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/compat.h>
37 #include <linux/bit_spinlock.h>
38 #include <linux/security.h>
39 #include <linux/xattr.h>
41 #include <linux/slab.h>
42 #include <linux/blkdev.h>
43 #include <linux/uuid.h>
44 #include <linux/btrfs.h>
45 #include <linux/uaccess.h>
48 #include "transaction.h"
49 #include "btrfs_inode.h"
50 #include "print-tree.h"
53 #include "inode-map.h"
55 #include "rcu-string.h"
57 #include "dev-replace.h"
62 #include "compression.h"
65 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
66 * structures are incorrect, as the timespec structure from userspace
67 * is 4 bytes too small. We define these alternatives here to teach
68 * the kernel about the 32-bit struct packing.
70 struct btrfs_ioctl_timespec_32 {
73 } __attribute__ ((__packed__));
75 struct btrfs_ioctl_received_subvol_args_32 {
76 char uuid[BTRFS_UUID_SIZE]; /* in */
77 __u64 stransid; /* in */
78 __u64 rtransid; /* out */
79 struct btrfs_ioctl_timespec_32 stime; /* in */
80 struct btrfs_ioctl_timespec_32 rtime; /* out */
82 __u64 reserved[16]; /* in */
83 } __attribute__ ((__packed__));
85 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
86 struct btrfs_ioctl_received_subvol_args_32)
90 static int btrfs_clone(struct inode *src, struct inode *inode,
91 u64 off, u64 olen, u64 olen_aligned, u64 destoff,
94 /* Mask out flags that are inappropriate for the given type of inode. */
95 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
99 else if (S_ISREG(mode))
100 return flags & ~FS_DIRSYNC_FL;
102 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
106 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
108 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
110 unsigned int iflags = 0;
112 if (flags & BTRFS_INODE_SYNC)
113 iflags |= FS_SYNC_FL;
114 if (flags & BTRFS_INODE_IMMUTABLE)
115 iflags |= FS_IMMUTABLE_FL;
116 if (flags & BTRFS_INODE_APPEND)
117 iflags |= FS_APPEND_FL;
118 if (flags & BTRFS_INODE_NODUMP)
119 iflags |= FS_NODUMP_FL;
120 if (flags & BTRFS_INODE_NOATIME)
121 iflags |= FS_NOATIME_FL;
122 if (flags & BTRFS_INODE_DIRSYNC)
123 iflags |= FS_DIRSYNC_FL;
124 if (flags & BTRFS_INODE_NODATACOW)
125 iflags |= FS_NOCOW_FL;
127 if (flags & BTRFS_INODE_NOCOMPRESS)
128 iflags |= FS_NOCOMP_FL;
129 else if (flags & BTRFS_INODE_COMPRESS)
130 iflags |= FS_COMPR_FL;
136 * Update inode->i_flags based on the btrfs internal flags.
138 void btrfs_update_iflags(struct inode *inode)
140 struct btrfs_inode *ip = BTRFS_I(inode);
141 unsigned int new_fl = 0;
143 if (ip->flags & BTRFS_INODE_SYNC)
145 if (ip->flags & BTRFS_INODE_IMMUTABLE)
146 new_fl |= S_IMMUTABLE;
147 if (ip->flags & BTRFS_INODE_APPEND)
149 if (ip->flags & BTRFS_INODE_NOATIME)
151 if (ip->flags & BTRFS_INODE_DIRSYNC)
154 set_mask_bits(&inode->i_flags,
155 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC,
159 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
161 struct btrfs_inode *ip = BTRFS_I(file_inode(file));
162 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
164 if (copy_to_user(arg, &flags, sizeof(flags)))
169 static int check_flags(unsigned int flags)
171 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
172 FS_NOATIME_FL | FS_NODUMP_FL | \
173 FS_SYNC_FL | FS_DIRSYNC_FL | \
174 FS_NOCOMP_FL | FS_COMPR_FL |
178 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
184 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
186 struct inode *inode = file_inode(file);
187 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
188 struct btrfs_inode *ip = BTRFS_I(inode);
189 struct btrfs_root *root = ip->root;
190 struct btrfs_trans_handle *trans;
191 unsigned int flags, oldflags;
194 unsigned int i_oldflags;
197 if (!inode_owner_or_capable(inode))
200 if (btrfs_root_readonly(root))
203 if (copy_from_user(&flags, arg, sizeof(flags)))
206 ret = check_flags(flags);
210 ret = mnt_want_write_file(file);
216 ip_oldflags = ip->flags;
217 i_oldflags = inode->i_flags;
218 mode = inode->i_mode;
220 flags = btrfs_mask_flags(inode->i_mode, flags);
221 oldflags = btrfs_flags_to_ioctl(ip->flags);
222 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
223 if (!capable(CAP_LINUX_IMMUTABLE)) {
229 if (flags & FS_SYNC_FL)
230 ip->flags |= BTRFS_INODE_SYNC;
232 ip->flags &= ~BTRFS_INODE_SYNC;
233 if (flags & FS_IMMUTABLE_FL)
234 ip->flags |= BTRFS_INODE_IMMUTABLE;
236 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
237 if (flags & FS_APPEND_FL)
238 ip->flags |= BTRFS_INODE_APPEND;
240 ip->flags &= ~BTRFS_INODE_APPEND;
241 if (flags & FS_NODUMP_FL)
242 ip->flags |= BTRFS_INODE_NODUMP;
244 ip->flags &= ~BTRFS_INODE_NODUMP;
245 if (flags & FS_NOATIME_FL)
246 ip->flags |= BTRFS_INODE_NOATIME;
248 ip->flags &= ~BTRFS_INODE_NOATIME;
249 if (flags & FS_DIRSYNC_FL)
250 ip->flags |= BTRFS_INODE_DIRSYNC;
252 ip->flags &= ~BTRFS_INODE_DIRSYNC;
253 if (flags & FS_NOCOW_FL) {
256 * It's safe to turn csums off here, no extents exist.
257 * Otherwise we want the flag to reflect the real COW
258 * status of the file and will not set it.
260 if (inode->i_size == 0)
261 ip->flags |= BTRFS_INODE_NODATACOW
262 | BTRFS_INODE_NODATASUM;
264 ip->flags |= BTRFS_INODE_NODATACOW;
268 * Revert back under same assumptions as above
271 if (inode->i_size == 0)
272 ip->flags &= ~(BTRFS_INODE_NODATACOW
273 | BTRFS_INODE_NODATASUM);
275 ip->flags &= ~BTRFS_INODE_NODATACOW;
280 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
281 * flag may be changed automatically if compression code won't make
284 if (flags & FS_NOCOMP_FL) {
285 ip->flags &= ~BTRFS_INODE_COMPRESS;
286 ip->flags |= BTRFS_INODE_NOCOMPRESS;
288 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
289 if (ret && ret != -ENODATA)
291 } else if (flags & FS_COMPR_FL) {
294 ip->flags |= BTRFS_INODE_COMPRESS;
295 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
297 if (fs_info->compress_type == BTRFS_COMPRESS_LZO)
299 else if (fs_info->compress_type == BTRFS_COMPRESS_ZLIB)
303 ret = btrfs_set_prop(inode, "btrfs.compression",
304 comp, strlen(comp), 0);
309 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
310 if (ret && ret != -ENODATA)
312 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
315 trans = btrfs_start_transaction(root, 1);
317 ret = PTR_ERR(trans);
321 btrfs_update_iflags(inode);
322 inode_inc_iversion(inode);
323 inode->i_ctime = current_time(inode);
324 ret = btrfs_update_inode(trans, root, inode);
326 btrfs_end_transaction(trans);
329 ip->flags = ip_oldflags;
330 inode->i_flags = i_oldflags;
335 mnt_drop_write_file(file);
339 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
341 struct inode *inode = file_inode(file);
343 return put_user(inode->i_generation, arg);
346 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
348 struct inode *inode = file_inode(file);
349 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
350 struct btrfs_device *device;
351 struct request_queue *q;
352 struct fstrim_range range;
353 u64 minlen = ULLONG_MAX;
355 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
358 if (!capable(CAP_SYS_ADMIN))
362 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
366 q = bdev_get_queue(device->bdev);
367 if (blk_queue_discard(q)) {
369 minlen = min_t(u64, q->limits.discard_granularity,
377 if (copy_from_user(&range, arg, sizeof(range)))
379 if (range.start > total_bytes ||
380 range.len < fs_info->sb->s_blocksize)
383 range.len = min(range.len, total_bytes - range.start);
384 range.minlen = max(range.minlen, minlen);
385 ret = btrfs_trim_fs(fs_info, &range);
389 if (copy_to_user(arg, &range, sizeof(range)))
395 int btrfs_is_empty_uuid(u8 *uuid)
399 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
406 static noinline int create_subvol(struct inode *dir,
407 struct dentry *dentry,
408 const char *name, int namelen,
410 struct btrfs_qgroup_inherit *inherit)
412 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
413 struct btrfs_trans_handle *trans;
414 struct btrfs_key key;
415 struct btrfs_root_item *root_item;
416 struct btrfs_inode_item *inode_item;
417 struct extent_buffer *leaf;
418 struct btrfs_root *root = BTRFS_I(dir)->root;
419 struct btrfs_root *new_root;
420 struct btrfs_block_rsv block_rsv;
421 struct timespec cur_time = current_time(dir);
426 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
431 root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
435 ret = btrfs_find_free_objectid(fs_info->tree_root, &objectid);
440 * Don't create subvolume whose level is not zero. Or qgroup will be
441 * screwed up since it assumes subvolume qgroup's level to be 0.
443 if (btrfs_qgroup_level(objectid)) {
448 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
450 * The same as the snapshot creation, please see the comment
451 * of create_snapshot().
453 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
454 8, &qgroup_reserved, false);
458 trans = btrfs_start_transaction(root, 0);
460 ret = PTR_ERR(trans);
461 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
464 trans->block_rsv = &block_rsv;
465 trans->bytes_reserved = block_rsv.size;
467 ret = btrfs_qgroup_inherit(trans, fs_info, 0, objectid, inherit);
471 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
477 memzero_extent_buffer(leaf, 0, sizeof(struct btrfs_header));
478 btrfs_set_header_bytenr(leaf, leaf->start);
479 btrfs_set_header_generation(leaf, trans->transid);
480 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
481 btrfs_set_header_owner(leaf, objectid);
483 write_extent_buffer_fsid(leaf, fs_info->fsid);
484 write_extent_buffer_chunk_tree_uuid(leaf, fs_info->chunk_tree_uuid);
485 btrfs_mark_buffer_dirty(leaf);
487 inode_item = &root_item->inode;
488 btrfs_set_stack_inode_generation(inode_item, 1);
489 btrfs_set_stack_inode_size(inode_item, 3);
490 btrfs_set_stack_inode_nlink(inode_item, 1);
491 btrfs_set_stack_inode_nbytes(inode_item,
493 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
495 btrfs_set_root_flags(root_item, 0);
496 btrfs_set_root_limit(root_item, 0);
497 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
499 btrfs_set_root_bytenr(root_item, leaf->start);
500 btrfs_set_root_generation(root_item, trans->transid);
501 btrfs_set_root_level(root_item, 0);
502 btrfs_set_root_refs(root_item, 1);
503 btrfs_set_root_used(root_item, leaf->len);
504 btrfs_set_root_last_snapshot(root_item, 0);
506 btrfs_set_root_generation_v2(root_item,
507 btrfs_root_generation(root_item));
508 uuid_le_gen(&new_uuid);
509 memcpy(root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE);
510 btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
511 btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
512 root_item->ctime = root_item->otime;
513 btrfs_set_root_ctransid(root_item, trans->transid);
514 btrfs_set_root_otransid(root_item, trans->transid);
516 btrfs_tree_unlock(leaf);
517 free_extent_buffer(leaf);
520 btrfs_set_root_dirid(root_item, new_dirid);
522 key.objectid = objectid;
524 key.type = BTRFS_ROOT_ITEM_KEY;
525 ret = btrfs_insert_root(trans, fs_info->tree_root, &key,
530 key.offset = (u64)-1;
531 new_root = btrfs_read_fs_root_no_name(fs_info, &key);
532 if (IS_ERR(new_root)) {
533 ret = PTR_ERR(new_root);
534 btrfs_abort_transaction(trans, ret);
538 btrfs_record_root_in_trans(trans, new_root);
540 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
542 /* We potentially lose an unused inode item here */
543 btrfs_abort_transaction(trans, ret);
547 mutex_lock(&new_root->objectid_mutex);
548 new_root->highest_objectid = new_dirid;
549 mutex_unlock(&new_root->objectid_mutex);
552 * insert the directory item
554 ret = btrfs_set_inode_index(BTRFS_I(dir), &index);
556 btrfs_abort_transaction(trans, ret);
560 ret = btrfs_insert_dir_item(trans, root,
561 name, namelen, BTRFS_I(dir), &key,
562 BTRFS_FT_DIR, index);
564 btrfs_abort_transaction(trans, ret);
568 btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2);
569 ret = btrfs_update_inode(trans, root, dir);
572 ret = btrfs_add_root_ref(trans, fs_info,
573 objectid, root->root_key.objectid,
574 btrfs_ino(BTRFS_I(dir)), index, name, namelen);
577 ret = btrfs_uuid_tree_add(trans, fs_info, root_item->uuid,
578 BTRFS_UUID_KEY_SUBVOL, objectid);
580 btrfs_abort_transaction(trans, ret);
584 trans->block_rsv = NULL;
585 trans->bytes_reserved = 0;
586 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
589 *async_transid = trans->transid;
590 err = btrfs_commit_transaction_async(trans, 1);
592 err = btrfs_commit_transaction(trans);
594 err = btrfs_commit_transaction(trans);
600 inode = btrfs_lookup_dentry(dir, dentry);
602 return PTR_ERR(inode);
603 d_instantiate(dentry, inode);
612 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
613 struct dentry *dentry,
614 u64 *async_transid, bool readonly,
615 struct btrfs_qgroup_inherit *inherit)
617 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
619 struct btrfs_pending_snapshot *pending_snapshot;
620 struct btrfs_trans_handle *trans;
623 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
626 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
627 if (!pending_snapshot)
630 pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
632 pending_snapshot->path = btrfs_alloc_path();
633 if (!pending_snapshot->root_item || !pending_snapshot->path) {
638 atomic_inc(&root->will_be_snapshotted);
639 smp_mb__after_atomic();
640 /* wait for no snapshot writes */
641 wait_event(root->subv_writers->wait,
642 percpu_counter_sum(&root->subv_writers->counter) == 0);
644 ret = btrfs_start_delalloc_inodes(root, 0);
648 btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1);
650 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
651 BTRFS_BLOCK_RSV_TEMP);
653 * 1 - parent dir inode
656 * 2 - root ref/backref
657 * 1 - root of snapshot
660 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
661 &pending_snapshot->block_rsv, 8,
662 &pending_snapshot->qgroup_reserved,
667 pending_snapshot->dentry = dentry;
668 pending_snapshot->root = root;
669 pending_snapshot->readonly = readonly;
670 pending_snapshot->dir = dir;
671 pending_snapshot->inherit = inherit;
673 trans = btrfs_start_transaction(root, 0);
675 ret = PTR_ERR(trans);
679 spin_lock(&fs_info->trans_lock);
680 list_add(&pending_snapshot->list,
681 &trans->transaction->pending_snapshots);
682 spin_unlock(&fs_info->trans_lock);
684 *async_transid = trans->transid;
685 ret = btrfs_commit_transaction_async(trans, 1);
687 ret = btrfs_commit_transaction(trans);
689 ret = btrfs_commit_transaction(trans);
694 ret = pending_snapshot->error;
698 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
702 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
704 ret = PTR_ERR(inode);
708 d_instantiate(dentry, inode);
711 btrfs_subvolume_release_metadata(fs_info, &pending_snapshot->block_rsv);
713 if (atomic_dec_and_test(&root->will_be_snapshotted))
714 wake_up_atomic_t(&root->will_be_snapshotted);
716 kfree(pending_snapshot->root_item);
717 btrfs_free_path(pending_snapshot->path);
718 kfree(pending_snapshot);
723 /* copy of may_delete in fs/namei.c()
724 * Check whether we can remove a link victim from directory dir, check
725 * whether the type of victim is right.
