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 may_delete in fs/namei.c()
769 * Check whether we can remove a link victim from directory dir, check
770 * whether the type of victim is right.
771 * 1. We can't do it if dir is read-only (done in permission())
772 * 2. We should have write and exec permissions on dir
773 * 3. We can't remove anything from append-only dir
774 * 4. We can't do anything with immutable dir (done in permission())
775 * 5. If the sticky bit on dir is set we should either
776 * a. be owner of dir, or
777 * b. be owner of victim, or
778 * c. have CAP_FOWNER capability
779 * 6. If the victim is append-only or immutable we can't do antyhing with
780 * links pointing to it.
781 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
782 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
783 * 9. We can't remove a root or mountpoint.
784 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
785 * nfs_async_unlink().
788 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
792 if (!victim->d_inode)
795 BUG_ON(victim->d_parent->d_inode != dir);
796 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
798 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
803 if (check_sticky(dir, victim->d_inode) || IS_APPEND(victim->d_inode) ||
804 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
807 if (!S_ISDIR(victim->d_inode->i_mode))
811 } else if (S_ISDIR(victim->d_inode->i_mode))
815 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
820 /* copy of may_create in fs/namei.c() */
821 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
827 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
831 * Create a new subvolume below @parent. This is largely modeled after
832 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
833 * inside this filesystem so it's quite a bit simpler.
835 static noinline int btrfs_mksubvol(struct path *parent,
836 char *name, int namelen,
837 struct btrfs_root *snap_src,
838 u64 *async_transid, bool readonly,
839 struct btrfs_qgroup_inherit *inherit)
841 struct inode *dir = parent->dentry->d_inode;
842 struct dentry *dentry;
845 error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
849 dentry = lookup_one_len(name, parent->dentry, namelen);
850 error = PTR_ERR(dentry);
858 error = btrfs_may_create(dir, dentry);
863 * even if this name doesn't exist, we may get hash collisions.
864 * check for them now when we can safely fail
866 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
872 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
874 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
878 error = create_snapshot(snap_src, dir, dentry, name, namelen,
879 async_transid, readonly, inherit);
881 error = create_subvol(dir, dentry, name, namelen,
882 async_transid, inherit);
885 fsnotify_mkdir(dir, dentry);
887 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
891 mutex_unlock(&dir->i_mutex);
896 * When we're defragging a range, we don't want to kick it off again
897 * if it is really just waiting for delalloc to send it down.
898 * If we find a nice big extent or delalloc range for the bytes in the
899 * file you want to defrag, we return 0 to let you know to skip this
902 static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
904 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
905 struct extent_map *em = NULL;
906 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
909 read_lock(&em_tree->lock);
910 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
911 read_unlock(&em_tree->lock);
914 end = extent_map_end(em);
916 if (end - offset > thresh)
919 /* if we already have a nice delalloc here, just stop */
921 end = count_range_bits(io_tree, &offset, offset + thresh,
922 thresh, EXTENT_DELALLOC, 1);
929 * helper function to walk through a file and find extents
930 * newer than a specific transid, and smaller than thresh.
932 * This is used by the defragging code to find new and small
935 static int find_new_extents(struct btrfs_root *root,
936 struct inode *inode, u64 newer_than,
937 u64 *off, u32 thresh)
939 struct btrfs_path *path;
940 struct btrfs_key min_key;
941 struct extent_buffer *leaf;
942 struct btrfs_file_extent_item *extent;
945 u64 ino = btrfs_ino(inode);
947 path = btrfs_alloc_path();
951 min_key.objectid = ino;
952 min_key.type = BTRFS_EXTENT_DATA_KEY;
953 min_key.offset = *off;
956 ret = btrfs_search_forward(root, &min_key, path, newer_than);
960 if (min_key.objectid != ino)
962 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
965 leaf = path->nodes[0];
966 extent = btrfs_item_ptr(leaf, path->slots[0],
967 struct btrfs_file_extent_item);
969 type = btrfs_file_extent_type(leaf, extent);
970 if (type == BTRFS_FILE_EXTENT_REG &&
971 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
972 check_defrag_in_cache(inode, min_key.offset, thresh)) {
973 *off = min_key.offset;
974 btrfs_free_path(path);
979 if (path->slots[0] < btrfs_header_nritems(leaf)) {
980 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
984 if (min_key.offset == (u64)-1)
988 btrfs_release_path(path);
991 btrfs_free_path(path);
995 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
997 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
998 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
999 struct extent_map *em;
1000 u64 len = PAGE_CACHE_SIZE;
1003 * hopefully we have this extent in the tree already, try without
1004 * the full extent lock
1006 read_lock(&em_tree->lock);
1007 em = lookup_extent_mapping(em_tree, start, len);
1008 read_unlock(&em_tree->lock);
1011 struct extent_state *cached = NULL;
1012 u64 end = start + len - 1;
1014 /* get the big lock and read metadata off disk */
1015 lock_extent_bits(io_tree, start, end, 0, &cached);
1016 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
1017 unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS);
1026 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1028 struct extent_map *next;
1031 /* this is the last extent */
1032 if (em->start + em->len >= i_size_read(inode))
1035 next = defrag_lookup_extent(inode, em->start + em->len);
1036 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1038 else if ((em->block_start + em->block_len == next->block_start) &&
1039 (em->block_len > 128 * 1024 && next->block_len > 128 * 1024))
1042 free_extent_map(next);
1046 static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
1047 u64 *last_len, u64 *skip, u64 *defrag_end,
1050 struct extent_map *em;
1052 bool next_mergeable = true;
1055 * make sure that once we start defragging an extent, we keep on
1058 if (start < *defrag_end)
1063 em = defrag_lookup_extent(inode, start);
1067 /* this will cover holes, and inline extents */
1068 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1073 next_mergeable = defrag_check_next_extent(inode, em);
1075 * we hit a real extent, if it is big or the next extent is not a
1076 * real extent, don't bother defragging it
1078 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1079 (em->len >= thresh || !next_mergeable))
1083 * last_len ends up being a counter of how many bytes we've defragged.
1084 * every time we choose not to defrag an extent, we reset *last_len
1085 * so that the next tiny extent will force a defrag.
1087 * The end result of this is that tiny extents before a single big
1088 * extent will force at least part of that big extent to be defragged.
1091 *defrag_end = extent_map_end(em);
1094 *skip = extent_map_end(em);
1098 free_extent_map(em);
1103 * it doesn't do much good to defrag one or two pages
1104 * at a time. This pulls in a nice chunk of pages
1105 * to COW and defrag.
1107 * It also makes sure the delalloc code has enough
1108 * dirty data to avoid making new small extents as part
1111 * It's a good idea to start RA on this range
1112 * before calling this.
1114 static int cluster_pages_for_defrag(struct inode *inode,
1115 struct page **pages,
1116 unsigned long start_index,
1117 unsigned long num_pages)
1119 unsigned long file_end;
1120 u64 isize = i_size_read(inode);
1127 struct btrfs_ordered_extent *ordered;
1128 struct extent_state *cached_state = NULL;
1129 struct extent_io_tree *tree;
1130 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1132 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1133 if (!isize || start_index > file_end)
1136 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1138 ret = btrfs_delalloc_reserve_space(inode,
1139 page_cnt << PAGE_CACHE_SHIFT);
1143 tree = &BTRFS_I(inode)->io_tree;
1145 /* step one, lock all the pages */
1146 for (i = 0; i < page_cnt; i++) {
1149 page = find_or_create_page(inode->i_mapping,
1150 start_index + i, mask);
1154 page_start = page_offset(page);
1155 page_end = page_start + PAGE_CACHE_SIZE - 1;
1157 lock_extent_bits(tree, page_start, page_end,
1159 ordered = btrfs_lookup_ordered_extent(inode,
1161 unlock_extent_cached(tree, page_start, page_end,
1162 &cached_state, GFP_NOFS);
1167 btrfs_start_ordered_extent(inode, ordered, 1);
1168 btrfs_put_ordered_extent(ordered);
1171 * we unlocked the page above, so we need check if
1172 * it was released or not.
1174 if (page->mapping != inode->i_mapping) {
1176 page_cache_release(page);
1181 if (!PageUptodate(page)) {
1182 btrfs_readpage(NULL, page);
1184 if (!PageUptodate(page)) {
1186 page_cache_release(page);
1192 if (page->mapping != inode->i_mapping) {
1194 page_cache_release(page);
1204 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1208 * so now we have a nice long stream of locked
1209 * and up to date pages, lets wait on them
1211 for (i = 0; i < i_done; i++)
1212 wait_on_page_writeback(pages[i]);
1214 page_start = page_offset(pages[0]);
1215 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1217 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1218 page_start, page_end - 1, 0, &cached_state);
1219 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1220 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1221 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1222 &cached_state, GFP_NOFS);
1224 if (i_done != page_cnt) {
1225 spin_lock(&BTRFS_I(inode)->lock);
1226 BTRFS_I(inode)->outstanding_extents++;
1227 spin_unlock(&BTRFS_I(inode)->lock);
1228 btrfs_delalloc_release_space(inode,
1229 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1233 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1234 &cached_state, GFP_NOFS);
1236 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1237 page_start, page_end - 1, &cached_state,
1240 for (i = 0; i < i_done; i++) {
1241 clear_page_dirty_for_io(pages[i]);
1242 ClearPageChecked(pages[i]);
1243 set_page_extent_mapped(pages[i]);
1244 set_page_dirty(pages[i]);
1245 unlock_page(pages[i]);
1246 page_cache_release(pages[i]);
1250 for (i = 0; i < i_done; i++) {
1251 unlock_page(pages[i]);
1252 page_cache_release(pages[i]);
1254 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1259 int btrfs_defrag_file(struct inode *inode, struct file *file,
1260 struct btrfs_ioctl_defrag_range_args *range,
1261 u64 newer_than, unsigned long max_to_defrag)
1263 struct btrfs_root *root = BTRFS_I(inode)->root;
1264 struct file_ra_state *ra = NULL;
1265 unsigned long last_index;
1266 u64 isize = i_size_read(inode);
1270 u64 newer_off = range->start;
1272 unsigned long ra_index = 0;
1274 int defrag_count = 0;
1275 int compress_type = BTRFS_COMPRESS_ZLIB;
1276 u32 extent_thresh = range->extent_thresh;
1277 unsigned long max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1278 unsigned long cluster = max_cluster;
1279 u64 new_align = ~((u64)128 * 1024 - 1);
1280 struct page **pages = NULL;
1285 if (range->start >= isize)
1288 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1289 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1291 if (range->compress_type)
1292 compress_type = range->compress_type;
1295 if (extent_thresh == 0)
1296 extent_thresh = 256 * 1024;
1299 * if we were not given a file, allocate a readahead
1303 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1306 file_ra_state_init(ra, inode->i_mapping);
1311 pages = kmalloc_array(max_cluster, sizeof(struct page *),
1318 /* find the last page to defrag */
1319 if (range->start + range->len > range->start) {
1320 last_index = min_t(u64, isize - 1,
1321 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1323 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1327 ret = find_new_extents(root, inode, newer_than,
1328 &newer_off, 64 * 1024);
1330 range->start = newer_off;
1332 * we always align our defrag to help keep
1333 * the extents in the file evenly spaced
1335 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1339 i = range->start >> PAGE_CACHE_SHIFT;
1342 max_to_defrag = last_index + 1;
1345 * make writeback starts from i, so the defrag range can be
1346 * written sequentially.
