2 * Copyright (C) 2007 Oracle. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
19 #include <linux/kernel.h>
20 #include <linux/bio.h>
21 #include <linux/buffer_head.h>
22 #include <linux/file.h>
24 #include <linux/fsnotify.h>
25 #include <linux/pagemap.h>
26 #include <linux/highmem.h>
27 #include <linux/time.h>
28 #include <linux/init.h>
29 #include <linux/string.h>
30 #include <linux/backing-dev.h>
31 #include <linux/mount.h>
32 #include <linux/mpage.h>
33 #include <linux/namei.h>
34 #include <linux/swap.h>
35 #include <linux/writeback.h>
36 #include <linux/statfs.h>
37 #include <linux/compat.h>
38 #include <linux/bit_spinlock.h>
39 #include <linux/security.h>
40 #include <linux/xattr.h>
41 #include <linux/vmalloc.h>
42 #include <linux/slab.h>
43 #include <linux/blkdev.h>
44 #include <linux/uuid.h>
45 #include <linux/btrfs.h>
46 #include <linux/uaccess.h>
49 #include "transaction.h"
50 #include "btrfs_inode.h"
51 #include "print-tree.h"
54 #include "inode-map.h"
56 #include "rcu-string.h"
58 #include "dev-replace.h"
64 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
65 * structures are incorrect, as the timespec structure from userspace
66 * is 4 bytes too small. We define these alternatives here to teach
67 * the kernel about the 32-bit struct packing.
69 struct btrfs_ioctl_timespec_32 {
72 } __attribute__ ((__packed__));
74 struct btrfs_ioctl_received_subvol_args_32 {
75 char uuid[BTRFS_UUID_SIZE]; /* in */
76 __u64 stransid; /* in */
77 __u64 rtransid; /* out */
78 struct btrfs_ioctl_timespec_32 stime; /* in */
79 struct btrfs_ioctl_timespec_32 rtime; /* out */
81 __u64 reserved[16]; /* in */
82 } __attribute__ ((__packed__));
84 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
85 struct btrfs_ioctl_received_subvol_args_32)
89 static int btrfs_clone(struct inode *src, struct inode *inode,
90 u64 off, u64 olen, u64 olen_aligned, u64 destoff,
93 /* Mask out flags that are inappropriate for the given type of inode. */
94 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
98 else if (S_ISREG(mode))
99 return flags & ~FS_DIRSYNC_FL;
101 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
105 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
107 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
109 unsigned int iflags = 0;
111 if (flags & BTRFS_INODE_SYNC)
112 iflags |= FS_SYNC_FL;
113 if (flags & BTRFS_INODE_IMMUTABLE)
114 iflags |= FS_IMMUTABLE_FL;
115 if (flags & BTRFS_INODE_APPEND)
116 iflags |= FS_APPEND_FL;
117 if (flags & BTRFS_INODE_NODUMP)
118 iflags |= FS_NODUMP_FL;
119 if (flags & BTRFS_INODE_NOATIME)
120 iflags |= FS_NOATIME_FL;
121 if (flags & BTRFS_INODE_DIRSYNC)
122 iflags |= FS_DIRSYNC_FL;
123 if (flags & BTRFS_INODE_NODATACOW)
124 iflags |= FS_NOCOW_FL;
126 if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
127 iflags |= FS_COMPR_FL;
128 else if (flags & BTRFS_INODE_NOCOMPRESS)
129 iflags |= FS_NOCOMP_FL;
135 * Update inode->i_flags based on the btrfs internal flags.
137 void btrfs_update_iflags(struct inode *inode)
139 struct btrfs_inode *ip = BTRFS_I(inode);
140 unsigned int new_fl = 0;
142 if (ip->flags & BTRFS_INODE_SYNC)
144 if (ip->flags & BTRFS_INODE_IMMUTABLE)
145 new_fl |= S_IMMUTABLE;
146 if (ip->flags & BTRFS_INODE_APPEND)
148 if (ip->flags & BTRFS_INODE_NOATIME)
150 if (ip->flags & BTRFS_INODE_DIRSYNC)
153 set_mask_bits(&inode->i_flags,
154 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC,
159 * Inherit flags from the parent inode.
161 * Currently only the compression flags and the cow flags are inherited.
163 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
170 flags = BTRFS_I(dir)->flags;
172 if (flags & BTRFS_INODE_NOCOMPRESS) {
173 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
174 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
175 } else if (flags & BTRFS_INODE_COMPRESS) {
176 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
177 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
180 if (flags & BTRFS_INODE_NODATACOW) {
181 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
182 if (S_ISREG(inode->i_mode))
183 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
186 btrfs_update_iflags(inode);
189 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
191 struct btrfs_inode *ip = BTRFS_I(file_inode(file));
192 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
194 if (copy_to_user(arg, &flags, sizeof(flags)))
199 static int check_flags(unsigned int flags)
201 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
202 FS_NOATIME_FL | FS_NODUMP_FL | \
203 FS_SYNC_FL | FS_DIRSYNC_FL | \
204 FS_NOCOMP_FL | FS_COMPR_FL |
208 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
214 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
216 struct inode *inode = file_inode(file);
217 struct btrfs_inode *ip = BTRFS_I(inode);
218 struct btrfs_root *root = ip->root;
219 struct btrfs_trans_handle *trans;
220 unsigned int flags, oldflags;
223 unsigned int i_oldflags;
226 if (!inode_owner_or_capable(inode))
229 if (btrfs_root_readonly(root))
232 if (copy_from_user(&flags, arg, sizeof(flags)))
235 ret = check_flags(flags);
239 ret = mnt_want_write_file(file);
243 mutex_lock(&inode->i_mutex);
245 ip_oldflags = ip->flags;
246 i_oldflags = inode->i_flags;
247 mode = inode->i_mode;
249 flags = btrfs_mask_flags(inode->i_mode, flags);
250 oldflags = btrfs_flags_to_ioctl(ip->flags);
251 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
252 if (!capable(CAP_LINUX_IMMUTABLE)) {
258 if (flags & FS_SYNC_FL)
259 ip->flags |= BTRFS_INODE_SYNC;
261 ip->flags &= ~BTRFS_INODE_SYNC;
262 if (flags & FS_IMMUTABLE_FL)
263 ip->flags |= BTRFS_INODE_IMMUTABLE;
265 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
266 if (flags & FS_APPEND_FL)
267 ip->flags |= BTRFS_INODE_APPEND;
269 ip->flags &= ~BTRFS_INODE_APPEND;
270 if (flags & FS_NODUMP_FL)
271 ip->flags |= BTRFS_INODE_NODUMP;
273 ip->flags &= ~BTRFS_INODE_NODUMP;
274 if (flags & FS_NOATIME_FL)
275 ip->flags |= BTRFS_INODE_NOATIME;
277 ip->flags &= ~BTRFS_INODE_NOATIME;
278 if (flags & FS_DIRSYNC_FL)
279 ip->flags |= BTRFS_INODE_DIRSYNC;
281 ip->flags &= ~BTRFS_INODE_DIRSYNC;
282 if (flags & FS_NOCOW_FL) {
285 * It's safe to turn csums off here, no extents exist.
286 * Otherwise we want the flag to reflect the real COW
287 * status of the file and will not set it.
289 if (inode->i_size == 0)
290 ip->flags |= BTRFS_INODE_NODATACOW
291 | BTRFS_INODE_NODATASUM;
293 ip->flags |= BTRFS_INODE_NODATACOW;
297 * Revert back under same assuptions as above
300 if (inode->i_size == 0)
301 ip->flags &= ~(BTRFS_INODE_NODATACOW
302 | BTRFS_INODE_NODATASUM);
304 ip->flags &= ~BTRFS_INODE_NODATACOW;
309 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
310 * flag may be changed automatically if compression code won't make
313 if (flags & FS_NOCOMP_FL) {
314 ip->flags &= ~BTRFS_INODE_COMPRESS;
315 ip->flags |= BTRFS_INODE_NOCOMPRESS;
317 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
318 if (ret && ret != -ENODATA)
320 } else if (flags & FS_COMPR_FL) {
323 ip->flags |= BTRFS_INODE_COMPRESS;
324 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
326 if (root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
330 ret = btrfs_set_prop(inode, "btrfs.compression",
331 comp, strlen(comp), 0);
336 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
337 if (ret && ret != -ENODATA)
339 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
342 trans = btrfs_start_transaction(root, 1);
344 ret = PTR_ERR(trans);
348 btrfs_update_iflags(inode);
349 inode_inc_iversion(inode);
350 inode->i_ctime = CURRENT_TIME;
351 ret = btrfs_update_inode(trans, root, inode);
353 btrfs_end_transaction(trans, root);
356 ip->flags = ip_oldflags;
357 inode->i_flags = i_oldflags;
361 mutex_unlock(&inode->i_mutex);
362 mnt_drop_write_file(file);
366 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
368 struct inode *inode = file_inode(file);
370 return put_user(inode->i_generation, arg);
373 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
375 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
376 struct btrfs_device *device;
377 struct request_queue *q;
378 struct fstrim_range range;
379 u64 minlen = ULLONG_MAX;
381 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
384 if (!capable(CAP_SYS_ADMIN))
388 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
392 q = bdev_get_queue(device->bdev);
393 if (blk_queue_discard(q)) {
395 minlen = min((u64)q->limits.discard_granularity,
403 if (copy_from_user(&range, arg, sizeof(range)))
405 if (range.start > total_bytes ||
406 range.len < fs_info->sb->s_blocksize)
409 range.len = min(range.len, total_bytes - range.start);
410 range.minlen = max(range.minlen, minlen);
411 ret = btrfs_trim_fs(fs_info->tree_root, &range);
415 if (copy_to_user(arg, &range, sizeof(range)))
421 int btrfs_is_empty_uuid(u8 *uuid)
425 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
432 static noinline int create_subvol(struct inode *dir,
433 struct dentry *dentry,
434 char *name, int namelen,
436 struct btrfs_qgroup_inherit *inherit)
438 struct btrfs_trans_handle *trans;
439 struct btrfs_key key;
440 struct btrfs_root_item root_item;
441 struct btrfs_inode_item *inode_item;
442 struct extent_buffer *leaf;
443 struct btrfs_root *root = BTRFS_I(dir)->root;
444 struct btrfs_root *new_root;
445 struct btrfs_block_rsv block_rsv;
446 struct timespec cur_time = CURRENT_TIME;
451 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
456 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
461 * Don't create subvolume whose level is not zero. Or qgroup will be
462 * screwed up since it assume subvolme qgroup's level to be 0.
464 if (btrfs_qgroup_level(objectid))
467 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
469 * The same as the snapshot creation, please see the comment
470 * of create_snapshot().
472 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
473 8, &qgroup_reserved, false);
477 trans = btrfs_start_transaction(root, 0);
479 ret = PTR_ERR(trans);
480 btrfs_subvolume_release_metadata(root, &block_rsv,
484 trans->block_rsv = &block_rsv;
485 trans->bytes_reserved = block_rsv.size;
487 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
491 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
497 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
498 btrfs_set_header_bytenr(leaf, leaf->start);
499 btrfs_set_header_generation(leaf, trans->transid);
500 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
501 btrfs_set_header_owner(leaf, objectid);
503 write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(),
505 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
506 btrfs_header_chunk_tree_uuid(leaf),
508 btrfs_mark_buffer_dirty(leaf);
510 memset(&root_item, 0, sizeof(root_item));
512 inode_item = &root_item.inode;
513 btrfs_set_stack_inode_generation(inode_item, 1);
514 btrfs_set_stack_inode_size(inode_item, 3);
515 btrfs_set_stack_inode_nlink(inode_item, 1);
516 btrfs_set_stack_inode_nbytes(inode_item, root->nodesize);
517 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
519 btrfs_set_root_flags(&root_item, 0);
520 btrfs_set_root_limit(&root_item, 0);
521 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
523 btrfs_set_root_bytenr(&root_item, leaf->start);
524 btrfs_set_root_generation(&root_item, trans->transid);
525 btrfs_set_root_level(&root_item, 0);
526 btrfs_set_root_refs(&root_item, 1);
527 btrfs_set_root_used(&root_item, leaf->len);
528 btrfs_set_root_last_snapshot(&root_item, 0);
530 btrfs_set_root_generation_v2(&root_item,
531 btrfs_root_generation(&root_item));
532 uuid_le_gen(&new_uuid);
533 memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
534 btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec);
535 btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec);
536 root_item.ctime = root_item.otime;
537 btrfs_set_root_ctransid(&root_item, trans->transid);
538 btrfs_set_root_otransid(&root_item, trans->transid);
540 btrfs_tree_unlock(leaf);
541 free_extent_buffer(leaf);
544 btrfs_set_root_dirid(&root_item, new_dirid);
546 key.objectid = objectid;
548 key.type = BTRFS_ROOT_ITEM_KEY;
549 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
554 key.offset = (u64)-1;
555 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
556 if (IS_ERR(new_root)) {
557 ret = PTR_ERR(new_root);
558 btrfs_abort_transaction(trans, root, ret);
562 btrfs_record_root_in_trans(trans, new_root);
564 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
566 /* We potentially lose an unused inode item here */
567 btrfs_abort_transaction(trans, root, ret);
572 * insert the directory item
574 ret = btrfs_set_inode_index(dir, &index);
576 btrfs_abort_transaction(trans, root, ret);
580 ret = btrfs_insert_dir_item(trans, root,
581 name, namelen, dir, &key,
582 BTRFS_FT_DIR, index);
584 btrfs_abort_transaction(trans, root, ret);
588 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
589 ret = btrfs_update_inode(trans, root, dir);
592 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
593 objectid, root->root_key.objectid,
594 btrfs_ino(dir), index, name, namelen);
597 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
598 root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
601 btrfs_abort_transaction(trans, root, ret);
604 trans->block_rsv = NULL;
605 trans->bytes_reserved = 0;
606 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
609 *async_transid = trans->transid;
610 err = btrfs_commit_transaction_async(trans, root, 1);
612 err = btrfs_commit_transaction(trans, root);
614 err = btrfs_commit_transaction(trans, root);
620 inode = btrfs_lookup_dentry(dir, dentry);
622 return PTR_ERR(inode);
623 d_instantiate(dentry, inode);
628 static void btrfs_wait_for_no_snapshoting_writes(struct btrfs_root *root)
634 prepare_to_wait(&root->subv_writers->wait, &wait,
635 TASK_UNINTERRUPTIBLE);
637 writers = percpu_counter_sum(&root->subv_writers->counter);
641 finish_wait(&root->subv_writers->wait, &wait);
645 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
646 struct dentry *dentry, char *name, int namelen,
647 u64 *async_transid, bool readonly,
648 struct btrfs_qgroup_inherit *inherit)
651 struct btrfs_pending_snapshot *pending_snapshot;
652 struct btrfs_trans_handle *trans;
655 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
658 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
659 if (!pending_snapshot)
662 pending_snapshot->root_item = kzalloc(sizeof(struct btrfs_root_item),
664 pending_snapshot->path = btrfs_alloc_path();
665 if (!pending_snapshot->root_item || !pending_snapshot->path) {
670 atomic_inc(&root->will_be_snapshoted);
671 smp_mb__after_atomic();
672 btrfs_wait_for_no_snapshoting_writes(root);
674 ret = btrfs_start_delalloc_inodes(root, 0);
678 btrfs_wait_ordered_extents(root, -1);
680 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
681 BTRFS_BLOCK_RSV_TEMP);
683 * 1 - parent dir inode
686 * 2 - root ref/backref
687 * 1 - root of snapshot
690 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
691 &pending_snapshot->block_rsv, 8,
692 &pending_snapshot->qgroup_reserved,
697 pending_snapshot->dentry = dentry;
698 pending_snapshot->root = root;
699 pending_snapshot->readonly = readonly;
700 pending_snapshot->dir = dir;
701 pending_snapshot->inherit = inherit;
703 trans = btrfs_start_transaction(root, 0);
705 ret = PTR_ERR(trans);
709 spin_lock(&root->fs_info->trans_lock);
710 list_add(&pending_snapshot->list,
711 &trans->transaction->pending_snapshots);
712 spin_unlock(&root->fs_info->trans_lock);
714 *async_transid = trans->transid;
715 ret = btrfs_commit_transaction_async(trans,
716 root->fs_info->extent_root, 1);
718 ret = btrfs_commit_transaction(trans, root);
720 ret = btrfs_commit_transaction(trans,
721 root->fs_info->extent_root);
726 ret = pending_snapshot->error;
730 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
734 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
736 ret = PTR_ERR(inode);
740 d_instantiate(dentry, inode);
743 btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
744 &pending_snapshot->block_rsv,
745 pending_snapshot->qgroup_reserved);
747 if (atomic_dec_and_test(&root->will_be_snapshoted))
748 wake_up_atomic_t(&root->will_be_snapshoted);
750 kfree(pending_snapshot->root_item);
751 btrfs_free_path(pending_snapshot->path);
752 kfree(pending_snapshot);
757 /* copy of may_delete in fs/namei.c()
758 * Check whether we can remove a link victim from directory dir, check
759 * whether the type of victim is right.