726 * 1. We can't do it if dir is read-only (done in permission())
727 * 2. We should have write and exec permissions on dir
728 * 3. We can't remove anything from append-only dir
729 * 4. We can't do anything with immutable dir (done in permission())
730 * 5. If the sticky bit on dir is set we should either
731 * a. be owner of dir, or
732 * b. be owner of victim, or
733 * c. have CAP_FOWNER capability
734 * 6. If the victim is append-only or immutable we can't do anything with
735 * links pointing to it.
736 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
737 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
738 * 9. We can't remove a root or mountpoint.
739 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
740 * nfs_async_unlink().
743 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
747 if (d_really_is_negative(victim))
750 BUG_ON(d_inode(victim->d_parent) != dir);
751 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
753 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
758 if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) ||
759 IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim)))
762 if (!d_is_dir(victim))
766 } else if (d_is_dir(victim))
770 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
775 /* copy of may_create in fs/namei.c() */
776 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
778 if (d_really_is_positive(child))
782 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
786 * Create a new subvolume below @parent. This is largely modeled after
787 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
788 * inside this filesystem so it's quite a bit simpler.
790 static noinline int btrfs_mksubvol(const struct path *parent,
791 const char *name, int namelen,
792 struct btrfs_root *snap_src,
793 u64 *async_transid, bool readonly,
794 struct btrfs_qgroup_inherit *inherit)
796 struct inode *dir = d_inode(parent->dentry);
797 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
798 struct dentry *dentry;
801 error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
805 dentry = lookup_one_len(name, parent->dentry, namelen);
806 error = PTR_ERR(dentry);
810 error = btrfs_may_create(dir, dentry);
815 * even if this name doesn't exist, we may get hash collisions.
816 * check for them now when we can safely fail
818 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
824 down_read(&fs_info->subvol_sem);
826 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
830 error = create_snapshot(snap_src, dir, dentry,
831 async_transid, readonly, inherit);
833 error = create_subvol(dir, dentry, name, namelen,
834 async_transid, inherit);
837 fsnotify_mkdir(dir, dentry);
839 up_read(&fs_info->subvol_sem);
848 * When we're defragging a range, we don't want to kick it off again
849 * if it is really just waiting for delalloc to send it down.
850 * If we find a nice big extent or delalloc range for the bytes in the
851 * file you want to defrag, we return 0 to let you know to skip this
854 static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
856 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
857 struct extent_map *em = NULL;
858 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
861 read_lock(&em_tree->lock);
862 em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE);
863 read_unlock(&em_tree->lock);
866 end = extent_map_end(em);
868 if (end - offset > thresh)
871 /* if we already have a nice delalloc here, just stop */
873 end = count_range_bits(io_tree, &offset, offset + thresh,
874 thresh, EXTENT_DELALLOC, 1);
881 * helper function to walk through a file and find extents
882 * newer than a specific transid, and smaller than thresh.
884 * This is used by the defragging code to find new and small
887 static int find_new_extents(struct btrfs_root *root,
888 struct inode *inode, u64 newer_than,
889 u64 *off, u32 thresh)
891 struct btrfs_path *path;
892 struct btrfs_key min_key;
893 struct extent_buffer *leaf;
894 struct btrfs_file_extent_item *extent;
897 u64 ino = btrfs_ino(BTRFS_I(inode));
899 path = btrfs_alloc_path();
903 min_key.objectid = ino;
904 min_key.type = BTRFS_EXTENT_DATA_KEY;
905 min_key.offset = *off;
908 ret = btrfs_search_forward(root, &min_key, path, newer_than);
912 if (min_key.objectid != ino)
914 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
917 leaf = path->nodes[0];
918 extent = btrfs_item_ptr(leaf, path->slots[0],
919 struct btrfs_file_extent_item);
921 type = btrfs_file_extent_type(leaf, extent);
922 if (type == BTRFS_FILE_EXTENT_REG &&
923 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
924 check_defrag_in_cache(inode, min_key.offset, thresh)) {
925 *off = min_key.offset;
926 btrfs_free_path(path);
931 if (path->slots[0] < btrfs_header_nritems(leaf)) {
932 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
936 if (min_key.offset == (u64)-1)
940 btrfs_release_path(path);
943 btrfs_free_path(path);
947 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
949 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
950 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
951 struct extent_map *em;
955 * hopefully we have this extent in the tree already, try without
956 * the full extent lock
958 read_lock(&em_tree->lock);
959 em = lookup_extent_mapping(em_tree, start, len);
960 read_unlock(&em_tree->lock);
963 struct extent_state *cached = NULL;
964 u64 end = start + len - 1;
966 /* get the big lock and read metadata off disk */
967 lock_extent_bits(io_tree, start, end, &cached);
968 em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len, 0);
969 unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS);
978 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
980 struct extent_map *next;
983 /* this is the last extent */
984 if (em->start + em->len >= i_size_read(inode))
987 next = defrag_lookup_extent(inode, em->start + em->len);
988 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
990 else if ((em->block_start + em->block_len == next->block_start) &&
991 (em->block_len > SZ_128K && next->block_len > SZ_128K))
994 free_extent_map(next);
998 static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
999 u64 *last_len, u64 *skip, u64 *defrag_end,
1002 struct extent_map *em;
1004 bool next_mergeable = true;
1005 bool prev_mergeable = true;
1008 * make sure that once we start defragging an extent, we keep on
1011 if (start < *defrag_end)
1016 em = defrag_lookup_extent(inode, start);
1020 /* this will cover holes, and inline extents */
1021 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1027 prev_mergeable = false;
1029 next_mergeable = defrag_check_next_extent(inode, em);
1031 * we hit a real extent, if it is big or the next extent is not a
1032 * real extent, don't bother defragging it
1034 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1035 (em->len >= thresh || (!next_mergeable && !prev_mergeable)))
1039 * last_len ends up being a counter of how many bytes we've defragged.
1040 * every time we choose not to defrag an extent, we reset *last_len
1041 * so that the next tiny extent will force a defrag.
1043 * The end result of this is that tiny extents before a single big
1044 * extent will force at least part of that big extent to be defragged.
1047 *defrag_end = extent_map_end(em);
1050 *skip = extent_map_end(em);
1054 free_extent_map(em);
1059 * it doesn't do much good to defrag one or two pages
1060 * at a time. This pulls in a nice chunk of pages
1061 * to COW and defrag.
1063 * It also makes sure the delalloc code has enough
1064 * dirty data to avoid making new small extents as part
1067 * It's a good idea to start RA on this range
1068 * before calling this.
1070 static int cluster_pages_for_defrag(struct inode *inode,
1071 struct page **pages,
1072 unsigned long start_index,
1073 unsigned long num_pages)
1075 unsigned long file_end;
1076 u64 isize = i_size_read(inode);
1083 struct btrfs_ordered_extent *ordered;
1084 struct extent_state *cached_state = NULL;
1085 struct extent_io_tree *tree;
1086 struct extent_changeset *data_reserved = NULL;
1087 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1089 file_end = (isize - 1) >> PAGE_SHIFT;
1090 if (!isize || start_index > file_end)
1093 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1095 ret = btrfs_delalloc_reserve_space(inode, &data_reserved,
1096 start_index << PAGE_SHIFT,
1097 page_cnt << PAGE_SHIFT);
1101 tree = &BTRFS_I(inode)->io_tree;
1103 /* step one, lock all the pages */
1104 for (i = 0; i < page_cnt; i++) {
1107 page = find_or_create_page(inode->i_mapping,
1108 start_index + i, mask);
1112 page_start = page_offset(page);
1113 page_end = page_start + PAGE_SIZE - 1;
1115 lock_extent_bits(tree, page_start, page_end,
1117 ordered = btrfs_lookup_ordered_extent(inode,
1119 unlock_extent_cached(tree, page_start, page_end,
1120 &cached_state, GFP_NOFS);
1125 btrfs_start_ordered_extent(inode, ordered, 1);
1126 btrfs_put_ordered_extent(ordered);
1129 * we unlocked the page above, so we need check if
1130 * it was released or not.
1132 if (page->mapping != inode->i_mapping) {
1139 if (!PageUptodate(page)) {
1140 btrfs_readpage(NULL, page);
1142 if (!PageUptodate(page)) {
1150 if (page->mapping != inode->i_mapping) {
1162 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1166 * so now we have a nice long stream of locked
1167 * and up to date pages, lets wait on them
1169 for (i = 0; i < i_done; i++)
1170 wait_on_page_writeback(pages[i]);
1172 page_start = page_offset(pages[0]);
1173 page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE;
1175 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1176 page_start, page_end - 1, &cached_state);
1177 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1178 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1179 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1180 &cached_state, GFP_NOFS);
1182 if (i_done != page_cnt) {
1183 spin_lock(&BTRFS_I(inode)->lock);
1184 BTRFS_I(inode)->outstanding_extents++;
1185 spin_unlock(&BTRFS_I(inode)->lock);
1186 btrfs_delalloc_release_space(inode, data_reserved,
1187 start_index << PAGE_SHIFT,
1188 (page_cnt - i_done) << PAGE_SHIFT);
1192 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1195 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1196 page_start, page_end - 1, &cached_state,
1199 for (i = 0; i < i_done; i++) {
1200 clear_page_dirty_for_io(pages[i]);
1201 ClearPageChecked(pages[i]);
1202 set_page_extent_mapped(pages[i]);
1203 set_page_dirty(pages[i]);
1204 unlock_page(pages[i]);
1207 extent_changeset_free(data_reserved);
1210 for (i = 0; i < i_done; i++) {
1211 unlock_page(pages[i]);
1214 btrfs_delalloc_release_space(inode, data_reserved,
1215 start_index << PAGE_SHIFT,
1216 page_cnt << PAGE_SHIFT);
1217 extent_changeset_free(data_reserved);
1222 int btrfs_defrag_file(struct inode *inode, struct file *file,
1223 struct btrfs_ioctl_defrag_range_args *range,
1224 u64 newer_than, unsigned long max_to_defrag)
1226 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1227 struct btrfs_root *root = BTRFS_I(inode)->root;
1228 struct file_ra_state *ra = NULL;
1229 unsigned long last_index;
1230 u64 isize = i_size_read(inode);
1234 u64 newer_off = range->start;
1236 unsigned long ra_index = 0;
1238 int defrag_count = 0;
1239 int compress_type = BTRFS_COMPRESS_ZLIB;
1240 u32 extent_thresh = range->extent_thresh;
1241 unsigned long max_cluster = SZ_256K >> PAGE_SHIFT;
1242 unsigned long cluster = max_cluster;
1243 u64 new_align = ~((u64)SZ_128K - 1);
1244 struct page **pages = NULL;
1245 bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS;
1250 if (range->start >= isize)
1254 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1256 if (range->compress_type)
1257 compress_type = range->compress_type;
1260 if (extent_thresh == 0)
1261 extent_thresh = SZ_256K;
1264 * If we were not given a file, allocate a readahead context. As
1265 * readahead is just an optimization, defrag will work without it so
1266 * we don't error out.
1269 ra = kzalloc(sizeof(*ra), GFP_KERNEL);
1271 file_ra_state_init(ra, inode->i_mapping);
1276 pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL);
1282 /* find the last page to defrag */
1283 if (range->start + range->len > range->start) {
1284 last_index = min_t(u64, isize - 1,
1285 range->start + range->len - 1) >> PAGE_SHIFT;
1287 last_index = (isize - 1) >> PAGE_SHIFT;
1291 ret = find_new_extents(root, inode, newer_than,
1292 &newer_off, SZ_64K);
1294 range->start = newer_off;
1296 * we always align our defrag to help keep
1297 * the extents in the file evenly spaced
1299 i = (newer_off & new_align) >> PAGE_SHIFT;
1303 i = range->start >> PAGE_SHIFT;
1306 max_to_defrag = last_index - i + 1;
1309 * make writeback starts from i, so the defrag range can be
1310 * written sequentially.
