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;
357 if (!capable(CAP_SYS_ADMIN))
361 * If the fs is mounted with nologreplay, which requires it to be
362 * mounted in RO mode as well, we can not allow discard on free space
363 * inside block groups, because log trees refer to extents that are not
364 * pinned in a block group's free space cache (pinning the extents is
365 * precisely the first phase of replaying a log tree).
367 if (btrfs_test_opt(fs_info, NOLOGREPLAY))
371 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
375 q = bdev_get_queue(device->bdev);
376 if (blk_queue_discard(q)) {
378 minlen = min_t(u64, q->limits.discard_granularity,
386 if (copy_from_user(&range, arg, sizeof(range)))
390 * NOTE: Don't truncate the range using super->total_bytes. Bytenr of
391 * block group is in the logical address space, which can be any
392 * sectorsize aligned bytenr in the range [0, U64_MAX].
394 if (range.len < fs_info->sb->s_blocksize)
397 range.minlen = max(range.minlen, minlen);
398 ret = btrfs_trim_fs(fs_info, &range);
402 if (copy_to_user(arg, &range, sizeof(range)))
408 int btrfs_is_empty_uuid(u8 *uuid)
412 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
419 static noinline int create_subvol(struct inode *dir,
420 struct dentry *dentry,
421 const char *name, int namelen,
423 struct btrfs_qgroup_inherit *inherit)
425 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
426 struct btrfs_trans_handle *trans;
427 struct btrfs_key key;
428 struct btrfs_root_item *root_item;
429 struct btrfs_inode_item *inode_item;
430 struct extent_buffer *leaf;
431 struct btrfs_root *root = BTRFS_I(dir)->root;
432 struct btrfs_root *new_root;
433 struct btrfs_block_rsv block_rsv;
434 struct timespec cur_time = current_time(dir);
439 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
444 root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
448 ret = btrfs_find_free_objectid(fs_info->tree_root, &objectid);
453 * Don't create subvolume whose level is not zero. Or qgroup will be
454 * screwed up since it assumes subvolume qgroup's level to be 0.
456 if (btrfs_qgroup_level(objectid)) {
461 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
463 * The same as the snapshot creation, please see the comment
464 * of create_snapshot().
466 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
467 8, &qgroup_reserved, false);
471 trans = btrfs_start_transaction(root, 0);
473 ret = PTR_ERR(trans);
474 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
477 trans->block_rsv = &block_rsv;
478 trans->bytes_reserved = block_rsv.size;
480 ret = btrfs_qgroup_inherit(trans, fs_info, 0, objectid, inherit);
484 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
490 memzero_extent_buffer(leaf, 0, sizeof(struct btrfs_header));
491 btrfs_set_header_bytenr(leaf, leaf->start);
492 btrfs_set_header_generation(leaf, trans->transid);
493 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
494 btrfs_set_header_owner(leaf, objectid);
496 write_extent_buffer_fsid(leaf, fs_info->fsid);
497 write_extent_buffer_chunk_tree_uuid(leaf, fs_info->chunk_tree_uuid);
498 btrfs_mark_buffer_dirty(leaf);
500 inode_item = &root_item->inode;
501 btrfs_set_stack_inode_generation(inode_item, 1);
502 btrfs_set_stack_inode_size(inode_item, 3);
503 btrfs_set_stack_inode_nlink(inode_item, 1);
504 btrfs_set_stack_inode_nbytes(inode_item,
506 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
508 btrfs_set_root_flags(root_item, 0);
509 btrfs_set_root_limit(root_item, 0);
510 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
512 btrfs_set_root_bytenr(root_item, leaf->start);
513 btrfs_set_root_generation(root_item, trans->transid);
514 btrfs_set_root_level(root_item, 0);
515 btrfs_set_root_refs(root_item, 1);
516 btrfs_set_root_used(root_item, leaf->len);
517 btrfs_set_root_last_snapshot(root_item, 0);
519 btrfs_set_root_generation_v2(root_item,
520 btrfs_root_generation(root_item));
521 uuid_le_gen(&new_uuid);
522 memcpy(root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE);
523 btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
524 btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
525 root_item->ctime = root_item->otime;
526 btrfs_set_root_ctransid(root_item, trans->transid);
527 btrfs_set_root_otransid(root_item, trans->transid);
529 btrfs_tree_unlock(leaf);
531 btrfs_set_root_dirid(root_item, new_dirid);
533 key.objectid = objectid;
535 key.type = BTRFS_ROOT_ITEM_KEY;
536 ret = btrfs_insert_root(trans, fs_info->tree_root, &key,
540 * Since we don't abort the transaction in this case, free the
541 * tree block so that we don't leak space and leave the
542 * filesystem in an inconsistent state (an extent item in the
543 * extent tree without backreferences). Also no need to have
544 * the tree block locked since it is not in any tree at this
545 * point, so no other task can find it and use it.
547 btrfs_free_tree_block(trans, root, leaf, 0, 1);
548 free_extent_buffer(leaf);
552 free_extent_buffer(leaf);
555 key.offset = (u64)-1;
556 new_root = btrfs_read_fs_root_no_name(fs_info, &key);
557 if (IS_ERR(new_root)) {
558 ret = PTR_ERR(new_root);
559 btrfs_abort_transaction(trans, ret);
563 btrfs_record_root_in_trans(trans, new_root);
565 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
567 /* We potentially lose an unused inode item here */
568 btrfs_abort_transaction(trans, ret);
572 mutex_lock(&new_root->objectid_mutex);
573 new_root->highest_objectid = new_dirid;
574 mutex_unlock(&new_root->objectid_mutex);
577 * insert the directory item
579 ret = btrfs_set_inode_index(BTRFS_I(dir), &index);
581 btrfs_abort_transaction(trans, ret);
585 ret = btrfs_insert_dir_item(trans, root,
586 name, namelen, BTRFS_I(dir), &key,
587 BTRFS_FT_DIR, index);
589 btrfs_abort_transaction(trans, ret);
593 btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2);
594 ret = btrfs_update_inode(trans, root, dir);
596 btrfs_abort_transaction(trans, ret);
600 ret = btrfs_add_root_ref(trans, fs_info,
601 objectid, root->root_key.objectid,
602 btrfs_ino(BTRFS_I(dir)), index, name, namelen);
604 btrfs_abort_transaction(trans, ret);
608 ret = btrfs_uuid_tree_add(trans, fs_info, root_item->uuid,
609 BTRFS_UUID_KEY_SUBVOL, objectid);
611 btrfs_abort_transaction(trans, ret);
615 trans->block_rsv = NULL;
616 trans->bytes_reserved = 0;
617 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
620 *async_transid = trans->transid;
621 err = btrfs_commit_transaction_async(trans, 1);
623 err = btrfs_commit_transaction(trans);
625 err = btrfs_commit_transaction(trans);
631 inode = btrfs_lookup_dentry(dir, dentry);
633 return PTR_ERR(inode);
634 d_instantiate(dentry, inode);
643 static void btrfs_wait_for_no_snapshotting_writes(struct btrfs_root *root)
649 prepare_to_wait(&root->subv_writers->wait, &wait,
650 TASK_UNINTERRUPTIBLE);
652 writers = percpu_counter_sum(&root->subv_writers->counter);
656 finish_wait(&root->subv_writers->wait, &wait);
660 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
661 struct dentry *dentry,
662 u64 *async_transid, bool readonly,
663 struct btrfs_qgroup_inherit *inherit)
665 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
667 struct btrfs_pending_snapshot *pending_snapshot;
668 struct btrfs_trans_handle *trans;
670 bool snapshot_force_cow = false;
672 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
675 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
676 if (!pending_snapshot)
679 pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
681 pending_snapshot->path = btrfs_alloc_path();
682 if (!pending_snapshot->root_item || !pending_snapshot->path) {
688 * Force new buffered writes to reserve space even when NOCOW is
689 * possible. This is to avoid later writeback (running dealloc) to
690 * fallback to COW mode and unexpectedly fail with ENOSPC.
692 atomic_inc(&root->will_be_snapshotted);
693 smp_mb__after_atomic();
694 btrfs_wait_for_no_snapshotting_writes(root);
696 ret = btrfs_start_delalloc_inodes(root, 0);
701 * All previous writes have started writeback in NOCOW mode, so now
702 * we force future writes to fallback to COW mode during snapshot
705 atomic_inc(&root->snapshot_force_cow);
706 snapshot_force_cow = true;
708 btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1);
710 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
711 BTRFS_BLOCK_RSV_TEMP);
713 * 1 - parent dir inode
716 * 2 - root ref/backref
717 * 1 - root of snapshot
720 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
721 &pending_snapshot->block_rsv, 8,
722 &pending_snapshot->qgroup_reserved,
727 pending_snapshot->dentry = dentry;
728 pending_snapshot->root = root;
729 pending_snapshot->readonly = readonly;
730 pending_snapshot->dir = dir;
731 pending_snapshot->inherit = inherit;
733 trans = btrfs_start_transaction(root, 0);
735 ret = PTR_ERR(trans);
739 spin_lock(&fs_info->trans_lock);
740 list_add(&pending_snapshot->list,
741 &trans->transaction->pending_snapshots);
742 spin_unlock(&fs_info->trans_lock);
744 *async_transid = trans->transid;
745 ret = btrfs_commit_transaction_async(trans, 1);
747 ret = btrfs_commit_transaction(trans);
749 ret = btrfs_commit_transaction(trans);
754 ret = pending_snapshot->error;
758 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
762 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
764 ret = PTR_ERR(inode);
768 d_instantiate(dentry, inode);
771 btrfs_subvolume_release_metadata(fs_info, &pending_snapshot->block_rsv);
773 if (snapshot_force_cow)
774 atomic_dec(&root->snapshot_force_cow);
775 if (atomic_dec_and_test(&root->will_be_snapshotted))
776 wake_up_atomic_t(&root->will_be_snapshotted);
778 kfree(pending_snapshot->root_item);
779 btrfs_free_path(pending_snapshot->path);
780 kfree(pending_snapshot);
785 /* copy of may_delete in fs/namei.c()
786 * Check whether we can remove a link victim from directory dir, check
787 * whether the type of victim is right.
788 * 1. We can't do it if dir is read-only (done in permission())
789 * 2. We should have write and exec permissions on dir
790 * 3. We can't remove anything from append-only dir
791 * 4. We can't do anything with immutable dir (done in permission())
792 * 5. If the sticky bit on dir is set we should either
793 * a. be owner of dir, or
794 * b. be owner of victim, or
795 * c. have CAP_FOWNER capability
796 * 6. If the victim is append-only or immutable we can't do anything with
797 * links pointing to it.
798 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
799 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
800 * 9. We can't remove a root or mountpoint.
801 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
802 * nfs_async_unlink().
805 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
809 if (d_really_is_negative(victim))
812 BUG_ON(d_inode(victim->d_parent) != dir);
813 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
815 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
820 if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) ||
821 IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim)))
824 if (!d_is_dir(victim))
828 } else if (d_is_dir(victim))
832 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
837 /* copy of may_create in fs/namei.c() */
838 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
840 if (d_really_is_positive(child))
844 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
848 * Create a new subvolume below @parent. This is largely modeled after
849 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
850 * inside this filesystem so it's quite a bit simpler.
852 static noinline int btrfs_mksubvol(const struct path *parent,
853 const char *name, int namelen,
854 struct btrfs_root *snap_src,
855 u64 *async_transid, bool readonly,
856 struct btrfs_qgroup_inherit *inherit)
858 struct inode *dir = d_inode(parent->dentry);
859 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
860 struct dentry *dentry;
863 error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
867 dentry = lookup_one_len(name, parent->dentry, namelen);
868 error = PTR_ERR(dentry);
872 error = btrfs_may_create(dir, dentry);
877 * even if this name doesn't exist, we may get hash collisions.
878 * check for them now when we can safely fail
880 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
886 down_read(&fs_info->subvol_sem);
888 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
892 error = create_snapshot(snap_src, dir, dentry,
893 async_transid, readonly, inherit);
895 error = create_subvol(dir, dentry, name, namelen,
896 async_transid, inherit);
899 fsnotify_mkdir(dir, dentry);
901 up_read(&fs_info->subvol_sem);
910 * When we're defragging a range, we don't want to kick it off again
911 * if it is really just waiting for delalloc to send it down.
912 * If we find a nice big extent or delalloc range for the bytes in the
913 * file you want to defrag, we return 0 to let you know to skip this
916 static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
918 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
919 struct extent_map *em = NULL;
920 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
923 read_lock(&em_tree->lock);
924 em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE);
925 read_unlock(&em_tree->lock);
928 end = extent_map_end(em);
930 if (end - offset > thresh)
933 /* if we already have a nice delalloc here, just stop */
935 end = count_range_bits(io_tree, &offset, offset + thresh,
936 thresh, EXTENT_DELALLOC, 1);
943 * helper function to walk through a file and find extents
944 * newer than a specific transid, and smaller than thresh.
946 * This is used by the defragging code to find new and small
949 static int find_new_extents(struct btrfs_root *root,
950 struct inode *inode, u64 newer_than,
951 u64 *off, u32 thresh)
953 struct btrfs_path *path;
954 struct btrfs_key min_key;
955 struct extent_buffer *leaf;
956 struct btrfs_file_extent_item *extent;
959 u64 ino = btrfs_ino(BTRFS_I(inode));
961 path = btrfs_alloc_path();
965 min_key.objectid = ino;
966 min_key.type = BTRFS_EXTENT_DATA_KEY;
967 min_key.offset = *off;
970 ret = btrfs_search_forward(root, &min_key, path, newer_than);
974 if (min_key.objectid != ino)
976 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
979 leaf = path->nodes[0];
980 extent = btrfs_item_ptr(leaf, path->slots[0],
981 struct btrfs_file_extent_item);
983 type = btrfs_file_extent_type(leaf, extent);
984 if (type == BTRFS_FILE_EXTENT_REG &&
985 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
986 check_defrag_in_cache(inode, min_key.offset, thresh)) {
987 *off = min_key.offset;
988 btrfs_free_path(path);
993 if (path->slots[0] < btrfs_header_nritems(leaf)) {
994 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
998 if (min_key.offset == (u64)-1)
1002 btrfs_release_path(path);
1005 btrfs_free_path(path);
1009 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
1011 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1012 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1013 struct extent_map *em;
1014 u64 len = PAGE_SIZE;
1017 * hopefully we have this extent in the tree already, try without
1018 * the full extent lock
1020 read_lock(&em_tree->lock);
1021 em = lookup_extent_mapping(em_tree, start, len);
1022 read_unlock(&em_tree->lock);
1025 struct extent_state *cached = NULL;
1026 u64 end = start + len - 1;
1028 /* get the big lock and read metadata off disk */
1029 lock_extent_bits(io_tree, start, end, &cached);
1030 em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len, 0);
1031 unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS);
1040 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1042 struct extent_map *next;
1045 /* this is the last extent */
1046 if (em->start + em->len >= i_size_read(inode))
1049 next = defrag_lookup_extent(inode, em->start + em->len);
1050 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1052 else if ((em->block_start + em->block_len == next->block_start) &&
1053 (em->block_len > SZ_128K && next->block_len > SZ_128K))
1056 free_extent_map(next);
1060 static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
1061 u64 *last_len, u64 *skip, u64 *defrag_end,
1064 struct extent_map *em;
1066 bool next_mergeable = true;
1067 bool prev_mergeable = true;
1070 * make sure that once we start defragging an extent, we keep on
1073 if (start < *defrag_end)
1078 em = defrag_lookup_extent(inode, start);
1082 /* this will cover holes, and inline extents */
1083 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1089 prev_mergeable = false;
1091 next_mergeable = defrag_check_next_extent(inode, em);
1093 * we hit a real extent, if it is big or the next extent is not a
1094 * real extent, don't bother defragging it
1096 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1097 (em->len >= thresh || (!next_mergeable && !prev_mergeable)))
1101 * last_len ends up being a counter of how many bytes we've defragged.
