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);
92 /* Mask out flags that are inappropriate for the given type of inode. */
93 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
97 else if (S_ISREG(mode))
98 return flags & ~FS_DIRSYNC_FL;
100 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
104 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
106 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
108 unsigned int iflags = 0;
110 if (flags & BTRFS_INODE_SYNC)
111 iflags |= FS_SYNC_FL;
112 if (flags & BTRFS_INODE_IMMUTABLE)
113 iflags |= FS_IMMUTABLE_FL;
114 if (flags & BTRFS_INODE_APPEND)
115 iflags |= FS_APPEND_FL;
116 if (flags & BTRFS_INODE_NODUMP)
117 iflags |= FS_NODUMP_FL;
118 if (flags & BTRFS_INODE_NOATIME)
119 iflags |= FS_NOATIME_FL;
120 if (flags & BTRFS_INODE_DIRSYNC)
121 iflags |= FS_DIRSYNC_FL;
122 if (flags & BTRFS_INODE_NODATACOW)
123 iflags |= FS_NOCOW_FL;
125 if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
126 iflags |= FS_COMPR_FL;
127 else if (flags & BTRFS_INODE_NOCOMPRESS)
128 iflags |= FS_NOCOMP_FL;
134 * Update inode->i_flags based on the btrfs internal flags.
136 void btrfs_update_iflags(struct inode *inode)
138 struct btrfs_inode *ip = BTRFS_I(inode);
139 unsigned int new_fl = 0;
141 if (ip->flags & BTRFS_INODE_SYNC)
143 if (ip->flags & BTRFS_INODE_IMMUTABLE)
144 new_fl |= S_IMMUTABLE;
145 if (ip->flags & BTRFS_INODE_APPEND)
147 if (ip->flags & BTRFS_INODE_NOATIME)
149 if (ip->flags & BTRFS_INODE_DIRSYNC)
152 set_mask_bits(&inode->i_flags,
153 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC,
158 * Inherit flags from the parent inode.
160 * Currently only the compression flags and the cow flags are inherited.
162 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
169 flags = BTRFS_I(dir)->flags;
171 if (flags & BTRFS_INODE_NOCOMPRESS) {
172 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
173 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
174 } else if (flags & BTRFS_INODE_COMPRESS) {
175 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
176 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
179 if (flags & BTRFS_INODE_NODATACOW) {
180 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
181 if (S_ISREG(inode->i_mode))
182 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
185 btrfs_update_iflags(inode);
188 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
190 struct btrfs_inode *ip = BTRFS_I(file_inode(file));
191 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
193 if (copy_to_user(arg, &flags, sizeof(flags)))
198 static int check_flags(unsigned int flags)
200 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
201 FS_NOATIME_FL | FS_NODUMP_FL | \
202 FS_SYNC_FL | FS_DIRSYNC_FL | \
203 FS_NOCOMP_FL | FS_COMPR_FL |
207 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
213 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
215 struct inode *inode = file_inode(file);
216 struct btrfs_inode *ip = BTRFS_I(inode);
217 struct btrfs_root *root = ip->root;
218 struct btrfs_trans_handle *trans;
219 unsigned int flags, oldflags;
222 unsigned int i_oldflags;
225 if (!inode_owner_or_capable(inode))
228 if (btrfs_root_readonly(root))
231 if (copy_from_user(&flags, arg, sizeof(flags)))
234 ret = check_flags(flags);
238 ret = mnt_want_write_file(file);
242 mutex_lock(&inode->i_mutex);
244 ip_oldflags = ip->flags;
245 i_oldflags = inode->i_flags;
246 mode = inode->i_mode;
248 flags = btrfs_mask_flags(inode->i_mode, flags);
249 oldflags = btrfs_flags_to_ioctl(ip->flags);
250 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
251 if (!capable(CAP_LINUX_IMMUTABLE)) {
257 if (flags & FS_SYNC_FL)
258 ip->flags |= BTRFS_INODE_SYNC;
260 ip->flags &= ~BTRFS_INODE_SYNC;
261 if (flags & FS_IMMUTABLE_FL)
262 ip->flags |= BTRFS_INODE_IMMUTABLE;
264 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
265 if (flags & FS_APPEND_FL)
266 ip->flags |= BTRFS_INODE_APPEND;
268 ip->flags &= ~BTRFS_INODE_APPEND;
269 if (flags & FS_NODUMP_FL)
270 ip->flags |= BTRFS_INODE_NODUMP;
272 ip->flags &= ~BTRFS_INODE_NODUMP;
273 if (flags & FS_NOATIME_FL)
274 ip->flags |= BTRFS_INODE_NOATIME;
276 ip->flags &= ~BTRFS_INODE_NOATIME;
277 if (flags & FS_DIRSYNC_FL)
278 ip->flags |= BTRFS_INODE_DIRSYNC;
280 ip->flags &= ~BTRFS_INODE_DIRSYNC;
281 if (flags & FS_NOCOW_FL) {
284 * It's safe to turn csums off here, no extents exist.
285 * Otherwise we want the flag to reflect the real COW
286 * status of the file and will not set it.
288 if (inode->i_size == 0)
289 ip->flags |= BTRFS_INODE_NODATACOW
290 | BTRFS_INODE_NODATASUM;
292 ip->flags |= BTRFS_INODE_NODATACOW;
296 * Revert back under same assuptions as above
299 if (inode->i_size == 0)
300 ip->flags &= ~(BTRFS_INODE_NODATACOW
301 | BTRFS_INODE_NODATASUM);
303 ip->flags &= ~BTRFS_INODE_NODATACOW;
308 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
309 * flag may be changed automatically if compression code won't make
312 if (flags & FS_NOCOMP_FL) {
313 ip->flags &= ~BTRFS_INODE_COMPRESS;
314 ip->flags |= BTRFS_INODE_NOCOMPRESS;
316 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
317 if (ret && ret != -ENODATA)
319 } else if (flags & FS_COMPR_FL) {
322 ip->flags |= BTRFS_INODE_COMPRESS;
323 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
325 if (root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
329 ret = btrfs_set_prop(inode, "btrfs.compression",
330 comp, strlen(comp), 0);
335 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
336 if (ret && ret != -ENODATA)
338 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
341 trans = btrfs_start_transaction(root, 1);
343 ret = PTR_ERR(trans);
347 btrfs_update_iflags(inode);
348 inode_inc_iversion(inode);
349 inode->i_ctime = CURRENT_TIME;
350 ret = btrfs_update_inode(trans, root, inode);
352 btrfs_end_transaction(trans, root);
355 ip->flags = ip_oldflags;
356 inode->i_flags = i_oldflags;
360 mutex_unlock(&inode->i_mutex);
361 mnt_drop_write_file(file);
365 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
367 struct inode *inode = file_inode(file);
369 return put_user(inode->i_generation, arg);
372 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
374 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
375 struct btrfs_device *device;
376 struct request_queue *q;
377 struct fstrim_range range;
378 u64 minlen = ULLONG_MAX;
380 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
383 if (!capable(CAP_SYS_ADMIN))
387 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
391 q = bdev_get_queue(device->bdev);
392 if (blk_queue_discard(q)) {
394 minlen = min((u64)q->limits.discard_granularity,
402 if (copy_from_user(&range, arg, sizeof(range)))
404 if (range.start > total_bytes ||
405 range.len < fs_info->sb->s_blocksize)
408 range.len = min(range.len, total_bytes - range.start);
409 range.minlen = max(range.minlen, minlen);
410 ret = btrfs_trim_fs(fs_info->tree_root, &range);
414 if (copy_to_user(arg, &range, sizeof(range)))
420 int btrfs_is_empty_uuid(u8 *uuid)
424 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
431 static noinline int create_subvol(struct inode *dir,
432 struct dentry *dentry,
433 char *name, int namelen,
435 struct btrfs_qgroup_inherit *inherit)
437 struct btrfs_trans_handle *trans;
438 struct btrfs_key key;
439 struct btrfs_root_item root_item;
440 struct btrfs_inode_item *inode_item;
441 struct extent_buffer *leaf;
442 struct btrfs_root *root = BTRFS_I(dir)->root;
443 struct btrfs_root *new_root;
444 struct btrfs_block_rsv block_rsv;
445 struct timespec cur_time = CURRENT_TIME;
450 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
455 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
459 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
461 * The same as the snapshot creation, please see the comment
462 * of create_snapshot().
464 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
465 8, &qgroup_reserved, false);
469 trans = btrfs_start_transaction(root, 0);
471 ret = PTR_ERR(trans);
472 btrfs_subvolume_release_metadata(root, &block_rsv,
476 trans->block_rsv = &block_rsv;
477 trans->bytes_reserved = block_rsv.size;
479 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
483 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
489 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
490 btrfs_set_header_bytenr(leaf, leaf->start);
491 btrfs_set_header_generation(leaf, trans->transid);
492 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
493 btrfs_set_header_owner(leaf, objectid);
495 write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(),
497 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
498 btrfs_header_chunk_tree_uuid(leaf),
500 btrfs_mark_buffer_dirty(leaf);
502 memset(&root_item, 0, sizeof(root_item));
504 inode_item = &root_item.inode;
505 btrfs_set_stack_inode_generation(inode_item, 1);
506 btrfs_set_stack_inode_size(inode_item, 3);
507 btrfs_set_stack_inode_nlink(inode_item, 1);
508 btrfs_set_stack_inode_nbytes(inode_item, root->nodesize);
509 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
511 btrfs_set_root_flags(&root_item, 0);
512 btrfs_set_root_limit(&root_item, 0);
513 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
515 btrfs_set_root_bytenr(&root_item, leaf->start);
516 btrfs_set_root_generation(&root_item, trans->transid);
517 btrfs_set_root_level(&root_item, 0);
518 btrfs_set_root_refs(&root_item, 1);
519 btrfs_set_root_used(&root_item, leaf->len);
520 btrfs_set_root_last_snapshot(&root_item, 0);
522 btrfs_set_root_generation_v2(&root_item,
523 btrfs_root_generation(&root_item));
524 uuid_le_gen(&new_uuid);
525 memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
526 btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec);
527 btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec);
528 root_item.ctime = root_item.otime;
529 btrfs_set_root_ctransid(&root_item, trans->transid);
530 btrfs_set_root_otransid(&root_item, trans->transid);
532 btrfs_tree_unlock(leaf);
533 free_extent_buffer(leaf);
536 btrfs_set_root_dirid(&root_item, new_dirid);
538 key.objectid = objectid;
540 key.type = BTRFS_ROOT_ITEM_KEY;
541 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
546 key.offset = (u64)-1;
547 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
548 if (IS_ERR(new_root)) {
549 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
550 ret = PTR_ERR(new_root);
554 btrfs_record_root_in_trans(trans, new_root);
556 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
558 /* We potentially lose an unused inode item here */
559 btrfs_abort_transaction(trans, root, ret);
564 * insert the directory item
566 ret = btrfs_set_inode_index(dir, &index);
568 btrfs_abort_transaction(trans, root, ret);
572 ret = btrfs_insert_dir_item(trans, root,
573 name, namelen, dir, &key,
574 BTRFS_FT_DIR, index);
576 btrfs_abort_transaction(trans, root, ret);
580 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
581 ret = btrfs_update_inode(trans, root, dir);
584 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
585 objectid, root->root_key.objectid,
586 btrfs_ino(dir), index, name, namelen);
589 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
590 root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
593 btrfs_abort_transaction(trans, root, ret);
596 trans->block_rsv = NULL;
597 trans->bytes_reserved = 0;
598 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
601 *async_transid = trans->transid;
602 err = btrfs_commit_transaction_async(trans, root, 1);
604 err = btrfs_commit_transaction(trans, root);
606 err = btrfs_commit_transaction(trans, root);
612 inode = btrfs_lookup_dentry(dir, dentry);
614 return PTR_ERR(inode);
615 d_instantiate(dentry, inode);
620 static void btrfs_wait_nocow_write(struct btrfs_root *root)
626 prepare_to_wait(&root->subv_writers->wait, &wait,
627 TASK_UNINTERRUPTIBLE);
629 writers = percpu_counter_sum(&root->subv_writers->counter);
633 finish_wait(&root->subv_writers->wait, &wait);
637 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
638 struct dentry *dentry, char *name, int namelen,
639 u64 *async_transid, bool readonly,
640 struct btrfs_qgroup_inherit *inherit)
643 struct btrfs_pending_snapshot *pending_snapshot;
644 struct btrfs_trans_handle *trans;
647 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
650 atomic_inc(&root->will_be_snapshoted);
651 smp_mb__after_atomic();
652 btrfs_wait_nocow_write(root);
654 ret = btrfs_start_delalloc_inodes(root, 0);
658 btrfs_wait_ordered_extents(root, -1);
660 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
661 if (!pending_snapshot) {
666 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
667 BTRFS_BLOCK_RSV_TEMP);
669 * 1 - parent dir inode
672 * 2 - root ref/backref
673 * 1 - root of snapshot
676 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
677 &pending_snapshot->block_rsv, 8,
678 &pending_snapshot->qgroup_reserved,
683 pending_snapshot->dentry = dentry;
684 pending_snapshot->root = root;
685 pending_snapshot->readonly = readonly;
686 pending_snapshot->dir = dir;
687 pending_snapshot->inherit = inherit;
689 trans = btrfs_start_transaction(root, 0);
691 ret = PTR_ERR(trans);
695 spin_lock(&root->fs_info->trans_lock);
696 list_add(&pending_snapshot->list,
697 &trans->transaction->pending_snapshots);
698 spin_unlock(&root->fs_info->trans_lock);
700 *async_transid = trans->transid;
701 ret = btrfs_commit_transaction_async(trans,
702 root->fs_info->extent_root, 1);
704 ret = btrfs_commit_transaction(trans, root);
706 ret = btrfs_commit_transaction(trans,
707 root->fs_info->extent_root);
712 ret = pending_snapshot->error;
716 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
721 * If orphan cleanup did remove any orphans, it means the tree was
722 * modified and therefore the commit root is not the same as the
723 * current root anymore. This is a problem, because send uses the
724 * commit root and therefore can see inode items that don't exist
725 * in the current root anymore, and for example make calls to
726 * btrfs_iget, which will do tree lookups based on the current root
727 * and not on the commit root. Those lookups will fail, returning a
728 * -ESTALE error, and making send fail with that error. So make sure
729 * a send does not see any orphans we have just removed, and that it
730 * will see the same inodes regardless of whether a transaction
731 * commit happened before it started (meaning that the commit root
732 * will be the same as the current root) or not.
734 if (readonly && pending_snapshot->snap->node !=
735 pending_snapshot->snap->commit_root) {
736 trans = btrfs_join_transaction(pending_snapshot->snap);
737 if (IS_ERR(trans) && PTR_ERR(trans) != -ENOENT) {
738 ret = PTR_ERR(trans);
741 if (!IS_ERR(trans)) {
742 ret = btrfs_commit_transaction(trans,
743 pending_snapshot->snap);
749 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
751 ret = PTR_ERR(inode);
755 d_instantiate(dentry, inode);
758 btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
759 &pending_snapshot->block_rsv,
760 pending_snapshot->qgroup_reserved);
762 kfree(pending_snapshot);
764 atomic_dec(&root->will_be_snapshoted);
768 /* copy of check_sticky in fs/namei.c()
769 * It's inline, so penalty for filesystems that don't use sticky bit is
772 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
774 kuid_t fsuid = current_fsuid();
776 if (!(dir->i_mode & S_ISVTX))
778 if (uid_eq(inode->i_uid, fsuid))
780 if (uid_eq(dir->i_uid, fsuid))
782 return !capable(CAP_FOWNER);
785 /* copy of may_delete in fs/namei.c()
786 * Check whether we can remove a link victim from directory dir, check
787 * whether the type of victim is right.
788 * 1. We can't do it if dir is read-only (done in permission())
789 * 2. We should have write and exec permissions on dir
790 * 3. We can't remove anything from append-only dir
791 * 4. We can't do anything with immutable dir (done in permission())
792 * 5. If the sticky bit on dir is set we should either
793 * a. be owner of dir, or
794 * b. be owner of victim, or
795 * c. have CAP_FOWNER capability
796 * 6. If the victim is append-only or immutable we can't do antyhing with
797 * links pointing to it.
798 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
799 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
800 * 9. We can't remove a root or mountpoint.
801 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
802 * nfs_async_unlink().
805 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
809 if (!victim->d_inode)
812 BUG_ON(victim->d_parent->d_inode != dir);
813 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
815 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
820 if (btrfs_check_sticky(dir, victim->d_inode)||
821 IS_APPEND(victim->d_inode)||
822 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
825 if (!S_ISDIR(victim->d_inode->i_mode))
829 } else if (S_ISDIR(victim->d_inode->i_mode))
833 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
838 /* copy of may_create in fs/namei.c() */
839 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
845 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
849 * Create a new subvolume below @parent. This is largely modeled after
850 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
851 * inside this filesystem so it's quite a bit simpler.
853 static noinline int btrfs_mksubvol(struct path *parent,
854 char *name, int namelen,
855 struct btrfs_root *snap_src,
856 u64 *async_transid, bool readonly,
857 struct btrfs_qgroup_inherit *inherit)
859 struct inode *dir = parent->dentry->d_inode;
860 struct dentry *dentry;
863 error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
867 dentry = lookup_one_len(name, parent->dentry, namelen);
868 error = PTR_ERR(dentry);
876 error = btrfs_may_create(dir, dentry);
881 * even if this name doesn't exist, we may get hash collisions.
882 * check for them now when we can safely fail
884 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
890 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
892 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
896 error = create_snapshot(snap_src, dir, dentry, name, namelen,
897 async_transid, readonly, inherit);
899 error = create_subvol(dir, dentry, name, namelen,
900 async_transid, inherit);
903 fsnotify_mkdir(dir, dentry);
905 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
909 mutex_unlock(&dir->i_mutex);
914 * When we're defragging a range, we don't want to kick it off again
915 * if it is really just waiting for delalloc to send it down.
916 * If we find a nice big extent or delalloc range for the bytes in the
917 * file you want to defrag, we return 0 to let you know to skip this
920 static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
922 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
923 struct extent_map *em = NULL;
924 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
927 read_lock(&em_tree->lock);
928 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
929 read_unlock(&em_tree->lock);
932 end = extent_map_end(em);
934 if (end - offset > thresh)
937 /* if we already have a nice delalloc here, just stop */
939 end = count_range_bits(io_tree, &offset, offset + thresh,
940 thresh, EXTENT_DELALLOC, 1);
947 * helper function to walk through a file and find extents
948 * newer than a specific transid, and smaller than thresh.
950 * This is used by the defragging code to find new and small
953 static int find_new_extents(struct btrfs_root *root,
954 struct inode *inode, u64 newer_than,
955 u64 *off, u32 thresh)
957 struct btrfs_path *path;
958 struct btrfs_key min_key;
959 struct extent_buffer *leaf;
960 struct btrfs_file_extent_item *extent;
963 u64 ino = btrfs_ino(inode);
965 path = btrfs_alloc_path();
969 min_key.objectid = ino;
970 min_key.type = BTRFS_EXTENT_DATA_KEY;
971 min_key.offset = *off;
974 ret = btrfs_search_forward(root, &min_key, path, newer_than);
978 if (min_key.objectid != ino)
980 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
983 leaf = path->nodes[0];
984 extent = btrfs_item_ptr(leaf, path->slots[0],
985 struct btrfs_file_extent_item);
987 type = btrfs_file_extent_type(leaf, extent);
988 if (type == BTRFS_FILE_EXTENT_REG &&
989 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
990 check_defrag_in_cache(inode, min_key.offset, thresh)) {
991 *off = min_key.offset;
992 btrfs_free_path(path);
997 if (path->slots[0] < btrfs_header_nritems(leaf)) {
998 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
1002 if (min_key.offset == (u64)-1)
1006 btrfs_release_path(path);
1009 btrfs_free_path(path);
1013 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
1015 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1016 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1017 struct extent_map *em;
1018 u64 len = PAGE_CACHE_SIZE;
1021 * hopefully we have this extent in the tree already, try without
1022 * the full extent lock
1024 read_lock(&em_tree->lock);
1025 em = lookup_extent_mapping(em_tree, start, len);
1026 read_unlock(&em_tree->lock);
1029 struct extent_state *cached = NULL;
1030 u64 end = start + len - 1;
1032 /* get the big lock and read metadata off disk */
1033 lock_extent_bits(io_tree, start, end, 0, &cached);
1034 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
1035 unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS);
1044 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1046 struct extent_map *next;
1049 /* this is the last extent */
1050 if (em->start + em->len >= i_size_read(inode))
1053 next = defrag_lookup_extent(inode, em->start + em->len);
1054 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1056 else if ((em->block_start + em->block_len == next->block_start) &&
1057 (em->block_len > 128 * 1024 && next->block_len > 128 * 1024))
1060 free_extent_map(next);
1064 static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
1065 u64 *last_len, u64 *skip, u64 *defrag_end,
1068 struct extent_map *em;
1070 bool next_mergeable = true;
1073 * make sure that once we start defragging an extent, we keep on
1076 if (start < *defrag_end)
1081 em = defrag_lookup_extent(inode, start);
1085 /* this will cover holes, and inline extents */
1086 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1091 next_mergeable = defrag_check_next_extent(inode, em);
1093 * we hit a real extent, if it is big or the next extent is not a
1094 * real extent, don't bother defragging it
1096 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1097 (em->len >= thresh || !next_mergeable))
1101 * last_len ends up being a counter of how many bytes we've defragged.
1102 * every time we choose not to defrag an extent, we reset *last_len
1103 * so that the next tiny extent will force a defrag.
1105 * The end result of this is that tiny extents before a single big
1106 * extent will force at least part of that big extent to be defragged.
1109 *defrag_end = extent_map_end(em);
1112 *skip = extent_map_end(em);
1116 free_extent_map(em);
1121 * it doesn't do much good to defrag one or two pages
1122 * at a time. This pulls in a nice chunk of pages
1123 * to COW and defrag.
1125 * It also makes sure the delalloc code has enough
1126 * dirty data to avoid making new small extents as part
1129 * It's a good idea to start RA on this range
1130 * before calling this.
1132 static int cluster_pages_for_defrag(struct inode *inode,
1133 struct page **pages,
1134 unsigned long start_index,
1135 unsigned long num_pages)
1137 unsigned long file_end;
1138 u64 isize = i_size_read(inode);
1145 struct btrfs_ordered_extent *ordered;
1146 struct extent_state *cached_state = NULL;
1147 struct extent_io_tree *tree;
1148 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1150 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1151 if (!isize || start_index > file_end)
1154 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1156 ret = btrfs_delalloc_reserve_space(inode,
1157 page_cnt << PAGE_CACHE_SHIFT);
1161 tree = &BTRFS_I(inode)->io_tree;
1163 /* step one, lock all the pages */
1164 for (i = 0; i < page_cnt; i++) {
1167 page = find_or_create_page(inode->i_mapping,
1168 start_index + i, mask);
1172 page_start = page_offset(page);
1173 page_end = page_start + PAGE_CACHE_SIZE - 1;
1175 lock_extent_bits(tree, page_start, page_end,
1177 ordered = btrfs_lookup_ordered_extent(inode,
1179 unlock_extent_cached(tree, page_start, page_end,
1180 &cached_state, GFP_NOFS);
1185 btrfs_start_ordered_extent(inode, ordered, 1);
1186 btrfs_put_ordered_extent(ordered);
1189 * we unlocked the page above, so we need check if
1190 * it was released or not.
1192 if (page->mapping != inode->i_mapping) {
1194 page_cache_release(page);
1199 if (!PageUptodate(page)) {
1200 btrfs_readpage(NULL, page);
1202 if (!PageUptodate(page)) {
1204 page_cache_release(page);
1210 if (page->mapping != inode->i_mapping) {
1212 page_cache_release(page);
1222 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1226 * so now we have a nice long stream of locked
1227 * and up to date pages, lets wait on them
1229 for (i = 0; i < i_done; i++)
1230 wait_on_page_writeback(pages[i]);
1232 page_start = page_offset(pages[0]);
1233 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1235 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1236 page_start, page_end - 1, 0, &cached_state);
1237 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1238 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1239 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1240 &cached_state, GFP_NOFS);
1242 if (i_done != page_cnt) {
1243 spin_lock(&BTRFS_I(inode)->lock);
1244 BTRFS_I(inode)->outstanding_extents++;
1245 spin_unlock(&BTRFS_I(inode)->lock);
1246 btrfs_delalloc_release_space(inode,
1247 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1251 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1252 &cached_state, GFP_NOFS);
1254 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1255 page_start, page_end - 1, &cached_state,
1258 for (i = 0; i < i_done; i++) {
1259 clear_page_dirty_for_io(pages[i]);
1260 ClearPageChecked(pages[i]);
1261 set_page_extent_mapped(pages[i]);
1262 set_page_dirty(pages[i]);
1263 unlock_page(pages[i]);
1264 page_cache_release(pages[i]);
1268 for (i = 0; i < i_done; i++) {
1269 unlock_page(pages[i]);
1270 page_cache_release(pages[i]);
1272 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1277 int btrfs_defrag_file(struct inode *inode, struct file *file,
1278 struct btrfs_ioctl_defrag_range_args *range,
1279 u64 newer_than, unsigned long max_to_defrag)
1281 struct btrfs_root *root = BTRFS_I(inode)->root;
1282 struct file_ra_state *ra = NULL;
1283 unsigned long last_index;
1284 u64 isize = i_size_read(inode);
1288 u64 newer_off = range->start;
1290 unsigned long ra_index = 0;
1292 int defrag_count = 0;
1293 int compress_type = BTRFS_COMPRESS_ZLIB;
1294 u32 extent_thresh = range->extent_thresh;
1295 unsigned long max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1296 unsigned long cluster = max_cluster;
1297 u64 new_align = ~((u64)128 * 1024 - 1);
1298 struct page **pages = NULL;
1303 if (range->start >= isize)
1306 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1307 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1309 if (range->compress_type)
1310 compress_type = range->compress_type;
1313 if (extent_thresh == 0)
1314 extent_thresh = 256 * 1024;
1317 * if we were not given a file, allocate a readahead
1321 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1324 file_ra_state_init(ra, inode->i_mapping);
1329 pages = kmalloc_array(max_cluster, sizeof(struct page *),
1336 /* find the last page to defrag */
1337 if (range->start + range->len > range->start) {
1338 last_index = min_t(u64, isize - 1,
1339 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1341 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1345 ret = find_new_extents(root, inode, newer_than,
1346 &newer_off, 64 * 1024);
1348 range->start = newer_off;
1350 * we always align our defrag to help keep
1351 * the extents in the file evenly spaced
1353 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1357 i = range->start >> PAGE_CACHE_SHIFT;
1360 max_to_defrag = last_index + 1;
1363 * make writeback starts from i, so the defrag range can be
1364 * written sequentially.
1366 if (i < inode->i_mapping->writeback_index)
1367 inode->i_mapping->writeback_index = i;
1369 while (i <= last_index && defrag_count < max_to_defrag &&
1370 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_CACHE_SIZE))) {
1372 * make sure we stop running if someone unmounts
1375 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1378 if (btrfs_defrag_cancelled(root->fs_info)) {
1379 printk(KERN_DEBUG "BTRFS: defrag_file cancelled\n");
1384 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1385 extent_thresh, &last_len, &skip,
1386 &defrag_end, range->flags &
1387 BTRFS_DEFRAG_RANGE_COMPRESS)) {
1390 * the should_defrag function tells us how much to skip
1391 * bump our counter by the suggested amount
1393 next = DIV_ROUND_UP(skip, PAGE_CACHE_SIZE);
1394 i = max(i + 1, next);
1399 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1400 PAGE_CACHE_SHIFT) - i;
1401 cluster = min(cluster, max_cluster);
1403 cluster = max_cluster;
1406 if (i + cluster > ra_index) {
1407 ra_index = max(i, ra_index);
1408 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1410 ra_index += max_cluster;
1413 mutex_lock(&inode->i_mutex);
1414 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1415 BTRFS_I(inode)->force_compress = compress_type;
1416 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1418 mutex_unlock(&inode->i_mutex);
1422 defrag_count += ret;
1423 balance_dirty_pages_ratelimited(inode->i_mapping);
1424 mutex_unlock(&inode->i_mutex);
1427 if (newer_off == (u64)-1)
1433 newer_off = max(newer_off + 1,
1434 (u64)i << PAGE_CACHE_SHIFT);
1436 ret = find_new_extents(root, inode,
1437 newer_than, &newer_off,
1440 range->start = newer_off;
1441 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1448 last_len += ret << PAGE_CACHE_SHIFT;
1456 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1457 filemap_flush(inode->i_mapping);
1458 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1459 &BTRFS_I(inode)->runtime_flags))
1460 filemap_flush(inode->i_mapping);
1463 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1464 /* the filemap_flush will queue IO into the worker threads, but
1465 * we have to make sure the IO is actually started and that
1466 * ordered extents get created before we return
1468 atomic_inc(&root->fs_info->async_submit_draining);
1469 while (atomic_read(&root->fs_info->nr_async_submits) ||
1470 atomic_read(&root->fs_info->async_delalloc_pages)) {
1471 wait_event(root->fs_info->async_submit_wait,
1472 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1473 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1475 atomic_dec(&root->fs_info->async_submit_draining);
1478 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1479 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1485 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1486 mutex_lock(&inode->i_mutex);
1487 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1488 mutex_unlock(&inode->i_mutex);
1496 static noinline int btrfs_ioctl_resize(struct file *file,
1502 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
1503 struct btrfs_ioctl_vol_args *vol_args;
1504 struct btrfs_trans_handle *trans;
1505 struct btrfs_device *device = NULL;
1508 char *devstr = NULL;
1512 if (!capable(CAP_SYS_ADMIN))
1515 ret = mnt_want_write_file(file);
1519 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1521 mnt_drop_write_file(file);
1522 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1525 mutex_lock(&root->fs_info->volume_mutex);
1526 vol_args = memdup_user(arg, sizeof(*vol_args));
1527 if (IS_ERR(vol_args)) {
1528 ret = PTR_ERR(vol_args);
1532 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1534 sizestr = vol_args->name;
1535 devstr = strchr(sizestr, ':');
1537 sizestr = devstr + 1;
1539 devstr = vol_args->name;
1540 ret = kstrtoull(devstr, 10, &devid);
1547 btrfs_info(root->fs_info, "resizing devid %llu", devid);
1550 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
1552 btrfs_info(root->fs_info, "resizer unable to find device %llu",
1558 if (!device->writeable) {
1559 btrfs_info(root->fs_info,
1560 "resizer unable to apply on readonly device %llu",
1566 if (!strcmp(sizestr, "max"))
1567 new_size = device->bdev->bd_inode->i_size;
1569 if (sizestr[0] == '-') {
1572 } else if (sizestr[0] == '+') {
1576 new_size = memparse(sizestr, &retptr);
1577 if (*retptr != '\0' || new_size == 0) {
1583 if (device->is_tgtdev_for_dev_replace) {
1588 old_size = btrfs_device_get_total_bytes(device);
1591 if (new_size > old_size) {
1595 new_size = old_size - new_size;
1596 } else if (mod > 0) {
1597 if (new_size > ULLONG_MAX - old_size) {
1601 new_size = old_size + new_size;
1604 if (new_size < 256 * 1024 * 1024) {
1608 if (new_size > device->bdev->bd_inode->i_size) {
1613 do_div(new_size, root->sectorsize);
1614 new_size *= root->sectorsize;
1616 printk_in_rcu(KERN_INFO "BTRFS: new size for %s is %llu\n",
1617 rcu_str_deref(device->name), new_size);
1619 if (new_size > old_size) {
1620 trans = btrfs_start_transaction(root, 0);
1621 if (IS_ERR(trans)) {
1622 ret = PTR_ERR(trans);
1625 ret = btrfs_grow_device(trans, device, new_size);
1626 btrfs_commit_transaction(trans, root);
1627 } else if (new_size < old_size) {
1628 ret = btrfs_shrink_device(device, new_size);
1629 } /* equal, nothing need to do */
1634 mutex_unlock(&root->fs_info->volume_mutex);
1635 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
1636 mnt_drop_write_file(file);
1640 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1641 char *name, unsigned long fd, int subvol,
1642 u64 *transid, bool readonly,
1643 struct btrfs_qgroup_inherit *inherit)
1648 ret = mnt_want_write_file(file);
1652 namelen = strlen(name);
1653 if (strchr(name, '/')) {
1655 goto out_drop_write;
1658 if (name[0] == '.' &&
1659 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1661 goto out_drop_write;
1665 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1666 NULL, transid, readonly, inherit);
1668 struct fd src = fdget(fd);
1669 struct inode *src_inode;
1672 goto out_drop_write;
1675 src_inode = file_inode(src.file);
1676 if (src_inode->i_sb != file_inode(file)->i_sb) {
1677 btrfs_info(BTRFS_I(src_inode)->root->fs_info,
1678 "Snapshot src from another FS");
1680 } else if (!inode_owner_or_capable(src_inode)) {
1682 * Subvolume creation is not restricted, but snapshots
1683 * are limited to own subvolumes only
1687 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1688 BTRFS_I(src_inode)->root,
1689 transid, readonly, inherit);
1694 mnt_drop_write_file(file);
1699 static noinline int btrfs_ioctl_snap_create(struct file *file,
1700 void __user *arg, int subvol)
1702 struct btrfs_ioctl_vol_args *vol_args;
1705 vol_args = memdup_user(arg, sizeof(*vol_args));
1706 if (IS_ERR(vol_args))
1707 return PTR_ERR(vol_args);
1708 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1710 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1711 vol_args->fd, subvol,
1718 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1719 void __user *arg, int subvol)
1721 struct btrfs_ioctl_vol_args_v2 *vol_args;
1725 bool readonly = false;
1726 struct btrfs_qgroup_inherit *inherit = NULL;
1728 vol_args = memdup_user(arg, sizeof(*vol_args));
1729 if (IS_ERR(vol_args))
1730 return PTR_ERR(vol_args);
1731 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1733 if (vol_args->flags &
1734 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1735 BTRFS_SUBVOL_QGROUP_INHERIT)) {
1740 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1742 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1744 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1745 if (vol_args->size > PAGE_CACHE_SIZE) {
1749 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1750 if (IS_ERR(inherit)) {
1751 ret = PTR_ERR(inherit);
1756 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1757 vol_args->fd, subvol, ptr,
1762 if (ptr && copy_to_user(arg +
1763 offsetof(struct btrfs_ioctl_vol_args_v2,
1775 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1778 struct inode *inode = file_inode(file);
1779 struct btrfs_root *root = BTRFS_I(inode)->root;
1783 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1786 down_read(&root->fs_info->subvol_sem);
1787 if (btrfs_root_readonly(root))
1788 flags |= BTRFS_SUBVOL_RDONLY;
1789 up_read(&root->fs_info->subvol_sem);
1791 if (copy_to_user(arg, &flags, sizeof(flags)))
1797 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1800 struct inode *inode = file_inode(file);
1801 struct btrfs_root *root = BTRFS_I(inode)->root;
1802 struct btrfs_trans_handle *trans;
1807 if (!inode_owner_or_capable(inode))
1810 ret = mnt_want_write_file(file);
1814 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1816 goto out_drop_write;
1819 if (copy_from_user(&flags, arg, sizeof(flags))) {
1821 goto out_drop_write;
1824 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1826 goto out_drop_write;
1829 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1831 goto out_drop_write;
1834 down_write(&root->fs_info->subvol_sem);
1837 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1840 root_flags = btrfs_root_flags(&root->root_item);
1841 if (flags & BTRFS_SUBVOL_RDONLY) {
1842 btrfs_set_root_flags(&root->root_item,
1843 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1846 * Block RO -> RW transition if this subvolume is involved in
1849 spin_lock(&root->root_item_lock);
1850 if (root->send_in_progress == 0) {
1851 btrfs_set_root_flags(&root->root_item,
1852 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1853 spin_unlock(&root->root_item_lock);
1855 spin_unlock(&root->root_item_lock);
1856 btrfs_warn(root->fs_info,
1857 "Attempt to set subvolume %llu read-write during send",
1858 root->root_key.objectid);
1864 trans = btrfs_start_transaction(root, 1);
1865 if (IS_ERR(trans)) {
1866 ret = PTR_ERR(trans);
1870 ret = btrfs_update_root(trans, root->fs_info->tree_root,
1871 &root->root_key, &root->root_item);
1873 btrfs_commit_transaction(trans, root);
1876 btrfs_set_root_flags(&root->root_item, root_flags);
1878 up_write(&root->fs_info->subvol_sem);
1880 mnt_drop_write_file(file);
1886 * helper to check if the subvolume references other subvolumes
1888 static noinline int may_destroy_subvol(struct btrfs_root *root)
1890 struct btrfs_path *path;
1891 struct btrfs_dir_item *di;
1892 struct btrfs_key key;
1896 path = btrfs_alloc_path();
1900 /* Make sure this root isn't set as the default subvol */
1901 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1902 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1903 dir_id, "default", 7, 0);
1904 if (di && !IS_ERR(di)) {
1905 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1906 if (key.objectid == root->root_key.objectid) {
1908 btrfs_err(root->fs_info, "deleting default subvolume "
1909 "%llu is not allowed", key.objectid);
1912 btrfs_release_path(path);
1915 key.objectid = root->root_key.objectid;
1916 key.type = BTRFS_ROOT_REF_KEY;
1917 key.offset = (u64)-1;
1919 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1926 if (path->slots[0] > 0) {
1928 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1929 if (key.objectid == root->root_key.objectid &&
1930 key.type == BTRFS_ROOT_REF_KEY)
1934 btrfs_free_path(path);
1938 static noinline int key_in_sk(struct btrfs_key *key,
1939 struct btrfs_ioctl_search_key *sk)
1941 struct btrfs_key test;
1944 test.objectid = sk->min_objectid;
1945 test.type = sk->min_type;
1946 test.offset = sk->min_offset;
1948 ret = btrfs_comp_cpu_keys(key, &test);
1952 test.objectid = sk->max_objectid;
1953 test.type = sk->max_type;
1954 test.offset = sk->max_offset;
1956 ret = btrfs_comp_cpu_keys(key, &test);
1962 static noinline int copy_to_sk(struct btrfs_root *root,
1963 struct btrfs_path *path,
1964 struct btrfs_key *key,
1965 struct btrfs_ioctl_search_key *sk,
1968 unsigned long *sk_offset,
1972 struct extent_buffer *leaf;
1973 struct btrfs_ioctl_search_header sh;
1974 unsigned long item_off;
1975 unsigned long item_len;
1981 leaf = path->nodes[0];
1982 slot = path->slots[0];
1983 nritems = btrfs_header_nritems(leaf);
1985 if (btrfs_header_generation(leaf) > sk->max_transid) {
1989 found_transid = btrfs_header_generation(leaf);
1991 for (i = slot; i < nritems; i++) {
1992 item_off = btrfs_item_ptr_offset(leaf, i);
1993 item_len = btrfs_item_size_nr(leaf, i);
1995 btrfs_item_key_to_cpu(leaf, key, i);
1996 if (!key_in_sk(key, sk))
1999 if (sizeof(sh) + item_len > *buf_size) {
2006 * return one empty item back for v1, which does not
2010 *buf_size = sizeof(sh) + item_len;
2015 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
2020 sh.objectid = key->objectid;
2021 sh.offset = key->offset;
2022 sh.type = key->type;
2024 sh.transid = found_transid;
2026 /* copy search result header */
2027 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
2032 *sk_offset += sizeof(sh);
2035 char __user *up = ubuf + *sk_offset;
2037 if (read_extent_buffer_to_user(leaf, up,
2038 item_off, item_len)) {
2043 *sk_offset += item_len;
2047 if (ret) /* -EOVERFLOW from above */
2050 if (*num_found >= sk->nr_items) {
2057 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
2059 else if (key->type < (u8)-1 && key->type < sk->max_type) {
2062 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
2070 * 0: all items from this leaf copied, continue with next
2071 * 1: * more items can be copied, but unused buffer is too small
2072 * * all items were found
2073 * Either way, it will stops the loop which iterates to the next
2075 * -EOVERFLOW: item was to large for buffer
2076 * -EFAULT: could not copy extent buffer back to userspace
2081 static noinline int search_ioctl(struct inode *inode,
2082 struct btrfs_ioctl_search_key *sk,
2086 struct btrfs_root *root;
2087 struct btrfs_key key;
2088 struct btrfs_path *path;
2089 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
2092 unsigned long sk_offset = 0;
2094 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2095 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2099 path = btrfs_alloc_path();
2103 if (sk->tree_id == 0) {
2104 /* search the root of the inode that was passed */
2105 root = BTRFS_I(inode)->root;
2107 key.objectid = sk->tree_id;
2108 key.type = BTRFS_ROOT_ITEM_KEY;
2109 key.offset = (u64)-1;
2110 root = btrfs_read_fs_root_no_name(info, &key);
2112 printk(KERN_ERR "BTRFS: could not find root %llu\n",
2114 btrfs_free_path(path);
2119 key.objectid = sk->min_objectid;
2120 key.type = sk->min_type;
2121 key.offset = sk->min_offset;
2124 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2130 ret = copy_to_sk(root, path, &key, sk, buf_size, ubuf,
2131 &sk_offset, &num_found);
2132 btrfs_release_path(path);
2140 sk->nr_items = num_found;
2141 btrfs_free_path(path);
2145 static noinline int btrfs_ioctl_tree_search(struct file *file,
2148 struct btrfs_ioctl_search_args __user *uargs;
2149 struct btrfs_ioctl_search_key sk;
2150 struct inode *inode;
2154 if (!capable(CAP_SYS_ADMIN))
2157 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2159 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2162 buf_size = sizeof(uargs->buf);
2164 inode = file_inode(file);
2165 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2168 * In the origin implementation an overflow is handled by returning a
2169 * search header with a len of zero, so reset ret.
2171 if (ret == -EOVERFLOW)
2174 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2179 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2182 struct btrfs_ioctl_search_args_v2 __user *uarg;
2183 struct btrfs_ioctl_search_args_v2 args;
2184 struct inode *inode;
2187 const size_t buf_limit = 16 * 1024 * 1024;
2189 if (!capable(CAP_SYS_ADMIN))
2192 /* copy search header and buffer size */
2193 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2194 if (copy_from_user(&args, uarg, sizeof(args)))
2197 buf_size = args.buf_size;
2199 if (buf_size < sizeof(struct btrfs_ioctl_search_header))
2202 /* limit result size to 16MB */
2203 if (buf_size > buf_limit)
2204 buf_size = buf_limit;
2206 inode = file_inode(file);
2207 ret = search_ioctl(inode, &args.key, &buf_size,
2208 (char *)(&uarg->buf[0]));
2209 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2211 else if (ret == -EOVERFLOW &&
2212 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2219 * Search INODE_REFs to identify path name of 'dirid' directory
2220 * in a 'tree_id' tree. and sets path name to 'name'.
2222 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2223 u64 tree_id, u64 dirid, char *name)
2225 struct btrfs_root *root;
2226 struct btrfs_key key;
2232 struct btrfs_inode_ref *iref;
2233 struct extent_buffer *l;
2234 struct btrfs_path *path;
2236 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2241 path = btrfs_alloc_path();
2245 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
2247 key.objectid = tree_id;
2248 key.type = BTRFS_ROOT_ITEM_KEY;
2249 key.offset = (u64)-1;
2250 root = btrfs_read_fs_root_no_name(info, &key);
2252 printk(KERN_ERR "BTRFS: could not find root %llu\n", tree_id);
2257 key.objectid = dirid;
2258 key.type = BTRFS_INODE_REF_KEY;
2259 key.offset = (u64)-1;
2262 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2266 ret = btrfs_previous_item(root, path, dirid,
2267 BTRFS_INODE_REF_KEY);
2277 slot = path->slots[0];
2278 btrfs_item_key_to_cpu(l, &key, slot);
2280 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2281 len = btrfs_inode_ref_name_len(l, iref);
2283 total_len += len + 1;
2285 ret = -ENAMETOOLONG;
2290 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2292 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2295 btrfs_release_path(path);
2296 key.objectid = key.offset;
2297 key.offset = (u64)-1;
2298 dirid = key.objectid;
2300 memmove(name, ptr, total_len);
2301 name[total_len] = '\0';
2304 btrfs_free_path(path);
2308 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2311 struct btrfs_ioctl_ino_lookup_args *args;
2312 struct inode *inode;
2315 if (!capable(CAP_SYS_ADMIN))
2318 args = memdup_user(argp, sizeof(*args));
2320 return PTR_ERR(args);
2322 inode = file_inode(file);
2324 if (args->treeid == 0)
2325 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2327 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2328 args->treeid, args->objectid,
2331 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2338 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2341 struct dentry *parent = file->f_path.dentry;
2342 struct dentry *dentry;
2343 struct inode *dir = parent->d_inode;
2344 struct inode *inode;
2345 struct btrfs_root *root = BTRFS_I(dir)->root;
2346 struct btrfs_root *dest = NULL;
2347 struct btrfs_ioctl_vol_args *vol_args;
2348 struct btrfs_trans_handle *trans;
2349 struct btrfs_block_rsv block_rsv;
2351 u64 qgroup_reserved;
2356 vol_args = memdup_user(arg, sizeof(*vol_args));
2357 if (IS_ERR(vol_args))
2358 return PTR_ERR(vol_args);
2360 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2361 namelen = strlen(vol_args->name);
2362 if (strchr(vol_args->name, '/') ||
2363 strncmp(vol_args->name, "..", namelen) == 0) {
2368 err = mnt_want_write_file(file);
2373 err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2375 goto out_drop_write;
2376 dentry = lookup_one_len(vol_args->name, parent, namelen);
2377 if (IS_ERR(dentry)) {
2378 err = PTR_ERR(dentry);
2379 goto out_unlock_dir;
2382 if (!dentry->d_inode) {
2387 inode = dentry->d_inode;
2388 dest = BTRFS_I(inode)->root;
2389 if (!capable(CAP_SYS_ADMIN)) {
2391 * Regular user. Only allow this with a special mount
2392 * option, when the user has write+exec access to the
2393 * subvol root, and when rmdir(2) would have been
2396 * Note that this is _not_ check that the subvol is
2397 * empty or doesn't contain data that we wouldn't
2398 * otherwise be able to delete.
2400 * Users who want to delete empty subvols should try
2404 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2408 * Do not allow deletion if the parent dir is the same
2409 * as the dir to be deleted. That means the ioctl
2410 * must be called on the dentry referencing the root
2411 * of the subvol, not a random directory contained
2418 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2423 /* check if subvolume may be deleted by a user */
2424 err = btrfs_may_delete(dir, dentry, 1);
2428 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2433 mutex_lock(&inode->i_mutex);
2436 * Don't allow to delete a subvolume with send in progress. This is
2437 * inside the i_mutex so the error handling that has to drop the bit
2438 * again is not run concurrently.
2440 spin_lock(&dest->root_item_lock);
2441 root_flags = btrfs_root_flags(&dest->root_item);
2442 if (dest->send_in_progress == 0) {
2443 btrfs_set_root_flags(&dest->root_item,
2444 root_flags | BTRFS_ROOT_SUBVOL_DEAD);
2445 spin_unlock(&dest->root_item_lock);
2447 spin_unlock(&dest->root_item_lock);
2448 btrfs_warn(root->fs_info,
2449 "Attempt to delete subvolume %llu during send",
2450 dest->root_key.objectid);
2455 d_invalidate(dentry);
2457 down_write(&root->fs_info->subvol_sem);
2459 err = may_destroy_subvol(dest);
2463 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2465 * One for dir inode, two for dir entries, two for root
2468 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2469 5, &qgroup_reserved, true);
2473 trans = btrfs_start_transaction(root, 0);
2474 if (IS_ERR(trans)) {
2475 err = PTR_ERR(trans);
2478 trans->block_rsv = &block_rsv;
2479 trans->bytes_reserved = block_rsv.size;
2481 ret = btrfs_unlink_subvol(trans, root, dir,
2482 dest->root_key.objectid,
2483 dentry->d_name.name,
2484 dentry->d_name.len);
2487 btrfs_abort_transaction(trans, root, ret);
2491 btrfs_record_root_in_trans(trans, dest);
2493 memset(&dest->root_item.drop_progress, 0,
2494 sizeof(dest->root_item.drop_progress));
2495 dest->root_item.drop_level = 0;
2496 btrfs_set_root_refs(&dest->root_item, 0);
2498 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
2499 ret = btrfs_insert_orphan_item(trans,
2500 root->fs_info->tree_root,
2501 dest->root_key.objectid);
2503 btrfs_abort_transaction(trans, root, ret);
2509 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2510 dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
2511 dest->root_key.objectid);
2512 if (ret && ret != -ENOENT) {
2513 btrfs_abort_transaction(trans, root, ret);
2517 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2518 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2519 dest->root_item.received_uuid,
2520 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2521 dest->root_key.objectid);
2522 if (ret && ret != -ENOENT) {
2523 btrfs_abort_transaction(trans, root, ret);
2530 trans->block_rsv = NULL;
2531 trans->bytes_reserved = 0;
2532 ret = btrfs_end_transaction(trans, root);
2535 inode->i_flags |= S_DEAD;
2537 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
2539 up_write(&root->fs_info->subvol_sem);
2541 spin_lock(&dest->root_item_lock);
2542 root_flags = btrfs_root_flags(&dest->root_item);
2543 btrfs_set_root_flags(&dest->root_item,
2544 root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
2545 spin_unlock(&dest->root_item_lock);
2547 mutex_unlock(&inode->i_mutex);
2549 shrink_dcache_sb(root->fs_info->sb);
2550 btrfs_invalidate_inodes(dest);
2552 ASSERT(dest->send_in_progress == 0);
2555 if (dest->ino_cache_inode) {
2556 iput(dest->ino_cache_inode);
2557 dest->ino_cache_inode = NULL;
2563 mutex_unlock(&dir->i_mutex);
2565 mnt_drop_write_file(file);
2571 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2573 struct inode *inode = file_inode(file);
2574 struct btrfs_root *root = BTRFS_I(inode)->root;
2575 struct btrfs_ioctl_defrag_range_args *range;
2578 ret = mnt_want_write_file(file);
2582 if (btrfs_root_readonly(root)) {
2587 switch (inode->i_mode & S_IFMT) {
2589 if (!capable(CAP_SYS_ADMIN)) {
2593 ret = btrfs_defrag_root(root);
2596 ret = btrfs_defrag_root(root->fs_info->extent_root);
2599 if (!(file->f_mode & FMODE_WRITE)) {
2604 range = kzalloc(sizeof(*range), GFP_KERNEL);
2611 if (copy_from_user(range, argp,
2617 /* compression requires us to start the IO */
2618 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2619 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2620 range->extent_thresh = (u32)-1;
2623 /* the rest are all set to zero by kzalloc */
2624 range->len = (u64)-1;
2626 ret = btrfs_defrag_file(file_inode(file), file,
2636 mnt_drop_write_file(file);
2640 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2642 struct btrfs_ioctl_vol_args *vol_args;
2645 if (!capable(CAP_SYS_ADMIN))
2648 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2650 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2653 mutex_lock(&root->fs_info->volume_mutex);
2654 vol_args = memdup_user(arg, sizeof(*vol_args));
2655 if (IS_ERR(vol_args)) {
2656 ret = PTR_ERR(vol_args);
2660 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2661 ret = btrfs_init_new_device(root, vol_args->name);
2664 btrfs_info(root->fs_info, "disk added %s",vol_args->name);
2668 mutex_unlock(&root->fs_info->volume_mutex);
2669 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2673 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2675 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2676 struct btrfs_ioctl_vol_args *vol_args;
2679 if (!capable(CAP_SYS_ADMIN))
2682 ret = mnt_want_write_file(file);
2686 vol_args = memdup_user(arg, sizeof(*vol_args));
2687 if (IS_ERR(vol_args)) {
2688 ret = PTR_ERR(vol_args);
2692 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2694 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2696 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2700 mutex_lock(&root->fs_info->volume_mutex);
2701 ret = btrfs_rm_device(root, vol_args->name);
2702 mutex_unlock(&root->fs_info->volume_mutex);
2703 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2706 btrfs_info(root->fs_info, "disk deleted %s",vol_args->name);
2711 mnt_drop_write_file(file);
2715 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2717 struct btrfs_ioctl_fs_info_args *fi_args;
2718 struct btrfs_device *device;
2719 struct btrfs_device *next;
2720 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2723 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2727 mutex_lock(&fs_devices->device_list_mutex);
2728 fi_args->num_devices = fs_devices->num_devices;
2729 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2731 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2732 if (device->devid > fi_args->max_id)
2733 fi_args->max_id = device->devid;
2735 mutex_unlock(&fs_devices->device_list_mutex);
2737 fi_args->nodesize = root->fs_info->super_copy->nodesize;
2738 fi_args->sectorsize = root->fs_info->super_copy->sectorsize;
2739 fi_args->clone_alignment = root->fs_info->super_copy->sectorsize;
2741 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2748 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2750 struct btrfs_ioctl_dev_info_args *di_args;
2751 struct btrfs_device *dev;
2752 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2754 char *s_uuid = NULL;
2756 di_args = memdup_user(arg, sizeof(*di_args));
2757 if (IS_ERR(di_args))
2758 return PTR_ERR(di_args);
2760 if (!btrfs_is_empty_uuid(di_args->uuid))
2761 s_uuid = di_args->uuid;
2763 mutex_lock(&fs_devices->device_list_mutex);
2764 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
2771 di_args->devid = dev->devid;
2772 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2773 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
2774 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2776 struct rcu_string *name;
2779 name = rcu_dereference(dev->name);
2780 strncpy(di_args->path, name->str, sizeof(di_args->path));
2782 di_args->path[sizeof(di_args->path) - 1] = 0;
2784 di_args->path[0] = '\0';
2788 mutex_unlock(&fs_devices->device_list_mutex);
2789 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2796 static struct page *extent_same_get_page(struct inode *inode, u64 off)
2800 struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
2802 index = off >> PAGE_CACHE_SHIFT;
2804 page = grab_cache_page(inode->i_mapping, index);
2808 if (!PageUptodate(page)) {
2809 if (extent_read_full_page_nolock(tree, page, btrfs_get_extent,
2813 if (!PageUptodate(page)) {
2815 page_cache_release(page);
2824 static inline void lock_extent_range(struct inode *inode, u64 off, u64 len)
2826 /* do any pending delalloc/csum calc on src, one way or
2827 another, and lock file content */
2829 struct btrfs_ordered_extent *ordered;
2830 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2831 ordered = btrfs_lookup_first_ordered_extent(inode,
2834 ordered->file_offset + ordered->len <= off ||
2835 ordered->file_offset >= off + len) &&
2836 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2837 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2839 btrfs_put_ordered_extent(ordered);
2842 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2844 btrfs_put_ordered_extent(ordered);
2845 btrfs_wait_ordered_range(inode, off, len);
2849 static void btrfs_double_unlock(struct inode *inode1, u64 loff1,
2850 struct inode *inode2, u64 loff2, u64 len)
2852 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2853 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2855 mutex_unlock(&inode1->i_mutex);
2856 mutex_unlock(&inode2->i_mutex);
2859 static void btrfs_double_lock(struct inode *inode1, u64 loff1,
2860 struct inode *inode2, u64 loff2, u64 len)
2862 if (inode1 < inode2) {
2863 swap(inode1, inode2);
2867 mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT);
2868 lock_extent_range(inode1, loff1, len);
2869 if (inode1 != inode2) {
2870 mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD);
2871 lock_extent_range(inode2, loff2, len);
2875 static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
2876 u64 dst_loff, u64 len)
2879 struct page *src_page, *dst_page;
2880 unsigned int cmp_len = PAGE_CACHE_SIZE;
2881 void *addr, *dst_addr;
2884 if (len < PAGE_CACHE_SIZE)
2887 src_page = extent_same_get_page(src, loff);
2890 dst_page = extent_same_get_page(dst, dst_loff);
2892 page_cache_release(src_page);
2895 addr = kmap_atomic(src_page);
2896 dst_addr = kmap_atomic(dst_page);
2898 flush_dcache_page(src_page);
2899 flush_dcache_page(dst_page);
2901 if (memcmp(addr, dst_addr, cmp_len))
2902 ret = BTRFS_SAME_DATA_DIFFERS;
2904 kunmap_atomic(addr);
2905 kunmap_atomic(dst_addr);
2906 page_cache_release(src_page);
2907 page_cache_release(dst_page);
2913 dst_loff += cmp_len;
2920 static int extent_same_check_offsets(struct inode *inode, u64 off, u64 len)
2922 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
2924 if (off + len > inode->i_size || off + len < off)
2926 /* Check that we are block aligned - btrfs_clone() requires this */
2927 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
2933 static int btrfs_extent_same(struct inode *src, u64 loff, u64 len,
2934 struct inode *dst, u64 dst_loff)
2939 * btrfs_clone() can't handle extents in the same file
2940 * yet. Once that works, we can drop this check and replace it
2941 * with a check for the same inode, but overlapping extents.
2946 btrfs_double_lock(src, loff, dst, dst_loff, len);
2948 ret = extent_same_check_offsets(src, loff, len);
2952 ret = extent_same_check_offsets(dst, dst_loff, len);
2956 /* don't make the dst file partly checksummed */
2957 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2958 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
2963 ret = btrfs_cmp_data(src, loff, dst, dst_loff, len);
2965 ret = btrfs_clone(src, dst, loff, len, len, dst_loff);
2968 btrfs_double_unlock(src, loff, dst, dst_loff, len);
2973 #define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024)
2975 static long btrfs_ioctl_file_extent_same(struct file *file,
2976 struct btrfs_ioctl_same_args __user *argp)
2978 struct btrfs_ioctl_same_args *same;
2979 struct btrfs_ioctl_same_extent_info *info;
2980 struct inode *src = file_inode(file);
2986 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
2987 bool is_admin = capable(CAP_SYS_ADMIN);
2990 if (!(file->f_mode & FMODE_READ))
2993 ret = mnt_want_write_file(file);
2997 if (get_user(count, &argp->dest_count)) {
3002 size = offsetof(struct btrfs_ioctl_same_args __user, info[count]);
3004 same = memdup_user(argp, size);
3007 ret = PTR_ERR(same);
3011 off = same->logical_offset;
3015 * Limit the total length we will dedupe for each operation.
3016 * This is intended to bound the total time spent in this
3017 * ioctl to something sane.
3019 if (len > BTRFS_MAX_DEDUPE_LEN)
3020 len = BTRFS_MAX_DEDUPE_LEN;
3022 if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) {
3024 * Btrfs does not support blocksize < page_size. As a
3025 * result, btrfs_cmp_data() won't correctly handle
3026 * this situation without an update.
3033 if (S_ISDIR(src->i_mode))
3037 if (!S_ISREG(src->i_mode))
3040 /* pre-format output fields to sane values */
3041 for (i = 0; i < count; i++) {
3042 same->info[i].bytes_deduped = 0ULL;
3043 same->info[i].status = 0;
3046 for (i = 0, info = same->info; i < count; i++, info++) {
3048 struct fd dst_file = fdget(info->fd);
3049 if (!dst_file.file) {
3050 info->status = -EBADF;
3053 dst = file_inode(dst_file.file);
3055 if (!(is_admin || (dst_file.file->f_mode & FMODE_WRITE))) {
3056 info->status = -EINVAL;
3057 } else if (file->f_path.mnt != dst_file.file->f_path.mnt) {
3058 info->status = -EXDEV;
3059 } else if (S_ISDIR(dst->i_mode)) {
3060 info->status = -EISDIR;
3061 } else if (!S_ISREG(dst->i_mode)) {
3062 info->status = -EACCES;
3064 info->status = btrfs_extent_same(src, off, len, dst,
3065 info->logical_offset);
3066 if (info->status == 0)
3067 info->bytes_deduped += len;
3072 ret = copy_to_user(argp, same, size);
3077 mnt_drop_write_file(file);
3081 /* Helper to check and see if this root currently has a ref on the given disk
3082 * bytenr. If it does then we need to update the quota for this root. This
3083 * doesn't do anything if quotas aren't enabled.
3085 static int check_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
3088 struct seq_list tree_mod_seq_elem = {};
3089 struct ulist *roots;
3090 struct ulist_iterator uiter;
3091 struct ulist_node *root_node = NULL;
3094 if (!root->fs_info->quota_enabled)
3097 btrfs_get_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3098 ret = btrfs_find_all_roots(trans, root->fs_info, disko,
3099 tree_mod_seq_elem.seq, &roots);
3103 ULIST_ITER_INIT(&uiter);
3104 while ((root_node = ulist_next(roots, &uiter))) {
3105 if (root_node->val == root->objectid) {
3112 btrfs_put_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3116 static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3117 struct inode *inode,
3122 struct btrfs_root *root = BTRFS_I(inode)->root;
3125 inode_inc_iversion(inode);
3126 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3128 * We round up to the block size at eof when determining which
3129 * extents to clone above, but shouldn't round up the file size.
3131 if (endoff > destoff + olen)
3132 endoff = destoff + olen;
3133 if (endoff > inode->i_size)
3134 btrfs_i_size_write(inode, endoff);
3136 ret = btrfs_update_inode(trans, root, inode);
3138 btrfs_abort_transaction(trans, root, ret);
3139 btrfs_end_transaction(trans, root);
3142 ret = btrfs_end_transaction(trans, root);
3147 static void clone_update_extent_map(struct inode *inode,
3148 const struct btrfs_trans_handle *trans,
3149 const struct btrfs_path *path,
3150 const u64 hole_offset,
3153 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
3154 struct extent_map *em;
3157 em = alloc_extent_map();
3159 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3160 &BTRFS_I(inode)->runtime_flags);
3165 struct btrfs_file_extent_item *fi;
3167 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3168 struct btrfs_file_extent_item);
3169 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3170 em->generation = -1;
3171 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3172 BTRFS_FILE_EXTENT_INLINE)
3173 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3174 &BTRFS_I(inode)->runtime_flags);
3176 em->start = hole_offset;
3178 em->ram_bytes = em->len;
3179 em->orig_start = hole_offset;
3180 em->block_start = EXTENT_MAP_HOLE;
3182 em->orig_block_len = 0;
3183 em->compress_type = BTRFS_COMPRESS_NONE;
3184 em->generation = trans->transid;
3188 write_lock(&em_tree->lock);
3189 ret = add_extent_mapping(em_tree, em, 1);
3190 write_unlock(&em_tree->lock);
3191 if (ret != -EEXIST) {
3192 free_extent_map(em);
3195 btrfs_drop_extent_cache(inode, em->start,
3196 em->start + em->len - 1, 0);
3200 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3201 &BTRFS_I(inode)->runtime_flags);
3205 * btrfs_clone() - clone a range from inode file to another
3207 * @src: Inode to clone from
3208 * @inode: Inode to clone to
3209 * @off: Offset within source to start clone from
3210 * @olen: Original length, passed by user, of range to clone
3211 * @olen_aligned: Block-aligned value of olen, extent_same uses
3212 * identical values here
3213 * @destoff: Offset within @inode to start clone
3215 static int btrfs_clone(struct inode *src, struct inode *inode,
3216 const u64 off, const u64 olen, const u64 olen_aligned,
3219 struct btrfs_root *root = BTRFS_I(inode)->root;
3220 struct btrfs_path *path = NULL;
3221 struct extent_buffer *leaf;
3222 struct btrfs_trans_handle *trans;
3224 struct btrfs_key key;
3229 const u64 len = olen_aligned;
3231 u64 last_dest_end = destoff;
3234 buf = vmalloc(root->nodesize);
3238 path = btrfs_alloc_path();
3246 key.objectid = btrfs_ino(src);
3247 key.type = BTRFS_EXTENT_DATA_KEY;
3252 * note the key will change type as we walk through the
3255 path->leave_spinning = 1;
3256 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3261 * First search, if no extent item that starts at offset off was
3262 * found but the previous item is an extent item, it's possible
3263 * it might overlap our target range, therefore process it.
3265 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3266 btrfs_item_key_to_cpu(path->nodes[0], &key,
3267 path->slots[0] - 1);
3268 if (key.type == BTRFS_EXTENT_DATA_KEY)
3272 nritems = btrfs_header_nritems(path->nodes[0]);
3275 if (path->slots[0] >= nritems) {
3276 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
3281 nritems = btrfs_header_nritems(path->nodes[0]);
3283 leaf = path->nodes[0];
3284 slot = path->slots[0];
3286 btrfs_item_key_to_cpu(leaf, &key, slot);
3287 if (key.type > BTRFS_EXTENT_DATA_KEY ||
3288 key.objectid != btrfs_ino(src))
3291 if (key.type == BTRFS_EXTENT_DATA_KEY) {
3292 struct btrfs_file_extent_item *extent;
3295 struct btrfs_key new_key;
3296 u64 disko = 0, diskl = 0;
3297 u64 datao = 0, datal = 0;
3301 extent = btrfs_item_ptr(leaf, slot,
3302 struct btrfs_file_extent_item);
3303 comp = btrfs_file_extent_compression(leaf, extent);
3304 type = btrfs_file_extent_type(leaf, extent);
3305 if (type == BTRFS_FILE_EXTENT_REG ||
3306 type == BTRFS_FILE_EXTENT_PREALLOC) {
3307 disko = btrfs_file_extent_disk_bytenr(leaf,
3309 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3311 datao = btrfs_file_extent_offset(leaf, extent);
3312 datal = btrfs_file_extent_num_bytes(leaf,
3314 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3315 /* take upper bound, may be compressed */
3316 datal = btrfs_file_extent_ram_bytes(leaf,
3321 * The first search might have left us at an extent
3322 * item that ends before our target range's start, can
3323 * happen if we have holes and NO_HOLES feature enabled.
3325 if (key.offset + datal <= off) {
3328 } else if (key.offset >= off + len) {
3332 size = btrfs_item_size_nr(leaf, slot);
3333 read_extent_buffer(leaf, buf,
3334 btrfs_item_ptr_offset(leaf, slot),
3337 btrfs_release_path(path);
3338 path->leave_spinning = 0;
3340 memcpy(&new_key, &key, sizeof(new_key));
3341 new_key.objectid = btrfs_ino(inode);
3342 if (off <= key.offset)
3343 new_key.offset = key.offset + destoff - off;
3345 new_key.offset = destoff;
3348 * Deal with a hole that doesn't have an extent item
3349 * that represents it (NO_HOLES feature enabled).
3350 * This hole is either in the middle of the cloning
3351 * range or at the beginning (fully overlaps it or
3352 * partially overlaps it).
3354 if (new_key.offset != last_dest_end)
3355 drop_start = last_dest_end;
3357 drop_start = new_key.offset;
3360 * 1 - adjusting old extent (we may have to split it)
3361 * 1 - add new extent
3364 trans = btrfs_start_transaction(root, 3);
3365 if (IS_ERR(trans)) {
3366 ret = PTR_ERR(trans);
3370 if (type == BTRFS_FILE_EXTENT_REG ||
3371 type == BTRFS_FILE_EXTENT_PREALLOC) {
3373 * a | --- range to clone ---| b
3374 * | ------------- extent ------------- |
3377 /* subtract range b */
3378 if (key.offset + datal > off + len)
3379 datal = off + len - key.offset;
3381 /* subtract range a */
3382 if (off > key.offset) {
3383 datao += off - key.offset;
3384 datal -= off - key.offset;
3387 ret = btrfs_drop_extents(trans, root, inode,
3389 new_key.offset + datal,
3392 if (ret != -EOPNOTSUPP)
3393 btrfs_abort_transaction(trans,
3395 btrfs_end_transaction(trans, root);
3399 ret = btrfs_insert_empty_item(trans, root, path,
3402 btrfs_abort_transaction(trans, root,
3404 btrfs_end_transaction(trans, root);
3408 leaf = path->nodes[0];
3409 slot = path->slots[0];
3410 write_extent_buffer(leaf, buf,
3411 btrfs_item_ptr_offset(leaf, slot),
3414 extent = btrfs_item_ptr(leaf, slot,
3415 struct btrfs_file_extent_item);
3417 /* disko == 0 means it's a hole */
3421 btrfs_set_file_extent_offset(leaf, extent,
3423 btrfs_set_file_extent_num_bytes(leaf, extent,
3427 * We need to look up the roots that point at
3428 * this bytenr and see if the new root does. If
3429 * it does not we need to make sure we update
3430 * quotas appropriately.
3432 if (disko && root != BTRFS_I(src)->root &&
3433 disko != last_disko) {
3434 no_quota = check_ref(trans, root,
3437 btrfs_abort_transaction(trans,
3440 btrfs_end_transaction(trans,
3448 inode_add_bytes(inode, datal);
3449 ret = btrfs_inc_extent_ref(trans, root,
3451 root->root_key.objectid,
3453 new_key.offset - datao,
3456 btrfs_abort_transaction(trans,
3459 btrfs_end_transaction(trans,
3465 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3468 u64 aligned_end = 0;
3470 if (off > key.offset) {
3471 skip = off - key.offset;
3472 new_key.offset += skip;
3475 if (key.offset + datal > off + len)
3476 trim = key.offset + datal - (off + len);
3478 if (comp && (skip || trim)) {
3480 btrfs_end_transaction(trans, root);
3483 size -= skip + trim;
3484 datal -= skip + trim;
3486 aligned_end = ALIGN(new_key.offset + datal,
3488 ret = btrfs_drop_extents(trans, root, inode,
3493 if (ret != -EOPNOTSUPP)
3494 btrfs_abort_transaction(trans,
3496 btrfs_end_transaction(trans, root);
3500 ret = btrfs_insert_empty_item(trans, root, path,
3503 btrfs_abort_transaction(trans, root,
3505 btrfs_end_transaction(trans, root);
3511 btrfs_file_extent_calc_inline_size(0);
3512 memmove(buf+start, buf+start+skip,
3516 leaf = path->nodes[0];
3517 slot = path->slots[0];
3518 write_extent_buffer(leaf, buf,
3519 btrfs_item_ptr_offset(leaf, slot),
3521 inode_add_bytes(inode, datal);
3524 /* If we have an implicit hole (NO_HOLES feature). */
3525 if (drop_start < new_key.offset)
3526 clone_update_extent_map(inode, trans,
3528 new_key.offset - drop_start);
3530 clone_update_extent_map(inode, trans, path, 0, 0);
3532 btrfs_mark_buffer_dirty(leaf);
3533 btrfs_release_path(path);
3535 last_dest_end = ALIGN(new_key.offset + datal,
3537 ret = clone_finish_inode_update(trans, inode,
3542 if (new_key.offset + datal >= destoff + len)
3545 btrfs_release_path(path);
3550 if (last_dest_end < destoff + len) {
3552 * We have an implicit hole (NO_HOLES feature is enabled) that
3553 * fully or partially overlaps our cloning range at its end.
3555 btrfs_release_path(path);
3558 * 1 - remove extent(s)
3561 trans = btrfs_start_transaction(root, 2);
3562 if (IS_ERR(trans)) {
3563 ret = PTR_ERR(trans);
3566 ret = btrfs_drop_extents(trans, root, inode,
3567 last_dest_end, destoff + len, 1);
3569 if (ret != -EOPNOTSUPP)
3570 btrfs_abort_transaction(trans, root, ret);
3571 btrfs_end_transaction(trans, root);
3574 clone_update_extent_map(inode, trans, NULL, last_dest_end,
3575 destoff + len - last_dest_end);
3576 ret = clone_finish_inode_update(trans, inode, destoff + len,
3581 btrfs_free_path(path);
3586 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
3587 u64 off, u64 olen, u64 destoff)
3589 struct inode *inode = file_inode(file);
3590 struct btrfs_root *root = BTRFS_I(inode)->root;
3595 u64 bs = root->fs_info->sb->s_blocksize;
3600 * - split compressed inline extents. annoying: we need to
3601 * decompress into destination's address_space (the file offset
3602 * may change, so source mapping won't do), then recompress (or
3603 * otherwise reinsert) a subrange.
3605 * - split destination inode's inline extents. The inline extents can
3606 * be either compressed or non-compressed.
3609 /* the destination must be opened for writing */
3610 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
3613 if (btrfs_root_readonly(root))
3616 ret = mnt_want_write_file(file);
3620 src_file = fdget(srcfd);
3621 if (!src_file.file) {
3623 goto out_drop_write;
3627 if (src_file.file->f_path.mnt != file->f_path.mnt)
3630 src = file_inode(src_file.file);
3636 /* the src must be open for reading */
3637 if (!(src_file.file->f_mode & FMODE_READ))
3640 /* don't make the dst file partly checksummed */
3641 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3642 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3646 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3650 if (src->i_sb != inode->i_sb)
3655 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
3656 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
3658 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
3659 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
3662 mutex_lock(&src->i_mutex);
3665 /* determine range to clone */
3667 if (off + len > src->i_size || off + len < off)
3670 olen = len = src->i_size - off;
3671 /* if we extend to eof, continue to block boundary */
3672 if (off + len == src->i_size)
3673 len = ALIGN(src->i_size, bs) - off;
3675 /* verify the end result is block aligned */
3676 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3677 !IS_ALIGNED(destoff, bs))
3680 /* verify if ranges are overlapped within the same file */
3682 if (destoff + len > off && destoff < off + len)
3686 if (destoff > inode->i_size) {
3687 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3693 * Lock the target range too. Right after we replace the file extent
3694 * items in the fs tree (which now point to the cloned data), we might
3695 * have a worker replace them with extent items relative to a write
3696 * operation that was issued before this clone operation (i.e. confront
3697 * with inode.c:btrfs_finish_ordered_io).
3700 u64 lock_start = min_t(u64, off, destoff);
3701 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
3703 lock_extent_range(src, lock_start, lock_len);
3705 lock_extent_range(src, off, len);
3706 lock_extent_range(inode, destoff, len);
3709 ret = btrfs_clone(src, inode, off, olen, len, destoff);
3712 u64 lock_start = min_t(u64, off, destoff);
3713 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3715 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3717 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
3718 unlock_extent(&BTRFS_I(inode)->io_tree, destoff,
3722 * Truncate page cache pages so that future reads will see the cloned
3723 * data immediately and not the previous data.
3725 truncate_inode_pages_range(&inode->i_data, destoff,
3726 PAGE_CACHE_ALIGN(destoff + len) - 1);
3730 mutex_unlock(&src->i_mutex);
3731 mutex_unlock(&inode->i_mutex);
3733 mutex_unlock(&inode->i_mutex);
3734 mutex_unlock(&src->i_mutex);
3737 mutex_unlock(&src->i_mutex);
3742 mnt_drop_write_file(file);
3746 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
3748 struct btrfs_ioctl_clone_range_args args;
3750 if (copy_from_user(&args, argp, sizeof(args)))
3752 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
3753 args.src_length, args.dest_offset);
3757 * there are many ways the trans_start and trans_end ioctls can lead
3758 * to deadlocks. They should only be used by applications that
3759 * basically own the machine, and have a very in depth understanding
3760 * of all the possible deadlocks and enospc problems.
3762 static long btrfs_ioctl_trans_start(struct file *file)
3764 struct inode *inode = file_inode(file);
3765 struct btrfs_root *root = BTRFS_I(inode)->root;
3766 struct btrfs_trans_handle *trans;
3770 if (!capable(CAP_SYS_ADMIN))
3774 if (file->private_data)
3778 if (btrfs_root_readonly(root))
3781 ret = mnt_want_write_file(file);
3785 atomic_inc(&root->fs_info->open_ioctl_trans);
3788 trans = btrfs_start_ioctl_transaction(root);
3792 file->private_data = trans;
3796 atomic_dec(&root->fs_info->open_ioctl_trans);
3797 mnt_drop_write_file(file);
3802 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3804 struct inode *inode = file_inode(file);
3805 struct btrfs_root *root = BTRFS_I(inode)->root;
3806 struct btrfs_root *new_root;
3807 struct btrfs_dir_item *di;
3808 struct btrfs_trans_handle *trans;
3809 struct btrfs_path *path;
3810 struct btrfs_key location;
3811 struct btrfs_disk_key disk_key;
3816 if (!capable(CAP_SYS_ADMIN))
3819 ret = mnt_want_write_file(file);
3823 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3829 objectid = BTRFS_FS_TREE_OBJECTID;
3831 location.objectid = objectid;
3832 location.type = BTRFS_ROOT_ITEM_KEY;
3833 location.offset = (u64)-1;
3835 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
3836 if (IS_ERR(new_root)) {
3837 ret = PTR_ERR(new_root);
3841 path = btrfs_alloc_path();
3846 path->leave_spinning = 1;
3848 trans = btrfs_start_transaction(root, 1);
3849 if (IS_ERR(trans)) {
3850 btrfs_free_path(path);
3851 ret = PTR_ERR(trans);
3855 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
3856 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
3857 dir_id, "default", 7, 1);
3858 if (IS_ERR_OR_NULL(di)) {
3859 btrfs_free_path(path);
3860 btrfs_end_transaction(trans, root);
3861 btrfs_err(new_root->fs_info, "Umm, you don't have the default dir"
3862 "item, this isn't going to work");
3867 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3868 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3869 btrfs_mark_buffer_dirty(path->nodes[0]);
3870 btrfs_free_path(path);
3872 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
3873 btrfs_end_transaction(trans, root);
3875 mnt_drop_write_file(file);
3879 void btrfs_get_block_group_info(struct list_head *groups_list,
3880 struct btrfs_ioctl_space_info *space)
3882 struct btrfs_block_group_cache *block_group;
3884 space->total_bytes = 0;
3885 space->used_bytes = 0;
3887 list_for_each_entry(block_group, groups_list, list) {
3888 space->flags = block_group->flags;
3889 space->total_bytes += block_group->key.offset;
3890 space->used_bytes +=
3891 btrfs_block_group_used(&block_group->item);
3895 static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
3897 struct btrfs_ioctl_space_args space_args;
3898 struct btrfs_ioctl_space_info space;
3899 struct btrfs_ioctl_space_info *dest;
3900 struct btrfs_ioctl_space_info *dest_orig;
3901 struct btrfs_ioctl_space_info __user *user_dest;
3902 struct btrfs_space_info *info;
3903 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3904 BTRFS_BLOCK_GROUP_SYSTEM,
3905 BTRFS_BLOCK_GROUP_METADATA,
3906 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3913 if (copy_from_user(&space_args,
3914 (struct btrfs_ioctl_space_args __user *)arg,
3915 sizeof(space_args)))
3918 for (i = 0; i < num_types; i++) {
3919 struct btrfs_space_info *tmp;
3923 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3925 if (tmp->flags == types[i]) {
3935 down_read(&info->groups_sem);
3936 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3937 if (!list_empty(&info->block_groups[c]))
3940 up_read(&info->groups_sem);
3944 * Global block reserve, exported as a space_info
3948 /* space_slots == 0 means they are asking for a count */
3949 if (space_args.space_slots == 0) {
3950 space_args.total_spaces = slot_count;
3954 slot_count = min_t(u64, space_args.space_slots, slot_count);
3956 alloc_size = sizeof(*dest) * slot_count;
3958 /* we generally have at most 6 or so space infos, one for each raid
3959 * level. So, a whole page should be more than enough for everyone
3961 if (alloc_size > PAGE_CACHE_SIZE)
3964 space_args.total_spaces = 0;
3965 dest = kmalloc(alloc_size, GFP_NOFS);
3970 /* now we have a buffer to copy into */
3971 for (i = 0; i < num_types; i++) {
3972 struct btrfs_space_info *tmp;
3979 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3981 if (tmp->flags == types[i]) {
3990 down_read(&info->groups_sem);
3991 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3992 if (!list_empty(&info->block_groups[c])) {
3993 btrfs_get_block_group_info(
3994 &info->block_groups[c], &space);
3995 memcpy(dest, &space, sizeof(space));
3997 space_args.total_spaces++;
4003 up_read(&info->groups_sem);
4007 * Add global block reserve
4010 struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv;
4012 spin_lock(&block_rsv->lock);
4013 space.total_bytes = block_rsv->size;
4014 space.used_bytes = block_rsv->size - block_rsv->reserved;
4015 spin_unlock(&block_rsv->lock);
4016 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
4017 memcpy(dest, &space, sizeof(space));
4018 space_args.total_spaces++;
4021 user_dest = (struct btrfs_ioctl_space_info __user *)
4022 (arg + sizeof(struct btrfs_ioctl_space_args));
4024 if (copy_to_user(user_dest, dest_orig, alloc_size))
4029 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
4036 * there are many ways the trans_start and trans_end ioctls can lead
4037 * to deadlocks. They should only be used by applications that
4038 * basically own the machine, and have a very in depth understanding
4039 * of all the possible deadlocks and enospc problems.
4041 long btrfs_ioctl_trans_end(struct file *file)
4043 struct inode *inode = file_inode(file);
4044 struct btrfs_root *root = BTRFS_I(inode)->root;
4045 struct btrfs_trans_handle *trans;
4047 trans = file->private_data;
4050 file->private_data = NULL;
4052 btrfs_end_transaction(trans, root);
4054 atomic_dec(&root->fs_info->open_ioctl_trans);
4056 mnt_drop_write_file(file);
4060 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4063 struct btrfs_trans_handle *trans;
4067 trans = btrfs_attach_transaction_barrier(root);
4068 if (IS_ERR(trans)) {
4069 if (PTR_ERR(trans) != -ENOENT)
4070 return PTR_ERR(trans);
4072 /* No running transaction, don't bother */
4073 transid = root->fs_info->last_trans_committed;
4076 transid = trans->transid;
4077 ret = btrfs_commit_transaction_async(trans, root, 0);
4079 btrfs_end_transaction(trans, root);
4084 if (copy_to_user(argp, &transid, sizeof(transid)))
4089 static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
4095 if (copy_from_user(&transid, argp, sizeof(transid)))
4098 transid = 0; /* current trans */
4100 return btrfs_wait_for_commit(root, transid);
4103 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
4105 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4106 struct btrfs_ioctl_scrub_args *sa;
4109 if (!capable(CAP_SYS_ADMIN))
4112 sa = memdup_user(arg, sizeof(*sa));
4116 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4117 ret = mnt_want_write_file(file);
4122 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
4123 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4126 if (copy_to_user(arg, sa, sizeof(*sa)))
4129 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4130 mnt_drop_write_file(file);
4136 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
4138 if (!capable(CAP_SYS_ADMIN))
4141 return btrfs_scrub_cancel(root->fs_info);
4144 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
4147 struct btrfs_ioctl_scrub_args *sa;
4150 if (!capable(CAP_SYS_ADMIN))
4153 sa = memdup_user(arg, sizeof(*sa));
4157 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
4159 if (copy_to_user(arg, sa, sizeof(*sa)))
4166 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
4169 struct btrfs_ioctl_get_dev_stats *sa;
4172 sa = memdup_user(arg, sizeof(*sa));
4176 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4181 ret = btrfs_get_dev_stats(root, sa);
4183 if (copy_to_user(arg, sa, sizeof(*sa)))
4190 static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
4192 struct btrfs_ioctl_dev_replace_args *p;
4195 if (!capable(CAP_SYS_ADMIN))
4198 p = memdup_user(arg, sizeof(*p));
4203 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
4204 if (root->fs_info->sb->s_flags & MS_RDONLY) {
4209 &root->fs_info->mutually_exclusive_operation_running,
4211 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4213 ret = btrfs_dev_replace_start(root, p);
4215 &root->fs_info->mutually_exclusive_operation_running,
4219 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4220 btrfs_dev_replace_status(root->fs_info, p);
4223 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4224 ret = btrfs_dev_replace_cancel(root->fs_info, p);
4231 if (copy_to_user(arg, p, sizeof(*p)))
4238 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4244 struct btrfs_ioctl_ino_path_args *ipa = NULL;
4245 struct inode_fs_paths *ipath = NULL;
4246 struct btrfs_path *path;
4248 if (!capable(CAP_DAC_READ_SEARCH))
4251 path = btrfs_alloc_path();
4257 ipa = memdup_user(arg, sizeof(*ipa));
4264 size = min_t(u32, ipa->size, 4096);
4265 ipath = init_ipath(size, root, path);
4266 if (IS_ERR(ipath)) {
4267 ret = PTR_ERR(ipath);
4272 ret = paths_from_inode(ipa->inum, ipath);
4276 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
4277 rel_ptr = ipath->fspath->val[i] -
4278 (u64)(unsigned long)ipath->fspath->val;
4279 ipath->fspath->val[i] = rel_ptr;
4282 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
4283 (void *)(unsigned long)ipath->fspath, size);
4290 btrfs_free_path(path);
4297 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4299 struct btrfs_data_container *inodes = ctx;
4300 const size_t c = 3 * sizeof(u64);
4302 if (inodes->bytes_left >= c) {
4303 inodes->bytes_left -= c;
4304 inodes->val[inodes->elem_cnt] = inum;
4305 inodes->val[inodes->elem_cnt + 1] = offset;
4306 inodes->val[inodes->elem_cnt + 2] = root;
4307 inodes->elem_cnt += 3;
4309 inodes->bytes_missing += c - inodes->bytes_left;
4310 inodes->bytes_left = 0;
4311 inodes->elem_missed += 3;
4317 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
4322 struct btrfs_ioctl_logical_ino_args *loi;
4323 struct btrfs_data_container *inodes = NULL;
4324 struct btrfs_path *path = NULL;
4326 if (!capable(CAP_SYS_ADMIN))
4329 loi = memdup_user(arg, sizeof(*loi));
4336 path = btrfs_alloc_path();
4342 size = min_t(u32, loi->size, 64 * 1024);
4343 inodes = init_data_container(size);
4344 if (IS_ERR(inodes)) {
4345 ret = PTR_ERR(inodes);
4350 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
4351 build_ino_list, inodes);
4357 ret = copy_to_user((void *)(unsigned long)loi->inodes,
4358 (void *)(unsigned long)inodes, size);
4363 btrfs_free_path(path);
4370 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
4371 struct btrfs_ioctl_balance_args *bargs)
4373 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4375 bargs->flags = bctl->flags;
4377 if (atomic_read(&fs_info->balance_running))
4378 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4379 if (atomic_read(&fs_info->balance_pause_req))
4380 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
4381 if (atomic_read(&fs_info->balance_cancel_req))
4382 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
4384 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4385 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4386 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4389 spin_lock(&fs_info->balance_lock);
4390 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4391 spin_unlock(&fs_info->balance_lock);
4393 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4397 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4399 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4400 struct btrfs_fs_info *fs_info = root->fs_info;
4401 struct btrfs_ioctl_balance_args *bargs;
4402 struct btrfs_balance_control *bctl;
4403 bool need_unlock; /* for mut. excl. ops lock */
4406 if (!capable(CAP_SYS_ADMIN))
4409 ret = mnt_want_write_file(file);
4414 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
4415 mutex_lock(&fs_info->volume_mutex);
4416 mutex_lock(&fs_info->balance_mutex);
4422 * mut. excl. ops lock is locked. Three possibilites:
4423 * (1) some other op is running
4424 * (2) balance is running
4425 * (3) balance is paused -- special case (think resume)
4427 mutex_lock(&fs_info->balance_mutex);
4428 if (fs_info->balance_ctl) {
4429 /* this is either (2) or (3) */
4430 if (!atomic_read(&fs_info->balance_running)) {
4431 mutex_unlock(&fs_info->balance_mutex);
4432 if (!mutex_trylock(&fs_info->volume_mutex))
4434 mutex_lock(&fs_info->balance_mutex);
4436 if (fs_info->balance_ctl &&
4437 !atomic_read(&fs_info->balance_running)) {
4439 need_unlock = false;
4443 mutex_unlock(&fs_info->balance_mutex);
4444 mutex_unlock(&fs_info->volume_mutex);
4448 mutex_unlock(&fs_info->balance_mutex);
4454 mutex_unlock(&fs_info->balance_mutex);
4455 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4460 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
4463 bargs = memdup_user(arg, sizeof(*bargs));
4464 if (IS_ERR(bargs)) {
4465 ret = PTR_ERR(bargs);
4469 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4470 if (!fs_info->balance_ctl) {
4475 bctl = fs_info->balance_ctl;
4476 spin_lock(&fs_info->balance_lock);
4477 bctl->flags |= BTRFS_BALANCE_RESUME;
4478 spin_unlock(&fs_info->balance_lock);
4486 if (fs_info->balance_ctl) {
4491 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
4497 bctl->fs_info = fs_info;
4499 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4500 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4501 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4503 bctl->flags = bargs->flags;
4505 /* balance everything - no filters */
4506 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4511 * Ownership of bctl and mutually_exclusive_operation_running
4512 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4513 * or, if restriper was paused all the way until unmount, in
4514 * free_fs_info. mutually_exclusive_operation_running is
4515 * cleared in __cancel_balance.
4517 need_unlock = false;
4519 ret = btrfs_balance(bctl, bargs);
4522 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4529 mutex_unlock(&fs_info->balance_mutex);
4530 mutex_unlock(&fs_info->volume_mutex);
4532 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4534 mnt_drop_write_file(file);
4538 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4540 if (!capable(CAP_SYS_ADMIN))
4544 case BTRFS_BALANCE_CTL_PAUSE:
4545 return btrfs_pause_balance(root->fs_info);
4546 case BTRFS_BALANCE_CTL_CANCEL:
4547 return btrfs_cancel_balance(root->fs_info);
4553 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4556 struct btrfs_fs_info *fs_info = root->fs_info;
4557 struct btrfs_ioctl_balance_args *bargs;
4560 if (!capable(CAP_SYS_ADMIN))
4563 mutex_lock(&fs_info->balance_mutex);
4564 if (!fs_info->balance_ctl) {
4569 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
4575 update_ioctl_balance_args(fs_info, 1, bargs);
4577 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4582 mutex_unlock(&fs_info->balance_mutex);
4586 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4588 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4589 struct btrfs_ioctl_quota_ctl_args *sa;
4590 struct btrfs_trans_handle *trans = NULL;
4594 if (!capable(CAP_SYS_ADMIN))
4597 ret = mnt_want_write_file(file);
4601 sa = memdup_user(arg, sizeof(*sa));
4607 down_write(&root->fs_info->subvol_sem);
4608 trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4609 if (IS_ERR(trans)) {
4610 ret = PTR_ERR(trans);
4615 case BTRFS_QUOTA_CTL_ENABLE:
4616 ret = btrfs_quota_enable(trans, root->fs_info);
4618 case BTRFS_QUOTA_CTL_DISABLE:
4619 ret = btrfs_quota_disable(trans, root->fs_info);
4626 err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4631 up_write(&root->fs_info->subvol_sem);
4633 mnt_drop_write_file(file);
4637 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4639 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4640 struct btrfs_ioctl_qgroup_assign_args *sa;
4641 struct btrfs_trans_handle *trans;
4645 if (!capable(CAP_SYS_ADMIN))
4648 ret = mnt_want_write_file(file);
4652 sa = memdup_user(arg, sizeof(*sa));
4658 trans = btrfs_join_transaction(root);
4659 if (IS_ERR(trans)) {
4660 ret = PTR_ERR(trans);
4664 /* FIXME: check if the IDs really exist */
4666 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4669 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4673 err = btrfs_end_transaction(trans, root);
4680 mnt_drop_write_file(file);
4684 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4686 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4687 struct btrfs_ioctl_qgroup_create_args *sa;
4688 struct btrfs_trans_handle *trans;
4692 if (!capable(CAP_SYS_ADMIN))
4695 ret = mnt_want_write_file(file);
4699 sa = memdup_user(arg, sizeof(*sa));
4705 if (!sa->qgroupid) {
4710 trans = btrfs_join_transaction(root);
4711 if (IS_ERR(trans)) {
4712 ret = PTR_ERR(trans);
4716 /* FIXME: check if the IDs really exist */
4718 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
4721 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4724 err = btrfs_end_transaction(trans, root);
4731 mnt_drop_write_file(file);
4735 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4737 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4738 struct btrfs_ioctl_qgroup_limit_args *sa;
4739 struct btrfs_trans_handle *trans;
4744 if (!capable(CAP_SYS_ADMIN))
4747 ret = mnt_want_write_file(file);
4751 sa = memdup_user(arg, sizeof(*sa));
4757 trans = btrfs_join_transaction(root);
4758 if (IS_ERR(trans)) {
4759 ret = PTR_ERR(trans);
4763 qgroupid = sa->qgroupid;
4765 /* take the current subvol as qgroup */
4766 qgroupid = root->root_key.objectid;
4769 /* FIXME: check if the IDs really exist */
4770 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
4772 err = btrfs_end_transaction(trans, root);
4779 mnt_drop_write_file(file);
4783 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4785 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4786 struct btrfs_ioctl_quota_rescan_args *qsa;
4789 if (!capable(CAP_SYS_ADMIN))
4792 ret = mnt_want_write_file(file);
4796 qsa = memdup_user(arg, sizeof(*qsa));
4807 ret = btrfs_qgroup_rescan(root->fs_info);
4812 mnt_drop_write_file(file);
4816 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
4818 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4819 struct btrfs_ioctl_quota_rescan_args *qsa;
4822 if (!capable(CAP_SYS_ADMIN))
4825 qsa = kzalloc(sizeof(*qsa), GFP_NOFS);
4829 if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4831 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
4834 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4841 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
4843 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4845 if (!capable(CAP_SYS_ADMIN))
4848 return btrfs_qgroup_wait_for_completion(root->fs_info);
4851 static long _btrfs_ioctl_set_received_subvol(struct file *file,
4852 struct btrfs_ioctl_received_subvol_args *sa)
4854 struct inode *inode = file_inode(file);
4855 struct btrfs_root *root = BTRFS_I(inode)->root;
4856 struct btrfs_root_item *root_item = &root->root_item;
4857 struct btrfs_trans_handle *trans;
4858 struct timespec ct = CURRENT_TIME;
4860 int received_uuid_changed;
4862 if (!inode_owner_or_capable(inode))
4865 ret = mnt_want_write_file(file);
4869 down_write(&root->fs_info->subvol_sem);
4871 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
4876 if (btrfs_root_readonly(root)) {
4883 * 2 - uuid items (received uuid + subvol uuid)
4885 trans = btrfs_start_transaction(root, 3);
4886 if (IS_ERR(trans)) {
4887 ret = PTR_ERR(trans);
4892 sa->rtransid = trans->transid;
4893 sa->rtime.sec = ct.tv_sec;
4894 sa->rtime.nsec = ct.tv_nsec;
4896 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4898 if (received_uuid_changed &&
4899 !btrfs_is_empty_uuid(root_item->received_uuid))
4900 btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
4901 root_item->received_uuid,
4902 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4903 root->root_key.objectid);
4904 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4905 btrfs_set_root_stransid(root_item, sa->stransid);
4906 btrfs_set_root_rtransid(root_item, sa->rtransid);
4907 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4908 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4909 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4910 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4912 ret = btrfs_update_root(trans, root->fs_info->tree_root,
4913 &root->root_key, &root->root_item);
4915 btrfs_end_transaction(trans, root);
4918 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4919 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
4921 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4922 root->root_key.objectid);
4923 if (ret < 0 && ret != -EEXIST) {
4924 btrfs_abort_transaction(trans, root, ret);
4928 ret = btrfs_commit_transaction(trans, root);
4930 btrfs_abort_transaction(trans, root, ret);
4935 up_write(&root->fs_info->subvol_sem);
4936 mnt_drop_write_file(file);
4941 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4944 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4945 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4948 args32 = memdup_user(arg, sizeof(*args32));
4949 if (IS_ERR(args32)) {
4950 ret = PTR_ERR(args32);
4955 args64 = kmalloc(sizeof(*args64), GFP_NOFS);
4961 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4962 args64->stransid = args32->stransid;
4963 args64->rtransid = args32->rtransid;
4964 args64->stime.sec = args32->stime.sec;
4965 args64->stime.nsec = args32->stime.nsec;
4966 args64->rtime.sec = args32->rtime.sec;
4967 args64->rtime.nsec = args32->rtime.nsec;
4968 args64->flags = args32->flags;
4970 ret = _btrfs_ioctl_set_received_subvol(file, args64);
4974 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4975 args32->stransid = args64->stransid;
4976 args32->rtransid = args64->rtransid;
4977 args32->stime.sec = args64->stime.sec;
4978 args32->stime.nsec = args64->stime.nsec;
4979 args32->rtime.sec = args64->rtime.sec;
4980 args32->rtime.nsec = args64->rtime.nsec;
4981 args32->flags = args64->flags;
4983 ret = copy_to_user(arg, args32, sizeof(*args32));
4994 static long btrfs_ioctl_set_received_subvol(struct file *file,
4997 struct btrfs_ioctl_received_subvol_args *sa = NULL;
5000 sa = memdup_user(arg, sizeof(*sa));
5007 ret = _btrfs_ioctl_set_received_subvol(file, sa);
5012 ret = copy_to_user(arg, sa, sizeof(*sa));
5021 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
5023 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5026 char label[BTRFS_LABEL_SIZE];
5028 spin_lock(&root->fs_info->super_lock);
5029 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
5030 spin_unlock(&root->fs_info->super_lock);
5032 len = strnlen(label, BTRFS_LABEL_SIZE);
5034 if (len == BTRFS_LABEL_SIZE) {
5035 btrfs_warn(root->fs_info,
5036 "label is too long, return the first %zu bytes", --len);
5039 ret = copy_to_user(arg, label, len);
5041 return ret ? -EFAULT : 0;
5044 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5046 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5047 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5048 struct btrfs_trans_handle *trans;
5049 char label[BTRFS_LABEL_SIZE];
5052 if (!capable(CAP_SYS_ADMIN))
5055 if (copy_from_user(label, arg, sizeof(label)))
5058 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
5059 btrfs_err(root->fs_info, "unable to set label with more than %d bytes",
5060 BTRFS_LABEL_SIZE - 1);
5064 ret = mnt_want_write_file(file);
5068 trans = btrfs_start_transaction(root, 0);
5069 if (IS_ERR(trans)) {
5070 ret = PTR_ERR(trans);
5074 spin_lock(&root->fs_info->super_lock);
5075 strcpy(super_block->label, label);
5076 spin_unlock(&root->fs_info->super_lock);
5077 ret = btrfs_commit_transaction(trans, root);
5080 mnt_drop_write_file(file);
5084 #define INIT_FEATURE_FLAGS(suffix) \
5085 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5086 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5087 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5089 static int btrfs_ioctl_get_supported_features(struct file *file,
5092 static struct btrfs_ioctl_feature_flags features[3] = {
5093 INIT_FEATURE_FLAGS(SUPP),
5094 INIT_FEATURE_FLAGS(SAFE_SET),
5095 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5098 if (copy_to_user(arg, &features, sizeof(features)))
5104 static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5106 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5107 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5108 struct btrfs_ioctl_feature_flags features;
5110 features.compat_flags = btrfs_super_compat_flags(super_block);
5111 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5112 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5114 if (copy_to_user(arg, &features, sizeof(features)))
5120 static int check_feature_bits(struct btrfs_root *root,
5121 enum btrfs_feature_set set,
5122 u64 change_mask, u64 flags, u64 supported_flags,
5123 u64 safe_set, u64 safe_clear)
5125 const char *type = btrfs_feature_set_names[set];
5127 u64 disallowed, unsupported;
5128 u64 set_mask = flags & change_mask;
5129 u64 clear_mask = ~flags & change_mask;
5131 unsupported = set_mask & ~supported_flags;
5133 names = btrfs_printable_features(set, unsupported);
5135 btrfs_warn(root->fs_info,
5136 "this kernel does not support the %s feature bit%s",
5137 names, strchr(names, ',') ? "s" : "");
5140 btrfs_warn(root->fs_info,
5141 "this kernel does not support %s bits 0x%llx",
5146 disallowed = set_mask & ~safe_set;
5148 names = btrfs_printable_features(set, disallowed);
5150 btrfs_warn(root->fs_info,
5151 "can't set the %s feature bit%s while mounted",
5152 names, strchr(names, ',') ? "s" : "");
5155 btrfs_warn(root->fs_info,
5156 "can't set %s bits 0x%llx while mounted",
5161 disallowed = clear_mask & ~safe_clear;
5163 names = btrfs_printable_features(set, disallowed);
5165 btrfs_warn(root->fs_info,
5166 "can't clear the %s feature bit%s while mounted",
5167 names, strchr(names, ',') ? "s" : "");
5170 btrfs_warn(root->fs_info,
5171 "can't clear %s bits 0x%llx while mounted",
5179 #define check_feature(root, change_mask, flags, mask_base) \
5180 check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
5181 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5182 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5183 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5185 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5187 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5188 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5189 struct btrfs_ioctl_feature_flags flags[2];
5190 struct btrfs_trans_handle *trans;
5194 if (!capable(CAP_SYS_ADMIN))
5197 if (copy_from_user(flags, arg, sizeof(flags)))
5201 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5202 !flags[0].incompat_flags)
5205 ret = check_feature(root, flags[0].compat_flags,
5206 flags[1].compat_flags, COMPAT);
5210 ret = check_feature(root, flags[0].compat_ro_flags,
5211 flags[1].compat_ro_flags, COMPAT_RO);
5215 ret = check_feature(root, flags[0].incompat_flags,
5216 flags[1].incompat_flags, INCOMPAT);
5220 trans = btrfs_start_transaction(root, 0);
5222 return PTR_ERR(trans);
5224 spin_lock(&root->fs_info->super_lock);
5225 newflags = btrfs_super_compat_flags(super_block);
5226 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5227 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5228 btrfs_set_super_compat_flags(super_block, newflags);
5230 newflags = btrfs_super_compat_ro_flags(super_block);
5231 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5232 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5233 btrfs_set_super_compat_ro_flags(super_block, newflags);
5235 newflags = btrfs_super_incompat_flags(super_block);
5236 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5237 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5238 btrfs_set_super_incompat_flags(super_block, newflags);
5239 spin_unlock(&root->fs_info->super_lock);
5241 return btrfs_commit_transaction(trans, root);
5244 long btrfs_ioctl(struct file *file, unsigned int
5245 cmd, unsigned long arg)
5247 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5248 void __user *argp = (void __user *)arg;
5251 case FS_IOC_GETFLAGS:
5252 return btrfs_ioctl_getflags(file, argp);
5253 case FS_IOC_SETFLAGS:
5254 return btrfs_ioctl_setflags(file, argp);
5255 case FS_IOC_GETVERSION:
5256 return btrfs_ioctl_getversion(file, argp);
5258 return btrfs_ioctl_fitrim(file, argp);
5259 case BTRFS_IOC_SNAP_CREATE:
5260 return btrfs_ioctl_snap_create(file, argp, 0);
5261 case BTRFS_IOC_SNAP_CREATE_V2:
5262 return btrfs_ioctl_snap_create_v2(file, argp, 0);
5263 case BTRFS_IOC_SUBVOL_CREATE:
5264 return btrfs_ioctl_snap_create(file, argp, 1);
5265 case BTRFS_IOC_SUBVOL_CREATE_V2:
5266 return btrfs_ioctl_snap_create_v2(file, argp, 1);
5267 case BTRFS_IOC_SNAP_DESTROY:
5268 return btrfs_ioctl_snap_destroy(file, argp);
5269 case BTRFS_IOC_SUBVOL_GETFLAGS:
5270 return btrfs_ioctl_subvol_getflags(file, argp);
5271 case BTRFS_IOC_SUBVOL_SETFLAGS:
5272 return btrfs_ioctl_subvol_setflags(file, argp);
5273 case BTRFS_IOC_DEFAULT_SUBVOL:
5274 return btrfs_ioctl_default_subvol(file, argp);
5275 case BTRFS_IOC_DEFRAG:
5276 return btrfs_ioctl_defrag(file, NULL);
5277 case BTRFS_IOC_DEFRAG_RANGE:
5278 return btrfs_ioctl_defrag(file, argp);
5279 case BTRFS_IOC_RESIZE:
5280 return btrfs_ioctl_resize(file, argp);
5281 case BTRFS_IOC_ADD_DEV:
5282 return btrfs_ioctl_add_dev(root, argp);
5283 case BTRFS_IOC_RM_DEV:
5284 return btrfs_ioctl_rm_dev(file, argp);
5285 case BTRFS_IOC_FS_INFO:
5286 return btrfs_ioctl_fs_info(root, argp);
5287 case BTRFS_IOC_DEV_INFO:
5288 return btrfs_ioctl_dev_info(root, argp);
5289 case BTRFS_IOC_BALANCE:
5290 return btrfs_ioctl_balance(file, NULL);
5291 case BTRFS_IOC_CLONE:
5292 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
5293 case BTRFS_IOC_CLONE_RANGE:
5294 return btrfs_ioctl_clone_range(file, argp);
5295 case BTRFS_IOC_TRANS_START:
5296 return btrfs_ioctl_trans_start(file);
5297 case BTRFS_IOC_TRANS_END:
5298 return btrfs_ioctl_trans_end(file);
5299 case BTRFS_IOC_TREE_SEARCH:
5300 return btrfs_ioctl_tree_search(file, argp);
5301 case BTRFS_IOC_TREE_SEARCH_V2:
5302 return btrfs_ioctl_tree_search_v2(file, argp);
5303 case BTRFS_IOC_INO_LOOKUP:
5304 return btrfs_ioctl_ino_lookup(file, argp);
5305 case BTRFS_IOC_INO_PATHS:
5306 return btrfs_ioctl_ino_to_path(root, argp);
5307 case BTRFS_IOC_LOGICAL_INO:
5308 return btrfs_ioctl_logical_to_ino(root, argp);
5309 case BTRFS_IOC_SPACE_INFO:
5310 return btrfs_ioctl_space_info(root, argp);
5311 case BTRFS_IOC_SYNC: {
5314 ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1);
5317 ret = btrfs_sync_fs(file->f_dentry->d_sb, 1);
5319 * The transaction thread may want to do more work,
5320 * namely it pokes the cleaner ktread that will start
5321 * processing uncleaned subvols.
5323 wake_up_process(root->fs_info->transaction_kthread);
5326 case BTRFS_IOC_START_SYNC:
5327 return btrfs_ioctl_start_sync(root, argp);
5328 case BTRFS_IOC_WAIT_SYNC:
5329 return btrfs_ioctl_wait_sync(root, argp);
5330 case BTRFS_IOC_SCRUB:
5331 return btrfs_ioctl_scrub(file, argp);
5332 case BTRFS_IOC_SCRUB_CANCEL:
5333 return btrfs_ioctl_scrub_cancel(root, argp);
5334 case BTRFS_IOC_SCRUB_PROGRESS:
5335 return btrfs_ioctl_scrub_progress(root, argp);
5336 case BTRFS_IOC_BALANCE_V2:
5337 return btrfs_ioctl_balance(file, argp);
5338 case BTRFS_IOC_BALANCE_CTL:
5339 return btrfs_ioctl_balance_ctl(root, arg);
5340 case BTRFS_IOC_BALANCE_PROGRESS:
5341 return btrfs_ioctl_balance_progress(root, argp);
5342 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5343 return btrfs_ioctl_set_received_subvol(file, argp);
5345 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5346 return btrfs_ioctl_set_received_subvol_32(file, argp);
5348 case BTRFS_IOC_SEND:
5349 return btrfs_ioctl_send(file, argp);
5350 case BTRFS_IOC_GET_DEV_STATS:
5351 return btrfs_ioctl_get_dev_stats(root, argp);
5352 case BTRFS_IOC_QUOTA_CTL:
5353 return btrfs_ioctl_quota_ctl(file, argp);
5354 case BTRFS_IOC_QGROUP_ASSIGN:
5355 return btrfs_ioctl_qgroup_assign(file, argp);
5356 case BTRFS_IOC_QGROUP_CREATE:
5357 return btrfs_ioctl_qgroup_create(file, argp);
5358 case BTRFS_IOC_QGROUP_LIMIT:
5359 return btrfs_ioctl_qgroup_limit(file, argp);
5360 case BTRFS_IOC_QUOTA_RESCAN:
5361 return btrfs_ioctl_quota_rescan(file, argp);
5362 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5363 return btrfs_ioctl_quota_rescan_status(file, argp);
5364 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5365 return btrfs_ioctl_quota_rescan_wait(file, argp);
5366 case BTRFS_IOC_DEV_REPLACE:
5367 return btrfs_ioctl_dev_replace(root, argp);
5368 case BTRFS_IOC_GET_FSLABEL:
5369 return btrfs_ioctl_get_fslabel(file, argp);
5370 case BTRFS_IOC_SET_FSLABEL:
5371 return btrfs_ioctl_set_fslabel(file, argp);
5372 case BTRFS_IOC_FILE_EXTENT_SAME:
5373 return btrfs_ioctl_file_extent_same(file, argp);
5374 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5375 return btrfs_ioctl_get_supported_features(file, argp);
5376 case BTRFS_IOC_GET_FEATURES:
5377 return btrfs_ioctl_get_features(file, argp);
5378 case BTRFS_IOC_SET_FEATURES:
5379 return btrfs_ioctl_set_features(file, argp);