1312 if (i < inode->i_mapping->writeback_index)
1313 inode->i_mapping->writeback_index = i;
1315 while (i <= last_index && defrag_count < max_to_defrag &&
1316 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) {
1318 * make sure we stop running if someone unmounts
1321 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1324 if (btrfs_defrag_cancelled(fs_info)) {
1325 btrfs_debug(fs_info, "defrag_file cancelled");
1330 if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT,
1331 extent_thresh, &last_len, &skip,
1332 &defrag_end, do_compress)){
1335 * the should_defrag function tells us how much to skip
1336 * bump our counter by the suggested amount
1338 next = DIV_ROUND_UP(skip, PAGE_SIZE);
1339 i = max(i + 1, next);
1344 cluster = (PAGE_ALIGN(defrag_end) >>
1346 cluster = min(cluster, max_cluster);
1348 cluster = max_cluster;
1351 if (i + cluster > ra_index) {
1352 ra_index = max(i, ra_index);
1354 page_cache_sync_readahead(inode->i_mapping, ra,
1355 file, ra_index, cluster);
1356 ra_index += cluster;
1361 BTRFS_I(inode)->defrag_compress = compress_type;
1362 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1364 inode_unlock(inode);
1368 defrag_count += ret;
1369 balance_dirty_pages_ratelimited(inode->i_mapping);
1370 inode_unlock(inode);
1373 if (newer_off == (u64)-1)
1379 newer_off = max(newer_off + 1,
1380 (u64)i << PAGE_SHIFT);
1382 ret = find_new_extents(root, inode, newer_than,
1383 &newer_off, SZ_64K);
1385 range->start = newer_off;
1386 i = (newer_off & new_align) >> PAGE_SHIFT;
1393 last_len += ret << PAGE_SHIFT;
1401 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1402 filemap_flush(inode->i_mapping);
1403 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1404 &BTRFS_I(inode)->runtime_flags))
1405 filemap_flush(inode->i_mapping);
1408 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1409 btrfs_set_fs_incompat(fs_info, COMPRESS_LZO);
1410 } else if (range->compress_type == BTRFS_COMPRESS_ZSTD) {
1411 btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD);
1419 BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE;
1420 inode_unlock(inode);
1428 static noinline int btrfs_ioctl_resize(struct file *file,
1431 struct inode *inode = file_inode(file);
1432 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1436 struct btrfs_root *root = BTRFS_I(inode)->root;
1437 struct btrfs_ioctl_vol_args *vol_args;
1438 struct btrfs_trans_handle *trans;
1439 struct btrfs_device *device = NULL;
1442 char *devstr = NULL;
1446 if (!capable(CAP_SYS_ADMIN))
1449 ret = mnt_want_write_file(file);
1453 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
1454 mnt_drop_write_file(file);
1455 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1458 mutex_lock(&fs_info->volume_mutex);
1459 vol_args = memdup_user(arg, sizeof(*vol_args));
1460 if (IS_ERR(vol_args)) {
1461 ret = PTR_ERR(vol_args);
1465 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1467 sizestr = vol_args->name;
1468 devstr = strchr(sizestr, ':');
1470 sizestr = devstr + 1;
1472 devstr = vol_args->name;
1473 ret = kstrtoull(devstr, 10, &devid);
1480 btrfs_info(fs_info, "resizing devid %llu", devid);
1483 device = btrfs_find_device(fs_info, devid, NULL, NULL);
1485 btrfs_info(fs_info, "resizer unable to find device %llu",
1491 if (!device->writeable) {
1493 "resizer unable to apply on readonly device %llu",
1499 if (!strcmp(sizestr, "max"))
1500 new_size = device->bdev->bd_inode->i_size;
1502 if (sizestr[0] == '-') {
1505 } else if (sizestr[0] == '+') {
1509 new_size = memparse(sizestr, &retptr);
1510 if (*retptr != '\0' || new_size == 0) {
1516 if (device->is_tgtdev_for_dev_replace) {
1521 old_size = btrfs_device_get_total_bytes(device);
1524 if (new_size > old_size) {
1528 new_size = old_size - new_size;
1529 } else if (mod > 0) {
1530 if (new_size > ULLONG_MAX - old_size) {
1534 new_size = old_size + new_size;
1537 if (new_size < SZ_256M) {
1541 if (new_size > device->bdev->bd_inode->i_size) {
1546 new_size = round_down(new_size, fs_info->sectorsize);
1548 btrfs_info_in_rcu(fs_info, "new size for %s is %llu",
1549 rcu_str_deref(device->name), new_size);
1551 if (new_size > old_size) {
1552 trans = btrfs_start_transaction(root, 0);
1553 if (IS_ERR(trans)) {
1554 ret = PTR_ERR(trans);
1557 ret = btrfs_grow_device(trans, device, new_size);
1558 btrfs_commit_transaction(trans);
1559 } else if (new_size < old_size) {
1560 ret = btrfs_shrink_device(device, new_size);
1561 } /* equal, nothing need to do */
1566 mutex_unlock(&fs_info->volume_mutex);
1567 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
1568 mnt_drop_write_file(file);
1572 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1573 const char *name, unsigned long fd, int subvol,
1574 u64 *transid, bool readonly,
1575 struct btrfs_qgroup_inherit *inherit)
1580 if (!S_ISDIR(file_inode(file)->i_mode))
1583 ret = mnt_want_write_file(file);
1587 namelen = strlen(name);
1588 if (strchr(name, '/')) {
1590 goto out_drop_write;
1593 if (name[0] == '.' &&
1594 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1596 goto out_drop_write;
1600 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1601 NULL, transid, readonly, inherit);
1603 struct fd src = fdget(fd);
1604 struct inode *src_inode;
1607 goto out_drop_write;
1610 src_inode = file_inode(src.file);
1611 if (src_inode->i_sb != file_inode(file)->i_sb) {
1612 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
1613 "Snapshot src from another FS");
1615 } else if (!inode_owner_or_capable(src_inode)) {
1617 * Subvolume creation is not restricted, but snapshots
1618 * are limited to own subvolumes only
1622 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1623 BTRFS_I(src_inode)->root,
1624 transid, readonly, inherit);
1629 mnt_drop_write_file(file);
1634 static noinline int btrfs_ioctl_snap_create(struct file *file,
1635 void __user *arg, int subvol)
1637 struct btrfs_ioctl_vol_args *vol_args;
1640 if (!S_ISDIR(file_inode(file)->i_mode))
1643 vol_args = memdup_user(arg, sizeof(*vol_args));
1644 if (IS_ERR(vol_args))
1645 return PTR_ERR(vol_args);
1646 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1648 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1649 vol_args->fd, subvol,
1656 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1657 void __user *arg, int subvol)
1659 struct btrfs_ioctl_vol_args_v2 *vol_args;
1663 bool readonly = false;
1664 struct btrfs_qgroup_inherit *inherit = NULL;
1666 if (!S_ISDIR(file_inode(file)->i_mode))
1669 vol_args = memdup_user(arg, sizeof(*vol_args));
1670 if (IS_ERR(vol_args))
1671 return PTR_ERR(vol_args);
1672 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1674 if (vol_args->flags &
1675 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1676 BTRFS_SUBVOL_QGROUP_INHERIT)) {
1681 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1683 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1685 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1686 if (vol_args->size > PAGE_SIZE) {
1690 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1691 if (IS_ERR(inherit)) {
1692 ret = PTR_ERR(inherit);
1697 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1698 vol_args->fd, subvol, ptr,
1703 if (ptr && copy_to_user(arg +
1704 offsetof(struct btrfs_ioctl_vol_args_v2,
1716 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1719 struct inode *inode = file_inode(file);
1720 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1721 struct btrfs_root *root = BTRFS_I(inode)->root;
1725 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID)
1728 down_read(&fs_info->subvol_sem);
1729 if (btrfs_root_readonly(root))
1730 flags |= BTRFS_SUBVOL_RDONLY;
1731 up_read(&fs_info->subvol_sem);
1733 if (copy_to_user(arg, &flags, sizeof(flags)))
1739 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1742 struct inode *inode = file_inode(file);
1743 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1744 struct btrfs_root *root = BTRFS_I(inode)->root;
1745 struct btrfs_trans_handle *trans;
1750 if (!inode_owner_or_capable(inode))
1753 ret = mnt_want_write_file(file);
1757 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1759 goto out_drop_write;
1762 if (copy_from_user(&flags, arg, sizeof(flags))) {
1764 goto out_drop_write;
1767 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1769 goto out_drop_write;
1772 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1774 goto out_drop_write;
1777 down_write(&fs_info->subvol_sem);
1780 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1783 root_flags = btrfs_root_flags(&root->root_item);
1784 if (flags & BTRFS_SUBVOL_RDONLY) {
1785 btrfs_set_root_flags(&root->root_item,
1786 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1789 * Block RO -> RW transition if this subvolume is involved in
1792 spin_lock(&root->root_item_lock);
1793 if (root->send_in_progress == 0) {
1794 btrfs_set_root_flags(&root->root_item,
1795 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1796 spin_unlock(&root->root_item_lock);
1798 spin_unlock(&root->root_item_lock);
1800 "Attempt to set subvolume %llu read-write during send",
1801 root->root_key.objectid);
1807 trans = btrfs_start_transaction(root, 1);
1808 if (IS_ERR(trans)) {
1809 ret = PTR_ERR(trans);
1813 ret = btrfs_update_root(trans, fs_info->tree_root,
1814 &root->root_key, &root->root_item);
1816 btrfs_end_transaction(trans);
1820 ret = btrfs_commit_transaction(trans);
1824 btrfs_set_root_flags(&root->root_item, root_flags);
1826 up_write(&fs_info->subvol_sem);
1828 mnt_drop_write_file(file);
1834 * helper to check if the subvolume references other subvolumes
1836 static noinline int may_destroy_subvol(struct btrfs_root *root)
1838 struct btrfs_fs_info *fs_info = root->fs_info;
1839 struct btrfs_path *path;
1840 struct btrfs_dir_item *di;
1841 struct btrfs_key key;
1845 path = btrfs_alloc_path();
1849 /* Make sure this root isn't set as the default subvol */
1850 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1851 di = btrfs_lookup_dir_item(NULL, fs_info->tree_root, path,
1852 dir_id, "default", 7, 0);
1853 if (di && !IS_ERR(di)) {
1854 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1855 if (key.objectid == root->root_key.objectid) {
1858 "deleting default subvolume %llu is not allowed",
1862 btrfs_release_path(path);
1865 key.objectid = root->root_key.objectid;
1866 key.type = BTRFS_ROOT_REF_KEY;
1867 key.offset = (u64)-1;
1869 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
1875 if (path->slots[0] > 0) {
1877 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1878 if (key.objectid == root->root_key.objectid &&
1879 key.type == BTRFS_ROOT_REF_KEY)
1883 btrfs_free_path(path);
1887 static noinline int key_in_sk(struct btrfs_key *key,
1888 struct btrfs_ioctl_search_key *sk)
1890 struct btrfs_key test;
1893 test.objectid = sk->min_objectid;
1894 test.type = sk->min_type;
1895 test.offset = sk->min_offset;
1897 ret = btrfs_comp_cpu_keys(key, &test);
1901 test.objectid = sk->max_objectid;
1902 test.type = sk->max_type;
1903 test.offset = sk->max_offset;
1905 ret = btrfs_comp_cpu_keys(key, &test);
1911 static noinline int copy_to_sk(struct btrfs_path *path,
1912 struct btrfs_key *key,
1913 struct btrfs_ioctl_search_key *sk,
1916 unsigned long *sk_offset,
1920 struct extent_buffer *leaf;
1921 struct btrfs_ioctl_search_header sh;
1922 struct btrfs_key test;
1923 unsigned long item_off;
1924 unsigned long item_len;
1930 leaf = path->nodes[0];
1931 slot = path->slots[0];
1932 nritems = btrfs_header_nritems(leaf);
1934 if (btrfs_header_generation(leaf) > sk->max_transid) {
1938 found_transid = btrfs_header_generation(leaf);
1940 for (i = slot; i < nritems; i++) {
1941 item_off = btrfs_item_ptr_offset(leaf, i);
1942 item_len = btrfs_item_size_nr(leaf, i);
1944 btrfs_item_key_to_cpu(leaf, key, i);
1945 if (!key_in_sk(key, sk))
1948 if (sizeof(sh) + item_len > *buf_size) {
1955 * return one empty item back for v1, which does not
1959 *buf_size = sizeof(sh) + item_len;
1964 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
1969 sh.objectid = key->objectid;
1970 sh.offset = key->offset;
1971 sh.type = key->type;
1973 sh.transid = found_transid;
1975 /* copy search result header */
1976 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
1981 *sk_offset += sizeof(sh);
1984 char __user *up = ubuf + *sk_offset;
1986 if (read_extent_buffer_to_user(leaf, up,
1987 item_off, item_len)) {
1992 *sk_offset += item_len;
1996 if (ret) /* -EOVERFLOW from above */
1999 if (*num_found >= sk->nr_items) {
2006 test.objectid = sk->max_objectid;
2007 test.type = sk->max_type;
2008 test.offset = sk->max_offset;
2009 if (btrfs_comp_cpu_keys(key, &test) >= 0)
2011 else if (key->offset < (u64)-1)
2013 else if (key->type < (u8)-1) {
2016 } else if (key->objectid < (u64)-1) {
2024 * 0: all items from this leaf copied, continue with next
2025 * 1: * more items can be copied, but unused buffer is too small
2026 * * all items were found
2027 * Either way, it will stops the loop which iterates to the next
2029 * -EOVERFLOW: item was to large for buffer
2030 * -EFAULT: could not copy extent buffer back to userspace
2035 static noinline int search_ioctl(struct inode *inode,
2036 struct btrfs_ioctl_search_key *sk,
2040 struct btrfs_fs_info *info = btrfs_sb(inode->i_sb);
2041 struct btrfs_root *root;
2042 struct btrfs_key key;
2043 struct btrfs_path *path;
2046 unsigned long sk_offset = 0;
2048 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2049 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2053 path = btrfs_alloc_path();
2057 if (sk->tree_id == 0) {
2058 /* search the root of the inode that was passed */
2059 root = BTRFS_I(inode)->root;
2061 key.objectid = sk->tree_id;
2062 key.type = BTRFS_ROOT_ITEM_KEY;
2063 key.offset = (u64)-1;
2064 root = btrfs_read_fs_root_no_name(info, &key);
2066 btrfs_free_path(path);
2071 key.objectid = sk->min_objectid;
2072 key.type = sk->min_type;
2073 key.offset = sk->min_offset;
2076 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2082 ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
2083 &sk_offset, &num_found);
2084 btrfs_release_path(path);
2092 sk->nr_items = num_found;
2093 btrfs_free_path(path);
2097 static noinline int btrfs_ioctl_tree_search(struct file *file,
2100 struct btrfs_ioctl_search_args __user *uargs;
2101 struct btrfs_ioctl_search_key sk;
2102 struct inode *inode;
2106 if (!capable(CAP_SYS_ADMIN))
2109 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2111 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2114 buf_size = sizeof(uargs->buf);
2116 inode = file_inode(file);
2117 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2120 * In the origin implementation an overflow is handled by returning a
2121 * search header with a len of zero, so reset ret.
2123 if (ret == -EOVERFLOW)
2126 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2131 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2134 struct btrfs_ioctl_search_args_v2 __user *uarg;
2135 struct btrfs_ioctl_search_args_v2 args;
2136 struct inode *inode;
2139 const size_t buf_limit = SZ_16M;
2141 if (!capable(CAP_SYS_ADMIN))
2144 /* copy search header and buffer size */
2145 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2146 if (copy_from_user(&args, uarg, sizeof(args)))
2149 buf_size = args.buf_size;
2151 /* limit result size to 16MB */
2152 if (buf_size > buf_limit)
2153 buf_size = buf_limit;
2155 inode = file_inode(file);
2156 ret = search_ioctl(inode, &args.key, &buf_size,
2157 (char __user *)(&uarg->buf[0]));
2158 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2160 else if (ret == -EOVERFLOW &&
2161 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2168 * Search INODE_REFs to identify path name of 'dirid' directory
2169 * in a 'tree_id' tree. and sets path name to 'name'.
2171 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2172 u64 tree_id, u64 dirid, char *name)
2174 struct btrfs_root *root;
2175 struct btrfs_key key;
2181 struct btrfs_inode_ref *iref;
2182 struct extent_buffer *l;
2183 struct btrfs_path *path;
2185 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2190 path = btrfs_alloc_path();
2194 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
2196 key.objectid = tree_id;
2197 key.type = BTRFS_ROOT_ITEM_KEY;
2198 key.offset = (u64)-1;
2199 root = btrfs_read_fs_root_no_name(info, &key);
2201 btrfs_err(info, "could not find root %llu", tree_id);
2206 key.objectid = dirid;
2207 key.type = BTRFS_INODE_REF_KEY;
2208 key.offset = (u64)-1;
2211 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2215 ret = btrfs_previous_item(root, path, dirid,
2216 BTRFS_INODE_REF_KEY);
2226 slot = path->slots[0];
2227 btrfs_item_key_to_cpu(l, &key, slot);
2229 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2230 len = btrfs_inode_ref_name_len(l, iref);
2232 total_len += len + 1;
2234 ret = -ENAMETOOLONG;
2239 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2241 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2244 btrfs_release_path(path);
2245 key.objectid = key.offset;
2246 key.offset = (u64)-1;
2247 dirid = key.objectid;
2249 memmove(name, ptr, total_len);
2250 name[total_len] = '\0';
2253 btrfs_free_path(path);
2257 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2260 struct btrfs_ioctl_ino_lookup_args *args;
2261 struct inode *inode;
2264 args = memdup_user(argp, sizeof(*args));
2266 return PTR_ERR(args);
2268 inode = file_inode(file);
2271 * Unprivileged query to obtain the containing subvolume root id. The
2272 * path is reset so it's consistent with btrfs_search_path_in_tree.