760 * 1. We can't do it if dir is read-only (done in permission())
761 * 2. We should have write and exec permissions on dir
762 * 3. We can't remove anything from append-only dir
763 * 4. We can't do anything with immutable dir (done in permission())
764 * 5. If the sticky bit on dir is set we should either
765 * a. be owner of dir, or
766 * b. be owner of victim, or
767 * c. have CAP_FOWNER capability
768 * 6. If the victim is append-only or immutable we can't do antyhing with
769 * links pointing to it.
770 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
771 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
772 * 9. We can't remove a root or mountpoint.
773 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
774 * nfs_async_unlink().
777 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
781 if (d_really_is_negative(victim))
784 BUG_ON(d_inode(victim->d_parent) != dir);
785 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
787 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
792 if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) ||
793 IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim)))
796 if (!d_is_dir(victim))
800 } else if (d_is_dir(victim))
804 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
809 /* copy of may_create in fs/namei.c() */
810 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
812 if (d_really_is_positive(child))
816 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
820 * Create a new subvolume below @parent. This is largely modeled after
821 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
822 * inside this filesystem so it's quite a bit simpler.
824 static noinline int btrfs_mksubvol(struct path *parent,
825 char *name, int namelen,
826 struct btrfs_root *snap_src,
827 u64 *async_transid, bool readonly,
828 struct btrfs_qgroup_inherit *inherit)
830 struct inode *dir = d_inode(parent->dentry);
831 struct dentry *dentry;
834 error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
838 dentry = lookup_one_len(name, parent->dentry, namelen);
839 error = PTR_ERR(dentry);
844 if (d_really_is_positive(dentry))
847 error = btrfs_may_create(dir, dentry);
852 * even if this name doesn't exist, we may get hash collisions.
853 * check for them now when we can safely fail
855 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
861 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
863 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
867 error = create_snapshot(snap_src, dir, dentry, name, namelen,
868 async_transid, readonly, inherit);
870 error = create_subvol(dir, dentry, name, namelen,
871 async_transid, inherit);
874 fsnotify_mkdir(dir, dentry);
876 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
880 mutex_unlock(&dir->i_mutex);
885 * When we're defragging a range, we don't want to kick it off again
886 * if it is really just waiting for delalloc to send it down.
887 * If we find a nice big extent or delalloc range for the bytes in the
888 * file you want to defrag, we return 0 to let you know to skip this
891 static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
893 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
894 struct extent_map *em = NULL;
895 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
898 read_lock(&em_tree->lock);
899 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
900 read_unlock(&em_tree->lock);
903 end = extent_map_end(em);
905 if (end - offset > thresh)
908 /* if we already have a nice delalloc here, just stop */
910 end = count_range_bits(io_tree, &offset, offset + thresh,
911 thresh, EXTENT_DELALLOC, 1);
918 * helper function to walk through a file and find extents
919 * newer than a specific transid, and smaller than thresh.
921 * This is used by the defragging code to find new and small
924 static int find_new_extents(struct btrfs_root *root,
925 struct inode *inode, u64 newer_than,
926 u64 *off, u32 thresh)
928 struct btrfs_path *path;
929 struct btrfs_key min_key;
930 struct extent_buffer *leaf;
931 struct btrfs_file_extent_item *extent;
934 u64 ino = btrfs_ino(inode);
936 path = btrfs_alloc_path();
940 min_key.objectid = ino;
941 min_key.type = BTRFS_EXTENT_DATA_KEY;
942 min_key.offset = *off;
945 ret = btrfs_search_forward(root, &min_key, path, newer_than);
949 if (min_key.objectid != ino)
951 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
954 leaf = path->nodes[0];
955 extent = btrfs_item_ptr(leaf, path->slots[0],
956 struct btrfs_file_extent_item);
958 type = btrfs_file_extent_type(leaf, extent);
959 if (type == BTRFS_FILE_EXTENT_REG &&
960 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
961 check_defrag_in_cache(inode, min_key.offset, thresh)) {
962 *off = min_key.offset;
963 btrfs_free_path(path);
968 if (path->slots[0] < btrfs_header_nritems(leaf)) {
969 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
973 if (min_key.offset == (u64)-1)
977 btrfs_release_path(path);
980 btrfs_free_path(path);
984 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
986 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
987 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
988 struct extent_map *em;
989 u64 len = PAGE_CACHE_SIZE;
992 * hopefully we have this extent in the tree already, try without
993 * the full extent lock
995 read_lock(&em_tree->lock);
996 em = lookup_extent_mapping(em_tree, start, len);
997 read_unlock(&em_tree->lock);
1000 struct extent_state *cached = NULL;
1001 u64 end = start + len - 1;
1003 /* get the big lock and read metadata off disk */
1004 lock_extent_bits(io_tree, start, end, &cached);
1005 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
1006 unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS);
1015 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1017 struct extent_map *next;
1020 /* this is the last extent */
1021 if (em->start + em->len >= i_size_read(inode))
1024 next = defrag_lookup_extent(inode, em->start + em->len);
1025 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1027 else if ((em->block_start + em->block_len == next->block_start) &&
1028 (em->block_len > SZ_128K && next->block_len > SZ_128K))
1031 free_extent_map(next);
1035 static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
1036 u64 *last_len, u64 *skip, u64 *defrag_end,
1039 struct extent_map *em;
1041 bool next_mergeable = true;
1042 bool prev_mergeable = true;
1045 * make sure that once we start defragging an extent, we keep on
1048 if (start < *defrag_end)
1053 em = defrag_lookup_extent(inode, start);
1057 /* this will cover holes, and inline extents */
1058 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1064 prev_mergeable = false;
1066 next_mergeable = defrag_check_next_extent(inode, em);
1068 * we hit a real extent, if it is big or the next extent is not a
1069 * real extent, don't bother defragging it
1071 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1072 (em->len >= thresh || (!next_mergeable && !prev_mergeable)))
1076 * last_len ends up being a counter of how many bytes we've defragged.
1077 * every time we choose not to defrag an extent, we reset *last_len
1078 * so that the next tiny extent will force a defrag.
1080 * The end result of this is that tiny extents before a single big
1081 * extent will force at least part of that big extent to be defragged.
1084 *defrag_end = extent_map_end(em);
1087 *skip = extent_map_end(em);
1091 free_extent_map(em);
1096 * it doesn't do much good to defrag one or two pages
1097 * at a time. This pulls in a nice chunk of pages
1098 * to COW and defrag.
1100 * It also makes sure the delalloc code has enough
1101 * dirty data to avoid making new small extents as part
1104 * It's a good idea to start RA on this range
1105 * before calling this.
1107 static int cluster_pages_for_defrag(struct inode *inode,
1108 struct page **pages,
1109 unsigned long start_index,
1110 unsigned long num_pages)
1112 unsigned long file_end;
1113 u64 isize = i_size_read(inode);
1120 struct btrfs_ordered_extent *ordered;
1121 struct extent_state *cached_state = NULL;
1122 struct extent_io_tree *tree;
1123 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1125 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1126 if (!isize || start_index > file_end)
1129 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1131 ret = btrfs_delalloc_reserve_space(inode,
1132 start_index << PAGE_CACHE_SHIFT,
1133 page_cnt << PAGE_CACHE_SHIFT);
1137 tree = &BTRFS_I(inode)->io_tree;
1139 /* step one, lock all the pages */
1140 for (i = 0; i < page_cnt; i++) {
1143 page = find_or_create_page(inode->i_mapping,
1144 start_index + i, mask);
1148 page_start = page_offset(page);
1149 page_end = page_start + PAGE_CACHE_SIZE - 1;
1151 lock_extent_bits(tree, page_start, page_end,
1153 ordered = btrfs_lookup_ordered_extent(inode,
1155 unlock_extent_cached(tree, page_start, page_end,
1156 &cached_state, GFP_NOFS);
1161 btrfs_start_ordered_extent(inode, ordered, 1);
1162 btrfs_put_ordered_extent(ordered);
1165 * we unlocked the page above, so we need check if
1166 * it was released or not.
1168 if (page->mapping != inode->i_mapping) {
1170 page_cache_release(page);
1175 if (!PageUptodate(page)) {
1176 btrfs_readpage(NULL, page);
1178 if (!PageUptodate(page)) {
1180 page_cache_release(page);
1186 if (page->mapping != inode->i_mapping) {
1188 page_cache_release(page);
1198 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1202 * so now we have a nice long stream of locked
1203 * and up to date pages, lets wait on them
1205 for (i = 0; i < i_done; i++)
1206 wait_on_page_writeback(pages[i]);
1208 page_start = page_offset(pages[0]);
1209 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1211 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1212 page_start, page_end - 1, &cached_state);
1213 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1214 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1215 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1216 &cached_state, GFP_NOFS);
1218 if (i_done != page_cnt) {
1219 spin_lock(&BTRFS_I(inode)->lock);
1220 BTRFS_I(inode)->outstanding_extents++;
1221 spin_unlock(&BTRFS_I(inode)->lock);
1222 btrfs_delalloc_release_space(inode,
1223 start_index << PAGE_CACHE_SHIFT,
1224 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1228 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1229 &cached_state, GFP_NOFS);
1231 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1232 page_start, page_end - 1, &cached_state,
1235 for (i = 0; i < i_done; i++) {
1236 clear_page_dirty_for_io(pages[i]);
1237 ClearPageChecked(pages[i]);
1238 set_page_extent_mapped(pages[i]);
1239 set_page_dirty(pages[i]);
1240 unlock_page(pages[i]);
1241 page_cache_release(pages[i]);
1245 for (i = 0; i < i_done; i++) {
1246 unlock_page(pages[i]);
1247 page_cache_release(pages[i]);
1249 btrfs_delalloc_release_space(inode,
1250 start_index << PAGE_CACHE_SHIFT,
1251 page_cnt << PAGE_CACHE_SHIFT);
1256 int btrfs_defrag_file(struct inode *inode, struct file *file,
1257 struct btrfs_ioctl_defrag_range_args *range,
1258 u64 newer_than, unsigned long max_to_defrag)
1260 struct btrfs_root *root = BTRFS_I(inode)->root;
1261 struct file_ra_state *ra = NULL;
1262 unsigned long last_index;
1263 u64 isize = i_size_read(inode);
1267 u64 newer_off = range->start;
1269 unsigned long ra_index = 0;
1271 int defrag_count = 0;
1272 int compress_type = BTRFS_COMPRESS_ZLIB;
1273 u32 extent_thresh = range->extent_thresh;
1274 unsigned long max_cluster = SZ_256K >> PAGE_CACHE_SHIFT;
1275 unsigned long cluster = max_cluster;
1276 u64 new_align = ~((u64)SZ_128K - 1);
1277 struct page **pages = NULL;
1282 if (range->start >= isize)
1285 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1286 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1288 if (range->compress_type)
1289 compress_type = range->compress_type;
1292 if (extent_thresh == 0)
1293 extent_thresh = SZ_256K;
1296 * if we were not given a file, allocate a readahead
1300 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1303 file_ra_state_init(ra, inode->i_mapping);
1308 pages = kmalloc_array(max_cluster, sizeof(struct page *),
1315 /* find the last page to defrag */
1316 if (range->start + range->len > range->start) {
1317 last_index = min_t(u64, isize - 1,
1318 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1320 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1324 ret = find_new_extents(root, inode, newer_than,
1325 &newer_off, SZ_64K);
1327 range->start = newer_off;
1329 * we always align our defrag to help keep
1330 * the extents in the file evenly spaced
1332 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1336 i = range->start >> PAGE_CACHE_SHIFT;
1339 max_to_defrag = last_index - i + 1;
1342 * make writeback starts from i, so the defrag range can be
1343 * written sequentially.
