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 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
365 q = bdev_get_queue(device->bdev);
366 if (blk_queue_discard(q)) {
368 minlen = min_t(u64, q->limits.discard_granularity,
376 if (copy_from_user(&range, arg, sizeof(range)))
380 * NOTE: Don't truncate the range using super->total_bytes. Bytenr of
381 * block group is in the logical address space, which can be any
382 * sectorsize aligned bytenr in the range [0, U64_MAX].
384 if (range.len < fs_info->sb->s_blocksize)
387 range.minlen = max(range.minlen, minlen);
388 ret = btrfs_trim_fs(fs_info, &range);
392 if (copy_to_user(arg, &range, sizeof(range)))
398 int btrfs_is_empty_uuid(u8 *uuid)
402 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
409 static noinline int create_subvol(struct inode *dir,
410 struct dentry *dentry,
411 const char *name, int namelen,
413 struct btrfs_qgroup_inherit *inherit)
415 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
416 struct btrfs_trans_handle *trans;
417 struct btrfs_key key;
418 struct btrfs_root_item *root_item;
419 struct btrfs_inode_item *inode_item;
420 struct extent_buffer *leaf;
421 struct btrfs_root *root = BTRFS_I(dir)->root;
422 struct btrfs_root *new_root;
423 struct btrfs_block_rsv block_rsv;
424 struct timespec cur_time = current_time(dir);
429 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
434 root_item = kzalloc(sizeof(*root_item), GFP_KERNEL);
438 ret = btrfs_find_free_objectid(fs_info->tree_root, &objectid);
443 * Don't create subvolume whose level is not zero. Or qgroup will be
444 * screwed up since it assumes subvolume qgroup's level to be 0.
446 if (btrfs_qgroup_level(objectid)) {
451 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
453 * The same as the snapshot creation, please see the comment
454 * of create_snapshot().
456 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
457 8, &qgroup_reserved, false);
461 trans = btrfs_start_transaction(root, 0);
463 ret = PTR_ERR(trans);
464 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
467 trans->block_rsv = &block_rsv;
468 trans->bytes_reserved = block_rsv.size;
470 ret = btrfs_qgroup_inherit(trans, fs_info, 0, objectid, inherit);
474 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
480 memzero_extent_buffer(leaf, 0, sizeof(struct btrfs_header));
481 btrfs_set_header_bytenr(leaf, leaf->start);
482 btrfs_set_header_generation(leaf, trans->transid);
483 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
484 btrfs_set_header_owner(leaf, objectid);
486 write_extent_buffer_fsid(leaf, fs_info->fsid);
487 write_extent_buffer_chunk_tree_uuid(leaf, fs_info->chunk_tree_uuid);
488 btrfs_mark_buffer_dirty(leaf);
490 inode_item = &root_item->inode;
491 btrfs_set_stack_inode_generation(inode_item, 1);
492 btrfs_set_stack_inode_size(inode_item, 3);
493 btrfs_set_stack_inode_nlink(inode_item, 1);
494 btrfs_set_stack_inode_nbytes(inode_item,
496 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
498 btrfs_set_root_flags(root_item, 0);
499 btrfs_set_root_limit(root_item, 0);
500 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
502 btrfs_set_root_bytenr(root_item, leaf->start);
503 btrfs_set_root_generation(root_item, trans->transid);
504 btrfs_set_root_level(root_item, 0);
505 btrfs_set_root_refs(root_item, 1);
506 btrfs_set_root_used(root_item, leaf->len);
507 btrfs_set_root_last_snapshot(root_item, 0);
509 btrfs_set_root_generation_v2(root_item,
510 btrfs_root_generation(root_item));
511 uuid_le_gen(&new_uuid);
512 memcpy(root_item->uuid, new_uuid.b, BTRFS_UUID_SIZE);
513 btrfs_set_stack_timespec_sec(&root_item->otime, cur_time.tv_sec);
514 btrfs_set_stack_timespec_nsec(&root_item->otime, cur_time.tv_nsec);
515 root_item->ctime = root_item->otime;
516 btrfs_set_root_ctransid(root_item, trans->transid);
517 btrfs_set_root_otransid(root_item, trans->transid);
519 btrfs_tree_unlock(leaf);
520 free_extent_buffer(leaf);
523 btrfs_set_root_dirid(root_item, new_dirid);
525 key.objectid = objectid;
527 key.type = BTRFS_ROOT_ITEM_KEY;
528 ret = btrfs_insert_root(trans, fs_info->tree_root, &key,
533 key.offset = (u64)-1;
534 new_root = btrfs_read_fs_root_no_name(fs_info, &key);
535 if (IS_ERR(new_root)) {
536 ret = PTR_ERR(new_root);
537 btrfs_abort_transaction(trans, ret);
541 btrfs_record_root_in_trans(trans, new_root);
543 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
545 /* We potentially lose an unused inode item here */
546 btrfs_abort_transaction(trans, ret);
550 mutex_lock(&new_root->objectid_mutex);
551 new_root->highest_objectid = new_dirid;
552 mutex_unlock(&new_root->objectid_mutex);
555 * insert the directory item
557 ret = btrfs_set_inode_index(BTRFS_I(dir), &index);
559 btrfs_abort_transaction(trans, ret);
563 ret = btrfs_insert_dir_item(trans, root,
564 name, namelen, BTRFS_I(dir), &key,
565 BTRFS_FT_DIR, index);
567 btrfs_abort_transaction(trans, ret);
571 btrfs_i_size_write(BTRFS_I(dir), dir->i_size + namelen * 2);
572 ret = btrfs_update_inode(trans, root, dir);
575 ret = btrfs_add_root_ref(trans, fs_info,
576 objectid, root->root_key.objectid,
577 btrfs_ino(BTRFS_I(dir)), index, name, namelen);
580 ret = btrfs_uuid_tree_add(trans, fs_info, root_item->uuid,
581 BTRFS_UUID_KEY_SUBVOL, objectid);
583 btrfs_abort_transaction(trans, ret);
587 trans->block_rsv = NULL;
588 trans->bytes_reserved = 0;
589 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
592 *async_transid = trans->transid;
593 err = btrfs_commit_transaction_async(trans, 1);
595 err = btrfs_commit_transaction(trans);
597 err = btrfs_commit_transaction(trans);
603 inode = btrfs_lookup_dentry(dir, dentry);
605 return PTR_ERR(inode);
606 d_instantiate(dentry, inode);
615 static void btrfs_wait_for_no_snapshotting_writes(struct btrfs_root *root)
621 prepare_to_wait(&root->subv_writers->wait, &wait,
622 TASK_UNINTERRUPTIBLE);
624 writers = percpu_counter_sum(&root->subv_writers->counter);
628 finish_wait(&root->subv_writers->wait, &wait);
632 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
633 struct dentry *dentry,
634 u64 *async_transid, bool readonly,
635 struct btrfs_qgroup_inherit *inherit)
637 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
639 struct btrfs_pending_snapshot *pending_snapshot;
640 struct btrfs_trans_handle *trans;
643 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
646 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_KERNEL);
647 if (!pending_snapshot)
650 pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
652 pending_snapshot->path = btrfs_alloc_path();
653 if (!pending_snapshot->root_item || !pending_snapshot->path) {
658 atomic_inc(&root->will_be_snapshotted);
659 smp_mb__after_atomic();
660 btrfs_wait_for_no_snapshotting_writes(root);
662 ret = btrfs_start_delalloc_inodes(root, 0);
666 btrfs_wait_ordered_extents(root, U64_MAX, 0, (u64)-1);
668 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
669 BTRFS_BLOCK_RSV_TEMP);
671 * 1 - parent dir inode
674 * 2 - root ref/backref
675 * 1 - root of snapshot
678 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
679 &pending_snapshot->block_rsv, 8,
680 &pending_snapshot->qgroup_reserved,
685 pending_snapshot->dentry = dentry;
686 pending_snapshot->root = root;
687 pending_snapshot->readonly = readonly;
688 pending_snapshot->dir = dir;
689 pending_snapshot->inherit = inherit;
691 trans = btrfs_start_transaction(root, 0);
693 ret = PTR_ERR(trans);
697 spin_lock(&fs_info->trans_lock);
698 list_add(&pending_snapshot->list,
699 &trans->transaction->pending_snapshots);
700 spin_unlock(&fs_info->trans_lock);
702 *async_transid = trans->transid;
703 ret = btrfs_commit_transaction_async(trans, 1);
705 ret = btrfs_commit_transaction(trans);
707 ret = btrfs_commit_transaction(trans);
712 ret = pending_snapshot->error;
716 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
720 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
722 ret = PTR_ERR(inode);
726 d_instantiate(dentry, inode);
729 btrfs_subvolume_release_metadata(fs_info, &pending_snapshot->block_rsv);
731 if (atomic_dec_and_test(&root->will_be_snapshotted))
732 wake_up_atomic_t(&root->will_be_snapshotted);
734 kfree(pending_snapshot->root_item);
735 btrfs_free_path(pending_snapshot->path);
736 kfree(pending_snapshot);
741 /* copy of may_delete in fs/namei.c()
742 * Check whether we can remove a link victim from directory dir, check
743 * whether the type of victim is right.
744 * 1. We can't do it if dir is read-only (done in permission())
745 * 2. We should have write and exec permissions on dir
746 * 3. We can't remove anything from append-only dir
747 * 4. We can't do anything with immutable dir (done in permission())
748 * 5. If the sticky bit on dir is set we should either
749 * a. be owner of dir, or
750 * b. be owner of victim, or
751 * c. have CAP_FOWNER capability
752 * 6. If the victim is append-only or immutable we can't do anything with
753 * links pointing to it.
754 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
755 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
756 * 9. We can't remove a root or mountpoint.
757 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
758 * nfs_async_unlink().
761 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
765 if (d_really_is_negative(victim))
768 BUG_ON(d_inode(victim->d_parent) != dir);
769 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
771 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
776 if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) ||
777 IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim)))
780 if (!d_is_dir(victim))
784 } else if (d_is_dir(victim))
788 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
793 /* copy of may_create in fs/namei.c() */
794 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
796 if (d_really_is_positive(child))
800 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
804 * Create a new subvolume below @parent. This is largely modeled after
805 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
806 * inside this filesystem so it's quite a bit simpler.
808 static noinline int btrfs_mksubvol(const struct path *parent,
809 const char *name, int namelen,
810 struct btrfs_root *snap_src,
811 u64 *async_transid, bool readonly,
812 struct btrfs_qgroup_inherit *inherit)
814 struct inode *dir = d_inode(parent->dentry);
815 struct btrfs_fs_info *fs_info = btrfs_sb(dir->i_sb);
816 struct dentry *dentry;
819 error = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
823 dentry = lookup_one_len(name, parent->dentry, namelen);
824 error = PTR_ERR(dentry);
828 error = btrfs_may_create(dir, dentry);
833 * even if this name doesn't exist, we may get hash collisions.
834 * check for them now when we can safely fail
836 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
842 down_read(&fs_info->subvol_sem);
844 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
848 error = create_snapshot(snap_src, dir, dentry,
849 async_transid, readonly, inherit);
851 error = create_subvol(dir, dentry, name, namelen,
852 async_transid, inherit);
855 fsnotify_mkdir(dir, dentry);
857 up_read(&fs_info->subvol_sem);
866 * When we're defragging a range, we don't want to kick it off again
867 * if it is really just waiting for delalloc to send it down.
868 * If we find a nice big extent or delalloc range for the bytes in the
869 * file you want to defrag, we return 0 to let you know to skip this
872 static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
874 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
875 struct extent_map *em = NULL;
876 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
879 read_lock(&em_tree->lock);
880 em = lookup_extent_mapping(em_tree, offset, PAGE_SIZE);
881 read_unlock(&em_tree->lock);
884 end = extent_map_end(em);
886 if (end - offset > thresh)
889 /* if we already have a nice delalloc here, just stop */
891 end = count_range_bits(io_tree, &offset, offset + thresh,
892 thresh, EXTENT_DELALLOC, 1);
899 * helper function to walk through a file and find extents
900 * newer than a specific transid, and smaller than thresh.
902 * This is used by the defragging code to find new and small
905 static int find_new_extents(struct btrfs_root *root,
906 struct inode *inode, u64 newer_than,
907 u64 *off, u32 thresh)
909 struct btrfs_path *path;
910 struct btrfs_key min_key;
911 struct extent_buffer *leaf;
912 struct btrfs_file_extent_item *extent;
915 u64 ino = btrfs_ino(BTRFS_I(inode));
917 path = btrfs_alloc_path();
921 min_key.objectid = ino;
922 min_key.type = BTRFS_EXTENT_DATA_KEY;
923 min_key.offset = *off;
926 ret = btrfs_search_forward(root, &min_key, path, newer_than);
930 if (min_key.objectid != ino)
932 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
935 leaf = path->nodes[0];
936 extent = btrfs_item_ptr(leaf, path->slots[0],
937 struct btrfs_file_extent_item);
939 type = btrfs_file_extent_type(leaf, extent);
940 if (type == BTRFS_FILE_EXTENT_REG &&
941 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
942 check_defrag_in_cache(inode, min_key.offset, thresh)) {
943 *off = min_key.offset;
944 btrfs_free_path(path);
949 if (path->slots[0] < btrfs_header_nritems(leaf)) {
950 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
954 if (min_key.offset == (u64)-1)
958 btrfs_release_path(path);
961 btrfs_free_path(path);
965 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
967 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
968 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
969 struct extent_map *em;
973 * hopefully we have this extent in the tree already, try without
974 * the full extent lock
976 read_lock(&em_tree->lock);
977 em = lookup_extent_mapping(em_tree, start, len);
978 read_unlock(&em_tree->lock);
981 struct extent_state *cached = NULL;
982 u64 end = start + len - 1;
984 /* get the big lock and read metadata off disk */
985 lock_extent_bits(io_tree, start, end, &cached);
986 em = btrfs_get_extent(BTRFS_I(inode), NULL, 0, start, len, 0);
987 unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS);
996 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
998 struct extent_map *next;
1001 /* this is the last extent */
1002 if (em->start + em->len >= i_size_read(inode))
1005 next = defrag_lookup_extent(inode, em->start + em->len);
1006 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1008 else if ((em->block_start + em->block_len == next->block_start) &&
1009 (em->block_len > SZ_128K && next->block_len > SZ_128K))
1012 free_extent_map(next);
1016 static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
1017 u64 *last_len, u64 *skip, u64 *defrag_end,
1020 struct extent_map *em;
1022 bool next_mergeable = true;
1023 bool prev_mergeable = true;
1026 * make sure that once we start defragging an extent, we keep on
1029 if (start < *defrag_end)
1034 em = defrag_lookup_extent(inode, start);
1038 /* this will cover holes, and inline extents */
1039 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1045 prev_mergeable = false;
1047 next_mergeable = defrag_check_next_extent(inode, em);
1049 * we hit a real extent, if it is big or the next extent is not a
1050 * real extent, don't bother defragging it
1052 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1053 (em->len >= thresh || (!next_mergeable && !prev_mergeable)))
1057 * last_len ends up being a counter of how many bytes we've defragged.