1102 * every time we choose not to defrag an extent, we reset *last_len
1103 * so that the next tiny extent will force a defrag.
1105 * The end result of this is that tiny extents before a single big
1106 * extent will force at least part of that big extent to be defragged.
1109 *defrag_end = extent_map_end(em);
1112 *skip = extent_map_end(em);
1116 free_extent_map(em);
1121 * it doesn't do much good to defrag one or two pages
1122 * at a time. This pulls in a nice chunk of pages
1123 * to COW and defrag.
1125 * It also makes sure the delalloc code has enough
1126 * dirty data to avoid making new small extents as part
1129 * It's a good idea to start RA on this range
1130 * before calling this.
1132 static int cluster_pages_for_defrag(struct inode *inode,
1133 struct page **pages,
1134 unsigned long start_index,
1135 unsigned long num_pages)
1137 unsigned long file_end;
1138 u64 isize = i_size_read(inode);
1145 struct btrfs_ordered_extent *ordered;
1146 struct extent_state *cached_state = NULL;
1147 struct extent_io_tree *tree;
1148 struct extent_changeset *data_reserved = NULL;
1149 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1151 file_end = (isize - 1) >> PAGE_SHIFT;
1152 if (!isize || start_index > file_end)
1155 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1157 ret = btrfs_delalloc_reserve_space(inode, &data_reserved,
1158 start_index << PAGE_SHIFT,
1159 page_cnt << PAGE_SHIFT);
1163 tree = &BTRFS_I(inode)->io_tree;
1165 /* step one, lock all the pages */
1166 for (i = 0; i < page_cnt; i++) {
1169 page = find_or_create_page(inode->i_mapping,
1170 start_index + i, mask);
1174 page_start = page_offset(page);
1175 page_end = page_start + PAGE_SIZE - 1;
1177 lock_extent_bits(tree, page_start, page_end,
1179 ordered = btrfs_lookup_ordered_extent(inode,
1181 unlock_extent_cached(tree, page_start, page_end,
1182 &cached_state, GFP_NOFS);
1187 btrfs_start_ordered_extent(inode, ordered, 1);
1188 btrfs_put_ordered_extent(ordered);
1191 * we unlocked the page above, so we need check if
1192 * it was released or not.
1194 if (page->mapping != inode->i_mapping) {
1201 if (!PageUptodate(page)) {
1202 btrfs_readpage(NULL, page);
1204 if (!PageUptodate(page)) {
1212 if (page->mapping != inode->i_mapping) {
1224 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1228 * so now we have a nice long stream of locked
1229 * and up to date pages, lets wait on them
1231 for (i = 0; i < i_done; i++)
1232 wait_on_page_writeback(pages[i]);
1234 page_start = page_offset(pages[0]);
1235 page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE;
1237 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1238 page_start, page_end - 1, &cached_state);
1239 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1240 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1241 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1242 &cached_state, GFP_NOFS);
1244 if (i_done != page_cnt) {
1245 spin_lock(&BTRFS_I(inode)->lock);
1246 BTRFS_I(inode)->outstanding_extents++;
1247 spin_unlock(&BTRFS_I(inode)->lock);
1248 btrfs_delalloc_release_space(inode, data_reserved,
1249 start_index << PAGE_SHIFT,
1250 (page_cnt - i_done) << PAGE_SHIFT);
1254 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1257 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1258 page_start, page_end - 1, &cached_state,
1261 for (i = 0; i < i_done; i++) {
1262 clear_page_dirty_for_io(pages[i]);
1263 ClearPageChecked(pages[i]);
1264 set_page_extent_mapped(pages[i]);
1265 set_page_dirty(pages[i]);
1266 unlock_page(pages[i]);
1269 extent_changeset_free(data_reserved);
1272 for (i = 0; i < i_done; i++) {
1273 unlock_page(pages[i]);
1276 btrfs_delalloc_release_space(inode, data_reserved,
1277 start_index << PAGE_SHIFT,
1278 page_cnt << PAGE_SHIFT);
1279 extent_changeset_free(data_reserved);
1284 int btrfs_defrag_file(struct inode *inode, struct file *file,
1285 struct btrfs_ioctl_defrag_range_args *range,
1286 u64 newer_than, unsigned long max_to_defrag)
1288 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1289 struct btrfs_root *root = BTRFS_I(inode)->root;
1290 struct file_ra_state *ra = NULL;
1291 unsigned long last_index;
1292 u64 isize = i_size_read(inode);
1296 u64 newer_off = range->start;
1298 unsigned long ra_index = 0;
1300 int defrag_count = 0;
1301 int compress_type = BTRFS_COMPRESS_ZLIB;
1302 u32 extent_thresh = range->extent_thresh;
1303 unsigned long max_cluster = SZ_256K >> PAGE_SHIFT;
1304 unsigned long cluster = max_cluster;
1305 u64 new_align = ~((u64)SZ_128K - 1);
1306 struct page **pages = NULL;
1307 bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS;
1312 if (range->start >= isize)
1316 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1318 if (range->compress_type)
1319 compress_type = range->compress_type;
1322 if (extent_thresh == 0)
1323 extent_thresh = SZ_256K;
1326 * If we were not given a file, allocate a readahead context. As
1327 * readahead is just an optimization, defrag will work without it so
1328 * we don't error out.
1331 ra = kzalloc(sizeof(*ra), GFP_KERNEL);
1333 file_ra_state_init(ra, inode->i_mapping);
1338 pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL);
1344 /* find the last page to defrag */
1345 if (range->start + range->len > range->start) {
1346 last_index = min_t(u64, isize - 1,
1347 range->start + range->len - 1) >> PAGE_SHIFT;
1349 last_index = (isize - 1) >> PAGE_SHIFT;
1353 ret = find_new_extents(root, inode, newer_than,
1354 &newer_off, SZ_64K);
1356 range->start = newer_off;
1358 * we always align our defrag to help keep
1359 * the extents in the file evenly spaced
1361 i = (newer_off & new_align) >> PAGE_SHIFT;
1365 i = range->start >> PAGE_SHIFT;
1368 max_to_defrag = last_index - i + 1;
1371 * make writeback starts from i, so the defrag range can be
1372 * written sequentially.
1374 if (i < inode->i_mapping->writeback_index)
1375 inode->i_mapping->writeback_index = i;
1377 while (i <= last_index && defrag_count < max_to_defrag &&
1378 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) {
1380 * make sure we stop running if someone unmounts
1383 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1386 if (btrfs_defrag_cancelled(fs_info)) {
1387 btrfs_debug(fs_info, "defrag_file cancelled");
1392 if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT,
1393 extent_thresh, &last_len, &skip,
1394 &defrag_end, do_compress)){
1397 * the should_defrag function tells us how much to skip
1398 * bump our counter by the suggested amount
1400 next = DIV_ROUND_UP(skip, PAGE_SIZE);
1401 i = max(i + 1, next);
1406 cluster = (PAGE_ALIGN(defrag_end) >>
1408 cluster = min(cluster, max_cluster);
1410 cluster = max_cluster;
1413 if (i + cluster > ra_index) {
1414 ra_index = max(i, ra_index);
1416 page_cache_sync_readahead(inode->i_mapping, ra,
1417 file, ra_index, cluster);
1418 ra_index += cluster;
1423 BTRFS_I(inode)->defrag_compress = compress_type;
1424 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1426 inode_unlock(inode);
1430 defrag_count += ret;
1431 balance_dirty_pages_ratelimited(inode->i_mapping);
1432 inode_unlock(inode);
1435 if (newer_off == (u64)-1)
1441 newer_off = max(newer_off + 1,
1442 (u64)i << PAGE_SHIFT);
1444 ret = find_new_extents(root, inode, newer_than,
1445 &newer_off, SZ_64K);
1447 range->start = newer_off;
1448 i = (newer_off & new_align) >> PAGE_SHIFT;
1455 last_len += ret << PAGE_SHIFT;
1463 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1464 filemap_flush(inode->i_mapping);
1465 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1466 &BTRFS_I(inode)->runtime_flags))
1467 filemap_flush(inode->i_mapping);
1471 /* the filemap_flush will queue IO into the worker threads, but
1472 * we have to make sure the IO is actually started and that
1473 * ordered extents get created before we return
1475 atomic_inc(&fs_info->async_submit_draining);
1476 while (atomic_read(&fs_info->nr_async_submits) ||
1477 atomic_read(&fs_info->async_delalloc_pages)) {
1478 wait_event(fs_info->async_submit_wait,
1479 (atomic_read(&fs_info->nr_async_submits) == 0 &&
1480 atomic_read(&fs_info->async_delalloc_pages) == 0));
1482 atomic_dec(&fs_info->async_submit_draining);
1485 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1486 btrfs_set_fs_incompat(fs_info, COMPRESS_LZO);
1487 } else if (range->compress_type == BTRFS_COMPRESS_ZSTD) {
1488 btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD);
1496 BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE;
1497 inode_unlock(inode);
1505 static noinline int btrfs_ioctl_resize(struct file *file,
1508 struct inode *inode = file_inode(file);
1509 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1513 struct btrfs_root *root = BTRFS_I(inode)->root;
1514 struct btrfs_ioctl_vol_args *vol_args;
1515 struct btrfs_trans_handle *trans;
1516 struct btrfs_device *device = NULL;
1519 char *devstr = NULL;
1523 if (!capable(CAP_SYS_ADMIN))
1526 ret = mnt_want_write_file(file);
1530 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
1531 mnt_drop_write_file(file);
1532 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1535 mutex_lock(&fs_info->volume_mutex);
1536 vol_args = memdup_user(arg, sizeof(*vol_args));
1537 if (IS_ERR(vol_args)) {
1538 ret = PTR_ERR(vol_args);
1542 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1544 sizestr = vol_args->name;
1545 devstr = strchr(sizestr, ':');
1547 sizestr = devstr + 1;
1549 devstr = vol_args->name;
1550 ret = kstrtoull(devstr, 10, &devid);
1557 btrfs_info(fs_info, "resizing devid %llu", devid);
1560 device = btrfs_find_device(fs_info, devid, NULL, NULL);
1562 btrfs_info(fs_info, "resizer unable to find device %llu",
1568 if (!device->writeable) {
1570 "resizer unable to apply on readonly device %llu",
1576 if (!strcmp(sizestr, "max"))
1577 new_size = device->bdev->bd_inode->i_size;
1579 if (sizestr[0] == '-') {
1582 } else if (sizestr[0] == '+') {
1586 new_size = memparse(sizestr, &retptr);
1587 if (*retptr != '\0' || new_size == 0) {
1593 if (device->is_tgtdev_for_dev_replace) {
1598 old_size = btrfs_device_get_total_bytes(device);
1601 if (new_size > old_size) {
1605 new_size = old_size - new_size;
1606 } else if (mod > 0) {
1607 if (new_size > ULLONG_MAX - old_size) {
1611 new_size = old_size + new_size;
1614 if (new_size < SZ_256M) {
1618 if (new_size > device->bdev->bd_inode->i_size) {
1623 new_size = round_down(new_size, fs_info->sectorsize);
1625 btrfs_info_in_rcu(fs_info, "new size for %s is %llu",
1626 rcu_str_deref(device->name), new_size);
1628 if (new_size > old_size) {
1629 trans = btrfs_start_transaction(root, 0);
1630 if (IS_ERR(trans)) {
1631 ret = PTR_ERR(trans);
1634 ret = btrfs_grow_device(trans, device, new_size);
1635 btrfs_commit_transaction(trans);
1636 } else if (new_size < old_size) {
1637 ret = btrfs_shrink_device(device, new_size);
1638 } /* equal, nothing need to do */
1643 mutex_unlock(&fs_info->volume_mutex);
1644 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
1645 mnt_drop_write_file(file);
1649 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1650 const char *name, unsigned long fd, int subvol,
1651 u64 *transid, bool readonly,
1652 struct btrfs_qgroup_inherit *inherit)
1657 if (!S_ISDIR(file_inode(file)->i_mode))
1660 ret = mnt_want_write_file(file);
1664 namelen = strlen(name);
1665 if (strchr(name, '/')) {
1667 goto out_drop_write;
1670 if (name[0] == '.' &&
1671 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1673 goto out_drop_write;
1677 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1678 NULL, transid, readonly, inherit);
1680 struct fd src = fdget(fd);
1681 struct inode *src_inode;
1684 goto out_drop_write;
1687 src_inode = file_inode(src.file);
1688 if (src_inode->i_sb != file_inode(file)->i_sb) {
1689 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
1690 "Snapshot src from another FS");
1692 } else if (!inode_owner_or_capable(src_inode)) {
1694 * Subvolume creation is not restricted, but snapshots
1695 * are limited to own subvolumes only
1699 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1700 BTRFS_I(src_inode)->root,
1701 transid, readonly, inherit);
1706 mnt_drop_write_file(file);
1711 static noinline int btrfs_ioctl_snap_create(struct file *file,
1712 void __user *arg, int subvol)
1714 struct btrfs_ioctl_vol_args *vol_args;
1717 if (!S_ISDIR(file_inode(file)->i_mode))
1720 vol_args = memdup_user(arg, sizeof(*vol_args));
1721 if (IS_ERR(vol_args))
1722 return PTR_ERR(vol_args);
1723 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1725 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1726 vol_args->fd, subvol,
1733 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1734 void __user *arg, int subvol)
1736 struct btrfs_ioctl_vol_args_v2 *vol_args;
1740 bool readonly = false;
1741 struct btrfs_qgroup_inherit *inherit = NULL;
1743 if (!S_ISDIR(file_inode(file)->i_mode))
1746 vol_args = memdup_user(arg, sizeof(*vol_args));
1747 if (IS_ERR(vol_args))
1748 return PTR_ERR(vol_args);
1749 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1751 if (vol_args->flags &
1752 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1753 BTRFS_SUBVOL_QGROUP_INHERIT)) {
1758 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1760 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1762 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1763 if (vol_args->size > PAGE_SIZE) {
1767 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1768 if (IS_ERR(inherit)) {
1769 ret = PTR_ERR(inherit);
