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 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
718 ret = PTR_ERR(inode);
722 d_instantiate(dentry, inode);
725 btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
726 &pending_snapshot->block_rsv,
727 pending_snapshot->qgroup_reserved);
729 kfree(pending_snapshot);
731 atomic_dec(&root->will_be_snapshoted);
735 /* copy of check_sticky in fs/namei.c()
736 * It's inline, so penalty for filesystems that don't use sticky bit is
739 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
741 kuid_t fsuid = current_fsuid();
743 if (!(dir->i_mode & S_ISVTX))
745 if (uid_eq(inode->i_uid, fsuid))
747 if (uid_eq(dir->i_uid, fsuid))
749 return !capable(CAP_FOWNER);
752 /* copy of may_delete in fs/namei.c()
753 * Check whether we can remove a link victim from directory dir, check
754 * whether the type of victim is right.
755 * 1. We can't do it if dir is read-only (done in permission())
756 * 2. We should have write and exec permissions on dir
757 * 3. We can't remove anything from append-only dir
758 * 4. We can't do anything with immutable dir (done in permission())
759 * 5. If the sticky bit on dir is set we should either
760 * a. be owner of dir, or
761 * b. be owner of victim, or
762 * c. have CAP_FOWNER capability
763 * 6. If the victim is append-only or immutable we can't do antyhing with
764 * links pointing to it.
765 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
766 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
767 * 9. We can't remove a root or mountpoint.
768 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
769 * nfs_async_unlink().
772 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
776 if (!victim->d_inode)
779 BUG_ON(victim->d_parent->d_inode != dir);
780 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
782 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
787 if (btrfs_check_sticky(dir, victim->d_inode)||
788 IS_APPEND(victim->d_inode)||
789 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
792 if (!S_ISDIR(victim->d_inode->i_mode))
796 } else if (S_ISDIR(victim->d_inode->i_mode))
800 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
805 /* copy of may_create in fs/namei.c() */
806 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
812 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
816 * Create a new subvolume below @parent. This is largely modeled after
817 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
818 * inside this filesystem so it's quite a bit simpler.
820 static noinline int btrfs_mksubvol(struct path *parent,
821 char *name, int namelen,
822 struct btrfs_root *snap_src,
823 u64 *async_transid, bool readonly,
824 struct btrfs_qgroup_inherit *inherit)
826 struct inode *dir = parent->dentry->d_inode;
827 struct dentry *dentry;
830 error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
834 dentry = lookup_one_len(name, parent->dentry, namelen);
835 error = PTR_ERR(dentry);
843 error = btrfs_may_create(dir, dentry);
848 * even if this name doesn't exist, we may get hash collisions.
849 * check for them now when we can safely fail
851 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
857 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
859 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
863 error = create_snapshot(snap_src, dir, dentry, name, namelen,
864 async_transid, readonly, inherit);
866 error = create_subvol(dir, dentry, name, namelen,
867 async_transid, inherit);
870 fsnotify_mkdir(dir, dentry);
872 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
876 mutex_unlock(&dir->i_mutex);
881 * When we're defragging a range, we don't want to kick it off again
882 * if it is really just waiting for delalloc to send it down.
883 * If we find a nice big extent or delalloc range for the bytes in the
884 * file you want to defrag, we return 0 to let you know to skip this
887 static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
889 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
890 struct extent_map *em = NULL;
891 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
894 read_lock(&em_tree->lock);
895 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
896 read_unlock(&em_tree->lock);
899 end = extent_map_end(em);
901 if (end - offset > thresh)
904 /* if we already have a nice delalloc here, just stop */
906 end = count_range_bits(io_tree, &offset, offset + thresh,
907 thresh, EXTENT_DELALLOC, 1);
914 * helper function to walk through a file and find extents
915 * newer than a specific transid, and smaller than thresh.
917 * This is used by the defragging code to find new and small
920 static int find_new_extents(struct btrfs_root *root,
921 struct inode *inode, u64 newer_than,
922 u64 *off, u32 thresh)
924 struct btrfs_path *path;
925 struct btrfs_key min_key;
926 struct extent_buffer *leaf;
927 struct btrfs_file_extent_item *extent;
930 u64 ino = btrfs_ino(inode);
932 path = btrfs_alloc_path();
936 min_key.objectid = ino;
937 min_key.type = BTRFS_EXTENT_DATA_KEY;
938 min_key.offset = *off;
941 ret = btrfs_search_forward(root, &min_key, path, newer_than);
945 if (min_key.objectid != ino)
947 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
950 leaf = path->nodes[0];
951 extent = btrfs_item_ptr(leaf, path->slots[0],
952 struct btrfs_file_extent_item);
954 type = btrfs_file_extent_type(leaf, extent);
955 if (type == BTRFS_FILE_EXTENT_REG &&
956 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
957 check_defrag_in_cache(inode, min_key.offset, thresh)) {
958 *off = min_key.offset;
959 btrfs_free_path(path);
964 if (path->slots[0] < btrfs_header_nritems(leaf)) {
965 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
969 if (min_key.offset == (u64)-1)
973 btrfs_release_path(path);
976 btrfs_free_path(path);
980 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
982 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
983 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
984 struct extent_map *em;
985 u64 len = PAGE_CACHE_SIZE;
988 * hopefully we have this extent in the tree already, try without
989 * the full extent lock
991 read_lock(&em_tree->lock);
992 em = lookup_extent_mapping(em_tree, start, len);
993 read_unlock(&em_tree->lock);
996 struct extent_state *cached = NULL;
997 u64 end = start + len - 1;
999 /* get the big lock and read metadata off disk */
1000 lock_extent_bits(io_tree, start, end, 0, &cached);
1001 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
1002 unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS);
1011 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1013 struct extent_map *next;
1016 /* this is the last extent */
1017 if (em->start + em->len >= i_size_read(inode))
1020 next = defrag_lookup_extent(inode, em->start + em->len);
1021 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1023 else if ((em->block_start + em->block_len == next->block_start) &&
1024 (em->block_len > 128 * 1024 && next->block_len > 128 * 1024))
1027 free_extent_map(next);
1031 static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
1032 u64 *last_len, u64 *skip, u64 *defrag_end,
1035 struct extent_map *em;
1037 bool next_mergeable = true;
1040 * make sure that once we start defragging an extent, we keep on
1043 if (start < *defrag_end)
1048 em = defrag_lookup_extent(inode, start);
1052 /* this will cover holes, and inline extents */
1053 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1058 next_mergeable = defrag_check_next_extent(inode, em);
1060 * we hit a real extent, if it is big or the next extent is not a
1061 * real extent, don't bother defragging it
1063 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1064 (em->len >= thresh || !next_mergeable))
1068 * last_len ends up being a counter of how many bytes we've defragged.
1069 * every time we choose not to defrag an extent, we reset *last_len
1070 * so that the next tiny extent will force a defrag.
1072 * The end result of this is that tiny extents before a single big
1073 * extent will force at least part of that big extent to be defragged.
1076 *defrag_end = extent_map_end(em);
1079 *skip = extent_map_end(em);
1083 free_extent_map(em);
1088 * it doesn't do much good to defrag one or two pages
1089 * at a time. This pulls in a nice chunk of pages
1090 * to COW and defrag.
1092 * It also makes sure the delalloc code has enough
1093 * dirty data to avoid making new small extents as part
1096 * It's a good idea to start RA on this range
1097 * before calling this.