1058 * every time we choose not to defrag an extent, we reset *last_len
1059 * so that the next tiny extent will force a defrag.
1061 * The end result of this is that tiny extents before a single big
1062 * extent will force at least part of that big extent to be defragged.
1065 *defrag_end = extent_map_end(em);
1068 *skip = extent_map_end(em);
1072 free_extent_map(em);
1077 * it doesn't do much good to defrag one or two pages
1078 * at a time. This pulls in a nice chunk of pages
1079 * to COW and defrag.
1081 * It also makes sure the delalloc code has enough
1082 * dirty data to avoid making new small extents as part
1085 * It's a good idea to start RA on this range
1086 * before calling this.
1088 static int cluster_pages_for_defrag(struct inode *inode,
1089 struct page **pages,
1090 unsigned long start_index,
1091 unsigned long num_pages)
1093 unsigned long file_end;
1094 u64 isize = i_size_read(inode);
1101 struct btrfs_ordered_extent *ordered;
1102 struct extent_state *cached_state = NULL;
1103 struct extent_io_tree *tree;
1104 struct extent_changeset *data_reserved = NULL;
1105 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1107 file_end = (isize - 1) >> PAGE_SHIFT;
1108 if (!isize || start_index > file_end)
1111 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1113 ret = btrfs_delalloc_reserve_space(inode, &data_reserved,
1114 start_index << PAGE_SHIFT,
1115 page_cnt << PAGE_SHIFT);
1119 tree = &BTRFS_I(inode)->io_tree;
1121 /* step one, lock all the pages */
1122 for (i = 0; i < page_cnt; i++) {
1125 page = find_or_create_page(inode->i_mapping,
1126 start_index + i, mask);
1130 page_start = page_offset(page);
1131 page_end = page_start + PAGE_SIZE - 1;
1133 lock_extent_bits(tree, page_start, page_end,
1135 ordered = btrfs_lookup_ordered_extent(inode,
1137 unlock_extent_cached(tree, page_start, page_end,
1138 &cached_state, GFP_NOFS);
1143 btrfs_start_ordered_extent(inode, ordered, 1);
1144 btrfs_put_ordered_extent(ordered);
1147 * we unlocked the page above, so we need check if
1148 * it was released or not.
1150 if (page->mapping != inode->i_mapping) {
1157 if (!PageUptodate(page)) {
1158 btrfs_readpage(NULL, page);
1160 if (!PageUptodate(page)) {
1168 if (page->mapping != inode->i_mapping) {
1180 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1184 * so now we have a nice long stream of locked
1185 * and up to date pages, lets wait on them
1187 for (i = 0; i < i_done; i++)
1188 wait_on_page_writeback(pages[i]);
1190 page_start = page_offset(pages[0]);
1191 page_end = page_offset(pages[i_done - 1]) + PAGE_SIZE;
1193 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1194 page_start, page_end - 1, &cached_state);
1195 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1196 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1197 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1198 &cached_state, GFP_NOFS);
1200 if (i_done != page_cnt) {
1201 spin_lock(&BTRFS_I(inode)->lock);
1202 BTRFS_I(inode)->outstanding_extents++;
1203 spin_unlock(&BTRFS_I(inode)->lock);
1204 btrfs_delalloc_release_space(inode, data_reserved,
1205 start_index << PAGE_SHIFT,
1206 (page_cnt - i_done) << PAGE_SHIFT);
1210 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1213 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1214 page_start, page_end - 1, &cached_state,
1217 for (i = 0; i < i_done; i++) {
1218 clear_page_dirty_for_io(pages[i]);
1219 ClearPageChecked(pages[i]);
1220 set_page_extent_mapped(pages[i]);
1221 set_page_dirty(pages[i]);
1222 unlock_page(pages[i]);
1225 extent_changeset_free(data_reserved);
1228 for (i = 0; i < i_done; i++) {
1229 unlock_page(pages[i]);
1232 btrfs_delalloc_release_space(inode, data_reserved,
1233 start_index << PAGE_SHIFT,
1234 page_cnt << PAGE_SHIFT);
1235 extent_changeset_free(data_reserved);
1240 int btrfs_defrag_file(struct inode *inode, struct file *file,
1241 struct btrfs_ioctl_defrag_range_args *range,
1242 u64 newer_than, unsigned long max_to_defrag)
1244 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1245 struct btrfs_root *root = BTRFS_I(inode)->root;
1246 struct file_ra_state *ra = NULL;
1247 unsigned long last_index;
1248 u64 isize = i_size_read(inode);
1252 u64 newer_off = range->start;
1254 unsigned long ra_index = 0;
1256 int defrag_count = 0;
1257 int compress_type = BTRFS_COMPRESS_ZLIB;
1258 u32 extent_thresh = range->extent_thresh;
1259 unsigned long max_cluster = SZ_256K >> PAGE_SHIFT;
1260 unsigned long cluster = max_cluster;
1261 u64 new_align = ~((u64)SZ_128K - 1);
1262 struct page **pages = NULL;
1263 bool do_compress = range->flags & BTRFS_DEFRAG_RANGE_COMPRESS;
1268 if (range->start >= isize)
1272 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1274 if (range->compress_type)
1275 compress_type = range->compress_type;
1278 if (extent_thresh == 0)
1279 extent_thresh = SZ_256K;
1282 * If we were not given a file, allocate a readahead context. As
1283 * readahead is just an optimization, defrag will work without it so
1284 * we don't error out.
1287 ra = kzalloc(sizeof(*ra), GFP_KERNEL);
1289 file_ra_state_init(ra, inode->i_mapping);
1294 pages = kmalloc_array(max_cluster, sizeof(struct page *), GFP_KERNEL);
1300 /* find the last page to defrag */
1301 if (range->start + range->len > range->start) {
1302 last_index = min_t(u64, isize - 1,
1303 range->start + range->len - 1) >> PAGE_SHIFT;
1305 last_index = (isize - 1) >> PAGE_SHIFT;
1309 ret = find_new_extents(root, inode, newer_than,
1310 &newer_off, SZ_64K);
1312 range->start = newer_off;
1314 * we always align our defrag to help keep
1315 * the extents in the file evenly spaced
1317 i = (newer_off & new_align) >> PAGE_SHIFT;
1321 i = range->start >> PAGE_SHIFT;
1324 max_to_defrag = last_index - i + 1;
1327 * make writeback starts from i, so the defrag range can be
1328 * written sequentially.
1330 if (i < inode->i_mapping->writeback_index)
1331 inode->i_mapping->writeback_index = i;
1333 while (i <= last_index && defrag_count < max_to_defrag &&
1334 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE))) {
1336 * make sure we stop running if someone unmounts
1339 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1342 if (btrfs_defrag_cancelled(fs_info)) {
1343 btrfs_debug(fs_info, "defrag_file cancelled");
1348 if (!should_defrag_range(inode, (u64)i << PAGE_SHIFT,
1349 extent_thresh, &last_len, &skip,
1350 &defrag_end, do_compress)){
1353 * the should_defrag function tells us how much to skip
1354 * bump our counter by the suggested amount
1356 next = DIV_ROUND_UP(skip, PAGE_SIZE);
1357 i = max(i + 1, next);
1362 cluster = (PAGE_ALIGN(defrag_end) >>
1364 cluster = min(cluster, max_cluster);
1366 cluster = max_cluster;
1369 if (i + cluster > ra_index) {
1370 ra_index = max(i, ra_index);
1372 page_cache_sync_readahead(inode->i_mapping, ra,
1373 file, ra_index, cluster);
1374 ra_index += cluster;
1379 BTRFS_I(inode)->defrag_compress = compress_type;
1380 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1382 inode_unlock(inode);
1386 defrag_count += ret;
1387 balance_dirty_pages_ratelimited(inode->i_mapping);
1388 inode_unlock(inode);
1391 if (newer_off == (u64)-1)
1397 newer_off = max(newer_off + 1,
1398 (u64)i << PAGE_SHIFT);
1400 ret = find_new_extents(root, inode, newer_than,
1401 &newer_off, SZ_64K);
1403 range->start = newer_off;
1404 i = (newer_off & new_align) >> PAGE_SHIFT;
1411 last_len += ret << PAGE_SHIFT;
1419 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1420 filemap_flush(inode->i_mapping);
1421 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1422 &BTRFS_I(inode)->runtime_flags))
1423 filemap_flush(inode->i_mapping);
1427 /* the filemap_flush will queue IO into the worker threads, but
1428 * we have to make sure the IO is actually started and that
1429 * ordered extents get created before we return
1431 atomic_inc(&fs_info->async_submit_draining);
1432 while (atomic_read(&fs_info->nr_async_submits) ||
1433 atomic_read(&fs_info->async_delalloc_pages)) {
1434 wait_event(fs_info->async_submit_wait,
1435 (atomic_read(&fs_info->nr_async_submits) == 0 &&
1436 atomic_read(&fs_info->async_delalloc_pages) == 0));
1438 atomic_dec(&fs_info->async_submit_draining);
1441 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1442 btrfs_set_fs_incompat(fs_info, COMPRESS_LZO);
1443 } else if (range->compress_type == BTRFS_COMPRESS_ZSTD) {
1444 btrfs_set_fs_incompat(fs_info, COMPRESS_ZSTD);
1452 BTRFS_I(inode)->defrag_compress = BTRFS_COMPRESS_NONE;
1453 inode_unlock(inode);
1461 static noinline int btrfs_ioctl_resize(struct file *file,
1464 struct inode *inode = file_inode(file);
1465 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1469 struct btrfs_root *root = BTRFS_I(inode)->root;
1470 struct btrfs_ioctl_vol_args *vol_args;
1471 struct btrfs_trans_handle *trans;
1472 struct btrfs_device *device = NULL;
1475 char *devstr = NULL;
1479 if (!capable(CAP_SYS_ADMIN))
1482 ret = mnt_want_write_file(file);
1486 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
1487 mnt_drop_write_file(file);
1488 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1491 mutex_lock(&fs_info->volume_mutex);
1492 vol_args = memdup_user(arg, sizeof(*vol_args));
1493 if (IS_ERR(vol_args)) {
1494 ret = PTR_ERR(vol_args);
1498 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1500 sizestr = vol_args->name;
1501 devstr = strchr(sizestr, ':');
1503 sizestr = devstr + 1;
1505 devstr = vol_args->name;
1506 ret = kstrtoull(devstr, 10, &devid);
1513 btrfs_info(fs_info, "resizing devid %llu", devid);
1516 device = btrfs_find_device(fs_info, devid, NULL, NULL);
1518 btrfs_info(fs_info, "resizer unable to find device %llu",
1524 if (!device->writeable) {
1526 "resizer unable to apply on readonly device %llu",
1532 if (!strcmp(sizestr, "max"))
1533 new_size = device->bdev->bd_inode->i_size;
1535 if (sizestr[0] == '-') {
1538 } else if (sizestr[0] == '+') {
1542 new_size = memparse(sizestr, &retptr);
1543 if (*retptr != '\0' || new_size == 0) {
1549 if (device->is_tgtdev_for_dev_replace) {
1554 old_size = btrfs_device_get_total_bytes(device);
1557 if (new_size > old_size) {
1561 new_size = old_size - new_size;
1562 } else if (mod > 0) {
1563 if (new_size > ULLONG_MAX - old_size) {
1567 new_size = old_size + new_size;
1570 if (new_size < SZ_256M) {
1574 if (new_size > device->bdev->bd_inode->i_size) {
1579 new_size = round_down(new_size, fs_info->sectorsize);
1581 btrfs_info_in_rcu(fs_info, "new size for %s is %llu",
1582 rcu_str_deref(device->name), new_size);
1584 if (new_size > old_size) {
1585 trans = btrfs_start_transaction(root, 0);
1586 if (IS_ERR(trans)) {
1587 ret = PTR_ERR(trans);
1590 ret = btrfs_grow_device(trans, device, new_size);
1591 btrfs_commit_transaction(trans);
1592 } else if (new_size < old_size) {
1593 ret = btrfs_shrink_device(device, new_size);
1594 } /* equal, nothing need to do */
1599 mutex_unlock(&fs_info->volume_mutex);
1600 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
1601 mnt_drop_write_file(file);
1605 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1606 const char *name, unsigned long fd, int subvol,
1607 u64 *transid, bool readonly,
1608 struct btrfs_qgroup_inherit *inherit)
1613 if (!S_ISDIR(file_inode(file)->i_mode))
1616 ret = mnt_want_write_file(file);
1620 namelen = strlen(name);
1621 if (strchr(name, '/')) {
1623 goto out_drop_write;
1626 if (name[0] == '.' &&
1627 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1629 goto out_drop_write;
1633 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1634 NULL, transid, readonly, inherit);
1636 struct fd src = fdget(fd);
1637 struct inode *src_inode;
1640 goto out_drop_write;
1643 src_inode = file_inode(src.file);
1644 if (src_inode->i_sb != file_inode(file)->i_sb) {
1645 btrfs_info(BTRFS_I(file_inode(file))->root->fs_info,
1646 "Snapshot src from another FS");
1648 } else if (!inode_owner_or_capable(src_inode)) {
1650 * Subvolume creation is not restricted, but snapshots
1651 * are limited to own subvolumes only
1655 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1656 BTRFS_I(src_inode)->root,
1657 transid, readonly, inherit);
1662 mnt_drop_write_file(file);
1667 static noinline int btrfs_ioctl_snap_create(struct file *file,
1668 void __user *arg, int subvol)
1670 struct btrfs_ioctl_vol_args *vol_args;
1673 if (!S_ISDIR(file_inode(file)->i_mode))
1676 vol_args = memdup_user(arg, sizeof(*vol_args));
1677 if (IS_ERR(vol_args))
1678 return PTR_ERR(vol_args);
1679 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1681 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1682 vol_args->fd, subvol,
1689 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1690 void __user *arg, int subvol)
1692 struct btrfs_ioctl_vol_args_v2 *vol_args;
1696 bool readonly = false;
1697 struct btrfs_qgroup_inherit *inherit = NULL;
1699 if (!S_ISDIR(file_inode(file)->i_mode))
1702 vol_args = memdup_user(arg, sizeof(*vol_args));
1703 if (IS_ERR(vol_args))
1704 return PTR_ERR(vol_args);
1705 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1707 if (vol_args->flags &
1708 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1709 BTRFS_SUBVOL_QGROUP_INHERIT)) {
1714 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1716 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1718 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1719 if (vol_args->size > PAGE_SIZE) {