1774 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1775 vol_args->fd, subvol, ptr,
1780 if (ptr && copy_to_user(arg +
1781 offsetof(struct btrfs_ioctl_vol_args_v2,
1793 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1796 struct inode *inode = file_inode(file);
1797 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1798 struct btrfs_root *root = BTRFS_I(inode)->root;
1802 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID)
1805 down_read(&fs_info->subvol_sem);
1806 if (btrfs_root_readonly(root))
1807 flags |= BTRFS_SUBVOL_RDONLY;
1808 up_read(&fs_info->subvol_sem);
1810 if (copy_to_user(arg, &flags, sizeof(flags)))
1816 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1819 struct inode *inode = file_inode(file);
1820 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1821 struct btrfs_root *root = BTRFS_I(inode)->root;
1822 struct btrfs_trans_handle *trans;
1827 if (!inode_owner_or_capable(inode))
1830 ret = mnt_want_write_file(file);
1834 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1836 goto out_drop_write;
1839 if (copy_from_user(&flags, arg, sizeof(flags))) {
1841 goto out_drop_write;
1844 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1846 goto out_drop_write;
1849 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1851 goto out_drop_write;
1854 down_write(&fs_info->subvol_sem);
1857 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1860 root_flags = btrfs_root_flags(&root->root_item);
1861 if (flags & BTRFS_SUBVOL_RDONLY) {
1862 btrfs_set_root_flags(&root->root_item,
1863 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1866 * Block RO -> RW transition if this subvolume is involved in
1869 spin_lock(&root->root_item_lock);
1870 if (root->send_in_progress == 0) {
1871 btrfs_set_root_flags(&root->root_item,
1872 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1873 spin_unlock(&root->root_item_lock);
1875 spin_unlock(&root->root_item_lock);
1877 "Attempt to set subvolume %llu read-write during send",
1878 root->root_key.objectid);
1884 trans = btrfs_start_transaction(root, 1);
1885 if (IS_ERR(trans)) {
1886 ret = PTR_ERR(trans);
1890 ret = btrfs_update_root(trans, fs_info->tree_root,
1891 &root->root_key, &root->root_item);
1893 btrfs_end_transaction(trans);
1897 ret = btrfs_commit_transaction(trans);
1901 btrfs_set_root_flags(&root->root_item, root_flags);
1903 up_write(&fs_info->subvol_sem);
1905 mnt_drop_write_file(file);
1911 * helper to check if the subvolume references other subvolumes
1913 static noinline int may_destroy_subvol(struct btrfs_root *root)
1915 struct btrfs_fs_info *fs_info = root->fs_info;
1916 struct btrfs_path *path;
1917 struct btrfs_dir_item *di;
1918 struct btrfs_key key;
1922 path = btrfs_alloc_path();
1926 /* Make sure this root isn't set as the default subvol */
1927 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1928 di = btrfs_lookup_dir_item(NULL, fs_info->tree_root, path,
1929 dir_id, "default", 7, 0);
1930 if (di && !IS_ERR(di)) {
1931 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1932 if (key.objectid == root->root_key.objectid) {
1935 "deleting default subvolume %llu is not allowed",
1939 btrfs_release_path(path);
1942 key.objectid = root->root_key.objectid;
1943 key.type = BTRFS_ROOT_REF_KEY;
1944 key.offset = (u64)-1;
1946 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
1952 if (path->slots[0] > 0) {
1954 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1955 if (key.objectid == root->root_key.objectid &&
1956 key.type == BTRFS_ROOT_REF_KEY)
1960 btrfs_free_path(path);
1964 static noinline int key_in_sk(struct btrfs_key *key,
1965 struct btrfs_ioctl_search_key *sk)
1967 struct btrfs_key test;
1970 test.objectid = sk->min_objectid;
1971 test.type = sk->min_type;
1972 test.offset = sk->min_offset;
1974 ret = btrfs_comp_cpu_keys(key, &test);
1978 test.objectid = sk->max_objectid;
1979 test.type = sk->max_type;
1980 test.offset = sk->max_offset;
1982 ret = btrfs_comp_cpu_keys(key, &test);
1988 static noinline int copy_to_sk(struct btrfs_path *path,
1989 struct btrfs_key *key,
1990 struct btrfs_ioctl_search_key *sk,
1993 unsigned long *sk_offset,
1997 struct extent_buffer *leaf;
1998 struct btrfs_ioctl_search_header sh;
1999 struct btrfs_key test;
2000 unsigned long item_off;
2001 unsigned long item_len;
2007 leaf = path->nodes[0];
2008 slot = path->slots[0];
2009 nritems = btrfs_header_nritems(leaf);
2011 if (btrfs_header_generation(leaf) > sk->max_transid) {
2015 found_transid = btrfs_header_generation(leaf);
2017 for (i = slot; i < nritems; i++) {
2018 item_off = btrfs_item_ptr_offset(leaf, i);
2019 item_len = btrfs_item_size_nr(leaf, i);
2021 btrfs_item_key_to_cpu(leaf, key, i);
2022 if (!key_in_sk(key, sk))
2025 if (sizeof(sh) + item_len > *buf_size) {
2032 * return one empty item back for v1, which does not
2036 *buf_size = sizeof(sh) + item_len;
2041 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
2046 sh.objectid = key->objectid;
2047 sh.offset = key->offset;
2048 sh.type = key->type;
2050 sh.transid = found_transid;
2053 * Copy search result header. If we fault then loop again so we
2054 * can fault in the pages and -EFAULT there if there's a
2055 * problem. Otherwise we'll fault and then copy the buffer in
2056 * properly this next time through
2058 if (probe_user_write(ubuf + *sk_offset, &sh, sizeof(sh))) {
2063 *sk_offset += sizeof(sh);
2066 char __user *up = ubuf + *sk_offset;
2068 * Copy the item, same behavior as above, but reset the
2069 * * sk_offset so we copy the full thing again.
2071 if (read_extent_buffer_to_user_nofault(leaf, up,
2072 item_off, item_len)) {
2074 *sk_offset -= sizeof(sh);
2078 *sk_offset += item_len;
2082 if (ret) /* -EOVERFLOW from above */
2085 if (*num_found >= sk->nr_items) {
2092 test.objectid = sk->max_objectid;
2093 test.type = sk->max_type;
2094 test.offset = sk->max_offset;
2095 if (btrfs_comp_cpu_keys(key, &test) >= 0)
2097 else if (key->offset < (u64)-1)
2099 else if (key->type < (u8)-1) {
2102 } else if (key->objectid < (u64)-1) {
2110 * 0: all items from this leaf copied, continue with next
2111 * 1: * more items can be copied, but unused buffer is too small
2112 * * all items were found
2113 * Either way, it will stops the loop which iterates to the next
2115 * -EOVERFLOW: item was to large for buffer
2116 * -EFAULT: could not copy extent buffer back to userspace
2121 static noinline int search_ioctl(struct inode *inode,
2122 struct btrfs_ioctl_search_key *sk,
2126 struct btrfs_fs_info *info = btrfs_sb(inode->i_sb);
2127 struct btrfs_root *root;
2128 struct btrfs_key key;
2129 struct btrfs_path *path;
2132 unsigned long sk_offset = 0;
2134 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2135 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2139 path = btrfs_alloc_path();
2143 if (sk->tree_id == 0) {
2144 /* search the root of the inode that was passed */
2145 root = BTRFS_I(inode)->root;
2147 key.objectid = sk->tree_id;
2148 key.type = BTRFS_ROOT_ITEM_KEY;
2149 key.offset = (u64)-1;
2150 root = btrfs_read_fs_root_no_name(info, &key);
2152 btrfs_free_path(path);
2157 key.objectid = sk->min_objectid;
2158 key.type = sk->min_type;
2159 key.offset = sk->min_offset;
2162 ret = fault_in_pages_writeable(ubuf + sk_offset,
2163 *buf_size - sk_offset);
2167 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2173 ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
2174 &sk_offset, &num_found);
2175 btrfs_release_path(path);
2183 sk->nr_items = num_found;
2184 btrfs_free_path(path);
2188 static noinline int btrfs_ioctl_tree_search(struct file *file,
2191 struct btrfs_ioctl_search_args __user *uargs;
2192 struct btrfs_ioctl_search_key sk;
2193 struct inode *inode;
2197 if (!capable(CAP_SYS_ADMIN))
2200 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2202 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2205 buf_size = sizeof(uargs->buf);
2207 inode = file_inode(file);
2208 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2211 * In the origin implementation an overflow is handled by returning a
2212 * search header with a len of zero, so reset ret.
2214 if (ret == -EOVERFLOW)
2217 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2222 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2225 struct btrfs_ioctl_search_args_v2 __user *uarg;
2226 struct btrfs_ioctl_search_args_v2 args;
2227 struct inode *inode;
2230 const size_t buf_limit = SZ_16M;
2232 if (!capable(CAP_SYS_ADMIN))
2235 /* copy search header and buffer size */
2236 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2237 if (copy_from_user(&args, uarg, sizeof(args)))
2240 buf_size = args.buf_size;
2242 /* limit result size to 16MB */
2243 if (buf_size > buf_limit)
2244 buf_size = buf_limit;
2246 inode = file_inode(file);
2247 ret = search_ioctl(inode, &args.key, &buf_size,
2248 (char *)(&uarg->buf[0]));
2249 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2251 else if (ret == -EOVERFLOW &&
2252 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2259 * Search INODE_REFs to identify path name of 'dirid' directory
2260 * in a 'tree_id' tree. and sets path name to 'name'.
2262 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2263 u64 tree_id, u64 dirid, char *name)
2265 struct btrfs_root *root;
2266 struct btrfs_key key;
2272 struct btrfs_inode_ref *iref;
2273 struct extent_buffer *l;
2274 struct btrfs_path *path;
2276 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2281 path = btrfs_alloc_path();
2285 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1];
2287 key.objectid = tree_id;
2288 key.type = BTRFS_ROOT_ITEM_KEY;
2289 key.offset = (u64)-1;
2290 root = btrfs_read_fs_root_no_name(info, &key);
2292 btrfs_err(info, "could not find root %llu", tree_id);
2297 key.objectid = dirid;
2298 key.type = BTRFS_INODE_REF_KEY;
2299 key.offset = (u64)-1;
2302 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2306 ret = btrfs_previous_item(root, path, dirid,
2307 BTRFS_INODE_REF_KEY);
2317 slot = path->slots[0];
2318 btrfs_item_key_to_cpu(l, &key, slot);
2320 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2321 len = btrfs_inode_ref_name_len(l, iref);
2323 total_len += len + 1;
2325 ret = -ENAMETOOLONG;
2330 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2332 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2335 btrfs_release_path(path);
2336 key.objectid = key.offset;
2337 key.offset = (u64)-1;
2338 dirid = key.objectid;
2340 memmove(name, ptr, total_len);
2341 name[total_len] = '\0';
2344 btrfs_free_path(path);
2348 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2351 struct btrfs_ioctl_ino_lookup_args *args;
2352 struct inode *inode;
2355 args = memdup_user(argp, sizeof(*args));
2357 return PTR_ERR(args);
2359 inode = file_inode(file);
2362 * Unprivileged query to obtain the containing subvolume root id. The
2363 * path is reset so it's consistent with btrfs_search_path_in_tree.
2365 if (args->treeid == 0)
2366 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2368 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2373 if (!capable(CAP_SYS_ADMIN)) {
2378 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2379 args->treeid, args->objectid,
2383 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2390 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2393 struct dentry *parent = file->f_path.dentry;
2394 struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb);
2395 struct dentry *dentry;
2396 struct inode *dir = d_inode(parent);
2397 struct inode *inode;
2398 struct btrfs_root *root = BTRFS_I(dir)->root;
2399 struct btrfs_root *dest = NULL;
2400 struct btrfs_ioctl_vol_args *vol_args;
2401 struct btrfs_trans_handle *trans;
2402 struct btrfs_block_rsv block_rsv;
2404 u64 qgroup_reserved;
2409 if (!S_ISDIR(dir->i_mode))
2412 vol_args = memdup_user(arg, sizeof(*vol_args));
2413 if (IS_ERR(vol_args))
2414 return PTR_ERR(vol_args);
2416 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2417 namelen = strlen(vol_args->name);
2418 if (strchr(vol_args->name, '/') ||
2419 strncmp(vol_args->name, "..", namelen) == 0) {
2424 err = mnt_want_write_file(file);
2429 err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
2431 goto out_drop_write;
2432 dentry = lookup_one_len(vol_args->name, parent, namelen);
2433 if (IS_ERR(dentry)) {
2434 err = PTR_ERR(dentry);
2435 goto out_unlock_dir;
2438 if (d_really_is_negative(dentry)) {
2443 inode = d_inode(dentry);
2444 dest = BTRFS_I(inode)->root;
2445 if (!capable(CAP_SYS_ADMIN)) {
2447 * Regular user. Only allow this with a special mount
2448 * option, when the user has write+exec access to the
2449 * subvol root, and when rmdir(2) would have been
2452 * Note that this is _not_ check that the subvol is
2453 * empty or doesn't contain data that we wouldn't
2454 * otherwise be able to delete.
2456 * Users who want to delete empty subvols should try
2460 if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
2464 * Do not allow deletion if the parent dir is the same
2465 * as the dir to be deleted. That means the ioctl
2466 * must be called on the dentry referencing the root
2467 * of the subvol, not a random directory contained
2474 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2479 /* check if subvolume may be deleted by a user */
2480 err = btrfs_may_delete(dir, dentry, 1);
2484 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
2492 * Don't allow to delete a subvolume with send in progress. This is
2493 * inside the i_mutex so the error handling that has to drop the bit
2494 * again is not run concurrently.