1099 static int cluster_pages_for_defrag(struct inode *inode,
1100 struct page **pages,
1101 unsigned long start_index,
1102 unsigned long num_pages)
1104 unsigned long file_end;
1105 u64 isize = i_size_read(inode);
1112 struct btrfs_ordered_extent *ordered;
1113 struct extent_state *cached_state = NULL;
1114 struct extent_io_tree *tree;
1115 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1117 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1118 if (!isize || start_index > file_end)
1121 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1123 ret = btrfs_delalloc_reserve_space(inode,
1124 page_cnt << PAGE_CACHE_SHIFT);
1128 tree = &BTRFS_I(inode)->io_tree;
1130 /* step one, lock all the pages */
1131 for (i = 0; i < page_cnt; i++) {
1134 page = find_or_create_page(inode->i_mapping,
1135 start_index + i, mask);
1139 page_start = page_offset(page);
1140 page_end = page_start + PAGE_CACHE_SIZE - 1;
1142 lock_extent_bits(tree, page_start, page_end,
1144 ordered = btrfs_lookup_ordered_extent(inode,
1146 unlock_extent_cached(tree, page_start, page_end,
1147 &cached_state, GFP_NOFS);
1152 btrfs_start_ordered_extent(inode, ordered, 1);
1153 btrfs_put_ordered_extent(ordered);
1156 * we unlocked the page above, so we need check if
1157 * it was released or not.
1159 if (page->mapping != inode->i_mapping) {
1161 page_cache_release(page);
1166 if (!PageUptodate(page)) {
1167 btrfs_readpage(NULL, page);
1169 if (!PageUptodate(page)) {
1171 page_cache_release(page);
1177 if (page->mapping != inode->i_mapping) {
1179 page_cache_release(page);
1189 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1193 * so now we have a nice long stream of locked
1194 * and up to date pages, lets wait on them
1196 for (i = 0; i < i_done; i++)
1197 wait_on_page_writeback(pages[i]);
1199 page_start = page_offset(pages[0]);
1200 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1202 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1203 page_start, page_end - 1, 0, &cached_state);
1204 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1205 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1206 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1207 &cached_state, GFP_NOFS);
1209 if (i_done != page_cnt) {
1210 spin_lock(&BTRFS_I(inode)->lock);
1211 BTRFS_I(inode)->outstanding_extents++;
1212 spin_unlock(&BTRFS_I(inode)->lock);
1213 btrfs_delalloc_release_space(inode,
1214 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1218 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1219 &cached_state, GFP_NOFS);
1221 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1222 page_start, page_end - 1, &cached_state,
1225 for (i = 0; i < i_done; i++) {
1226 clear_page_dirty_for_io(pages[i]);
1227 ClearPageChecked(pages[i]);
1228 set_page_extent_mapped(pages[i]);
1229 set_page_dirty(pages[i]);
1230 unlock_page(pages[i]);
1231 page_cache_release(pages[i]);
1235 for (i = 0; i < i_done; i++) {
1236 unlock_page(pages[i]);
1237 page_cache_release(pages[i]);
1239 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1244 int btrfs_defrag_file(struct inode *inode, struct file *file,
1245 struct btrfs_ioctl_defrag_range_args *range,
1246 u64 newer_than, unsigned long max_to_defrag)
1248 struct btrfs_root *root = BTRFS_I(inode)->root;
1249 struct file_ra_state *ra = NULL;
1250 unsigned long last_index;
1251 u64 isize = i_size_read(inode);
1255 u64 newer_off = range->start;
1257 unsigned long ra_index = 0;
1259 int defrag_count = 0;
1260 int compress_type = BTRFS_COMPRESS_ZLIB;
1261 u32 extent_thresh = range->extent_thresh;
1262 unsigned long max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1263 unsigned long cluster = max_cluster;
1264 u64 new_align = ~((u64)128 * 1024 - 1);
1265 struct page **pages = NULL;
1270 if (range->start >= isize)
1273 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1274 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1276 if (range->compress_type)
1277 compress_type = range->compress_type;
1280 if (extent_thresh == 0)
1281 extent_thresh = 256 * 1024;
1284 * if we were not given a file, allocate a readahead
1288 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1291 file_ra_state_init(ra, inode->i_mapping);
1296 pages = kmalloc_array(max_cluster, sizeof(struct page *),
1303 /* find the last page to defrag */
1304 if (range->start + range->len > range->start) {
1305 last_index = min_t(u64, isize - 1,
1306 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1308 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1312 ret = find_new_extents(root, inode, newer_than,
1313 &newer_off, 64 * 1024);
1315 range->start = newer_off;
1317 * we always align our defrag to help keep
1318 * the extents in the file evenly spaced
1320 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1324 i = range->start >> PAGE_CACHE_SHIFT;
1327 max_to_defrag = last_index + 1;
1330 * make writeback starts from i, so the defrag range can be
1331 * written sequentially.
1333 if (i < inode->i_mapping->writeback_index)
1334 inode->i_mapping->writeback_index = i;
1336 while (i <= last_index && defrag_count < max_to_defrag &&
1337 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_CACHE_SIZE))) {
1339 * make sure we stop running if someone unmounts
1342 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1345 if (btrfs_defrag_cancelled(root->fs_info)) {
1346 printk(KERN_DEBUG "BTRFS: defrag_file cancelled\n");
1351 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1352 extent_thresh, &last_len, &skip,
1353 &defrag_end, range->flags &
1354 BTRFS_DEFRAG_RANGE_COMPRESS)) {
1357 * the should_defrag function tells us how much to skip
1358 * bump our counter by the suggested amount
1360 next = DIV_ROUND_UP(skip, PAGE_CACHE_SIZE);
1361 i = max(i + 1, next);
1366 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1367 PAGE_CACHE_SHIFT) - i;
1368 cluster = min(cluster, max_cluster);
1370 cluster = max_cluster;
1373 if (i + cluster > ra_index) {
1374 ra_index = max(i, ra_index);
1375 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1377 ra_index += max_cluster;
1380 mutex_lock(&inode->i_mutex);
1381 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1382 BTRFS_I(inode)->force_compress = compress_type;
1383 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1385 mutex_unlock(&inode->i_mutex);
1389 defrag_count += ret;
1390 balance_dirty_pages_ratelimited(inode->i_mapping);
1391 mutex_unlock(&inode->i_mutex);
1394 if (newer_off == (u64)-1)
1400 newer_off = max(newer_off + 1,
1401 (u64)i << PAGE_CACHE_SHIFT);
1403 ret = find_new_extents(root, inode,
1404 newer_than, &newer_off,
1407 range->start = newer_off;
1408 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1415 last_len += ret << PAGE_CACHE_SHIFT;
1423 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1424 filemap_flush(inode->i_mapping);
1425 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1426 &BTRFS_I(inode)->runtime_flags))
1427 filemap_flush(inode->i_mapping);
1430 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1431 /* the filemap_flush will queue IO into the worker threads, but