1723 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1724 if (IS_ERR(inherit)) {
1725 ret = PTR_ERR(inherit);
1730 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1731 vol_args->fd, subvol, ptr,
1736 if (ptr && copy_to_user(arg +
1737 offsetof(struct btrfs_ioctl_vol_args_v2,
1749 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1752 struct inode *inode = file_inode(file);
1753 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1754 struct btrfs_root *root = BTRFS_I(inode)->root;
1758 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID)
1761 down_read(&fs_info->subvol_sem);
1762 if (btrfs_root_readonly(root))
1763 flags |= BTRFS_SUBVOL_RDONLY;
1764 up_read(&fs_info->subvol_sem);
1766 if (copy_to_user(arg, &flags, sizeof(flags)))
1772 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1775 struct inode *inode = file_inode(file);
1776 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
1777 struct btrfs_root *root = BTRFS_I(inode)->root;
1778 struct btrfs_trans_handle *trans;
1783 if (!inode_owner_or_capable(inode))
1786 ret = mnt_want_write_file(file);
1790 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
1792 goto out_drop_write;
1795 if (copy_from_user(&flags, arg, sizeof(flags))) {
1797 goto out_drop_write;
1800 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1802 goto out_drop_write;
1805 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1807 goto out_drop_write;
1810 down_write(&fs_info->subvol_sem);
1813 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1816 root_flags = btrfs_root_flags(&root->root_item);
1817 if (flags & BTRFS_SUBVOL_RDONLY) {
1818 btrfs_set_root_flags(&root->root_item,
1819 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1822 * Block RO -> RW transition if this subvolume is involved in
1825 spin_lock(&root->root_item_lock);
1826 if (root->send_in_progress == 0) {
1827 btrfs_set_root_flags(&root->root_item,
1828 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1829 spin_unlock(&root->root_item_lock);
1831 spin_unlock(&root->root_item_lock);
1833 "Attempt to set subvolume %llu read-write during send",
1834 root->root_key.objectid);
1840 trans = btrfs_start_transaction(root, 1);
1841 if (IS_ERR(trans)) {
1842 ret = PTR_ERR(trans);
1846 ret = btrfs_update_root(trans, fs_info->tree_root,
1847 &root->root_key, &root->root_item);
1849 btrfs_end_transaction(trans);
1853 ret = btrfs_commit_transaction(trans);
1857 btrfs_set_root_flags(&root->root_item, root_flags);
1859 up_write(&fs_info->subvol_sem);
1861 mnt_drop_write_file(file);
1867 * helper to check if the subvolume references other subvolumes
1869 static noinline int may_destroy_subvol(struct btrfs_root *root)
1871 struct btrfs_fs_info *fs_info = root->fs_info;
1872 struct btrfs_path *path;
1873 struct btrfs_dir_item *di;
1874 struct btrfs_key key;
1878 path = btrfs_alloc_path();
1882 /* Make sure this root isn't set as the default subvol */
1883 dir_id = btrfs_super_root_dir(fs_info->super_copy);
1884 di = btrfs_lookup_dir_item(NULL, fs_info->tree_root, path,
1885 dir_id, "default", 7, 0);
1886 if (di && !IS_ERR(di)) {
1887 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1888 if (key.objectid == root->root_key.objectid) {
1891 "deleting default subvolume %llu is not allowed",
1895 btrfs_release_path(path);
1898 key.objectid = root->root_key.objectid;
1899 key.type = BTRFS_ROOT_REF_KEY;
1900 key.offset = (u64)-1;
1902 ret = btrfs_search_slot(NULL, fs_info->tree_root, &key, path, 0, 0);
1908 if (path->slots[0] > 0) {
1910 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1911 if (key.objectid == root->root_key.objectid &&
1912 key.type == BTRFS_ROOT_REF_KEY)
1916 btrfs_free_path(path);
1920 static noinline int key_in_sk(struct btrfs_key *key,
1921 struct btrfs_ioctl_search_key *sk)
1923 struct btrfs_key test;
1926 test.objectid = sk->min_objectid;
1927 test.type = sk->min_type;
1928 test.offset = sk->min_offset;
1930 ret = btrfs_comp_cpu_keys(key, &test);
1934 test.objectid = sk->max_objectid;
1935 test.type = sk->max_type;
1936 test.offset = sk->max_offset;
1938 ret = btrfs_comp_cpu_keys(key, &test);
1944 static noinline int copy_to_sk(struct btrfs_path *path,
1945 struct btrfs_key *key,
1946 struct btrfs_ioctl_search_key *sk,
1949 unsigned long *sk_offset,
1953 struct extent_buffer *leaf;
1954 struct btrfs_ioctl_search_header sh;
1955 struct btrfs_key test;
1956 unsigned long item_off;
1957 unsigned long item_len;
1963 leaf = path->nodes[0];
1964 slot = path->slots[0];
1965 nritems = btrfs_header_nritems(leaf);
1967 if (btrfs_header_generation(leaf) > sk->max_transid) {
1971 found_transid = btrfs_header_generation(leaf);
1973 for (i = slot; i < nritems; i++) {
1974 item_off = btrfs_item_ptr_offset(leaf, i);
1975 item_len = btrfs_item_size_nr(leaf, i);
1977 btrfs_item_key_to_cpu(leaf, key, i);
1978 if (!key_in_sk(key, sk))
1981 if (sizeof(sh) + item_len > *buf_size) {
1988 * return one empty item back for v1, which does not
1992 *buf_size = sizeof(sh) + item_len;
1997 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
2002 sh.objectid = key->objectid;
2003 sh.offset = key->offset;
2004 sh.type = key->type;
2006 sh.transid = found_transid;
2008 /* copy search result header */
2009 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
2014 *sk_offset += sizeof(sh);
2017 char __user *up = ubuf + *sk_offset;
2019 if (read_extent_buffer_to_user(leaf, up,
2020 item_off, item_len)) {
2025 *sk_offset += item_len;
2029 if (ret) /* -EOVERFLOW from above */
2032 if (*num_found >= sk->nr_items) {
2039 test.objectid = sk->max_objectid;
2040 test.type = sk->max_type;
2041 test.offset = sk->max_offset;
2042 if (btrfs_comp_cpu_keys(key, &test) >= 0)
2044 else if (key->offset < (u64)-1)
2046 else if (key->type < (u8)-1) {
2049 } else if (key->objectid < (u64)-1) {
2057 * 0: all items from this leaf copied, continue with next
2058 * 1: * more items can be copied, but unused buffer is too small
2059 * * all items were found
2060 * Either way, it will stops the loop which iterates to the next
2062 * -EOVERFLOW: item was to large for buffer
2063 * -EFAULT: could not copy extent buffer back to userspace
2068 static noinline int search_ioctl(struct inode *inode,
2069 struct btrfs_ioctl_search_key *sk,
2073 struct btrfs_fs_info *info = btrfs_sb(inode->i_sb);
2074 struct btrfs_root *root;
2075 struct btrfs_key key;
2076 struct btrfs_path *path;
2079 unsigned long sk_offset = 0;
2081 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2082 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2086 path = btrfs_alloc_path();
2090 if (sk->tree_id == 0) {
2091 /* search the root of the inode that was passed */
2092 root = BTRFS_I(inode)->root;
2094 key.objectid = sk->tree_id;
2095 key.type = BTRFS_ROOT_ITEM_KEY;
2096 key.offset = (u64)-1;
2097 root = btrfs_read_fs_root_no_name(info, &key);
2099 btrfs_free_path(path);
2104 key.objectid = sk->min_objectid;
2105 key.type = sk->min_type;
2106 key.offset = sk->min_offset;
2109 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2115 ret = copy_to_sk(path, &key, sk, buf_size, ubuf,
2116 &sk_offset, &num_found);
2117 btrfs_release_path(path);
2125 sk->nr_items = num_found;
2126 btrfs_free_path(path);
2130 static noinline int btrfs_ioctl_tree_search(struct file *file,
2133 struct btrfs_ioctl_search_args __user *uargs;
2134 struct btrfs_ioctl_search_key sk;
2135 struct inode *inode;
2139 if (!capable(CAP_SYS_ADMIN))
2142 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2144 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2147 buf_size = sizeof(uargs->buf);
2149 inode = file_inode(file);
2150 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2153 * In the origin implementation an overflow is handled by returning a
2154 * search header with a len of zero, so reset ret.
2156 if (ret == -EOVERFLOW)
2159 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2164 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2167 struct btrfs_ioctl_search_args_v2 __user *uarg;
2168 struct btrfs_ioctl_search_args_v2 args;
2169 struct inode *inode;
2172 const size_t buf_limit = SZ_16M;
2174 if (!capable(CAP_SYS_ADMIN))
2177 /* copy search header and buffer size */
2178 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2179 if (copy_from_user(&args, uarg, sizeof(args)))
2182 buf_size = args.buf_size;
2184 /* limit result size to 16MB */
2185 if (buf_size > buf_limit)
2186 buf_size = buf_limit;
2188 inode = file_inode(file);
2189 ret = search_ioctl(inode, &args.key, &buf_size,
2190 (char *)(&uarg->buf[0]));
2191 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2193 else if (ret == -EOVERFLOW &&
2194 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2201 * Search INODE_REFs to identify path name of 'dirid' directory
2202 * in a 'tree_id' tree. and sets path name to 'name'.
2204 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2205 u64 tree_id, u64 dirid, char *name)
2207 struct btrfs_root *root;
2208 struct btrfs_key key;
2214 struct btrfs_inode_ref *iref;
2215 struct extent_buffer *l;
2216 struct btrfs_path *path;
2218 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2223 path = btrfs_alloc_path();
2227 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX - 1];
2229 key.objectid = tree_id;
2230 key.type = BTRFS_ROOT_ITEM_KEY;
2231 key.offset = (u64)-1;
2232 root = btrfs_read_fs_root_no_name(info, &key);
2234 btrfs_err(info, "could not find root %llu", tree_id);
2239 key.objectid = dirid;
2240 key.type = BTRFS_INODE_REF_KEY;
2241 key.offset = (u64)-1;
2244 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2248 ret = btrfs_previous_item(root, path, dirid,
2249 BTRFS_INODE_REF_KEY);
2259 slot = path->slots[0];
2260 btrfs_item_key_to_cpu(l, &key, slot);
2262 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2263 len = btrfs_inode_ref_name_len(l, iref);
2265 total_len += len + 1;
2267 ret = -ENAMETOOLONG;
2272 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2274 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2277 btrfs_release_path(path);
2278 key.objectid = key.offset;
2279 key.offset = (u64)-1;
2280 dirid = key.objectid;
2282 memmove(name, ptr, total_len);
2283 name[total_len] = '\0';
2286 btrfs_free_path(path);
2290 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2293 struct btrfs_ioctl_ino_lookup_args *args;
2294 struct inode *inode;
2297 args = memdup_user(argp, sizeof(*args));
2299 return PTR_ERR(args);
2301 inode = file_inode(file);
2304 * Unprivileged query to obtain the containing subvolume root id. The
2305 * path is reset so it's consistent with btrfs_search_path_in_tree.
2307 if (args->treeid == 0)
2308 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2310 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2315 if (!capable(CAP_SYS_ADMIN)) {
2320 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2321 args->treeid, args->objectid,
2325 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2332 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2335 struct dentry *parent = file->f_path.dentry;
2336 struct btrfs_fs_info *fs_info = btrfs_sb(parent->d_sb);
2337 struct dentry *dentry;
2338 struct inode *dir = d_inode(parent);
2339 struct inode *inode;
2340 struct btrfs_root *root = BTRFS_I(dir)->root;
2341 struct btrfs_root *dest = NULL;
2342 struct btrfs_ioctl_vol_args *vol_args;
2343 struct btrfs_trans_handle *trans;
2344 struct btrfs_block_rsv block_rsv;
2346 u64 qgroup_reserved;
2351 if (!S_ISDIR(dir->i_mode))
2354 vol_args = memdup_user(arg, sizeof(*vol_args));
2355 if (IS_ERR(vol_args))
2356 return PTR_ERR(vol_args);
2358 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2359 namelen = strlen(vol_args->name);
2360 if (strchr(vol_args->name, '/') ||
2361 strncmp(vol_args->name, "..", namelen) == 0) {
2366 err = mnt_want_write_file(file);
2371 err = down_write_killable_nested(&dir->i_rwsem, I_MUTEX_PARENT);
2373 goto out_drop_write;
2374 dentry = lookup_one_len(vol_args->name, parent, namelen);
2375 if (IS_ERR(dentry)) {
2376 err = PTR_ERR(dentry);
2377 goto out_unlock_dir;
2380 if (d_really_is_negative(dentry)) {
2385 inode = d_inode(dentry);
2386 dest = BTRFS_I(inode)->root;
2387 if (!capable(CAP_SYS_ADMIN)) {
2389 * Regular user. Only allow this with a special mount
2390 * option, when the user has write+exec access to the
2391 * subvol root, and when rmdir(2) would have been
2394 * Note that this is _not_ check that the subvol is
2395 * empty or doesn't contain data that we wouldn't
2396 * otherwise be able to delete.
2398 * Users who want to delete empty subvols should try
2402 if (!btrfs_test_opt(fs_info, USER_SUBVOL_RM_ALLOWED))
2406 * Do not allow deletion if the parent dir is the same
2407 * as the dir to be deleted. That means the ioctl
2408 * must be called on the dentry referencing the root
2409 * of the subvol, not a random directory contained
2416 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2421 /* check if subvolume may be deleted by a user */
2422 err = btrfs_may_delete(dir, dentry, 1);
2426 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
2434 * Don't allow to delete a subvolume with send in progress. This is
2435 * inside the i_mutex so the error handling that has to drop the bit
2436 * again is not run concurrently.