2496 spin_lock(&dest->root_item_lock);
2497 root_flags = btrfs_root_flags(&dest->root_item);
2498 if (dest->send_in_progress == 0) {
2499 btrfs_set_root_flags(&dest->root_item,
2500 root_flags | BTRFS_ROOT_SUBVOL_DEAD);
2501 spin_unlock(&dest->root_item_lock);
2503 spin_unlock(&dest->root_item_lock);
2505 "Attempt to delete subvolume %llu during send",
2506 dest->root_key.objectid);
2508 goto out_unlock_inode;
2511 down_write(&fs_info->subvol_sem);
2513 err = may_destroy_subvol(dest);
2517 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2519 * One for dir inode, two for dir entries, two for root
2522 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2523 5, &qgroup_reserved, true);
2527 trans = btrfs_start_transaction(root, 0);
2528 if (IS_ERR(trans)) {
2529 err = PTR_ERR(trans);
2532 trans->block_rsv = &block_rsv;
2533 trans->bytes_reserved = block_rsv.size;
2535 btrfs_record_snapshot_destroy(trans, BTRFS_I(dir));
2537 ret = btrfs_unlink_subvol(trans, root, dir,
2538 dest->root_key.objectid,
2539 dentry->d_name.name,
2540 dentry->d_name.len);
2543 btrfs_abort_transaction(trans, ret);
2547 btrfs_record_root_in_trans(trans, dest);
2549 memset(&dest->root_item.drop_progress, 0,
2550 sizeof(dest->root_item.drop_progress));
2551 dest->root_item.drop_level = 0;
2552 btrfs_set_root_refs(&dest->root_item, 0);
2554 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
2555 ret = btrfs_insert_orphan_item(trans,
2557 dest->root_key.objectid);
2559 btrfs_abort_transaction(trans, ret);
2565 ret = btrfs_uuid_tree_rem(trans, fs_info, dest->root_item.uuid,
2566 BTRFS_UUID_KEY_SUBVOL,
2567 dest->root_key.objectid);
2568 if (ret && ret != -ENOENT) {
2569 btrfs_abort_transaction(trans, ret);
2573 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2574 ret = btrfs_uuid_tree_rem(trans, fs_info,
2575 dest->root_item.received_uuid,
2576 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2577 dest->root_key.objectid);
2578 if (ret && ret != -ENOENT) {
2579 btrfs_abort_transaction(trans, ret);
2586 trans->block_rsv = NULL;
2587 trans->bytes_reserved = 0;
2588 ret = btrfs_end_transaction(trans);
2591 inode->i_flags |= S_DEAD;
2593 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
2595 up_write(&fs_info->subvol_sem);
2597 spin_lock(&dest->root_item_lock);
2598 root_flags = btrfs_root_flags(&dest->root_item);
2599 btrfs_set_root_flags(&dest->root_item,
2600 root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
2601 spin_unlock(&dest->root_item_lock);
2604 inode_unlock(inode);
2606 d_invalidate(dentry);
2607 btrfs_invalidate_inodes(dest);
2609 ASSERT(dest->send_in_progress == 0);
2612 if (dest->ino_cache_inode) {
2613 iput(dest->ino_cache_inode);
2614 dest->ino_cache_inode = NULL;
2622 mnt_drop_write_file(file);
2628 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2630 struct inode *inode = file_inode(file);
2631 struct btrfs_root *root = BTRFS_I(inode)->root;
2632 struct btrfs_ioctl_defrag_range_args *range;
2635 ret = mnt_want_write_file(file);
2639 if (btrfs_root_readonly(root)) {
2644 switch (inode->i_mode & S_IFMT) {
2646 if (!capable(CAP_SYS_ADMIN)) {
2650 ret = btrfs_defrag_root(root);
2653 if (!(file->f_mode & FMODE_WRITE)) {
2658 range = kzalloc(sizeof(*range), GFP_KERNEL);
2665 if (copy_from_user(range, argp,
2671 /* compression requires us to start the IO */
2672 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2673 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2674 range->extent_thresh = (u32)-1;
2677 /* the rest are all set to zero by kzalloc */
2678 range->len = (u64)-1;
2680 ret = btrfs_defrag_file(file_inode(file), file,
2690 mnt_drop_write_file(file);
2694 static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg)
2696 struct btrfs_ioctl_vol_args *vol_args;
2699 if (!capable(CAP_SYS_ADMIN))
2702 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags))
2703 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2705 mutex_lock(&fs_info->volume_mutex);
2706 vol_args = memdup_user(arg, sizeof(*vol_args));
2707 if (IS_ERR(vol_args)) {
2708 ret = PTR_ERR(vol_args);
2712 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2713 ret = btrfs_init_new_device(fs_info, vol_args->name);
2716 btrfs_info(fs_info, "disk added %s", vol_args->name);
2720 mutex_unlock(&fs_info->volume_mutex);
2721 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
2725 static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
2727 struct inode *inode = file_inode(file);
2728 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2729 struct btrfs_ioctl_vol_args_v2 *vol_args;
2732 if (!capable(CAP_SYS_ADMIN))
2735 ret = mnt_want_write_file(file);
2739 vol_args = memdup_user(arg, sizeof(*vol_args));
2740 if (IS_ERR(vol_args)) {
2741 ret = PTR_ERR(vol_args);
2745 /* Check for compatibility reject unknown flags */
2746 if (vol_args->flags & ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED) {
2751 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
2752 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2756 mutex_lock(&fs_info->volume_mutex);
2757 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
2758 ret = btrfs_rm_device(fs_info, NULL, vol_args->devid);
2760 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
2761 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
2763 mutex_unlock(&fs_info->volume_mutex);
2764 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
2767 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
2768 btrfs_info(fs_info, "device deleted: id %llu",
2771 btrfs_info(fs_info, "device deleted: %s",
2777 mnt_drop_write_file(file);
2781 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2783 struct inode *inode = file_inode(file);
2784 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2785 struct btrfs_ioctl_vol_args *vol_args;
2788 if (!capable(CAP_SYS_ADMIN))
2791 ret = mnt_want_write_file(file);
2795 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
2796 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2797 goto out_drop_write;
2800 vol_args = memdup_user(arg, sizeof(*vol_args));
2801 if (IS_ERR(vol_args)) {
2802 ret = PTR_ERR(vol_args);
2806 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2807 mutex_lock(&fs_info->volume_mutex);
2808 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
2809 mutex_unlock(&fs_info->volume_mutex);
2812 btrfs_info(fs_info, "disk deleted %s", vol_args->name);
2815 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
2817 mnt_drop_write_file(file);
2822 static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
2825 struct btrfs_ioctl_fs_info_args *fi_args;
2826 struct btrfs_device *device;
2827 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2830 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2834 mutex_lock(&fs_devices->device_list_mutex);
2835 fi_args->num_devices = fs_devices->num_devices;
2836 memcpy(&fi_args->fsid, fs_info->fsid, sizeof(fi_args->fsid));
2838 list_for_each_entry(device, &fs_devices->devices, dev_list) {
2839 if (device->devid > fi_args->max_id)
2840 fi_args->max_id = device->devid;
2842 mutex_unlock(&fs_devices->device_list_mutex);
2844 fi_args->nodesize = fs_info->nodesize;
2845 fi_args->sectorsize = fs_info->sectorsize;
2846 fi_args->clone_alignment = fs_info->sectorsize;
2848 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2855 static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
2858 struct btrfs_ioctl_dev_info_args *di_args;
2859 struct btrfs_device *dev;
2860 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2862 char *s_uuid = NULL;
2864 di_args = memdup_user(arg, sizeof(*di_args));
2865 if (IS_ERR(di_args))
2866 return PTR_ERR(di_args);
2868 if (!btrfs_is_empty_uuid(di_args->uuid))
2869 s_uuid = di_args->uuid;
2871 mutex_lock(&fs_devices->device_list_mutex);
2872 dev = btrfs_find_device(fs_info, di_args->devid, s_uuid, NULL);
2879 di_args->devid = dev->devid;
2880 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2881 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
2882 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2884 struct rcu_string *name;
2887 name = rcu_dereference(dev->name);
2888 strncpy(di_args->path, name->str, sizeof(di_args->path));
2890 di_args->path[sizeof(di_args->path) - 1] = 0;
2892 di_args->path[0] = '\0';
2896 mutex_unlock(&fs_devices->device_list_mutex);
2897 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2904 static struct page *extent_same_get_page(struct inode *inode, pgoff_t index)
2908 page = grab_cache_page(inode->i_mapping, index);
2910 return ERR_PTR(-ENOMEM);
2912 if (!PageUptodate(page)) {
2915 ret = btrfs_readpage(NULL, page);
2917 return ERR_PTR(ret);
2919 if (!PageUptodate(page)) {
2922 return ERR_PTR(-EIO);
2924 if (page->mapping != inode->i_mapping) {
2927 return ERR_PTR(-EAGAIN);
2934 static int gather_extent_pages(struct inode *inode, struct page **pages,
2935 int num_pages, u64 off)
2938 pgoff_t index = off >> PAGE_SHIFT;
2940 for (i = 0; i < num_pages; i++) {
2942 pages[i] = extent_same_get_page(inode, index + i);
2943 if (IS_ERR(pages[i])) {
2944 int err = PTR_ERR(pages[i]);
2955 static int lock_extent_range(struct inode *inode, u64 off, u64 len,
2956 bool retry_range_locking)
2959 * Do any pending delalloc/csum calculations on inode, one way or
2960 * another, and lock file content.
2961 * The locking order is:
2964 * 2) range in the inode's io tree
2967 struct btrfs_ordered_extent *ordered;
2968 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2969 ordered = btrfs_lookup_first_ordered_extent(inode,
2972 ordered->file_offset + ordered->len <= off ||
2973 ordered->file_offset >= off + len) &&
2974 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2975 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2977 btrfs_put_ordered_extent(ordered);
2980 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2982 btrfs_put_ordered_extent(ordered);
2983 if (!retry_range_locking)
2985 btrfs_wait_ordered_range(inode, off, len);
2990 static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2)
2992 inode_unlock(inode1);
2993 inode_unlock(inode2);
2996 static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2)
2998 if (inode1 < inode2)
2999 swap(inode1, inode2);
3001 inode_lock_nested(inode1, I_MUTEX_PARENT);
3002 inode_lock_nested(inode2, I_MUTEX_CHILD);
3005 static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1,
3006 struct inode *inode2, u64 loff2, u64 len)
3008 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
3009 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
3012 static int btrfs_double_extent_lock(struct inode *inode1, u64 loff1,
3013 struct inode *inode2, u64 loff2, u64 len,
3014 bool retry_range_locking)
3018 if (inode1 < inode2) {
3019 swap(inode1, inode2);
3022 ret = lock_extent_range(inode1, loff1, len, retry_range_locking);
3025 ret = lock_extent_range(inode2, loff2, len, retry_range_locking);
3027 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1,
3034 struct page **src_pages;
3035 struct page **dst_pages;
3038 static void btrfs_cmp_data_free(struct cmp_pages *cmp)
3043 for (i = 0; i < cmp->num_pages; i++) {
3044 pg = cmp->src_pages[i];
3049 pg = cmp->dst_pages[i];
3055 kfree(cmp->src_pages);
3056 kfree(cmp->dst_pages);
3059 static int btrfs_cmp_data_prepare(struct inode *src, u64 loff,
3060 struct inode *dst, u64 dst_loff,
3061 u64 len, struct cmp_pages *cmp)
3064 int num_pages = PAGE_ALIGN(len) >> PAGE_SHIFT;
3065 struct page **src_pgarr, **dst_pgarr;
3068 * We must gather up all the pages before we initiate our
3069 * extent locking. We use an array for the page pointers. Size
3070 * of the array is bounded by len, which is in turn bounded by
3071 * BTRFS_MAX_DEDUPE_LEN.
3073 src_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL);
3074 dst_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL);
3075 if (!src_pgarr || !dst_pgarr) {
3080 cmp->num_pages = num_pages;
3081 cmp->src_pages = src_pgarr;
3082 cmp->dst_pages = dst_pgarr;
3085 * If deduping ranges in the same inode, locking rules make it mandatory
3086 * to always lock pages in ascending order to avoid deadlocks with
3087 * concurrent tasks (such as starting writeback/delalloc).
3089 if (src == dst && dst_loff < loff) {
3090 swap(src_pgarr, dst_pgarr);
3091 swap(loff, dst_loff);
3094 ret = gather_extent_pages(src, src_pgarr, cmp->num_pages, loff);
3098 ret = gather_extent_pages(dst, dst_pgarr, cmp->num_pages, dst_loff);
3102 btrfs_cmp_data_free(cmp);
3106 static int btrfs_cmp_data(u64 len, struct cmp_pages *cmp)
3110 struct page *src_page, *dst_page;
3111 unsigned int cmp_len = PAGE_SIZE;
3112 void *addr, *dst_addr;
3116 if (len < PAGE_SIZE)
3119 BUG_ON(i >= cmp->num_pages);
3121 src_page = cmp->src_pages[i];
3122 dst_page = cmp->dst_pages[i];
3123 ASSERT(PageLocked(src_page));
3124 ASSERT(PageLocked(dst_page));
3126 addr = kmap_atomic(src_page);
3127 dst_addr = kmap_atomic(dst_page);
3129 flush_dcache_page(src_page);
3130 flush_dcache_page(dst_page);
3132 if (memcmp(addr, dst_addr, cmp_len))
3135 kunmap_atomic(addr);
3136 kunmap_atomic(dst_addr);
3148 static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen,
3152 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
3154 if (off + olen > inode->i_size || off + olen < off)
3157 /* if we extend to eof, continue to block boundary */
3158 if (off + len == inode->i_size)
3159 *plen = len = ALIGN(inode->i_size, bs) - off;
3161 /* Check that we are block aligned - btrfs_clone() requires this */
3162 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
3168 static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen,
3169 struct inode *dst, u64 dst_loff)
3173 struct cmp_pages cmp;
3174 bool same_inode = (src == dst);
3175 u64 same_lock_start = 0;
3176 u64 same_lock_len = 0;
3184 btrfs_double_inode_lock(src, dst);
3186 ret = extent_same_check_offsets(src, loff, &len, olen);
3190 ret = extent_same_check_offsets(dst, dst_loff, &len, olen);
3196 * Single inode case wants the same checks, except we
3197 * don't want our length pushed out past i_size as
3198 * comparing that data range makes no sense.
3200 * extent_same_check_offsets() will do this for an
3201 * unaligned length at i_size, so catch it here and
3202 * reject the request.