1432 * we have to make sure the IO is actually started and that
1433 * ordered extents get created before we return
1435 atomic_inc(&root->fs_info->async_submit_draining);
1436 while (atomic_read(&root->fs_info->nr_async_submits) ||
1437 atomic_read(&root->fs_info->async_delalloc_pages)) {
1438 wait_event(root->fs_info->async_submit_wait,
1439 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1440 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1442 atomic_dec(&root->fs_info->async_submit_draining);
1445 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1446 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1452 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1453 mutex_lock(&inode->i_mutex);
1454 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1455 mutex_unlock(&inode->i_mutex);
1463 static noinline int btrfs_ioctl_resize(struct file *file,
1469 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
1470 struct btrfs_ioctl_vol_args *vol_args;
1471 struct btrfs_trans_handle *trans;
1472 struct btrfs_device *device = NULL;
1475 char *devstr = NULL;
1479 if (!capable(CAP_SYS_ADMIN))
1482 ret = mnt_want_write_file(file);
1486 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1488 mnt_drop_write_file(file);
1489 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1492 mutex_lock(&root->fs_info->volume_mutex);
1493 vol_args = memdup_user(arg, sizeof(*vol_args));
1494 if (IS_ERR(vol_args)) {
1495 ret = PTR_ERR(vol_args);
1499 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1501 sizestr = vol_args->name;
1502 devstr = strchr(sizestr, ':');
1504 sizestr = devstr + 1;
1506 devstr = vol_args->name;
1507 ret = kstrtoull(devstr, 10, &devid);
1514 btrfs_info(root->fs_info, "resizing devid %llu", devid);
1517 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
1519 btrfs_info(root->fs_info, "resizer unable to find device %llu",
1525 if (!device->writeable) {
1526 btrfs_info(root->fs_info,
1527 "resizer unable to apply on readonly device %llu",
1533 if (!strcmp(sizestr, "max"))
1534 new_size = device->bdev->bd_inode->i_size;
1536 if (sizestr[0] == '-') {
1539 } else if (sizestr[0] == '+') {
1543 new_size = memparse(sizestr, &retptr);
1544 if (*retptr != '\0' || new_size == 0) {
1550 if (device->is_tgtdev_for_dev_replace) {
1555 old_size = btrfs_device_get_total_bytes(device);
1558 if (new_size > old_size) {
1562 new_size = old_size - new_size;
1563 } else if (mod > 0) {
1564 if (new_size > ULLONG_MAX - old_size) {
1568 new_size = old_size + new_size;
1571 if (new_size < 256 * 1024 * 1024) {
1575 if (new_size > device->bdev->bd_inode->i_size) {
1580 do_div(new_size, root->sectorsize);
1581 new_size *= root->sectorsize;
1583 printk_in_rcu(KERN_INFO "BTRFS: new size for %s is %llu\n",
1584 rcu_str_deref(device->name), new_size);
1586 if (new_size > old_size) {
1587 trans = btrfs_start_transaction(root, 0);
1588 if (IS_ERR(trans)) {
1589 ret = PTR_ERR(trans);
1592 ret = btrfs_grow_device(trans, device, new_size);
1593 btrfs_commit_transaction(trans, root);
1594 } else if (new_size < old_size) {
1595 ret = btrfs_shrink_device(device, new_size);
1596 } /* equal, nothing need to do */
1601 mutex_unlock(&root->fs_info->volume_mutex);
1602 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
1603 mnt_drop_write_file(file);
1607 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1608 char *name, unsigned long fd, int subvol,
1609 u64 *transid, bool readonly,
1610 struct btrfs_qgroup_inherit *inherit)
1615 ret = mnt_want_write_file(file);
1619 namelen = strlen(name);
1620 if (strchr(name, '/')) {
1622 goto out_drop_write;
1625 if (name[0] == '.' &&
1626 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1628 goto out_drop_write;
1632 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1633 NULL, transid, readonly, inherit);
1635 struct fd src = fdget(fd);
1636 struct inode *src_inode;
1639 goto out_drop_write;
1642 src_inode = file_inode(src.file);
1643 if (src_inode->i_sb != file_inode(file)->i_sb) {
1644 btrfs_info(BTRFS_I(src_inode)->root->fs_info,
1645 "Snapshot src from another FS");
1647 } else if (!inode_owner_or_capable(src_inode)) {
1649 * Subvolume creation is not restricted, but snapshots
1650 * are limited to own subvolumes only
1654 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1655 BTRFS_I(src_inode)->root,
1656 transid, readonly, inherit);
1661 mnt_drop_write_file(file);
1666 static noinline int btrfs_ioctl_snap_create(struct file *file,
1667 void __user *arg, int subvol)
1669 struct btrfs_ioctl_vol_args *vol_args;
1672 vol_args = memdup_user(arg, sizeof(*vol_args));
1673 if (IS_ERR(vol_args))
1674 return PTR_ERR(vol_args);
1675 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1677 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1678 vol_args->fd, subvol,
1685 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1686 void __user *arg, int subvol)
1688 struct btrfs_ioctl_vol_args_v2 *vol_args;
1692 bool readonly = false;
1693 struct btrfs_qgroup_inherit *inherit = NULL;
1695 vol_args = memdup_user(arg, sizeof(*vol_args));
1696 if (IS_ERR(vol_args))
1697 return PTR_ERR(vol_args);
1698 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1700 if (vol_args->flags &
1701 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1702 BTRFS_SUBVOL_QGROUP_INHERIT)) {
1707 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1709 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1711 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1712 if (vol_args->size > PAGE_CACHE_SIZE) {
1716 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1717 if (IS_ERR(inherit)) {
1718 ret = PTR_ERR(inherit);
1723 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1724 vol_args->fd, subvol, ptr,
1729 if (ptr && copy_to_user(arg +
1730 offsetof(struct btrfs_ioctl_vol_args_v2,
1742 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1745 struct inode *inode = file_inode(file);
1746 struct btrfs_root *root = BTRFS_I(inode)->root;
1750 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1753 down_read(&root->fs_info->subvol_sem);
1754 if (btrfs_root_readonly(root))
1755 flags |= BTRFS_SUBVOL_RDONLY;
1756 up_read(&root->fs_info->subvol_sem);
1758 if (copy_to_user(arg, &flags, sizeof(flags)))
1764 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1767 struct inode *inode = file_inode(file);
1768 struct btrfs_root *root = BTRFS_I(inode)->root;
1769 struct btrfs_trans_handle *trans;
1774 if (!inode_owner_or_capable(inode))
1777 ret = mnt_want_write_file(file);
1781 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1783 goto out_drop_write;
1786 if (copy_from_user(&flags, arg, sizeof(flags))) {
1788 goto out_drop_write;
1791 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1793 goto out_drop_write;
1796 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1798 goto out_drop_write;
1801 down_write(&root->fs_info->subvol_sem);
1804 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1807 root_flags = btrfs_root_flags(&root->root_item);
1808 if (flags & BTRFS_SUBVOL_RDONLY) {
1809 btrfs_set_root_flags(&root->root_item,
1810 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1813 * Block RO -> RW transition if this subvolume is involved in
1816 spin_lock(&root->root_item_lock);
1817 if (root->send_in_progress == 0) {
1818 btrfs_set_root_flags(&root->root_item,
1819 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1820 spin_unlock(&root->root_item_lock);
1822 spin_unlock(&root->root_item_lock);
1823 btrfs_warn(root->fs_info,
1824 "Attempt to set subvolume %llu read-write during send",
1825 root->root_key.objectid);
1831 trans = btrfs_start_transaction(root, 1);
1832 if (IS_ERR(trans)) {
1833 ret = PTR_ERR(trans);
1837 ret = btrfs_update_root(trans, root->fs_info->tree_root,
1838 &root->root_key, &root->root_item);
1840 btrfs_commit_transaction(trans, root);
1843 btrfs_set_root_flags(&root->root_item, root_flags);
1845 up_write(&root->fs_info->subvol_sem);
1847 mnt_drop_write_file(file);
1853 * helper to check if the subvolume references other subvolumes
1855 static noinline int may_destroy_subvol(struct btrfs_root *root)
1857 struct btrfs_path *path;
1858 struct btrfs_dir_item *di;
1859 struct btrfs_key key;
1863 path = btrfs_alloc_path();
1867 /* Make sure this root isn't set as the default subvol */
1868 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1869 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1870 dir_id, "default", 7, 0);
1871 if (di && !IS_ERR(di)) {
1872 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1873 if (key.objectid == root->root_key.objectid) {
1875 btrfs_err(root->fs_info, "deleting default subvolume "
1876 "%llu is not allowed", key.objectid);
1879 btrfs_release_path(path);
1882 key.objectid = root->root_key.objectid;
1883 key.type = BTRFS_ROOT_REF_KEY;
1884 key.offset = (u64)-1;
1886 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1893 if (path->slots[0] > 0) {
1895 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1896 if (key.objectid == root->root_key.objectid &&
1897 key.type == BTRFS_ROOT_REF_KEY)
1901 btrfs_free_path(path);
1905 static noinline int key_in_sk(struct btrfs_key *key,
1906 struct btrfs_ioctl_search_key *sk)
1908 struct btrfs_key test;
1911 test.objectid = sk->min_objectid;
1912 test.type = sk->min_type;
1913 test.offset = sk->min_offset;
1915 ret = btrfs_comp_cpu_keys(key, &test);
1919 test.objectid = sk->max_objectid;
1920 test.type = sk->max_type;
1921 test.offset = sk->max_offset;
1923 ret = btrfs_comp_cpu_keys(key, &test);
1929 static noinline int copy_to_sk(struct btrfs_root *root,
1930 struct btrfs_path *path,
1931 struct btrfs_key *key,
1932 struct btrfs_ioctl_search_key *sk,
1935 unsigned long *sk_offset,
1939 struct extent_buffer *leaf;
1940 struct btrfs_ioctl_search_header sh;
1941 unsigned long item_off;
1942 unsigned long item_len;
1948 leaf = path->nodes[0];
1949 slot = path->slots[0];
1950 nritems = btrfs_header_nritems(leaf);
1952 if (btrfs_header_generation(leaf) > sk->max_transid) {
1956 found_transid = btrfs_header_generation(leaf);
1958 for (i = slot; i < nritems; i++) {
1959 item_off = btrfs_item_ptr_offset(leaf, i);
1960 item_len = btrfs_item_size_nr(leaf, i);
1962 btrfs_item_key_to_cpu(leaf, key, i);
1963 if (!key_in_sk(key, sk))
1966 if (sizeof(sh) + item_len > *buf_size) {
1973 * return one empty item back for v1, which does not
1977 *buf_size = sizeof(sh) + item_len;
1982 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
1987 sh.objectid = key->objectid;
1988 sh.offset = key->offset;
1989 sh.type = key->type;
1991 sh.transid = found_transid;
1993 /* copy search result header */
1994 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
1999 *sk_offset += sizeof(sh);
2002 char __user *up = ubuf + *sk_offset;
2004 if (read_extent_buffer_to_user(leaf, up,
2005 item_off, item_len)) {
2010 *sk_offset += item_len;
2014 if (ret) /* -EOVERFLOW from above */
2017 if (*num_found >= sk->nr_items) {
2024 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
2026 else if (key->type < (u8)-1 && key->type < sk->max_type) {
2029 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
2037 * 0: all items from this leaf copied, continue with next
2038 * 1: * more items can be copied, but unused buffer is too small
2039 * * all items were found
2040 * Either way, it will stops the loop which iterates to the next
2042 * -EOVERFLOW: item was to large for buffer
2043 * -EFAULT: could not copy extent buffer back to userspace
2048 static noinline int search_ioctl(struct inode *inode,
2049 struct btrfs_ioctl_search_key *sk,
2053 struct btrfs_root *root;
2054 struct btrfs_key key;
2055 struct btrfs_path *path;
2056 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
2059 unsigned long sk_offset = 0;
2061 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2062 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2066 path = btrfs_alloc_path();
2070 if (sk->tree_id == 0) {
2071 /* search the root of the inode that was passed */
2072 root = BTRFS_I(inode)->root;
2074 key.objectid = sk->tree_id;
2075 key.type = BTRFS_ROOT_ITEM_KEY;
2076 key.offset = (u64)-1;
2077 root = btrfs_read_fs_root_no_name(info, &key);
2079 printk(KERN_ERR "BTRFS: could not find root %llu\n",
2081 btrfs_free_path(path);
2086 key.objectid = sk->min_objectid;
2087 key.type = sk->min_type;
2088 key.offset = sk->min_offset;
2091 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2097 ret = copy_to_sk(root, path, &key, sk, buf_size, ubuf,
2098 &sk_offset, &num_found);
2099 btrfs_release_path(path);
2107 sk->nr_items = num_found;
2108 btrfs_free_path(path);
2112 static noinline int btrfs_ioctl_tree_search(struct file *file,
2115 struct btrfs_ioctl_search_args __user *uargs;
2116 struct btrfs_ioctl_search_key sk;
2117 struct inode *inode;
2121 if (!capable(CAP_SYS_ADMIN))
2124 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2126 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2129 buf_size = sizeof(uargs->buf);
2131 inode = file_inode(file);
2132 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2135 * In the origin implementation an overflow is handled by returning a
2136 * search header with a len of zero, so reset ret.
2138 if (ret == -EOVERFLOW)
2141 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2146 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2149 struct btrfs_ioctl_search_args_v2 __user *uarg;
2150 struct btrfs_ioctl_search_args_v2 args;
2151 struct inode *inode;
2154 const size_t buf_limit = 16 * 1024 * 1024;
2156 if (!capable(CAP_SYS_ADMIN))
2159 /* copy search header and buffer size */
2160 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2161 if (copy_from_user(&args, uarg, sizeof(args)))
2164 buf_size = args.buf_size;
2166 if (buf_size < sizeof(struct btrfs_ioctl_search_header))
2169 /* limit result size to 16MB */
2170 if (buf_size > buf_limit)
2171 buf_size = buf_limit;
2173 inode = file_inode(file);
2174 ret = search_ioctl(inode, &args.key, &buf_size,
2175 (char *)(&uarg->buf[0]));
2176 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2178 else if (ret == -EOVERFLOW &&
2179 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2186 * Search INODE_REFs to identify path name of 'dirid' directory
2187 * in a 'tree_id' tree. and sets path name to 'name'.
2189 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2190 u64 tree_id, u64 dirid, char *name)
2192 struct btrfs_root *root;
2193 struct btrfs_key key;
2199 struct btrfs_inode_ref *iref;
2200 struct extent_buffer *l;
2201 struct btrfs_path *path;
2203 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2208 path = btrfs_alloc_path();
2212 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
2214 key.objectid = tree_id;
2215 key.type = BTRFS_ROOT_ITEM_KEY;
2216 key.offset = (u64)-1;
2217 root = btrfs_read_fs_root_no_name(info, &key);
2219 printk(KERN_ERR "BTRFS: could not find root %llu\n", tree_id);
2224 key.objectid = dirid;
2225 key.type = BTRFS_INODE_REF_KEY;
2226 key.offset = (u64)-1;
2229 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2233 ret = btrfs_previous_item(root, path, dirid,
2234 BTRFS_INODE_REF_KEY);
2244 slot = path->slots[0];
2245 btrfs_item_key_to_cpu(l, &key, slot);
2247 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2248 len = btrfs_inode_ref_name_len(l, iref);
2250 total_len += len + 1;
2252 ret = -ENAMETOOLONG;
2257 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2259 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2262 btrfs_release_path(path);
2263 key.objectid = key.offset;
2264 key.offset = (u64)-1;
2265 dirid = key.objectid;
2267 memmove(name, ptr, total_len);
2268 name[total_len] = '\0';
2271 btrfs_free_path(path);
2275 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2278 struct btrfs_ioctl_ino_lookup_args *args;
2279 struct inode *inode;
2282 if (!capable(CAP_SYS_ADMIN))
2285 args = memdup_user(argp, sizeof(*args));
2287 return PTR_ERR(args);
2289 inode = file_inode(file);
2291 if (args->treeid == 0)
2292 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2294 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2295 args->treeid, args->objectid,
2298 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2305 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2308 struct dentry *parent = file->f_path.dentry;
2309 struct dentry *dentry;
2310 struct inode *dir = parent->d_inode;
2311 struct inode *inode;
2312 struct btrfs_root *root = BTRFS_I(dir)->root;
2313 struct btrfs_root *dest = NULL;
2314 struct btrfs_ioctl_vol_args *vol_args;
2315 struct btrfs_trans_handle *trans;
2316 struct btrfs_block_rsv block_rsv;
2318 u64 qgroup_reserved;
2323 vol_args = memdup_user(arg, sizeof(*vol_args));
2324 if (IS_ERR(vol_args))
2325 return PTR_ERR(vol_args);
2327 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2328 namelen = strlen(vol_args->name);
2329 if (strchr(vol_args->name, '/') ||
2330 strncmp(vol_args->name, "..", namelen) == 0) {
2335 err = mnt_want_write_file(file);
2340 err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2342 goto out_drop_write;
2343 dentry = lookup_one_len(vol_args->name, parent, namelen);
2344 if (IS_ERR(dentry)) {
2345 err = PTR_ERR(dentry);
2346 goto out_unlock_dir;
2349 if (!dentry->d_inode) {
2354 inode = dentry->d_inode;
2355 dest = BTRFS_I(inode)->root;
2356 if (!capable(CAP_SYS_ADMIN)) {
2358 * Regular user. Only allow this with a special mount
2359 * option, when the user has write+exec access to the
2360 * subvol root, and when rmdir(2) would have been
2363 * Note that this is _not_ check that the subvol is
2364 * empty or doesn't contain data that we wouldn't
2365 * otherwise be able to delete.
2367 * Users who want to delete empty subvols should try
2371 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2375 * Do not allow deletion if the parent dir is the same
2376 * as the dir to be deleted. That means the ioctl
2377 * must be called on the dentry referencing the root
2378 * of the subvol, not a random directory contained
2385 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2390 /* check if subvolume may be deleted by a user */
2391 err = btrfs_may_delete(dir, dentry, 1);
2395 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2400 mutex_lock(&inode->i_mutex);
2403 * Don't allow to delete a subvolume with send in progress. This is
2404 * inside the i_mutex so the error handling that has to drop the bit
2405 * again is not run concurrently.
2407 spin_lock(&dest->root_item_lock);
2408 root_flags = btrfs_root_flags(&dest->root_item);
2409 if (dest->send_in_progress == 0) {
2410 btrfs_set_root_flags(&dest->root_item,
2411 root_flags | BTRFS_ROOT_SUBVOL_DEAD);
2412 spin_unlock(&dest->root_item_lock);
2414 spin_unlock(&dest->root_item_lock);
2415 btrfs_warn(root->fs_info,
2416 "Attempt to delete subvolume %llu during send",
2417 dest->root_key.objectid);
2422 err = d_invalidate(dentry);
2426 down_write(&root->fs_info->subvol_sem);
2428 err = may_destroy_subvol(dest);
2432 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2434 * One for dir inode, two for dir entries, two for root
2437 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2438 5, &qgroup_reserved, true);
2442 trans = btrfs_start_transaction(root, 0);
2443 if (IS_ERR(trans)) {
2444 err = PTR_ERR(trans);
2447 trans->block_rsv = &block_rsv;
2448 trans->bytes_reserved = block_rsv.size;
2450 ret = btrfs_unlink_subvol(trans, root, dir,
2451 dest->root_key.objectid,
2452 dentry->d_name.name,
2453 dentry->d_name.len);
2456 btrfs_abort_transaction(trans, root, ret);
2460 btrfs_record_root_in_trans(trans, dest);
2462 memset(&dest->root_item.drop_progress, 0,
2463 sizeof(dest->root_item.drop_progress));
2464 dest->root_item.drop_level = 0;
2465 btrfs_set_root_refs(&dest->root_item, 0);
2467 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
2468 ret = btrfs_insert_orphan_item(trans,
2469 root->fs_info->tree_root,
2470 dest->root_key.objectid);
2472 btrfs_abort_transaction(trans, root, ret);
2478 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2479 dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
2480 dest->root_key.objectid);
2481 if (ret && ret != -ENOENT) {
2482 btrfs_abort_transaction(trans, root, ret);
2486 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2487 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2488 dest->root_item.received_uuid,
2489 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2490 dest->root_key.objectid);
2491 if (ret && ret != -ENOENT) {
2492 btrfs_abort_transaction(trans, root, ret);
2499 trans->block_rsv = NULL;
2500 trans->bytes_reserved = 0;
2501 ret = btrfs_end_transaction(trans, root);
2504 inode->i_flags |= S_DEAD;
2506 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
2508 up_write(&root->fs_info->subvol_sem);
2511 spin_lock(&dest->root_item_lock);
2512 root_flags = btrfs_root_flags(&dest->root_item);
2513 btrfs_set_root_flags(&dest->root_item,
2514 root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
2515 spin_unlock(&dest->root_item_lock);
2517 mutex_unlock(&inode->i_mutex);
2519 shrink_dcache_sb(root->fs_info->sb);
2520 btrfs_invalidate_inodes(dest);
2522 ASSERT(dest->send_in_progress == 0);
2525 if (dest->ino_cache_inode) {
2526 iput(dest->ino_cache_inode);
2527 dest->ino_cache_inode = NULL;
2533 mutex_unlock(&dir->i_mutex);
2535 mnt_drop_write_file(file);
2541 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2543 struct inode *inode = file_inode(file);
2544 struct btrfs_root *root = BTRFS_I(inode)->root;
2545 struct btrfs_ioctl_defrag_range_args *range;
2548 ret = mnt_want_write_file(file);
2552 if (btrfs_root_readonly(root)) {
2557 switch (inode->i_mode & S_IFMT) {
2559 if (!capable(CAP_SYS_ADMIN)) {
2563 ret = btrfs_defrag_root(root);
2566 ret = btrfs_defrag_root(root->fs_info->extent_root);
2569 if (!(file->f_mode & FMODE_WRITE)) {
2574 range = kzalloc(sizeof(*range), GFP_KERNEL);
2581 if (copy_from_user(range, argp,
2587 /* compression requires us to start the IO */
2588 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2589 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2590 range->extent_thresh = (u32)-1;
2593 /* the rest are all set to zero by kzalloc */
2594 range->len = (u64)-1;
2596 ret = btrfs_defrag_file(file_inode(file), file,
2606 mnt_drop_write_file(file);
2610 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2612 struct btrfs_ioctl_vol_args *vol_args;
2615 if (!capable(CAP_SYS_ADMIN))
2618 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2620 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2623 mutex_lock(&root->fs_info->volume_mutex);
2624 vol_args = memdup_user(arg, sizeof(*vol_args));
2625 if (IS_ERR(vol_args)) {
2626 ret = PTR_ERR(vol_args);
2630 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2631 ret = btrfs_init_new_device(root, vol_args->name);
2634 btrfs_info(root->fs_info, "disk added %s",vol_args->name);
2638 mutex_unlock(&root->fs_info->volume_mutex);
2639 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2643 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2645 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2646 struct btrfs_ioctl_vol_args *vol_args;
2649 if (!capable(CAP_SYS_ADMIN))
2652 ret = mnt_want_write_file(file);
2656 vol_args = memdup_user(arg, sizeof(*vol_args));
2657 if (IS_ERR(vol_args)) {
2658 ret = PTR_ERR(vol_args);
2662 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2664 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2666 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2670 mutex_lock(&root->fs_info->volume_mutex);
2671 ret = btrfs_rm_device(root, vol_args->name);
2672 mutex_unlock(&root->fs_info->volume_mutex);
2673 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2676 btrfs_info(root->fs_info, "disk deleted %s",vol_args->name);
2681 mnt_drop_write_file(file);
2685 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2687 struct btrfs_ioctl_fs_info_args *fi_args;
2688 struct btrfs_device *device;
2689 struct btrfs_device *next;
2690 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2693 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2697 mutex_lock(&fs_devices->device_list_mutex);
2698 fi_args->num_devices = fs_devices->num_devices;
2699 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2701 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2702 if (device->devid > fi_args->max_id)
2703 fi_args->max_id = device->devid;
2705 mutex_unlock(&fs_devices->device_list_mutex);
2707 fi_args->nodesize = root->fs_info->super_copy->nodesize;
2708 fi_args->sectorsize = root->fs_info->super_copy->sectorsize;
2709 fi_args->clone_alignment = root->fs_info->super_copy->sectorsize;
2711 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2718 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2720 struct btrfs_ioctl_dev_info_args *di_args;
2721 struct btrfs_device *dev;
2722 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2724 char *s_uuid = NULL;
2726 di_args = memdup_user(arg, sizeof(*di_args));
2727 if (IS_ERR(di_args))
2728 return PTR_ERR(di_args);
2730 if (!btrfs_is_empty_uuid(di_args->uuid))
2731 s_uuid = di_args->uuid;
2733 mutex_lock(&fs_devices->device_list_mutex);
2734 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
2741 di_args->devid = dev->devid;
2742 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2743 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
2744 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2746 struct rcu_string *name;
2749 name = rcu_dereference(dev->name);
2750 strncpy(di_args->path, name->str, sizeof(di_args->path));
2752 di_args->path[sizeof(di_args->path) - 1] = 0;
2754 di_args->path[0] = '\0';
2758 mutex_unlock(&fs_devices->device_list_mutex);
2759 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2766 static struct page *extent_same_get_page(struct inode *inode, u64 off)
2770 struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
2772 index = off >> PAGE_CACHE_SHIFT;
2774 page = grab_cache_page(inode->i_mapping, index);
2778 if (!PageUptodate(page)) {
2779 if (extent_read_full_page_nolock(tree, page, btrfs_get_extent,
2783 if (!PageUptodate(page)) {
2785 page_cache_release(page);
2794 static inline void lock_extent_range(struct inode *inode, u64 off, u64 len)
2796 /* do any pending delalloc/csum calc on src, one way or
2797 another, and lock file content */
2799 struct btrfs_ordered_extent *ordered;
2800 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2801 ordered = btrfs_lookup_first_ordered_extent(inode,
2804 ordered->file_offset + ordered->len <= off ||
2805 ordered->file_offset >= off + len) &&
2806 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2807 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2809 btrfs_put_ordered_extent(ordered);
2812 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2814 btrfs_put_ordered_extent(ordered);
2815 btrfs_wait_ordered_range(inode, off, len);
2819 static void btrfs_double_unlock(struct inode *inode1, u64 loff1,
2820 struct inode *inode2, u64 loff2, u64 len)
2822 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2823 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2825 mutex_unlock(&inode1->i_mutex);
2826 mutex_unlock(&inode2->i_mutex);
2829 static void btrfs_double_lock(struct inode *inode1, u64 loff1,
2830 struct inode *inode2, u64 loff2, u64 len)
2832 if (inode1 < inode2) {
2833 swap(inode1, inode2);
2837 mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT);
2838 lock_extent_range(inode1, loff1, len);
2839 if (inode1 != inode2) {
2840 mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD);
2841 lock_extent_range(inode2, loff2, len);
2845 static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
2846 u64 dst_loff, u64 len)
2849 struct page *src_page, *dst_page;
2850 unsigned int cmp_len = PAGE_CACHE_SIZE;
2851 void *addr, *dst_addr;
2854 if (len < PAGE_CACHE_SIZE)
2857 src_page = extent_same_get_page(src, loff);
2860 dst_page = extent_same_get_page(dst, dst_loff);
2862 page_cache_release(src_page);
2865 addr = kmap_atomic(src_page);
2866 dst_addr = kmap_atomic(dst_page);
2868 flush_dcache_page(src_page);
2869 flush_dcache_page(dst_page);
2871 if (memcmp(addr, dst_addr, cmp_len))
2872 ret = BTRFS_SAME_DATA_DIFFERS;
2874 kunmap_atomic(addr);
2875 kunmap_atomic(dst_addr);
2876 page_cache_release(src_page);
2877 page_cache_release(dst_page);
2883 dst_loff += cmp_len;
2890 static int extent_same_check_offsets(struct inode *inode, u64 off, u64 len)
2892 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
2894 if (off + len > inode->i_size || off + len < off)
2896 /* Check that we are block aligned - btrfs_clone() requires this */
2897 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
2903 static int btrfs_extent_same(struct inode *src, u64 loff, u64 len,
2904 struct inode *dst, u64 dst_loff)
2909 * btrfs_clone() can't handle extents in the same file
2910 * yet. Once that works, we can drop this check and replace it
2911 * with a check for the same inode, but overlapping extents.
2916 btrfs_double_lock(src, loff, dst, dst_loff, len);
2918 ret = extent_same_check_offsets(src, loff, len);
2922 ret = extent_same_check_offsets(dst, dst_loff, len);
2926 /* don't make the dst file partly checksummed */
2927 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2928 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
2933 ret = btrfs_cmp_data(src, loff, dst, dst_loff, len);
2935 ret = btrfs_clone(src, dst, loff, len, len, dst_loff);
2938 btrfs_double_unlock(src, loff, dst, dst_loff, len);
2943 #define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024)
2945 static long btrfs_ioctl_file_extent_same(struct file *file,
2946 struct btrfs_ioctl_same_args __user *argp)
2948 struct btrfs_ioctl_same_args *same;
2949 struct btrfs_ioctl_same_extent_info *info;
2950 struct inode *src = file_inode(file);
2956 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
2957 bool is_admin = capable(CAP_SYS_ADMIN);
2960 if (!(file->f_mode & FMODE_READ))
2963 ret = mnt_want_write_file(file);
2967 if (get_user(count, &argp->dest_count)) {
2972 size = offsetof(struct btrfs_ioctl_same_args __user, info[count]);
2974 same = memdup_user(argp, size);
2977 ret = PTR_ERR(same);
2981 off = same->logical_offset;
2985 * Limit the total length we will dedupe for each operation.
2986 * This is intended to bound the total time spent in this
2987 * ioctl to something sane.
2989 if (len > BTRFS_MAX_DEDUPE_LEN)
2990 len = BTRFS_MAX_DEDUPE_LEN;
2992 if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) {
2994 * Btrfs does not support blocksize < page_size. As a
2995 * result, btrfs_cmp_data() won't correctly handle
2996 * this situation without an update.
3003 if (S_ISDIR(src->i_mode))
3007 if (!S_ISREG(src->i_mode))
3010 /* pre-format output fields to sane values */
3011 for (i = 0; i < count; i++) {
3012 same->info[i].bytes_deduped = 0ULL;
3013 same->info[i].status = 0;
3016 for (i = 0, info = same->info; i < count; i++, info++) {
3018 struct fd dst_file = fdget(info->fd);
3019 if (!dst_file.file) {
3020 info->status = -EBADF;
3023 dst = file_inode(dst_file.file);
3025 if (!(is_admin || (dst_file.file->f_mode & FMODE_WRITE))) {
3026 info->status = -EINVAL;
3027 } else if (file->f_path.mnt != dst_file.file->f_path.mnt) {
3028 info->status = -EXDEV;
3029 } else if (S_ISDIR(dst->i_mode)) {
3030 info->status = -EISDIR;
3031 } else if (!S_ISREG(dst->i_mode)) {
3032 info->status = -EACCES;
3034 info->status = btrfs_extent_same(src, off, len, dst,
3035 info->logical_offset);
3036 if (info->status == 0)
3037 info->bytes_deduped += len;
3042 ret = copy_to_user(argp, same, size);
3047 mnt_drop_write_file(file);
3051 /* Helper to check and see if this root currently has a ref on the given disk
3052 * bytenr. If it does then we need to update the quota for this root. This
3053 * doesn't do anything if quotas aren't enabled.
3055 static int check_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
3058 struct seq_list tree_mod_seq_elem = {};
3059 struct ulist *roots;
3060 struct ulist_iterator uiter;
3061 struct ulist_node *root_node = NULL;
3064 if (!root->fs_info->quota_enabled)
3067 btrfs_get_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3068 ret = btrfs_find_all_roots(trans, root->fs_info, disko,
3069 tree_mod_seq_elem.seq, &roots);
3073 ULIST_ITER_INIT(&uiter);
3074 while ((root_node = ulist_next(roots, &uiter))) {
3075 if (root_node->val == root->objectid) {
3082 btrfs_put_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3086 static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3087 struct inode *inode,
3092 struct btrfs_root *root = BTRFS_I(inode)->root;
3095 inode_inc_iversion(inode);
3096 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3098 * We round up to the block size at eof when determining which
3099 * extents to clone above, but shouldn't round up the file size.
3101 if (endoff > destoff + olen)
3102 endoff = destoff + olen;
3103 if (endoff > inode->i_size)
3104 btrfs_i_size_write(inode, endoff);
3106 ret = btrfs_update_inode(trans, root, inode);
3108 btrfs_abort_transaction(trans, root, ret);
3109 btrfs_end_transaction(trans, root);
3112 ret = btrfs_end_transaction(trans, root);
3117 static void clone_update_extent_map(struct inode *inode,
3118 const struct btrfs_trans_handle *trans,
3119 const struct btrfs_path *path,
3120 const u64 hole_offset,
3123 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
3124 struct extent_map *em;
3127 em = alloc_extent_map();
3129 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3130 &BTRFS_I(inode)->runtime_flags);
3135 struct btrfs_file_extent_item *fi;
3137 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3138 struct btrfs_file_extent_item);
3139 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3140 em->generation = -1;
3141 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3142 BTRFS_FILE_EXTENT_INLINE)
3143 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3144 &BTRFS_I(inode)->runtime_flags);
3146 em->start = hole_offset;
3148 em->ram_bytes = em->len;
3149 em->orig_start = hole_offset;
3150 em->block_start = EXTENT_MAP_HOLE;
3152 em->orig_block_len = 0;
3153 em->compress_type = BTRFS_COMPRESS_NONE;
3154 em->generation = trans->transid;
3158 write_lock(&em_tree->lock);
3159 ret = add_extent_mapping(em_tree, em, 1);
3160 write_unlock(&em_tree->lock);
3161 if (ret != -EEXIST) {
3162 free_extent_map(em);
3165 btrfs_drop_extent_cache(inode, em->start,
3166 em->start + em->len - 1, 0);
3170 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3171 &BTRFS_I(inode)->runtime_flags);
3175 * btrfs_clone() - clone a range from inode file to another
3177 * @src: Inode to clone from
3178 * @inode: Inode to clone to
3179 * @off: Offset within source to start clone from
3180 * @olen: Original length, passed by user, of range to clone
3181 * @olen_aligned: Block-aligned value of olen, extent_same uses
3182 * identical values here
3183 * @destoff: Offset within @inode to start clone
3185 static int btrfs_clone(struct inode *src, struct inode *inode,
3186 const u64 off, const u64 olen, const u64 olen_aligned,
3189 struct btrfs_root *root = BTRFS_I(inode)->root;
3190 struct btrfs_path *path = NULL;
3191 struct extent_buffer *leaf;
3192 struct btrfs_trans_handle *trans;
3194 struct btrfs_key key;
3199 const u64 len = olen_aligned;
3201 u64 last_dest_end = destoff;
3204 buf = vmalloc(root->nodesize);
3208 path = btrfs_alloc_path();
3216 key.objectid = btrfs_ino(src);
3217 key.type = BTRFS_EXTENT_DATA_KEY;
3222 * note the key will change type as we walk through the
3225 path->leave_spinning = 1;
3226 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3231 * First search, if no extent item that starts at offset off was
3232 * found but the previous item is an extent item, it's possible
3233 * it might overlap our target range, therefore process it.