2438 spin_lock(&dest->root_item_lock);
2439 root_flags = btrfs_root_flags(&dest->root_item);
2440 if (dest->send_in_progress == 0) {
2441 btrfs_set_root_flags(&dest->root_item,
2442 root_flags | BTRFS_ROOT_SUBVOL_DEAD);
2443 spin_unlock(&dest->root_item_lock);
2445 spin_unlock(&dest->root_item_lock);
2447 "Attempt to delete subvolume %llu during send",
2448 dest->root_key.objectid);
2450 goto out_unlock_inode;
2453 down_write(&fs_info->subvol_sem);
2455 err = may_destroy_subvol(dest);
2459 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2461 * One for dir inode, two for dir entries, two for root
2464 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2465 5, &qgroup_reserved, true);
2469 trans = btrfs_start_transaction(root, 0);
2470 if (IS_ERR(trans)) {
2471 err = PTR_ERR(trans);
2474 trans->block_rsv = &block_rsv;
2475 trans->bytes_reserved = block_rsv.size;
2477 btrfs_record_snapshot_destroy(trans, BTRFS_I(dir));
2479 ret = btrfs_unlink_subvol(trans, root, dir,
2480 dest->root_key.objectid,
2481 dentry->d_name.name,
2482 dentry->d_name.len);
2485 btrfs_abort_transaction(trans, ret);
2489 btrfs_record_root_in_trans(trans, dest);
2491 memset(&dest->root_item.drop_progress, 0,
2492 sizeof(dest->root_item.drop_progress));
2493 dest->root_item.drop_level = 0;
2494 btrfs_set_root_refs(&dest->root_item, 0);
2496 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
2497 ret = btrfs_insert_orphan_item(trans,
2499 dest->root_key.objectid);
2501 btrfs_abort_transaction(trans, ret);
2507 ret = btrfs_uuid_tree_rem(trans, fs_info, dest->root_item.uuid,
2508 BTRFS_UUID_KEY_SUBVOL,
2509 dest->root_key.objectid);
2510 if (ret && ret != -ENOENT) {
2511 btrfs_abort_transaction(trans, ret);
2515 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2516 ret = btrfs_uuid_tree_rem(trans, fs_info,
2517 dest->root_item.received_uuid,
2518 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2519 dest->root_key.objectid);
2520 if (ret && ret != -ENOENT) {
2521 btrfs_abort_transaction(trans, ret);
2528 trans->block_rsv = NULL;
2529 trans->bytes_reserved = 0;
2530 ret = btrfs_end_transaction(trans);
2533 inode->i_flags |= S_DEAD;
2535 btrfs_subvolume_release_metadata(fs_info, &block_rsv);
2537 up_write(&fs_info->subvol_sem);
2539 spin_lock(&dest->root_item_lock);
2540 root_flags = btrfs_root_flags(&dest->root_item);
2541 btrfs_set_root_flags(&dest->root_item,
2542 root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
2543 spin_unlock(&dest->root_item_lock);
2546 inode_unlock(inode);
2548 d_invalidate(dentry);
2549 btrfs_invalidate_inodes(dest);
2551 ASSERT(dest->send_in_progress == 0);
2554 if (dest->ino_cache_inode) {
2555 iput(dest->ino_cache_inode);
2556 dest->ino_cache_inode = NULL;
2564 mnt_drop_write_file(file);
2570 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2572 struct inode *inode = file_inode(file);
2573 struct btrfs_root *root = BTRFS_I(inode)->root;
2574 struct btrfs_ioctl_defrag_range_args *range;
2577 ret = mnt_want_write_file(file);
2581 if (btrfs_root_readonly(root)) {
2586 switch (inode->i_mode & S_IFMT) {
2588 if (!capable(CAP_SYS_ADMIN)) {
2592 ret = btrfs_defrag_root(root);
2595 if (!(file->f_mode & FMODE_WRITE)) {
2600 range = kzalloc(sizeof(*range), GFP_KERNEL);
2607 if (copy_from_user(range, argp,
2613 /* compression requires us to start the IO */
2614 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2615 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2616 range->extent_thresh = (u32)-1;
2619 /* the rest are all set to zero by kzalloc */
2620 range->len = (u64)-1;
2622 ret = btrfs_defrag_file(file_inode(file), file,
2632 mnt_drop_write_file(file);
2636 static long btrfs_ioctl_add_dev(struct btrfs_fs_info *fs_info, void __user *arg)
2638 struct btrfs_ioctl_vol_args *vol_args;
2641 if (!capable(CAP_SYS_ADMIN))
2644 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags))
2645 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2647 mutex_lock(&fs_info->volume_mutex);
2648 vol_args = memdup_user(arg, sizeof(*vol_args));
2649 if (IS_ERR(vol_args)) {
2650 ret = PTR_ERR(vol_args);
2654 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2655 ret = btrfs_init_new_device(fs_info, vol_args->name);
2658 btrfs_info(fs_info, "disk added %s", vol_args->name);
2662 mutex_unlock(&fs_info->volume_mutex);
2663 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
2667 static long btrfs_ioctl_rm_dev_v2(struct file *file, void __user *arg)
2669 struct inode *inode = file_inode(file);
2670 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2671 struct btrfs_ioctl_vol_args_v2 *vol_args;
2674 if (!capable(CAP_SYS_ADMIN))
2677 ret = mnt_want_write_file(file);
2681 vol_args = memdup_user(arg, sizeof(*vol_args));
2682 if (IS_ERR(vol_args)) {
2683 ret = PTR_ERR(vol_args);
2687 /* Check for compatibility reject unknown flags */
2688 if (vol_args->flags & ~BTRFS_VOL_ARG_V2_FLAGS_SUPPORTED) {
2693 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
2694 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2698 mutex_lock(&fs_info->volume_mutex);
2699 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID) {
2700 ret = btrfs_rm_device(fs_info, NULL, vol_args->devid);
2702 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
2703 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
2705 mutex_unlock(&fs_info->volume_mutex);
2706 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
2709 if (vol_args->flags & BTRFS_DEVICE_SPEC_BY_ID)
2710 btrfs_info(fs_info, "device deleted: id %llu",
2713 btrfs_info(fs_info, "device deleted: %s",
2719 mnt_drop_write_file(file);
2723 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2725 struct inode *inode = file_inode(file);
2726 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
2727 struct btrfs_ioctl_vol_args *vol_args;
2730 if (!capable(CAP_SYS_ADMIN))
2733 ret = mnt_want_write_file(file);
2737 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
2738 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2739 goto out_drop_write;
2742 vol_args = memdup_user(arg, sizeof(*vol_args));
2743 if (IS_ERR(vol_args)) {
2744 ret = PTR_ERR(vol_args);
2748 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2749 mutex_lock(&fs_info->volume_mutex);
2750 ret = btrfs_rm_device(fs_info, vol_args->name, 0);
2751 mutex_unlock(&fs_info->volume_mutex);
2754 btrfs_info(fs_info, "disk deleted %s", vol_args->name);
2757 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
2759 mnt_drop_write_file(file);
2764 static long btrfs_ioctl_fs_info(struct btrfs_fs_info *fs_info,
2767 struct btrfs_ioctl_fs_info_args *fi_args;
2768 struct btrfs_device *device;
2769 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2772 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2776 mutex_lock(&fs_devices->device_list_mutex);
2777 fi_args->num_devices = fs_devices->num_devices;
2778 memcpy(&fi_args->fsid, fs_info->fsid, sizeof(fi_args->fsid));
2780 list_for_each_entry(device, &fs_devices->devices, dev_list) {
2781 if (device->devid > fi_args->max_id)
2782 fi_args->max_id = device->devid;
2784 mutex_unlock(&fs_devices->device_list_mutex);
2786 fi_args->nodesize = fs_info->nodesize;
2787 fi_args->sectorsize = fs_info->sectorsize;
2788 fi_args->clone_alignment = fs_info->sectorsize;
2790 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2797 static long btrfs_ioctl_dev_info(struct btrfs_fs_info *fs_info,
2800 struct btrfs_ioctl_dev_info_args *di_args;
2801 struct btrfs_device *dev;
2802 struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
2804 char *s_uuid = NULL;
2806 di_args = memdup_user(arg, sizeof(*di_args));
2807 if (IS_ERR(di_args))
2808 return PTR_ERR(di_args);
2810 if (!btrfs_is_empty_uuid(di_args->uuid))
2811 s_uuid = di_args->uuid;
2813 mutex_lock(&fs_devices->device_list_mutex);
2814 dev = btrfs_find_device(fs_info, di_args->devid, s_uuid, NULL);
2821 di_args->devid = dev->devid;
2822 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2823 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
2824 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2826 struct rcu_string *name;
2829 name = rcu_dereference(dev->name);
2830 strncpy(di_args->path, name->str, sizeof(di_args->path));
2832 di_args->path[sizeof(di_args->path) - 1] = 0;
2834 di_args->path[0] = '\0';
2838 mutex_unlock(&fs_devices->device_list_mutex);
2839 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2846 static struct page *extent_same_get_page(struct inode *inode, pgoff_t index)
2850 page = grab_cache_page(inode->i_mapping, index);
2852 return ERR_PTR(-ENOMEM);
2854 if (!PageUptodate(page)) {
2857 ret = btrfs_readpage(NULL, page);
2859 return ERR_PTR(ret);
2861 if (!PageUptodate(page)) {
2864 return ERR_PTR(-EIO);
2866 if (page->mapping != inode->i_mapping) {
2869 return ERR_PTR(-EAGAIN);
2876 static int gather_extent_pages(struct inode *inode, struct page **pages,
2877 int num_pages, u64 off)
2880 pgoff_t index = off >> PAGE_SHIFT;
2882 for (i = 0; i < num_pages; i++) {
2884 pages[i] = extent_same_get_page(inode, index + i);
2885 if (IS_ERR(pages[i])) {
2886 int err = PTR_ERR(pages[i]);
2897 static int lock_extent_range(struct inode *inode, u64 off, u64 len,
2898 bool retry_range_locking)
2901 * Do any pending delalloc/csum calculations on inode, one way or
2902 * another, and lock file content.
2903 * The locking order is:
2906 * 2) range in the inode's io tree
2909 struct btrfs_ordered_extent *ordered;
2910 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2911 ordered = btrfs_lookup_first_ordered_extent(inode,
2914 ordered->file_offset + ordered->len <= off ||
2915 ordered->file_offset >= off + len) &&
2916 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2917 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2919 btrfs_put_ordered_extent(ordered);
2922 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2924 btrfs_put_ordered_extent(ordered);
2925 if (!retry_range_locking)
2927 btrfs_wait_ordered_range(inode, off, len);
2932 static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2)
2934 inode_unlock(inode1);
2935 inode_unlock(inode2);
2938 static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2)
2940 if (inode1 < inode2)
2941 swap(inode1, inode2);
2943 inode_lock_nested(inode1, I_MUTEX_PARENT);
2944 inode_lock_nested(inode2, I_MUTEX_CHILD);
2947 static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1,
2948 struct inode *inode2, u64 loff2, u64 len)
2950 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2951 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2954 static int btrfs_double_extent_lock(struct inode *inode1, u64 loff1,
2955 struct inode *inode2, u64 loff2, u64 len,
2956 bool retry_range_locking)
2960 if (inode1 < inode2) {
2961 swap(inode1, inode2);
2964 ret = lock_extent_range(inode1, loff1, len, retry_range_locking);
2967 ret = lock_extent_range(inode2, loff2, len, retry_range_locking);
2969 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1,
2976 struct page **src_pages;
2977 struct page **dst_pages;
2980 static void btrfs_cmp_data_free(struct cmp_pages *cmp)
2985 for (i = 0; i < cmp->num_pages; i++) {
2986 pg = cmp->src_pages[i];
2991 pg = cmp->dst_pages[i];
2997 kfree(cmp->src_pages);
2998 kfree(cmp->dst_pages);
3001 static int btrfs_cmp_data_prepare(struct inode *src, u64 loff,
3002 struct inode *dst, u64 dst_loff,
3003 u64 len, struct cmp_pages *cmp)
3006 int num_pages = PAGE_ALIGN(len) >> PAGE_SHIFT;
3007 struct page **src_pgarr, **dst_pgarr;
3010 * We must gather up all the pages before we initiate our
3011 * extent locking. We use an array for the page pointers. Size
3012 * of the array is bounded by len, which is in turn bounded by
3013 * BTRFS_MAX_DEDUPE_LEN.
3015 src_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL);
3016 dst_pgarr = kcalloc(num_pages, sizeof(struct page *), GFP_KERNEL);
3017 if (!src_pgarr || !dst_pgarr) {
3022 cmp->num_pages = num_pages;
3023 cmp->src_pages = src_pgarr;
3024 cmp->dst_pages = dst_pgarr;
3027 * If deduping ranges in the same inode, locking rules make it mandatory
3028 * to always lock pages in ascending order to avoid deadlocks with
3029 * concurrent tasks (such as starting writeback/delalloc).
3031 if (src == dst && dst_loff < loff) {
3032 swap(src_pgarr, dst_pgarr);
3033 swap(loff, dst_loff);
3036 ret = gather_extent_pages(src, src_pgarr, cmp->num_pages, loff);
3040 ret = gather_extent_pages(dst, dst_pgarr, cmp->num_pages, dst_loff);
3044 btrfs_cmp_data_free(cmp);
3048 static int btrfs_cmp_data(u64 len, struct cmp_pages *cmp)
3052 struct page *src_page, *dst_page;
3053 unsigned int cmp_len = PAGE_SIZE;
3054 void *addr, *dst_addr;
3058 if (len < PAGE_SIZE)
3061 BUG_ON(i >= cmp->num_pages);
3063 src_page = cmp->src_pages[i];
3064 dst_page = cmp->dst_pages[i];
3065 ASSERT(PageLocked(src_page));
3066 ASSERT(PageLocked(dst_page));
3068 addr = kmap_atomic(src_page);
3069 dst_addr = kmap_atomic(dst_page);
3071 flush_dcache_page(src_page);
3072 flush_dcache_page(dst_page);
3074 if (memcmp(addr, dst_addr, cmp_len))
3077 kunmap_atomic(addr);
3078 kunmap_atomic(dst_addr);
3090 static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen,
3094 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
3096 if (off + olen > inode->i_size || off + olen < off)
3099 /* if we extend to eof, continue to block boundary */
3100 if (off + len == inode->i_size)
3101 *plen = len = ALIGN(inode->i_size, bs) - off;
3103 /* Check that we are block aligned - btrfs_clone() requires this */
3104 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
3110 static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen,
3111 struct inode *dst, u64 dst_loff)
3115 struct cmp_pages cmp;
3116 bool same_inode = (src == dst);
3117 u64 same_lock_start = 0;
3118 u64 same_lock_len = 0;
3126 btrfs_double_inode_lock(src, dst);
3128 ret = extent_same_check_offsets(src, loff, &len, olen);
3132 ret = extent_same_check_offsets(dst, dst_loff, &len, olen);
3138 * Single inode case wants the same checks, except we
3139 * don't want our length pushed out past i_size as
3140 * comparing that data range makes no sense.
3142 * extent_same_check_offsets() will do this for an
3143 * unaligned length at i_size, so catch it here and
3144 * reject the request.
3146 * This effectively means we require aligned extents
3147 * for the single-inode case, whereas the other cases
3148 * allow an unaligned length so long as it ends at
3156 /* Check for overlapping ranges */
3157 if (dst_loff + len > loff && dst_loff < loff + len) {
3162 same_lock_start = min_t(u64, loff, dst_loff);
3163 same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start;
3166 * If the source and destination inodes are different, the
3167 * source's range end offset matches the source's i_size, that
3168 * i_size is not a multiple of the sector size, and the
3169 * destination range does not go past the destination's i_size,
3170 * we must round down the length to the nearest sector size
3171 * multiple. If we don't do this adjustment we end replacing
3172 * with zeroes the bytes in the range that starts at the
3173 * deduplication range's end offset and ends at the next sector
3176 if (loff + olen == i_size_read(src) &&
3177 dst_loff + len < i_size_read(dst)) {
3178 const u64 sz = BTRFS_I(src)->root->fs_info->sectorsize;
3180 len = round_down(i_size_read(src), sz) - loff;
3187 /* don't make the dst file partly checksummed */
3188 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3189 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
3195 ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, &cmp);
3200 ret = lock_extent_range(src, same_lock_start, same_lock_len,
3203 ret = btrfs_double_extent_lock(src, loff, dst, dst_loff, len,
3206 * If one of the inodes has dirty pages in the respective range or
3207 * ordered extents, we need to flush dellaloc and wait for all ordered
3208 * extents in the range. We must unlock the pages and the ranges in the
3209 * io trees to avoid deadlocks when flushing delalloc (requires locking
3210 * pages) and when waiting for ordered extents to complete (they require
3213 if (ret == -EAGAIN) {
3215 * Ranges in the io trees already unlocked. Now unlock all
3216 * pages before waiting for all IO to complete.