3204 * This effectively means we require aligned extents
3205 * for the single-inode case, whereas the other cases
3206 * allow an unaligned length so long as it ends at
3214 /* Check for overlapping ranges */
3215 if (dst_loff + len > loff && dst_loff < loff + len) {
3220 same_lock_start = min_t(u64, loff, dst_loff);
3221 same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start;
3224 * If the source and destination inodes are different, the
3225 * source's range end offset matches the source's i_size, that
3226 * i_size is not a multiple of the sector size, and the
3227 * destination range does not go past the destination's i_size,
3228 * we must round down the length to the nearest sector size
3229 * multiple. If we don't do this adjustment we end replacing
3230 * with zeroes the bytes in the range that starts at the
3231 * deduplication range's end offset and ends at the next sector
3234 if (loff + olen == i_size_read(src) &&
3235 dst_loff + len < i_size_read(dst)) {
3236 const u64 sz = BTRFS_I(src)->root->fs_info->sectorsize;
3238 len = round_down(i_size_read(src), sz) - loff;
3245 /* don't make the dst file partly checksummed */
3246 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3247 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
3253 ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, &cmp);
3258 ret = lock_extent_range(src, same_lock_start, same_lock_len,
3261 ret = btrfs_double_extent_lock(src, loff, dst, dst_loff, len,
3264 * If one of the inodes has dirty pages in the respective range or
3265 * ordered extents, we need to flush dellaloc and wait for all ordered
3266 * extents in the range. We must unlock the pages and the ranges in the
3267 * io trees to avoid deadlocks when flushing delalloc (requires locking
3268 * pages) and when waiting for ordered extents to complete (they require
3271 if (ret == -EAGAIN) {
3273 * Ranges in the io trees already unlocked. Now unlock all
3274 * pages before waiting for all IO to complete.
3276 btrfs_cmp_data_free(&cmp);
3278 btrfs_wait_ordered_range(src, same_lock_start,
3281 btrfs_wait_ordered_range(src, loff, len);
3282 btrfs_wait_ordered_range(dst, dst_loff, len);
3288 /* ranges in the io trees already unlocked */
3289 btrfs_cmp_data_free(&cmp);
3293 /* pass original length for comparison so we stay within i_size */
3294 ret = btrfs_cmp_data(olen, &cmp);
3296 ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1);
3299 unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start,
3300 same_lock_start + same_lock_len - 1);
3302 btrfs_double_extent_unlock(src, loff, dst, dst_loff, len);
3304 btrfs_cmp_data_free(&cmp);
3309 btrfs_double_inode_unlock(src, dst);
3314 #define BTRFS_MAX_DEDUPE_LEN SZ_16M
3316 ssize_t btrfs_dedupe_file_range(struct file *src_file, u64 loff, u64 olen,
3317 struct file *dst_file, u64 dst_loff)
3319 struct inode *src = file_inode(src_file);
3320 struct inode *dst = file_inode(dst_file);
3321 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
3324 if (olen > BTRFS_MAX_DEDUPE_LEN)
3325 olen = BTRFS_MAX_DEDUPE_LEN;
3327 if (WARN_ON_ONCE(bs < PAGE_SIZE)) {
3329 * Btrfs does not support blocksize < page_size. As a
3330 * result, btrfs_cmp_data() won't correctly handle
3331 * this situation without an update.
3336 res = btrfs_extent_same(src, loff, olen, dst, dst_loff);
3342 static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3343 struct inode *inode,
3349 struct btrfs_root *root = BTRFS_I(inode)->root;
3352 inode_inc_iversion(inode);
3353 if (!no_time_update)
3354 inode->i_mtime = inode->i_ctime = current_time(inode);
3356 * We round up to the block size at eof when determining which
3357 * extents to clone above, but shouldn't round up the file size.
3359 if (endoff > destoff + olen)
3360 endoff = destoff + olen;
3361 if (endoff > inode->i_size)
3362 btrfs_i_size_write(BTRFS_I(inode), endoff);
3364 ret = btrfs_update_inode(trans, root, inode);
3366 btrfs_abort_transaction(trans, ret);
3367 btrfs_end_transaction(trans);
3370 ret = btrfs_end_transaction(trans);
3375 static void clone_update_extent_map(struct btrfs_inode *inode,
3376 const struct btrfs_trans_handle *trans,
3377 const struct btrfs_path *path,
3378 const u64 hole_offset,
3381 struct extent_map_tree *em_tree = &inode->extent_tree;
3382 struct extent_map *em;
3385 em = alloc_extent_map();
3387 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags);
3392 struct btrfs_file_extent_item *fi;
3394 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3395 struct btrfs_file_extent_item);
3396 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3397 em->generation = -1;
3398 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3399 BTRFS_FILE_EXTENT_INLINE)
3400 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3401 &inode->runtime_flags);
3403 em->start = hole_offset;
3405 em->ram_bytes = em->len;
3406 em->orig_start = hole_offset;
3407 em->block_start = EXTENT_MAP_HOLE;
3409 em->orig_block_len = 0;
3410 em->compress_type = BTRFS_COMPRESS_NONE;
3411 em->generation = trans->transid;
3415 write_lock(&em_tree->lock);
3416 ret = add_extent_mapping(em_tree, em, 1);
3417 write_unlock(&em_tree->lock);
3418 if (ret != -EEXIST) {
3419 free_extent_map(em);
3422 btrfs_drop_extent_cache(inode, em->start,
3423 em->start + em->len - 1, 0);
3427 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags);
3431 * Make sure we do not end up inserting an inline extent into a file that has
3432 * already other (non-inline) extents. If a file has an inline extent it can
3433 * not have any other extents and the (single) inline extent must start at the
3434 * file offset 0. Failing to respect these rules will lead to file corruption,
3435 * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc
3437 * We can have extents that have been already written to disk or we can have
3438 * dirty ranges still in delalloc, in which case the extent maps and items are
3439 * created only when we run delalloc, and the delalloc ranges might fall outside
3440 * the range we are currently locking in the inode's io tree. So we check the
3441 * inode's i_size because of that (i_size updates are done while holding the
3442 * i_mutex, which we are holding here).
3443 * We also check to see if the inode has a size not greater than "datal" but has
3444 * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are
3445 * protected against such concurrent fallocate calls by the i_mutex).
3447 * If the file has no extents but a size greater than datal, do not allow the
3448 * copy because we would need turn the inline extent into a non-inline one (even
3449 * with NO_HOLES enabled). If we find our destination inode only has one inline
3450 * extent, just overwrite it with the source inline extent if its size is less
3451 * than the source extent's size, or we could copy the source inline extent's
3452 * data into the destination inode's inline extent if the later is greater then
3455 static int clone_copy_inline_extent(struct inode *dst,
3456 struct btrfs_trans_handle *trans,
3457 struct btrfs_path *path,
3458 struct btrfs_key *new_key,
3459 const u64 drop_start,
3465 struct btrfs_fs_info *fs_info = btrfs_sb(dst->i_sb);
3466 struct btrfs_root *root = BTRFS_I(dst)->root;
3467 const u64 aligned_end = ALIGN(new_key->offset + datal,
3468 fs_info->sectorsize);
3470 struct btrfs_key key;
3472 if (new_key->offset > 0)
3475 key.objectid = btrfs_ino(BTRFS_I(dst));
3476 key.type = BTRFS_EXTENT_DATA_KEY;
3478 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3481 } else if (ret > 0) {
3482 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
3483 ret = btrfs_next_leaf(root, path);
3487 goto copy_inline_extent;
3489 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
3490 if (key.objectid == btrfs_ino(BTRFS_I(dst)) &&
3491 key.type == BTRFS_EXTENT_DATA_KEY) {
3492 ASSERT(key.offset > 0);
3495 } else if (i_size_read(dst) <= datal) {
3496 struct btrfs_file_extent_item *ei;
3500 * If the file size is <= datal, make sure there are no other
3501 * extents following (can happen do to an fallocate call with
3502 * the flag FALLOC_FL_KEEP_SIZE).
3504 ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
3505 struct btrfs_file_extent_item);
3507 * If it's an inline extent, it can not have other extents
3510 if (btrfs_file_extent_type(path->nodes[0], ei) ==
3511 BTRFS_FILE_EXTENT_INLINE)
3512 goto copy_inline_extent;
3514 ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei);
3515 if (ext_len > aligned_end)
3518 ret = btrfs_next_item(root, path);
3521 } else if (ret == 0) {
3522 btrfs_item_key_to_cpu(path->nodes[0], &key,
3524 if (key.objectid == btrfs_ino(BTRFS_I(dst)) &&
3525 key.type == BTRFS_EXTENT_DATA_KEY)
3532 * We have no extent items, or we have an extent at offset 0 which may
3533 * or may not be inlined. All these cases are dealt the same way.
3535 if (i_size_read(dst) > datal) {
3537 * If the destination inode has an inline extent...
3538 * This would require copying the data from the source inline
3539 * extent into the beginning of the destination's inline extent.
3540 * But this is really complex, both extents can be compressed
3541 * or just one of them, which would require decompressing and
3542 * re-compressing data (which could increase the new compressed
3543 * size, not allowing the compressed data to fit anymore in an
3545 * So just don't support this case for now (it should be rare,
3546 * we are not really saving space when cloning inline extents).
3551 btrfs_release_path(path);
3552 ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1);
3555 ret = btrfs_insert_empty_item(trans, root, path, new_key, size);
3560 const u32 start = btrfs_file_extent_calc_inline_size(0);
3562 memmove(inline_data + start, inline_data + start + skip, datal);
3565 write_extent_buffer(path->nodes[0], inline_data,
3566 btrfs_item_ptr_offset(path->nodes[0],
3569 inode_add_bytes(dst, datal);
3575 * btrfs_clone() - clone a range from inode file to another
3577 * @src: Inode to clone from
3578 * @inode: Inode to clone to
3579 * @off: Offset within source to start clone from
3580 * @olen: Original length, passed by user, of range to clone
3581 * @olen_aligned: Block-aligned value of olen
3582 * @destoff: Offset within @inode to start clone
3583 * @no_time_update: Whether to update mtime/ctime on the target inode
3585 static int btrfs_clone(struct inode *src, struct inode *inode,
3586 const u64 off, const u64 olen, const u64 olen_aligned,
3587 const u64 destoff, int no_time_update)
3589 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3590 struct btrfs_root *root = BTRFS_I(inode)->root;
3591 struct btrfs_path *path = NULL;
3592 struct extent_buffer *leaf;
3593 struct btrfs_trans_handle *trans;
3595 struct btrfs_key key;
3599 const u64 len = olen_aligned;
3600 u64 last_dest_end = destoff;
3603 buf = kvmalloc(fs_info->nodesize, GFP_KERNEL);
3607 path = btrfs_alloc_path();
3613 path->reada = READA_FORWARD;
3615 key.objectid = btrfs_ino(BTRFS_I(src));
3616 key.type = BTRFS_EXTENT_DATA_KEY;
3620 u64 next_key_min_offset = key.offset + 1;
3623 * note the key will change type as we walk through the
3626 path->leave_spinning = 1;
3627 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3632 * First search, if no extent item that starts at offset off was
3633 * found but the previous item is an extent item, it's possible
3634 * it might overlap our target range, therefore process it.
3636 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3637 btrfs_item_key_to_cpu(path->nodes[0], &key,
3638 path->slots[0] - 1);
3639 if (key.type == BTRFS_EXTENT_DATA_KEY)
3643 nritems = btrfs_header_nritems(path->nodes[0]);
3645 if (path->slots[0] >= nritems) {
3646 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
3651 nritems = btrfs_header_nritems(path->nodes[0]);
3653 leaf = path->nodes[0];
3654 slot = path->slots[0];
3656 btrfs_item_key_to_cpu(leaf, &key, slot);
3657 if (key.type > BTRFS_EXTENT_DATA_KEY ||
3658 key.objectid != btrfs_ino(BTRFS_I(src)))
3661 if (key.type == BTRFS_EXTENT_DATA_KEY) {
3662 struct btrfs_file_extent_item *extent;
3665 struct btrfs_key new_key;
3666 u64 disko = 0, diskl = 0;
3667 u64 datao = 0, datal = 0;
3671 extent = btrfs_item_ptr(leaf, slot,
3672 struct btrfs_file_extent_item);
3673 comp = btrfs_file_extent_compression(leaf, extent);
3674 type = btrfs_file_extent_type(leaf, extent);
3675 if (type == BTRFS_FILE_EXTENT_REG ||
3676 type == BTRFS_FILE_EXTENT_PREALLOC) {
3677 disko = btrfs_file_extent_disk_bytenr(leaf,
3679 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3681 datao = btrfs_file_extent_offset(leaf, extent);
3682 datal = btrfs_file_extent_num_bytes(leaf,
3684 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3685 /* take upper bound, may be compressed */
3686 datal = btrfs_file_extent_ram_bytes(leaf,
3691 * The first search might have left us at an extent
3692 * item that ends before our target range's start, can
3693 * happen if we have holes and NO_HOLES feature enabled.
3695 if (key.offset + datal <= off) {
3698 } else if (key.offset >= off + len) {
3701 next_key_min_offset = key.offset + datal;
3702 size = btrfs_item_size_nr(leaf, slot);
3703 read_extent_buffer(leaf, buf,
3704 btrfs_item_ptr_offset(leaf, slot),
3707 btrfs_release_path(path);
3708 path->leave_spinning = 0;
3710 memcpy(&new_key, &key, sizeof(new_key));
3711 new_key.objectid = btrfs_ino(BTRFS_I(inode));
3712 if (off <= key.offset)
3713 new_key.offset = key.offset + destoff - off;
3715 new_key.offset = destoff;
3718 * Deal with a hole that doesn't have an extent item
3719 * that represents it (NO_HOLES feature enabled).
3720 * This hole is either in the middle of the cloning
3721 * range or at the beginning (fully overlaps it or
3722 * partially overlaps it).