2274 if (args->treeid == 0)
2275 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2277 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2282 if (!capable(CAP_SYS_ADMIN)) {
2287 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2288 args->treeid, args->objectid,
2292 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2299 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2302 struct dentry *parent = file->f_path.dentry;
2303 struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb);
2304 struct dentry *dentry;
2305 struct inode *dir = d_inode(parent);
2306 struct inode *inode;
2307 struct btrfs_root *root = BTRFS_I(dir)->root;
2308 struct btrfs_root *dest = NULL;
2309 struct btrfs_ioctl_vol_args *vol_args;
2310 struct btrfs_trans_handle *trans;
2311 struct btrfs_block_rsv block_rsv;
2313 u64 qgroup_reserved;
2318 if (!S_ISDIR(dir->i_mode))
2321 vol_args = memdup_user(arg, sizeof(*vol_args));
2322 if (IS_ERR(vol_args))
2323 return PTR_ERR(vol_args);
2325 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2326 namelen = strlen(vol_args->name);
2327 if (strchr(vol_args->name, '/') ||
2328 strncmp(vol_args->name, "..", namelen) == 0) {
2333 err = mnt_want_write_file(file);
2338 err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
2340 goto out_drop_write;
2341 dentry = lookup_one_len(vol_args->name, parent, namelen);
2342 if (IS_ERR(dentry)) {
2343 err = PTR_ERR(dentry);
2344 goto out_unlock_dir;
2347 if (d_really_is_negative(dentry)) {
2352 inode = d_inode(dentry);
2353 dest = BTRFS_I(inode)->root;
2354 if (!capable(CAP_SYS_ADMIN)) {
2356 * Regular user. Only allow this with a special mount
2357 * option, when the user has write+exec access to the
2358 * subvol root, and when rmdir(2) would have been
2361 * Note that this is _not_ check that the subvol is
2362 * empty or doesn't contain data that we wouldn't
2363 * otherwise be able to delete.
2365 * Users who want to delete empty subvols should try
2369 if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
2373 * Do not allow deletion if the parent dir is the same
2374 * as the dir to be deleted. That means the ioctl
2375 * must be called on the dentry referencing the root
2376 * of the subvol, not a random directory contained
2383 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2388 /* check if subvolume may be deleted by a user */
2389 err = btrfs_may_delete(dir, dentry, 1);
2393 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
2401 * Don't allow to delete a subvolume with send in progress. This is
2402 * inside the i_mutex so the error handling that has to drop the bit
2403 * again is not run concurrently.
2405 spin_lock(&dest->root_item_lock);
2406 root_flags = btrfs_root_flags(&dest->root_item);
2407 if (dest->send_in_progress == 0) {
2408 btrfs_set_root_flags(&dest->root_item,
2409 root_flags | BTRFS_ROOT_SUBVOL_DEAD);
2410 spin_unlock(&dest->root_item_lock);
2412 spin_unlock(&dest->root_item_lock);
2414 "Attempt to delete subvolume %llu during send",
2415 dest->root_key.objectid);
2417 goto out_unlock_inode;
2420 down_write(&fs_info->subvol_sem);
2422 err = may_destroy_subvol(dest);
2426 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2428 * One for dir inode, two for dir entries, two for root
2431 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2432 5, &qgroup_reserved, true);
2436 trans = btrfs_start_transaction(root, 0);
2437 if (IS_ERR(trans)) {
2438 err = PTR_ERR(trans);
2441 trans->block_rsv = &block_rsv;
2442 trans->bytes_reserved = block_rsv.size;
2444 btrfs_record_snapshot_destroy(trans, BTRFS_I(dir));
2446 ret = btrfs_unlink_subvol(trans, root, dir,
2447 dest->root_key.objectid,
2448 dentry->d_name.name,
2449 dentry->d_name.len);
2452 btrfs_abort_transaction(trans, ret);
2456 btrfs_record_root_in_trans(trans, dest);
2458 memset(&dest->root_item.drop_progress, 0,
2459 sizeof(dest->root_item.drop_progress));
2460 dest->root_item.drop_level = 0;
2461 btrfs_set_root_refs(&dest->root_item, 0);
2463 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
2464 ret = btrfs_insert_orphan_item(trans,
2466 dest->root_key.objectid);
2468 btrfs_abort_transaction(trans, ret);
2474 ret = btrfs_uuid_tree_rem(trans, fs_info, dest->root_item.uuid,
2475 BTRFS_UUID_KEY_SUBVOL,
2476 dest->root_key.objectid);
2477 if (ret && ret != -ENOENT) {
2478 btrfs_abort_transaction(trans, ret);
2482 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2483 ret = btrfs_uuid_tree_rem(trans, fs_info,
2484 dest->root_item.received_uuid,
2485 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2486 dest->root_key.objectid);
2487 if (ret && ret != -ENOENT) {
2488 btrfs_abort_transaction(trans, ret);
2495 trans->block_rsv = NULL;
2496 trans->bytes_reserved = 0;
2497 ret = btrfs_end_transaction(trans);
2500 inode->i_flags |= S_DEAD;
2502 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
2504 up_write(&fs_info->subvol_sem);
2506 spin_lock(&dest->root_item_lock);
2507 root_flags = btrfs_root_flags(&dest->root_item);
2508 btrfs_set_root_flags(&dest->root_item,
2509 root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
2510 spin_unlock(&dest->root_item_lock);
2513 inode_unlock(inode);
2515 d_invalidate(dentry);
2516 btrfs_invalidate_inodes(dest);
2518 ASSERT(dest->send_in_progress == 0);
2521 if (dest->ino_cache_inode) {
2522 iput(dest->ino_cache_inode);
2523 dest->ino_cache_inode = NULL;
2531 mnt_drop_write_file(file);
2537 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2539 struct inode *inode = file_inode(file);
2540 struct btrfs_root *root = BTRFS_I(inode)->root;
2541 struct btrfs_ioctl_defrag_range_args *range;
2544 ret = mnt_want_write_file(file);
2548 if (btrfs_root_readonly(root)) {
2553 switch (inode->i_mode & S_IFMT) {
2555 if (!capable(CAP_SYS_ADMIN)) {
2559 ret = btrfs_defrag_root(root);
2562 if (!(file->f_mode & FMODE_WRITE)) {
2567 range = kzalloc(sizeof(*range), GFP_KERNEL);
2574 if (copy_from_user(range, argp,
2580 /* compression requires us to start the IO */
2581 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2582 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2583 range->extent_thresh = (u32)-1;
2586 /* the rest are all set to zero by kzalloc */
2587 range->len = (u64)-1;
2589 ret = btrfs_defrag_file(file_inode(file), file,
2599 mnt_drop_write_file(file);
2603 static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg)
2605 struct btrfs_ioctl_vol_args *vol_args;
2608 if (!capable(CAP_SYS_ADMIN))
2611 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags))
2612 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2614 mutex_lock(&fs_info->volume_mutex);
2615 vol_args = memdup_user(arg, sizeof(*vol_args));
2616 if (IS_ERR(vol_args)) {
2617 ret = PTR_ERR(vol_args);
2621 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2622 ret = btrfs_init_new_device(fs_info, vol_args->name);
2625 btrfs_info(fs_info, "disk added %s", vol_args->name);
2629 mutex_unlock(&fs_info->volume_mutex);
2630 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
2634 static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
2636 struct inode *inode = file_inode(file);
2637 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2638 struct btrfs_ioctl_vol_args_v2 *vol_args;
2641 if (!capable(CAP_SYS_ADMIN))
2644 ret = mnt_want_write_file(file);
2648 vol_args = memdup_user(arg, sizeof(*vol_args));
2649 if (IS_ERR(vol_args)) {
2650 ret = PTR_ERR(vol_args);
2654 /* Check for compatibility reject unknown flags */
2655 if (vol_args->flags & ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED)
2658 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
2659 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2663 mutex_lock(&fs_info->volume_mutex);
2664 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
2665 ret = btrfs_rm_device(fs_info, NULL, vol_args->devid);
2667 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
2668 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
2670 mutex_unlock(&fs_info->volume_mutex);
2671 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
2674 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
2675 btrfs_info(fs_info, "device deleted: id %llu",
2678 btrfs_info(fs_info, "device deleted: %s",
2684 mnt_drop_write_file(file);
2688 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2690 struct inode *inode = file_inode(file);
2691 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2692 struct btrfs_ioctl_vol_args *vol_args;
2695 if (!capable(CAP_SYS_ADMIN))
2698 ret = mnt_want_write_file(file);
2702 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
2703 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2704 goto out_drop_write;
2707 vol_args = memdup_user(arg, sizeof(*vol_args));
2708 if (IS_ERR(vol_args)) {
2709 ret = PTR_ERR(vol_args);
2713 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2714 mutex_lock(&fs_info->volume_mutex);
2715 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
2716 mutex_unlock(&fs_info->volume_mutex);
2719 btrfs_info(fs_info, "disk deleted %s", vol_args->name);
2722 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
2724 mnt_drop_write_file(file);
2729 static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
2732 struct btrfs_ioctl_fs_info_args *fi_args;
2733 struct btrfs_device *device;
2734 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2737 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2741 mutex_lock(&fs_devices->device_list_mutex);
2742 fi_args->num_devices = fs_devices->num_devices;
2743 memcpy(&fi_args->fsid, fs_info->fsid, sizeof(fi_args->fsid));
2745 list_for_each_entry(device, &fs_devices->devices, dev_list) {
2746 if (device->devid > fi_args->max_id)
2747 fi_args->max_id = device->devid;
2749 mutex_unlock(&fs_devices->device_list_mutex);
2751 fi_args->nodesize = fs_info->nodesize;
2752 fi_args->sectorsize = fs_info->sectorsize;
2753 fi_args->clone_alignment = fs_info->sectorsize;
2755 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2762 static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
2765 struct btrfs_ioctl_dev_info_args *di_args;
2766 struct btrfs_device *dev;
2767 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2769 char *s_uuid = NULL;
2771 di_args = memdup_user(arg, sizeof(*di_args));
2772 if (IS_ERR(di_args))
2773 return PTR_ERR(di_args);
2775 if (!btrfs_is_empty_uuid(di_args->uuid))
2776 s_uuid = di_args->uuid;
2778 mutex_lock(&fs_devices->device_list_mutex);
2779 dev = btrfs_find_device(fs_info, di_args->devid, s_uuid, NULL);
2786 di_args->devid = dev->devid;
2787 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2788 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
2789 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2791 struct rcu_string *name;
2794 name = rcu_dereference(dev->name);
2795 strncpy(di_args->path, name->str, sizeof(di_args->path));
2797 di_args->path[sizeof(di_args->path) - 1] = 0;
2799 di_args->path[0] = '\0';
2803 mutex_unlock(&fs_devices->device_list_mutex);
2804 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2811 static struct page *extent_same_get_page(struct inode *inode, pgoff_t index)
2815 page = grab_cache_page(inode->i_mapping, index);
2817 return ERR_PTR(-ENOMEM);
2819 if (!PageUptodate(page)) {
2822 ret = btrfs_readpage(NULL, page);
2824 return ERR_PTR(ret);
2826 if (!PageUptodate(page)) {
2829 return ERR_PTR(-EIO);
2831 if (page->mapping != inode->i_mapping) {
2834 return ERR_PTR(-EAGAIN);
2841 static int gather_extent_pages(struct inode *inode, struct page **pages,
2842 int num_pages, u64 off)
2845 pgoff_t index = off >> PAGE_SHIFT;
2847 for (i = 0; i < num_pages; i++) {
2849 pages[i] = extent_same_get_page(inode, index + i);
2850 if (IS_ERR(pages[i])) {
2851 int err = PTR_ERR(pages[i]);
2862 static int lock_extent_range(struct inode *inode, u64 off, u64 len,
2863 bool retry_range_locking)
2866 * Do any pending delalloc/csum calculations on inode, one way or
2867 * another, and lock file content.
2868 * The locking order is:
2871 * 2) range in the inode's io tree
2874 struct btrfs_ordered_extent *ordered;
2875 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2876 ordered = btrfs_lookup_first_ordered_extent(inode,
2879 ordered->file_offset + ordered->len <= off ||
2880 ordered->file_offset >= off + len) &&
2881 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2882 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2884 btrfs_put_ordered_extent(ordered);
2887 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2889 btrfs_put_ordered_extent(ordered);
2890 if (!retry_range_locking)
2892 btrfs_wait_ordered_range(inode, off, len);
2897 static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2)
2899 inode_unlock(inode1);
2900 inode_unlock(inode2);
2903 static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2)
2905 if (inode1 < inode2)
2906 swap(inode1, inode2);
2908 inode_lock_nested(inode1, I_MUTEX_PARENT);
2909 inode_lock_nested(inode2, I_MUTEX_CHILD);
2912 static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1,
2913 struct inode *inode2, u64 loff2, u64 len)
2915 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2916 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2919 static int btrfs_double_extent_lock(struct inode *inode1, u64 loff1,
2920 struct inode *inode2, u64 loff2, u64 len,
2921 bool retry_range_locking)
2925 if (inode1 < inode2) {
2926 swap(inode1, inode2);
2929 ret = lock_extent_range(inode1, loff1, len, retry_range_locking);
2932 ret = lock_extent_range(inode2, loff2, len, retry_range_locking);
2934 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1,
2941 struct page **src_pages;
2942 struct page **dst_pages;
2945 static void btrfs_cmp_data_free(struct cmp_pages *cmp)
2950 for (i = 0; i < cmp->num_pages; i++) {
2951 pg = cmp->src_pages[i];
2956 pg = cmp->dst_pages[i];
2962 kfree(cmp->src_pages);
2963 kfree(cmp->dst_pages);
2966 static int btrfs_cmp_data_prepare(struct inode *src, u64 loff,
2967 struct inode *dst, u64 dst_loff,
2968 u64 len, struct cmp_pages *cmp)
2971 int num_pages = PAGE_ALIGN(len) >> PAGE_SHIFT;
2972 struct page **src_pgarr, **dst_pgarr;
2975 * We must gather up all the pages before we initiate our
2976 * extent locking. We use an array for the page pointers. Size
2977 * of the array is bounded by len, which is in turn bounded by
2978 * BTRFS_MAX_DEDUPE_LEN.