3218 btrfs_cmp_data_free(&cmp);
3220 btrfs_wait_ordered_range(src, same_lock_start,
3223 btrfs_wait_ordered_range(src, loff, len);
3224 btrfs_wait_ordered_range(dst, dst_loff, len);
3230 /* ranges in the io trees already unlocked */
3231 btrfs_cmp_data_free(&cmp);
3235 /* pass original length for comparison so we stay within i_size */
3236 ret = btrfs_cmp_data(olen, &cmp);
3238 ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1);
3241 unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start,
3242 same_lock_start + same_lock_len - 1);
3244 btrfs_double_extent_unlock(src, loff, dst, dst_loff, len);
3246 btrfs_cmp_data_free(&cmp);
3251 btrfs_double_inode_unlock(src, dst);
3256 #define BTRFS_MAX_DEDUPE_LEN SZ_16M
3258 ssize_t btrfs_dedupe_file_range(struct file *src_file, u64 loff, u64 olen,
3259 struct file *dst_file, u64 dst_loff)
3261 struct inode *src = file_inode(src_file);
3262 struct inode *dst = file_inode(dst_file);
3263 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
3266 if (olen > BTRFS_MAX_DEDUPE_LEN)
3267 olen = BTRFS_MAX_DEDUPE_LEN;
3269 if (WARN_ON_ONCE(bs < PAGE_SIZE)) {
3271 * Btrfs does not support blocksize < page_size. As a
3272 * result, btrfs_cmp_data() won't correctly handle
3273 * this situation without an update.
3278 res = btrfs_extent_same(src, loff, olen, dst, dst_loff);
3284 static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3285 struct inode *inode,
3291 struct btrfs_root *root = BTRFS_I(inode)->root;
3294 inode_inc_iversion(inode);
3295 if (!no_time_update)
3296 inode->i_mtime = inode->i_ctime = current_time(inode);
3298 * We round up to the block size at eof when determining which
3299 * extents to clone above, but shouldn't round up the file size.
3301 if (endoff > destoff + olen)
3302 endoff = destoff + olen;
3303 if (endoff > inode->i_size)
3304 btrfs_i_size_write(BTRFS_I(inode), endoff);
3306 ret = btrfs_update_inode(trans, root, inode);
3308 btrfs_abort_transaction(trans, ret);
3309 btrfs_end_transaction(trans);
3312 ret = btrfs_end_transaction(trans);
3317 static void clone_update_extent_map(struct btrfs_inode *inode,
3318 const struct btrfs_trans_handle *trans,
3319 const struct btrfs_path *path,
3320 const u64 hole_offset,
3323 struct extent_map_tree *em_tree = &inode->extent_tree;
3324 struct extent_map *em;
3327 em = alloc_extent_map();
3329 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags);
3334 struct btrfs_file_extent_item *fi;
3336 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3337 struct btrfs_file_extent_item);
3338 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3339 em->generation = -1;
3340 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3341 BTRFS_FILE_EXTENT_INLINE)
3342 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3343 &inode->runtime_flags);
3345 em->start = hole_offset;
3347 em->ram_bytes = em->len;
3348 em->orig_start = hole_offset;
3349 em->block_start = EXTENT_MAP_HOLE;
3351 em->orig_block_len = 0;
3352 em->compress_type = BTRFS_COMPRESS_NONE;
3353 em->generation = trans->transid;
3357 write_lock(&em_tree->lock);
3358 ret = add_extent_mapping(em_tree, em, 1);
3359 write_unlock(&em_tree->lock);
3360 if (ret != -EEXIST) {
3361 free_extent_map(em);
3364 btrfs_drop_extent_cache(inode, em->start,
3365 em->start + em->len - 1, 0);
3369 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags);
3373 * Make sure we do not end up inserting an inline extent into a file that has
3374 * already other (non-inline) extents. If a file has an inline extent it can
3375 * not have any other extents and the (single) inline extent must start at the
3376 * file offset 0. Failing to respect these rules will lead to file corruption,
3377 * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc
3379 * We can have extents that have been already written to disk or we can have
3380 * dirty ranges still in delalloc, in which case the extent maps and items are
3381 * created only when we run delalloc, and the delalloc ranges might fall outside
3382 * the range we are currently locking in the inode's io tree. So we check the
3383 * inode's i_size because of that (i_size updates are done while holding the
3384 * i_mutex, which we are holding here).
3385 * We also check to see if the inode has a size not greater than "datal" but has
3386 * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are
3387 * protected against such concurrent fallocate calls by the i_mutex).
3389 * If the file has no extents but a size greater than datal, do not allow the
3390 * copy because we would need turn the inline extent into a non-inline one (even
3391 * with NO_HOLES enabled). If we find our destination inode only has one inline
3392 * extent, just overwrite it with the source inline extent if its size is less
3393 * than the source extent's size, or we could copy the source inline extent's
3394 * data into the destination inode's inline extent if the later is greater then
3397 static int clone_copy_inline_extent(struct inode *dst,
3398 struct btrfs_trans_handle *trans,
3399 struct btrfs_path *path,
3400 struct btrfs_key *new_key,
3401 const u64 drop_start,
3407 struct btrfs_fs_info *fs_info = btrfs_sb(dst->i_sb);
3408 struct btrfs_root *root = BTRFS_I(dst)->root;
3409 const u64 aligned_end = ALIGN(new_key->offset + datal,
3410 fs_info->sectorsize);
3412 struct btrfs_key key;
3414 if (new_key->offset > 0)
3417 key.objectid = btrfs_ino(BTRFS_I(dst));
3418 key.type = BTRFS_EXTENT_DATA_KEY;
3420 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3423 } else if (ret > 0) {
3424 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
3425 ret = btrfs_next_leaf(root, path);
3429 goto copy_inline_extent;
3431 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
3432 if (key.objectid == btrfs_ino(BTRFS_I(dst)) &&
3433 key.type == BTRFS_EXTENT_DATA_KEY) {
3434 ASSERT(key.offset > 0);
3437 } else if (i_size_read(dst) <= datal) {
3438 struct btrfs_file_extent_item *ei;
3442 * If the file size is <= datal, make sure there are no other
3443 * extents following (can happen do to an fallocate call with
3444 * the flag FALLOC_FL_KEEP_SIZE).
3446 ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
3447 struct btrfs_file_extent_item);
3449 * If it's an inline extent, it can not have other extents
3452 if (btrfs_file_extent_type(path->nodes[0], ei) ==
3453 BTRFS_FILE_EXTENT_INLINE)
3454 goto copy_inline_extent;
3456 ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei);
3457 if (ext_len > aligned_end)
3460 ret = btrfs_next_item(root, path);
3463 } else if (ret == 0) {
3464 btrfs_item_key_to_cpu(path->nodes[0], &key,
3466 if (key.objectid == btrfs_ino(BTRFS_I(dst)) &&
3467 key.type == BTRFS_EXTENT_DATA_KEY)
3474 * We have no extent items, or we have an extent at offset 0 which may
3475 * or may not be inlined. All these cases are dealt the same way.
3477 if (i_size_read(dst) > datal) {
3479 * If the destination inode has an inline extent...
3480 * This would require copying the data from the source inline
3481 * extent into the beginning of the destination's inline extent.
3482 * But this is really complex, both extents can be compressed
3483 * or just one of them, which would require decompressing and
3484 * re-compressing data (which could increase the new compressed
3485 * size, not allowing the compressed data to fit anymore in an
3487 * So just don't support this case for now (it should be rare,
3488 * we are not really saving space when cloning inline extents).
3493 btrfs_release_path(path);
3494 ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1);
3497 ret = btrfs_insert_empty_item(trans, root, path, new_key, size);
3502 const u32 start = btrfs_file_extent_calc_inline_size(0);
3504 memmove(inline_data + start, inline_data + start + skip, datal);
3507 write_extent_buffer(path->nodes[0], inline_data,
3508 btrfs_item_ptr_offset(path->nodes[0],
3511 inode_add_bytes(dst, datal);
3517 * btrfs_clone() - clone a range from inode file to another
3519 * @src: Inode to clone from
3520 * @inode: Inode to clone to
3521 * @off: Offset within source to start clone from
3522 * @olen: Original length, passed by user, of range to clone
3523 * @olen_aligned: Block-aligned value of olen
3524 * @destoff: Offset within @inode to start clone
3525 * @no_time_update: Whether to update mtime/ctime on the target inode
3527 static int btrfs_clone(struct inode *src, struct inode *inode,
3528 const u64 off, const u64 olen, const u64 olen_aligned,
3529 const u64 destoff, int no_time_update)
3531 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3532 struct btrfs_root *root = BTRFS_I(inode)->root;
3533 struct btrfs_path *path = NULL;
3534 struct extent_buffer *leaf;
3535 struct btrfs_trans_handle *trans;
3537 struct btrfs_key key;
3541 const u64 len = olen_aligned;
3542 u64 last_dest_end = destoff;
3545 buf = kvmalloc(fs_info->nodesize, GFP_KERNEL);
3549 path = btrfs_alloc_path();
3555 path->reada = READA_FORWARD;
3557 key.objectid = btrfs_ino(BTRFS_I(src));
3558 key.type = BTRFS_EXTENT_DATA_KEY;
3562 u64 next_key_min_offset = key.offset + 1;
3565 * note the key will change type as we walk through the
3568 path->leave_spinning = 1;
3569 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3574 * First search, if no extent item that starts at offset off was
3575 * found but the previous item is an extent item, it's possible
3576 * it might overlap our target range, therefore process it.
3578 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3579 btrfs_item_key_to_cpu(path->nodes[0], &key,
3580 path->slots[0] - 1);
3581 if (key.type == BTRFS_EXTENT_DATA_KEY)
3585 nritems = btrfs_header_nritems(path->nodes[0]);
3587 if (path->slots[0] >= nritems) {
3588 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
3593 nritems = btrfs_header_nritems(path->nodes[0]);
3595 leaf = path->nodes[0];
3596 slot = path->slots[0];
3598 btrfs_item_key_to_cpu(leaf, &key, slot);
3599 if (key.type > BTRFS_EXTENT_DATA_KEY ||
3600 key.objectid != btrfs_ino(BTRFS_I(src)))
3603 if (key.type == BTRFS_EXTENT_DATA_KEY) {
3604 struct btrfs_file_extent_item *extent;
3607 struct btrfs_key new_key;
3608 u64 disko = 0, diskl = 0;
3609 u64 datao = 0, datal = 0;
3613 extent = btrfs_item_ptr(leaf, slot,
3614 struct btrfs_file_extent_item);
3615 comp = btrfs_file_extent_compression(leaf, extent);
3616 type = btrfs_file_extent_type(leaf, extent);
3617 if (type == BTRFS_FILE_EXTENT_REG ||
3618 type == BTRFS_FILE_EXTENT_PREALLOC) {
3619 disko = btrfs_file_extent_disk_bytenr(leaf,
3621 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3623 datao = btrfs_file_extent_offset(leaf, extent);
3624 datal = btrfs_file_extent_num_bytes(leaf,
3626 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3627 /* take upper bound, may be compressed */
3628 datal = btrfs_file_extent_ram_bytes(leaf,
3633 * The first search might have left us at an extent
3634 * item that ends before our target range's start, can
3635 * happen if we have holes and NO_HOLES feature enabled.
3637 if (key.offset + datal <= off) {
3640 } else if (key.offset >= off + len) {
3643 next_key_min_offset = key.offset + datal;
3644 size = btrfs_item_size_nr(leaf, slot);
3645 read_extent_buffer(leaf, buf,
3646 btrfs_item_ptr_offset(leaf, slot),
3649 btrfs_release_path(path);
3650 path->leave_spinning = 0;
3652 memcpy(&new_key, &key, sizeof(new_key));
3653 new_key.objectid = btrfs_ino(BTRFS_I(inode));
3654 if (off <= key.offset)
3655 new_key.offset = key.offset + destoff - off;
3657 new_key.offset = destoff;
3660 * Deal with a hole that doesn't have an extent item
3661 * that represents it (NO_HOLES feature enabled).
3662 * This hole is either in the middle of the cloning
3663 * range or at the beginning (fully overlaps it or
3664 * partially overlaps it).
3666 if (new_key.offset != last_dest_end)
3667 drop_start = last_dest_end;
3669 drop_start = new_key.offset;
3672 * 1 - adjusting old extent (we may have to split it)
3673 * 1 - add new extent
3676 trans = btrfs_start_transaction(root, 3);
3677 if (IS_ERR(trans)) {
3678 ret = PTR_ERR(trans);
3682 if (type == BTRFS_FILE_EXTENT_REG ||
3683 type == BTRFS_FILE_EXTENT_PREALLOC) {
3685 * a | --- range to clone ---| b
3686 * | ------------- extent ------------- |
3689 /* subtract range b */
3690 if (key.offset + datal > off + len)
3691 datal = off + len - key.offset;
3693 /* subtract range a */
3694 if (off > key.offset) {
3695 datao += off - key.offset;
3696 datal -= off - key.offset;
3699 ret = btrfs_drop_extents(trans, root, inode,
3701 new_key.offset + datal,
3704 if (ret != -EOPNOTSUPP)
3705 btrfs_abort_transaction(trans,
3707 btrfs_end_transaction(trans);
3711 ret = btrfs_insert_empty_item(trans, root, path,
3714 btrfs_abort_transaction(trans, ret);
3715 btrfs_end_transaction(trans);
3719 leaf = path->nodes[0];
3720 slot = path->slots[0];
3721 write_extent_buffer(leaf, buf,
3722 btrfs_item_ptr_offset(leaf, slot),
3725 extent = btrfs_item_ptr(leaf, slot,
3726 struct btrfs_file_extent_item);
3728 /* disko == 0 means it's a hole */
3732 btrfs_set_file_extent_offset(leaf, extent,
3734 btrfs_set_file_extent_num_bytes(leaf, extent,
3738 inode_add_bytes(inode, datal);
3739 ret = btrfs_inc_extent_ref(trans,
3742 root->root_key.objectid,
3743 btrfs_ino(BTRFS_I(inode)),
3744 new_key.offset - datao);
3746 btrfs_abort_transaction(trans,
3748 btrfs_end_transaction(trans);
3753 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3757 if (off > key.offset) {
3758 skip = off - key.offset;
3759 new_key.offset += skip;
3762 if (key.offset + datal > off + len)
3763 trim = key.offset + datal - (off + len);
3765 if (comp && (skip || trim)) {
3767 btrfs_end_transaction(trans);
3770 size -= skip + trim;
3771 datal -= skip + trim;
3773 ret = clone_copy_inline_extent(inode,
3780 if (ret != -EOPNOTSUPP)
3781 btrfs_abort_transaction(trans,
3783 btrfs_end_transaction(trans);
3786 leaf = path->nodes[0];
3787 slot = path->slots[0];
3790 /* If we have an implicit hole (NO_HOLES feature). */
3791 if (drop_start < new_key.offset)
3792 clone_update_extent_map(BTRFS_I(inode), trans,
3794 new_key.offset - drop_start);
3796 clone_update_extent_map(BTRFS_I(inode), trans,
3799 btrfs_mark_buffer_dirty(leaf);
3800 btrfs_release_path(path);
3802 last_dest_end = ALIGN(new_key.offset + datal,
3803 fs_info->sectorsize);
3804 ret = clone_finish_inode_update(trans, inode,
3810 if (new_key.offset + datal >= destoff + len)
3813 btrfs_release_path(path);
3814 key.offset = next_key_min_offset;
3816 if (fatal_signal_pending(current)) {
3823 if (last_dest_end < destoff + len) {
3825 * We have an implicit hole (NO_HOLES feature is enabled) that
3826 * fully or partially overlaps our cloning range at its end.