3724 if (new_key.offset != last_dest_end)
3725 drop_start = last_dest_end;
3727 drop_start = new_key.offset;
3730 * 1 - adjusting old extent (we may have to split it)
3731 * 1 - add new extent
3734 trans = btrfs_start_transaction(root, 3);
3735 if (IS_ERR(trans)) {
3736 ret = PTR_ERR(trans);
3740 if (type == BTRFS_FILE_EXTENT_REG ||
3741 type == BTRFS_FILE_EXTENT_PREALLOC) {
3743 * a | --- range to clone ---| b
3744 * | ------------- extent ------------- |
3747 /* subtract range b */
3748 if (key.offset + datal > off + len)
3749 datal = off + len - key.offset;
3751 /* subtract range a */
3752 if (off > key.offset) {
3753 datao += off - key.offset;
3754 datal -= off - key.offset;
3757 ret = btrfs_drop_extents(trans, root, inode,
3759 new_key.offset + datal,
3762 if (ret != -EOPNOTSUPP)
3763 btrfs_abort_transaction(trans,
3765 btrfs_end_transaction(trans);
3769 ret = btrfs_insert_empty_item(trans, root, path,
3772 btrfs_abort_transaction(trans, ret);
3773 btrfs_end_transaction(trans);
3777 leaf = path->nodes[0];
3778 slot = path->slots[0];
3779 write_extent_buffer(leaf, buf,
3780 btrfs_item_ptr_offset(leaf, slot),
3783 extent = btrfs_item_ptr(leaf, slot,
3784 struct btrfs_file_extent_item);
3786 /* disko == 0 means it's a hole */
3790 btrfs_set_file_extent_offset(leaf, extent,
3792 btrfs_set_file_extent_num_bytes(leaf, extent,
3796 inode_add_bytes(inode, datal);
3797 ret = btrfs_inc_extent_ref(trans,
3800 root->root_key.objectid,
3801 btrfs_ino(BTRFS_I(inode)),
3802 new_key.offset - datao);
3804 btrfs_abort_transaction(trans,
3806 btrfs_end_transaction(trans);
3811 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3815 if (off > key.offset) {
3816 skip = off - key.offset;
3817 new_key.offset += skip;
3820 if (key.offset + datal > off + len)
3821 trim = key.offset + datal - (off + len);
3823 if (comp && (skip || trim)) {
3825 btrfs_end_transaction(trans);
3828 size -= skip + trim;
3829 datal -= skip + trim;
3831 ret = clone_copy_inline_extent(inode,
3838 if (ret != -EOPNOTSUPP)
3839 btrfs_abort_transaction(trans,
3841 btrfs_end_transaction(trans);
3844 leaf = path->nodes[0];
3845 slot = path->slots[0];
3848 /* If we have an implicit hole (NO_HOLES feature). */
3849 if (drop_start < new_key.offset)
3850 clone_update_extent_map(BTRFS_I(inode), trans,
3852 new_key.offset - drop_start);
3854 clone_update_extent_map(BTRFS_I(inode), trans,
3857 btrfs_mark_buffer_dirty(leaf);
3858 btrfs_release_path(path);
3860 last_dest_end = ALIGN(new_key.offset + datal,
3861 fs_info->sectorsize);
3862 ret = clone_finish_inode_update(trans, inode,
3868 if (new_key.offset + datal >= destoff + len)
3871 btrfs_release_path(path);
3872 key.offset = next_key_min_offset;
3874 if (fatal_signal_pending(current)) {
3883 if (last_dest_end < destoff + len) {
3885 * We have an implicit hole (NO_HOLES feature is enabled) that
3886 * fully or partially overlaps our cloning range at its end.
3888 btrfs_release_path(path);
3891 * 1 - remove extent(s)
3894 trans = btrfs_start_transaction(root, 2);
3895 if (IS_ERR(trans)) {
3896 ret = PTR_ERR(trans);
3899 ret = btrfs_drop_extents(trans, root, inode,
3900 last_dest_end, destoff + len, 1);
3902 if (ret != -EOPNOTSUPP)
3903 btrfs_abort_transaction(trans, ret);
3904 btrfs_end_transaction(trans);
3907 clone_update_extent_map(BTRFS_I(inode), trans, NULL,
3909 destoff + len - last_dest_end);
3910 ret = clone_finish_inode_update(trans, inode, destoff + len,
3911 destoff, olen, no_time_update);
3915 btrfs_free_path(path);
3920 static noinline int btrfs_clone_files(struct file *file, struct file *file_src,
3921 u64 off, u64 olen, u64 destoff)
3923 struct inode *inode = file_inode(file);
3924 struct inode *src = file_inode(file_src);
3925 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3926 struct btrfs_root *root = BTRFS_I(inode)->root;
3929 u64 bs = fs_info->sb->s_blocksize;
3930 int same_inode = src == inode;
3934 * - split compressed inline extents. annoying: we need to
3935 * decompress into destination's address_space (the file offset
3936 * may change, so source mapping won't do), then recompress (or
3937 * otherwise reinsert) a subrange.
3939 * - split destination inode's inline extents. The inline extents can
3940 * be either compressed or non-compressed.
3943 if (btrfs_root_readonly(root))
3946 if (file_src->f_path.mnt != file->f_path.mnt ||
3947 src->i_sb != inode->i_sb)
3950 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3954 btrfs_double_inode_lock(src, inode);
3959 /* don't make the dst file partly checksummed */
3960 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3961 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) {
3966 /* determine range to clone */
3968 if (off + len > src->i_size || off + len < off)
3971 olen = len = src->i_size - off;
3973 * If we extend to eof, continue to block boundary if and only if the
3974 * destination end offset matches the destination file's size, otherwise
3975 * we would be corrupting data by placing the eof block into the middle
3978 if (off + len == src->i_size) {
3979 if (!IS_ALIGNED(len, bs) && destoff + len < inode->i_size)
3981 len = ALIGN(src->i_size, bs) - off;
3989 /* verify the end result is block aligned */
3990 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3991 !IS_ALIGNED(destoff, bs))
3994 /* verify if ranges are overlapped within the same file */
3996 if (destoff + len > off && destoff < off + len)
4000 if (destoff > inode->i_size) {
4001 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
4007 * Lock the target range too. Right after we replace the file extent
4008 * items in the fs tree (which now point to the cloned data), we might
4009 * have a worker replace them with extent items relative to a write
4010 * operation that was issued before this clone operation (i.e. confront
4011 * with inode.c:btrfs_finish_ordered_io).
4014 u64 lock_start = min_t(u64, off, destoff);
4015 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
4017 ret = lock_extent_range(src, lock_start, lock_len, true);
4019 ret = btrfs_double_extent_lock(src, off, inode, destoff, len,
4024 /* ranges in the io trees already unlocked */
4028 ret = btrfs_clone(src, inode, off, olen, len, destoff, 0);
4031 u64 lock_start = min_t(u64, off, destoff);
4032 u64 lock_end = max_t(u64, off, destoff) + len - 1;
4034 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
4036 btrfs_double_extent_unlock(src, off, inode, destoff, len);
4039 * Truncate page cache pages so that future reads will see the cloned
4040 * data immediately and not the previous data.
4042 truncate_inode_pages_range(&inode->i_data,
4043 round_down(destoff, PAGE_SIZE),
4044 round_up(destoff + len, PAGE_SIZE) - 1);
4047 btrfs_double_inode_unlock(src, inode);
4053 int btrfs_clone_file_range(struct file *src_file, loff_t off,
4054 struct file *dst_file, loff_t destoff, u64 len)
4056 return btrfs_clone_files(dst_file, src_file, off, len, destoff);
4060 * there are many ways the trans_start and trans_end ioctls can lead
4061 * to deadlocks. They should only be used by applications that
4062 * basically own the machine, and have a very in depth understanding
4063 * of all the possible deadlocks and enospc problems.
4065 static long btrfs_ioctl_trans_start(struct file *file)
4067 struct inode *inode = file_inode(file);
4068 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4069 struct btrfs_root *root = BTRFS_I(inode)->root;
4070 struct btrfs_trans_handle *trans;
4071 struct btrfs_file_private *private;
4073 static bool warned = false;
4076 if (!capable(CAP_SYS_ADMIN))
4081 "Userspace transaction mechanism is considered "
4082 "deprecated and slated to be removed in 4.17. "
4083 "If you have a valid use case please "
4084 "speak up on the mailing list");
4090 private = file->private_data;
4091 if (private && private->trans)
4094 private = kzalloc(sizeof(struct btrfs_file_private),
4098 file->private_data = private;
4102 if (btrfs_root_readonly(root))
4105 ret = mnt_want_write_file(file);
4109 atomic_inc(&fs_info->open_ioctl_trans);
4112 trans = btrfs_start_ioctl_transaction(root);
4116 private->trans = trans;
4120 atomic_dec(&fs_info->open_ioctl_trans);
4121 mnt_drop_write_file(file);
4126 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
4128 struct inode *inode = file_inode(file);
4129 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4130 struct btrfs_root *root = BTRFS_I(inode)->root;
4131 struct btrfs_root *new_root;
4132 struct btrfs_dir_item *di;
4133 struct btrfs_trans_handle *trans;
4134 struct btrfs_path *path;
4135 struct btrfs_key location;
4136 struct btrfs_disk_key disk_key;
4141 if (!capable(CAP_SYS_ADMIN))
4144 ret = mnt_want_write_file(file);
4148 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
4154 objectid = BTRFS_FS_TREE_OBJECTID;
4156 location.objectid = objectid;
4157 location.type = BTRFS_ROOT_ITEM_KEY;
4158 location.offset = (u64)-1;
4160 new_root = btrfs_read_fs_root_no_name(fs_info, &location);
4161 if (IS_ERR(new_root)) {
4162 ret = PTR_ERR(new_root);
4165 if (!is_fstree(new_root->objectid)) {
4170 path = btrfs_alloc_path();
4175 path->leave_spinning = 1;
4177 trans = btrfs_start_transaction(root, 1);
4178 if (IS_ERR(trans)) {
4179 btrfs_free_path(path);
4180 ret = PTR_ERR(trans);
4184 dir_id = btrfs_super_root_dir(fs_info->super_copy);
4185 di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
4186 dir_id, "default", 7, 1);
4187 if (IS_ERR_OR_NULL(di)) {
4188 btrfs_free_path(path);
4189 btrfs_end_transaction(trans);
4191 "Umm, you don't have the default diritem, this isn't going to work");
4196 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
4197 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
4198 btrfs_mark_buffer_dirty(path->nodes[0]);
4199 btrfs_free_path(path);
4201 btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
4202 btrfs_end_transaction(trans);
4204 mnt_drop_write_file(file);
4208 void btrfs_get_block_group_info(struct list_head *groups_list,
4209 struct btrfs_ioctl_space_info *space)
4211 struct btrfs_block_group_cache *block_group;
4213 space->total_bytes = 0;
4214 space->used_bytes = 0;
4216 list_for_each_entry(block_group, groups_list, list) {
4217 space->flags = block_group->flags;
4218 space->total_bytes += block_group->key.offset;
4219 space->used_bytes +=
4220 btrfs_block_group_used(&block_group->item);
4224 static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
4227 struct btrfs_ioctl_space_args space_args;
4228 struct btrfs_ioctl_space_info space;
4229 struct btrfs_ioctl_space_info *dest;
4230 struct btrfs_ioctl_space_info *dest_orig;
4231 struct btrfs_ioctl_space_info __user *user_dest;
4232 struct btrfs_space_info *info;
4233 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
4234 BTRFS_BLOCK_GROUP_SYSTEM,
4235 BTRFS_BLOCK_GROUP_METADATA,
4236 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
4243 if (copy_from_user(&space_args,
4244 (struct btrfs_ioctl_space_args __user *)arg,
4245 sizeof(space_args)))
4248 for (i = 0; i < num_types; i++) {
4249 struct btrfs_space_info *tmp;
4253 list_for_each_entry_rcu(tmp, &fs_info->space_info,
4255 if (tmp->flags == types[i]) {
4265 down_read(&info->groups_sem);
4266 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4267 if (!list_empty(&info->block_groups[c]))
4270 up_read(&info->groups_sem);
4274 * Global block reserve, exported as a space_info
4278 /* space_slots == 0 means they are asking for a count */
4279 if (space_args.space_slots == 0) {
4280 space_args.total_spaces = slot_count;
4284 slot_count = min_t(u64, space_args.space_slots, slot_count);
4286 alloc_size = sizeof(*dest) * slot_count;
4288 /* we generally have at most 6 or so space infos, one for each raid
4289 * level. So, a whole page should be more than enough for everyone
4291 if (alloc_size > PAGE_SIZE)
4294 space_args.total_spaces = 0;
4295 dest = kmalloc(alloc_size, GFP_KERNEL);
4300 /* now we have a buffer to copy into */
4301 for (i = 0; i < num_types; i++) {
4302 struct btrfs_space_info *tmp;
4309 list_for_each_entry_rcu(tmp, &fs_info->space_info,
4311 if (tmp->flags == types[i]) {
4320 down_read(&info->groups_sem);
4321 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4322 if (!list_empty(&info->block_groups[c])) {
4323 btrfs_get_block_group_info(
4324 &info->block_groups[c], &space);
4325 memcpy(dest, &space, sizeof(space));
4327 space_args.total_spaces++;
4333 up_read(&info->groups_sem);
4337 * Add global block reserve
4340 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
4342 spin_lock(&block_rsv->lock);
4343 space.total_bytes = block_rsv->size;
4344 space.used_bytes = block_rsv->size - block_rsv->reserved;
4345 spin_unlock(&block_rsv->lock);
4346 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
4347 memcpy(dest, &space, sizeof(space));
4348 space_args.total_spaces++;
4351 user_dest = (struct btrfs_ioctl_space_info __user *)
4352 (arg + sizeof(struct btrfs_ioctl_space_args));
4354 if (copy_to_user(user_dest, dest_orig, alloc_size))
4359 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
4366 * there are many ways the trans_start and trans_end ioctls can lead
4367 * to deadlocks. They should only be used by applications that
4368 * basically own the machine, and have a very in depth understanding
4369 * of all the possible deadlocks and enospc problems.