1345 if (i < inode->i_mapping->writeback_index)
1346 inode->i_mapping->writeback_index = i;
1348 while (i <= last_index && defrag_count < max_to_defrag &&
1349 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_CACHE_SIZE))) {
1351 * make sure we stop running if someone unmounts
1354 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1357 if (btrfs_defrag_cancelled(root->fs_info)) {
1358 btrfs_debug(root->fs_info, "defrag_file cancelled");
1363 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1364 extent_thresh, &last_len, &skip,
1365 &defrag_end, range->flags &
1366 BTRFS_DEFRAG_RANGE_COMPRESS)) {
1369 * the should_defrag function tells us how much to skip
1370 * bump our counter by the suggested amount
1372 next = DIV_ROUND_UP(skip, PAGE_CACHE_SIZE);
1373 i = max(i + 1, next);
1378 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1379 PAGE_CACHE_SHIFT) - i;
1380 cluster = min(cluster, max_cluster);
1382 cluster = max_cluster;
1385 if (i + cluster > ra_index) {
1386 ra_index = max(i, ra_index);
1387 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1389 ra_index += cluster;
1392 mutex_lock(&inode->i_mutex);
1393 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1394 BTRFS_I(inode)->force_compress = compress_type;
1395 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1397 mutex_unlock(&inode->i_mutex);
1401 defrag_count += ret;
1402 balance_dirty_pages_ratelimited(inode->i_mapping);
1403 mutex_unlock(&inode->i_mutex);
1406 if (newer_off == (u64)-1)
1412 newer_off = max(newer_off + 1,
1413 (u64)i << PAGE_CACHE_SHIFT);
1415 ret = find_new_extents(root, inode, newer_than,
1416 &newer_off, SZ_64K);
1418 range->start = newer_off;
1419 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1426 last_len += ret << PAGE_CACHE_SHIFT;
1434 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1435 filemap_flush(inode->i_mapping);
1436 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1437 &BTRFS_I(inode)->runtime_flags))
1438 filemap_flush(inode->i_mapping);
1441 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1442 /* the filemap_flush will queue IO into the worker threads, but
1443 * we have to make sure the IO is actually started and that
1444 * ordered extents get created before we return
1446 atomic_inc(&root->fs_info->async_submit_draining);
1447 while (atomic_read(&root->fs_info->nr_async_submits) ||
1448 atomic_read(&root->fs_info->async_delalloc_pages)) {
1449 wait_event(root->fs_info->async_submit_wait,
1450 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1451 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1453 atomic_dec(&root->fs_info->async_submit_draining);
1456 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1457 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1463 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1464 mutex_lock(&inode->i_mutex);
1465 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1466 mutex_unlock(&inode->i_mutex);
1474 static noinline int btrfs_ioctl_resize(struct file *file,
1480 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
1481 struct btrfs_ioctl_vol_args *vol_args;
1482 struct btrfs_trans_handle *trans;
1483 struct btrfs_device *device = NULL;
1486 char *devstr = NULL;
1490 if (!capable(CAP_SYS_ADMIN))
1493 ret = mnt_want_write_file(file);
1497 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1499 mnt_drop_write_file(file);
1500 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1503 mutex_lock(&root->fs_info->volume_mutex);
1504 vol_args = memdup_user(arg, sizeof(*vol_args));
1505 if (IS_ERR(vol_args)) {
1506 ret = PTR_ERR(vol_args);
1510 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1512 sizestr = vol_args->name;
1513 devstr = strchr(sizestr, ':');
1515 sizestr = devstr + 1;
1517 devstr = vol_args->name;
1518 ret = kstrtoull(devstr, 10, &devid);
1525 btrfs_info(root->fs_info, "resizing devid %llu", devid);
1528 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
1530 btrfs_info(root->fs_info, "resizer unable to find device %llu",
1536 if (!device->writeable) {
1537 btrfs_info(root->fs_info,
1538 "resizer unable to apply on readonly device %llu",
1544 if (!strcmp(sizestr, "max"))
1545 new_size = device->bdev->bd_inode->i_size;
1547 if (sizestr[0] == '-') {
1550 } else if (sizestr[0] == '+') {
1554 new_size = memparse(sizestr, &retptr);
1555 if (*retptr != '\0' || new_size == 0) {
1561 if (device->is_tgtdev_for_dev_replace) {
1566 old_size = btrfs_device_get_total_bytes(device);
1569 if (new_size > old_size) {
1573 new_size = old_size - new_size;
1574 } else if (mod > 0) {
1575 if (new_size > ULLONG_MAX - old_size) {
1579 new_size = old_size + new_size;
1582 if (new_size < SZ_256M) {
1586 if (new_size > device->bdev->bd_inode->i_size) {
1591 new_size = div_u64(new_size, root->sectorsize);
1592 new_size *= root->sectorsize;
1594 btrfs_info_in_rcu(root->fs_info, "new size for %s is %llu",
1595 rcu_str_deref(device->name), new_size);
1597 if (new_size > old_size) {
1598 trans = btrfs_start_transaction(root, 0);
1599 if (IS_ERR(trans)) {
1600 ret = PTR_ERR(trans);
1603 ret = btrfs_grow_device(trans, device, new_size);
1604 btrfs_commit_transaction(trans, root);
1605 } else if (new_size < old_size) {
1606 ret = btrfs_shrink_device(device, new_size);
1607 } /* equal, nothing need to do */
1612 mutex_unlock(&root->fs_info->volume_mutex);
1613 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
1614 mnt_drop_write_file(file);
1618 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1619 char *name, unsigned long fd, int subvol,
1620 u64 *transid, bool readonly,
1621 struct btrfs_qgroup_inherit *inherit)
1626 ret = mnt_want_write_file(file);
1630 namelen = strlen(name);
1631 if (strchr(name, '/')) {
1633 goto out_drop_write;
1636 if (name[0] == '.' &&
1637 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1639 goto out_drop_write;
1643 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1644 NULL, transid, readonly, inherit);
1646 struct fd src = fdget(fd);
1647 struct inode *src_inode;
1650 goto out_drop_write;
1653 src_inode = file_inode(src.file);
1654 if (src_inode->i_sb != file_inode(file)->i_sb) {
1655 btrfs_info(BTRFS_I(src_inode)->root->fs_info,
1656 "Snapshot src from another FS");
1658 } else if (!inode_owner_or_capable(src_inode)) {
1660 * Subvolume creation is not restricted, but snapshots
1661 * are limited to own subvolumes only
1665 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1666 BTRFS_I(src_inode)->root,
1667 transid, readonly, inherit);
1672 mnt_drop_write_file(file);
1677 static noinline int btrfs_ioctl_snap_create(struct file *file,
1678 void __user *arg, int subvol)
1680 struct btrfs_ioctl_vol_args *vol_args;
1683 vol_args = memdup_user(arg, sizeof(*vol_args));
1684 if (IS_ERR(vol_args))
1685 return PTR_ERR(vol_args);
1686 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1688 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1689 vol_args->fd, subvol,
1696 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1697 void __user *arg, int subvol)
1699 struct btrfs_ioctl_vol_args_v2 *vol_args;
1703 bool readonly = false;
1704 struct btrfs_qgroup_inherit *inherit = NULL;
1706 vol_args = memdup_user(arg, sizeof(*vol_args));
1707 if (IS_ERR(vol_args))
1708 return PTR_ERR(vol_args);
1709 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1711 if (vol_args->flags &
1712 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1713 BTRFS_SUBVOL_QGROUP_INHERIT)) {
1718 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1720 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1722 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1723 if (vol_args->size > PAGE_CACHE_SIZE) {
1727 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1728 if (IS_ERR(inherit)) {
1729 ret = PTR_ERR(inherit);
1734 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1735 vol_args->fd, subvol, ptr,
1740 if (ptr && copy_to_user(arg +
1741 offsetof(struct btrfs_ioctl_vol_args_v2,
1753 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1756 struct inode *inode = file_inode(file);
1757 struct btrfs_root *root = BTRFS_I(inode)->root;
1761 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1764 down_read(&root->fs_info->subvol_sem);
1765 if (btrfs_root_readonly(root))
1766 flags |= BTRFS_SUBVOL_RDONLY;
1767 up_read(&root->fs_info->subvol_sem);
1769 if (copy_to_user(arg, &flags, sizeof(flags)))
1775 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1778 struct inode *inode = file_inode(file);
1779 struct btrfs_root *root = BTRFS_I(inode)->root;
1780 struct btrfs_trans_handle *trans;
1785 if (!inode_owner_or_capable(inode))
1788 ret = mnt_want_write_file(file);
1792 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1794 goto out_drop_write;
1797 if (copy_from_user(&flags, arg, sizeof(flags))) {
1799 goto out_drop_write;
1802 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1804 goto out_drop_write;
1807 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1809 goto out_drop_write;
1812 down_write(&root->fs_info->subvol_sem);
1815 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1818 root_flags = btrfs_root_flags(&root->root_item);
1819 if (flags & BTRFS_SUBVOL_RDONLY) {
1820 btrfs_set_root_flags(&root->root_item,
1821 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1824 * Block RO -> RW transition if this subvolume is involved in
1827 spin_lock(&root->root_item_lock);
1828 if (root->send_in_progress == 0) {
1829 btrfs_set_root_flags(&root->root_item,
1830 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1831 spin_unlock(&root->root_item_lock);
1833 spin_unlock(&root->root_item_lock);
1834 btrfs_warn(root->fs_info,
1835 "Attempt to set subvolume %llu read-write during send",
1836 root->root_key.objectid);
1842 trans = btrfs_start_transaction(root, 1);
1843 if (IS_ERR(trans)) {
1844 ret = PTR_ERR(trans);
1848 ret = btrfs_update_root(trans, root->fs_info->tree_root,
1849 &root->root_key, &root->root_item);
1851 btrfs_commit_transaction(trans, root);
1854 btrfs_set_root_flags(&root->root_item, root_flags);
1856 up_write(&root->fs_info->subvol_sem);
1858 mnt_drop_write_file(file);
1864 * helper to check if the subvolume references other subvolumes
1866 static noinline int may_destroy_subvol(struct btrfs_root *root)
1868 struct btrfs_path *path;
1869 struct btrfs_dir_item *di;
1870 struct btrfs_key key;
1874 path = btrfs_alloc_path();
1878 /* Make sure this root isn't set as the default subvol */
1879 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1880 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1881 dir_id, "default", 7, 0);
1882 if (di && !IS_ERR(di)) {
1883 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1884 if (key.objectid == root->root_key.objectid) {
1886 btrfs_err(root->fs_info, "deleting default subvolume "
1887 "%llu is not allowed", key.objectid);
1890 btrfs_release_path(path);
1893 key.objectid = root->root_key.objectid;
1894 key.type = BTRFS_ROOT_REF_KEY;
1895 key.offset = (u64)-1;
1897 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1904 if (path->slots[0] > 0) {
1906 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1907 if (key.objectid == root->root_key.objectid &&
1908 key.type == BTRFS_ROOT_REF_KEY)
1912 btrfs_free_path(path);
1916 static noinline int key_in_sk(struct btrfs_key *key,
1917 struct btrfs_ioctl_search_key *sk)
1919 struct btrfs_key test;
1922 test.objectid = sk->min_objectid;
1923 test.type = sk->min_type;
1924 test.offset = sk->min_offset;
1926 ret = btrfs_comp_cpu_keys(key, &test);
1930 test.objectid = sk->max_objectid;
1931 test.type = sk->max_type;
1932 test.offset = sk->max_offset;
1934 ret = btrfs_comp_cpu_keys(key, &test);
1940 static noinline int copy_to_sk(struct btrfs_root *root,
1941 struct btrfs_path *path,
1942 struct btrfs_key *key,
1943 struct btrfs_ioctl_search_key *sk,
1946 unsigned long *sk_offset,
1950 struct extent_buffer *leaf;
1951 struct btrfs_ioctl_search_header sh;
1952 struct btrfs_key test;
1953 unsigned long item_off;
1954 unsigned long item_len;
1960 leaf = path->nodes[0];
1961 slot = path->slots[0];
1962 nritems = btrfs_header_nritems(leaf);
1964 if (btrfs_header_generation(leaf) > sk->max_transid) {
1968 found_transid = btrfs_header_generation(leaf);
1970 for (i = slot; i < nritems; i++) {
1971 item_off = btrfs_item_ptr_offset(leaf, i);
1972 item_len = btrfs_item_size_nr(leaf, i);
1974 btrfs_item_key_to_cpu(leaf, key, i);
1975 if (!key_in_sk(key, sk))
1978 if (sizeof(sh) + item_len > *buf_size) {
1985 * return one empty item back for v1, which does not
1989 *buf_size = sizeof(sh) + item_len;
1994 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
1999 sh.objectid = key->objectid;
2000 sh.offset = key->offset;
2001 sh.type = key->type;
2003 sh.transid = found_transid;
2005 /* copy search result header */
2006 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
2011 *sk_offset += sizeof(sh);
2014 char __user *up = ubuf + *sk_offset;
2016 if (read_extent_buffer_to_user(leaf, up,
2017 item_off, item_len)) {
2022 *sk_offset += item_len;
2026 if (ret) /* -EOVERFLOW from above */
2029 if (*num_found >= sk->nr_items) {
2036 test.objectid = sk->max_objectid;
2037 test.type = sk->max_type;
2038 test.offset = sk->max_offset;
2039 if (btrfs_comp_cpu_keys(key, &test) >= 0)
2041 else if (key->offset < (u64)-1)
2043 else if (key->type < (u8)-1) {
2046 } else if (key->objectid < (u64)-1) {
2054 * 0: all items from this leaf copied, continue with next
2055 * 1: * more items can be copied, but unused buffer is too small
2056 * * all items were found
2057 * Either way, it will stops the loop which iterates to the next
2059 * -EOVERFLOW: item was to large for buffer
2060 * -EFAULT: could not copy extent buffer back to userspace
2065 static noinline int search_ioctl(struct inode *inode,
2066 struct btrfs_ioctl_search_key *sk,
2070 struct btrfs_root *root;
2071 struct btrfs_key key;
2072 struct btrfs_path *path;
2073 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
2076 unsigned long sk_offset = 0;
2078 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2079 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2083 path = btrfs_alloc_path();
2087 if (sk->tree_id == 0) {
2088 /* search the root of the inode that was passed */
2089 root = BTRFS_I(inode)->root;
2091 key.objectid = sk->tree_id;
2092 key.type = BTRFS_ROOT_ITEM_KEY;
2093 key.offset = (u64)-1;
2094 root = btrfs_read_fs_root_no_name(info, &key);
2096 btrfs_err(info, "could not find root %llu",
2098 btrfs_free_path(path);
2103 key.objectid = sk->min_objectid;
2104 key.type = sk->min_type;
2105 key.offset = sk->min_offset;
2108 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2114 ret = copy_to_sk(root, path, &key, sk, buf_size, ubuf,
2115 &sk_offset, &num_found);
2116 btrfs_release_path(path);
2124 sk->nr_items = num_found;
2125 btrfs_free_path(path);
2129 static noinline int btrfs_ioctl_tree_search(struct file *file,
2132 struct btrfs_ioctl_search_args __user *uargs;
2133 struct btrfs_ioctl_search_key sk;
2134 struct inode *inode;
2138 if (!capable(CAP_SYS_ADMIN))
2141 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2143 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2146 buf_size = sizeof(uargs->buf);
2148 inode = file_inode(file);
2149 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2152 * In the origin implementation an overflow is handled by returning a
2153 * search header with a len of zero, so reset ret.