1348 if (i < inode->i_mapping->writeback_index)
1349 inode->i_mapping->writeback_index = i;
1351 while (i <= last_index && defrag_count < max_to_defrag &&
1352 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_CACHE_SIZE))) {
1354 * make sure we stop running if someone unmounts
1357 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1360 if (btrfs_defrag_cancelled(root->fs_info)) {
1361 printk(KERN_DEBUG "BTRFS: defrag_file cancelled\n");
1366 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1367 extent_thresh, &last_len, &skip,
1368 &defrag_end, range->flags &
1369 BTRFS_DEFRAG_RANGE_COMPRESS)) {
1372 * the should_defrag function tells us how much to skip
1373 * bump our counter by the suggested amount
1375 next = DIV_ROUND_UP(skip, PAGE_CACHE_SIZE);
1376 i = max(i + 1, next);
1381 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1382 PAGE_CACHE_SHIFT) - i;
1383 cluster = min(cluster, max_cluster);
1385 cluster = max_cluster;
1388 if (i + cluster > ra_index) {
1389 ra_index = max(i, ra_index);
1390 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1392 ra_index += max_cluster;
1395 mutex_lock(&inode->i_mutex);
1396 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1397 BTRFS_I(inode)->force_compress = compress_type;
1398 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1400 mutex_unlock(&inode->i_mutex);
1404 defrag_count += ret;
1405 balance_dirty_pages_ratelimited(inode->i_mapping);
1406 mutex_unlock(&inode->i_mutex);
1409 if (newer_off == (u64)-1)
1415 newer_off = max(newer_off + 1,
1416 (u64)i << PAGE_CACHE_SHIFT);
1418 ret = find_new_extents(root, inode,
1419 newer_than, &newer_off,
1422 range->start = newer_off;
1423 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1430 last_len += ret << PAGE_CACHE_SHIFT;
1438 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1439 filemap_flush(inode->i_mapping);
1440 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1441 &BTRFS_I(inode)->runtime_flags))
1442 filemap_flush(inode->i_mapping);
1445 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1446 /* the filemap_flush will queue IO into the worker threads, but
1447 * we have to make sure the IO is actually started and that
1448 * ordered extents get created before we return
1450 atomic_inc(&root->fs_info->async_submit_draining);
1451 while (atomic_read(&root->fs_info->nr_async_submits) ||
1452 atomic_read(&root->fs_info->async_delalloc_pages)) {
1453 wait_event(root->fs_info->async_submit_wait,
1454 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1455 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1457 atomic_dec(&root->fs_info->async_submit_draining);
1460 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1461 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1467 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1468 mutex_lock(&inode->i_mutex);
1469 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1470 mutex_unlock(&inode->i_mutex);
1478 static noinline int btrfs_ioctl_resize(struct file *file,
1484 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
1485 struct btrfs_ioctl_vol_args *vol_args;
1486 struct btrfs_trans_handle *trans;
1487 struct btrfs_device *device = NULL;
1490 char *devstr = NULL;
1494 if (!capable(CAP_SYS_ADMIN))
1497 ret = mnt_want_write_file(file);
1501 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1503 mnt_drop_write_file(file);
1504 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1507 mutex_lock(&root->fs_info->volume_mutex);
1508 vol_args = memdup_user(arg, sizeof(*vol_args));
1509 if (IS_ERR(vol_args)) {
1510 ret = PTR_ERR(vol_args);
1514 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1516 sizestr = vol_args->name;
1517 devstr = strchr(sizestr, ':');
1519 sizestr = devstr + 1;
1521 devstr = vol_args->name;
1522 ret = kstrtoull(devstr, 10, &devid);
1529 btrfs_info(root->fs_info, "resizing devid %llu", devid);
1532 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
1534 btrfs_info(root->fs_info, "resizer unable to find device %llu",
1540 if (!device->writeable) {
1541 btrfs_info(root->fs_info,
1542 "resizer unable to apply on readonly device %llu",
1548 if (!strcmp(sizestr, "max"))
1549 new_size = device->bdev->bd_inode->i_size;
1551 if (sizestr[0] == '-') {
1554 } else if (sizestr[0] == '+') {
1558 new_size = memparse(sizestr, &retptr);
1559 if (*retptr != '\0' || new_size == 0) {
1565 if (device->is_tgtdev_for_dev_replace) {
1570 old_size = btrfs_device_get_total_bytes(device);
1573 if (new_size > old_size) {
1577 new_size = old_size - new_size;
1578 } else if (mod > 0) {
1579 if (new_size > ULLONG_MAX - old_size) {
1583 new_size = old_size + new_size;
1586 if (new_size < 256 * 1024 * 1024) {
1590 if (new_size > device->bdev->bd_inode->i_size) {
1595 do_div(new_size, root->sectorsize);
1596 new_size *= root->sectorsize;
1598 printk_in_rcu(KERN_INFO "BTRFS: new size for %s is %llu\n",
1599 rcu_str_deref(device->name), new_size);
1601 if (new_size > old_size) {
1602 trans = btrfs_start_transaction(root, 0);
1603 if (IS_ERR(trans)) {
1604 ret = PTR_ERR(trans);
1607 ret = btrfs_grow_device(trans, device, new_size);
1608 btrfs_commit_transaction(trans, root);
1609 } else if (new_size < old_size) {
1610 ret = btrfs_shrink_device(device, new_size);
1611 } /* equal, nothing need to do */
1616 mutex_unlock(&root->fs_info->volume_mutex);
1617 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
1618 mnt_drop_write_file(file);
1622 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1623 char *name, unsigned long fd, int subvol,
1624 u64 *transid, bool readonly,
1625 struct btrfs_qgroup_inherit *inherit)
1630 ret = mnt_want_write_file(file);
1634 namelen = strlen(name);
1635 if (strchr(name, '/')) {
1637 goto out_drop_write;
1640 if (name[0] == '.' &&
1641 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1643 goto out_drop_write;
1647 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1648 NULL, transid, readonly, inherit);
1650 struct fd src = fdget(fd);
1651 struct inode *src_inode;
1654 goto out_drop_write;
1657 src_inode = file_inode(src.file);
1658 if (src_inode->i_sb != file_inode(file)->i_sb) {
1659 btrfs_info(BTRFS_I(src_inode)->root->fs_info,
1660 "Snapshot src from another FS");
1662 } else if (!inode_owner_or_capable(src_inode)) {
1664 * Subvolume creation is not restricted, but snapshots
1665 * are limited to own subvolumes only
1669 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1670 BTRFS_I(src_inode)->root,
1671 transid, readonly, inherit);
1676 mnt_drop_write_file(file);
1681 static noinline int btrfs_ioctl_snap_create(struct file *file,
1682 void __user *arg, int subvol)
1684 struct btrfs_ioctl_vol_args *vol_args;
1687 vol_args = memdup_user(arg, sizeof(*vol_args));
1688 if (IS_ERR(vol_args))
1689 return PTR_ERR(vol_args);
1690 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1692 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1693 vol_args->fd, subvol,
1700 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1701 void __user *arg, int subvol)
1703 struct btrfs_ioctl_vol_args_v2 *vol_args;
1707 bool readonly = false;
1708 struct btrfs_qgroup_inherit *inherit = NULL;
1710 vol_args = memdup_user(arg, sizeof(*vol_args));
1711 if (IS_ERR(vol_args))
1712 return PTR_ERR(vol_args);
1713 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1715 if (vol_args->flags &
1716 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1717 BTRFS_SUBVOL_QGROUP_INHERIT)) {
1722 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1724 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1726 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1727 if (vol_args->size > PAGE_CACHE_SIZE) {
1731 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1732 if (IS_ERR(inherit)) {
1733 ret = PTR_ERR(inherit);
1738 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1739 vol_args->fd, subvol, ptr,
1744 if (ptr && copy_to_user(arg +
1745 offsetof(struct btrfs_ioctl_vol_args_v2,
1757 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1760 struct inode *inode = file_inode(file);
1761 struct btrfs_root *root = BTRFS_I(inode)->root;
1765 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1768 down_read(&root->fs_info->subvol_sem);
1769 if (btrfs_root_readonly(root))
1770 flags |= BTRFS_SUBVOL_RDONLY;
1771 up_read(&root->fs_info->subvol_sem);
1773 if (copy_to_user(arg, &flags, sizeof(flags)))
1779 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1782 struct inode *inode = file_inode(file);
1783 struct btrfs_root *root = BTRFS_I(inode)->root;
1784 struct btrfs_trans_handle *trans;
1789 if (!inode_owner_or_capable(inode))
1792 ret = mnt_want_write_file(file);
1796 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1798 goto out_drop_write;
1801 if (copy_from_user(&flags, arg, sizeof(flags))) {
1803 goto out_drop_write;
1806 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1808 goto out_drop_write;
1811 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1813 goto out_drop_write;
1816 down_write(&root->fs_info->subvol_sem);
1819 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1822 root_flags = btrfs_root_flags(&root->root_item);
1823 if (flags & BTRFS_SUBVOL_RDONLY) {
1824 btrfs_set_root_flags(&root->root_item,
1825 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1828 * Block RO -> RW transition if this subvolume is involved in
1831 spin_lock(&root->root_item_lock);
1832 if (root->send_in_progress == 0) {
1833 btrfs_set_root_flags(&root->root_item,
1834 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1835 spin_unlock(&root->root_item_lock);
1837 spin_unlock(&root->root_item_lock);
1838 btrfs_warn(root->fs_info,
1839 "Attempt to set subvolume %llu read-write during send",
1840 root->root_key.objectid);
1846 trans = btrfs_start_transaction(root, 1);
1847 if (IS_ERR(trans)) {
1848 ret = PTR_ERR(trans);
1852 ret = btrfs_update_root(trans, root->fs_info->tree_root,
1853 &root->root_key, &root->root_item);
1855 btrfs_commit_transaction(trans, root);
1858 btrfs_set_root_flags(&root->root_item, root_flags);
1860 up_write(&root->fs_info->subvol_sem);
1862 mnt_drop_write_file(file);
1868 * helper to check if the subvolume references other subvolumes
1870 static noinline int may_destroy_subvol(struct btrfs_root *root)
1872 struct btrfs_path *path;
1873 struct btrfs_dir_item *di;
1874 struct btrfs_key key;
1878 path = btrfs_alloc_path();
1882 /* Make sure this root isn't set as the default subvol */
1883 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1884 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1885 dir_id, "default", 7, 0);
1886 if (di && !IS_ERR(di)) {
1887 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1888 if (key.objectid == root->root_key.objectid) {
1890 btrfs_err(root->fs_info, "deleting default subvolume "
1891 "%llu is not allowed", key.objectid);
1894 btrfs_release_path(path);
1897 key.objectid = root->root_key.objectid;
1898 key.type = BTRFS_ROOT_REF_KEY;
1899 key.offset = (u64)-1;
1901 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1908 if (path->slots[0] > 0) {
1910 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1911 if (key.objectid == root->root_key.objectid &&
1912 key.type == BTRFS_ROOT_REF_KEY)
1916 btrfs_free_path(path);
1920 static noinline int key_in_sk(struct btrfs_key *key,
1921 struct btrfs_ioctl_search_key *sk)
1923 struct btrfs_key test;
1926 test.objectid = sk->min_objectid;
1927 test.type = sk->min_type;
1928 test.offset = sk->min_offset;
1930 ret = btrfs_comp_cpu_keys(key, &test);
1934 test.objectid = sk->max_objectid;
1935 test.type = sk->max_type;
1936 test.offset = sk->max_offset;
1938 ret = btrfs_comp_cpu_keys(key, &test);
1944 static noinline int copy_to_sk(struct btrfs_root *root,
1945 struct btrfs_path *path,
1946 struct btrfs_key *key,
1947 struct btrfs_ioctl_search_key *sk,
1950 unsigned long *sk_offset,
1954 struct extent_buffer *leaf;
1955 struct btrfs_ioctl_search_header sh;
1956 unsigned long item_off;
1957 unsigned long item_len;
1963 leaf = path->nodes[0];
1964 slot = path->slots[0];
1965 nritems = btrfs_header_nritems(leaf);
1967 if (btrfs_header_generation(leaf) > sk->max_transid) {
1971 found_transid = btrfs_header_generation(leaf);
1973 for (i = slot; i < nritems; i++) {
1974 item_off = btrfs_item_ptr_offset(leaf, i);
1975 item_len = btrfs_item_size_nr(leaf, i);
1977 btrfs_item_key_to_cpu(leaf, key, i);
1978 if (!key_in_sk(key, sk))
1981 if (sizeof(sh) + item_len > *buf_size) {
1988 * return one empty item back for v1, which does not
1992 *buf_size = sizeof(sh) + item_len;
1997 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
2002 sh.objectid = key->objectid;
2003 sh.offset = key->offset;
2004 sh.type = key->type;
2006 sh.transid = found_transid;
2008 /* copy search result header */
2009 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
2014 *sk_offset += sizeof(sh);
2017 char __user *up = ubuf + *sk_offset;
2019 if (read_extent_buffer_to_user(leaf, up,
2020 item_off, item_len)) {
2025 *sk_offset += item_len;
2029 if (ret) /* -EOVERFLOW from above */
2032 if (*num_found >= sk->nr_items) {
2039 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
2041 else if (key->type < (u8)-1 && key->type < sk->max_type) {
2044 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
2052 * 0: all items from this leaf copied, continue with next
2053 * 1: * more items can be copied, but unused buffer is too small
2054 * * all items were found
2055 * Either way, it will stops the loop which iterates to the next
2057 * -EOVERFLOW: item was to large for buffer
2058 * -EFAULT: could not copy extent buffer back to userspace
2063 static noinline int search_ioctl(struct inode *inode,
2064 struct btrfs_ioctl_search_key *sk,
2068 struct btrfs_root *root;
2069 struct btrfs_key key;
2070 struct btrfs_path *path;
2071 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
2074 unsigned long sk_offset = 0;
2076 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2077 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2081 path = btrfs_alloc_path();
2085 if (sk->tree_id == 0) {
2086 /* search the root of the inode that was passed */
2087 root = BTRFS_I(inode)->root;
2089 key.objectid = sk->tree_id;
2090 key.type = BTRFS_ROOT_ITEM_KEY;
2091 key.offset = (u64)-1;
2092 root = btrfs_read_fs_root_no_name(info, &key);
2094 printk(KERN_ERR "BTRFS: could not find root %llu\n",
2096 btrfs_free_path(path);
2101 key.objectid = sk->min_objectid;
2102 key.type = sk->min_type;
2103 key.offset = sk->min_offset;
2106 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2112 ret = copy_to_sk(root, path, &key, sk, buf_size, ubuf,
2113 &sk_offset, &num_found);
2114 btrfs_release_path(path);
2122 sk->nr_items = num_found;
2123 btrfs_free_path(path);
2127 static noinline int btrfs_ioctl_tree_search(struct file *file,
2130 struct btrfs_ioctl_search_args __user *uargs;
2131 struct btrfs_ioctl_search_key sk;
2132 struct inode *inode;
2136 if (!capable(CAP_SYS_ADMIN))
2139 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2141 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2144 buf_size = sizeof(uargs->buf);
2146 inode = file_inode(file);
2147 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2150 * In the origin implementation an overflow is handled by returning a
2151 * search header with a len of zero, so reset ret.