4371 long btrfs_ioctl_trans_end(struct file *file)
4373 struct inode *inode = file_inode(file);
4374 struct btrfs_root *root = BTRFS_I(inode)->root;
4375 struct btrfs_file_private *private = file->private_data;
4377 if (!private || !private->trans)
4380 btrfs_end_transaction(private->trans);
4381 private->trans = NULL;
4383 atomic_dec(&root->fs_info->open_ioctl_trans);
4385 mnt_drop_write_file(file);
4389 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4392 struct btrfs_trans_handle *trans;
4396 trans = btrfs_attach_transaction_barrier(root);
4397 if (IS_ERR(trans)) {
4398 if (PTR_ERR(trans) != -ENOENT)
4399 return PTR_ERR(trans);
4401 /* No running transaction, don't bother */
4402 transid = root->fs_info->last_trans_committed;
4405 transid = trans->transid;
4406 ret = btrfs_commit_transaction_async(trans, 0);
4408 btrfs_end_transaction(trans);
4413 if (copy_to_user(argp, &transid, sizeof(transid)))
4418 static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
4424 if (copy_from_user(&transid, argp, sizeof(transid)))
4427 transid = 0; /* current trans */
4429 return btrfs_wait_for_commit(fs_info, transid);
4432 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
4434 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
4435 struct btrfs_ioctl_scrub_args *sa;
4438 if (!capable(CAP_SYS_ADMIN))
4441 sa = memdup_user(arg, sizeof(*sa));
4445 if (sa->flags & ~BTRFS_SCRUB_SUPPORTED_FLAGS) {
4450 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4451 ret = mnt_want_write_file(file);
4456 ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
4457 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4460 if (copy_to_user(arg, sa, sizeof(*sa)))
4463 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4464 mnt_drop_write_file(file);
4470 static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
4472 if (!capable(CAP_SYS_ADMIN))
4475 return btrfs_scrub_cancel(fs_info);
4478 static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
4481 struct btrfs_ioctl_scrub_args *sa;
4484 if (!capable(CAP_SYS_ADMIN))
4487 sa = memdup_user(arg, sizeof(*sa));
4491 ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
4493 if (copy_to_user(arg, sa, sizeof(*sa)))
4500 static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
4503 struct btrfs_ioctl_get_dev_stats *sa;
4506 sa = memdup_user(arg, sizeof(*sa));
4510 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4515 ret = btrfs_get_dev_stats(fs_info, sa);
4517 if (copy_to_user(arg, sa, sizeof(*sa)))
4524 static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
4527 struct btrfs_ioctl_dev_replace_args *p;
4530 if (!capable(CAP_SYS_ADMIN))
4533 p = memdup_user(arg, sizeof(*p));
4538 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
4539 if (sb_rdonly(fs_info->sb)) {
4543 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
4544 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4546 ret = btrfs_dev_replace_by_ioctl(fs_info, p);
4547 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
4550 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4551 btrfs_dev_replace_status(fs_info, p);
4554 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4555 ret = btrfs_dev_replace_cancel(fs_info, p);
4562 if (copy_to_user(arg, p, sizeof(*p)))
4569 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4575 struct btrfs_ioctl_ino_path_args *ipa = NULL;
4576 struct inode_fs_paths *ipath = NULL;
4577 struct btrfs_path *path;
4579 if (!capable(CAP_DAC_READ_SEARCH))
4582 path = btrfs_alloc_path();
4588 ipa = memdup_user(arg, sizeof(*ipa));
4595 size = min_t(u32, ipa->size, 4096);
4596 ipath = init_ipath(size, root, path);
4597 if (IS_ERR(ipath)) {
4598 ret = PTR_ERR(ipath);
4603 ret = paths_from_inode(ipa->inum, ipath);
4607 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
4608 rel_ptr = ipath->fspath->val[i] -
4609 (u64)(unsigned long)ipath->fspath->val;
4610 ipath->fspath->val[i] = rel_ptr;
4613 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
4614 (void *)(unsigned long)ipath->fspath, size);
4621 btrfs_free_path(path);
4628 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4630 struct btrfs_data_container *inodes = ctx;
4631 const size_t c = 3 * sizeof(u64);
4633 if (inodes->bytes_left >= c) {
4634 inodes->bytes_left -= c;
4635 inodes->val[inodes->elem_cnt] = inum;
4636 inodes->val[inodes->elem_cnt + 1] = offset;
4637 inodes->val[inodes->elem_cnt + 2] = root;
4638 inodes->elem_cnt += 3;
4640 inodes->bytes_missing += c - inodes->bytes_left;
4641 inodes->bytes_left = 0;
4642 inodes->elem_missed += 3;
4648 static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
4653 struct btrfs_ioctl_logical_ino_args *loi;
4654 struct btrfs_data_container *inodes = NULL;
4655 struct btrfs_path *path = NULL;
4657 if (!capable(CAP_SYS_ADMIN))
4660 loi = memdup_user(arg, sizeof(*loi));
4662 return PTR_ERR(loi);
4664 path = btrfs_alloc_path();
4670 size = min_t(u32, loi->size, SZ_64K);
4671 inodes = init_data_container(size);
4672 if (IS_ERR(inodes)) {
4673 ret = PTR_ERR(inodes);
4678 ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
4679 build_ino_list, inodes);
4685 ret = copy_to_user((void *)(unsigned long)loi->inodes,
4686 (void *)(unsigned long)inodes, size);
4691 btrfs_free_path(path);
4698 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
4699 struct btrfs_ioctl_balance_args *bargs)
4701 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4703 bargs->flags = bctl->flags;
4705 if (atomic_read(&fs_info->balance_running))
4706 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4707 if (atomic_read(&fs_info->balance_pause_req))
4708 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
4709 if (atomic_read(&fs_info->balance_cancel_req))
4710 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
4712 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4713 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4714 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4717 spin_lock(&fs_info->balance_lock);
4718 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4719 spin_unlock(&fs_info->balance_lock);
4721 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4725 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4727 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4728 struct btrfs_fs_info *fs_info = root->fs_info;
4729 struct btrfs_ioctl_balance_args *bargs;
4730 struct btrfs_balance_control *bctl;
4731 bool need_unlock; /* for mut. excl. ops lock */
4734 if (!capable(CAP_SYS_ADMIN))
4737 ret = mnt_want_write_file(file);
4742 if (!test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
4743 mutex_lock(&fs_info->volume_mutex);
4744 mutex_lock(&fs_info->balance_mutex);
4750 * mut. excl. ops lock is locked. Three possibilities:
4751 * (1) some other op is running
4752 * (2) balance is running
4753 * (3) balance is paused -- special case (think resume)
4755 mutex_lock(&fs_info->balance_mutex);
4756 if (fs_info->balance_ctl) {
4757 /* this is either (2) or (3) */
4758 if (!atomic_read(&fs_info->balance_running)) {
4759 mutex_unlock(&fs_info->balance_mutex);
4760 if (!mutex_trylock(&fs_info->volume_mutex))
4762 mutex_lock(&fs_info->balance_mutex);
4764 if (fs_info->balance_ctl &&
4765 !atomic_read(&fs_info->balance_running)) {
4767 need_unlock = false;
4771 mutex_unlock(&fs_info->balance_mutex);
4772 mutex_unlock(&fs_info->volume_mutex);
4776 mutex_unlock(&fs_info->balance_mutex);
4782 mutex_unlock(&fs_info->balance_mutex);
4783 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4788 BUG_ON(!test_bit(BTRFS_FS_EXCL_OP, &fs_info->flags));
4791 bargs = memdup_user(arg, sizeof(*bargs));
4792 if (IS_ERR(bargs)) {
4793 ret = PTR_ERR(bargs);
4797 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4798 if (!fs_info->balance_ctl) {
4803 bctl = fs_info->balance_ctl;
4804 spin_lock(&fs_info->balance_lock);
4805 bctl->flags |= BTRFS_BALANCE_RESUME;
4806 spin_unlock(&fs_info->balance_lock);
4814 if (fs_info->balance_ctl) {
4819 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
4825 bctl->fs_info = fs_info;
4827 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4828 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4829 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4831 bctl->flags = bargs->flags;
4833 /* balance everything - no filters */
4834 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4837 if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
4844 * Ownership of bctl and filesystem flag BTRFS_FS_EXCL_OP
4845 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4846 * or, if restriper was paused all the way until unmount, in
4847 * free_fs_info. The flag is cleared in __cancel_balance.
4849 need_unlock = false;
4851 ret = btrfs_balance(bctl, bargs);
4855 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4864 mutex_unlock(&fs_info->balance_mutex);
4865 mutex_unlock(&fs_info->volume_mutex);
4867 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
4869 mnt_drop_write_file(file);
4873 static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
4875 if (!capable(CAP_SYS_ADMIN))
4879 case BTRFS_BALANCE_CTL_PAUSE:
4880 return btrfs_pause_balance(fs_info);
4881 case BTRFS_BALANCE_CTL_CANCEL:
4882 return btrfs_cancel_balance(fs_info);
4888 static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
4891 struct btrfs_ioctl_balance_args *bargs;
4894 if (!capable(CAP_SYS_ADMIN))
4897 mutex_lock(&fs_info->balance_mutex);
4898 if (!fs_info->balance_ctl) {
4903 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
4909 update_ioctl_balance_args(fs_info, 1, bargs);
4911 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4916 mutex_unlock(&fs_info->balance_mutex);
4920 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4922 struct inode *inode = file_inode(file);
4923 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4924 struct btrfs_ioctl_quota_ctl_args *sa;
4925 struct btrfs_trans_handle *trans = NULL;
4929 if (!capable(CAP_SYS_ADMIN))
4932 ret = mnt_want_write_file(file);
4936 sa = memdup_user(arg, sizeof(*sa));
4942 down_write(&fs_info->subvol_sem);
4943 trans = btrfs_start_transaction(fs_info->tree_root, 2);
4944 if (IS_ERR(trans)) {
4945 ret = PTR_ERR(trans);
4950 case BTRFS_QUOTA_CTL_ENABLE:
4951 ret = btrfs_quota_enable(trans, fs_info);
4953 case BTRFS_QUOTA_CTL_DISABLE:
4954 ret = btrfs_quota_disable(trans, fs_info);
4961 err = btrfs_commit_transaction(trans);
4966 up_write(&fs_info->subvol_sem);
4968 mnt_drop_write_file(file);
4972 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4974 struct inode *inode = file_inode(file);
4975 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4976 struct btrfs_root *root = BTRFS_I(inode)->root;
4977 struct btrfs_ioctl_qgroup_assign_args *sa;
4978 struct btrfs_trans_handle *trans;
4982 if (!capable(CAP_SYS_ADMIN))
4985 ret = mnt_want_write_file(file);
4989 sa = memdup_user(arg, sizeof(*sa));
4995 trans = btrfs_join_transaction(root);
4996 if (IS_ERR(trans)) {
4997 ret = PTR_ERR(trans);
5002 ret = btrfs_add_qgroup_relation(trans, fs_info,
5005 ret = btrfs_del_qgroup_relation(trans, fs_info,
5009 /* update qgroup status and info */
5010 err = btrfs_run_qgroups(trans, fs_info);
5012 btrfs_handle_fs_error(fs_info, err,
5013 "failed to update qgroup status and info");
5014 err = btrfs_end_transaction(trans);
5021 mnt_drop_write_file(file);
5025 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
5027 struct inode *inode = file_inode(file);
5028 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5029 struct btrfs_root *root = BTRFS_I(inode)->root;
5030 struct btrfs_ioctl_qgroup_create_args *sa;
5031 struct btrfs_trans_handle *trans;
5035 if (!capable(CAP_SYS_ADMIN))
5038 ret = mnt_want_write_file(file);
5042 sa = memdup_user(arg, sizeof(*sa));
5048 if (!sa->qgroupid) {
5053 trans = btrfs_join_transaction(root);
5054 if (IS_ERR(trans)) {
5055 ret = PTR_ERR(trans);
5060 ret = btrfs_create_qgroup(trans, fs_info, sa->qgroupid);
5062 ret = btrfs_remove_qgroup(trans, fs_info, sa->qgroupid);
5065 err = btrfs_end_transaction(trans);
5072 mnt_drop_write_file(file);
5076 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
5078 struct inode *inode = file_inode(file);
5079 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5080 struct btrfs_root *root = BTRFS_I(inode)->root;
5081 struct btrfs_ioctl_qgroup_limit_args *sa;
5082 struct btrfs_trans_handle *trans;
5087 if (!capable(CAP_SYS_ADMIN))
5090 ret = mnt_want_write_file(file);
5094 sa = memdup_user(arg, sizeof(*sa));
5100 trans = btrfs_join_transaction(root);
5101 if (IS_ERR(trans)) {
5102 ret = PTR_ERR(trans);
5106 qgroupid = sa->qgroupid;
5108 /* take the current subvol as qgroup */
5109 qgroupid = root->root_key.objectid;
5112 ret = btrfs_limit_qgroup(trans, fs_info, qgroupid, &sa->lim);
5114 err = btrfs_end_transaction(trans);
5121 mnt_drop_write_file(file);
5125 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
5127 struct inode *inode = file_inode(file);
5128 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5129 struct btrfs_ioctl_quota_rescan_args *qsa;
5132 if (!capable(CAP_SYS_ADMIN))
5135 ret = mnt_want_write_file(file);
5139 qsa = memdup_user(arg, sizeof(*qsa));
5150 ret = btrfs_qgroup_rescan(fs_info);
5155 mnt_drop_write_file(file);
5159 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
5161 struct inode *inode = file_inode(file);
5162 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5163 struct btrfs_ioctl_quota_rescan_args *qsa;
5166 if (!capable(CAP_SYS_ADMIN))
5169 qsa = kzalloc(sizeof(*qsa), GFP_KERNEL);
5173 if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
5175 qsa->progress = fs_info->qgroup_rescan_progress.objectid;
5178 if (copy_to_user(arg, qsa, sizeof(*qsa)))
5185 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
5187 struct inode *inode = file_inode(file);
5188 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5190 if (!capable(CAP_SYS_ADMIN))
5193 return btrfs_qgroup_wait_for_completion(fs_info, true);
5196 static long _btrfs_ioctl_set_received_subvol(struct file *file,
5197 struct btrfs_ioctl_received_subvol_args *sa)
5199 struct inode *inode = file_inode(file);
5200 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5201 struct btrfs_root *root = BTRFS_I(inode)->root;
5202 struct btrfs_root_item *root_item = &root->root_item;
5203 struct btrfs_trans_handle *trans;
5204 struct timespec ct = current_time(inode);
5206 int received_uuid_changed;
5208 if (!inode_owner_or_capable(inode))
5211 ret = mnt_want_write_file(file);