2980 src_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL);
2981 dst_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL);
2982 if (!src_pgarr || !dst_pgarr) {
2987 cmp->num_pages = num_pages;
2988 cmp->src_pages = src_pgarr;
2989 cmp->dst_pages = dst_pgarr;
2992 * If deduping ranges in the same inode, locking rules make it mandatory
2993 * to always lock pages in ascending order to avoid deadlocks with
2994 * concurrent tasks (such as starting writeback/delalloc).
2996 if (src == dst && dst_loff < loff) {
2997 swap(src_pgarr, dst_pgarr);
2998 swap(loff, dst_loff);
3001 ret = gather_extent_pages(src, src_pgarr, cmp->num_pages, loff);
3005 ret = gather_extent_pages(dst, dst_pgarr, cmp->num_pages, dst_loff);
3009 btrfs_cmp_data_free(cmp);
3013 static int btrfs_cmp_data(u64 len, struct cmp_pages *cmp)
3017 struct page *src_page, *dst_page;
3018 unsigned int cmp_len = PAGE_SIZE;
3019 void *addr, *dst_addr;
3023 if (len < PAGE_SIZE)
3026 BUG_ON(i >= cmp->num_pages);
3028 src_page = cmp->src_pages[i];
3029 dst_page = cmp->dst_pages[i];
3030 ASSERT(PageLocked(src_page));
3031 ASSERT(PageLocked(dst_page));
3033 addr = kmap_atomic(src_page);
3034 dst_addr = kmap_atomic(dst_page);
3036 flush_dcache_page(src_page);
3037 flush_dcache_page(dst_page);
3039 if (memcmp(addr, dst_addr, cmp_len))
3042 kunmap_atomic(addr);
3043 kunmap_atomic(dst_addr);
3055 static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen,
3059 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
3061 if (off + olen > inode->i_size || off + olen < off)
3064 /* if we extend to eof, continue to block boundary */
3065 if (off + len == inode->i_size)
3066 *plen = len = ALIGN(inode->i_size, bs) - off;
3068 /* Check that we are block aligned - btrfs_clone() requires this */
3069 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
3075 static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen,
3076 struct inode *dst, u64 dst_loff)
3080 struct cmp_pages cmp;
3081 bool same_inode = (src == dst);
3082 u64 same_lock_start = 0;
3083 u64 same_lock_len = 0;
3091 btrfs_double_inode_lock(src, dst);
3093 ret = extent_same_check_offsets(src, loff, &len, olen);
3097 ret = extent_same_check_offsets(dst, dst_loff, &len, olen);
3103 * Single inode case wants the same checks, except we
3104 * don't want our length pushed out past i_size as
3105 * comparing that data range makes no sense.
3107 * extent_same_check_offsets() will do this for an
3108 * unaligned length at i_size, so catch it here and
3109 * reject the request.
3111 * This effectively means we require aligned extents
3112 * for the single-inode case, whereas the other cases
3113 * allow an unaligned length so long as it ends at
3121 /* Check for overlapping ranges */
3122 if (dst_loff + len > loff && dst_loff < loff + len) {
3127 same_lock_start = min_t(u64, loff, dst_loff);
3128 same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start;
3131 /* don't make the dst file partly checksummed */
3132 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3133 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
3139 ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, &cmp);
3144 ret = lock_extent_range(src, same_lock_start, same_lock_len,
3147 ret = btrfs_double_extent_lock(src, loff, dst, dst_loff, len,
3150 * If one of the inodes has dirty pages in the respective range or
3151 * ordered extents, we need to flush dellaloc and wait for all ordered
3152 * extents in the range. We must unlock the pages and the ranges in the
3153 * io trees to avoid deadlocks when flushing delalloc (requires locking
3154 * pages) and when waiting for ordered extents to complete (they require
3157 if (ret == -EAGAIN) {
3159 * Ranges in the io trees already unlocked. Now unlock all
3160 * pages before waiting for all IO to complete.
3162 btrfs_cmp_data_free(&cmp);
3164 btrfs_wait_ordered_range(src, same_lock_start,
3167 btrfs_wait_ordered_range(src, loff, len);
3168 btrfs_wait_ordered_range(dst, dst_loff, len);
3174 /* ranges in the io trees already unlocked */
3175 btrfs_cmp_data_free(&cmp);
3179 /* pass original length for comparison so we stay within i_size */
3180 ret = btrfs_cmp_data(olen, &cmp);
3182 ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1);
3185 unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start,
3186 same_lock_start + same_lock_len - 1);
3188 btrfs_double_extent_unlock(src, loff, dst, dst_loff, len);
3190 btrfs_cmp_data_free(&cmp);
3195 btrfs_double_inode_unlock(src, dst);
3200 #define BTRFS_MAX_DEDUPE_LEN SZ_16M
3202 ssize_t btrfs_dedupe_file_range(struct file *src_file, u64 loff, u64 olen,
3203 struct file *dst_file, u64 dst_loff)
3205 struct inode *src = file_inode(src_file);
3206 struct inode *dst = file_inode(dst_file);
3207 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
3210 if (olen > BTRFS_MAX_DEDUPE_LEN)
3211 olen = BTRFS_MAX_DEDUPE_LEN;
3213 if (WARN_ON_ONCE(bs < PAGE_SIZE)) {
3215 * Btrfs does not support blocksize < page_size. As a
3216 * result, btrfs_cmp_data() won't correctly handle
3217 * this situation without an update.
3222 res = btrfs_extent_same(src, loff, olen, dst, dst_loff);
3228 static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3229 struct inode *inode,
3235 struct btrfs_root *root = BTRFS_I(inode)->root;
3238 inode_inc_iversion(inode);
3239 if (!no_time_update)
3240 inode->i_mtime = inode->i_ctime = current_time(inode);
3242 * We round up to the block size at eof when determining which
3243 * extents to clone above, but shouldn't round up the file size.
3245 if (endoff > destoff + olen)
3246 endoff = destoff + olen;
3247 if (endoff > inode->i_size)
3248 btrfs_i_size_write(BTRFS_I(inode), endoff);
3250 ret = btrfs_update_inode(trans, root, inode);
3252 btrfs_abort_transaction(trans, ret);
3253 btrfs_end_transaction(trans);
3256 ret = btrfs_end_transaction(trans);
3261 static void clone_update_extent_map(struct btrfs_inode *inode,
3262 const struct btrfs_trans_handle *trans,
3263 const struct btrfs_path *path,
3264 const u64 hole_offset,
3267 struct extent_map_tree *em_tree = &inode->extent_tree;
3268 struct extent_map *em;
3271 em = alloc_extent_map();
3273 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags);
3278 struct btrfs_file_extent_item *fi;
3280 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3281 struct btrfs_file_extent_item);
3282 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3283 em->generation = -1;
3284 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3285 BTRFS_FILE_EXTENT_INLINE)
3286 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3287 &inode->runtime_flags);
3289 em->start = hole_offset;
3291 em->ram_bytes = em->len;
3292 em->orig_start = hole_offset;
3293 em->block_start = EXTENT_MAP_HOLE;
3295 em->orig_block_len = 0;
3296 em->compress_type = BTRFS_COMPRESS_NONE;
3297 em->generation = trans->transid;
3301 write_lock(&em_tree->lock);
3302 ret = add_extent_mapping(em_tree, em, 1);
3303 write_unlock(&em_tree->lock);
3304 if (ret != -EEXIST) {
3305 free_extent_map(em);
3308 btrfs_drop_extent_cache(inode, em->start,
3309 em->start + em->len - 1, 0);
3313 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags);
3317 * Make sure we do not end up inserting an inline extent into a file that has
3318 * already other (non-inline) extents. If a file has an inline extent it can
3319 * not have any other extents and the (single) inline extent must start at the
3320 * file offset 0. Failing to respect these rules will lead to file corruption,
3321 * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc
3323 * We can have extents that have been already written to disk or we can have
3324 * dirty ranges still in delalloc, in which case the extent maps and items are
3325 * created only when we run delalloc, and the delalloc ranges might fall outside
3326 * the range we are currently locking in the inode's io tree. So we check the
3327 * inode's i_size because of that (i_size updates are done while holding the
3328 * i_mutex, which we are holding here).
3329 * We also check to see if the inode has a size not greater than "datal" but has
3330 * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are
3331 * protected against such concurrent fallocate calls by the i_mutex).
3333 * If the file has no extents but a size greater than datal, do not allow the
3334 * copy because we would need turn the inline extent into a non-inline one (even
3335 * with NO_HOLES enabled). If we find our destination inode only has one inline
3336 * extent, just overwrite it with the source inline extent if its size is less
3337 * than the source extent's size, or we could copy the source inline extent's
3338 * data into the destination inode's inline extent if the later is greater then
3341 static int clone_copy_inline_extent(struct inode *dst,
3342 struct btrfs_trans_handle *trans,
3343 struct btrfs_path *path,
3344 struct btrfs_key *new_key,
3345 const u64 drop_start,
3351 struct btrfs_fs_info *fs_info = btrfs_sb(dst->i_sb);
3352 struct btrfs_root *root = BTRFS_I(dst)->root;
3353 const u64 aligned_end = ALIGN(new_key->offset + datal,
3354 fs_info->sectorsize);
3356 struct btrfs_key key;
3358 if (new_key->offset > 0)
3361 key.objectid = btrfs_ino(BTRFS_I(dst));
3362 key.type = BTRFS_EXTENT_DATA_KEY;
3364 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3367 } else if (ret > 0) {
3368 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
3369 ret = btrfs_next_leaf(root, path);
3373 goto copy_inline_extent;
3375 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
3376 if (key.objectid == btrfs_ino(BTRFS_I(dst)) &&
3377 key.type == BTRFS_EXTENT_DATA_KEY) {
3378 ASSERT(key.offset > 0);
3381 } else if (i_size_read(dst) <= datal) {
3382 struct btrfs_file_extent_item *ei;
3386 * If the file size is <= datal, make sure there are no other
3387 * extents following (can happen do to an fallocate call with
3388 * the flag FALLOC_FL_KEEP_SIZE).
3390 ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
3391 struct btrfs_file_extent_item);
3393 * If it's an inline extent, it can not have other extents
3396 if (btrfs_file_extent_type(path->nodes[0], ei) ==
3397 BTRFS_FILE_EXTENT_INLINE)
3398 goto copy_inline_extent;
3400 ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei);
3401 if (ext_len > aligned_end)
3404 ret = btrfs_next_item(root, path);
3407 } else if (ret == 0) {
3408 btrfs_item_key_to_cpu(path->nodes[0], &key,
3410 if (key.objectid == btrfs_ino(BTRFS_I(dst)) &&
3411 key.type == BTRFS_EXTENT_DATA_KEY)
3418 * We have no extent items, or we have an extent at offset 0 which may
3419 * or may not be inlined. All these cases are dealt the same way.
3421 if (i_size_read(dst) > datal) {
3423 * If the destination inode has an inline extent...
3424 * This would require copying the data from the source inline
3425 * extent into the beginning of the destination's inline extent.
3426 * But this is really complex, both extents can be compressed
3427 * or just one of them, which would require decompressing and
3428 * re-compressing data (which could increase the new compressed
3429 * size, not allowing the compressed data to fit anymore in an
3431 * So just don't support this case for now (it should be rare,
3432 * we are not really saving space when cloning inline extents).
3437 btrfs_release_path(path);
3438 ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1);
3441 ret = btrfs_insert_empty_item(trans, root, path, new_key, size);
3446 const u32 start = btrfs_file_extent_calc_inline_size(0);
3448 memmove(inline_data + start, inline_data + start + skip, datal);
3451 write_extent_buffer(path->nodes[0], inline_data,
3452 btrfs_item_ptr_offset(path->nodes[0],
3455 inode_add_bytes(dst, datal);
3461 * btrfs_clone() - clone a range from inode file to another
3463 * @src: Inode to clone from
3464 * @inode: Inode to clone to
3465 * @off: Offset within source to start clone from
3466 * @olen: Original length, passed by user, of range to clone
3467 * @olen_aligned: Block-aligned value of olen
3468 * @destoff: Offset within @inode to start clone
3469 * @no_time_update: Whether to update mtime/ctime on the target inode
3471 static int btrfs_clone(struct inode *src, struct inode *inode,
3472 const u64 off, const u64 olen, const u64 olen_aligned,
3473 const u64 destoff, int no_time_update)
3475 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3476 struct btrfs_root *root = BTRFS_I(inode)->root;
3477 struct btrfs_path *path = NULL;
3478 struct extent_buffer *leaf;
3479 struct btrfs_trans_handle *trans;
3481 struct btrfs_key key;
3485 const u64 len = olen_aligned;
3486 u64 last_dest_end = destoff;
3489 buf = kvmalloc(fs_info->nodesize, GFP_KERNEL);
3493 path = btrfs_alloc_path();
3499 path->reada = READA_FORWARD;
3501 key.objectid = btrfs_ino(BTRFS_I(src));
3502 key.type = BTRFS_EXTENT_DATA_KEY;
3506 u64 next_key_min_offset = key.offset + 1;
3509 * note the key will change type as we walk through the
3512 path->leave_spinning = 1;
3513 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3518 * First search, if no extent item that starts at offset off was
3519 * found but the previous item is an extent item, it's possible
3520 * it might overlap our target range, therefore process it.
3522 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3523 btrfs_item_key_to_cpu(path->nodes[0], &key,
3524 path->slots[0] - 1);
3525 if (key.type == BTRFS_EXTENT_DATA_KEY)
3529 nritems = btrfs_header_nritems(path->nodes[0]);
3531 if (path->slots[0] >= nritems) {
3532 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
3537 nritems = btrfs_header_nritems(path->nodes[0]);
3539 leaf = path->nodes[0];
3540 slot = path->slots[0];
3542 btrfs_item_key_to_cpu(leaf, &key, slot);
3543 if (key.type > BTRFS_EXTENT_DATA_KEY ||
3544 key.objectid != btrfs_ino(BTRFS_I(src)))
3547 if (key.type == BTRFS_EXTENT_DATA_KEY) {
3548 struct btrfs_file_extent_item *extent;
3551 struct btrfs_key new_key;
3552 u64 disko = 0, diskl = 0;
3553 u64 datao = 0, datal = 0;
3557 extent = btrfs_item_ptr(leaf, slot,
3558 struct btrfs_file_extent_item);
3559 comp = btrfs_file_extent_compression(leaf, extent);
3560 type = btrfs_file_extent_type(leaf, extent);
3561 if (type == BTRFS_FILE_EXTENT_REG ||
3562 type == BTRFS_FILE_EXTENT_PREALLOC) {
3563 disko = btrfs_file_extent_disk_bytenr(leaf,
3565 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3567 datao = btrfs_file_extent_offset(leaf, extent);
3568 datal = btrfs_file_extent_num_bytes(leaf,
3570 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3571 /* take upper bound, may be compressed */
3572 datal = btrfs_file_extent_ram_bytes(leaf,
3577 * The first search might have left us at an extent
3578 * item that ends before our target range's start, can
3579 * happen if we have holes and NO_HOLES feature enabled.