3828 btrfs_release_path(path);
3831 * 1 - remove extent(s)
3834 trans = btrfs_start_transaction(root, 2);
3835 if (IS_ERR(trans)) {
3836 ret = PTR_ERR(trans);
3839 ret = btrfs_drop_extents(trans, root, inode,
3840 last_dest_end, destoff + len, 1);
3842 if (ret != -EOPNOTSUPP)
3843 btrfs_abort_transaction(trans, ret);
3844 btrfs_end_transaction(trans);
3847 clone_update_extent_map(BTRFS_I(inode), trans, NULL,
3849 destoff + len - last_dest_end);
3850 ret = clone_finish_inode_update(trans, inode, destoff + len,
3851 destoff, olen, no_time_update);
3855 btrfs_free_path(path);
3860 static noinline int btrfs_clone_files(struct file *file, struct file *file_src,
3861 u64 off, u64 olen, u64 destoff)
3863 struct inode *inode = file_inode(file);
3864 struct inode *src = file_inode(file_src);
3865 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
3866 struct btrfs_root *root = BTRFS_I(inode)->root;
3869 u64 bs = fs_info->sb->s_blocksize;
3870 int same_inode = src == inode;
3874 * - split compressed inline extents. annoying: we need to
3875 * decompress into destination's address_space (the file offset
3876 * may change, so source mapping won't do), then recompress (or
3877 * otherwise reinsert) a subrange.
3879 * - split destination inode's inline extents. The inline extents can
3880 * be either compressed or non-compressed.
3883 if (btrfs_root_readonly(root))
3886 if (file_src->f_path.mnt != file->f_path.mnt ||
3887 src->i_sb != inode->i_sb)
3890 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3894 btrfs_double_inode_lock(src, inode);
3899 /* don't make the dst file partly checksummed */
3900 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3901 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) {
3906 /* determine range to clone */
3908 if (off + len > src->i_size || off + len < off)
3911 olen = len = src->i_size - off;
3913 * If we extend to eof, continue to block boundary if and only if the
3914 * destination end offset matches the destination file's size, otherwise
3915 * we would be corrupting data by placing the eof block into the middle
3918 if (off + len == src->i_size) {
3919 if (!IS_ALIGNED(len, bs) && destoff + len < inode->i_size)
3921 len = ALIGN(src->i_size, bs) - off;
3929 /* verify the end result is block aligned */
3930 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3931 !IS_ALIGNED(destoff, bs))
3934 /* verify if ranges are overlapped within the same file */
3936 if (destoff + len > off && destoff < off + len)
3940 if (destoff > inode->i_size) {
3941 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3947 * Lock the target range too. Right after we replace the file extent
3948 * items in the fs tree (which now point to the cloned data), we might
3949 * have a worker replace them with extent items relative to a write
3950 * operation that was issued before this clone operation (i.e. confront
3951 * with inode.c:btrfs_finish_ordered_io).
3954 u64 lock_start = min_t(u64, off, destoff);
3955 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
3957 ret = lock_extent_range(src, lock_start, lock_len, true);
3959 ret = btrfs_double_extent_lock(src, off, inode, destoff, len,
3964 /* ranges in the io trees already unlocked */
3968 ret = btrfs_clone(src, inode, off, olen, len, destoff, 0);
3971 u64 lock_start = min_t(u64, off, destoff);
3972 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3974 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3976 btrfs_double_extent_unlock(src, off, inode, destoff, len);
3979 * Truncate page cache pages so that future reads will see the cloned
3980 * data immediately and not the previous data.
3982 truncate_inode_pages_range(&inode->i_data,
3983 round_down(destoff, PAGE_SIZE),
3984 round_up(destoff + len, PAGE_SIZE) - 1);
3987 btrfs_double_inode_unlock(src, inode);
3993 int btrfs_clone_file_range(struct file *src_file, loff_t off,
3994 struct file *dst_file, loff_t destoff, u64 len)
3996 return btrfs_clone_files(dst_file, src_file, off, len, destoff);
4000 * there are many ways the trans_start and trans_end ioctls can lead
4001 * to deadlocks. They should only be used by applications that
4002 * basically own the machine, and have a very in depth understanding
4003 * of all the possible deadlocks and enospc problems.
4005 static long btrfs_ioctl_trans_start(struct file *file)
4007 struct inode *inode = file_inode(file);
4008 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4009 struct btrfs_root *root = BTRFS_I(inode)->root;
4010 struct btrfs_trans_handle *trans;
4011 struct btrfs_file_private *private;
4013 static bool warned = false;
4016 if (!capable(CAP_SYS_ADMIN))
4021 "Userspace transaction mechanism is considered "
4022 "deprecated and slated to be removed in 4.17. "
4023 "If you have a valid use case please "
4024 "speak up on the mailing list");
4030 private = file->private_data;
4031 if (private && private->trans)
4034 private = kzalloc(sizeof(struct btrfs_file_private),
4038 file->private_data = private;
4042 if (btrfs_root_readonly(root))
4045 ret = mnt_want_write_file(file);
4049 atomic_inc(&fs_info->open_ioctl_trans);
4052 trans = btrfs_start_ioctl_transaction(root);
4056 private->trans = trans;
4060 atomic_dec(&fs_info->open_ioctl_trans);
4061 mnt_drop_write_file(file);
4066 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
4068 struct inode *inode = file_inode(file);
4069 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4070 struct btrfs_root *root = BTRFS_I(inode)->root;
4071 struct btrfs_root *new_root;
4072 struct btrfs_dir_item *di;
4073 struct btrfs_trans_handle *trans;
4074 struct btrfs_path *path;
4075 struct btrfs_key location;
4076 struct btrfs_disk_key disk_key;
4081 if (!capable(CAP_SYS_ADMIN))
4084 ret = mnt_want_write_file(file);
4088 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
4094 objectid = BTRFS_FS_TREE_OBJECTID;
4096 location.objectid = objectid;
4097 location.type = BTRFS_ROOT_ITEM_KEY;
4098 location.offset = (u64)-1;
4100 new_root = btrfs_read_fs_root_no_name(fs_info, &location);
4101 if (IS_ERR(new_root)) {
4102 ret = PTR_ERR(new_root);
4105 if (!is_fstree(new_root->objectid)) {
4110 path = btrfs_alloc_path();
4115 path->leave_spinning = 1;
4117 trans = btrfs_start_transaction(root, 1);
4118 if (IS_ERR(trans)) {
4119 btrfs_free_path(path);
4120 ret = PTR_ERR(trans);
4124 dir_id = btrfs_super_root_dir(fs_info->super_copy);
4125 di = btrfs_lookup_dir_item(trans, fs_info->tree_root, path,
4126 dir_id, "default", 7, 1);
4127 if (IS_ERR_OR_NULL(di)) {
4128 btrfs_free_path(path);
4129 btrfs_end_transaction(trans);
4131 "Umm, you don't have the default diritem, this isn't going to work");
4136 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
4137 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
4138 btrfs_mark_buffer_dirty(path->nodes[0]);
4139 btrfs_free_path(path);
4141 btrfs_set_fs_incompat(fs_info, DEFAULT_SUBVOL);
4142 btrfs_end_transaction(trans);
4144 mnt_drop_write_file(file);
4148 void btrfs_get_block_group_info(struct list_head *groups_list,
4149 struct btrfs_ioctl_space_info *space)
4151 struct btrfs_block_group_cache *block_group;
4153 space->total_bytes = 0;
4154 space->used_bytes = 0;
4156 list_for_each_entry(block_group, groups_list, list) {
4157 space->flags = block_group->flags;
4158 space->total_bytes += block_group->key.offset;
4159 space->used_bytes +=
4160 btrfs_block_group_used(&block_group->item);
4164 static long btrfs_ioctl_space_info(struct btrfs_fs_info *fs_info,
4167 struct btrfs_ioctl_space_args space_args;
4168 struct btrfs_ioctl_space_info space;
4169 struct btrfs_ioctl_space_info *dest;
4170 struct btrfs_ioctl_space_info *dest_orig;
4171 struct btrfs_ioctl_space_info __user *user_dest;
4172 struct btrfs_space_info *info;
4173 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
4174 BTRFS_BLOCK_GROUP_SYSTEM,
4175 BTRFS_BLOCK_GROUP_METADATA,
4176 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
4183 if (copy_from_user(&space_args,
4184 (struct btrfs_ioctl_space_args __user *)arg,
4185 sizeof(space_args)))
4188 for (i = 0; i < num_types; i++) {
4189 struct btrfs_space_info *tmp;
4193 list_for_each_entry_rcu(tmp, &fs_info->space_info,
4195 if (tmp->flags == types[i]) {
4205 down_read(&info->groups_sem);
4206 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4207 if (!list_empty(&info->block_groups[c]))
4210 up_read(&info->groups_sem);
4214 * Global block reserve, exported as a space_info
4218 /* space_slots == 0 means they are asking for a count */
4219 if (space_args.space_slots == 0) {
4220 space_args.total_spaces = slot_count;
4224 slot_count = min_t(u64, space_args.space_slots, slot_count);
4226 alloc_size = sizeof(*dest) * slot_count;
4228 /* we generally have at most 6 or so space infos, one for each raid
4229 * level. So, a whole page should be more than enough for everyone
4231 if (alloc_size > PAGE_SIZE)
4234 space_args.total_spaces = 0;
4235 dest = kmalloc(alloc_size, GFP_KERNEL);
4240 /* now we have a buffer to copy into */
4241 for (i = 0; i < num_types; i++) {
4242 struct btrfs_space_info *tmp;
4249 list_for_each_entry_rcu(tmp, &fs_info->space_info,
4251 if (tmp->flags == types[i]) {
4260 down_read(&info->groups_sem);
4261 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4262 if (!list_empty(&info->block_groups[c])) {
4263 btrfs_get_block_group_info(
4264 &info->block_groups[c], &space);
4265 memcpy(dest, &space, sizeof(space));
4267 space_args.total_spaces++;
4273 up_read(&info->groups_sem);
4277 * Add global block reserve
4280 struct btrfs_block_rsv *block_rsv = &fs_info->global_block_rsv;
4282 spin_lock(&block_rsv->lock);
4283 space.total_bytes = block_rsv->size;
4284 space.used_bytes = block_rsv->size - block_rsv->reserved;
4285 spin_unlock(&block_rsv->lock);
4286 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
4287 memcpy(dest, &space, sizeof(space));
4288 space_args.total_spaces++;
4291 user_dest = (struct btrfs_ioctl_space_info __user *)
4292 (arg + sizeof(struct btrfs_ioctl_space_args));
4294 if (copy_to_user(user_dest, dest_orig, alloc_size))
4299 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
4306 * there are many ways the trans_start and trans_end ioctls can lead
4307 * to deadlocks. They should only be used by applications that
4308 * basically own the machine, and have a very in depth understanding
4309 * of all the possible deadlocks and enospc problems.