5215 down_write(&fs_info->subvol_sem);
5217 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
5222 if (btrfs_root_readonly(root)) {
5229 * 2 - uuid items (received uuid + subvol uuid)
5231 trans = btrfs_start_transaction(root, 3);
5232 if (IS_ERR(trans)) {
5233 ret = PTR_ERR(trans);
5238 sa->rtransid = trans->transid;
5239 sa->rtime.sec = ct.tv_sec;
5240 sa->rtime.nsec = ct.tv_nsec;
5242 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
5244 if (received_uuid_changed &&
5245 !btrfs_is_empty_uuid(root_item->received_uuid))
5246 btrfs_uuid_tree_rem(trans, fs_info, root_item->received_uuid,
5247 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5248 root->root_key.objectid);
5249 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
5250 btrfs_set_root_stransid(root_item, sa->stransid);
5251 btrfs_set_root_rtransid(root_item, sa->rtransid);
5252 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
5253 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
5254 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
5255 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
5257 ret = btrfs_update_root(trans, fs_info->tree_root,
5258 &root->root_key, &root->root_item);
5260 btrfs_end_transaction(trans);
5263 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
5264 ret = btrfs_uuid_tree_add(trans, fs_info, sa->uuid,
5265 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5266 root->root_key.objectid);
5267 if (ret < 0 && ret != -EEXIST) {
5268 btrfs_abort_transaction(trans, ret);
5272 ret = btrfs_commit_transaction(trans);
5274 btrfs_abort_transaction(trans, ret);
5279 up_write(&fs_info->subvol_sem);
5280 mnt_drop_write_file(file);
5285 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
5288 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
5289 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
5292 args32 = memdup_user(arg, sizeof(*args32));
5294 return PTR_ERR(args32);
5296 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
5302 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
5303 args64->stransid = args32->stransid;
5304 args64->rtransid = args32->rtransid;
5305 args64->stime.sec = args32->stime.sec;
5306 args64->stime.nsec = args32->stime.nsec;
5307 args64->rtime.sec = args32->rtime.sec;
5308 args64->rtime.nsec = args32->rtime.nsec;
5309 args64->flags = args32->flags;
5311 ret = _btrfs_ioctl_set_received_subvol(file, args64);
5315 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
5316 args32->stransid = args64->stransid;
5317 args32->rtransid = args64->rtransid;
5318 args32->stime.sec = args64->stime.sec;
5319 args32->stime.nsec = args64->stime.nsec;
5320 args32->rtime.sec = args64->rtime.sec;
5321 args32->rtime.nsec = args64->rtime.nsec;
5322 args32->flags = args64->flags;
5324 ret = copy_to_user(arg, args32, sizeof(*args32));
5335 static long btrfs_ioctl_set_received_subvol(struct file *file,
5338 struct btrfs_ioctl_received_subvol_args *sa = NULL;
5341 sa = memdup_user(arg, sizeof(*sa));
5345 ret = _btrfs_ioctl_set_received_subvol(file, sa);
5350 ret = copy_to_user(arg, sa, sizeof(*sa));
5359 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
5361 struct inode *inode = file_inode(file);
5362 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5365 char label[BTRFS_LABEL_SIZE];
5367 spin_lock(&fs_info->super_lock);
5368 memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
5369 spin_unlock(&fs_info->super_lock);
5371 len = strnlen(label, BTRFS_LABEL_SIZE);
5373 if (len == BTRFS_LABEL_SIZE) {
5375 "label is too long, return the first %zu bytes",
5379 ret = copy_to_user(arg, label, len);
5381 return ret ? -EFAULT : 0;
5384 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5386 struct inode *inode = file_inode(file);
5387 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5388 struct btrfs_root *root = BTRFS_I(inode)->root;
5389 struct btrfs_super_block *super_block = fs_info->super_copy;
5390 struct btrfs_trans_handle *trans;
5391 char label[BTRFS_LABEL_SIZE];
5394 if (!capable(CAP_SYS_ADMIN))
5397 if (copy_from_user(label, arg, sizeof(label)))
5400 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
5402 "unable to set label with more than %d bytes",
5403 BTRFS_LABEL_SIZE - 1);
5407 ret = mnt_want_write_file(file);
5411 trans = btrfs_start_transaction(root, 0);
5412 if (IS_ERR(trans)) {
5413 ret = PTR_ERR(trans);
5417 spin_lock(&fs_info->super_lock);
5418 strcpy(super_block->label, label);
5419 spin_unlock(&fs_info->super_lock);
5420 ret = btrfs_commit_transaction(trans);
5423 mnt_drop_write_file(file);
5427 #define INIT_FEATURE_FLAGS(suffix) \
5428 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5429 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5430 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5432 int btrfs_ioctl_get_supported_features(void __user *arg)
5434 static const struct btrfs_ioctl_feature_flags features[3] = {
5435 INIT_FEATURE_FLAGS(SUPP),
5436 INIT_FEATURE_FLAGS(SAFE_SET),
5437 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5440 if (copy_to_user(arg, &features, sizeof(features)))
5446 static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5448 struct inode *inode = file_inode(file);
5449 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5450 struct btrfs_super_block *super_block = fs_info->super_copy;
5451 struct btrfs_ioctl_feature_flags features;
5453 features.compat_flags = btrfs_super_compat_flags(super_block);
5454 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5455 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5457 if (copy_to_user(arg, &features, sizeof(features)))
5463 static int check_feature_bits(struct btrfs_fs_info *fs_info,
5464 enum btrfs_feature_set set,
5465 u64 change_mask, u64 flags, u64 supported_flags,
5466 u64 safe_set, u64 safe_clear)
5468 const char *type = btrfs_feature_set_names[set];
5470 u64 disallowed, unsupported;
5471 u64 set_mask = flags & change_mask;
5472 u64 clear_mask = ~flags & change_mask;
5474 unsupported = set_mask & ~supported_flags;
5476 names = btrfs_printable_features(set, unsupported);
5479 "this kernel does not support the %s feature bit%s",
5480 names, strchr(names, ',') ? "s" : "");
5484 "this kernel does not support %s bits 0x%llx",
5489 disallowed = set_mask & ~safe_set;
5491 names = btrfs_printable_features(set, disallowed);
5494 "can't set the %s feature bit%s while mounted",
5495 names, strchr(names, ',') ? "s" : "");
5499 "can't set %s bits 0x%llx while mounted",
5504 disallowed = clear_mask & ~safe_clear;
5506 names = btrfs_printable_features(set, disallowed);
5509 "can't clear the %s feature bit%s while mounted",
5510 names, strchr(names, ',') ? "s" : "");
5514 "can't clear %s bits 0x%llx while mounted",
5522 #define check_feature(fs_info, change_mask, flags, mask_base) \
5523 check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \
5524 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5525 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5526 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5528 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5530 struct inode *inode = file_inode(file);
5531 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5532 struct btrfs_root *root = BTRFS_I(inode)->root;
5533 struct btrfs_super_block *super_block = fs_info->super_copy;
5534 struct btrfs_ioctl_feature_flags flags[2];
5535 struct btrfs_trans_handle *trans;
5539 if (!capable(CAP_SYS_ADMIN))
5542 if (copy_from_user(flags, arg, sizeof(flags)))
5546 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5547 !flags[0].incompat_flags)
5550 ret = check_feature(fs_info, flags[0].compat_flags,
5551 flags[1].compat_flags, COMPAT);
5555 ret = check_feature(fs_info, flags[0].compat_ro_flags,
5556 flags[1].compat_ro_flags, COMPAT_RO);
5560 ret = check_feature(fs_info, flags[0].incompat_flags,
5561 flags[1].incompat_flags, INCOMPAT);
5565 ret = mnt_want_write_file(file);
5569 trans = btrfs_start_transaction(root, 0);
5570 if (IS_ERR(trans)) {
5571 ret = PTR_ERR(trans);
5572 goto out_drop_write;
5575 spin_lock(&fs_info->super_lock);
5576 newflags = btrfs_super_compat_flags(super_block);
5577 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5578 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5579 btrfs_set_super_compat_flags(super_block, newflags);
5581 newflags = btrfs_super_compat_ro_flags(super_block);
5582 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5583 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5584 btrfs_set_super_compat_ro_flags(super_block, newflags);
5586 newflags = btrfs_super_incompat_flags(super_block);
5587 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5588 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5589 btrfs_set_super_incompat_flags(super_block, newflags);
5590 spin_unlock(&fs_info->super_lock);
5592 ret = btrfs_commit_transaction(trans);
5594 mnt_drop_write_file(file);
5599 long btrfs_ioctl(struct file *file, unsigned int
5600 cmd, unsigned long arg)
5602 struct inode *inode = file_inode(file);
5603 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5604 struct btrfs_root *root = BTRFS_I(inode)->root;
5605 void __user *argp = (void __user *)arg;
5608 case FS_IOC_GETFLAGS:
5609 return btrfs_ioctl_getflags(file, argp);
5610 case FS_IOC_SETFLAGS:
5611 return btrfs_ioctl_setflags(file, argp);
5612 case FS_IOC_GETVERSION:
5613 return btrfs_ioctl_getversion(file, argp);
5615 return btrfs_ioctl_fitrim(file, argp);
5616 case BTRFS_IOC_SNAP_CREATE:
5617 return btrfs_ioctl_snap_create(file, argp, 0);
5618 case BTRFS_IOC_SNAP_CREATE_V2:
5619 return btrfs_ioctl_snap_create_v2(file, argp, 0);
5620 case BTRFS_IOC_SUBVOL_CREATE:
5621 return btrfs_ioctl_snap_create(file, argp, 1);
5622 case BTRFS_IOC_SUBVOL_CREATE_V2:
5623 return btrfs_ioctl_snap_create_v2(file, argp, 1);
5624 case BTRFS_IOC_SNAP_DESTROY:
5625 return btrfs_ioctl_snap_destroy(file, argp);
5626 case BTRFS_IOC_SUBVOL_GETFLAGS:
5627 return btrfs_ioctl_subvol_getflags(file, argp);
5628 case BTRFS_IOC_SUBVOL_SETFLAGS:
5629 return btrfs_ioctl_subvol_setflags(file, argp);
5630 case BTRFS_IOC_DEFAULT_SUBVOL:
5631 return btrfs_ioctl_default_subvol(file, argp);
5632 case BTRFS_IOC_DEFRAG:
5633 return btrfs_ioctl_defrag(file, NULL);
5634 case BTRFS_IOC_DEFRAG_RANGE:
5635 return btrfs_ioctl_defrag(file, argp);
5636 case BTRFS_IOC_RESIZE:
5637 return btrfs_ioctl_resize(file, argp);
5638 case BTRFS_IOC_ADD_DEV:
5639 return btrfs_ioctl_add_dev(fs_info, argp);
5640 case BTRFS_IOC_RM_DEV:
5641 return btrfs_ioctl_rm_dev(file, argp);
5642 case BTRFS_IOC_RM_DEV_V2:
5643 return btrfs_ioctl_rm_dev_v2(file, argp);
5644 case BTRFS_IOC_FS_INFO:
5645 return btrfs_ioctl_fs_info(fs_info, argp);
5646 case BTRFS_IOC_DEV_INFO:
5647 return btrfs_ioctl_dev_info(fs_info, argp);
5648 case BTRFS_IOC_BALANCE:
5649 return btrfs_ioctl_balance(file, NULL);
5650 case BTRFS_IOC_TRANS_START:
5651 return btrfs_ioctl_trans_start(file);
5652 case BTRFS_IOC_TRANS_END:
5653 return btrfs_ioctl_trans_end(file);
5654 case BTRFS_IOC_TREE_SEARCH:
5655 return btrfs_ioctl_tree_search(file, argp);
5656 case BTRFS_IOC_TREE_SEARCH_V2:
5657 return btrfs_ioctl_tree_search_v2(file, argp);
5658 case BTRFS_IOC_INO_LOOKUP:
5659 return btrfs_ioctl_ino_lookup(file, argp);
5660 case BTRFS_IOC_INO_PATHS:
5661 return btrfs_ioctl_ino_to_path(root, argp);
5662 case BTRFS_IOC_LOGICAL_INO:
5663 return btrfs_ioctl_logical_to_ino(fs_info, argp);
5664 case BTRFS_IOC_SPACE_INFO:
5665 return btrfs_ioctl_space_info(fs_info, argp);
5666 case BTRFS_IOC_SYNC: {
5669 ret = btrfs_start_delalloc_roots(fs_info, 0, -1);
5672 ret = btrfs_sync_fs(inode->i_sb, 1);
5674 * The transaction thread may want to do more work,
5675 * namely it pokes the cleaner kthread that will start
5676 * processing uncleaned subvols.
5678 wake_up_process(fs_info->transaction_kthread);
5681 case BTRFS_IOC_START_SYNC:
5682 return btrfs_ioctl_start_sync(root, argp);
5683 case BTRFS_IOC_WAIT_SYNC:
5684 return btrfs_ioctl_wait_sync(fs_info, argp);
5685 case BTRFS_IOC_SCRUB:
5686 return btrfs_ioctl_scrub(file, argp);
5687 case BTRFS_IOC_SCRUB_CANCEL:
5688 return btrfs_ioctl_scrub_cancel(fs_info);
5689 case BTRFS_IOC_SCRUB_PROGRESS:
5690 return btrfs_ioctl_scrub_progress(fs_info, argp);
5691 case BTRFS_IOC_BALANCE_V2:
5692 return btrfs_ioctl_balance(file, argp);
5693 case BTRFS_IOC_BALANCE_CTL:
5694 return btrfs_ioctl_balance_ctl(fs_info, arg);
5695 case BTRFS_IOC_BALANCE_PROGRESS:
5696 return btrfs_ioctl_balance_progress(fs_info, argp);
5697 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5698 return btrfs_ioctl_set_received_subvol(file, argp);
5700 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5701 return btrfs_ioctl_set_received_subvol_32(file, argp);
5703 case BTRFS_IOC_SEND:
5704 return btrfs_ioctl_send(file, argp);
5705 case BTRFS_IOC_GET_DEV_STATS:
5706 return btrfs_ioctl_get_dev_stats(fs_info, argp);
5707 case BTRFS_IOC_QUOTA_CTL:
5708 return btrfs_ioctl_quota_ctl(file, argp);
5709 case BTRFS_IOC_QGROUP_ASSIGN:
5710 return btrfs_ioctl_qgroup_assign(file, argp);
5711 case BTRFS_IOC_QGROUP_CREATE:
5712 return btrfs_ioctl_qgroup_create(file, argp);
5713 case BTRFS_IOC_QGROUP_LIMIT:
5714 return btrfs_ioctl_qgroup_limit(file, argp);
5715 case BTRFS_IOC_QUOTA_RESCAN:
5716 return btrfs_ioctl_quota_rescan(file, argp);
5717 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5718 return btrfs_ioctl_quota_rescan_status(file, argp);
5719 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5720 return btrfs_ioctl_quota_rescan_wait(file, argp);
5721 case BTRFS_IOC_DEV_REPLACE:
5722 return btrfs_ioctl_dev_replace(fs_info, argp);
5723 case BTRFS_IOC_GET_FSLABEL:
5724 return btrfs_ioctl_get_fslabel(file, argp);
5725 case BTRFS_IOC_SET_FSLABEL:
5726 return btrfs_ioctl_set_fslabel(file, argp);
5727 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5728 return btrfs_ioctl_get_supported_features(argp);
5729 case BTRFS_IOC_GET_FEATURES:
5730 return btrfs_ioctl_get_features(file, argp);
5731 case BTRFS_IOC_SET_FEATURES:
5732 return btrfs_ioctl_set_features(file, argp);
5738 #ifdef CONFIG_COMPAT
5739 long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
5742 * These all access 32-bit values anyway so no further
5743 * handling is necessary.
5746 case FS_IOC32_GETFLAGS:
5747 cmd = FS_IOC_GETFLAGS;
5749 case FS_IOC32_SETFLAGS:
5750 cmd = FS_IOC_SETFLAGS;
5752 case FS_IOC32_GETVERSION:
5753 cmd = FS_IOC_GETVERSION;
5757 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));