2155 if (ret == -EOVERFLOW)
2158 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2163 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2166 struct btrfs_ioctl_search_args_v2 __user *uarg;
2167 struct btrfs_ioctl_search_args_v2 args;
2168 struct inode *inode;
2171 const size_t buf_limit = SZ_16M;
2173 if (!capable(CAP_SYS_ADMIN))
2176 /* copy search header and buffer size */
2177 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2178 if (copy_from_user(&args, uarg, sizeof(args)))
2181 buf_size = args.buf_size;
2183 if (buf_size < sizeof(struct btrfs_ioctl_search_header))
2186 /* limit result size to 16MB */
2187 if (buf_size > buf_limit)
2188 buf_size = buf_limit;
2190 inode = file_inode(file);
2191 ret = search_ioctl(inode, &args.key, &buf_size,
2192 (char *)(&uarg->buf[0]));
2193 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2195 else if (ret == -EOVERFLOW &&
2196 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2203 * Search INODE_REFs to identify path name of 'dirid' directory
2204 * in a 'tree_id' tree. and sets path name to 'name'.
2206 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2207 u64 tree_id, u64 dirid, char *name)
2209 struct btrfs_root *root;
2210 struct btrfs_key key;
2216 struct btrfs_inode_ref *iref;
2217 struct extent_buffer *l;
2218 struct btrfs_path *path;
2220 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2225 path = btrfs_alloc_path();
2229 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
2231 key.objectid = tree_id;
2232 key.type = BTRFS_ROOT_ITEM_KEY;
2233 key.offset = (u64)-1;
2234 root = btrfs_read_fs_root_no_name(info, &key);
2236 btrfs_err(info, "could not find root %llu", tree_id);
2241 key.objectid = dirid;
2242 key.type = BTRFS_INODE_REF_KEY;
2243 key.offset = (u64)-1;
2246 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2250 ret = btrfs_previous_item(root, path, dirid,
2251 BTRFS_INODE_REF_KEY);
2261 slot = path->slots[0];
2262 btrfs_item_key_to_cpu(l, &key, slot);
2264 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2265 len = btrfs_inode_ref_name_len(l, iref);
2267 total_len += len + 1;
2269 ret = -ENAMETOOLONG;
2274 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2276 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2279 btrfs_release_path(path);
2280 key.objectid = key.offset;
2281 key.offset = (u64)-1;
2282 dirid = key.objectid;
2284 memmove(name, ptr, total_len);
2285 name[total_len] = '\0';
2288 btrfs_free_path(path);
2292 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2295 struct btrfs_ioctl_ino_lookup_args *args;
2296 struct inode *inode;
2299 args = memdup_user(argp, sizeof(*args));
2301 return PTR_ERR(args);
2303 inode = file_inode(file);
2306 * Unprivileged query to obtain the containing subvolume root id. The
2307 * path is reset so it's consistent with btrfs_search_path_in_tree.
2309 if (args->treeid == 0)
2310 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2312 if (args->objectid == BTRFS_FIRST_FREE_OBJECTID) {
2317 if (!capable(CAP_SYS_ADMIN)) {
2322 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2323 args->treeid, args->objectid,
2327 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2334 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2337 struct dentry *parent = file->f_path.dentry;
2338 struct dentry *dentry;
2339 struct inode *dir = d_inode(parent);
2340 struct inode *inode;
2341 struct btrfs_root *root = BTRFS_I(dir)->root;
2342 struct btrfs_root *dest = NULL;
2343 struct btrfs_ioctl_vol_args *vol_args;
2344 struct btrfs_trans_handle *trans;
2345 struct btrfs_block_rsv block_rsv;
2347 u64 qgroup_reserved;
2352 vol_args = memdup_user(arg, sizeof(*vol_args));
2353 if (IS_ERR(vol_args))
2354 return PTR_ERR(vol_args);
2356 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2357 namelen = strlen(vol_args->name);
2358 if (strchr(vol_args->name, '/') ||
2359 strncmp(vol_args->name, "..", namelen) == 0) {
2364 err = mnt_want_write_file(file);
2369 err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2371 goto out_drop_write;
2372 dentry = lookup_one_len(vol_args->name, parent, namelen);
2373 if (IS_ERR(dentry)) {
2374 err = PTR_ERR(dentry);
2375 goto out_unlock_dir;
2378 if (d_really_is_negative(dentry)) {
2383 inode = d_inode(dentry);
2384 dest = BTRFS_I(inode)->root;
2385 if (!capable(CAP_SYS_ADMIN)) {
2387 * Regular user. Only allow this with a special mount
2388 * option, when the user has write+exec access to the
2389 * subvol root, and when rmdir(2) would have been
2392 * Note that this is _not_ check that the subvol is
2393 * empty or doesn't contain data that we wouldn't
2394 * otherwise be able to delete.
2396 * Users who want to delete empty subvols should try
2400 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2404 * Do not allow deletion if the parent dir is the same
2405 * as the dir to be deleted. That means the ioctl
2406 * must be called on the dentry referencing the root
2407 * of the subvol, not a random directory contained
2414 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2419 /* check if subvolume may be deleted by a user */
2420 err = btrfs_may_delete(dir, dentry, 1);
2424 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2429 mutex_lock(&inode->i_mutex);
2432 * Don't allow to delete a subvolume with send in progress. This is
2433 * inside the i_mutex so the error handling that has to drop the bit
2434 * again is not run concurrently.
2436 spin_lock(&dest->root_item_lock);
2437 root_flags = btrfs_root_flags(&dest->root_item);
2438 if (dest->send_in_progress == 0) {
2439 btrfs_set_root_flags(&dest->root_item,
2440 root_flags | BTRFS_ROOT_SUBVOL_DEAD);
2441 spin_unlock(&dest->root_item_lock);
2443 spin_unlock(&dest->root_item_lock);
2444 btrfs_warn(root->fs_info,
2445 "Attempt to delete subvolume %llu during send",
2446 dest->root_key.objectid);
2448 goto out_unlock_inode;
2451 down_write(&root->fs_info->subvol_sem);
2453 err = may_destroy_subvol(dest);
2457 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2459 * One for dir inode, two for dir entries, two for root
2462 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2463 5, &qgroup_reserved, true);
2467 trans = btrfs_start_transaction(root, 0);
2468 if (IS_ERR(trans)) {
2469 err = PTR_ERR(trans);
2472 trans->block_rsv = &block_rsv;
2473 trans->bytes_reserved = block_rsv.size;
2475 ret = btrfs_unlink_subvol(trans, root, dir,
2476 dest->root_key.objectid,
2477 dentry->d_name.name,
2478 dentry->d_name.len);
2481 btrfs_abort_transaction(trans, root, ret);
2485 btrfs_record_root_in_trans(trans, dest);
2487 memset(&dest->root_item.drop_progress, 0,
2488 sizeof(dest->root_item.drop_progress));
2489 dest->root_item.drop_level = 0;
2490 btrfs_set_root_refs(&dest->root_item, 0);
2492 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
2493 ret = btrfs_insert_orphan_item(trans,
2494 root->fs_info->tree_root,
2495 dest->root_key.objectid);
2497 btrfs_abort_transaction(trans, root, ret);
2503 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2504 dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
2505 dest->root_key.objectid);
2506 if (ret && ret != -ENOENT) {
2507 btrfs_abort_transaction(trans, root, ret);
2511 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2512 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2513 dest->root_item.received_uuid,
2514 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2515 dest->root_key.objectid);
2516 if (ret && ret != -ENOENT) {
2517 btrfs_abort_transaction(trans, root, ret);
2524 trans->block_rsv = NULL;
2525 trans->bytes_reserved = 0;
2526 ret = btrfs_end_transaction(trans, root);
2529 inode->i_flags |= S_DEAD;
2531 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
2533 up_write(&root->fs_info->subvol_sem);
2535 spin_lock(&dest->root_item_lock);
2536 root_flags = btrfs_root_flags(&dest->root_item);
2537 btrfs_set_root_flags(&dest->root_item,
2538 root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
2539 spin_unlock(&dest->root_item_lock);
2542 mutex_unlock(&inode->i_mutex);
2544 d_invalidate(dentry);
2545 btrfs_invalidate_inodes(dest);
2547 ASSERT(dest->send_in_progress == 0);
2550 if (dest->ino_cache_inode) {
2551 iput(dest->ino_cache_inode);
2552 dest->ino_cache_inode = NULL;
2558 mutex_unlock(&dir->i_mutex);
2560 mnt_drop_write_file(file);
2566 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2568 struct inode *inode = file_inode(file);
2569 struct btrfs_root *root = BTRFS_I(inode)->root;
2570 struct btrfs_ioctl_defrag_range_args *range;
2573 ret = mnt_want_write_file(file);
2577 if (btrfs_root_readonly(root)) {
2582 switch (inode->i_mode & S_IFMT) {
2584 if (!capable(CAP_SYS_ADMIN)) {
2588 ret = btrfs_defrag_root(root);
2591 ret = btrfs_defrag_root(root->fs_info->extent_root);
2594 if (!(file->f_mode & FMODE_WRITE)) {
2599 range = kzalloc(sizeof(*range), GFP_KERNEL);
2606 if (copy_from_user(range, argp,
2612 /* compression requires us to start the IO */
2613 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2614 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2615 range->extent_thresh = (u32)-1;
2618 /* the rest are all set to zero by kzalloc */
2619 range->len = (u64)-1;
2621 ret = btrfs_defrag_file(file_inode(file), file,
2631 mnt_drop_write_file(file);
2635 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2637 struct btrfs_ioctl_vol_args *vol_args;
2640 if (!capable(CAP_SYS_ADMIN))
2643 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2645 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2648 mutex_lock(&root->fs_info->volume_mutex);
2649 vol_args = memdup_user(arg, sizeof(*vol_args));
2650 if (IS_ERR(vol_args)) {
2651 ret = PTR_ERR(vol_args);
2655 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2656 ret = btrfs_init_new_device(root, vol_args->name);
2659 btrfs_info(root->fs_info, "disk added %s",vol_args->name);
2663 mutex_unlock(&root->fs_info->volume_mutex);
2664 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2668 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2670 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2671 struct btrfs_ioctl_vol_args *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 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2689 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2691 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2695 mutex_lock(&root->fs_info->volume_mutex);
2696 ret = btrfs_rm_device(root, vol_args->name);
2697 mutex_unlock(&root->fs_info->volume_mutex);
2698 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2701 btrfs_info(root->fs_info, "disk deleted %s",vol_args->name);
2706 mnt_drop_write_file(file);
2710 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2712 struct btrfs_ioctl_fs_info_args *fi_args;
2713 struct btrfs_device *device;
2714 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2717 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2721 mutex_lock(&fs_devices->device_list_mutex);
2722 fi_args->num_devices = fs_devices->num_devices;
2723 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2725 list_for_each_entry(device, &fs_devices->devices, dev_list) {
2726 if (device->devid > fi_args->max_id)
2727 fi_args->max_id = device->devid;
2729 mutex_unlock(&fs_devices->device_list_mutex);
2731 fi_args->nodesize = root->fs_info->super_copy->nodesize;
2732 fi_args->sectorsize = root->fs_info->super_copy->sectorsize;
2733 fi_args->clone_alignment = root->fs_info->super_copy->sectorsize;
2735 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2742 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2744 struct btrfs_ioctl_dev_info_args *di_args;
2745 struct btrfs_device *dev;
2746 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2748 char *s_uuid = NULL;
2750 di_args = memdup_user(arg, sizeof(*di_args));
2751 if (IS_ERR(di_args))
2752 return PTR_ERR(di_args);
2754 if (!btrfs_is_empty_uuid(di_args->uuid))
2755 s_uuid = di_args->uuid;
2757 mutex_lock(&fs_devices->device_list_mutex);
2758 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
2765 di_args->devid = dev->devid;
2766 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2767 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
2768 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2770 struct rcu_string *name;
2773 name = rcu_dereference(dev->name);
2774 strncpy(di_args->path, name->str, sizeof(di_args->path));
2776 di_args->path[sizeof(di_args->path) - 1] = 0;
2778 di_args->path[0] = '\0';
2782 mutex_unlock(&fs_devices->device_list_mutex);
2783 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2790 static struct page *extent_same_get_page(struct inode *inode, pgoff_t index)
2793 struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
2795 page = grab_cache_page(inode->i_mapping, index);
2799 if (!PageUptodate(page)) {
2800 if (extent_read_full_page_nolock(tree, page, btrfs_get_extent,
2804 if (!PageUptodate(page)) {
2806 page_cache_release(page);
2815 static int gather_extent_pages(struct inode *inode, struct page **pages,
2816 int num_pages, u64 off)
2819 pgoff_t index = off >> PAGE_CACHE_SHIFT;
2821 for (i = 0; i < num_pages; i++) {
2822 pages[i] = extent_same_get_page(inode, index + i);
2829 static inline void lock_extent_range(struct inode *inode, u64 off, u64 len)
2831 /* do any pending delalloc/csum calc on src, one way or
2832 another, and lock file content */
2834 struct btrfs_ordered_extent *ordered;
2835 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2836 ordered = btrfs_lookup_first_ordered_extent(inode,
2839 ordered->file_offset + ordered->len <= off ||
2840 ordered->file_offset >= off + len) &&
2841 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2842 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2844 btrfs_put_ordered_extent(ordered);
2847 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2849 btrfs_put_ordered_extent(ordered);
2850 btrfs_wait_ordered_range(inode, off, len);
2854 static void btrfs_double_inode_unlock(struct inode *inode1, struct inode *inode2)
2856 mutex_unlock(&inode1->i_mutex);
2857 mutex_unlock(&inode2->i_mutex);
2860 static void btrfs_double_inode_lock(struct inode *inode1, struct inode *inode2)
2862 if (inode1 < inode2)
2863 swap(inode1, inode2);
2865 mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT);
2866 mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD);
2869 static void btrfs_double_extent_unlock(struct inode *inode1, u64 loff1,
2870 struct inode *inode2, u64 loff2, u64 len)
2872 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2873 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2876 static void btrfs_double_extent_lock(struct inode *inode1, u64 loff1,
2877 struct inode *inode2, u64 loff2, u64 len)
2879 if (inode1 < inode2) {
2880 swap(inode1, inode2);
2883 lock_extent_range(inode1, loff1, len);
2884 lock_extent_range(inode2, loff2, len);
2889 struct page **src_pages;
2890 struct page **dst_pages;
2893 static void btrfs_cmp_data_free(struct cmp_pages *cmp)
2898 for (i = 0; i < cmp->num_pages; i++) {
2899 pg = cmp->src_pages[i];
2901 page_cache_release(pg);
2902 pg = cmp->dst_pages[i];
2904 page_cache_release(pg);
2906 kfree(cmp->src_pages);
2907 kfree(cmp->dst_pages);
2910 static int btrfs_cmp_data_prepare(struct inode *src, u64 loff,
2911 struct inode *dst, u64 dst_loff,
2912 u64 len, struct cmp_pages *cmp)
2915 int num_pages = PAGE_CACHE_ALIGN(len) >> PAGE_CACHE_SHIFT;
2916 struct page **src_pgarr, **dst_pgarr;
2919 * We must gather up all the pages before we initiate our
2920 * extent locking. We use an array for the page pointers. Size
2921 * of the array is bounded by len, which is in turn bounded by
2922 * BTRFS_MAX_DEDUPE_LEN.
2924 src_pgarr = kzalloc(num_pages * sizeof(struct page *), GFP_NOFS);
2925 dst_pgarr = kzalloc(num_pages * sizeof(struct page *), GFP_NOFS);
2926 if (!src_pgarr || !dst_pgarr) {
2931 cmp->num_pages = num_pages;
2932 cmp->src_pages = src_pgarr;
2933 cmp->dst_pages = dst_pgarr;
2935 ret = gather_extent_pages(src, cmp->src_pages, cmp->num_pages, loff);
2939 ret = gather_extent_pages(dst, cmp->dst_pages, cmp->num_pages, dst_loff);
2943 btrfs_cmp_data_free(cmp);
2947 static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
2948 u64 dst_loff, u64 len, struct cmp_pages *cmp)
2952 struct page *src_page, *dst_page;
2953 unsigned int cmp_len = PAGE_CACHE_SIZE;
2954 void *addr, *dst_addr;
2958 if (len < PAGE_CACHE_SIZE)
2961 BUG_ON(i >= cmp->num_pages);
2963 src_page = cmp->src_pages[i];
2964 dst_page = cmp->dst_pages[i];
2966 addr = kmap_atomic(src_page);
2967 dst_addr = kmap_atomic(dst_page);
2969 flush_dcache_page(src_page);
2970 flush_dcache_page(dst_page);
2972 if (memcmp(addr, dst_addr, cmp_len))
2975 kunmap_atomic(addr);
2976 kunmap_atomic(dst_addr);
2988 static int extent_same_check_offsets(struct inode *inode, u64 off, u64 *plen,
2992 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
2994 if (off + olen > inode->i_size || off + olen < off)
2997 /* if we extend to eof, continue to block boundary */
2998 if (off + len == inode->i_size)
2999 *plen = len = ALIGN(inode->i_size, bs) - off;
3001 /* Check that we are block aligned - btrfs_clone() requires this */
3002 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
3008 static int btrfs_extent_same(struct inode *src, u64 loff, u64 olen,
3009 struct inode *dst, u64 dst_loff)
3013 struct cmp_pages cmp;
3015 u64 same_lock_start = 0;
3016 u64 same_lock_len = 0;
3025 mutex_lock(&src->i_mutex);
3027 ret = extent_same_check_offsets(src, loff, &len, olen);
3032 * Single inode case wants the same checks, except we
3033 * don't want our length pushed out past i_size as
3034 * comparing that data range makes no sense.
3036 * extent_same_check_offsets() will do this for an
3037 * unaligned length at i_size, so catch it here and
3038 * reject the request.
3040 * This effectively means we require aligned extents
3041 * for the single-inode case, whereas the other cases
3042 * allow an unaligned length so long as it ends at
3050 /* Check for overlapping ranges */
3051 if (dst_loff + len > loff && dst_loff < loff + len) {
3056 same_lock_start = min_t(u64, loff, dst_loff);
3057 same_lock_len = max_t(u64, loff, dst_loff) + len - same_lock_start;
3059 btrfs_double_inode_lock(src, dst);
3061 ret = extent_same_check_offsets(src, loff, &len, olen);
3065 ret = extent_same_check_offsets(dst, dst_loff, &len, olen);
3070 /* don't make the dst file partly checksummed */
3071 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3072 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
3077 ret = btrfs_cmp_data_prepare(src, loff, dst, dst_loff, olen, &cmp);
3082 lock_extent_range(src, same_lock_start, same_lock_len);
3084 btrfs_double_extent_lock(src, loff, dst, dst_loff, len);
3086 /* pass original length for comparison so we stay within i_size */
3087 ret = btrfs_cmp_data(src, loff, dst, dst_loff, olen, &cmp);
3089 ret = btrfs_clone(src, dst, loff, olen, len, dst_loff, 1);
3092 unlock_extent(&BTRFS_I(src)->io_tree, same_lock_start,
3093 same_lock_start + same_lock_len - 1);
3095 btrfs_double_extent_unlock(src, loff, dst, dst_loff, len);
3097 btrfs_cmp_data_free(&cmp);
3100 mutex_unlock(&src->i_mutex);
3102 btrfs_double_inode_unlock(src, dst);
3107 #define BTRFS_MAX_DEDUPE_LEN SZ_16M
3109 ssize_t btrfs_dedupe_file_range(struct file *src_file, u64 loff, u64 olen,
3110 struct file *dst_file, u64 dst_loff)
3112 struct inode *src = file_inode(src_file);
3113 struct inode *dst = file_inode(dst_file);
3114 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
3117 if (olen > BTRFS_MAX_DEDUPE_LEN)
3118 olen = BTRFS_MAX_DEDUPE_LEN;
3120 if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) {
3122 * Btrfs does not support blocksize < page_size. As a
3123 * result, btrfs_cmp_data() won't correctly handle
3124 * this situation without an update.
3129 res = btrfs_extent_same(src, loff, olen, dst, dst_loff);
3135 static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3136 struct inode *inode,
3142 struct btrfs_root *root = BTRFS_I(inode)->root;
3145 inode_inc_iversion(inode);
3146 if (!no_time_update)
3147 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3149 * We round up to the block size at eof when determining which
3150 * extents to clone above, but shouldn't round up the file size.
3152 if (endoff > destoff + olen)
3153 endoff = destoff + olen;
3154 if (endoff > inode->i_size)
3155 btrfs_i_size_write(inode, endoff);
3157 ret = btrfs_update_inode(trans, root, inode);
3159 btrfs_abort_transaction(trans, root, ret);
3160 btrfs_end_transaction(trans, root);
3163 ret = btrfs_end_transaction(trans, root);
3168 static void clone_update_extent_map(struct inode *inode,
3169 const struct btrfs_trans_handle *trans,
3170 const struct btrfs_path *path,
3171 const u64 hole_offset,
3174 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
3175 struct extent_map *em;
3178 em = alloc_extent_map();
3180 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3181 &BTRFS_I(inode)->runtime_flags);
3186 struct btrfs_file_extent_item *fi;
3188 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3189 struct btrfs_file_extent_item);
3190 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3191 em->generation = -1;
3192 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3193 BTRFS_FILE_EXTENT_INLINE)
3194 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3195 &BTRFS_I(inode)->runtime_flags);
3197 em->start = hole_offset;
3199 em->ram_bytes = em->len;
3200 em->orig_start = hole_offset;
3201 em->block_start = EXTENT_MAP_HOLE;
3203 em->orig_block_len = 0;
3204 em->compress_type = BTRFS_COMPRESS_NONE;
3205 em->generation = trans->transid;
3209 write_lock(&em_tree->lock);
3210 ret = add_extent_mapping(em_tree, em, 1);
3211 write_unlock(&em_tree->lock);
3212 if (ret != -EEXIST) {
3213 free_extent_map(em);
3216 btrfs_drop_extent_cache(inode, em->start,
3217 em->start + em->len - 1, 0);
3221 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3222 &BTRFS_I(inode)->runtime_flags);
3226 * Make sure we do not end up inserting an inline extent into a file that has
3227 * already other (non-inline) extents. If a file has an inline extent it can
3228 * not have any other extents and the (single) inline extent must start at the
3229 * file offset 0. Failing to respect these rules will lead to file corruption,
3230 * resulting in EIO errors on read/write operations, hitting BUG_ON's in mm, etc
3232 * We can have extents that have been already written to disk or we can have
3233 * dirty ranges still in delalloc, in which case the extent maps and items are
3234 * created only when we run delalloc, and the delalloc ranges might fall outside
3235 * the range we are currently locking in the inode's io tree. So we check the
3236 * inode's i_size because of that (i_size updates are done while holding the
3237 * i_mutex, which we are holding here).
3238 * We also check to see if the inode has a size not greater than "datal" but has
3239 * extents beyond it, due to an fallocate with FALLOC_FL_KEEP_SIZE (and we are
3240 * protected against such concurrent fallocate calls by the i_mutex).
3242 * If the file has no extents but a size greater than datal, do not allow the
3243 * copy because we would need turn the inline extent into a non-inline one (even
3244 * with NO_HOLES enabled). If we find our destination inode only has one inline
3245 * extent, just overwrite it with the source inline extent if its size is less
3246 * than the source extent's size, or we could copy the source inline extent's
3247 * data into the destination inode's inline extent if the later is greater then
3250 static int clone_copy_inline_extent(struct inode *src,
3252 struct btrfs_trans_handle *trans,
3253 struct btrfs_path *path,
3254 struct btrfs_key *new_key,
3255 const u64 drop_start,
3261 struct btrfs_root *root = BTRFS_I(dst)->root;
3262 const u64 aligned_end = ALIGN(new_key->offset + datal,
3265 struct btrfs_key key;
3267 if (new_key->offset > 0)
3270 key.objectid = btrfs_ino(dst);
3271 key.type = BTRFS_EXTENT_DATA_KEY;
3273 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3276 } else if (ret > 0) {
3277 if (path->slots[0] >= btrfs_header_nritems(path->nodes[0])) {
3278 ret = btrfs_next_leaf(root, path);
3282 goto copy_inline_extent;
3284 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
3285 if (key.objectid == btrfs_ino(dst) &&
3286 key.type == BTRFS_EXTENT_DATA_KEY) {
3287 ASSERT(key.offset > 0);
3290 } else if (i_size_read(dst) <= datal) {
3291 struct btrfs_file_extent_item *ei;
3295 * If the file size is <= datal, make sure there are no other
3296 * extents following (can happen do to an fallocate call with
3297 * the flag FALLOC_FL_KEEP_SIZE).
3299 ei = btrfs_item_ptr(path->nodes[0], path->slots[0],
3300 struct btrfs_file_extent_item);
3302 * If it's an inline extent, it can not have other extents
3305 if (btrfs_file_extent_type(path->nodes[0], ei) ==
3306 BTRFS_FILE_EXTENT_INLINE)
3307 goto copy_inline_extent;
3309 ext_len = btrfs_file_extent_num_bytes(path->nodes[0], ei);
3310 if (ext_len > aligned_end)
3313 ret = btrfs_next_item(root, path);
3316 } else if (ret == 0) {
3317 btrfs_item_key_to_cpu(path->nodes[0], &key,
3319 if (key.objectid == btrfs_ino(dst) &&
3320 key.type == BTRFS_EXTENT_DATA_KEY)
3327 * We have no extent items, or we have an extent at offset 0 which may
3328 * or may not be inlined. All these cases are dealt the same way.
3330 if (i_size_read(dst) > datal) {
3332 * If the destination inode has an inline extent...
3333 * This would require copying the data from the source inline
3334 * extent into the beginning of the destination's inline extent.
3335 * But this is really complex, both extents can be compressed
3336 * or just one of them, which would require decompressing and
3337 * re-compressing data (which could increase the new compressed
3338 * size, not allowing the compressed data to fit anymore in an
3340 * So just don't support this case for now (it should be rare,
3341 * we are not really saving space when cloning inline extents).
3346 btrfs_release_path(path);
3347 ret = btrfs_drop_extents(trans, root, dst, drop_start, aligned_end, 1);
3350 ret = btrfs_insert_empty_item(trans, root, path, new_key, size);
3355 const u32 start = btrfs_file_extent_calc_inline_size(0);
3357 memmove(inline_data + start, inline_data + start + skip, datal);
3360 write_extent_buffer(path->nodes[0], inline_data,
3361 btrfs_item_ptr_offset(path->nodes[0],
3364 inode_add_bytes(dst, datal);
3370 * btrfs_clone() - clone a range from inode file to another
3372 * @src: Inode to clone from
3373 * @inode: Inode to clone to
3374 * @off: Offset within source to start clone from
3375 * @olen: Original length, passed by user, of range to clone
3376 * @olen_aligned: Block-aligned value of olen
3377 * @destoff: Offset within @inode to start clone
3378 * @no_time_update: Whether to update mtime/ctime on the target inode
3380 static int btrfs_clone(struct inode *src, struct inode *inode,
3381 const u64 off, const u64 olen, const u64 olen_aligned,
3382 const u64 destoff, int no_time_update)
3384 struct btrfs_root *root = BTRFS_I(inode)->root;
3385 struct btrfs_path *path = NULL;
3386 struct extent_buffer *leaf;
3387 struct btrfs_trans_handle *trans;
3389 struct btrfs_key key;
3393 const u64 len = olen_aligned;
3394 u64 last_dest_end = destoff;
3397 buf = vmalloc(root->nodesize);
3401 path = btrfs_alloc_path();
3407 path->reada = READA_FORWARD;
3409 key.objectid = btrfs_ino(src);
3410 key.type = BTRFS_EXTENT_DATA_KEY;
3414 u64 next_key_min_offset = key.offset + 1;
3417 * note the key will change type as we walk through the
3420 path->leave_spinning = 1;
3421 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3426 * First search, if no extent item that starts at offset off was
3427 * found but the previous item is an extent item, it's possible
3428 * it might overlap our target range, therefore process it.
3430 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3431 btrfs_item_key_to_cpu(path->nodes[0], &key,
3432 path->slots[0] - 1);
3433 if (key.type == BTRFS_EXTENT_DATA_KEY)
3437 nritems = btrfs_header_nritems(path->nodes[0]);
3439 if (path->slots[0] >= nritems) {
3440 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
3445 nritems = btrfs_header_nritems(path->nodes[0]);
3447 leaf = path->nodes[0];
3448 slot = path->slots[0];
3450 btrfs_item_key_to_cpu(leaf, &key, slot);
3451 if (key.type > BTRFS_EXTENT_DATA_KEY ||
3452 key.objectid != btrfs_ino(src))
3455 if (key.type == BTRFS_EXTENT_DATA_KEY) {
3456 struct btrfs_file_extent_item *extent;
3459 struct btrfs_key new_key;
3460 u64 disko = 0, diskl = 0;
3461 u64 datao = 0, datal = 0;
3465 extent = btrfs_item_ptr(leaf, slot,
3466 struct btrfs_file_extent_item);
3467 comp = btrfs_file_extent_compression(leaf, extent);
3468 type = btrfs_file_extent_type(leaf, extent);
3469 if (type == BTRFS_FILE_EXTENT_REG ||
3470 type == BTRFS_FILE_EXTENT_PREALLOC) {
3471 disko = btrfs_file_extent_disk_bytenr(leaf,
3473 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3475 datao = btrfs_file_extent_offset(leaf, extent);
3476 datal = btrfs_file_extent_num_bytes(leaf,
3478 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3479 /* take upper bound, may be compressed */
3480 datal = btrfs_file_extent_ram_bytes(leaf,
3485 * The first search might have left us at an extent
3486 * item that ends before our target range's start, can
3487 * happen if we have holes and NO_HOLES feature enabled.
3489 if (key.offset + datal <= off) {
3492 } else if (key.offset >= off + len) {
3495 next_key_min_offset = key.offset + datal;
3496 size = btrfs_item_size_nr(leaf, slot);
3497 read_extent_buffer(leaf, buf,
3498 btrfs_item_ptr_offset(leaf, slot),
3501 btrfs_release_path(path);
3502 path->leave_spinning = 0;
3504 memcpy(&new_key, &key, sizeof(new_key));
3505 new_key.objectid = btrfs_ino(inode);
3506 if (off <= key.offset)
3507 new_key.offset = key.offset + destoff - off;
3509 new_key.offset = destoff;
3512 * Deal with a hole that doesn't have an extent item
3513 * that represents it (NO_HOLES feature enabled).
3514 * This hole is either in the middle of the cloning
3515 * range or at the beginning (fully overlaps it or
3516 * partially overlaps it).
3518 if (new_key.offset != last_dest_end)
3519 drop_start = last_dest_end;
3521 drop_start = new_key.offset;
3524 * 1 - adjusting old extent (we may have to split it)
3525 * 1 - add new extent
3528 trans = btrfs_start_transaction(root, 3);
3529 if (IS_ERR(trans)) {
3530 ret = PTR_ERR(trans);
3534 if (type == BTRFS_FILE_EXTENT_REG ||
3535 type == BTRFS_FILE_EXTENT_PREALLOC) {
3537 * a | --- range to clone ---| b
3538 * | ------------- extent ------------- |
3541 /* subtract range b */
3542 if (key.offset + datal > off + len)
3543 datal = off + len - key.offset;
3545 /* subtract range a */
3546 if (off > key.offset) {
3547 datao += off - key.offset;
3548 datal -= off - key.offset;
3551 ret = btrfs_drop_extents(trans, root, inode,
3553 new_key.offset + datal,
3556 if (ret != -EOPNOTSUPP)
3557 btrfs_abort_transaction(trans,
3559 btrfs_end_transaction(trans, root);
3563 ret = btrfs_insert_empty_item(trans, root, path,
3566 btrfs_abort_transaction(trans, root,
3568 btrfs_end_transaction(trans, root);
3572 leaf = path->nodes[0];
3573 slot = path->slots[0];
3574 write_extent_buffer(leaf, buf,
3575 btrfs_item_ptr_offset(leaf, slot),
3578 extent = btrfs_item_ptr(leaf, slot,
3579 struct btrfs_file_extent_item);
3581 /* disko == 0 means it's a hole */
3585 btrfs_set_file_extent_offset(leaf, extent,
3587 btrfs_set_file_extent_num_bytes(leaf, extent,
3591 inode_add_bytes(inode, datal);
3592 ret = btrfs_inc_extent_ref(trans, root,
3594 root->root_key.objectid,
3596 new_key.offset - datao);
3598 btrfs_abort_transaction(trans,
3601 btrfs_end_transaction(trans,
3607 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3611 if (off > key.offset) {
3612 skip = off - key.offset;
3613 new_key.offset += skip;
3616 if (key.offset + datal > off + len)
3617 trim = key.offset + datal - (off + len);
3619 if (comp && (skip || trim)) {
3621 btrfs_end_transaction(trans, root);
3624 size -= skip + trim;
3625 datal -= skip + trim;
3627 ret = clone_copy_inline_extent(src, inode,
3634 if (ret != -EOPNOTSUPP)
3635 btrfs_abort_transaction(trans,
3638 btrfs_end_transaction(trans, root);
3641 leaf = path->nodes[0];
3642 slot = path->slots[0];
3645 /* If we have an implicit hole (NO_HOLES feature). */
3646 if (drop_start < new_key.offset)
3647 clone_update_extent_map(inode, trans,
3649 new_key.offset - drop_start);
3651 clone_update_extent_map(inode, trans, path, 0, 0);
3653 btrfs_mark_buffer_dirty(leaf);
3654 btrfs_release_path(path);
3656 last_dest_end = ALIGN(new_key.offset + datal,
3658 ret = clone_finish_inode_update(trans, inode,
3664 if (new_key.offset + datal >= destoff + len)
3667 btrfs_release_path(path);
3668 key.offset = next_key_min_offset;
3672 if (last_dest_end < destoff + len) {
3674 * We have an implicit hole (NO_HOLES feature is enabled) that
3675 * fully or partially overlaps our cloning range at its end.
3677 btrfs_release_path(path);
3680 * 1 - remove extent(s)
3683 trans = btrfs_start_transaction(root, 2);
3684 if (IS_ERR(trans)) {
3685 ret = PTR_ERR(trans);
3688 ret = btrfs_drop_extents(trans, root, inode,
3689 last_dest_end, destoff + len, 1);
3691 if (ret != -EOPNOTSUPP)
3692 btrfs_abort_transaction(trans, root, ret);
3693 btrfs_end_transaction(trans, root);
3696 clone_update_extent_map(inode, trans, NULL, last_dest_end,
3697 destoff + len - last_dest_end);
3698 ret = clone_finish_inode_update(trans, inode, destoff + len,
3699 destoff, olen, no_time_update);
3703 btrfs_free_path(path);
3708 static noinline int btrfs_clone_files(struct file *file, struct file *file_src,
3709 u64 off, u64 olen, u64 destoff)
3711 struct inode *inode = file_inode(file);
3712 struct inode *src = file_inode(file_src);
3713 struct btrfs_root *root = BTRFS_I(inode)->root;
3716 u64 bs = root->fs_info->sb->s_blocksize;
3717 int same_inode = src == inode;
3721 * - split compressed inline extents. annoying: we need to
3722 * decompress into destination's address_space (the file offset
3723 * may change, so source mapping won't do), then recompress (or
3724 * otherwise reinsert) a subrange.
3726 * - split destination inode's inline extents. The inline extents can
3727 * be either compressed or non-compressed.
3730 if (btrfs_root_readonly(root))
3733 if (file_src->f_path.mnt != file->f_path.mnt ||
3734 src->i_sb != inode->i_sb)
3737 /* don't make the dst file partly checksummed */
3738 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3739 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3742 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3746 btrfs_double_inode_lock(src, inode);
3748 mutex_lock(&src->i_mutex);
3751 /* determine range to clone */
3753 if (off + len > src->i_size || off + len < off)
3756 olen = len = src->i_size - off;
3757 /* if we extend to eof, continue to block boundary */
3758 if (off + len == src->i_size)
3759 len = ALIGN(src->i_size, bs) - off;
3766 /* verify the end result is block aligned */
3767 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3768 !IS_ALIGNED(destoff, bs))
3771 /* verify if ranges are overlapped within the same file */
3773 if (destoff + len > off && destoff < off + len)
3777 if (destoff > inode->i_size) {
3778 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3784 * Lock the target range too. Right after we replace the file extent
3785 * items in the fs tree (which now point to the cloned data), we might
3786 * have a worker replace them with extent items relative to a write
3787 * operation that was issued before this clone operation (i.e. confront
3788 * with inode.c:btrfs_finish_ordered_io).
3791 u64 lock_start = min_t(u64, off, destoff);
3792 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
3794 lock_extent_range(src, lock_start, lock_len);
3796 btrfs_double_extent_lock(src, off, inode, destoff, len);
3799 ret = btrfs_clone(src, inode, off, olen, len, destoff, 0);
3802 u64 lock_start = min_t(u64, off, destoff);
3803 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3805 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3807 btrfs_double_extent_unlock(src, off, inode, destoff, len);
3810 * Truncate page cache pages so that future reads will see the cloned
3811 * data immediately and not the previous data.
3813 truncate_inode_pages_range(&inode->i_data, destoff,
3814 PAGE_CACHE_ALIGN(destoff + len) - 1);
3817 btrfs_double_inode_unlock(src, inode);
3819 mutex_unlock(&src->i_mutex);
3823 ssize_t btrfs_copy_file_range(struct file *file_in, loff_t pos_in,
3824 struct file *file_out, loff_t pos_out,
3825 size_t len, unsigned int flags)
3829 ret = btrfs_clone_files(file_out, file_in, pos_in, len, pos_out);
3835 int btrfs_clone_file_range(struct file *src_file, loff_t off,
3836 struct file *dst_file, loff_t destoff, u64 len)
3838 return btrfs_clone_files(dst_file, src_file, off, len, destoff);
3842 * there are many ways the trans_start and trans_end ioctls can lead
3843 * to deadlocks. They should only be used by applications that
3844 * basically own the machine, and have a very in depth understanding
3845 * of all the possible deadlocks and enospc problems.
3847 static long btrfs_ioctl_trans_start(struct file *file)
3849 struct inode *inode = file_inode(file);
3850 struct btrfs_root *root = BTRFS_I(inode)->root;
3851 struct btrfs_trans_handle *trans;
3855 if (!capable(CAP_SYS_ADMIN))
3859 if (file->private_data)
3863 if (btrfs_root_readonly(root))
3866 ret = mnt_want_write_file(file);
3870 atomic_inc(&root->fs_info->open_ioctl_trans);
3873 trans = btrfs_start_ioctl_transaction(root);
3877 file->private_data = trans;
3881 atomic_dec(&root->fs_info->open_ioctl_trans);
3882 mnt_drop_write_file(file);
3887 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3889 struct inode *inode = file_inode(file);
3890 struct btrfs_root *root = BTRFS_I(inode)->root;
3891 struct btrfs_root *new_root;
3892 struct btrfs_dir_item *di;
3893 struct btrfs_trans_handle *trans;
3894 struct btrfs_path *path;
3895 struct btrfs_key location;
3896 struct btrfs_disk_key disk_key;
3901 if (!capable(CAP_SYS_ADMIN))
3904 ret = mnt_want_write_file(file);
3908 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3914 objectid = BTRFS_FS_TREE_OBJECTID;
3916 location.objectid = objectid;
3917 location.type = BTRFS_ROOT_ITEM_KEY;
3918 location.offset = (u64)-1;
3920 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
3921 if (IS_ERR(new_root)) {
3922 ret = PTR_ERR(new_root);
3926 path = btrfs_alloc_path();
3931 path->leave_spinning = 1;
3933 trans = btrfs_start_transaction(root, 1);
3934 if (IS_ERR(trans)) {
3935 btrfs_free_path(path);
3936 ret = PTR_ERR(trans);
3940 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
3941 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
3942 dir_id, "default", 7, 1);
3943 if (IS_ERR_OR_NULL(di)) {
3944 btrfs_free_path(path);
3945 btrfs_end_transaction(trans, root);
3946 btrfs_err(new_root->fs_info, "Umm, you don't have the default dir"
3947 "item, this isn't going to work");
3952 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3953 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3954 btrfs_mark_buffer_dirty(path->nodes[0]);
3955 btrfs_free_path(path);
3957 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
3958 btrfs_end_transaction(trans, root);
3960 mnt_drop_write_file(file);
3964 void btrfs_get_block_group_info(struct list_head *groups_list,
3965 struct btrfs_ioctl_space_info *space)
3967 struct btrfs_block_group_cache *block_group;
3969 space->total_bytes = 0;
3970 space->used_bytes = 0;
3972 list_for_each_entry(block_group, groups_list, list) {
3973 space->flags = block_group->flags;
3974 space->total_bytes += block_group->key.offset;
3975 space->used_bytes +=
3976 btrfs_block_group_used(&block_group->item);
3980 static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
3982 struct btrfs_ioctl_space_args space_args;
3983 struct btrfs_ioctl_space_info space;
3984 struct btrfs_ioctl_space_info *dest;
3985 struct btrfs_ioctl_space_info *dest_orig;
3986 struct btrfs_ioctl_space_info __user *user_dest;
3987 struct btrfs_space_info *info;
3988 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3989 BTRFS_BLOCK_GROUP_SYSTEM,
3990 BTRFS_BLOCK_GROUP_METADATA,
3991 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3998 if (copy_from_user(&space_args,
3999 (struct btrfs_ioctl_space_args __user *)arg,
4000 sizeof(space_args)))
4003 for (i = 0; i < num_types; i++) {
4004 struct btrfs_space_info *tmp;
4008 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
4010 if (tmp->flags == types[i]) {
4020 down_read(&info->groups_sem);
4021 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4022 if (!list_empty(&info->block_groups[c]))
4025 up_read(&info->groups_sem);
4029 * Global block reserve, exported as a space_info
4033 /* space_slots == 0 means they are asking for a count */
4034 if (space_args.space_slots == 0) {
4035 space_args.total_spaces = slot_count;
4039 slot_count = min_t(u64, space_args.space_slots, slot_count);
4041 alloc_size = sizeof(*dest) * slot_count;
4043 /* we generally have at most 6 or so space infos, one for each raid
4044 * level. So, a whole page should be more than enough for everyone
4046 if (alloc_size > PAGE_CACHE_SIZE)
4049 space_args.total_spaces = 0;
4050 dest = kmalloc(alloc_size, GFP_KERNEL);
4055 /* now we have a buffer to copy into */
4056 for (i = 0; i < num_types; i++) {
4057 struct btrfs_space_info *tmp;
4064 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
4066 if (tmp->flags == types[i]) {
4075 down_read(&info->groups_sem);
4076 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
4077 if (!list_empty(&info->block_groups[c])) {
4078 btrfs_get_block_group_info(
4079 &info->block_groups[c], &space);
4080 memcpy(dest, &space, sizeof(space));
4082 space_args.total_spaces++;
4088 up_read(&info->groups_sem);
4092 * Add global block reserve
4095 struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv;
4097 spin_lock(&block_rsv->lock);
4098 space.total_bytes = block_rsv->size;
4099 space.used_bytes = block_rsv->size - block_rsv->reserved;
4100 spin_unlock(&block_rsv->lock);
4101 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
4102 memcpy(dest, &space, sizeof(space));
4103 space_args.total_spaces++;
4106 user_dest = (struct btrfs_ioctl_space_info __user *)
4107 (arg + sizeof(struct btrfs_ioctl_space_args));
4109 if (copy_to_user(user_dest, dest_orig, alloc_size))
4114 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
4121 * there are many ways the trans_start and trans_end ioctls can lead
4122 * to deadlocks. They should only be used by applications that
4123 * basically own the machine, and have a very in depth understanding
4124 * of all the possible deadlocks and enospc problems.
4126 long btrfs_ioctl_trans_end(struct file *file)
4128 struct inode *inode = file_inode(file);
4129 struct btrfs_root *root = BTRFS_I(inode)->root;
4130 struct btrfs_trans_handle *trans;
4132 trans = file->private_data;
4135 file->private_data = NULL;
4137 btrfs_end_transaction(trans, root);
4139 atomic_dec(&root->fs_info->open_ioctl_trans);
4141 mnt_drop_write_file(file);
4145 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4148 struct btrfs_trans_handle *trans;
4152 trans = btrfs_attach_transaction_barrier(root);
4153 if (IS_ERR(trans)) {
4154 if (PTR_ERR(trans) != -ENOENT)
4155 return PTR_ERR(trans);
4157 /* No running transaction, don't bother */
4158 transid = root->fs_info->last_trans_committed;
4161 transid = trans->transid;
4162 ret = btrfs_commit_transaction_async(trans, root, 0);
4164 btrfs_end_transaction(trans, root);
4169 if (copy_to_user(argp, &transid, sizeof(transid)))
4174 static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
4180 if (copy_from_user(&transid, argp, sizeof(transid)))
4183 transid = 0; /* current trans */
4185 return btrfs_wait_for_commit(root, transid);
4188 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
4190 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4191 struct btrfs_ioctl_scrub_args *sa;
4194 if (!capable(CAP_SYS_ADMIN))
4197 sa = memdup_user(arg, sizeof(*sa));
4201 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4202 ret = mnt_want_write_file(file);
4207 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
4208 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4211 if (copy_to_user(arg, sa, sizeof(*sa)))
4214 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4215 mnt_drop_write_file(file);
4221 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
4223 if (!capable(CAP_SYS_ADMIN))
4226 return btrfs_scrub_cancel(root->fs_info);
4229 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
4232 struct btrfs_ioctl_scrub_args *sa;
4235 if (!capable(CAP_SYS_ADMIN))
4238 sa = memdup_user(arg, sizeof(*sa));
4242 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
4244 if (copy_to_user(arg, sa, sizeof(*sa)))
4251 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
4254 struct btrfs_ioctl_get_dev_stats *sa;
4257 sa = memdup_user(arg, sizeof(*sa));
4261 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4266 ret = btrfs_get_dev_stats(root, sa);
4268 if (copy_to_user(arg, sa, sizeof(*sa)))
4275 static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
4277 struct btrfs_ioctl_dev_replace_args *p;
4280 if (!capable(CAP_SYS_ADMIN))
4283 p = memdup_user(arg, sizeof(*p));
4288 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
4289 if (root->fs_info->sb->s_flags & MS_RDONLY) {
4294 &root->fs_info->mutually_exclusive_operation_running,
4296 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4298 ret = btrfs_dev_replace_start(root, p);
4300 &root->fs_info->mutually_exclusive_operation_running,
4304 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4305 btrfs_dev_replace_status(root->fs_info, p);
4308 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4309 ret = btrfs_dev_replace_cancel(root->fs_info, p);
4316 if (copy_to_user(arg, p, sizeof(*p)))
4323 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4329 struct btrfs_ioctl_ino_path_args *ipa = NULL;
4330 struct inode_fs_paths *ipath = NULL;
4331 struct btrfs_path *path;
4333 if (!capable(CAP_DAC_READ_SEARCH))
4336 path = btrfs_alloc_path();
4342 ipa = memdup_user(arg, sizeof(*ipa));
4349 size = min_t(u32, ipa->size, 4096);
4350 ipath = init_ipath(size, root, path);
4351 if (IS_ERR(ipath)) {
4352 ret = PTR_ERR(ipath);
4357 ret = paths_from_inode(ipa->inum, ipath);
4361 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
4362 rel_ptr = ipath->fspath->val[i] -
4363 (u64)(unsigned long)ipath->fspath->val;
4364 ipath->fspath->val[i] = rel_ptr;
4367 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
4368 (void *)(unsigned long)ipath->fspath, size);
4375 btrfs_free_path(path);
4382 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4384 struct btrfs_data_container *inodes = ctx;
4385 const size_t c = 3 * sizeof(u64);
4387 if (inodes->bytes_left >= c) {
4388 inodes->bytes_left -= c;
4389 inodes->val[inodes->elem_cnt] = inum;
4390 inodes->val[inodes->elem_cnt + 1] = offset;
4391 inodes->val[inodes->elem_cnt + 2] = root;
4392 inodes->elem_cnt += 3;
4394 inodes->bytes_missing += c - inodes->bytes_left;
4395 inodes->bytes_left = 0;
4396 inodes->elem_missed += 3;
4402 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
4407 struct btrfs_ioctl_logical_ino_args *loi;
4408 struct btrfs_data_container *inodes = NULL;
4409 struct btrfs_path *path = NULL;
4411 if (!capable(CAP_SYS_ADMIN))
4414 loi = memdup_user(arg, sizeof(*loi));
4421 path = btrfs_alloc_path();
4427 size = min_t(u32, loi->size, SZ_64K);
4428 inodes = init_data_container(size);
4429 if (IS_ERR(inodes)) {
4430 ret = PTR_ERR(inodes);
4435 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
4436 build_ino_list, inodes);
4442 ret = copy_to_user((void *)(unsigned long)loi->inodes,
4443 (void *)(unsigned long)inodes, size);
4448 btrfs_free_path(path);
4455 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
4456 struct btrfs_ioctl_balance_args *bargs)
4458 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4460 bargs->flags = bctl->flags;
4462 if (atomic_read(&fs_info->balance_running))
4463 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4464 if (atomic_read(&fs_info->balance_pause_req))
4465 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
4466 if (atomic_read(&fs_info->balance_cancel_req))
4467 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
4469 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4470 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4471 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4474 spin_lock(&fs_info->balance_lock);
4475 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4476 spin_unlock(&fs_info->balance_lock);
4478 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4482 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4484 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4485 struct btrfs_fs_info *fs_info = root->fs_info;
4486 struct btrfs_ioctl_balance_args *bargs;
4487 struct btrfs_balance_control *bctl;
4488 bool need_unlock; /* for mut. excl. ops lock */
4491 if (!capable(CAP_SYS_ADMIN))
4494 ret = mnt_want_write_file(file);
4499 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
4500 mutex_lock(&fs_info->volume_mutex);
4501 mutex_lock(&fs_info->balance_mutex);
4507 * mut. excl. ops lock is locked. Three possibilites:
4508 * (1) some other op is running
4509 * (2) balance is running
4510 * (3) balance is paused -- special case (think resume)
4512 mutex_lock(&fs_info->balance_mutex);
4513 if (fs_info->balance_ctl) {
4514 /* this is either (2) or (3) */
4515 if (!atomic_read(&fs_info->balance_running)) {
4516 mutex_unlock(&fs_info->balance_mutex);
4517 if (!mutex_trylock(&fs_info->volume_mutex))
4519 mutex_lock(&fs_info->balance_mutex);
4521 if (fs_info->balance_ctl &&
4522 !atomic_read(&fs_info->balance_running)) {
4524 need_unlock = false;
4528 mutex_unlock(&fs_info->balance_mutex);
4529 mutex_unlock(&fs_info->volume_mutex);
4533 mutex_unlock(&fs_info->balance_mutex);
4539 mutex_unlock(&fs_info->balance_mutex);
4540 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4545 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
4548 bargs = memdup_user(arg, sizeof(*bargs));
4549 if (IS_ERR(bargs)) {
4550 ret = PTR_ERR(bargs);
4554 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4555 if (!fs_info->balance_ctl) {
4560 bctl = fs_info->balance_ctl;
4561 spin_lock(&fs_info->balance_lock);
4562 bctl->flags |= BTRFS_BALANCE_RESUME;
4563 spin_unlock(&fs_info->balance_lock);
4571 if (fs_info->balance_ctl) {
4576 bctl = kzalloc(sizeof(*bctl), GFP_KERNEL);
4582 bctl->fs_info = fs_info;
4584 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4585 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4586 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4588 bctl->flags = bargs->flags;
4590 /* balance everything - no filters */
4591 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4594 if (bctl->flags & ~(BTRFS_BALANCE_ARGS_MASK | BTRFS_BALANCE_TYPE_MASK)) {
4601 * Ownership of bctl and mutually_exclusive_operation_running
4602 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4603 * or, if restriper was paused all the way until unmount, in
4604 * free_fs_info. mutually_exclusive_operation_running is
4605 * cleared in __cancel_balance.
4607 need_unlock = false;
4609 ret = btrfs_balance(bctl, bargs);
4613 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4622 mutex_unlock(&fs_info->balance_mutex);
4623 mutex_unlock(&fs_info->volume_mutex);
4625 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4627 mnt_drop_write_file(file);
4631 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4633 if (!capable(CAP_SYS_ADMIN))
4637 case BTRFS_BALANCE_CTL_PAUSE:
4638 return btrfs_pause_balance(root->fs_info);
4639 case BTRFS_BALANCE_CTL_CANCEL:
4640 return btrfs_cancel_balance(root->fs_info);
4646 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4649 struct btrfs_fs_info *fs_info = root->fs_info;
4650 struct btrfs_ioctl_balance_args *bargs;
4653 if (!capable(CAP_SYS_ADMIN))
4656 mutex_lock(&fs_info->balance_mutex);
4657 if (!fs_info->balance_ctl) {
4662 bargs = kzalloc(sizeof(*bargs), GFP_KERNEL);
4668 update_ioctl_balance_args(fs_info, 1, bargs);
4670 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4675 mutex_unlock(&fs_info->balance_mutex);
4679 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4681 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4682 struct btrfs_ioctl_quota_ctl_args *sa;
4683 struct btrfs_trans_handle *trans = NULL;
4687 if (!capable(CAP_SYS_ADMIN))
4690 ret = mnt_want_write_file(file);
4694 sa = memdup_user(arg, sizeof(*sa));
4700 down_write(&root->fs_info->subvol_sem);
4701 trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4702 if (IS_ERR(trans)) {
4703 ret = PTR_ERR(trans);
4708 case BTRFS_QUOTA_CTL_ENABLE:
4709 ret = btrfs_quota_enable(trans, root->fs_info);
4711 case BTRFS_QUOTA_CTL_DISABLE:
4712 ret = btrfs_quota_disable(trans, root->fs_info);
4719 err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4724 up_write(&root->fs_info->subvol_sem);
4726 mnt_drop_write_file(file);
4730 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4732 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4733 struct btrfs_ioctl_qgroup_assign_args *sa;
4734 struct btrfs_trans_handle *trans;
4738 if (!capable(CAP_SYS_ADMIN))
4741 ret = mnt_want_write_file(file);
4745 sa = memdup_user(arg, sizeof(*sa));
4751 trans = btrfs_join_transaction(root);
4752 if (IS_ERR(trans)) {
4753 ret = PTR_ERR(trans);
4757 /* FIXME: check if the IDs really exist */
4759 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4762 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4766 /* update qgroup status and info */
4767 err = btrfs_run_qgroups(trans, root->fs_info);
4769 btrfs_std_error(root->fs_info, ret,
4770 "failed to update qgroup status and info\n");
4771 err = btrfs_end_transaction(trans, root);
4778 mnt_drop_write_file(file);
4782 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4784 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4785 struct btrfs_ioctl_qgroup_create_args *sa;
4786 struct btrfs_trans_handle *trans;
4790 if (!capable(CAP_SYS_ADMIN))
4793 ret = mnt_want_write_file(file);
4797 sa = memdup_user(arg, sizeof(*sa));
4803 if (!sa->qgroupid) {
4808 trans = btrfs_join_transaction(root);
4809 if (IS_ERR(trans)) {
4810 ret = PTR_ERR(trans);
4814 /* FIXME: check if the IDs really exist */
4816 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid);
4818 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4821 err = btrfs_end_transaction(trans, root);
4828 mnt_drop_write_file(file);
4832 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4834 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4835 struct btrfs_ioctl_qgroup_limit_args *sa;
4836 struct btrfs_trans_handle *trans;
4841 if (!capable(CAP_SYS_ADMIN))
4844 ret = mnt_want_write_file(file);
4848 sa = memdup_user(arg, sizeof(*sa));
4854 trans = btrfs_join_transaction(root);
4855 if (IS_ERR(trans)) {
4856 ret = PTR_ERR(trans);
4860 qgroupid = sa->qgroupid;
4862 /* take the current subvol as qgroup */
4863 qgroupid = root->root_key.objectid;
4866 /* FIXME: check if the IDs really exist */
4867 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
4869 err = btrfs_end_transaction(trans, root);
4876 mnt_drop_write_file(file);
4880 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4882 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4883 struct btrfs_ioctl_quota_rescan_args *qsa;
4886 if (!capable(CAP_SYS_ADMIN))
4889 ret = mnt_want_write_file(file);
4893 qsa = memdup_user(arg, sizeof(*qsa));
4904 ret = btrfs_qgroup_rescan(root->fs_info);
4909 mnt_drop_write_file(file);
4913 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
4915 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4916 struct btrfs_ioctl_quota_rescan_args *qsa;
4919 if (!capable(CAP_SYS_ADMIN))
4922 qsa = kzalloc(sizeof(*qsa), GFP_KERNEL);
4926 if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4928 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
4931 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4938 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
4940 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4942 if (!capable(CAP_SYS_ADMIN))
4945 return btrfs_qgroup_wait_for_completion(root->fs_info);
4948 static long _btrfs_ioctl_set_received_subvol(struct file *file,
4949 struct btrfs_ioctl_received_subvol_args *sa)
4951 struct inode *inode = file_inode(file);
4952 struct btrfs_root *root = BTRFS_I(inode)->root;
4953 struct btrfs_root_item *root_item = &root->root_item;
4954 struct btrfs_trans_handle *trans;
4955 struct timespec ct = CURRENT_TIME;
4957 int received_uuid_changed;
4959 if (!inode_owner_or_capable(inode))
4962 ret = mnt_want_write_file(file);
4966 down_write(&root->fs_info->subvol_sem);
4968 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
4973 if (btrfs_root_readonly(root)) {
4980 * 2 - uuid items (received uuid + subvol uuid)
4982 trans = btrfs_start_transaction(root, 3);
4983 if (IS_ERR(trans)) {
4984 ret = PTR_ERR(trans);
4989 sa->rtransid = trans->transid;
4990 sa->rtime.sec = ct.tv_sec;
4991 sa->rtime.nsec = ct.tv_nsec;
4993 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4995 if (received_uuid_changed &&
4996 !btrfs_is_empty_uuid(root_item->received_uuid))
4997 btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
4998 root_item->received_uuid,
4999 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5000 root->root_key.objectid);
5001 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
5002 btrfs_set_root_stransid(root_item, sa->stransid);
5003 btrfs_set_root_rtransid(root_item, sa->rtransid);
5004 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
5005 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
5006 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
5007 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
5009 ret = btrfs_update_root(trans, root->fs_info->tree_root,
5010 &root->root_key, &root->root_item);
5012 btrfs_end_transaction(trans, root);
5015 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
5016 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
5018 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
5019 root->root_key.objectid);
5020 if (ret < 0 && ret != -EEXIST) {
5021 btrfs_abort_transaction(trans, root, ret);
5025 ret = btrfs_commit_transaction(trans, root);
5027 btrfs_abort_transaction(trans, root, ret);
5032 up_write(&root->fs_info->subvol_sem);
5033 mnt_drop_write_file(file);
5038 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
5041 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
5042 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
5045 args32 = memdup_user(arg, sizeof(*args32));
5046 if (IS_ERR(args32)) {
5047 ret = PTR_ERR(args32);
5052 args64 = kmalloc(sizeof(*args64), GFP_KERNEL);
5058 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
5059 args64->stransid = args32->stransid;
5060 args64->rtransid = args32->rtransid;
5061 args64->stime.sec = args32->stime.sec;
5062 args64->stime.nsec = args32->stime.nsec;
5063 args64->rtime.sec = args32->rtime.sec;
5064 args64->rtime.nsec = args32->rtime.nsec;
5065 args64->flags = args32->flags;
5067 ret = _btrfs_ioctl_set_received_subvol(file, args64);
5071 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
5072 args32->stransid = args64->stransid;
5073 args32->rtransid = args64->rtransid;
5074 args32->stime.sec = args64->stime.sec;
5075 args32->stime.nsec = args64->stime.nsec;
5076 args32->rtime.sec = args64->rtime.sec;
5077 args32->rtime.nsec = args64->rtime.nsec;
5078 args32->flags = args64->flags;
5080 ret = copy_to_user(arg, args32, sizeof(*args32));
5091 static long btrfs_ioctl_set_received_subvol(struct file *file,
5094 struct btrfs_ioctl_received_subvol_args *sa = NULL;
5097 sa = memdup_user(arg, sizeof(*sa));
5104 ret = _btrfs_ioctl_set_received_subvol(file, sa);
5109 ret = copy_to_user(arg, sa, sizeof(*sa));
5118 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
5120 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5123 char label[BTRFS_LABEL_SIZE];
5125 spin_lock(&root->fs_info->super_lock);
5126 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
5127 spin_unlock(&root->fs_info->super_lock);
5129 len = strnlen(label, BTRFS_LABEL_SIZE);
5131 if (len == BTRFS_LABEL_SIZE) {
5132 btrfs_warn(root->fs_info,
5133 "label is too long, return the first %zu bytes", --len);
5136 ret = copy_to_user(arg, label, len);
5138 return ret ? -EFAULT : 0;
5141 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5143 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5144 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5145 struct btrfs_trans_handle *trans;
5146 char label[BTRFS_LABEL_SIZE];
5149 if (!capable(CAP_SYS_ADMIN))
5152 if (copy_from_user(label, arg, sizeof(label)))
5155 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
5156 btrfs_err(root->fs_info, "unable to set label with more than %d bytes",
5157 BTRFS_LABEL_SIZE - 1);
5161 ret = mnt_want_write_file(file);
5165 trans = btrfs_start_transaction(root, 0);
5166 if (IS_ERR(trans)) {
5167 ret = PTR_ERR(trans);
5171 spin_lock(&root->fs_info->super_lock);
5172 strcpy(super_block->label, label);
5173 spin_unlock(&root->fs_info->super_lock);
5174 ret = btrfs_commit_transaction(trans, root);
5177 mnt_drop_write_file(file);
5181 #define INIT_FEATURE_FLAGS(suffix) \
5182 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5183 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5184 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5186 static int btrfs_ioctl_get_supported_features(struct file *file,
5189 static const struct btrfs_ioctl_feature_flags features[3] = {
5190 INIT_FEATURE_FLAGS(SUPP),
5191 INIT_FEATURE_FLAGS(SAFE_SET),
5192 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5195 if (copy_to_user(arg, &features, sizeof(features)))
5201 static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5203 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5204 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5205 struct btrfs_ioctl_feature_flags features;
5207 features.compat_flags = btrfs_super_compat_flags(super_block);
5208 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5209 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5211 if (copy_to_user(arg, &features, sizeof(features)))
5217 static int check_feature_bits(struct btrfs_root *root,
5218 enum btrfs_feature_set set,
5219 u64 change_mask, u64 flags, u64 supported_flags,
5220 u64 safe_set, u64 safe_clear)
5222 const char *type = btrfs_feature_set_names[set];
5224 u64 disallowed, unsupported;
5225 u64 set_mask = flags & change_mask;
5226 u64 clear_mask = ~flags & change_mask;
5228 unsupported = set_mask & ~supported_flags;
5230 names = btrfs_printable_features(set, unsupported);
5232 btrfs_warn(root->fs_info,
5233 "this kernel does not support the %s feature bit%s",
5234 names, strchr(names, ',') ? "s" : "");
5237 btrfs_warn(root->fs_info,
5238 "this kernel does not support %s bits 0x%llx",
5243 disallowed = set_mask & ~safe_set;
5245 names = btrfs_printable_features(set, disallowed);
5247 btrfs_warn(root->fs_info,
5248 "can't set the %s feature bit%s while mounted",
5249 names, strchr(names, ',') ? "s" : "");
5252 btrfs_warn(root->fs_info,
5253 "can't set %s bits 0x%llx while mounted",
5258 disallowed = clear_mask & ~safe_clear;
5260 names = btrfs_printable_features(set, disallowed);
5262 btrfs_warn(root->fs_info,
5263 "can't clear the %s feature bit%s while mounted",
5264 names, strchr(names, ',') ? "s" : "");
5267 btrfs_warn(root->fs_info,
5268 "can't clear %s bits 0x%llx while mounted",
5276 #define check_feature(root, change_mask, flags, mask_base) \
5277 check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
5278 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5279 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5280 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5282 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5284 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5285 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5286 struct btrfs_ioctl_feature_flags flags[2];
5287 struct btrfs_trans_handle *trans;
5291 if (!capable(CAP_SYS_ADMIN))
5294 if (copy_from_user(flags, arg, sizeof(flags)))
5298 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5299 !flags[0].incompat_flags)
5302 ret = check_feature(root, flags[0].compat_flags,
5303 flags[1].compat_flags, COMPAT);
5307 ret = check_feature(root, flags[0].compat_ro_flags,
5308 flags[1].compat_ro_flags, COMPAT_RO);
5312 ret = check_feature(root, flags[0].incompat_flags,
5313 flags[1].incompat_flags, INCOMPAT);
5317 trans = btrfs_start_transaction(root, 0);
5319 return PTR_ERR(trans);
5321 spin_lock(&root->fs_info->super_lock);
5322 newflags = btrfs_super_compat_flags(super_block);
5323 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5324 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5325 btrfs_set_super_compat_flags(super_block, newflags);
5327 newflags = btrfs_super_compat_ro_flags(super_block);
5328 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5329 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5330 btrfs_set_super_compat_ro_flags(super_block, newflags);
5332 newflags = btrfs_super_incompat_flags(super_block);
5333 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5334 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5335 btrfs_set_super_incompat_flags(super_block, newflags);
5336 spin_unlock(&root->fs_info->super_lock);
5338 return btrfs_commit_transaction(trans, root);
5341 long btrfs_ioctl(struct file *file, unsigned int
5342 cmd, unsigned long arg)
5344 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5345 void __user *argp = (void __user *)arg;
5348 case FS_IOC_GETFLAGS:
5349 return btrfs_ioctl_getflags(file, argp);
5350 case FS_IOC_SETFLAGS:
5351 return btrfs_ioctl_setflags(file, argp);
5352 case FS_IOC_GETVERSION:
5353 return btrfs_ioctl_getversion(file, argp);
5355 return btrfs_ioctl_fitrim(file, argp);
5356 case BTRFS_IOC_SNAP_CREATE:
5357 return btrfs_ioctl_snap_create(file, argp, 0);
5358 case BTRFS_IOC_SNAP_CREATE_V2:
5359 return btrfs_ioctl_snap_create_v2(file, argp, 0);
5360 case BTRFS_IOC_SUBVOL_CREATE:
5361 return btrfs_ioctl_snap_create(file, argp, 1);
5362 case BTRFS_IOC_SUBVOL_CREATE_V2:
5363 return btrfs_ioctl_snap_create_v2(file, argp, 1);
5364 case BTRFS_IOC_SNAP_DESTROY:
5365 return btrfs_ioctl_snap_destroy(file, argp);
5366 case BTRFS_IOC_SUBVOL_GETFLAGS:
5367 return btrfs_ioctl_subvol_getflags(file, argp);
5368 case BTRFS_IOC_SUBVOL_SETFLAGS:
5369 return btrfs_ioctl_subvol_setflags(file, argp);
5370 case BTRFS_IOC_DEFAULT_SUBVOL:
5371 return btrfs_ioctl_default_subvol(file, argp);
5372 case BTRFS_IOC_DEFRAG:
5373 return btrfs_ioctl_defrag(file, NULL);
5374 case BTRFS_IOC_DEFRAG_RANGE:
5375 return btrfs_ioctl_defrag(file, argp);
5376 case BTRFS_IOC_RESIZE:
5377 return btrfs_ioctl_resize(file, argp);
5378 case BTRFS_IOC_ADD_DEV:
5379 return btrfs_ioctl_add_dev(root, argp);
5380 case BTRFS_IOC_RM_DEV:
5381 return btrfs_ioctl_rm_dev(file, argp);
5382 case BTRFS_IOC_FS_INFO:
5383 return btrfs_ioctl_fs_info(root, argp);
5384 case BTRFS_IOC_DEV_INFO:
5385 return btrfs_ioctl_dev_info(root, argp);
5386 case BTRFS_IOC_BALANCE:
5387 return btrfs_ioctl_balance(file, NULL);
5388 case BTRFS_IOC_TRANS_START:
5389 return btrfs_ioctl_trans_start(file);
5390 case BTRFS_IOC_TRANS_END:
5391 return btrfs_ioctl_trans_end(file);
5392 case BTRFS_IOC_TREE_SEARCH:
5393 return btrfs_ioctl_tree_search(file, argp);
5394 case BTRFS_IOC_TREE_SEARCH_V2:
5395 return btrfs_ioctl_tree_search_v2(file, argp);
5396 case BTRFS_IOC_INO_LOOKUP:
5397 return btrfs_ioctl_ino_lookup(file, argp);
5398 case BTRFS_IOC_INO_PATHS:
5399 return btrfs_ioctl_ino_to_path(root, argp);
5400 case BTRFS_IOC_LOGICAL_INO:
5401 return btrfs_ioctl_logical_to_ino(root, argp);
5402 case BTRFS_IOC_SPACE_INFO:
5403 return btrfs_ioctl_space_info(root, argp);
5404 case BTRFS_IOC_SYNC: {
5407 ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1);
5410 ret = btrfs_sync_fs(file_inode(file)->i_sb, 1);
5412 * The transaction thread may want to do more work,
5413 * namely it pokes the cleaner ktread that will start
5414 * processing uncleaned subvols.
5416 wake_up_process(root->fs_info->transaction_kthread);
5419 case BTRFS_IOC_START_SYNC:
5420 return btrfs_ioctl_start_sync(root, argp);
5421 case BTRFS_IOC_WAIT_SYNC:
5422 return btrfs_ioctl_wait_sync(root, argp);
5423 case BTRFS_IOC_SCRUB:
5424 return btrfs_ioctl_scrub(file, argp);
5425 case BTRFS_IOC_SCRUB_CANCEL:
5426 return btrfs_ioctl_scrub_cancel(root, argp);
5427 case BTRFS_IOC_SCRUB_PROGRESS:
5428 return btrfs_ioctl_scrub_progress(root, argp);
5429 case BTRFS_IOC_BALANCE_V2:
5430 return btrfs_ioctl_balance(file, argp);
5431 case BTRFS_IOC_BALANCE_CTL:
5432 return btrfs_ioctl_balance_ctl(root, arg);
5433 case BTRFS_IOC_BALANCE_PROGRESS:
5434 return btrfs_ioctl_balance_progress(root, argp);
5435 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5436 return btrfs_ioctl_set_received_subvol(file, argp);
5438 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5439 return btrfs_ioctl_set_received_subvol_32(file, argp);
5441 case BTRFS_IOC_SEND:
5442 return btrfs_ioctl_send(file, argp);
5443 case BTRFS_IOC_GET_DEV_STATS:
5444 return btrfs_ioctl_get_dev_stats(root, argp);
5445 case BTRFS_IOC_QUOTA_CTL:
5446 return btrfs_ioctl_quota_ctl(file, argp);
5447 case BTRFS_IOC_QGROUP_ASSIGN:
5448 return btrfs_ioctl_qgroup_assign(file, argp);
5449 case BTRFS_IOC_QGROUP_CREATE:
5450 return btrfs_ioctl_qgroup_create(file, argp);
5451 case BTRFS_IOC_QGROUP_LIMIT:
5452 return btrfs_ioctl_qgroup_limit(file, argp);
5453 case BTRFS_IOC_QUOTA_RESCAN:
5454 return btrfs_ioctl_quota_rescan(file, argp);
5455 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5456 return btrfs_ioctl_quota_rescan_status(file, argp);
5457 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5458 return btrfs_ioctl_quota_rescan_wait(file, argp);
5459 case BTRFS_IOC_DEV_REPLACE:
5460 return btrfs_ioctl_dev_replace(root, argp);
5461 case BTRFS_IOC_GET_FSLABEL:
5462 return btrfs_ioctl_get_fslabel(file, argp);
5463 case BTRFS_IOC_SET_FSLABEL:
5464 return btrfs_ioctl_set_fslabel(file, argp);
5465 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5466 return btrfs_ioctl_get_supported_features(file, argp);
5467 case BTRFS_IOC_GET_FEATURES:
5468 return btrfs_ioctl_get_features(file, argp);
5469 case BTRFS_IOC_SET_FEATURES:
5470 return btrfs_ioctl_set_features(file, argp);