4311 long btrfs_ioctl_trans_end(struct file *file)
4313 struct inode *inode = file_inode(file);
4314 struct btrfs_root *root = BTRFS_I(inode)->root;
4315 struct btrfs_file_private *private = file->private_data;
4317 if (!private || !private->trans)
4320 btrfs_end_transaction(private->trans);
4321 private->trans = NULL;
4323 atomic_dec(&root->fs_info->open_ioctl_trans);
4325 mnt_drop_write_file(file);
4329 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4332 struct btrfs_trans_handle *trans;
4336 trans = btrfs_attach_transaction_barrier(root);
4337 if (IS_ERR(trans)) {
4338 if (PTR_ERR(trans) != -ENOENT)
4339 return PTR_ERR(trans);
4341 /* No running transaction, don't bother */
4342 transid = root->fs_info->last_trans_committed;
4345 transid = trans->transid;
4346 ret = btrfs_commit_transaction_async(trans, 0);
4348 btrfs_end_transaction(trans);
4353 if (copy_to_user(argp, &transid, sizeof(transid)))
4358 static noinline long btrfs_ioctl_wait_sync(struct btrfs_fs_info *fs_info,
4364 if (copy_from_user(&transid, argp, sizeof(transid)))
4367 transid = 0; /* current trans */
4369 return btrfs_wait_for_commit(fs_info, transid);
4372 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
4374 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
4375 struct btrfs_ioctl_scrub_args *sa;
4378 if (!capable(CAP_SYS_ADMIN))
4381 sa = memdup_user(arg, sizeof(*sa));
4385 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4386 ret = mnt_want_write_file(file);
4391 ret = btrfs_scrub_dev(fs_info, sa->devid, sa->start, sa->end,
4392 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4395 if (copy_to_user(arg, sa, sizeof(*sa)))
4398 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4399 mnt_drop_write_file(file);
4405 static long btrfs_ioctl_scrub_cancel(struct btrfs_fs_info *fs_info)
4407 if (!capable(CAP_SYS_ADMIN))
4410 return btrfs_scrub_cancel(fs_info);
4413 static long btrfs_ioctl_scrub_progress(struct btrfs_fs_info *fs_info,
4416 struct btrfs_ioctl_scrub_args *sa;
4419 if (!capable(CAP_SYS_ADMIN))
4422 sa = memdup_user(arg, sizeof(*sa));
4426 ret = btrfs_scrub_progress(fs_info, sa->devid, &sa->progress);
4428 if (copy_to_user(arg, sa, sizeof(*sa)))
4435 static long btrfs_ioctl_get_dev_stats(struct btrfs_fs_info *fs_info,
4438 struct btrfs_ioctl_get_dev_stats *sa;
4441 sa = memdup_user(arg, sizeof(*sa));
4445 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4450 ret = btrfs_get_dev_stats(fs_info, sa);
4452 if (copy_to_user(arg, sa, sizeof(*sa)))
4459 static long btrfs_ioctl_dev_replace(struct btrfs_fs_info *fs_info,
4462 struct btrfs_ioctl_dev_replace_args *p;
4465 if (!capable(CAP_SYS_ADMIN))
4468 p = memdup_user(arg, sizeof(*p));
4473 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
4474 if (sb_rdonly(fs_info->sb)) {
4478 if (test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
4479 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4481 ret = btrfs_dev_replace_by_ioctl(fs_info, p);
4482 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
4485 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4486 btrfs_dev_replace_status(fs_info, p);
4489 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4490 ret = btrfs_dev_replace_cancel(fs_info, p);
4497 if (copy_to_user(arg, p, sizeof(*p)))
4504 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4510 struct btrfs_ioctl_ino_path_args *ipa = NULL;
4511 struct inode_fs_paths *ipath = NULL;
4512 struct btrfs_path *path;
4514 if (!capable(CAP_DAC_READ_SEARCH))
4517 path = btrfs_alloc_path();
4523 ipa = memdup_user(arg, sizeof(*ipa));
4530 size = min_t(u32, ipa->size, 4096);
4531 ipath = init_ipath(size, root, path);
4532 if (IS_ERR(ipath)) {
4533 ret = PTR_ERR(ipath);
4538 ret = paths_from_inode(ipa->inum, ipath);
4542 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
4543 rel_ptr = ipath->fspath->val[i] -
4544 (u64)(unsigned long)ipath->fspath->val;
4545 ipath->fspath->val[i] = rel_ptr;
4548 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
4549 (void *)(unsigned long)ipath->fspath, size);
4556 btrfs_free_path(path);
4563 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4565 struct btrfs_data_container *inodes = ctx;
4566 const size_t c = 3 * sizeof(u64);
4568 if (inodes->bytes_left >= c) {
4569 inodes->bytes_left -= c;
4570 inodes->val[inodes->elem_cnt] = inum;
4571 inodes->val[inodes->elem_cnt + 1] = offset;
4572 inodes->val[inodes->elem_cnt + 2] = root;
4573 inodes->elem_cnt += 3;
4575 inodes->bytes_missing += c - inodes->bytes_left;
4576 inodes->bytes_left = 0;
4577 inodes->elem_missed += 3;
4583 static long btrfs_ioctl_logical_to_ino(struct btrfs_fs_info *fs_info,
4588 struct btrfs_ioctl_logical_ino_args *loi;
4589 struct btrfs_data_container *inodes = NULL;
4590 struct btrfs_path *path = NULL;
4592 if (!capable(CAP_SYS_ADMIN))
4595 loi = memdup_user(arg, sizeof(*loi));
4597 return PTR_ERR(loi);
4599 path = btrfs_alloc_path();
4605 size = min_t(u32, loi->size, SZ_64K);
4606 inodes = init_data_container(size);
4607 if (IS_ERR(inodes)) {
4608 ret = PTR_ERR(inodes);
4613 ret = iterate_inodes_from_logical(loi->logical, fs_info, path,
4614 build_ino_list, inodes);
4620 ret = copy_to_user((void *)(unsigned long)loi->inodes,
4621 (void *)(unsigned long)inodes, size);
4626 btrfs_free_path(path);
4633 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
4634 struct btrfs_ioctl_balance_args *bargs)
4636 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4638 bargs->flags = bctl->flags;
4640 if (atomic_read(&fs_info->balance_running))
4641 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4642 if (atomic_read(&fs_info->balance_pause_req))
4643 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
4644 if (atomic_read(&fs_info->balance_cancel_req))
4645 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
4647 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4648 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4649 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4652 spin_lock(&fs_info->balance_lock);
4653 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4654 spin_unlock(&fs_info->balance_lock);
4656 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4660 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4662 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4663 struct btrfs_fs_info *fs_info = root->fs_info;
4664 struct btrfs_ioctl_balance_args *bargs;
4665 struct btrfs_balance_control *bctl;
4666 bool need_unlock; /* for mut. excl. ops lock */
4669 if (!capable(CAP_SYS_ADMIN))
4672 ret = mnt_want_write_file(file);
4677 if (!test_and_set_bit(BTRFS_FS_EXCL_OP, &fs_info->flags)) {
4678 mutex_lock(&fs_info->volume_mutex);
4679 mutex_lock(&fs_info->balance_mutex);
4685 * mut. excl. ops lock is locked. Three possibilities:
4686 * (1) some other op is running
4687 * (2) balance is running
4688 * (3) balance is paused -- special case (think resume)
4690 mutex_lock(&fs_info->balance_mutex);
4691 if (fs_info->balance_ctl) {
4692 /* this is either (2) or (3) */
4693 if (!atomic_read(&fs_info->balance_running)) {
4694 mutex_unlock(&fs_info->balance_mutex);
4695 if (!mutex_trylock(&fs_info->volume_mutex))
4697 mutex_lock(&fs_info->balance_mutex);
4699 if (fs_info->balance_ctl &&
4700 !atomic_read(&fs_info->balance_running)) {
4702 need_unlock = false;
4706 mutex_unlock(&fs_info->balance_mutex);
4707 mutex_unlock(&fs_info->volume_mutex);
4711 mutex_unlock(&fs_info->balance_mutex);
4717 mutex_unlock(&fs_info->balance_mutex);
4718 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4723 BUG_ON(!test_bit(BTRFS_FS_EXCL_OP, &fs_info->flags));
4726 bargs = memdup_user(arg, sizeof(*bargs));
4727 if (IS_ERR(bargs)) {
4728 ret = PTR_ERR(bargs);
4732 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4733 if (!fs_info->balance_ctl) {
4738 bctl = fs_info->balance_ctl;
4739 spin_lock(&fs_info->balance_lock);
4740 bctl->flags |= BTRFS_BALANCE_RESUME;
4741 spin_unlock(&fs_info->balance_lock);
4749 if (fs_info->balance_ctl) {
4754 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
4760 bctl->fs_info = fs_info;
4762 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4763 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4764 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4766 bctl->flags = bargs->flags;
4768 /* balance everything - no filters */
4769 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4772 if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
4779 * Ownership of bctl and filesystem flag BTRFS_FS_EXCL_OP
4780 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4781 * or, if restriper was paused all the way until unmount, in
4782 * free_fs_info. The flag is cleared in __cancel_balance.
4784 need_unlock = false;
4786 ret = btrfs_balance(bctl, bargs);
4790 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4799 mutex_unlock(&fs_info->balance_mutex);
4800 mutex_unlock(&fs_info->volume_mutex);
4802 clear_bit(BTRFS_FS_EXCL_OP, &fs_info->flags);
4804 mnt_drop_write_file(file);
4808 static long btrfs_ioctl_balance_ctl(struct btrfs_fs_info *fs_info, int cmd)
4810 if (!capable(CAP_SYS_ADMIN))
4814 case BTRFS_BALANCE_CTL_PAUSE:
4815 return btrfs_pause_balance(fs_info);
4816 case BTRFS_BALANCE_CTL_CANCEL:
4817 return btrfs_cancel_balance(fs_info);
4823 static long btrfs_ioctl_balance_progress(struct btrfs_fs_info *fs_info,
4826 struct btrfs_ioctl_balance_args *bargs;
4829 if (!capable(CAP_SYS_ADMIN))
4832 mutex_lock(&fs_info->balance_mutex);
4833 if (!fs_info->balance_ctl) {
4838 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
4844 update_ioctl_balance_args(fs_info, 1, bargs);
4846 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4851 mutex_unlock(&fs_info->balance_mutex);
4855 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4857 struct inode *inode = file_inode(file);
4858 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4859 struct btrfs_ioctl_quota_ctl_args *sa;
4860 struct btrfs_trans_handle *trans = NULL;
4864 if (!capable(CAP_SYS_ADMIN))
4867 ret = mnt_want_write_file(file);
4871 sa = memdup_user(arg, sizeof(*sa));
4877 down_write(&fs_info->subvol_sem);
4878 trans = btrfs_start_transaction(fs_info->tree_root, 2);
4879 if (IS_ERR(trans)) {
4880 ret = PTR_ERR(trans);
4885 case BTRFS_QUOTA_CTL_ENABLE:
4886 ret = btrfs_quota_enable(trans, fs_info);
4888 case BTRFS_QUOTA_CTL_DISABLE:
4889 ret = btrfs_quota_disable(trans, fs_info);
4896 err = btrfs_commit_transaction(trans);
4901 up_write(&fs_info->subvol_sem);
4903 mnt_drop_write_file(file);
4907 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4909 struct inode *inode = file_inode(file);
4910 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4911 struct btrfs_root *root = BTRFS_I(inode)->root;
4912 struct btrfs_ioctl_qgroup_assign_args *sa;
4913 struct btrfs_trans_handle *trans;
4917 if (!capable(CAP_SYS_ADMIN))
4920 ret = mnt_want_write_file(file);
4924 sa = memdup_user(arg, sizeof(*sa));
4930 trans = btrfs_join_transaction(root);
4931 if (IS_ERR(trans)) {
4932 ret = PTR_ERR(trans);
4937 ret = btrfs_add_qgroup_relation(trans, fs_info,
4940 ret = btrfs_del_qgroup_relation(trans, fs_info,
4944 /* update qgroup status and info */
4945 err = btrfs_run_qgroups(trans, fs_info);
4947 btrfs_handle_fs_error(fs_info, err,
4948 "failed to update qgroup status and info");
4949 err = btrfs_end_transaction(trans);
4956 mnt_drop_write_file(file);
4960 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4962 struct inode *inode = file_inode(file);
4963 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
4964 struct btrfs_root *root = BTRFS_I(inode)->root;
4965 struct btrfs_ioctl_qgroup_create_args *sa;
4966 struct btrfs_trans_handle *trans;
4970 if (!capable(CAP_SYS_ADMIN))
4973 ret = mnt_want_write_file(file);
4977 sa = memdup_user(arg, sizeof(*sa));
4983 if (!sa->qgroupid) {
4988 trans = btrfs_join_transaction(root);
4989 if (IS_ERR(trans)) {
4990 ret = PTR_ERR(trans);
4995 ret = btrfs_create_qgroup(trans, fs_info, sa->qgroupid);
4997 ret = btrfs_remove_qgroup(trans, fs_info, sa->qgroupid);
5000 err = btrfs_end_transaction(trans);
5007 mnt_drop_write_file(file);
5011 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
5013 struct inode *inode = file_inode(file);
5014 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5015 struct btrfs_root *root = BTRFS_I(inode)->root;
5016 struct btrfs_ioctl_qgroup_limit_args *sa;
5017 struct btrfs_trans_handle *trans;
5022 if (!capable(CAP_SYS_ADMIN))
5025 ret = mnt_want_write_file(file);
5029 sa = memdup_user(arg, sizeof(*sa));
5035 trans = btrfs_join_transaction(root);
5036 if (IS_ERR(trans)) {
5037 ret = PTR_ERR(trans);
5041 qgroupid = sa->qgroupid;
5043 /* take the current subvol as qgroup */
5044 qgroupid = root->root_key.objectid;
5047 ret = btrfs_limit_qgroup(trans, fs_info, qgroupid, &sa->lim);
5049 err = btrfs_end_transaction(trans);
5056 mnt_drop_write_file(file);
5060 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
5062 struct inode *inode = file_inode(file);
5063 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5064 struct btrfs_ioctl_quota_rescan_args *qsa;
5067 if (!capable(CAP_SYS_ADMIN))
5070 ret = mnt_want_write_file(file);
5074 qsa = memdup_user(arg, sizeof(*qsa));
5085 ret = btrfs_qgroup_rescan(fs_info);
5090 mnt_drop_write_file(file);
5094 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
5096 struct inode *inode = file_inode(file);
5097 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5098 struct btrfs_ioctl_quota_rescan_args *qsa;
5101 if (!capable(CAP_SYS_ADMIN))
5104 qsa = kzalloc(sizeof(*qsa), GFP_KERNEL);
5108 if (fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
5110 qsa->progress = fs_info->qgroup_rescan_progress.objectid;
5113 if (copy_to_user(arg, qsa, sizeof(*qsa)))
5120 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
5122 struct inode *inode = file_inode(file);
5123 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5125 if (!capable(CAP_SYS_ADMIN))
5128 return btrfs_qgroup_wait_for_completion(fs_info, true);
5131 static long _btrfs_ioctl_set_received_subvol(struct file *file,
5132 struct btrfs_ioctl_received_subvol_args *sa)
5134 struct inode *inode = file_inode(file);
5135 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5136 struct btrfs_root *root = BTRFS_I(inode)->root;
5137 struct btrfs_root_item *root_item = &root->root_item;
5138 struct btrfs_trans_handle *trans;
5139 struct timespec ct = current_time(inode);
5141 int received_uuid_changed;
5143 if (!inode_owner_or_capable(inode))
5146 ret = mnt_want_write_file(file);