2153 if (ret == -EOVERFLOW)
2156 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2161 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2164 struct btrfs_ioctl_search_args_v2 __user *uarg;
2165 struct btrfs_ioctl_search_args_v2 args;
2166 struct inode *inode;
2169 const size_t buf_limit = 16 * 1024 * 1024;
2171 if (!capable(CAP_SYS_ADMIN))
2174 /* copy search header and buffer size */
2175 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2176 if (copy_from_user(&args, uarg, sizeof(args)))
2179 buf_size = args.buf_size;
2181 if (buf_size < sizeof(struct btrfs_ioctl_search_header))
2184 /* limit result size to 16MB */
2185 if (buf_size > buf_limit)
2186 buf_size = buf_limit;
2188 inode = file_inode(file);
2189 ret = search_ioctl(inode, &args.key, &buf_size,
2190 (char *)(&uarg->buf[0]));
2191 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2193 else if (ret == -EOVERFLOW &&
2194 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2201 * Search INODE_REFs to identify path name of 'dirid' directory
2202 * in a 'tree_id' tree. and sets path name to 'name'.
2204 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2205 u64 tree_id, u64 dirid, char *name)
2207 struct btrfs_root *root;
2208 struct btrfs_key key;
2214 struct btrfs_inode_ref *iref;
2215 struct extent_buffer *l;
2216 struct btrfs_path *path;
2218 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2223 path = btrfs_alloc_path();
2227 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
2229 key.objectid = tree_id;
2230 key.type = BTRFS_ROOT_ITEM_KEY;
2231 key.offset = (u64)-1;
2232 root = btrfs_read_fs_root_no_name(info, &key);
2234 printk(KERN_ERR "BTRFS: could not find root %llu\n", tree_id);
2239 key.objectid = dirid;
2240 key.type = BTRFS_INODE_REF_KEY;
2241 key.offset = (u64)-1;
2244 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2248 ret = btrfs_previous_item(root, path, dirid,
2249 BTRFS_INODE_REF_KEY);
2259 slot = path->slots[0];
2260 btrfs_item_key_to_cpu(l, &key, slot);
2262 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2263 len = btrfs_inode_ref_name_len(l, iref);
2265 total_len += len + 1;
2267 ret = -ENAMETOOLONG;
2272 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2274 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2277 btrfs_release_path(path);
2278 key.objectid = key.offset;
2279 key.offset = (u64)-1;
2280 dirid = key.objectid;
2282 memmove(name, ptr, total_len);
2283 name[total_len] = '\0';
2286 btrfs_free_path(path);
2290 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2293 struct btrfs_ioctl_ino_lookup_args *args;
2294 struct inode *inode;
2297 if (!capable(CAP_SYS_ADMIN))
2300 args = memdup_user(argp, sizeof(*args));
2302 return PTR_ERR(args);
2304 inode = file_inode(file);
2306 if (args->treeid == 0)
2307 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2309 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2310 args->treeid, args->objectid,
2313 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2320 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2323 struct dentry *parent = file->f_path.dentry;
2324 struct dentry *dentry;
2325 struct inode *dir = parent->d_inode;
2326 struct inode *inode;
2327 struct btrfs_root *root = BTRFS_I(dir)->root;
2328 struct btrfs_root *dest = NULL;
2329 struct btrfs_ioctl_vol_args *vol_args;
2330 struct btrfs_trans_handle *trans;
2331 struct btrfs_block_rsv block_rsv;
2333 u64 qgroup_reserved;
2338 vol_args = memdup_user(arg, sizeof(*vol_args));
2339 if (IS_ERR(vol_args))
2340 return PTR_ERR(vol_args);
2342 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2343 namelen = strlen(vol_args->name);
2344 if (strchr(vol_args->name, '/') ||
2345 strncmp(vol_args->name, "..", namelen) == 0) {
2350 err = mnt_want_write_file(file);
2355 err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2357 goto out_drop_write;
2358 dentry = lookup_one_len(vol_args->name, parent, namelen);
2359 if (IS_ERR(dentry)) {
2360 err = PTR_ERR(dentry);
2361 goto out_unlock_dir;
2364 if (!dentry->d_inode) {
2369 inode = dentry->d_inode;
2370 dest = BTRFS_I(inode)->root;
2371 if (!capable(CAP_SYS_ADMIN)) {
2373 * Regular user. Only allow this with a special mount
2374 * option, when the user has write+exec access to the
2375 * subvol root, and when rmdir(2) would have been
2378 * Note that this is _not_ check that the subvol is
2379 * empty or doesn't contain data that we wouldn't
2380 * otherwise be able to delete.
2382 * Users who want to delete empty subvols should try
2386 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2390 * Do not allow deletion if the parent dir is the same
2391 * as the dir to be deleted. That means the ioctl
2392 * must be called on the dentry referencing the root
2393 * of the subvol, not a random directory contained
2400 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2405 /* check if subvolume may be deleted by a user */
2406 err = btrfs_may_delete(dir, dentry, 1);
2410 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2415 mutex_lock(&inode->i_mutex);
2418 * Don't allow to delete a subvolume with send in progress. This is
2419 * inside the i_mutex so the error handling that has to drop the bit
2420 * again is not run concurrently.
2422 spin_lock(&dest->root_item_lock);
2423 root_flags = btrfs_root_flags(&dest->root_item);
2424 if (dest->send_in_progress == 0) {
2425 btrfs_set_root_flags(&dest->root_item,
2426 root_flags | BTRFS_ROOT_SUBVOL_DEAD);
2427 spin_unlock(&dest->root_item_lock);
2429 spin_unlock(&dest->root_item_lock);
2430 btrfs_warn(root->fs_info,
2431 "Attempt to delete subvolume %llu during send",
2432 dest->root_key.objectid);
2437 d_invalidate(dentry);
2439 down_write(&root->fs_info->subvol_sem);
2441 err = may_destroy_subvol(dest);
2445 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2447 * One for dir inode, two for dir entries, two for root
2450 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2451 5, &qgroup_reserved, true);
2455 trans = btrfs_start_transaction(root, 0);
2456 if (IS_ERR(trans)) {
2457 err = PTR_ERR(trans);
2460 trans->block_rsv = &block_rsv;
2461 trans->bytes_reserved = block_rsv.size;
2463 ret = btrfs_unlink_subvol(trans, root, dir,
2464 dest->root_key.objectid,
2465 dentry->d_name.name,
2466 dentry->d_name.len);
2469 btrfs_abort_transaction(trans, root, ret);
2473 btrfs_record_root_in_trans(trans, dest);
2475 memset(&dest->root_item.drop_progress, 0,
2476 sizeof(dest->root_item.drop_progress));
2477 dest->root_item.drop_level = 0;
2478 btrfs_set_root_refs(&dest->root_item, 0);
2480 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
2481 ret = btrfs_insert_orphan_item(trans,
2482 root->fs_info->tree_root,
2483 dest->root_key.objectid);
2485 btrfs_abort_transaction(trans, root, ret);
2491 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2492 dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
2493 dest->root_key.objectid);
2494 if (ret && ret != -ENOENT) {
2495 btrfs_abort_transaction(trans, root, ret);
2499 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2500 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2501 dest->root_item.received_uuid,
2502 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2503 dest->root_key.objectid);
2504 if (ret && ret != -ENOENT) {
2505 btrfs_abort_transaction(trans, root, ret);
2512 trans->block_rsv = NULL;
2513 trans->bytes_reserved = 0;
2514 ret = btrfs_end_transaction(trans, root);
2517 inode->i_flags |= S_DEAD;
2519 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
2521 up_write(&root->fs_info->subvol_sem);
2523 spin_lock(&dest->root_item_lock);
2524 root_flags = btrfs_root_flags(&dest->root_item);
2525 btrfs_set_root_flags(&dest->root_item,
2526 root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
2527 spin_unlock(&dest->root_item_lock);
2529 mutex_unlock(&inode->i_mutex);
2531 shrink_dcache_sb(root->fs_info->sb);
2532 btrfs_invalidate_inodes(dest);
2534 ASSERT(dest->send_in_progress == 0);
2537 if (dest->ino_cache_inode) {
2538 iput(dest->ino_cache_inode);
2539 dest->ino_cache_inode = NULL;
2545 mutex_unlock(&dir->i_mutex);
2547 mnt_drop_write_file(file);
2553 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2555 struct inode *inode = file_inode(file);
2556 struct btrfs_root *root = BTRFS_I(inode)->root;
2557 struct btrfs_ioctl_defrag_range_args *range;
2560 ret = mnt_want_write_file(file);
2564 if (btrfs_root_readonly(root)) {
2569 switch (inode->i_mode & S_IFMT) {
2571 if (!capable(CAP_SYS_ADMIN)) {
2575 ret = btrfs_defrag_root(root);
2578 ret = btrfs_defrag_root(root->fs_info->extent_root);
2581 if (!(file->f_mode & FMODE_WRITE)) {
2586 range = kzalloc(sizeof(*range), GFP_KERNEL);
2593 if (copy_from_user(range, argp,
2599 /* compression requires us to start the IO */
2600 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2601 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2602 range->extent_thresh = (u32)-1;
2605 /* the rest are all set to zero by kzalloc */
2606 range->len = (u64)-1;
2608 ret = btrfs_defrag_file(file_inode(file), file,
2618 mnt_drop_write_file(file);
2622 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2624 struct btrfs_ioctl_vol_args *vol_args;
2627 if (!capable(CAP_SYS_ADMIN))
2630 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2632 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2635 mutex_lock(&root->fs_info->volume_mutex);
2636 vol_args = memdup_user(arg, sizeof(*vol_args));
2637 if (IS_ERR(vol_args)) {
2638 ret = PTR_ERR(vol_args);
2642 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2643 ret = btrfs_init_new_device(root, vol_args->name);
2646 btrfs_info(root->fs_info, "disk added %s",vol_args->name);
2650 mutex_unlock(&root->fs_info->volume_mutex);
2651 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2655 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2657 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2658 struct btrfs_ioctl_vol_args *vol_args;
2661 if (!capable(CAP_SYS_ADMIN))
2664 ret = mnt_want_write_file(file);
2668 vol_args = memdup_user(arg, sizeof(*vol_args));
2669 if (IS_ERR(vol_args)) {
2670 ret = PTR_ERR(vol_args);
2674 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2676 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2678 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2682 mutex_lock(&root->fs_info->volume_mutex);
2683 ret = btrfs_rm_device(root, vol_args->name);
2684 mutex_unlock(&root->fs_info->volume_mutex);
2685 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2688 btrfs_info(root->fs_info, "disk deleted %s",vol_args->name);
2693 mnt_drop_write_file(file);
2697 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2699 struct btrfs_ioctl_fs_info_args *fi_args;
2700 struct btrfs_device *device;
2701 struct btrfs_device *next;
2702 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2705 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2709 mutex_lock(&fs_devices->device_list_mutex);
2710 fi_args->num_devices = fs_devices->num_devices;
2711 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2713 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2714 if (device->devid > fi_args->max_id)
2715 fi_args->max_id = device->devid;
2717 mutex_unlock(&fs_devices->device_list_mutex);
2719 fi_args->nodesize = root->fs_info->super_copy->nodesize;
2720 fi_args->sectorsize = root->fs_info->super_copy->sectorsize;
2721 fi_args->clone_alignment = root->fs_info->super_copy->sectorsize;
2723 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2730 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2732 struct btrfs_ioctl_dev_info_args *di_args;
2733 struct btrfs_device *dev;
2734 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2736 char *s_uuid = NULL;
2738 di_args = memdup_user(arg, sizeof(*di_args));
2739 if (IS_ERR(di_args))
2740 return PTR_ERR(di_args);
2742 if (!btrfs_is_empty_uuid(di_args->uuid))
2743 s_uuid = di_args->uuid;
2745 mutex_lock(&fs_devices->device_list_mutex);
2746 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
2753 di_args->devid = dev->devid;
2754 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2755 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
2756 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2758 struct rcu_string *name;
2761 name = rcu_dereference(dev->name);
2762 strncpy(di_args->path, name->str, sizeof(di_args->path));
2764 di_args->path[sizeof(di_args->path) - 1] = 0;
2766 di_args->path[0] = '\0';
2770 mutex_unlock(&fs_devices->device_list_mutex);
2771 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2778 static struct page *extent_same_get_page(struct inode *inode, u64 off)
2782 struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
2784 index = off >> PAGE_CACHE_SHIFT;
2786 page = grab_cache_page(inode->i_mapping, index);
2790 if (!PageUptodate(page)) {
2791 if (extent_read_full_page_nolock(tree, page, btrfs_get_extent,
2795 if (!PageUptodate(page)) {
2797 page_cache_release(page);
2806 static inline void lock_extent_range(struct inode *inode, u64 off, u64 len)
2808 /* do any pending delalloc/csum calc on src, one way or
2809 another, and lock file content */
2811 struct btrfs_ordered_extent *ordered;
2812 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2813 ordered = btrfs_lookup_first_ordered_extent(inode,
2816 ordered->file_offset + ordered->len <= off ||
2817 ordered->file_offset >= off + len) &&
2818 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2819 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2821 btrfs_put_ordered_extent(ordered);
2824 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2826 btrfs_put_ordered_extent(ordered);
2827 btrfs_wait_ordered_range(inode, off, len);
2831 static void btrfs_double_unlock(struct inode *inode1, u64 loff1,
2832 struct inode *inode2, u64 loff2, u64 len)
2834 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2835 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2837 mutex_unlock(&inode1->i_mutex);
2838 mutex_unlock(&inode2->i_mutex);
2841 static void btrfs_double_lock(struct inode *inode1, u64 loff1,
2842 struct inode *inode2, u64 loff2, u64 len)
2844 if (inode1 < inode2) {
2845 swap(inode1, inode2);
2849 mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT);
2850 lock_extent_range(inode1, loff1, len);
2851 if (inode1 != inode2) {
2852 mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD);
2853 lock_extent_range(inode2, loff2, len);
2857 static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
2858 u64 dst_loff, u64 len)
2861 struct page *src_page, *dst_page;
2862 unsigned int cmp_len = PAGE_CACHE_SIZE;
2863 void *addr, *dst_addr;
2866 if (len < PAGE_CACHE_SIZE)
2869 src_page = extent_same_get_page(src, loff);
2872 dst_page = extent_same_get_page(dst, dst_loff);
2874 page_cache_release(src_page);
2877 addr = kmap_atomic(src_page);
2878 dst_addr = kmap_atomic(dst_page);
2880 flush_dcache_page(src_page);
2881 flush_dcache_page(dst_page);
2883 if (memcmp(addr, dst_addr, cmp_len))
2884 ret = BTRFS_SAME_DATA_DIFFERS;
2886 kunmap_atomic(addr);
2887 kunmap_atomic(dst_addr);
2888 page_cache_release(src_page);
2889 page_cache_release(dst_page);
2895 dst_loff += cmp_len;
2902 static int extent_same_check_offsets(struct inode *inode, u64 off, u64 len)
2904 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
2906 if (off + len > inode->i_size || off + len < off)
2908 /* Check that we are block aligned - btrfs_clone() requires this */
2909 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
2915 static int btrfs_extent_same(struct inode *src, u64 loff, u64 len,
2916 struct inode *dst, u64 dst_loff)
2921 * btrfs_clone() can't handle extents in the same file
2922 * yet. Once that works, we can drop this check and replace it
2923 * with a check for the same inode, but overlapping extents.
2928 btrfs_double_lock(src, loff, dst, dst_loff, len);
2930 ret = extent_same_check_offsets(src, loff, len);
2934 ret = extent_same_check_offsets(dst, dst_loff, len);
2938 /* don't make the dst file partly checksummed */
2939 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2940 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
2945 ret = btrfs_cmp_data(src, loff, dst, dst_loff, len);
2947 ret = btrfs_clone(src, dst, loff, len, len, dst_loff);
2950 btrfs_double_unlock(src, loff, dst, dst_loff, len);
2955 #define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024)
2957 static long btrfs_ioctl_file_extent_same(struct file *file,
2958 struct btrfs_ioctl_same_args __user *argp)
2960 struct btrfs_ioctl_same_args *same;
2961 struct btrfs_ioctl_same_extent_info *info;
2962 struct inode *src = file_inode(file);
2968 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
2969 bool is_admin = capable(CAP_SYS_ADMIN);
2972 if (!(file->f_mode & FMODE_READ))
2975 ret = mnt_want_write_file(file);
2979 if (get_user(count, &argp->dest_count)) {
2984 size = offsetof(struct btrfs_ioctl_same_args __user, info[count]);
2986 same = memdup_user(argp, size);
2989 ret = PTR_ERR(same);
2993 off = same->logical_offset;
2997 * Limit the total length we will dedupe for each operation.
2998 * This is intended to bound the total time spent in this
2999 * ioctl to something sane.
3001 if (len > BTRFS_MAX_DEDUPE_LEN)
3002 len = BTRFS_MAX_DEDUPE_LEN;
3004 if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) {
3006 * Btrfs does not support blocksize < page_size. As a
3007 * result, btrfs_cmp_data() won't correctly handle
3008 * this situation without an update.
3015 if (S_ISDIR(src->i_mode))
3019 if (!S_ISREG(src->i_mode))
3022 /* pre-format output fields to sane values */
3023 for (i = 0; i < count; i++) {
3024 same->info[i].bytes_deduped = 0ULL;
3025 same->info[i].status = 0;
3028 for (i = 0, info = same->info; i < count; i++, info++) {
3030 struct fd dst_file = fdget(info->fd);
3031 if (!dst_file.file) {
3032 info->status = -EBADF;
3035 dst = file_inode(dst_file.file);
3037 if (!(is_admin || (dst_file.file->f_mode & FMODE_WRITE))) {
3038 info->status = -EINVAL;
3039 } else if (file->f_path.mnt != dst_file.file->f_path.mnt) {
3040 info->status = -EXDEV;
3041 } else if (S_ISDIR(dst->i_mode)) {
3042 info->status = -EISDIR;
3043 } else if (!S_ISREG(dst->i_mode)) {
3044 info->status = -EACCES;
3046 info->status = btrfs_extent_same(src, off, len, dst,
3047 info->logical_offset);
3048 if (info->status == 0)
3049 info->bytes_deduped += len;
3054 ret = copy_to_user(argp, same, size);
3059 mnt_drop_write_file(file);
3063 /* Helper to check and see if this root currently has a ref on the given disk
3064 * bytenr. If it does then we need to update the quota for this root. This
3065 * doesn't do anything if quotas aren't enabled.
3067 static int check_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
3070 struct seq_list tree_mod_seq_elem = {};
3071 struct ulist *roots;
3072 struct ulist_iterator uiter;
3073 struct ulist_node *root_node = NULL;
3076 if (!root->fs_info->quota_enabled)
3079 btrfs_get_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3080 ret = btrfs_find_all_roots(trans, root->fs_info, disko,
3081 tree_mod_seq_elem.seq, &roots);
3085 ULIST_ITER_INIT(&uiter);
3086 while ((root_node = ulist_next(roots, &uiter))) {
3087 if (root_node->val == root->objectid) {
3094 btrfs_put_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3098 static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3099 struct inode *inode,
3104 struct btrfs_root *root = BTRFS_I(inode)->root;
3107 inode_inc_iversion(inode);
3108 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3110 * We round up to the block size at eof when determining which
3111 * extents to clone above, but shouldn't round up the file size.
3113 if (endoff > destoff + olen)
3114 endoff = destoff + olen;
3115 if (endoff > inode->i_size)
3116 btrfs_i_size_write(inode, endoff);
3118 ret = btrfs_update_inode(trans, root, inode);
3120 btrfs_abort_transaction(trans, root, ret);
3121 btrfs_end_transaction(trans, root);
3124 ret = btrfs_end_transaction(trans, root);
3129 static void clone_update_extent_map(struct inode *inode,
3130 const struct btrfs_trans_handle *trans,
3131 const struct btrfs_path *path,
3132 const u64 hole_offset,
3135 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
3136 struct extent_map *em;
3139 em = alloc_extent_map();
3141 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3142 &BTRFS_I(inode)->runtime_flags);
3147 struct btrfs_file_extent_item *fi;
3149 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3150 struct btrfs_file_extent_item);
3151 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3152 em->generation = -1;
3153 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3154 BTRFS_FILE_EXTENT_INLINE)
3155 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3156 &BTRFS_I(inode)->runtime_flags);
3158 em->start = hole_offset;
3160 em->ram_bytes = em->len;
3161 em->orig_start = hole_offset;
3162 em->block_start = EXTENT_MAP_HOLE;
3164 em->orig_block_len = 0;
3165 em->compress_type = BTRFS_COMPRESS_NONE;
3166 em->generation = trans->transid;
3170 write_lock(&em_tree->lock);
3171 ret = add_extent_mapping(em_tree, em, 1);
3172 write_unlock(&em_tree->lock);
3173 if (ret != -EEXIST) {
3174 free_extent_map(em);
3177 btrfs_drop_extent_cache(inode, em->start,
3178 em->start + em->len - 1, 0);
3182 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3183 &BTRFS_I(inode)->runtime_flags);
3187 * btrfs_clone() - clone a range from inode file to another
3189 * @src: Inode to clone from
3190 * @inode: Inode to clone to
3191 * @off: Offset within source to start clone from
3192 * @olen: Original length, passed by user, of range to clone
3193 * @olen_aligned: Block-aligned value of olen, extent_same uses
3194 * identical values here
3195 * @destoff: Offset within @inode to start clone
3197 static int btrfs_clone(struct inode *src, struct inode *inode,
3198 const u64 off, const u64 olen, const u64 olen_aligned,
3201 struct btrfs_root *root = BTRFS_I(inode)->root;
3202 struct btrfs_path *path = NULL;
3203 struct extent_buffer *leaf;
3204 struct btrfs_trans_handle *trans;
3206 struct btrfs_key key;
3211 const u64 len = olen_aligned;
3213 u64 last_dest_end = destoff;
3216 buf = vmalloc(root->nodesize);
3220 path = btrfs_alloc_path();
3228 key.objectid = btrfs_ino(src);
3229 key.type = BTRFS_EXTENT_DATA_KEY;
3234 * note the key will change type as we walk through the
3237 path->leave_spinning = 1;
3238 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3243 * First search, if no extent item that starts at offset off was
3244 * found but the previous item is an extent item, it's possible
3245 * it might overlap our target range, therefore process it.
3247 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3248 btrfs_item_key_to_cpu(path->nodes[0], &key,
3249 path->slots[0] - 1);
3250 if (key.type == BTRFS_EXTENT_DATA_KEY)
3254 nritems = btrfs_header_nritems(path->nodes[0]);
3257 if (path->slots[0] >= nritems) {
3258 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
3263 nritems = btrfs_header_nritems(path->nodes[0]);
3265 leaf = path->nodes[0];
3266 slot = path->slots[0];
3268 btrfs_item_key_to_cpu(leaf, &key, slot);
3269 if (key.type > BTRFS_EXTENT_DATA_KEY ||
3270 key.objectid != btrfs_ino(src))
3273 if (key.type == BTRFS_EXTENT_DATA_KEY) {
3274 struct btrfs_file_extent_item *extent;
3277 struct btrfs_key new_key;
3278 u64 disko = 0, diskl = 0;
3279 u64 datao = 0, datal = 0;
3283 extent = btrfs_item_ptr(leaf, slot,
3284 struct btrfs_file_extent_item);
3285 comp = btrfs_file_extent_compression(leaf, extent);
3286 type = btrfs_file_extent_type(leaf, extent);
3287 if (type == BTRFS_FILE_EXTENT_REG ||
3288 type == BTRFS_FILE_EXTENT_PREALLOC) {
3289 disko = btrfs_file_extent_disk_bytenr(leaf,
3291 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3293 datao = btrfs_file_extent_offset(leaf, extent);
3294 datal = btrfs_file_extent_num_bytes(leaf,
3296 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3297 /* take upper bound, may be compressed */
3298 datal = btrfs_file_extent_ram_bytes(leaf,
3303 * The first search might have left us at an extent
3304 * item that ends before our target range's start, can
3305 * happen if we have holes and NO_HOLES feature enabled.
3307 if (key.offset + datal <= off) {
3310 } else if (key.offset >= off + len) {
3314 size = btrfs_item_size_nr(leaf, slot);
3315 read_extent_buffer(leaf, buf,
3316 btrfs_item_ptr_offset(leaf, slot),
3319 btrfs_release_path(path);
3320 path->leave_spinning = 0;
3322 memcpy(&new_key, &key, sizeof(new_key));
3323 new_key.objectid = btrfs_ino(inode);
3324 if (off <= key.offset)
3325 new_key.offset = key.offset + destoff - off;
3327 new_key.offset = destoff;
3330 * Deal with a hole that doesn't have an extent item
3331 * that represents it (NO_HOLES feature enabled).
3332 * This hole is either in the middle of the cloning
3333 * range or at the beginning (fully overlaps it or
3334 * partially overlaps it).
3336 if (new_key.offset != last_dest_end)
3337 drop_start = last_dest_end;
3339 drop_start = new_key.offset;
3342 * 1 - adjusting old extent (we may have to split it)
3343 * 1 - add new extent
3346 trans = btrfs_start_transaction(root, 3);
3347 if (IS_ERR(trans)) {
3348 ret = PTR_ERR(trans);
3352 if (type == BTRFS_FILE_EXTENT_REG ||
3353 type == BTRFS_FILE_EXTENT_PREALLOC) {
3355 * a | --- range to clone ---| b
3356 * | ------------- extent ------------- |
3359 /* subtract range b */
3360 if (key.offset + datal > off + len)
3361 datal = off + len - key.offset;
3363 /* subtract range a */
3364 if (off > key.offset) {
3365 datao += off - key.offset;
3366 datal -= off - key.offset;
3369 ret = btrfs_drop_extents(trans, root, inode,
3371 new_key.offset + datal,
3374 if (ret != -EOPNOTSUPP)
3375 btrfs_abort_transaction(trans,
3377 btrfs_end_transaction(trans, root);
3381 ret = btrfs_insert_empty_item(trans, root, path,
3384 btrfs_abort_transaction(trans, root,
3386 btrfs_end_transaction(trans, root);
3390 leaf = path->nodes[0];
3391 slot = path->slots[0];
3392 write_extent_buffer(leaf, buf,
3393 btrfs_item_ptr_offset(leaf, slot),
3396 extent = btrfs_item_ptr(leaf, slot,
3397 struct btrfs_file_extent_item);
3399 /* disko == 0 means it's a hole */
3403 btrfs_set_file_extent_offset(leaf, extent,
3405 btrfs_set_file_extent_num_bytes(leaf, extent,
3409 * We need to look up the roots that point at
3410 * this bytenr and see if the new root does. If
3411 * it does not we need to make sure we update
3412 * quotas appropriately.
3414 if (disko && root != BTRFS_I(src)->root &&
3415 disko != last_disko) {
3416 no_quota = check_ref(trans, root,
3419 btrfs_abort_transaction(trans,
3422 btrfs_end_transaction(trans,
3430 inode_add_bytes(inode, datal);
3431 ret = btrfs_inc_extent_ref(trans, root,
3433 root->root_key.objectid,
3435 new_key.offset - datao,
3438 btrfs_abort_transaction(trans,
3441 btrfs_end_transaction(trans,
3447 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3450 u64 aligned_end = 0;
3452 if (off > key.offset) {
3453 skip = off - key.offset;
3454 new_key.offset += skip;
3457 if (key.offset + datal > off + len)
3458 trim = key.offset + datal - (off + len);
3460 if (comp && (skip || trim)) {
3462 btrfs_end_transaction(trans, root);
3465 size -= skip + trim;
3466 datal -= skip + trim;
3468 aligned_end = ALIGN(new_key.offset + datal,
3470 ret = btrfs_drop_extents(trans, root, inode,
3475 if (ret != -EOPNOTSUPP)
3476 btrfs_abort_transaction(trans,
3478 btrfs_end_transaction(trans, root);
3482 ret = btrfs_insert_empty_item(trans, root, path,
3485 btrfs_abort_transaction(trans, root,
3487 btrfs_end_transaction(trans, root);
3493 btrfs_file_extent_calc_inline_size(0);
3494 memmove(buf+start, buf+start+skip,
3498 leaf = path->nodes[0];
3499 slot = path->slots[0];
3500 write_extent_buffer(leaf, buf,
3501 btrfs_item_ptr_offset(leaf, slot),
3503 inode_add_bytes(inode, datal);
3506 /* If we have an implicit hole (NO_HOLES feature). */
3507 if (drop_start < new_key.offset)
3508 clone_update_extent_map(inode, trans,
3510 new_key.offset - drop_start);
3512 clone_update_extent_map(inode, trans, path, 0, 0);
3514 btrfs_mark_buffer_dirty(leaf);
3515 btrfs_release_path(path);
3517 last_dest_end = ALIGN(new_key.offset + datal,
3519 ret = clone_finish_inode_update(trans, inode,
3524 if (new_key.offset + datal >= destoff + len)
3527 btrfs_release_path(path);
3532 if (last_dest_end < destoff + len) {
3534 * We have an implicit hole (NO_HOLES feature is enabled) that
3535 * fully or partially overlaps our cloning range at its end.
3537 btrfs_release_path(path);
3540 * 1 - remove extent(s)
3543 trans = btrfs_start_transaction(root, 2);
3544 if (IS_ERR(trans)) {
3545 ret = PTR_ERR(trans);
3548 ret = btrfs_drop_extents(trans, root, inode,
3549 last_dest_end, destoff + len, 1);
3551 if (ret != -EOPNOTSUPP)
3552 btrfs_abort_transaction(trans, root, ret);
3553 btrfs_end_transaction(trans, root);
3556 clone_update_extent_map(inode, trans, NULL, last_dest_end,
3557 destoff + len - last_dest_end);
3558 ret = clone_finish_inode_update(trans, inode, destoff + len,
3563 btrfs_free_path(path);
3568 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
3569 u64 off, u64 olen, u64 destoff)
3571 struct inode *inode = file_inode(file);
3572 struct btrfs_root *root = BTRFS_I(inode)->root;
3577 u64 bs = root->fs_info->sb->s_blocksize;
3582 * - split compressed inline extents. annoying: we need to
3583 * decompress into destination's address_space (the file offset
3584 * may change, so source mapping won't do), then recompress (or
3585 * otherwise reinsert) a subrange.
3587 * - split destination inode's inline extents. The inline extents can
3588 * be either compressed or non-compressed.
3591 /* the destination must be opened for writing */
3592 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
3595 if (btrfs_root_readonly(root))
3598 ret = mnt_want_write_file(file);
3602 src_file = fdget(srcfd);
3603 if (!src_file.file) {
3605 goto out_drop_write;
3609 if (src_file.file->f_path.mnt != file->f_path.mnt)
3612 src = file_inode(src_file.file);
3618 /* the src must be open for reading */
3619 if (!(src_file.file->f_mode & FMODE_READ))
3622 /* don't make the dst file partly checksummed */
3623 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3624 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3628 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3632 if (src->i_sb != inode->i_sb)
3637 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
3638 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
3640 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
3641 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
3644 mutex_lock(&src->i_mutex);
3647 /* determine range to clone */
3649 if (off + len > src->i_size || off + len < off)
3652 olen = len = src->i_size - off;
3653 /* if we extend to eof, continue to block boundary */
3654 if (off + len == src->i_size)
3655 len = ALIGN(src->i_size, bs) - off;
3657 /* verify the end result is block aligned */
3658 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3659 !IS_ALIGNED(destoff, bs))
3662 /* verify if ranges are overlapped within the same file */
3664 if (destoff + len > off && destoff < off + len)
3668 if (destoff > inode->i_size) {
3669 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3675 * Lock the target range too. Right after we replace the file extent
3676 * items in the fs tree (which now point to the cloned data), we might
3677 * have a worker replace them with extent items relative to a write
3678 * operation that was issued before this clone operation (i.e. confront
3679 * with inode.c:btrfs_finish_ordered_io).
3682 u64 lock_start = min_t(u64, off, destoff);
3683 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
3685 lock_extent_range(src, lock_start, lock_len);
3687 lock_extent_range(src, off, len);
3688 lock_extent_range(inode, destoff, len);
3691 ret = btrfs_clone(src, inode, off, olen, len, destoff);
3694 u64 lock_start = min_t(u64, off, destoff);
3695 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3697 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3699 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
3700 unlock_extent(&BTRFS_I(inode)->io_tree, destoff,
3704 * Truncate page cache pages so that future reads will see the cloned
3705 * data immediately and not the previous data.
3707 truncate_inode_pages_range(&inode->i_data, destoff,
3708 PAGE_CACHE_ALIGN(destoff + len) - 1);
3712 mutex_unlock(&src->i_mutex);
3713 mutex_unlock(&inode->i_mutex);
3715 mutex_unlock(&inode->i_mutex);
3716 mutex_unlock(&src->i_mutex);
3719 mutex_unlock(&src->i_mutex);
3724 mnt_drop_write_file(file);
3728 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
3730 struct btrfs_ioctl_clone_range_args args;
3732 if (copy_from_user(&args, argp, sizeof(args)))
3734 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
3735 args.src_length, args.dest_offset);
3739 * there are many ways the trans_start and trans_end ioctls can lead
3740 * to deadlocks. They should only be used by applications that
3741 * basically own the machine, and have a very in depth understanding
3742 * of all the possible deadlocks and enospc problems.
3744 static long btrfs_ioctl_trans_start(struct file *file)
3746 struct inode *inode = file_inode(file);
3747 struct btrfs_root *root = BTRFS_I(inode)->root;
3748 struct btrfs_trans_handle *trans;
3752 if (!capable(CAP_SYS_ADMIN))
3756 if (file->private_data)
3760 if (btrfs_root_readonly(root))
3763 ret = mnt_want_write_file(file);
3767 atomic_inc(&root->fs_info->open_ioctl_trans);
3770 trans = btrfs_start_ioctl_transaction(root);
3774 file->private_data = trans;
3778 atomic_dec(&root->fs_info->open_ioctl_trans);
3779 mnt_drop_write_file(file);
3784 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3786 struct inode *inode = file_inode(file);
3787 struct btrfs_root *root = BTRFS_I(inode)->root;
3788 struct btrfs_root *new_root;
3789 struct btrfs_dir_item *di;
3790 struct btrfs_trans_handle *trans;
3791 struct btrfs_path *path;
3792 struct btrfs_key location;
3793 struct btrfs_disk_key disk_key;
3798 if (!capable(CAP_SYS_ADMIN))
3801 ret = mnt_want_write_file(file);
3805 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3811 objectid = BTRFS_FS_TREE_OBJECTID;
3813 location.objectid = objectid;
3814 location.type = BTRFS_ROOT_ITEM_KEY;
3815 location.offset = (u64)-1;
3817 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
3818 if (IS_ERR(new_root)) {
3819 ret = PTR_ERR(new_root);
3823 path = btrfs_alloc_path();
3828 path->leave_spinning = 1;
3830 trans = btrfs_start_transaction(root, 1);
3831 if (IS_ERR(trans)) {
3832 btrfs_free_path(path);
3833 ret = PTR_ERR(trans);
3837 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
3838 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
3839 dir_id, "default", 7, 1);
3840 if (IS_ERR_OR_NULL(di)) {
3841 btrfs_free_path(path);
3842 btrfs_end_transaction(trans, root);
3843 btrfs_err(new_root->fs_info, "Umm, you don't have the default dir"
3844 "item, this isn't going to work");
3849 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3850 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3851 btrfs_mark_buffer_dirty(path->nodes[0]);
3852 btrfs_free_path(path);
3854 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
3855 btrfs_end_transaction(trans, root);
3857 mnt_drop_write_file(file);
3861 void btrfs_get_block_group_info(struct list_head *groups_list,
3862 struct btrfs_ioctl_space_info *space)
3864 struct btrfs_block_group_cache *block_group;
3866 space->total_bytes = 0;
3867 space->used_bytes = 0;
3869 list_for_each_entry(block_group, groups_list, list) {
3870 space->flags = block_group->flags;
3871 space->total_bytes += block_group->key.offset;
3872 space->used_bytes +=
3873 btrfs_block_group_used(&block_group->item);
3877 static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
3879 struct btrfs_ioctl_space_args space_args;
3880 struct btrfs_ioctl_space_info space;
3881 struct btrfs_ioctl_space_info *dest;
3882 struct btrfs_ioctl_space_info *dest_orig;
3883 struct btrfs_ioctl_space_info __user *user_dest;
3884 struct btrfs_space_info *info;
3885 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3886 BTRFS_BLOCK_GROUP_SYSTEM,
3887 BTRFS_BLOCK_GROUP_METADATA,
3888 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3895 if (copy_from_user(&space_args,
3896 (struct btrfs_ioctl_space_args __user *)arg,
3897 sizeof(space_args)))
3900 for (i = 0; i < num_types; i++) {
3901 struct btrfs_space_info *tmp;
3905 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3907 if (tmp->flags == types[i]) {
3917 down_read(&info->groups_sem);
3918 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3919 if (!list_empty(&info->block_groups[c]))
3922 up_read(&info->groups_sem);
3926 * Global block reserve, exported as a space_info
3930 /* space_slots == 0 means they are asking for a count */
3931 if (space_args.space_slots == 0) {
3932 space_args.total_spaces = slot_count;
3936 slot_count = min_t(u64, space_args.space_slots, slot_count);
3938 alloc_size = sizeof(*dest) * slot_count;
3940 /* we generally have at most 6 or so space infos, one for each raid
3941 * level. So, a whole page should be more than enough for everyone
3943 if (alloc_size > PAGE_CACHE_SIZE)
3946 space_args.total_spaces = 0;
3947 dest = kmalloc(alloc_size, GFP_NOFS);
3952 /* now we have a buffer to copy into */
3953 for (i = 0; i < num_types; i++) {
3954 struct btrfs_space_info *tmp;
3961 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3963 if (tmp->flags == types[i]) {
3972 down_read(&info->groups_sem);
3973 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3974 if (!list_empty(&info->block_groups[c])) {
3975 btrfs_get_block_group_info(
3976 &info->block_groups[c], &space);
3977 memcpy(dest, &space, sizeof(space));
3979 space_args.total_spaces++;
3985 up_read(&info->groups_sem);
3989 * Add global block reserve
3992 struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv;
3994 spin_lock(&block_rsv->lock);
3995 space.total_bytes = block_rsv->size;
3996 space.used_bytes = block_rsv->size - block_rsv->reserved;
3997 spin_unlock(&block_rsv->lock);
3998 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3999 memcpy(dest, &space, sizeof(space));
4000 space_args.total_spaces++;
4003 user_dest = (struct btrfs_ioctl_space_info __user *)
4004 (arg + sizeof(struct btrfs_ioctl_space_args));
4006 if (copy_to_user(user_dest, dest_orig, alloc_size))
4011 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
4018 * there are many ways the trans_start and trans_end ioctls can lead
4019 * to deadlocks. They should only be used by applications that
4020 * basically own the machine, and have a very in depth understanding
4021 * of all the possible deadlocks and enospc problems.
4023 long btrfs_ioctl_trans_end(struct file *file)
4025 struct inode *inode = file_inode(file);
4026 struct btrfs_root *root = BTRFS_I(inode)->root;
4027 struct btrfs_trans_handle *trans;
4029 trans = file->private_data;
4032 file->private_data = NULL;
4034 btrfs_end_transaction(trans, root);
4036 atomic_dec(&root->fs_info->open_ioctl_trans);
4038 mnt_drop_write_file(file);
4042 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4045 struct btrfs_trans_handle *trans;
4049 trans = btrfs_attach_transaction_barrier(root);
4050 if (IS_ERR(trans)) {
4051 if (PTR_ERR(trans) != -ENOENT)
4052 return PTR_ERR(trans);
4054 /* No running transaction, don't bother */
4055 transid = root->fs_info->last_trans_committed;
4058 transid = trans->transid;
4059 ret = btrfs_commit_transaction_async(trans, root, 0);
4061 btrfs_end_transaction(trans, root);
4066 if (copy_to_user(argp, &transid, sizeof(transid)))
4071 static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
4077 if (copy_from_user(&transid, argp, sizeof(transid)))
4080 transid = 0; /* current trans */
4082 return btrfs_wait_for_commit(root, transid);
4085 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
4087 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4088 struct btrfs_ioctl_scrub_args *sa;
4091 if (!capable(CAP_SYS_ADMIN))
4094 sa = memdup_user(arg, sizeof(*sa));
4098 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4099 ret = mnt_want_write_file(file);
4104 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
4105 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4108 if (copy_to_user(arg, sa, sizeof(*sa)))
4111 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4112 mnt_drop_write_file(file);
4118 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
4120 if (!capable(CAP_SYS_ADMIN))
4123 return btrfs_scrub_cancel(root->fs_info);
4126 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
4129 struct btrfs_ioctl_scrub_args *sa;
4132 if (!capable(CAP_SYS_ADMIN))
4135 sa = memdup_user(arg, sizeof(*sa));
4139 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
4141 if (copy_to_user(arg, sa, sizeof(*sa)))
4148 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
4151 struct btrfs_ioctl_get_dev_stats *sa;
4154 sa = memdup_user(arg, sizeof(*sa));
4158 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4163 ret = btrfs_get_dev_stats(root, sa);
4165 if (copy_to_user(arg, sa, sizeof(*sa)))
4172 static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
4174 struct btrfs_ioctl_dev_replace_args *p;
4177 if (!capable(CAP_SYS_ADMIN))
4180 p = memdup_user(arg, sizeof(*p));
4185 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
4186 if (root->fs_info->sb->s_flags & MS_RDONLY) {
4191 &root->fs_info->mutually_exclusive_operation_running,
4193 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4195 ret = btrfs_dev_replace_start(root, p);
4197 &root->fs_info->mutually_exclusive_operation_running,
4201 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4202 btrfs_dev_replace_status(root->fs_info, p);
4205 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4206 ret = btrfs_dev_replace_cancel(root->fs_info, p);
4213 if (copy_to_user(arg, p, sizeof(*p)))
4220 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4226 struct btrfs_ioctl_ino_path_args *ipa = NULL;
4227 struct inode_fs_paths *ipath = NULL;
4228 struct btrfs_path *path;
4230 if (!capable(CAP_DAC_READ_SEARCH))
4233 path = btrfs_alloc_path();
4239 ipa = memdup_user(arg, sizeof(*ipa));
4246 size = min_t(u32, ipa->size, 4096);
4247 ipath = init_ipath(size, root, path);
4248 if (IS_ERR(ipath)) {
4249 ret = PTR_ERR(ipath);
4254 ret = paths_from_inode(ipa->inum, ipath);
4258 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
4259 rel_ptr = ipath->fspath->val[i] -
4260 (u64)(unsigned long)ipath->fspath->val;
4261 ipath->fspath->val[i] = rel_ptr;
4264 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
4265 (void *)(unsigned long)ipath->fspath, size);
4272 btrfs_free_path(path);
4279 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4281 struct btrfs_data_container *inodes = ctx;
4282 const size_t c = 3 * sizeof(u64);
4284 if (inodes->bytes_left >= c) {
4285 inodes->bytes_left -= c;
4286 inodes->val[inodes->elem_cnt] = inum;
4287 inodes->val[inodes->elem_cnt + 1] = offset;
4288 inodes->val[inodes->elem_cnt + 2] = root;
4289 inodes->elem_cnt += 3;
4291 inodes->bytes_missing += c - inodes->bytes_left;
4292 inodes->bytes_left = 0;
4293 inodes->elem_missed += 3;
4299 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
4304 struct btrfs_ioctl_logical_ino_args *loi;
4305 struct btrfs_data_container *inodes = NULL;
4306 struct btrfs_path *path = NULL;
4308 if (!capable(CAP_SYS_ADMIN))
4311 loi = memdup_user(arg, sizeof(*loi));
4318 path = btrfs_alloc_path();
4324 size = min_t(u32, loi->size, 64 * 1024);
4325 inodes = init_data_container(size);
4326 if (IS_ERR(inodes)) {
4327 ret = PTR_ERR(inodes);
4332 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
4333 build_ino_list, inodes);
4339 ret = copy_to_user((void *)(unsigned long)loi->inodes,
4340 (void *)(unsigned long)inodes, size);
4345 btrfs_free_path(path);
4352 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
4353 struct btrfs_ioctl_balance_args *bargs)
4355 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4357 bargs->flags = bctl->flags;
4359 if (atomic_read(&fs_info->balance_running))
4360 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4361 if (atomic_read(&fs_info->balance_pause_req))
4362 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
4363 if (atomic_read(&fs_info->balance_cancel_req))
4364 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
4366 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4367 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4368 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4371 spin_lock(&fs_info->balance_lock);
4372 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4373 spin_unlock(&fs_info->balance_lock);
4375 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4379 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4381 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4382 struct btrfs_fs_info *fs_info = root->fs_info;
4383 struct btrfs_ioctl_balance_args *bargs;
4384 struct btrfs_balance_control *bctl;
4385 bool need_unlock; /* for mut. excl. ops lock */
4388 if (!capable(CAP_SYS_ADMIN))
4391 ret = mnt_want_write_file(file);
4396 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
4397 mutex_lock(&fs_info->volume_mutex);
4398 mutex_lock(&fs_info->balance_mutex);
4404 * mut. excl. ops lock is locked. Three possibilites:
4405 * (1) some other op is running
4406 * (2) balance is running
4407 * (3) balance is paused -- special case (think resume)
4409 mutex_lock(&fs_info->balance_mutex);
4410 if (fs_info->balance_ctl) {
4411 /* this is either (2) or (3) */
4412 if (!atomic_read(&fs_info->balance_running)) {
4413 mutex_unlock(&fs_info->balance_mutex);
4414 if (!mutex_trylock(&fs_info->volume_mutex))
4416 mutex_lock(&fs_info->balance_mutex);
4418 if (fs_info->balance_ctl &&
4419 !atomic_read(&fs_info->balance_running)) {
4421 need_unlock = false;
4425 mutex_unlock(&fs_info->balance_mutex);
4426 mutex_unlock(&fs_info->volume_mutex);
4430 mutex_unlock(&fs_info->balance_mutex);
4436 mutex_unlock(&fs_info->balance_mutex);
4437 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4442 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
4445 bargs = memdup_user(arg, sizeof(*bargs));
4446 if (IS_ERR(bargs)) {
4447 ret = PTR_ERR(bargs);
4451 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4452 if (!fs_info->balance_ctl) {
4457 bctl = fs_info->balance_ctl;
4458 spin_lock(&fs_info->balance_lock);
4459 bctl->flags |= BTRFS_BALANCE_RESUME;
4460 spin_unlock(&fs_info->balance_lock);
4468 if (fs_info->balance_ctl) {
4473 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
4479 bctl->fs_info = fs_info;
4481 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4482 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4483 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4485 bctl->flags = bargs->flags;
4487 /* balance everything - no filters */
4488 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4493 * Ownership of bctl and mutually_exclusive_operation_running
4494 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4495 * or, if restriper was paused all the way until unmount, in
4496 * free_fs_info. mutually_exclusive_operation_running is
4497 * cleared in __cancel_balance.
4499 need_unlock = false;
4501 ret = btrfs_balance(bctl, bargs);
4504 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4511 mutex_unlock(&fs_info->balance_mutex);
4512 mutex_unlock(&fs_info->volume_mutex);
4514 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4516 mnt_drop_write_file(file);
4520 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4522 if (!capable(CAP_SYS_ADMIN))
4526 case BTRFS_BALANCE_CTL_PAUSE:
4527 return btrfs_pause_balance(root->fs_info);
4528 case BTRFS_BALANCE_CTL_CANCEL:
4529 return btrfs_cancel_balance(root->fs_info);
4535 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4538 struct btrfs_fs_info *fs_info = root->fs_info;
4539 struct btrfs_ioctl_balance_args *bargs;
4542 if (!capable(CAP_SYS_ADMIN))
4545 mutex_lock(&fs_info->balance_mutex);
4546 if (!fs_info->balance_ctl) {
4551 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
4557 update_ioctl_balance_args(fs_info, 1, bargs);
4559 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4564 mutex_unlock(&fs_info->balance_mutex);
4568 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4570 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4571 struct btrfs_ioctl_quota_ctl_args *sa;
4572 struct btrfs_trans_handle *trans = NULL;
4576 if (!capable(CAP_SYS_ADMIN))
4579 ret = mnt_want_write_file(file);
4583 sa = memdup_user(arg, sizeof(*sa));
4589 down_write(&root->fs_info->subvol_sem);
4590 trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4591 if (IS_ERR(trans)) {
4592 ret = PTR_ERR(trans);
4597 case BTRFS_QUOTA_CTL_ENABLE:
4598 ret = btrfs_quota_enable(trans, root->fs_info);
4600 case BTRFS_QUOTA_CTL_DISABLE:
4601 ret = btrfs_quota_disable(trans, root->fs_info);
4608 err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4613 up_write(&root->fs_info->subvol_sem);
4615 mnt_drop_write_file(file);
4619 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4621 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4622 struct btrfs_ioctl_qgroup_assign_args *sa;
4623 struct btrfs_trans_handle *trans;
4627 if (!capable(CAP_SYS_ADMIN))
4630 ret = mnt_want_write_file(file);
4634 sa = memdup_user(arg, sizeof(*sa));
4640 trans = btrfs_join_transaction(root);
4641 if (IS_ERR(trans)) {
4642 ret = PTR_ERR(trans);
4646 /* FIXME: check if the IDs really exist */
4648 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4651 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4655 err = btrfs_end_transaction(trans, root);
4662 mnt_drop_write_file(file);
4666 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4668 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4669 struct btrfs_ioctl_qgroup_create_args *sa;
4670 struct btrfs_trans_handle *trans;
4674 if (!capable(CAP_SYS_ADMIN))
4677 ret = mnt_want_write_file(file);
4681 sa = memdup_user(arg, sizeof(*sa));
4687 if (!sa->qgroupid) {
4692 trans = btrfs_join_transaction(root);
4693 if (IS_ERR(trans)) {
4694 ret = PTR_ERR(trans);
4698 /* FIXME: check if the IDs really exist */
4700 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
4703 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4706 err = btrfs_end_transaction(trans, root);
4713 mnt_drop_write_file(file);
4717 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4719 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4720 struct btrfs_ioctl_qgroup_limit_args *sa;
4721 struct btrfs_trans_handle *trans;
4726 if (!capable(CAP_SYS_ADMIN))
4729 ret = mnt_want_write_file(file);
4733 sa = memdup_user(arg, sizeof(*sa));
4739 trans = btrfs_join_transaction(root);
4740 if (IS_ERR(trans)) {
4741 ret = PTR_ERR(trans);
4745 qgroupid = sa->qgroupid;
4747 /* take the current subvol as qgroup */
4748 qgroupid = root->root_key.objectid;
4751 /* FIXME: check if the IDs really exist */
4752 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
4754 err = btrfs_end_transaction(trans, root);
4761 mnt_drop_write_file(file);
4765 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4767 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4768 struct btrfs_ioctl_quota_rescan_args *qsa;
4771 if (!capable(CAP_SYS_ADMIN))
4774 ret = mnt_want_write_file(file);
4778 qsa = memdup_user(arg, sizeof(*qsa));
4789 ret = btrfs_qgroup_rescan(root->fs_info);
4794 mnt_drop_write_file(file);
4798 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
4800 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4801 struct btrfs_ioctl_quota_rescan_args *qsa;
4804 if (!capable(CAP_SYS_ADMIN))
4807 qsa = kzalloc(sizeof(*qsa), GFP_NOFS);
4811 if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4813 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
4816 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4823 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
4825 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4827 if (!capable(CAP_SYS_ADMIN))
4830 return btrfs_qgroup_wait_for_completion(root->fs_info);
4833 static long _btrfs_ioctl_set_received_subvol(struct file *file,
4834 struct btrfs_ioctl_received_subvol_args *sa)
4836 struct inode *inode = file_inode(file);
4837 struct btrfs_root *root = BTRFS_I(inode)->root;
4838 struct btrfs_root_item *root_item = &root->root_item;
4839 struct btrfs_trans_handle *trans;
4840 struct timespec ct = CURRENT_TIME;
4842 int received_uuid_changed;
4844 if (!inode_owner_or_capable(inode))
4847 ret = mnt_want_write_file(file);
4851 down_write(&root->fs_info->subvol_sem);
4853 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
4858 if (btrfs_root_readonly(root)) {
4865 * 2 - uuid items (received uuid + subvol uuid)
4867 trans = btrfs_start_transaction(root, 3);
4868 if (IS_ERR(trans)) {
4869 ret = PTR_ERR(trans);
4874 sa->rtransid = trans->transid;
4875 sa->rtime.sec = ct.tv_sec;
4876 sa->rtime.nsec = ct.tv_nsec;
4878 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4880 if (received_uuid_changed &&
4881 !btrfs_is_empty_uuid(root_item->received_uuid))
4882 btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
4883 root_item->received_uuid,
4884 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4885 root->root_key.objectid);
4886 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4887 btrfs_set_root_stransid(root_item, sa->stransid);
4888 btrfs_set_root_rtransid(root_item, sa->rtransid);
4889 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4890 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4891 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4892 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4894 ret = btrfs_update_root(trans, root->fs_info->tree_root,
4895 &root->root_key, &root->root_item);
4897 btrfs_end_transaction(trans, root);
4900 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4901 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
4903 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4904 root->root_key.objectid);
4905 if (ret < 0 && ret != -EEXIST) {
4906 btrfs_abort_transaction(trans, root, ret);
4910 ret = btrfs_commit_transaction(trans, root);
4912 btrfs_abort_transaction(trans, root, ret);
4917 up_write(&root->fs_info->subvol_sem);
4918 mnt_drop_write_file(file);
4923 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4926 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4927 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4930 args32 = memdup_user(arg, sizeof(*args32));
4931 if (IS_ERR(args32)) {
4932 ret = PTR_ERR(args32);
4937 args64 = kmalloc(sizeof(*args64), GFP_NOFS);
4943 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4944 args64->stransid = args32->stransid;
4945 args64->rtransid = args32->rtransid;
4946 args64->stime.sec = args32->stime.sec;
4947 args64->stime.nsec = args32->stime.nsec;
4948 args64->rtime.sec = args32->rtime.sec;
4949 args64->rtime.nsec = args32->rtime.nsec;
4950 args64->flags = args32->flags;
4952 ret = _btrfs_ioctl_set_received_subvol(file, args64);
4956 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4957 args32->stransid = args64->stransid;
4958 args32->rtransid = args64->rtransid;
4959 args32->stime.sec = args64->stime.sec;
4960 args32->stime.nsec = args64->stime.nsec;
4961 args32->rtime.sec = args64->rtime.sec;
4962 args32->rtime.nsec = args64->rtime.nsec;
4963 args32->flags = args64->flags;
4965 ret = copy_to_user(arg, args32, sizeof(*args32));
4976 static long btrfs_ioctl_set_received_subvol(struct file *file,
4979 struct btrfs_ioctl_received_subvol_args *sa = NULL;
4982 sa = memdup_user(arg, sizeof(*sa));
4989 ret = _btrfs_ioctl_set_received_subvol(file, sa);
4994 ret = copy_to_user(arg, sa, sizeof(*sa));
5003 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
5005 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5008 char label[BTRFS_LABEL_SIZE];
5010 spin_lock(&root->fs_info->super_lock);
5011 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
5012 spin_unlock(&root->fs_info->super_lock);
5014 len = strnlen(label, BTRFS_LABEL_SIZE);
5016 if (len == BTRFS_LABEL_SIZE) {
5017 btrfs_warn(root->fs_info,
5018 "label is too long, return the first %zu bytes", --len);
5021 ret = copy_to_user(arg, label, len);
5023 return ret ? -EFAULT : 0;
5026 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5028 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5029 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5030 struct btrfs_trans_handle *trans;
5031 char label[BTRFS_LABEL_SIZE];
5034 if (!capable(CAP_SYS_ADMIN))
5037 if (copy_from_user(label, arg, sizeof(label)))
5040 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
5041 btrfs_err(root->fs_info, "unable to set label with more than %d bytes",
5042 BTRFS_LABEL_SIZE - 1);
5046 ret = mnt_want_write_file(file);
5050 trans = btrfs_start_transaction(root, 0);
5051 if (IS_ERR(trans)) {
5052 ret = PTR_ERR(trans);
5056 spin_lock(&root->fs_info->super_lock);
5057 strcpy(super_block->label, label);
5058 spin_unlock(&root->fs_info->super_lock);
5059 ret = btrfs_commit_transaction(trans, root);
5062 mnt_drop_write_file(file);
5066 #define INIT_FEATURE_FLAGS(suffix) \
5067 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5068 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5069 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5071 static int btrfs_ioctl_get_supported_features(struct file *file,
5074 static struct btrfs_ioctl_feature_flags features[3] = {
5075 INIT_FEATURE_FLAGS(SUPP),
5076 INIT_FEATURE_FLAGS(SAFE_SET),
5077 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5080 if (copy_to_user(arg, &features, sizeof(features)))
5086 static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5088 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5089 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5090 struct btrfs_ioctl_feature_flags features;
5092 features.compat_flags = btrfs_super_compat_flags(super_block);
5093 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5094 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5096 if (copy_to_user(arg, &features, sizeof(features)))
5102 static int check_feature_bits(struct btrfs_root *root,
5103 enum btrfs_feature_set set,
5104 u64 change_mask, u64 flags, u64 supported_flags,
5105 u64 safe_set, u64 safe_clear)
5107 const char *type = btrfs_feature_set_names[set];
5109 u64 disallowed, unsupported;
5110 u64 set_mask = flags & change_mask;
5111 u64 clear_mask = ~flags & change_mask;
5113 unsupported = set_mask & ~supported_flags;
5115 names = btrfs_printable_features(set, unsupported);
5117 btrfs_warn(root->fs_info,
5118 "this kernel does not support the %s feature bit%s",
5119 names, strchr(names, ',') ? "s" : "");
5122 btrfs_warn(root->fs_info,
5123 "this kernel does not support %s bits 0x%llx",
5128 disallowed = set_mask & ~safe_set;
5130 names = btrfs_printable_features(set, disallowed);
5132 btrfs_warn(root->fs_info,
5133 "can't set the %s feature bit%s while mounted",
5134 names, strchr(names, ',') ? "s" : "");
5137 btrfs_warn(root->fs_info,
5138 "can't set %s bits 0x%llx while mounted",
5143 disallowed = clear_mask & ~safe_clear;
5145 names = btrfs_printable_features(set, disallowed);
5147 btrfs_warn(root->fs_info,
5148 "can't clear the %s feature bit%s while mounted",
5149 names, strchr(names, ',') ? "s" : "");
5152 btrfs_warn(root->fs_info,
5153 "can't clear %s bits 0x%llx while mounted",
5161 #define check_feature(root, change_mask, flags, mask_base) \
5162 check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
5163 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5164 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5165 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5167 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5169 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5170 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5171 struct btrfs_ioctl_feature_flags flags[2];
5172 struct btrfs_trans_handle *trans;
5176 if (!capable(CAP_SYS_ADMIN))
5179 if (copy_from_user(flags, arg, sizeof(flags)))
5183 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5184 !flags[0].incompat_flags)
5187 ret = check_feature(root, flags[0].compat_flags,
5188 flags[1].compat_flags, COMPAT);
5192 ret = check_feature(root, flags[0].compat_ro_flags,
5193 flags[1].compat_ro_flags, COMPAT_RO);
5197 ret = check_feature(root, flags[0].incompat_flags,
5198 flags[1].incompat_flags, INCOMPAT);
5202 trans = btrfs_start_transaction(root, 0);
5204 return PTR_ERR(trans);
5206 spin_lock(&root->fs_info->super_lock);
5207 newflags = btrfs_super_compat_flags(super_block);
5208 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5209 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5210 btrfs_set_super_compat_flags(super_block, newflags);
5212 newflags = btrfs_super_compat_ro_flags(super_block);
5213 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5214 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5215 btrfs_set_super_compat_ro_flags(super_block, newflags);
5217 newflags = btrfs_super_incompat_flags(super_block);
5218 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5219 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5220 btrfs_set_super_incompat_flags(super_block, newflags);
5221 spin_unlock(&root->fs_info->super_lock);
5223 return btrfs_commit_transaction(trans, root);
5226 long btrfs_ioctl(struct file *file, unsigned int
5227 cmd, unsigned long arg)
5229 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5230 void __user *argp = (void __user *)arg;
5233 case FS_IOC_GETFLAGS:
5234 return btrfs_ioctl_getflags(file, argp);
5235 case FS_IOC_SETFLAGS:
5236 return btrfs_ioctl_setflags(file, argp);
5237 case FS_IOC_GETVERSION:
5238 return btrfs_ioctl_getversion(file, argp);
5240 return btrfs_ioctl_fitrim(file, argp);
5241 case BTRFS_IOC_SNAP_CREATE:
5242 return btrfs_ioctl_snap_create(file, argp, 0);
5243 case BTRFS_IOC_SNAP_CREATE_V2:
5244 return btrfs_ioctl_snap_create_v2(file, argp, 0);
5245 case BTRFS_IOC_SUBVOL_CREATE:
5246 return btrfs_ioctl_snap_create(file, argp, 1);
5247 case BTRFS_IOC_SUBVOL_CREATE_V2:
5248 return btrfs_ioctl_snap_create_v2(file, argp, 1);
5249 case BTRFS_IOC_SNAP_DESTROY:
5250 return btrfs_ioctl_snap_destroy(file, argp);
5251 case BTRFS_IOC_SUBVOL_GETFLAGS:
5252 return btrfs_ioctl_subvol_getflags(file, argp);
5253 case BTRFS_IOC_SUBVOL_SETFLAGS:
5254 return btrfs_ioctl_subvol_setflags(file, argp);
5255 case BTRFS_IOC_DEFAULT_SUBVOL:
5256 return btrfs_ioctl_default_subvol(file, argp);
5257 case BTRFS_IOC_DEFRAG:
5258 return btrfs_ioctl_defrag(file, NULL);
5259 case BTRFS_IOC_DEFRAG_RANGE:
5260 return btrfs_ioctl_defrag(file, argp);
5261 case BTRFS_IOC_RESIZE:
5262 return btrfs_ioctl_resize(file, argp);
5263 case BTRFS_IOC_ADD_DEV:
5264 return btrfs_ioctl_add_dev(root, argp);
5265 case BTRFS_IOC_RM_DEV:
5266 return btrfs_ioctl_rm_dev(file, argp);
5267 case BTRFS_IOC_FS_INFO:
5268 return btrfs_ioctl_fs_info(root, argp);
5269 case BTRFS_IOC_DEV_INFO:
5270 return btrfs_ioctl_dev_info(root, argp);
5271 case BTRFS_IOC_BALANCE:
5272 return btrfs_ioctl_balance(file, NULL);
5273 case BTRFS_IOC_CLONE:
5274 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
5275 case BTRFS_IOC_CLONE_RANGE:
5276 return btrfs_ioctl_clone_range(file, argp);
5277 case BTRFS_IOC_TRANS_START:
5278 return btrfs_ioctl_trans_start(file);
5279 case BTRFS_IOC_TRANS_END:
5280 return btrfs_ioctl_trans_end(file);
5281 case BTRFS_IOC_TREE_SEARCH:
5282 return btrfs_ioctl_tree_search(file, argp);
5283 case BTRFS_IOC_TREE_SEARCH_V2:
5284 return btrfs_ioctl_tree_search_v2(file, argp);
5285 case BTRFS_IOC_INO_LOOKUP:
5286 return btrfs_ioctl_ino_lookup(file, argp);
5287 case BTRFS_IOC_INO_PATHS:
5288 return btrfs_ioctl_ino_to_path(root, argp);
5289 case BTRFS_IOC_LOGICAL_INO:
5290 return btrfs_ioctl_logical_to_ino(root, argp);
5291 case BTRFS_IOC_SPACE_INFO:
5292 return btrfs_ioctl_space_info(root, argp);
5293 case BTRFS_IOC_SYNC: {
5296 ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1);
5299 ret = btrfs_sync_fs(file->f_dentry->d_sb, 1);
5301 * The transaction thread may want to do more work,
5302 * namely it pokes the cleaner ktread that will start
5303 * processing uncleaned subvols.
5305 wake_up_process(root->fs_info->transaction_kthread);
5308 case BTRFS_IOC_START_SYNC:
5309 return btrfs_ioctl_start_sync(root, argp);
5310 case BTRFS_IOC_WAIT_SYNC:
5311 return btrfs_ioctl_wait_sync(root, argp);
5312 case BTRFS_IOC_SCRUB:
5313 return btrfs_ioctl_scrub(file, argp);
5314 case BTRFS_IOC_SCRUB_CANCEL:
5315 return btrfs_ioctl_scrub_cancel(root, argp);
5316 case BTRFS_IOC_SCRUB_PROGRESS:
5317 return btrfs_ioctl_scrub_progress(root, argp);
5318 case BTRFS_IOC_BALANCE_V2:
5319 return btrfs_ioctl_balance(file, argp);
5320 case BTRFS_IOC_BALANCE_CTL:
5321 return btrfs_ioctl_balance_ctl(root, arg);
5322 case BTRFS_IOC_BALANCE_PROGRESS:
5323 return btrfs_ioctl_balance_progress(root, argp);
5324 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5325 return btrfs_ioctl_set_received_subvol(file, argp);
5327 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5328 return btrfs_ioctl_set_received_subvol_32(file, argp);
5330 case BTRFS_IOC_SEND:
5331 return btrfs_ioctl_send(file, argp);
5332 case BTRFS_IOC_GET_DEV_STATS:
5333 return btrfs_ioctl_get_dev_stats(root, argp);
5334 case BTRFS_IOC_QUOTA_CTL:
5335 return btrfs_ioctl_quota_ctl(file, argp);
5336 case BTRFS_IOC_QGROUP_ASSIGN:
5337 return btrfs_ioctl_qgroup_assign(file, argp);
5338 case BTRFS_IOC_QGROUP_CREATE:
5339 return btrfs_ioctl_qgroup_create(file, argp);
5340 case BTRFS_IOC_QGROUP_LIMIT:
5341 return btrfs_ioctl_qgroup_limit(file, argp);
5342 case BTRFS_IOC_QUOTA_RESCAN:
5343 return btrfs_ioctl_quota_rescan(file, argp);
5344 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5345 return btrfs_ioctl_quota_rescan_status(file, argp);
5346 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5347 return btrfs_ioctl_quota_rescan_wait(file, argp);
5348 case BTRFS_IOC_DEV_REPLACE:
5349 return btrfs_ioctl_dev_replace(root, argp);
5350 case BTRFS_IOC_GET_FSLABEL:
5351 return btrfs_ioctl_get_fslabel(file, argp);
5352 case BTRFS_IOC_SET_FSLABEL:
5353 return btrfs_ioctl_set_fslabel(file, argp);
5354 case BTRFS_IOC_FILE_EXTENT_SAME:
5355 return btrfs_ioctl_file_extent_same(file, argp);
5356 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5357 return btrfs_ioctl_get_supported_features(file, argp);
5358 case BTRFS_IOC_GET_FEATURES:
5359 return btrfs_ioctl_get_features(file, argp);
5360 case BTRFS_IOC_SET_FEATURES:
5361 return btrfs_ioctl_set_features(file, argp);