3581 if (key.offset + datal <= off) {
3584 } else if (key.offset >= off + len) {
3587 next_key_min_offset = key.offset + datal;
3588 size = btrfs_item_size_nr(leaf, slot);
3589 read_extent_buffer(leaf, buf,
3590 btrfs_item_ptr_offset(leaf, slot),
3593 btrfs_release_path(path);
3594 path->leave_spinning = 0;
3596 memcpy(&new_key, &key, sizeof(new_key));
3597 new_key.objectid = btrfs_ino(BTRFS_I(inode));
3598 if (off <= key.offset)
3599 new_key.offset = key.offset + destoff - off;
3601 new_key.offset = destoff;
3604 * Deal with a hole that doesn't have an extent item
3605 * that represents it (NO_HOLES feature enabled).
3606 * This hole is either in the middle of the cloning
3607 * range or at the beginning (fully overlaps it or
3608 * partially overlaps it).
3610 if (new_key.offset != last_dest_end)
3611 drop_start = last_dest_end;
3613 drop_start = new_key.offset;
3616 * 1 - adjusting old extent (we may have to split it)
3617 * 1 - add new extent
3620 trans = btrfs_start_transaction(root, 3);
3621 if (IS_ERR(trans)) {
3622 ret = PTR_ERR(trans);
3626 if (type == BTRFS_FILE_EXTENT_REG ||
3627 type == BTRFS_FILE_EXTENT_PREALLOC) {
3629 * a | --- range to clone ---| b
3630 * | ------------- extent ------------- |
3633 /* subtract range b */
3634 if (key.offset + datal > off + len)
3635 datal = off + len - key.offset;
3637 /* subtract range a */
3638 if (off > key.offset) {
3639 datao += off - key.offset;
3640 datal -= off - key.offset;
3643 ret = btrfs_drop_extents(trans, root, inode,
3645 new_key.offset + datal,
3648 if (ret != -EOPNOTSUPP)
3649 btrfs_abort_transaction(trans,
3651 btrfs_end_transaction(trans);
3655 ret = btrfs_insert_empty_item(trans, root, path,
3658 btrfs_abort_transaction(trans, ret);
3659 btrfs_end_transaction(trans);
3663 leaf = path->nodes[0];
3664 slot = path->slots[0];
3665 write_extent_buffer(leaf, buf,
3666 btrfs_item_ptr_offset(leaf, slot),
3669 extent = btrfs_item_ptr(leaf, slot,
3670 struct btrfs_file_extent_item);
3672 /* disko == 0 means it's a hole */
3676 btrfs_set_file_extent_offset(leaf, extent,
3678 btrfs_set_file_extent_num_bytes(leaf, extent,
3682 inode_add_bytes(inode, datal);
3683 ret = btrfs_inc_extent_ref(trans,
3686 root->root_key.objectid,
3687 btrfs_ino(BTRFS_I(inode)),
3688 new_key.offset - datao);
3690 btrfs_abort_transaction(trans,
3692 btrfs_end_transaction(trans);
3697 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3701 if (off > key.offset) {
3702 skip = off - key.offset;
3703 new_key.offset += skip;
3706 if (key.offset + datal > off + len)
3707 trim = key.offset + datal - (off + len);
3709 if (comp && (skip || trim)) {
3711 btrfs_end_transaction(trans);
3714 size -= skip + trim;
3715 datal -= skip + trim;
3717 ret = clone_copy_inline_extent(inode,
3724 if (ret != -EOPNOTSUPP)
3725 btrfs_abort_transaction(trans,
3727 btrfs_end_transaction(trans);
3730 leaf = path->nodes[0];
3731 slot = path->slots[0];
3734 /* If we have an implicit hole (NO_HOLES feature). */
3735 if (drop_start < new_key.offset)
3736 clone_update_extent_map(BTRFS_I(inode), trans,
3738 new_key.offset - drop_start);
3740 clone_update_extent_map(BTRFS_I(inode), trans,
3743 btrfs_mark_buffer_dirty(leaf);
3744 btrfs_release_path(path);
3746 last_dest_end = ALIGN(new_key.offset + datal,
3747 fs_info->sectorsize);
3748 ret = clone_finish_inode_update(trans, inode,
3754 if (new_key.offset + datal >= destoff + len)
3757 btrfs_release_path(path);
3758 key.offset = next_key_min_offset;
3760 if (fatal_signal_pending(current)) {
3767 if (last_dest_end < destoff + len) {
3769 * We have an implicit hole (NO_HOLES feature is enabled) that
3770 * fully or partially overlaps our cloning range at its end.
3772 btrfs_release_path(path);
3775 * 1 - remove extent(s)
3778 trans = btrfs_start_transaction(root, 2);
3779 if (IS_ERR(trans)) {
3780 ret = PTR_ERR(trans);
3783 ret = btrfs_drop_extents(trans, root, inode,
3784 last_dest_end, destoff + len, 1);
3786 if (ret != -EOPNOTSUPP)
3787 btrfs_abort_transaction(trans, ret);
3788 btrfs_end_transaction(trans);
3791 clone_update_extent_map(BTRFS_I(inode), trans, NULL,
3793 destoff + len - last_dest_end);
3794 ret = clone_finish_inode_update(trans, inode, destoff + len,
3795 destoff, olen, no_time_update);
3799 btrfs_free_path(path);
3804 static noinline int btrfs_clone_files(struct file *file, struct file *file_src,
3805 u64 off, u64 olen, u64 destoff)
3807 struct inode *inode = file_inode(file);
3808 struct inode *src = file_inode(file_src);
3809 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3810 struct btrfs_root *root = BTRFS_I(inode)->root;
3813 u64 bs = fs_info->sb->s_blocksize;
3814 int same_inode = src == inode;
3818 * - split compressed inline extents. annoying: we need to
3819 * decompress into destination's address_space (the file offset
3820 * may change, so source mapping won't do), then recompress (or
3821 * otherwise reinsert) a subrange.
3823 * - split destination inode's inline extents. The inline extents can
3824 * be either compressed or non-compressed.
3827 if (btrfs_root_readonly(root))
3830 if (file_src->f_path.mnt != file->f_path.mnt ||
3831 src->i_sb != inode->i_sb)
3834 /* don't make the dst file partly checksummed */
3835 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3836 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3839 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3843 btrfs_double_inode_lock(src, inode);
3848 /* determine range to clone */
3850 if (off + len > src->i_size || off + len < off)
3853 olen = len = src->i_size - off;
3854 /* if we extend to eof, continue to block boundary */
3855 if (off + len == src->i_size)
3856 len = ALIGN(src->i_size, bs) - off;
3863 /* verify the end result is block aligned */
3864 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3865 !IS_ALIGNED(destoff, bs))
3868 /* verify if ranges are overlapped within the same file */
3870 if (destoff + len > off && destoff < off + len)
3874 if (destoff > inode->i_size) {
3875 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3881 * Lock the target range too. Right after we replace the file extent
3882 * items in the fs tree (which now point to the cloned data), we might
3883 * have a worker replace them with extent items relative to a write
3884 * operation that was issued before this clone operation (i.e. confront
3885 * with inode.c:btrfs_finish_ordered_io).
3888 u64 lock_start = min_t(u64, off, destoff);
3889 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
3891 ret = lock_extent_range(src, lock_start, lock_len, true);
3893 ret = btrfs_double_extent_lock(src, off, inode, destoff, len,
3898 /* ranges in the io trees already unlocked */
3902 ret = btrfs_clone(src, inode, off, olen, len, destoff, 0);
3905 u64 lock_start = min_t(u64, off, destoff);
3906 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3908 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3910 btrfs_double_extent_unlock(src, off, inode, destoff, len);
3913 * Truncate page cache pages so that future reads will see the cloned
3914 * data immediately and not the previous data.
3916 truncate_inode_pages_range(&inode->i_data,
3917 round_down(destoff, PAGE_SIZE),
3918 round_up(destoff + len, PAGE_SIZE) - 1);
3921 btrfs_double_inode_unlock(src, inode);
3927 int btrfs_clone_file_range(struct file *src_file, loff_t off,
3928 struct file *dst_file, loff_t destoff, u64 len)
3930 return btrfs_clone_files(dst_file, src_file, off, len, destoff);
3934 * there are many ways the trans_start and trans_end ioctls can lead
3935 * to deadlocks. They should only be used by applications that
3936 * basically own the machine, and have a very in depth understanding
3937 * of all the possible deadlocks and enospc problems.
3939 static long btrfs_ioctl_trans_start(struct file *file)
3941 struct inode *inode = file_inode(file);
3942 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3943 struct btrfs_root *root = BTRFS_I(inode)->root;
3944 struct btrfs_trans_handle *trans;
3945 struct btrfs_file_private *private;
3947 static bool warned = false;
3950 if (!capable(CAP_SYS_ADMIN))
3955 "Userspace transaction mechanism is considered "
3956 "deprecated and slated to be removed in 4.17. "
3957 "If you have a valid use case please "
3958 "speak up on the mailing list");
3964 private = file->private_data;
3965 if (private && private->trans)
3968 private = kzalloc(sizeof(struct btrfs_file_private),
3972 file->private_data = private;
3976 if (btrfs_root_readonly(root))
3979 ret = mnt_want_write_file(file);
3983 atomic_inc(&fs_info->open_ioctl_trans);
3986 trans = btrfs_start_ioctl_transaction(root);
3990 private->trans = trans;
3994 atomic_dec(&fs_info->open_ioctl_trans);
3995 mnt_drop_write_file(file);
4000 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
4002 struct inode *inode = file_inode(file);
4003 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4004 struct btrfs_root *root = BTRFS_I(inode)->root;
4005 struct btrfs_root *new_root;
4006 struct btrfs_dir_item *di;
4007 struct btrfs_trans_handle *trans;
4008 struct btrfs_path *path;
4009 struct btrfs_key location;
4010 struct btrfs_disk_key disk_key;
4015 if (!capable(CAP_SYS_ADMIN))
4018 ret = mnt_want_write_file(file);
4022 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
4028 objectid = BTRFS_FS_TREE_OBJECTID;
4030 location.objectid = objectid;
4031 location.type = BTRFS_ROOT_ITEM_KEY;
4032 location.offset = (u64)-1;
4034 new_root = btrfs_read_fs_root_no_name(fs_info, &location);
4035 if (IS_ERR(new_root)) {
4036 ret = PTR_ERR(new_root);
4039 if (!is_fstree(new_root->objectid)) {
4044 path = btrfs_alloc_path();
4049 path->leave_spinning = 1;
4051 trans = btrfs_start_transaction(root, 1);
4052 if (IS_ERR(trans)) {
4053 btrfs_free_path(path);
4054 ret = PTR_ERR(trans);
4058 dir_id = btrfs_super_root_dir(fs_info->super_copy);
4059 di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
4060 dir_id, "default", 7, 1);
4061 if (IS_ERR_OR_NULL(di)) {
4062 btrfs_free_path(path);
4063 btrfs_end_transaction(trans);
4065 "Umm, you don't have the default diritem, this isn't going to work");
4070 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
4071 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
4072 btrfs_mark_buffer_dirty(path->nodes[0]);
4073 btrfs_free_path(path);
4075 btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
4076 btrfs_end_transaction(trans);
4078 mnt_drop_write_file(file);
4082 void btrfs_get_block_group_info(struct list_head *groups_list,
4083 struct btrfs_ioctl_space_info *space)
4085 struct btrfs_block_group_cache *block_group;
4087 space->total_bytes = 0;
4088 space->used_bytes = 0;
4090 list_for_each_entry(block_group, groups_list, list) {
4091 space->flags = block_group->flags;
4092 space->total_bytes += block_group->key.offset;
4093 space->used_bytes +=
4094 btrfs_block_group_used(&block_group->item);
4098 static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
4101 struct btrfs_ioctl_space_args space_args;
4102 struct btrfs_ioctl_space_info space;
4103 struct btrfs_ioctl_space_info *dest;
4104 struct btrfs_ioctl_space_info *dest_orig;
4105 struct btrfs_ioctl_space_info __user *user_dest;
4106 struct btrfs_space_info *info;
4107 static const u64 types[] = {
4108 BTRFS_BLOCK_GROUP_DATA,
4109 BTRFS_BLOCK_GROUP_SYSTEM,
4110 BTRFS_BLOCK_GROUP_METADATA,
4111 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA
4119 if (copy_from_user(&space_args,
4120 (struct btrfs_ioctl_space_args __user *)arg,
4121 sizeof(space_args)))
4124 for (i = 0; i < num_types; i++) {
4125 struct btrfs_space_info *tmp;
4129 list_for_each_entry_rcu(tmp, &fs_info->space_info,
4131 if (tmp->flags == types[i]) {
4141 down_read(&info->groups_sem);
4142 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4143 if (!list_empty(&info->block_groups[c]))
4146 up_read(&info->groups_sem);
4150 * Global block reserve, exported as a space_info
4154 /* space_slots == 0 means they are asking for a count */
4155 if (space_args.space_slots == 0) {
4156 space_args.total_spaces = slot_count;
4160 slot_count = min_t(u64, space_args.space_slots, slot_count);
4162 alloc_size = sizeof(*dest) * slot_count;
4164 /* we generally have at most 6 or so space infos, one for each raid
4165 * level. So, a whole page should be more than enough for everyone
4167 if (alloc_size > PAGE_SIZE)
4170 space_args.total_spaces = 0;
4171 dest = kmalloc(alloc_size, GFP_KERNEL);
4176 /* now we have a buffer to copy into */
4177 for (i = 0; i < num_types; i++) {
4178 struct btrfs_space_info *tmp;
4185 list_for_each_entry_rcu(tmp, &fs_info->space_info,
4187 if (tmp->flags == types[i]) {
4196 down_read(&info->groups_sem);
4197 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4198 if (!list_empty(&info->block_groups[c])) {
4199 btrfs_get_block_group_info(
4200 &info->block_groups[c], &space);
4201 memcpy(dest, &space, sizeof(space));
4203 space_args.total_spaces++;
4209 up_read(&info->groups_sem);
4213 * Add global block reserve
4216 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
4218 spin_lock(&block_rsv->lock);
4219 space.total_bytes = block_rsv->size;
4220 space.used_bytes = block_rsv->size - block_rsv->reserved;
4221 spin_unlock(&block_rsv->lock);
4222 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
4223 memcpy(dest, &space, sizeof(space));
4224 space_args.total_spaces++;
4227 user_dest = (struct btrfs_ioctl_space_info __user *)
4228 (arg + sizeof(struct btrfs_ioctl_space_args));
4230 if (copy_to_user(user_dest, dest_orig, alloc_size))
4235 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
4242 * there are many ways the trans_start and trans_end ioctls can lead
4243 * to deadlocks. They should only be used by applications that
4244 * basically own the machine, and have a very in depth understanding
4245 * of all the possible deadlocks and enospc problems.