5150 down_write(&fs_info->subvol_sem);
5152 if (btrfs_ino(BTRFS_I(inode)) != BTRFS_FIRST_FREE_OBJECTID) {
5157 if (btrfs_root_readonly(root)) {
5164 * 2 - uuid items (received uuid + subvol uuid)
5166 trans = btrfs_start_transaction(root, 3);
5167 if (IS_ERR(trans)) {
5168 ret = PTR_ERR(trans);
5173 sa->rtransid = trans->transid;
5174 sa->rtime.sec = ct.tv_sec;
5175 sa->rtime.nsec = ct.tv_nsec;
5177 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
5179 if (received_uuid_changed &&
5180 !btrfs_is_empty_uuid(root_item->received_uuid))
5181 btrfs_uuid_tree_rem(trans, fs_info, root_item->received_uuid,
5182 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5183 root->root_key.objectid);
5184 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
5185 btrfs_set_root_stransid(root_item, sa->stransid);
5186 btrfs_set_root_rtransid(root_item, sa->rtransid);
5187 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
5188 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
5189 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
5190 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
5192 ret = btrfs_update_root(trans, fs_info->tree_root,
5193 &root->root_key, &root->root_item);
5195 btrfs_end_transaction(trans);
5198 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
5199 ret = btrfs_uuid_tree_add(trans, fs_info, sa->uuid,
5200 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5201 root->root_key.objectid);
5202 if (ret < 0 && ret != -EEXIST) {
5203 btrfs_abort_transaction(trans, ret);
5207 ret = btrfs_commit_transaction(trans);
5209 btrfs_abort_transaction(trans, ret);
5214 up_write(&fs_info->subvol_sem);
5215 mnt_drop_write_file(file);
5220 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
5223 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
5224 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
5227 args32 = memdup_user(arg, sizeof(*args32));
5229 return PTR_ERR(args32);
5231 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
5237 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
5238 args64->stransid = args32->stransid;
5239 args64->rtransid = args32->rtransid;
5240 args64->stime.sec = args32->stime.sec;
5241 args64->stime.nsec = args32->stime.nsec;
5242 args64->rtime.sec = args32->rtime.sec;
5243 args64->rtime.nsec = args32->rtime.nsec;
5244 args64->flags = args32->flags;
5246 ret = _btrfs_ioctl_set_received_subvol(file, args64);
5250 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
5251 args32->stransid = args64->stransid;
5252 args32->rtransid = args64->rtransid;
5253 args32->stime.sec = args64->stime.sec;
5254 args32->stime.nsec = args64->stime.nsec;
5255 args32->rtime.sec = args64->rtime.sec;
5256 args32->rtime.nsec = args64->rtime.nsec;
5257 args32->flags = args64->flags;
5259 ret = copy_to_user(arg, args32, sizeof(*args32));
5270 static long btrfs_ioctl_set_received_subvol(struct file *file,
5273 struct btrfs_ioctl_received_subvol_args *sa = NULL;
5276 sa = memdup_user(arg, sizeof(*sa));
5280 ret = _btrfs_ioctl_set_received_subvol(file, sa);
5285 ret = copy_to_user(arg, sa, sizeof(*sa));
5294 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
5296 struct inode *inode = file_inode(file);
5297 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5300 char label[BTRFS_LABEL_SIZE];
5302 spin_lock(&fs_info->super_lock);
5303 memcpy(label, fs_info->super_copy->label, BTRFS_LABEL_SIZE);
5304 spin_unlock(&fs_info->super_lock);
5306 len = strnlen(label, BTRFS_LABEL_SIZE);
5308 if (len == BTRFS_LABEL_SIZE) {
5310 "label is too long, return the first %zu bytes",
5314 ret = copy_to_user(arg, label, len);
5316 return ret ? -EFAULT : 0;
5319 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5321 struct inode *inode = file_inode(file);
5322 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5323 struct btrfs_root *root = BTRFS_I(inode)->root;
5324 struct btrfs_super_block *super_block = fs_info->super_copy;
5325 struct btrfs_trans_handle *trans;
5326 char label[BTRFS_LABEL_SIZE];
5329 if (!capable(CAP_SYS_ADMIN))
5332 if (copy_from_user(label, arg, sizeof(label)))
5335 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
5337 "unable to set label with more than %d bytes",
5338 BTRFS_LABEL_SIZE - 1);
5342 ret = mnt_want_write_file(file);
5346 trans = btrfs_start_transaction(root, 0);
5347 if (IS_ERR(trans)) {
5348 ret = PTR_ERR(trans);
5352 spin_lock(&fs_info->super_lock);
5353 strcpy(super_block->label, label);
5354 spin_unlock(&fs_info->super_lock);
5355 ret = btrfs_commit_transaction(trans);
5358 mnt_drop_write_file(file);
5362 #define INIT_FEATURE_FLAGS(suffix) \
5363 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5364 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5365 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5367 int btrfs_ioctl_get_supported_features(void __user *arg)
5369 static const struct btrfs_ioctl_feature_flags features[3] = {
5370 INIT_FEATURE_FLAGS(SUPP),
5371 INIT_FEATURE_FLAGS(SAFE_SET),
5372 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5375 if (copy_to_user(arg, &features, sizeof(features)))
5381 static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5383 struct inode *inode = file_inode(file);
5384 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5385 struct btrfs_super_block *super_block = fs_info->super_copy;
5386 struct btrfs_ioctl_feature_flags features;
5388 features.compat_flags = btrfs_super_compat_flags(super_block);
5389 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5390 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5392 if (copy_to_user(arg, &features, sizeof(features)))
5398 static int check_feature_bits(struct btrfs_fs_info *fs_info,
5399 enum btrfs_feature_set set,
5400 u64 change_mask, u64 flags, u64 supported_flags,
5401 u64 safe_set, u64 safe_clear)
5403 const char *type = btrfs_feature_set_names[set];
5405 u64 disallowed, unsupported;
5406 u64 set_mask = flags & change_mask;
5407 u64 clear_mask = ~flags & change_mask;
5409 unsupported = set_mask & ~supported_flags;
5411 names = btrfs_printable_features(set, unsupported);
5414 "this kernel does not support the %s feature bit%s",
5415 names, strchr(names, ',') ? "s" : "");
5419 "this kernel does not support %s bits 0x%llx",
5424 disallowed = set_mask & ~safe_set;
5426 names = btrfs_printable_features(set, disallowed);
5429 "can't set the %s feature bit%s while mounted",
5430 names, strchr(names, ',') ? "s" : "");
5434 "can't set %s bits 0x%llx while mounted",
5439 disallowed = clear_mask & ~safe_clear;
5441 names = btrfs_printable_features(set, disallowed);
5444 "can't clear the %s feature bit%s while mounted",
5445 names, strchr(names, ',') ? "s" : "");
5449 "can't clear %s bits 0x%llx while mounted",
5457 #define check_feature(fs_info, change_mask, flags, mask_base) \
5458 check_feature_bits(fs_info, FEAT_##mask_base, change_mask, flags, \
5459 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5460 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5461 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5463 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5465 struct inode *inode = file_inode(file);
5466 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5467 struct btrfs_root *root = BTRFS_I(inode)->root;
5468 struct btrfs_super_block *super_block = fs_info->super_copy;
5469 struct btrfs_ioctl_feature_flags flags[2];
5470 struct btrfs_trans_handle *trans;
5474 if (!capable(CAP_SYS_ADMIN))
5477 if (copy_from_user(flags, arg, sizeof(flags)))
5481 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5482 !flags[0].incompat_flags)
5485 ret = check_feature(fs_info, flags[0].compat_flags,
5486 flags[1].compat_flags, COMPAT);
5490 ret = check_feature(fs_info, flags[0].compat_ro_flags,
5491 flags[1].compat_ro_flags, COMPAT_RO);
5495 ret = check_feature(fs_info, flags[0].incompat_flags,
5496 flags[1].incompat_flags, INCOMPAT);
5500 ret = mnt_want_write_file(file);
5504 trans = btrfs_start_transaction(root, 0);
5505 if (IS_ERR(trans)) {
5506 ret = PTR_ERR(trans);
5507 goto out_drop_write;
5510 spin_lock(&fs_info->super_lock);
5511 newflags = btrfs_super_compat_flags(super_block);
5512 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5513 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5514 btrfs_set_super_compat_flags(super_block, newflags);
5516 newflags = btrfs_super_compat_ro_flags(super_block);
5517 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5518 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5519 btrfs_set_super_compat_ro_flags(super_block, newflags);
5521 newflags = btrfs_super_incompat_flags(super_block);
5522 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5523 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5524 btrfs_set_super_incompat_flags(super_block, newflags);
5525 spin_unlock(&fs_info->super_lock);
5527 ret = btrfs_commit_transaction(trans);
5529 mnt_drop_write_file(file);
5534 long btrfs_ioctl(struct file *file, unsigned int
5535 cmd, unsigned long arg)
5537 struct inode *inode = file_inode(file);
5538 struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
5539 struct btrfs_root *root = BTRFS_I(inode)->root;
5540 void __user *argp = (void __user *)arg;
5543 case FS_IOC_GETFLAGS:
5544 return btrfs_ioctl_getflags(file, argp);
5545 case FS_IOC_SETFLAGS:
5546 return btrfs_ioctl_setflags(file, argp);
5547 case FS_IOC_GETVERSION:
5548 return btrfs_ioctl_getversion(file, argp);
5550 return btrfs_ioctl_fitrim(file, argp);
5551 case BTRFS_IOC_SNAP_CREATE:
5552 return btrfs_ioctl_snap_create(file, argp, 0);
5553 case BTRFS_IOC_SNAP_CREATE_V2:
5554 return btrfs_ioctl_snap_create_v2(file, argp, 0);
5555 case BTRFS_IOC_SUBVOL_CREATE:
5556 return btrfs_ioctl_snap_create(file, argp, 1);
5557 case BTRFS_IOC_SUBVOL_CREATE_V2:
5558 return btrfs_ioctl_snap_create_v2(file, argp, 1);
5559 case BTRFS_IOC_SNAP_DESTROY:
5560 return btrfs_ioctl_snap_destroy(file, argp);
5561 case BTRFS_IOC_SUBVOL_GETFLAGS:
5562 return btrfs_ioctl_subvol_getflags(file, argp);
5563 case BTRFS_IOC_SUBVOL_SETFLAGS:
5564 return btrfs_ioctl_subvol_setflags(file, argp);
5565 case BTRFS_IOC_DEFAULT_SUBVOL:
5566 return btrfs_ioctl_default_subvol(file, argp);
5567 case BTRFS_IOC_DEFRAG:
5568 return btrfs_ioctl_defrag(file, NULL);
5569 case BTRFS_IOC_DEFRAG_RANGE:
5570 return btrfs_ioctl_defrag(file, argp);
5571 case BTRFS_IOC_RESIZE:
5572 return btrfs_ioctl_resize(file, argp);
5573 case BTRFS_IOC_ADD_DEV:
5574 return btrfs_ioctl_add_dev(fs_info, argp);
5575 case BTRFS_IOC_RM_DEV:
5576 return btrfs_ioctl_rm_dev(file, argp);
5577 case BTRFS_IOC_RM_DEV_V2:
5578 return btrfs_ioctl_rm_dev_v2(file, argp);
5579 case BTRFS_IOC_FS_INFO:
5580 return btrfs_ioctl_fs_info(fs_info, argp);
5581 case BTRFS_IOC_DEV_INFO:
5582 return btrfs_ioctl_dev_info(fs_info, argp);
5583 case BTRFS_IOC_BALANCE:
5584 return btrfs_ioctl_balance(file, NULL);
5585 case BTRFS_IOC_TRANS_START:
5586 return btrfs_ioctl_trans_start(file);
5587 case BTRFS_IOC_TRANS_END:
5588 return btrfs_ioctl_trans_end(file);
5589 case BTRFS_IOC_TREE_SEARCH:
5590 return btrfs_ioctl_tree_search(file, argp);
5591 case BTRFS_IOC_TREE_SEARCH_V2:
5592 return btrfs_ioctl_tree_search_v2(file, argp);
5593 case BTRFS_IOC_INO_LOOKUP:
5594 return btrfs_ioctl_ino_lookup(file, argp);
5595 case BTRFS_IOC_INO_PATHS:
5596 return btrfs_ioctl_ino_to_path(root, argp);
5597 case BTRFS_IOC_LOGICAL_INO:
5598 return btrfs_ioctl_logical_to_ino(fs_info, argp);
5599 case BTRFS_IOC_SPACE_INFO:
5600 return btrfs_ioctl_space_info(fs_info, argp);
5601 case BTRFS_IOC_SYNC: {
5604 ret = btrfs_start_delalloc_roots(fs_info, 0, -1);
5607 ret = btrfs_sync_fs(inode->i_sb, 1);
5609 * The transaction thread may want to do more work,
5610 * namely it pokes the cleaner kthread that will start
5611 * processing uncleaned subvols.
5613 wake_up_process(fs_info->transaction_kthread);
5616 case BTRFS_IOC_START_SYNC:
5617 return btrfs_ioctl_start_sync(root, argp);
5618 case BTRFS_IOC_WAIT_SYNC:
5619 return btrfs_ioctl_wait_sync(fs_info, argp);
5620 case BTRFS_IOC_SCRUB:
5621 return btrfs_ioctl_scrub(file, argp);
5622 case BTRFS_IOC_SCRUB_CANCEL:
5623 return btrfs_ioctl_scrub_cancel(fs_info);
5624 case BTRFS_IOC_SCRUB_PROGRESS:
5625 return btrfs_ioctl_scrub_progress(fs_info, argp);
5626 case BTRFS_IOC_BALANCE_V2:
5627 return btrfs_ioctl_balance(file, argp);
5628 case BTRFS_IOC_BALANCE_CTL:
5629 return btrfs_ioctl_balance_ctl(fs_info, arg);
5630 case BTRFS_IOC_BALANCE_PROGRESS:
5631 return btrfs_ioctl_balance_progress(fs_info, argp);
5632 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5633 return btrfs_ioctl_set_received_subvol(file, argp);
5635 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5636 return btrfs_ioctl_set_received_subvol_32(file, argp);
5638 case BTRFS_IOC_SEND:
5639 return btrfs_ioctl_send(file, argp);
5640 case BTRFS_IOC_GET_DEV_STATS:
5641 return btrfs_ioctl_get_dev_stats(fs_info, argp);
5642 case BTRFS_IOC_QUOTA_CTL:
5643 return btrfs_ioctl_quota_ctl(file, argp);
5644 case BTRFS_IOC_QGROUP_ASSIGN:
5645 return btrfs_ioctl_qgroup_assign(file, argp);
5646 case BTRFS_IOC_QGROUP_CREATE:
5647 return btrfs_ioctl_qgroup_create(file, argp);
5648 case BTRFS_IOC_QGROUP_LIMIT:
5649 return btrfs_ioctl_qgroup_limit(file, argp);
5650 case BTRFS_IOC_QUOTA_RESCAN:
5651 return btrfs_ioctl_quota_rescan(file, argp);
5652 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5653 return btrfs_ioctl_quota_rescan_status(file, argp);
5654 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5655 return btrfs_ioctl_quota_rescan_wait(file, argp);
5656 case BTRFS_IOC_DEV_REPLACE:
5657 return btrfs_ioctl_dev_replace(fs_info, argp);
5658 case BTRFS_IOC_GET_FSLABEL:
5659 return btrfs_ioctl_get_fslabel(file, argp);
5660 case BTRFS_IOC_SET_FSLABEL:
5661 return btrfs_ioctl_set_fslabel(file, argp);
5662 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5663 return btrfs_ioctl_get_supported_features(argp);
5664 case BTRFS_IOC_GET_FEATURES:
5665 return btrfs_ioctl_get_features(file, argp);
5666 case BTRFS_IOC_SET_FEATURES:
5667 return btrfs_ioctl_set_features(file, argp);
5673 #ifdef CONFIG_COMPAT
5674 long btrfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
5677 * These all access 32-bit values anyway so no further
5678 * handling is necessary.
5681 case FS_IOC32_GETFLAGS:
5682 cmd = FS_IOC_GETFLAGS;
5684 case FS_IOC32_SETFLAGS:
5685 cmd = FS_IOC_SETFLAGS;
5687 case FS_IOC32_GETVERSION:
5688 cmd = FS_IOC_GETVERSION;
5692 return btrfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg));