next_key_min_offset = key.offset + 1;
3495 * note the key will change type as we walk through the
3498 path->leave_spinning = 1;
3499 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3504 * First search, if no extent item that starts at offset off was
3505 * found but the previous item is an extent item, it's possible
3506 * it might overlap our target range, therefore process it.
3508 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3509 btrfs_item_key_to_cpu(path->nodes[0], &key,
3510 path->slots[0] - 1);
3511 if (key.type == BTRFS_EXTENT_DATA_KEY)
3515 nritems = btrfs_header_nritems(path->nodes[0]);
3517 if (path->slots[0] >= nritems) {
3518 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
3523 nritems = btrfs_header_nritems(path->nodes[0]);
3525 leaf = path->nodes[0];
3526 slot = path->slots[0];
3528 btrfs_item_key_to_cpu(leaf, &key, slot);
3529 if (key.type > BTRFS_EXTENT_DATA_KEY ||
3530 key.objectid != btrfs_ino(src))
3533 if (key.type == BTRFS_EXTENT_DATA_KEY) {
3534 struct btrfs_file_extent_item *extent;
3537 struct btrfs_key new_key;
3538 u64 disko = 0, diskl = 0;
3539 u64 datao = 0, datal = 0;
3543 extent = btrfs_item_ptr(leaf, slot,
3544 struct btrfs_file_extent_item);
3545 comp = btrfs_file_extent_compression(leaf, extent);
3546 type = btrfs_file_extent_type(leaf, extent);
3547 if (type == BTRFS_FILE_EXTENT_REG ||
3548 type == BTRFS_FILE_EXTENT_PREALLOC) {
3549 disko = btrfs_file_extent_disk_bytenr(leaf,
3551 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3553 datao = btrfs_file_extent_offset(leaf, extent);
3554 datal = btrfs_file_extent_num_bytes(leaf,
3556 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3557 /* take upper bound, may be compressed */
3558 datal = btrfs_file_extent_ram_bytes(leaf,
3563 * The first search might have left us at an extent
3564 * item that ends before our target range's start, can
3565 * happen if we have holes and NO_HOLES feature enabled.
3567 if (key.offset + datal <= off) {
3570 } else if (key.offset >= off + len) {
3573 next_key_min_offset = key.offset + datal;
3574 size = btrfs_item_size_nr(leaf, slot);
3575 read_extent_buffer(leaf, buf,
3576 btrfs_item_ptr_offset(leaf, slot),
3579 btrfs_release_path(path);
3580 path->leave_spinning = 0;
3582 memcpy(&new_key, &key, sizeof(new_key));
3583 new_key.objectid = btrfs_ino(inode);
3584 if (off <= key.offset)
3585 new_key.offset = key.offset + destoff - off;
3587 new_key.offset = destoff;
3590 * Deal with a hole that doesn't have an extent item
3591 * that represents it (NO_HOLES feature enabled).
3592 * This hole is either in the middle of the cloning
3593 * range or at the beginning (fully overlaps it or
3594 * partially overlaps it).
3596 if (new_key.offset != last_dest_end)
3597 drop_start = last_dest_end;
3599 drop_start = new_key.offset;
3602 * 1 - adjusting old extent (we may have to split it)
3603 * 1 - add new extent
3606 trans = btrfs_start_transaction(root, 3);
3607 if (IS_ERR(trans)) {
3608 ret = PTR_ERR(trans);
3612 if (type == BTRFS_FILE_EXTENT_REG ||
3613 type == BTRFS_FILE_EXTENT_PREALLOC) {
3615 * a | --- range to clone ---| b
3616 * | ------------- extent ------------- |
3619 /* subtract range b */
3620 if (key.offset + datal > off + len)
3621 datal = off + len - key.offset;
3623 /* subtract range a */
3624 if (off > key.offset) {
3625 datao += off - key.offset;
3626 datal -= off - key.offset;
3629 ret = btrfs_drop_extents(trans, root, inode,
3631 new_key.offset + datal,
3634 if (ret != -EOPNOTSUPP)
3635 btrfs_abort_transaction(trans,
3637 btrfs_end_transaction(trans, root);
3641 ret = btrfs_insert_empty_item(trans, root, path,
3644 btrfs_abort_transaction(trans, root,
3646 btrfs_end_transaction(trans, root);
3650 leaf = path->nodes[0];
3651 slot = path->slots[0];
3652 write_extent_buffer(leaf, buf,
3653 btrfs_item_ptr_offset(leaf, slot),
3656 extent = btrfs_item_ptr(leaf, slot,
3657 struct btrfs_file_extent_item);
3659 /* disko == 0 means it's a hole */
3663 btrfs_set_file_extent_offset(leaf, extent,
3665 btrfs_set_file_extent_num_bytes(leaf, extent,
3669 inode_add_bytes(inode, datal);
3670 ret = btrfs_inc_extent_ref(trans, root,
3672 root->root_key.objectid,
3674 new_key.offset - datao);
3676 btrfs_abort_transaction(trans,
3679 btrfs_end_transaction(trans,
3685 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3689 if (off > key.offset) {
3690 skip = off - key.offset;
3691 new_key.offset += skip;
3694 if (key.offset + datal > off + len)
3695 trim = key.offset + datal - (off + len);
3697 if (comp && (skip || trim)) {
3699 btrfs_end_transaction(trans, root);
3702 size -= skip + trim;
3703 datal -= skip + trim;
3705 ret = clone_copy_inline_extent(src, inode,
3712 if (ret != -EOPNOTSUPP)
3713 btrfs_abort_transaction(trans,
3716 btrfs_end_transaction(trans, root);
3719 leaf = path->nodes[0];
3720 slot = path->slots[0];
3723 /* If we have an implicit hole (NO_HOLES feature). */
3724 if (drop_start < new_key.offset)
3725 clone_update_extent_map(inode, trans,
3727 new_key.offset - drop_start);
3729 clone_update_extent_map(inode, trans, path, 0, 0);
3731 btrfs_mark_buffer_dirty(leaf);
3732 btrfs_release_path(path);
3734 last_dest_end = ALIGN(new_key.offset + datal,
3736 ret = clone_finish_inode_update(trans, inode,
3742 if (new_key.offset + datal >= destoff + len)
3745 btrfs_release_path(path);
3746 key.offset = next_key_min_offset;
3750 if (last_dest_end < destoff + len) {
3752 * We have an implicit hole (NO_HOLES feature is enabled) that
3753 * fully or partially overlaps our cloning range at its end.
3755 btrfs_release_path(path);
3758 * 1 - remove extent(s)
3761 trans = btrfs_start_transaction(root, 2);
3762 if (IS_ERR(trans)) {
3763 ret = PTR_ERR(trans);
3766 ret = btrfs_drop_extents(trans, root, inode,
3767 last_dest_end, destoff + len, 1);
3769 if (ret != -EOPNOTSUPP)
3770 btrfs_abort_transaction(trans, root, ret);
3771 btrfs_end_transaction(trans, root);
3774 clone_update_extent_map(inode, trans, NULL, last_dest_end,
3775 destoff + len - last_dest_end);
3776 ret = clone_finish_inode_update(trans, inode, destoff + len,
3777 destoff, olen, no_time_update);
3781 btrfs_free_path(path);
3786 static noinline int btrfs_clone_files(struct file *file, struct file *file_src,
3787 u64 off, u64 olen, u64 destoff)
3789 struct inode *inode = file_inode(file);
3790 struct inode *src = file_inode(file_src);
3791 struct btrfs_root *root = BTRFS_I(inode)->root;
3794 u64 bs = root->fs_info->sb->s_blocksize;
3795 int same_inode = src == inode;
3799 * - split compressed inline extents. annoying: we need to
3800 * decompress into destination's address_space (the file offset
3801 * may change, so source mapping won't do), then recompress (or
3802 * otherwise reinsert) a subrange.
3804 * - split destination inode's inline extents. The inline extents can
3805 * be either compressed or non-compressed.
3808 if (btrfs_root_readonly(root))
3811 if (file_src->f_path.mnt != file->f_path.mnt ||
3812 src->i_sb != inode->i_sb)
3815 /* don't make the dst file partly checksummed */
3816 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3817 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3820 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3824 btrfs_double_inode_lock(src, inode);
3829 /* determine range to clone */
3831 if (off + len > src->i_size || off + len < off)
3834 olen = len = src->i_size - off;
3835 /* if we extend to eof, continue to block boundary */
3836 if (off + len == src->i_size)
3837 len = ALIGN(src->i_size, bs) - off;
3844 /* verify the end result is block aligned */
3845 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3846 !IS_ALIGNED(destoff, bs))
3849 /* verify if ranges are overlapped within the same file */
3851 if (destoff + len > off && destoff < off + len)
3855 if (destoff > inode->i_size) {
3856 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3862 * Lock the target range too. Right after we replace the file extent
3863 * items in the fs tree (which now point to the cloned data), we might
3864 * have a worker replace them with extent items relative to a write
3865 * operation that was issued before this clone operation (i.e. confront
3866 * with inode.c:btrfs_finish_ordered_io).
3869 u64 lock_start = min_t(u64, off, destoff);
3870 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
3872 ret = lock_extent_range(src, lock_start, lock_len, true);
3874 ret = btrfs_double_extent_lock(src, off, inode, destoff, len,
3879 /* ranges in the io trees already unlocked */
3883 ret = btrfs_clone(src, inode, off, olen, len, destoff, 0);
3886 u64 lock_start = min_t(u64, off, destoff);
3887 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3889 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3891 btrfs_double_extent_unlock(src, off, inode, destoff, len);
3894 * Truncate page cache pages so that future reads will see the cloned
3895 * data immediately and not the previous data.
3897 truncate_inode_pages_range(&inode->i_data,
3898 round_down(destoff, PAGE_SIZE),
3899 round_up(destoff + len, PAGE_SIZE) - 1);
3902 btrfs_double_inode_unlock(src, inode);
3908 ssize_t btrfs_copy_file_range(struct file *file_in, loff_t pos_in,
3909 struct file *file_out, loff_t pos_out,
3910 size_t len, unsigned int flags)
3914 ret = btrfs_clone_files(file_out, file_in, pos_in, len, pos_out);
3920 int btrfs_clone_file_range(struct file *src_file, loff_t off,
3921 struct file *dst_file, loff_t destoff, u64 len)
3923 return btrfs_clone_files(dst_file, src_file, off, len, destoff);
3927 * there are many ways the trans_start and trans_end ioctls can lead
3928 * to deadlocks. They should only be used by applications that
3929 * basically own the machine, and have a very in depth understanding
3930 * of all the possible deadlocks and enospc problems.
3932 static long btrfs_ioctl_trans_start(struct file *file)
3934 struct inode *inode = file_inode(file);
3935 struct btrfs_root *root = BTRFS_I(inode)->root;
3936 struct btrfs_trans_handle *trans;
3940 if (!capable(CAP_SYS_ADMIN))
3944 if (file->private_data)
3948 if (btrfs_root_readonly(root))
3951 ret = mnt_want_write_file(file);
3955 atomic_inc(&root->fs_info->open_ioctl_trans);
3958 trans = btrfs_start_ioctl_transaction(root);
3962 file->private_data = trans;
3966 atomic_dec(&root->fs_info->open_ioctl_trans);
3967 mnt_drop_write_file(file);
3972 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3974 struct inode *inode = file_inode(file);
3975 struct btrfs_root *root = BTRFS_I(inode)->root;
3976 struct btrfs_root *new_root;
3977 struct btrfs_dir_item *di;
3978 struct btrfs_trans_handle *trans;
3979 struct btrfs_path *path;
3980 struct btrfs_key location;
3981 struct btrfs_disk_key disk_key;
3986 if (!capable(CAP_SYS_ADMIN))
3989 ret = mnt_want_write_file(file);
3993 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3999 objectid = BTRFS_FS_TREE_OBJECTID;
4001 location.objectid = objectid;
4002 location.type = BTRFS_ROOT_ITEM_KEY;
4003 location.offset = (u64)-1;
4005 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
4006 if (IS_ERR(new_root)) {
4007 ret = PTR_ERR(new_root);
4011 path = btrfs_alloc_path();
4016 path->leave_spinning = 1;
4018 trans = btrfs_start_transaction(root, 1);
4019 if (IS_ERR(trans)) {
4020 btrfs_free_path(path);
4021 ret = PTR_ERR(trans);
4025 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
4026 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
4027 dir_id, "default", 7, 1);
4028 if (IS_ERR_OR_NULL(di)) {
4029 btrfs_free_path(path);
4030 btrfs_end_transaction(trans, root);
4031 btrfs_err(new_root->fs_info, "Umm, you don't have the default dir"
4032 "item, this isn't going to work");
4037 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
4038 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
4039 btrfs_mark_buffer_dirty(path->nodes[0]);
4040 btrfs_free_path(path);
4042 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
4043 btrfs_end_transaction(trans, root);
4045 mnt_drop_write_file(file);
4049 void btrfs_get_block_group_info(struct list_head *groups_list,
4050 struct btrfs_ioctl_space_info *space)
4052 struct btrfs_block_group_cache *block_group;
4054 space->total_bytes = 0;
4055 space->used_bytes = 0;
4057 list_for_each_entry(block_group, groups_list, list) {
4058 space->flags = block_group->flags;
4059 space->total_bytes += block_group->key.offset;
4060 space->used_bytes +=
4061 btrfs_block_group_used(&block_group->item);
4065 static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
4067 struct btrfs_ioctl_space_args space_args;
4068 struct btrfs_ioctl_space_info space;
4069 struct btrfs_ioctl_space_info *dest;
4070 struct btrfs_ioctl_space_info *dest_orig;
4071 struct btrfs_ioctl_space_info __user *user_dest;
4072 struct btrfs_space_info *info;
4073 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
4074 BTRFS_BLOCK_GROUP_SYSTEM,
4075 BTRFS_BLOCK_GROUP_METADATA,
4076 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
4083 if (copy_from_user(&space_args,
4084 (struct btrfs_ioctl_space_args __user *)arg,
4085 sizeof(space_args)))
4088 for (i = 0; i < num_types; i++) {
4089 struct btrfs_space_info *tmp;
4093 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
4095 if (tmp->flags == types[i]) {
4105 down_read(&info->groups_sem);
4106 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4107 if (!list_empty(&info->block_groups[c]))
4110 up_read(&info->groups_sem);
4114 * Global block reserve, exported as a space_info
4118 /* space_slots == 0 means they are asking for a count */
4119 if (space_args.space_slots == 0) {
4120 space_args.total_spaces = slot_count;
4124 slot_count = min_t(u64, space_args.space_slots, slot_count);
4126 alloc_size = sizeof(*dest) * slot_count;
4128 /* we generally have at most 6 or so space infos, one for each raid
4129 * level. So, a whole page should be more than enough for everyone
4131 if (alloc_size > PAGE_SIZE)
4134 space_args.total_spaces = 0;
4135 dest = kmalloc(alloc_size, GFP_KERNEL);
4140 /* now we have a buffer to copy into */
4141 for (i = 0; i < num_types; i++) {
4142 struct btrfs_space_info *tmp;
4149 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
4151 if (tmp->flags == types[i]) {
4160 down_read(&info->groups_sem);
4161 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4162 if (!list_empty(&info->block_groups[c])) {
4163 btrfs_get_block_group_info(
4164 &info->block_groups[c], &space);
4165 memcpy(dest, &space, sizeof(space));
4167 space_args.total_spaces++;
4173 up_read(&info->groups_sem);
4177 * Add global block reserve
4180 struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv;
4182 spin_lock(&block_rsv->lock);
4183 space.total_bytes = block_rsv->size;
4184 space.used_bytes = block_rsv->size - block_rsv->reserved;
4185 spin_unlock(&block_rsv->lock);
4186 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
4187 memcpy(dest, &space, sizeof(space));
4188 space_args.total_spaces++;
4191 user_dest = (struct btrfs_ioctl_space_info __user *)
4192 (arg + sizeof(struct btrfs_ioctl_space_args));
4194 if (copy_to_user(user_dest, dest_orig, alloc_size))
4199 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
4206 * there are many ways the trans_start and trans_end ioctls can lead
4207 * to deadlocks. They should only be used by applications that
4208 * basically own the machine, and have a very in depth understanding
4209 * of all the possible deadlocks and enospc problems.
4211 long btrfs_ioctl_trans_end(struct file *file)
4213 struct inode *inode = file_inode(file);
4214 struct btrfs_root *root = BTRFS_I(inode)->root;
4215 struct btrfs_trans_handle *trans;
4217 trans = file->private_data;
4220 file->private_data = NULL;
4222 btrfs_end_transaction(trans, root);
4224 atomic_dec(&root->fs_info->open_ioctl_trans);
4226 mnt_drop_write_file(file);
4230 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4233 struct btrfs_trans_handle *trans;
4237 trans = btrfs_attach_transaction_barrier(root);
4238 if (IS_ERR(trans)) {
4239 if (PTR_ERR(trans) != -ENOENT)
4240 return PTR_ERR(trans);
4242 /* No running transaction, don't bother */
4243 transid = root->fs_info->last_trans_committed;
4246 transid = trans->transid;
4247 ret = btrfs_commit_transaction_async(trans, root, 0);
4249 btrfs_end_transaction(trans, root);
4254 if (copy_to_user(argp, &transid, sizeof(transid)))
4259 static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
4265 if (copy_from_user(&transid, argp, sizeof(transid)))
4268 transid = 0; /* current trans */
4270 return btrfs_wait_for_commit(root, transid);
4273 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
4275 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4276 struct btrfs_ioctl_scrub_args *sa;
4279 if (!capable(CAP_SYS_ADMIN))
4282 sa = memdup_user(arg, sizeof(*sa));
4286 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4287 ret = mnt_want_write_file(file);
4292 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
4293 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4296 if (copy_to_user(arg, sa, sizeof(*sa)))
4299 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4300 mnt_drop_write_file(file);
4306 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
4308 if (!capable(CAP_SYS_ADMIN))
4311 return btrfs_scrub_cancel(root->fs_info);
4314 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
4317 struct btrfs_ioctl_scrub_args *sa;
4320 if (!capable(CAP_SYS_ADMIN))
4323 sa = memdup_user(arg, sizeof(*sa));
4327 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
4329 if (copy_to_user(arg, sa, sizeof(*sa)))
4336 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
4339 struct btrfs_ioctl_get_dev_stats *sa;
4342 sa = memdup_user(arg, sizeof(*sa));
4346 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4351 ret = btrfs_get_dev_stats(root, sa);
4353 if (copy_to_user(arg, sa, sizeof(*sa)))
4360 static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
4362 struct btrfs_ioctl_dev_replace_args *p;
4365 if (!capable(CAP_SYS_ADMIN))
4368 p = memdup_user(arg, sizeof(*p));
4373 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
4374 if (root->fs_info->sb->s_flags & MS_RDONLY) {
4379 &root->fs_info->mutually_exclusive_operation_running,
4381 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4383 ret = btrfs_dev_replace_start(root, p);
4385 &root->fs_info->mutually_exclusive_operation_running,
4389 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4390 btrfs_dev_replace_status(root->fs_info, p);
4393 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4394 ret = btrfs_dev_replace_cancel(root->fs_info, p);
4401 if (copy_to_user(arg, p, sizeof(*p)))
4408 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4414 struct btrfs_ioctl_ino_path_args *ipa = NULL;
4415 struct inode_fs_paths *ipath = NULL;
4416 struct btrfs_path *path;
4418 if (!capable(CAP_DAC_READ_SEARCH))
4421 path = btrfs_alloc_path();
4427 ipa = memdup_user(arg, sizeof(*ipa));
4434 size = min_t(u32, ipa->size, 4096);
4435 ipath = init_ipath(size, root, path);
4436 if (IS_ERR(ipath)) {
4437 ret = PTR_ERR(ipath);
4442 ret = paths_from_inode(ipa->inum, ipath);
4446 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
4447 rel_ptr = ipath->fspath->val[i] -
4448 (u64)(unsigned long)ipath->fspath->val;
4449 ipath->fspath->val[i] = rel_ptr;
4452 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
4453 (void *)(unsigned long)ipath->fspath, size);
4460 btrfs_free_path(path);
4467 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4469 struct btrfs_data_container *inodes = ctx;
4470 const size_t c = 3 * sizeof(u64);
4472 if (inodes->bytes_left >= c) {
4473 inodes->bytes_left -= c;
4474 inodes->val[inodes->elem_cnt] = inum;
4475 inodes->val[inodes->elem_cnt + 1] = offset;
4476 inodes->val[inodes->elem_cnt + 2] = root;
4477 inodes->elem_cnt += 3;
4479 inodes->bytes_missing += c - inodes->bytes_left;
4480 inodes->bytes_left = 0;
4481 inodes->elem_missed += 3;
4487 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
4492 struct btrfs_ioctl_logical_ino_args *loi;
4493 struct btrfs_data_container *inodes = NULL;
4494 struct btrfs_path *path = NULL;
4496 if (!capable(CAP_SYS_ADMIN))
4499 loi = memdup_user(arg, sizeof(*loi));
4506 path = btrfs_alloc_path();
4512 size = min_t(u32, loi->size, SZ_64K);
4513 inodes = init_data_container(size);
4514 if (IS_ERR(inodes)) {
4515 ret = PTR_ERR(inodes);
4520 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
4521 build_ino_list, inodes);
4527 ret = copy_to_user((void *)(unsigned long)loi->inodes,
4528 (void *)(unsigned long)inodes, size);
4533 btrfs_free_path(path);
4540 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
4541 struct btrfs_ioctl_balance_args *bargs)
4543 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4545 bargs->flags = bctl->flags;
4547 if (atomic_read(&fs_info->balance_running))
4548 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4549 if (atomic_read(&fs_info->balance_pause_req))
4550 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
4551 if (atomic_read(&fs_info->balance_cancel_req))
4552 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
4554 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4555 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4556 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4559 spin_lock(&fs_info->balance_lock);
4560 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4561 spin_unlock(&fs_info->balance_lock);
4563 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4567 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4569 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4570 struct btrfs_fs_info *fs_info = root->fs_info;
4571 struct btrfs_ioctl_balance_args *bargs;
4572 struct btrfs_balance_control *bctl;
4573 bool need_unlock; /* for mut. excl. ops lock */
4576 if (!capable(CAP_SYS_ADMIN))
4579 ret = mnt_want_write_file(file);
4584 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
4585 mutex_lock(&fs_info->volume_mutex);
4586 mutex_lock(&fs_info->balance_mutex);
4592 * mut. excl. ops lock is locked. Three possibilites:
4593 * (1) some other op is running
4594 * (2) balance is running
4595 * (3) balance is paused -- special case (think resume)
4597 mutex_lock(&fs_info->balance_mutex);
4598 if (fs_info->balance_ctl) {
4599 /* this is either (2) or (3) */
4600 if (!atomic_read(&fs_info->balance_running)) {
4601 mutex_unlock(&fs_info->balance_mutex);
4602 if (!mutex_trylock(&fs_info->volume_mutex))
4604 mutex_lock(&fs_info->balance_mutex);
4606 if (fs_info->balance_ctl &&
4607 !atomic_read(&fs_info->balance_running)) {
4609 need_unlock = false;
4613 mutex_unlock(&fs_info->balance_mutex);
4614 mutex_unlock(&fs_info->volume_mutex);
4618 mutex_unlock(&fs_info->balance_mutex);
4624 mutex_unlock(&fs_info->balance_mutex);
4625 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4630 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
4633 bargs = memdup_user(arg, sizeof(*bargs));
4634 if (IS_ERR(bargs)) {
4635 ret = PTR_ERR(bargs);
4639 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4640 if (!fs_info->balance_ctl) {
4645 bctl = fs_info->balance_ctl;
4646 spin_lock(&fs_info->balance_lock);
4647 bctl->flags |= BTRFS_BALANCE_RESUME;
4648 spin_unlock(&fs_info->balance_lock);
4656 if (fs_info->balance_ctl) {
4661 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
4667 bctl->fs_info = fs_info;
4669 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4670 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4671 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4673 bctl->flags = bargs->flags;
4675 /* balance everything - no filters */
4676 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4679 if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
4686 * Ownership of bctl and mutually_exclusive_operation_running
4687 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4688 * or, if restriper was paused all the way until unmount, in
4689 * free_fs_info. mutually_exclusive_operation_running is
4690 * cleared in __cancel_balance.
4692 need_unlock = false;
4694 ret = btrfs_balance(bctl, bargs);
4698 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4707 mutex_unlock(&fs_info->balance_mutex);
4708 mutex_unlock(&fs_info->volume_mutex);
4710 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4712 mnt_drop_write_file(file);
4716 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4718 if (!capable(CAP_SYS_ADMIN))
4722 case BTRFS_BALANCE_CTL_PAUSE:
4723 return btrfs_pause_balance(root->fs_info);
4724 case BTRFS_BALANCE_CTL_CANCEL:
4725 return btrfs_cancel_balance(root->fs_info);
4731 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4734 struct btrfs_fs_info *fs_info = root->fs_info;
4735 struct btrfs_ioctl_balance_args *bargs;
4738 if (!capable(CAP_SYS_ADMIN))
4741 mutex_lock(&fs_info->balance_mutex);
4742 if (!fs_info->balance_ctl) {
4747 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
4753 update_ioctl_balance_args(fs_info, 1, bargs);
4755 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4760 mutex_unlock(&fs_info->balance_mutex);
4764 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4766 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4767 struct btrfs_ioctl_quota_ctl_args *sa;
4768 struct btrfs_trans_handle *trans = NULL;
4772 if (!capable(CAP_SYS_ADMIN))
4775 ret = mnt_want_write_file(file);
4779 sa = memdup_user(arg, sizeof(*sa));
4785 down_write(&root->fs_info->subvol_sem);
4786 trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4787 if (IS_ERR(trans)) {
4788 ret = PTR_ERR(trans);
4793 case BTRFS_QUOTA_CTL_ENABLE:
4794 ret = btrfs_quota_enable(trans, root->fs_info);
4796 case BTRFS_QUOTA_CTL_DISABLE:
4797 ret = btrfs_quota_disable(trans, root->fs_info);
4804 err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4809 up_write(&root->fs_info->subvol_sem);
4811 mnt_drop_write_file(file);
4815 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4817 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4818 struct btrfs_ioctl_qgroup_assign_args *sa;
4819 struct btrfs_trans_handle *trans;
4823 if (!capable(CAP_SYS_ADMIN))
4826 ret = mnt_want_write_file(file);
4830 sa = memdup_user(arg, sizeof(*sa));
4836 trans = btrfs_join_transaction(root);
4837 if (IS_ERR(trans)) {
4838 ret = PTR_ERR(trans);
4842 /* FIXME: check if the IDs really exist */
4844 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4847 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4851 /* update qgroup status and info */
4852 err = btrfs_run_qgroups(trans, root->fs_info);
4854 btrfs_std_error(root->fs_info, ret,
4855 "failed to update qgroup status and info\n");
4856 err = btrfs_end_transaction(trans, root);
4863 mnt_drop_write_file(file);
4867 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4869 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4870 struct btrfs_ioctl_qgroup_create_args *sa;
4871 struct btrfs_trans_handle *trans;
4875 if (!capable(CAP_SYS_ADMIN))
4878 ret = mnt_want_write_file(file);
4882 sa = memdup_user(arg, sizeof(*sa));
4888 if (!sa->qgroupid) {
4893 trans = btrfs_join_transaction(root);
4894 if (IS_ERR(trans)) {
4895 ret = PTR_ERR(trans);
4899 /* FIXME: check if the IDs really exist */
4901 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid);
4903 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4906 err = btrfs_end_transaction(trans, root);
4913 mnt_drop_write_file(file);
4917 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4919 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4920 struct btrfs_ioctl_qgroup_limit_args *sa;
4921 struct btrfs_trans_handle *trans;
4926 if (!capable(CAP_SYS_ADMIN))
4929 ret = mnt_want_write_file(file);
4933 sa = memdup_user(arg, sizeof(*sa));
4939 trans = btrfs_join_transaction(root);
4940 if (IS_ERR(trans)) {
4941 ret = PTR_ERR(trans);
4945 qgroupid = sa->qgroupid;
4947 /* take the current subvol as qgroup */
4948 qgroupid = root->root_key.objectid;
4951 /* FIXME: check if the IDs really exist */
4952 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
4954 err = btrfs_end_transaction(trans, root);
4961 mnt_drop_write_file(file);
4965 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4967 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4968 struct btrfs_ioctl_quota_rescan_args *qsa;
4971 if (!capable(CAP_SYS_ADMIN))
4974 ret = mnt_want_write_file(file);
4978 qsa = memdup_user(arg, sizeof(*qsa));
4989 ret = btrfs_qgroup_rescan(root->fs_info);
4994 mnt_drop_write_file(file);
4998 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
5000 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5001 struct btrfs_ioctl_quota_rescan_args *qsa;
5004 if (!capable(CAP_SYS_ADMIN))
5007 qsa = kzalloc(sizeof(*qsa), GFP_KERNEL);
5011 if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
5013 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
5016 if (copy_to_user(arg, qsa, sizeof(*qsa)))
5023 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
5025 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5027 if (!capable(CAP_SYS_ADMIN))
5030 return btrfs_qgroup_wait_for_completion(root->fs_info);
5033 static long _btrfs_ioctl_set_received_subvol(struct file *file,
5034 struct btrfs_ioctl_received_subvol_args *sa)
5036 struct inode *inode = file_inode(file);
5037 struct btrfs_root *root = BTRFS_I(inode)->root;
5038 struct btrfs_root_item *root_item = &root->root_item;
5039 struct btrfs_trans_handle *trans;
5040 struct timespec ct = current_fs_time(inode->i_sb);
5042 int received_uuid_changed;
5044 if (!inode_owner_or_capable(inode))
5047 ret = mnt_want_write_file(file);
5051 down_write(&root->fs_info->subvol_sem);
5053 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
5058 if (btrfs_root_readonly(root)) {
5065 * 2 - uuid items (received uuid + subvol uuid)
5067 trans = btrfs_start_transaction(root, 3);
5068 if (IS_ERR(trans)) {
5069 ret = PTR_ERR(trans);
5074 sa->rtransid = trans->transid;
5075 sa->rtime.sec = ct.tv_sec;
5076 sa->rtime.nsec = ct.tv_nsec;
5078 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
5080 if (received_uuid_changed &&
5081 !btrfs_is_empty_uuid(root_item->received_uuid))
5082 btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
5083 root_item->received_uuid,
5084 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5085 root->root_key.objectid);
5086 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
5087 btrfs_set_root_stransid(root_item, sa->stransid);
5088 btrfs_set_root_rtransid(root_item, sa->rtransid);
5089 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
5090 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
5091 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
5092 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
5094 ret = btrfs_update_root(trans, root->fs_info->tree_root,
5095 &root->root_key, &root->root_item);
5097 btrfs_end_transaction(trans, root);
5100 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
5101 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
5103 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5104 root->root_key.objectid);
5105 if (ret < 0 && ret != -EEXIST) {
5106 btrfs_abort_transaction(trans, root, ret);
5110 ret = btrfs_commit_transaction(trans, root);
5112 btrfs_abort_transaction(trans, root, ret);
5117 up_write(&root->fs_info->subvol_sem);
5118 mnt_drop_write_file(file);
5123 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
5126 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
5127 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
5130 args32 = memdup_user(arg, sizeof(*args32));
5131 if (IS_ERR(args32)) {
5132 ret = PTR_ERR(args32);
5137 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
5143 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
5144 args64->stransid = args32->stransid;
5145 args64->rtransid = args32->rtransid;
5146 args64->stime.sec = args32->stime.sec;
5147 args64->stime.nsec = args32->stime.nsec;
5148 args64->rtime.sec = args32->rtime.sec;
5149 args64->rtime.nsec = args32->rtime.nsec;
5150 args64->flags = args32->flags;
5152 ret = _btrfs_ioctl_set_received_subvol(file, args64);
5156 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
5157 args32->stransid = args64->stransid;
5158 args32->rtransid = args64->rtransid;
5159 args32->stime.sec = args64->stime.sec;
5160 args32->stime.nsec = args64->stime.nsec;
5161 args32->rtime.sec = args64->rtime.sec;
5162 args32->rtime.nsec = args64->rtime.nsec;
5163 args32->flags = args64->flags;
5165 ret = copy_to_user(arg, args32, sizeof(*args32));
5176 static long btrfs_ioctl_set_received_subvol(struct file *file,
5179 struct btrfs_ioctl_received_subvol_args *sa = NULL;
5182 sa = memdup_user(arg, sizeof(*sa));
5189 ret = _btrfs_ioctl_set_received_subvol(file, sa);
5194 ret = copy_to_user(arg, sa, sizeof(*sa));
5203 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
5205 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5208 char label[BTRFS_LABEL_SIZE];
5210 spin_lock(&root->fs_info->super_lock);
5211 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
5212 spin_unlock(&root->fs_info->super_lock);
5214 len = strnlen(label, BTRFS_LABEL_SIZE);
5216 if (len == BTRFS_LABEL_SIZE) {
5217 btrfs_warn(root->fs_info,
5218 "label is too long, return the first %zu bytes", --len);
5221 ret = copy_to_user(arg, label, len);
5223 return ret ? -EFAULT : 0;
5226 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5228 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5229 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5230 struct btrfs_trans_handle *trans;
5231 char label[BTRFS_LABEL_SIZE];
5234 if (!capable(CAP_SYS_ADMIN))
5237 if (copy_from_user(label, arg, sizeof(label)))
5240 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
5241 btrfs_err(root->fs_info, "unable to set label with more than %d bytes",
5242 BTRFS_LABEL_SIZE - 1);
5246 ret = mnt_want_write_file(file);
5250 trans = btrfs_start_transaction(root, 0);
5251 if (IS_ERR(trans)) {
5252 ret = PTR_ERR(trans);
5256 spin_lock(&root->fs_info->super_lock);
5257 strcpy(super_block->label, label);
5258 spin_unlock(&root->fs_info->super_lock);
5259 ret = btrfs_commit_transaction(trans, root);
5262 mnt_drop_write_file(file);
5266 #define INIT_FEATURE_FLAGS(suffix) \
5267 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5268 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5269 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5271 int btrfs_ioctl_get_supported_features(void __user *arg)
5273 static const struct btrfs_ioctl_feature_flags features[3] = {
5274 INIT_FEATURE_FLAGS(SUPP),
5275 INIT_FEATURE_FLAGS(SAFE_SET),
5276 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5279 if (copy_to_user(arg, &features, sizeof(features)))
5285 static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5287 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5288 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5289 struct btrfs_ioctl_feature_flags features;
5291 features.compat_flags = btrfs_super_compat_flags(super_block);
5292 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5293 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5295 if (copy_to_user(arg, &features, sizeof(features)))
5301 static int check_feature_bits(struct btrfs_root *root,
5302 enum btrfs_feature_set set,
5303 u64 change_mask, u64 flags, u64 supported_flags,
5304 u64 safe_set, u64 safe_clear)
5306 const char *type = btrfs_feature_set_names[set];
5308 u64 disallowed, unsupported;
5309 u64 set_mask = flags & change_mask;
5310 u64 clear_mask = ~flags & change_mask;
5312 unsupported = set_mask & ~supported_flags;
5314 names = btrfs_printable_features(set, unsupported);
5316 btrfs_warn(root->fs_info,
5317 "this kernel does not support the %s feature bit%s",
5318 names, strchr(names, ',') ? "s" : "");
5321 btrfs_warn(root->fs_info,
5322 "this kernel does not support %s bits 0x%llx",
5327 disallowed = set_mask & ~safe_set;
5329 names = btrfs_printable_features(set, disallowed);
5331 btrfs_warn(root->fs_info,
5332 "can't set the %s feature bit%s while mounted",
5333 names, strchr(names, ',') ? "s" : "");
5336 btrfs_warn(root->fs_info,
5337 "can't set %s bits 0x%llx while mounted",
5342 disallowed = clear_mask & ~safe_clear;
5344 names = btrfs_printable_features(set, disallowed);
5346 btrfs_warn(root->fs_info,
5347 "can't clear the %s feature bit%s while mounted",
5348 names, strchr(names, ',') ? "s" : "");
5351 btrfs_warn(root->fs_info,
5352 "can't clear %s bits 0x%llx while mounted",
5360 #define check_feature(root, change_mask, flags, mask_base) \
5361 check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
5362 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5363 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5364 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5366 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5368 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5369 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5370 struct btrfs_ioctl_feature_flags flags[2];
5371 struct btrfs_trans_handle *trans;
5375 if (!capable(CAP_SYS_ADMIN))
5378 if (copy_from_user(flags, arg, sizeof(flags)))
5382 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5383 !flags[0].incompat_flags)
5386 ret = check_feature(root, flags[0].compat_flags,
5387 flags[1].compat_flags, COMPAT);
5391 ret = check_feature(root, flags[0].compat_ro_flags,
5392 flags[1].compat_ro_flags, COMPAT_RO);
5396 ret = check_feature(root, flags[0].incompat_flags,
5397 flags[1].incompat_flags, INCOMPAT);
5401 trans = btrfs_start_transaction(root, 0);
5403 return PTR_ERR(trans);
5405 spin_lock(&root->fs_info->super_lock);
5406 newflags = btrfs_super_compat_flags(super_block);
5407 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5408 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5409 btrfs_set_super_compat_flags(super_block, newflags);