4247 long btrfs_ioctl_trans_end(struct file *file)
4249 struct inode *inode = file_inode(file);
4250 struct btrfs_root *root = BTRFS_I(inode)->root;
4251 struct btrfs_file_private *private = file->private_data;
4253 if (!private || !private->trans)
4256 btrfs_end_transaction(private->trans);
4257 private->trans = NULL;
4259 atomic_dec(&root->fs_info->open_ioctl_trans);
4261 mnt_drop_write_file(file);
4265 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4268 struct btrfs_trans_handle *trans;
4272 trans = btrfs_attach_transaction_barrier(root);
4273 if (IS_ERR(trans)) {
4274 if (PTR_ERR(trans) != -ENOENT)
4275 return PTR_ERR(trans);
4277 /* No running transaction, don't bother */
4278 transid = root->fs_info->last_trans_committed;
4281 transid = trans->transid;
4282 ret = btrfs_commit_transaction_async(trans, 0);
4284 btrfs_end_transaction(trans);
4289 if (copy_to_user(argp, &transid, sizeof(transid)))
4294 static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
4300 if (copy_from_user(&transid, argp, sizeof(transid)))
4303 transid = 0; /* current trans */
4305 return btrfs_wait_for_commit(fs_info, transid);
4308 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
4310 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
4311 struct btrfs_ioctl_scrub_args *sa;
4314 if (!capable(CAP_SYS_ADMIN))
4317 sa = memdup_user(arg, sizeof(*sa));
4321 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4322 ret = mnt_want_write_file(file);
4327 ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
4328 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4331 if (copy_to_user(arg, sa, sizeof(*sa)))
4334 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4335 mnt_drop_write_file(file);
4341 static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
4343 if (!capable(CAP_SYS_ADMIN))
4346 return btrfs_scrub_cancel(fs_info);
4349 static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
4352 struct btrfs_ioctl_scrub_args *sa;
4355 if (!capable(CAP_SYS_ADMIN))
4358 sa = memdup_user(arg, sizeof(*sa));
4362 ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
4364 if (copy_to_user(arg, sa, sizeof(*sa)))
4371 static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
4374 struct btrfs_ioctl_get_dev_stats *sa;
4377 sa = memdup_user(arg, sizeof(*sa));
4381 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4386 ret = btrfs_get_dev_stats(fs_info, sa);
4388 if (copy_to_user(arg, sa, sizeof(*sa)))
4395 static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
4398 struct btrfs_ioctl_dev_replace_args *p;
4401 if (!capable(CAP_SYS_ADMIN))
4404 p = memdup_user(arg, sizeof(*p));
4409 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
4410 if (sb_rdonly(fs_info->sb)) {
4414 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
4415 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4417 ret = btrfs_dev_replace_by_ioctl(fs_info, p);
4418 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
4421 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4422 btrfs_dev_replace_status(fs_info, p);
4425 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4426 ret = btrfs_dev_replace_cancel(fs_info, p);
4433 if (copy_to_user(arg, p, sizeof(*p)))
4440 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4446 struct btrfs_ioctl_ino_path_args *ipa = NULL;
4447 struct inode_fs_paths *ipath = NULL;
4448 struct btrfs_path *path;
4450 if (!capable(CAP_DAC_READ_SEARCH))
4453 path = btrfs_alloc_path();
4459 ipa = memdup_user(arg, sizeof(*ipa));
4466 size = min_t(u32, ipa->size, 4096);
4467 ipath = init_ipath(size, root, path);
4468 if (IS_ERR(ipath)) {
4469 ret = PTR_ERR(ipath);
4474 ret = paths_from_inode(ipa->inum, ipath);
4478 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
4479 rel_ptr = ipath->fspath->val[i] -
4480 (u64)(unsigned long)ipath->fspath->val;
4481 ipath->fspath->val[i] = rel_ptr;
4484 ret = copy_to_user((void __user *)(unsigned long)ipa->fspath,
4485 ipath->fspath, size);
4492 btrfs_free_path(path);
4499 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4501 struct btrfs_data_container *inodes = ctx;
4502 const size_t c = 3 * sizeof(u64);
4504 if (inodes->bytes_left >= c) {
4505 inodes->bytes_left -= c;
4506 inodes->val[inodes->elem_cnt] = inum;
4507 inodes->val[inodes->elem_cnt + 1] = offset;
4508 inodes->val[inodes->elem_cnt + 2] = root;
4509 inodes->elem_cnt += 3;
4511 inodes->bytes_missing += c - inodes->bytes_left;
4512 inodes->bytes_left = 0;
4513 inodes->elem_missed += 3;
4519 static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
4524 struct btrfs_ioctl_logical_ino_args *loi;
4525 struct btrfs_data_container *inodes = NULL;
4526 struct btrfs_path *path = NULL;
4528 if (!capable(CAP_SYS_ADMIN))
4531 loi = memdup_user(arg, sizeof(*loi));
4533 return PTR_ERR(loi);
4535 path = btrfs_alloc_path();
4541 size = min_t(u32, loi->size, SZ_64K);
4542 inodes = init_data_container(size);
4543 if (IS_ERR(inodes)) {
4544 ret = PTR_ERR(inodes);
4549 ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
4550 build_ino_list, inodes);
4556 ret = copy_to_user((void __user *)(unsigned long)loi->inodes, inodes,
4562 btrfs_free_path(path);
4569 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
4570 struct btrfs_ioctl_balance_args *bargs)
4572 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4574 bargs->flags = bctl->flags;
4576 if (atomic_read(&fs_info->balance_running))
4577 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4578 if (atomic_read(&fs_info->balance_pause_req))
4579 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
4580 if (atomic_read(&fs_info->balance_cancel_req))
4581 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
4583 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4584 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4585 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4588 spin_lock(&fs_info->balance_lock);
4589 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4590 spin_unlock(&fs_info->balance_lock);
4592 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4596 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4598 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4599 struct btrfs_fs_info *fs_info = root->fs_info;
4600 struct btrfs_ioctl_balance_args *bargs;
4601 struct btrfs_balance_control *bctl;
4602 bool need_unlock; /* for mut. excl. ops lock */
4605 if (!capable(CAP_SYS_ADMIN))
4608 ret = mnt_want_write_file(file);
4613 if (!test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
4614 mutex_lock(&fs_info->volume_mutex);
4615 mutex_lock(&fs_info->balance_mutex);
4621 * mut. excl. ops lock is locked. Three possibilities:
4622 * (1) some other op is running
4623 * (2) balance is running
4624 * (3) balance is paused -- special case (think resume)
4626 mutex_lock(&fs_info->balance_mutex);
4627 if (fs_info->balance_ctl) {
4628 /* this is either (2) or (3) */
4629 if (!atomic_read(&fs_info->balance_running)) {
4630 mutex_unlock(&fs_info->balance_mutex);
4631 if (!mutex_trylock(&fs_info->volume_mutex))
4633 mutex_lock(&fs_info->balance_mutex);
4635 if (fs_info->balance_ctl &&
4636 !atomic_read(&fs_info->balance_running)) {
4638 need_unlock = false;
4642 mutex_unlock(&fs_info->balance_mutex);
4643 mutex_unlock(&fs_info->volume_mutex);
4647 mutex_unlock(&fs_info->balance_mutex);
4653 mutex_unlock(&fs_info->balance_mutex);
4654 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4659 BUG_ON(!test_bit(BTRFS_FS_EXCL_OP, &fs_info->flags));
4662 bargs = memdup_user(arg, sizeof(*bargs));
4663 if (IS_ERR(bargs)) {
4664 ret = PTR_ERR(bargs);
4668 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4669 if (!fs_info->balance_ctl) {
4674 bctl = fs_info->balance_ctl;
4675 spin_lock(&fs_info->balance_lock);
4676 bctl->flags |= BTRFS_BALANCE_RESUME;
4677 spin_unlock(&fs_info->balance_lock);
4685 if (fs_info->balance_ctl) {
4690 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
4696 bctl->fs_info = fs_info;
4698 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4699 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4700 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4702 bctl->flags = bargs->flags;
4704 /* balance everything - no filters */
4705 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4708 if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
4715 * Ownership of bctl and filesystem flag BTRFS_FS_EXCL_OP
4716 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4717 * or, if restriper was paused all the way until unmount, in
4718 * free_fs_info. The flag is cleared in __cancel_balance.
4720 need_unlock = false;
4722 ret = btrfs_balance(bctl, bargs);
4726 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4735 mutex_unlock(&fs_info->balance_mutex);
4736 mutex_unlock(&fs_info->volume_mutex);
4738 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
4740 mnt_drop_write_file(file);
4744 static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
4746 if (!capable(CAP_SYS_ADMIN))
4750 case BTRFS_BALANCE_CTL_PAUSE:
4751 return btrfs_pause_balance(fs_info);
4752 case BTRFS_BALANCE_CTL_CANCEL:
4753 return btrfs_cancel_balance(fs_info);
4759 static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
4762 struct btrfs_ioctl_balance_args *bargs;
4765 if (!capable(CAP_SYS_ADMIN))
4768 mutex_lock(&fs_info->balance_mutex);
4769 if (!fs_info->balance_ctl) {
4774 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
4780 update_ioctl_balance_args(fs_info, 1, bargs);
4782 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4787 mutex_unlock(&fs_info->balance_mutex);
4791 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4793 struct inode *inode = file_inode(file);
4794 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4795 struct btrfs_ioctl_quota_ctl_args *sa;
4796 struct btrfs_trans_handle *trans = NULL;
4800 if (!capable(CAP_SYS_ADMIN))
4803 ret = mnt_want_write_file(file);
4807 sa = memdup_user(arg, sizeof(*sa));
4813 down_write(&fs_info->subvol_sem);
4814 trans = btrfs_start_transaction(fs_info->tree_root, 2);
4815 if (IS_ERR(trans)) {
4816 ret = PTR_ERR(trans);
4821 case BTRFS_QUOTA_CTL_ENABLE:
4822 ret = btrfs_quota_enable(trans, fs_info);
4824 case BTRFS_QUOTA_CTL_DISABLE:
4825 ret = btrfs_quota_disable(trans, fs_info);
4832 err = btrfs_commit_transaction(trans);
4837 up_write(&fs_info->subvol_sem);
4839 mnt_drop_write_file(file);
4843 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4845 struct inode *inode = file_inode(file);
4846 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4847 struct btrfs_root *root = BTRFS_I(inode)->root;
4848 struct btrfs_ioctl_qgroup_assign_args *sa;
4849 struct btrfs_trans_handle *trans;
4853 if (!capable(CAP_SYS_ADMIN))
4856 ret = mnt_want_write_file(file);
4860 sa = memdup_user(arg, sizeof(*sa));
4866 trans = btrfs_join_transaction(root);
4867 if (IS_ERR(trans)) {
4868 ret = PTR_ERR(trans);
4873 ret = btrfs_add_qgroup_relation(trans, fs_info,
4876 ret = btrfs_del_qgroup_relation(trans, fs_info,
4880 /* update qgroup status and info */
4881 err = btrfs_run_qgroups(trans, fs_info);
4883 btrfs_handle_fs_error(fs_info, err,
4884 "failed to update qgroup status and info");
4885 err = btrfs_end_transaction(trans);
4892 mnt_drop_write_file(file);
4896 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4898 struct inode *inode = file_inode(file);
4899 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4900 struct btrfs_root *root = BTRFS_I(inode)->root;
4901 struct btrfs_ioctl_qgroup_create_args *sa;
4902 struct btrfs_trans_handle *trans;
4906 if (!capable(CAP_SYS_ADMIN))
4909 ret = mnt_want_write_file(file);
4913 sa = memdup_user(arg, sizeof(*sa));
4919 if (!sa->qgroupid) {
4924 trans = btrfs_join_transaction(root);
4925 if (IS_ERR(trans)) {
4926 ret = PTR_ERR(trans);
4931 ret = btrfs_create_qgroup(trans, fs_info, sa->qgroupid);
4933 ret = btrfs_remove_qgroup(trans, fs_info, sa->qgroupid);
4936 err = btrfs_end_transaction(trans);
4943 mnt_drop_write_file(file);
4947 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4949 struct inode *inode = file_inode(file);
4950 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4951 struct btrfs_root *root = BTRFS_I(inode)->root;
4952 struct btrfs_ioctl_qgroup_limit_args *sa;
4953 struct btrfs_trans_handle *trans;
4958 if (!capable(CAP_SYS_ADMIN))
4961 ret = mnt_want_write_file(file);
4965 sa = memdup_user(arg, sizeof(*sa));
4971 trans = btrfs_join_transaction(root);
4972 if (IS_ERR(trans)) {
4973 ret = PTR_ERR(trans);
4977 qgroupid = sa->qgroupid;
4979 /* take the current subvol as qgroup */
4980 qgroupid = root->root_key.objectid;
4983 ret = btrfs_limit_qgroup(trans, fs_info, qgroupid, &sa->lim);
4985 err = btrfs_end_transaction(trans);
4992 mnt_drop_write_file(file);
4996 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4998 struct inode *inode = file_inode(file);
4999 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5000 struct btrfs_ioctl_quota_rescan_args *qsa;
5003 if (!capable(CAP_SYS_ADMIN))
5006 ret = mnt_want_write_file(file);
5010 qsa = memdup_user(arg, sizeof(*qsa));
5021 ret = btrfs_qgroup_rescan(fs_info);
5026 mnt_drop_write_file(file);
5030 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
5032 struct inode *inode = file_inode(file);
5033 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5034 struct btrfs_ioctl_quota_rescan_args *qsa;
5037 if (!capable(CAP_SYS_ADMIN))
5040 qsa = kzalloc(sizeof(*qsa), GFP_KERNEL);
5044 if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
5046 qsa->progress = fs_info->qgroup_rescan_progress.objectid;
5049 if (copy_to_user(arg, qsa, sizeof(*qsa)))
5056 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
5058 struct inode *inode = file_inode(file);
5059 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5061 if (!capable(CAP_SYS_ADMIN))
5064 return btrfs_qgroup_wait_for_completion(fs_info, true);
5067 static long _btrfs_ioctl_set_received_subvol(struct file *file,
5068 struct btrfs_ioctl_received_subvol_args *sa)
5070 struct inode *inode = file_inode(file);
5071 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5072 struct btrfs_root *root = BTRFS_I(inode)->root;
5073 struct btrfs_root_item *root_item = &root->root_item;
5074 struct btrfs_trans_handle *trans;
5075 struct timespec ct = current_time(inode);
5077 int received_uuid_changed;
5079 if (!inode_owner_or_capable(inode))
5082 ret = mnt_want_write_file(file);
5086 down_write(&fs_info->subvol_sem);
5088 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
5093 if (btrfs_root_readonly(root)) {
5100 * 2 - uuid items (received uuid + subvol uuid)
5102 trans = btrfs_start_transaction(root, 3);
5103 if (IS_ERR(trans)) {
5104 ret = PTR_ERR(trans);
5109 sa->rtransid = trans->transid;
5110 sa->rtime.sec = ct.tv_sec;
5111 sa->rtime.nsec = ct.tv_nsec;
5113 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
5115 if (received_uuid_changed &&
5116 !btrfs_is_empty_uuid(root_item->received_uuid))
5117 btrfs_uuid_tree_rem(trans, fs_info, root_item->received_uuid,
5118 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5119 root->root_key.objectid);
5120 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
5121 btrfs_set_root_stransid(root_item, sa->stransid);
5122 btrfs_set_root_rtransid(root_item, sa->rtransid);
5123 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
5124 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
5125 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
5126 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
5128 ret = btrfs_update_root(trans, fs_info->tree_root,
5129 &root->root_key, &root->root_item);
5131 btrfs_end_transaction(trans);
5134 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
5135 ret = btrfs_uuid_tree_add(trans, fs_info, sa->uuid,
5136 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5137 root->root_key.objectid);
5138 if (ret < 0 && ret != -EEXIST) {
5139 btrfs_abort_transaction(trans, ret);
5140 btrfs_end_transaction(trans);
5144 ret = btrfs_commit_transaction(trans);
5146 up_write(&fs_info->subvol_sem);
5147 mnt_drop_write_file(file);
5152 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
5155 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
5156 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
5159 args32 = memdup_user(arg, sizeof(*args32));
5161 return PTR_ERR(args32);
5163 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
5169 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
5170 args64->stransid = args32->stransid;
5171 args64->rtransid = args32->rtransid;
5172 args64->stime.sec = args32->stime.sec;
5173 args64->stime.nsec = args32->stime.nsec;
5174 args64->rtime.sec = args32->rtime.sec;
5175 args64->rtime.nsec = args32->rtime.nsec;
5176 args64->flags = args32->flags;
5178 ret = _btrfs_ioctl_set_received_subvol(file, args64);
5182 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
5183 args32->stransid = args64->stransid;
5184 args32->rtransid = args64->rtransid;
5185 args32->stime.sec = args64->stime.sec;
5186 args32->stime.nsec = args64->stime.nsec;
5187 args32->rtime.sec = args64->rtime.sec;
5188 args32->rtime.nsec = args64->rtime.nsec;
5189 args32->flags = args64->flags;
5191 ret = copy_to_user(arg, args32, sizeof(*args32));
5202 static long btrfs_ioctl_set_received_subvol(struct file *file,
5205 struct btrfs_ioctl_received_subvol_args *sa = NULL;
5208 sa = memdup_user(arg, sizeof(*sa));
5212 ret = _btrfs_ioctl_set_received_subvol(file, sa);
5217 ret = copy_to_user(arg, sa, sizeof(*sa));
5226 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
5228 struct inode *inode = file_inode(file);
5229 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5232 char label[BTRFS_LABEL_SIZE];
5234 spin_lock(&fs_info->super_lock);
5235 memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
5236 spin_unlock(&fs_info->super_lock);
5238 len = strnlen(label, BTRFS_LABEL_SIZE);
5240 if (len == BTRFS_LABEL_SIZE) {
5242 "label is too long, return the first %zu bytes",
5246 ret = copy_to_user(arg, label, len);
5248 return ret ? -EFAULT : 0;
5251 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5253 struct inode *inode = file_inode(file);
5254 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5255 struct btrfs_root *root = BTRFS_I(inode)->root;
5256 struct btrfs_super_block *super_block = fs_info->super_copy;
5257 struct btrfs_trans_handle *trans;
5258 char label[BTRFS_LABEL_SIZE];
5261 if (!capable(CAP_SYS_ADMIN))
5264 if (copy_from_user(label, arg, sizeof(label)))
5267 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
5269 "unable to set label with more than %d bytes",
5270 BTRFS_LABEL_SIZE - 1);
5274 ret = mnt_want_write_file(file);
5278 trans = btrfs_start_transaction(root, 0);
5279 if (IS_ERR(trans)) {
5280 ret = PTR_ERR(trans);
5284 spin_lock(&fs_info->super_lock);
5285 strcpy(super_block->label, label);
5286 spin_unlock(&fs_info->super_lock);
5287 ret = btrfs_commit_transaction(trans);
5290 mnt_drop_write_file(file);
5294 #define INIT_FEATURE_FLAGS(suffix) \
5295 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5296 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5297 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5299 int btrfs_ioctl_get_supported_features(void __user *arg)
5301 static const struct btrfs_ioctl_feature_flags features[3] = {
5302 INIT_FEATURE_FLAGS(SUPP),
5303 INIT_FEATURE_FLAGS(SAFE_SET),
5304 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5307 if (copy_to_user(arg, &features, sizeof(features)))
5313 static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5315 struct inode *inode = file_inode(file);
5316 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5317 struct btrfs_super_block *super_block = fs_info->super_copy;
5318 struct btrfs_ioctl_feature_flags features;
5320 features.compat_flags = btrfs_super_compat_flags(super_block);
5321 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5322 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5324 if (copy_to_user(arg, &features, sizeof(features)))
5330 static int check_feature_bits(struct btrfs_fs_info *fs_info,
5331 enum btrfs_feature_set set,
5332 u64 change_mask, u64 flags, u64 supported_flags,
5333 u64 safe_set, u64 safe_clear)
5335 const char *type = btrfs_feature_set_names[set];
5337 u64 disallowed, unsupported;
5338 u64 set_mask = flags & change_mask;
5339 u64 clear_mask = ~flags & change_mask;
5341 unsupported = set_mask & ~supported_flags;
5343 names = btrfs_printable_features(set, unsupported);
5346 "this kernel does not support the %s feature bit%s",
5347 names, strchr(names, ',') ? "s" : "");
5351 "this kernel does not support %s bits 0x%llx",
5356 disallowed = set_mask & ~safe_set;
5358 names = btrfs_printable_features(set, disallowed);
5361 "can't set the %s feature bit%s while mounted",
5362 names, strchr(names, ',') ? "s" : "");
5366 "can't set %s bits 0x%llx while mounted",
5371 disallowed = clear_mask & ~safe_clear;
5373 names = btrfs_printable_features(set, disallowed);
5376 "can't clear the %s feature bit%s while mounted",
5377 names, strchr(names, ',') ? "s" : "");
5381 "can't clear %s bits 0x%llx while mounted",
5389 #define check_feature(fs_info, change_mask, flags, mask_base) \
5390 check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \
5391 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5392 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5393 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5395 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5397 struct inode *inode = file_inode(file);
5398 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5399 struct btrfs_root *root = BTRFS_I(inode)->root;
5400 struct btrfs_super_block *super_block = fs_info->super_copy;
5401 struct btrfs_ioctl_feature_flags flags[2];
5402 struct btrfs_trans_handle *trans;
5406 if (!capable(CAP_SYS_ADMIN))
5409 if (copy_from_user(flags, arg, sizeof(flags)))
5413 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5414 !flags[0].incompat_flags)
5417 ret = check_feature(fs_info, flags[0].compat_flags,
5418 flags[1].compat_flags, COMPAT);
5422 ret = check_feature(fs_info, flags[0].compat_ro_flags,
5423 flags[1].compat_ro_flags, COMPAT_RO);
5427 ret = check_feature(fs_info, flags[0].incompat_flags,
5428 flags[1].incompat_flags, INCOMPAT);
5432 ret = mnt_want_write_file(file);
5436 trans = btrfs_start_transaction(root, 0);
5437 if (IS_ERR(trans)) {
5438 ret = PTR_ERR(trans);
5439 goto out_drop_write;
5442 spin_lock(&fs_info->super_lock);
5443 newflags = btrfs_super_compat_flags(super_block);
5444 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5445 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5446 btrfs_set_super_compat_flags(super_block, newflags);
5448 newflags = btrfs_super_compat_ro_flags(super_block);
5449 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5450 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5451 btrfs_set_super_compat_ro_flags(super_block, newflags);
5453 newflags = btrfs_super_incompat_flags(super_block);
5454 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5455 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5456 btrfs_set_super_incompat_flags(super_block, newflags);
5457 spin_unlock(&fs_info->super_lock);
5459 ret = btrfs_commit_transaction(trans);
5461 mnt_drop_write_file(file);
5466 long btrfs_ioctl(struct file *file, unsigned int
5467 cmd, unsigned long arg)
5469 struct inode *inode = file_inode(file);
5470 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5471 struct btrfs_root *root = BTRFS_I(inode)->root;
5472 void __user *argp = (void __user *)arg;
5475 case FS_IOC_GETFLAGS:
5476 return btrfs_ioctl_getflags(file, argp);
5477 case FS_IOC_SETFLAGS:
5478 return btrfs_ioctl_setflags(file, argp);
5479 case FS_IOC_GETVERSION:
5480 return btrfs_ioctl_getversion(file, argp);
5482 return btrfs_ioctl_fitrim(file, argp);
5483 case BTRFS_IOC_SNAP_CREATE:
5484 return btrfs_ioctl_snap_create(file, argp, 0);
5485 case BTRFS_IOC_SNAP_CREATE_V2:
5486 return btrfs_ioctl_snap_create_v2(file, argp, 0);
5487 case BTRFS_IOC_SUBVOL_CREATE:
5488 return btrfs_ioctl_snap_create(file, argp, 1);
5489 case BTRFS_IOC_SUBVOL_CREATE_V2:
5490 return btrfs_ioctl_snap_create_v2(file, argp, 1);
5491 case BTRFS_IOC_SNAP_DESTROY:
5492 return btrfs_ioctl_snap_destroy(file, argp);
5493 case BTRFS_IOC_SUBVOL_GETFLAGS:
5494 return btrfs_ioctl_subvol_getflags(file, argp);
5495 case BTRFS_IOC_SUBVOL_SETFLAGS:
5496 return btrfs_ioctl_subvol_setflags(file, argp);
5497 case BTRFS_IOC_DEFAULT_SUBVOL:
5498 return btrfs_ioctl_default_subvol(file, argp);
5499 case BTRFS_IOC_DEFRAG:
5500 return btrfs_ioctl_defrag(file, NULL);
5501 case BTRFS_IOC_DEFRAG_RANGE:
5502 return btrfs_ioctl_defrag(file, argp);
5503 case BTRFS_IOC_RESIZE:
5504 return btrfs_ioctl_resize(file, argp);
5505 case BTRFS_IOC_ADD_DEV:
5506 return btrfs_ioctl_add_dev(fs_info, argp);
5507 case BTRFS_IOC_RM_DEV:
5508 return btrfs_ioctl_rm_dev(file, argp);
5509 case BTRFS_IOC_RM_DEV_V2:
5510 return btrfs_ioctl_rm_dev_v2(file, argp);
5511 case BTRFS_IOC_FS_INFO:
5512 return btrfs_ioctl_fs_info(fs_info, argp);
5513 case BTRFS_IOC_DEV_INFO:
5514 return btrfs_ioctl_dev_info(fs_info, argp);
5515 case BTRFS_IOC_BALANCE:
5516 return btrfs_ioctl_balance(file, NULL);
5517 case BTRFS_IOC_TRANS_START:
5518 return btrfs_ioctl_trans_start(file);
5519 case BTRFS_IOC_TRANS_END:
5520 return btrfs_ioctl_trans_end(file);
5521 case BTRFS_IOC_TREE_SEARCH:
5522 return btrfs_ioctl_tree_search(file, argp);
5523 case BTRFS_IOC_TREE_SEARCH_V2:
5524 return btrfs_ioctl_tree_search_v2(file, argp);
5525 case BTRFS_IOC_INO_LOOKUP:
5526 return btrfs_ioctl_ino_lookup(file, argp);
5527 case BTRFS_IOC_INO_PATHS:
5528 return btrfs_ioctl_ino_to_path(root, argp);
5529 case BTRFS_IOC_LOGICAL_INO:
5530 return btrfs_ioctl_logical_to_ino(fs_info, argp);
5531 case BTRFS_IOC_SPACE_INFO:
5532 return btrfs_ioctl_space_info(fs_info, argp);
5533 case BTRFS_IOC_SYNC: {
5536 ret = btrfs_start_delalloc_roots(fs_info, 0, -1);
5539 ret = btrfs_sync_fs(inode->i_sb, 1);
5541 * The transaction thread may want to do more work,
5542 * namely it pokes the cleaner kthread that will start
5543 * processing uncleaned subvols.
5545 wake_up_process(fs_info->transaction_kthread);
5548 case BTRFS_IOC_START_SYNC:
5549 return btrfs_ioctl_start_sync(root, argp);
5550 case BTRFS_IOC_WAIT_SYNC:
5551 return btrfs_ioctl_wait_sync(fs_info, argp);
5552 case BTRFS_IOC_SCRUB:
5553 return btrfs_ioctl_scrub(file, argp);
5554 case BTRFS_IOC_SCRUB_CANCEL:
5555 return btrfs_ioctl_scrub_cancel(fs_info);
5556 case BTRFS_IOC_SCRUB_PROGRESS:
5557 return btrfs_ioctl_scrub_progress(fs_info, argp);
5558 case BTRFS_IOC_BALANCE_V2:
5559 return btrfs_ioctl_balance(file, argp);
5560 case BTRFS_IOC_BALANCE_CTL:
5561 return btrfs_ioctl_balance_ctl(fs_info, arg);
5562 case BTRFS_IOC_BALANCE_PROGRESS:
5563 return btrfs_ioctl_balance_progress(fs_info, argp);
5564 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5565 return btrfs_ioctl_set_received_subvol(file, argp);
5567 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5568 return btrfs_ioctl_set_received_subvol_32(file, argp);
5570 case BTRFS_IOC_SEND:
5571 return btrfs_ioctl_send(file, argp);
5572 case BTRFS_IOC_GET_DEV_STATS:
5573 return btrfs_ioctl_get_dev_stats(fs_info, argp);
5574 case BTRFS_IOC_QUOTA_CTL:
5575 return btrfs_ioctl_quota_ctl(file, argp);
5576 case BTRFS_IOC_QGROUP_ASSIGN:
5577 return btrfs_ioctl_qgroup_assign(file, argp);
5578 case BTRFS_IOC_QGROUP_CREATE:
5579 return btrfs_ioctl_qgroup_create(file, argp);
5580 case BTRFS_IOC_QGROUP_LIMIT:
5581 return btrfs_ioctl_qgroup_limit(file, argp);
5582 case BTRFS_IOC_QUOTA_RESCAN:
5583 return btrfs_ioctl_quota_rescan(file, argp);
5584 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5585 return btrfs_ioctl_quota_rescan_status(file, argp);
5586 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5587 return btrfs_ioctl_quota_rescan_wait(file, argp);
5588 case BTRFS_IOC_DEV_REPLACE:
5589 return btrfs_ioctl_dev_replace(fs_info, argp);
5590 case BTRFS_IOC_GET_FSLABEL:
5591 return btrfs_ioctl_get_fslabel(file, argp);
5592 case BTRFS_IOC_SET_FSLABEL:
5593 return btrfs_ioctl_set_fslabel(file, argp);
5594 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5595 return btrfs_ioctl_get_supported_features(argp);
5596 case BTRFS_IOC_GET_FEATURES:
5597 return btrfs_ioctl_get_features(file, argp);
5598 case BTRFS_IOC_SET_FEATURES:
5599 return btrfs_ioctl_set_features(file, argp);
5605 #ifdef CONFIG_COMPAT
5606 long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
5609 * These all access 32-bit values anyway so no further
5610 * handling is necessary.
5613 case FS_IOC32_GETFLAGS:
5614 cmd = FS_IOC_GETFLAGS;
5616 case FS_IOC32_SETFLAGS:
5617 cmd = FS_IOC_SETFLAGS;
5619 case FS_IOC32_GETVERSION:
5620 cmd = FS_IOC_GETVERSION;
5624 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));