5411 newflags = btrfs_super_compat_ro_flags(super_block);
5412 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5413 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5414 btrfs_set_super_compat_ro_flags(super_block, newflags);
5416 newflags = btrfs_super_incompat_flags(super_block);
5417 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5418 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5419 btrfs_set_super_incompat_flags(super_block, newflags);
5420 spin_unlock(&root->fs_info->super_lock);
5422 return btrfs_commit_transaction(trans, root);
5425 long btrfs_ioctl(struct file *file, unsigned int
5426 cmd, unsigned long arg)
5428 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5429 void __user *argp = (void __user *)arg;
5432 case FS_IOC_GETFLAGS:
5433 return btrfs_ioctl_getflags(file, argp);
5434 case FS_IOC_SETFLAGS:
5435 return btrfs_ioctl_setflags(file, argp);
5436 case FS_IOC_GETVERSION:
5437 return btrfs_ioctl_getversion(file, argp);
5439 return btrfs_ioctl_fitrim(file, argp);
5440 case BTRFS_IOC_SNAP_CREATE:
5441 return btrfs_ioctl_snap_create(file, argp, 0);
5442 case BTRFS_IOC_SNAP_CREATE_V2:
5443 return btrfs_ioctl_snap_create_v2(file, argp, 0);
5444 case BTRFS_IOC_SUBVOL_CREATE:
5445 return btrfs_ioctl_snap_create(file, argp, 1);
5446 case BTRFS_IOC_SUBVOL_CREATE_V2:
5447 return btrfs_ioctl_snap_create_v2(file, argp, 1);
5448 case BTRFS_IOC_SNAP_DESTROY:
5449 return btrfs_ioctl_snap_destroy(file, argp);
5450 case BTRFS_IOC_SUBVOL_GETFLAGS:
5451 return btrfs_ioctl_subvol_getflags(file, argp);
5452 case BTRFS_IOC_SUBVOL_SETFLAGS:
5453 return btrfs_ioctl_subvol_setflags(file, argp);
5454 case BTRFS_IOC_DEFAULT_SUBVOL:
5455 return btrfs_ioctl_default_subvol(file, argp);
5456 case BTRFS_IOC_DEFRAG:
5457 return btrfs_ioctl_defrag(file, NULL);
5458 case BTRFS_IOC_DEFRAG_RANGE:
5459 return btrfs_ioctl_defrag(file, argp);
5460 case BTRFS_IOC_RESIZE:
5461 return btrfs_ioctl_resize(file, argp);
5462 case BTRFS_IOC_ADD_DEV:
5463 return btrfs_ioctl_add_dev(root, argp);
5464 case BTRFS_IOC_RM_DEV:
5465 return btrfs_ioctl_rm_dev(file, argp);
5466 case BTRFS_IOC_FS_INFO:
5467 return btrfs_ioctl_fs_info(root, argp);
5468 case BTRFS_IOC_DEV_INFO:
5469 return btrfs_ioctl_dev_info(root, argp);
5470 case BTRFS_IOC_BALANCE:
5471 return btrfs_ioctl_balance(file, NULL);
5472 case BTRFS_IOC_TRANS_START:
5473 return btrfs_ioctl_trans_start(file);
5474 case BTRFS_IOC_TRANS_END:
5475 return btrfs_ioctl_trans_end(file);
5476 case BTRFS_IOC_TREE_SEARCH:
5477 return btrfs_ioctl_tree_search(file, argp);
5478 case BTRFS_IOC_TREE_SEARCH_V2:
5479 return btrfs_ioctl_tree_search_v2(file, argp);
5480 case BTRFS_IOC_INO_LOOKUP:
5481 return btrfs_ioctl_ino_lookup(file, argp);
5482 case BTRFS_IOC_INO_PATHS:
5483 return btrfs_ioctl_ino_to_path(root, argp);
5484 case BTRFS_IOC_LOGICAL_INO:
5485 return btrfs_ioctl_logical_to_ino(root, argp);
5486 case BTRFS_IOC_SPACE_INFO:
5487 return btrfs_ioctl_space_info(root, argp);
5488 case BTRFS_IOC_SYNC: {
5491 ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1);
5494 ret = btrfs_sync_fs(file_inode(file)->i_sb, 1);
5496 * The transaction thread may want to do more work,
5497 * namely it pokes the cleaner ktread that will start
5498 * processing uncleaned subvols.
5500 wake_up_process(root->fs_info->transaction_kthread);
5503 case BTRFS_IOC_START_SYNC:
5504 return btrfs_ioctl_start_sync(root, argp);
5505 case BTRFS_IOC_WAIT_SYNC:
5506 return btrfs_ioctl_wait_sync(root, argp);
5507 case BTRFS_IOC_SCRUB:
5508 return btrfs_ioctl_scrub(file, argp);
5509 case BTRFS_IOC_SCRUB_CANCEL:
5510 return btrfs_ioctl_scrub_cancel(root, argp);
5511 case BTRFS_IOC_SCRUB_PROGRESS:
5512 return btrfs_ioctl_scrub_progress(root, argp);
5513 case BTRFS_IOC_BALANCE_V2:
5514 return btrfs_ioctl_balance(file, argp);
5515 case BTRFS_IOC_BALANCE_CTL:
5516 return btrfs_ioctl_balance_ctl(root, arg);
5517 case BTRFS_IOC_BALANCE_PROGRESS:
5518 return btrfs_ioctl_balance_progress(root, argp);
5519 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5520 return btrfs_ioctl_set_received_subvol(file, argp);
5522 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5523 return btrfs_ioctl_set_received_subvol_32(file, argp);
5525 case BTRFS_IOC_SEND:
5526 return btrfs_ioctl_send(file, argp);
5527 case BTRFS_IOC_GET_DEV_STATS:
5528 return btrfs_ioctl_get_dev_stats(root, argp);
5529 case BTRFS_IOC_QUOTA_CTL:
5530 return btrfs_ioctl_quota_ctl(file, argp);
5531 case BTRFS_IOC_QGROUP_ASSIGN:
5532 return btrfs_ioctl_qgroup_assign(file, argp);
5533 case BTRFS_IOC_QGROUP_CREATE:
5534 return btrfs_ioctl_qgroup_create(file, argp);
5535 case BTRFS_IOC_QGROUP_LIMIT:
5536 return btrfs_ioctl_qgroup_limit(file, argp);
5537 case BTRFS_IOC_QUOTA_RESCAN:
5538 return btrfs_ioctl_quota_rescan(file, argp);
5539 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5540 return btrfs_ioctl_quota_rescan_status(file, argp);
5541 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5542 return btrfs_ioctl_quota_rescan_wait(file, argp);
5543 case BTRFS_IOC_DEV_REPLACE:
5544 return btrfs_ioctl_dev_replace(root, argp);
5545 case BTRFS_IOC_GET_FSLABEL:
5546 return btrfs_ioctl_get_fslabel(file, argp);
5547 case BTRFS_IOC_SET_FSLABEL:
5548 return btrfs_ioctl_set_fslabel(file, argp);
5549 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5550 return btrfs_ioctl_get_supported_features(argp);
5551 case BTRFS_IOC_GET_FEATURES:
5552 return btrfs_ioctl_get_features(file, argp);
5553 case BTRFS_IOC_SET_FEATURES:
5554 return btrfs_ioctl_set_features(file, argp);