3235 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3236 btrfs_item_key_to_cpu(path->nodes[0], &key,
3237 path->slots[0] - 1);
3238 if (key.type == BTRFS_EXTENT_DATA_KEY)
3242 nritems = btrfs_header_nritems(path->nodes[0]);
3245 if (path->slots[0] >= nritems) {
3246 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
3251 nritems = btrfs_header_nritems(path->nodes[0]);
3253 leaf = path->nodes[0];
3254 slot = path->slots[0];
3256 btrfs_item_key_to_cpu(leaf, &key, slot);
3257 if (key.type > BTRFS_EXTENT_DATA_KEY ||
3258 key.objectid != btrfs_ino(src))
3261 if (key.type == BTRFS_EXTENT_DATA_KEY) {
3262 struct btrfs_file_extent_item *extent;
3265 struct btrfs_key new_key;
3266 u64 disko = 0, diskl = 0;
3267 u64 datao = 0, datal = 0;
3271 extent = btrfs_item_ptr(leaf, slot,
3272 struct btrfs_file_extent_item);
3273 comp = btrfs_file_extent_compression(leaf, extent);
3274 type = btrfs_file_extent_type(leaf, extent);
3275 if (type == BTRFS_FILE_EXTENT_REG ||
3276 type == BTRFS_FILE_EXTENT_PREALLOC) {
3277 disko = btrfs_file_extent_disk_bytenr(leaf,
3279 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3281 datao = btrfs_file_extent_offset(leaf, extent);
3282 datal = btrfs_file_extent_num_bytes(leaf,
3284 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3285 /* take upper bound, may be compressed */
3286 datal = btrfs_file_extent_ram_bytes(leaf,
3291 * The first search might have left us at an extent
3292 * item that ends before our target range's start, can
3293 * happen if we have holes and NO_HOLES feature enabled.
3295 if (key.offset + datal <= off) {
3298 } else if (key.offset >= off + len) {
3302 size = btrfs_item_size_nr(leaf, slot);
3303 read_extent_buffer(leaf, buf,
3304 btrfs_item_ptr_offset(leaf, slot),
3307 btrfs_release_path(path);
3308 path->leave_spinning = 0;
3310 memcpy(&new_key, &key, sizeof(new_key));
3311 new_key.objectid = btrfs_ino(inode);
3312 if (off <= key.offset)
3313 new_key.offset = key.offset + destoff - off;
3315 new_key.offset = destoff;
3318 * Deal with a hole that doesn't have an extent item
3319 * that represents it (NO_HOLES feature enabled).
3320 * This hole is either in the middle of the cloning
3321 * range or at the beginning (fully overlaps it or
3322 * partially overlaps it).
3324 if (new_key.offset != last_dest_end)
3325 drop_start = last_dest_end;
3327 drop_start = new_key.offset;
3330 * 1 - adjusting old extent (we may have to split it)
3331 * 1 - add new extent
3334 trans = btrfs_start_transaction(root, 3);
3335 if (IS_ERR(trans)) {
3336 ret = PTR_ERR(trans);
3340 if (type == BTRFS_FILE_EXTENT_REG ||
3341 type == BTRFS_FILE_EXTENT_PREALLOC) {
3343 * a | --- range to clone ---| b
3344 * | ------------- extent ------------- |
3347 /* subtract range b */
3348 if (key.offset + datal > off + len)
3349 datal = off + len - key.offset;
3351 /* subtract range a */
3352 if (off > key.offset) {
3353 datao += off - key.offset;
3354 datal -= off - key.offset;
3357 ret = btrfs_drop_extents(trans, root, inode,
3359 new_key.offset + datal,
3362 if (ret != -EOPNOTSUPP)
3363 btrfs_abort_transaction(trans,
3365 btrfs_end_transaction(trans, root);
3369 ret = btrfs_insert_empty_item(trans, root, path,
3372 btrfs_abort_transaction(trans, root,
3374 btrfs_end_transaction(trans, root);
3378 leaf = path->nodes[0];
3379 slot = path->slots[0];
3380 write_extent_buffer(leaf, buf,
3381 btrfs_item_ptr_offset(leaf, slot),
3384 extent = btrfs_item_ptr(leaf, slot,
3385 struct btrfs_file_extent_item);
3387 /* disko == 0 means it's a hole */
3391 btrfs_set_file_extent_offset(leaf, extent,
3393 btrfs_set_file_extent_num_bytes(leaf, extent,
3397 * We need to look up the roots that point at
3398 * this bytenr and see if the new root does. If
3399 * it does not we need to make sure we update
3400 * quotas appropriately.
3402 if (disko && root != BTRFS_I(src)->root &&
3403 disko != last_disko) {
3404 no_quota = check_ref(trans, root,
3407 btrfs_abort_transaction(trans,
3410 btrfs_end_transaction(trans,
3418 inode_add_bytes(inode, datal);
3419 ret = btrfs_inc_extent_ref(trans, root,
3421 root->root_key.objectid,
3423 new_key.offset - datao,
3426 btrfs_abort_transaction(trans,
3429 btrfs_end_transaction(trans,
3435 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3438 u64 aligned_end = 0;
3440 if (off > key.offset) {
3441 skip = off - key.offset;
3442 new_key.offset += skip;
3445 if (key.offset + datal > off + len)
3446 trim = key.offset + datal - (off + len);
3448 if (comp && (skip || trim)) {
3450 btrfs_end_transaction(trans, root);
3453 size -= skip + trim;
3454 datal -= skip + trim;
3456 aligned_end = ALIGN(new_key.offset + datal,
3458 ret = btrfs_drop_extents(trans, root, inode,
3463 if (ret != -EOPNOTSUPP)
3464 btrfs_abort_transaction(trans,
3466 btrfs_end_transaction(trans, root);
3470 ret = btrfs_insert_empty_item(trans, root, path,
3473 btrfs_abort_transaction(trans, root,
3475 btrfs_end_transaction(trans, root);
3481 btrfs_file_extent_calc_inline_size(0);
3482 memmove(buf+start, buf+start+skip,
3486 leaf = path->nodes[0];
3487 slot = path->slots[0];
3488 write_extent_buffer(leaf, buf,
3489 btrfs_item_ptr_offset(leaf, slot),
3491 inode_add_bytes(inode, datal);
3494 /* If we have an implicit hole (NO_HOLES feature). */
3495 if (drop_start < new_key.offset)
3496 clone_update_extent_map(inode, trans,
3498 new_key.offset - drop_start);
3500 clone_update_extent_map(inode, trans, path, 0, 0);
3502 btrfs_mark_buffer_dirty(leaf);
3503 btrfs_release_path(path);
3505 last_dest_end = ALIGN(new_key.offset + datal,
3507 ret = clone_finish_inode_update(trans, inode,
3512 if (new_key.offset + datal >= destoff + len)
3515 btrfs_release_path(path);
3520 if (last_dest_end < destoff + len) {
3522 * We have an implicit hole (NO_HOLES feature is enabled) that
3523 * fully or partially overlaps our cloning range at its end.
3525 btrfs_release_path(path);
3528 * 1 - remove extent(s)
3531 trans = btrfs_start_transaction(root, 2);
3532 if (IS_ERR(trans)) {
3533 ret = PTR_ERR(trans);
3536 ret = btrfs_drop_extents(trans, root, inode,
3537 last_dest_end, destoff + len, 1);
3539 if (ret != -EOPNOTSUPP)
3540 btrfs_abort_transaction(trans, root, ret);
3541 btrfs_end_transaction(trans, root);
3544 clone_update_extent_map(inode, trans, NULL, last_dest_end,
3545 destoff + len - last_dest_end);
3546 ret = clone_finish_inode_update(trans, inode, destoff + len,
3551 btrfs_free_path(path);
3556 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
3557 u64 off, u64 olen, u64 destoff)
3559 struct inode *inode = file_inode(file);
3560 struct btrfs_root *root = BTRFS_I(inode)->root;
3565 u64 bs = root->fs_info->sb->s_blocksize;
3570 * - split compressed inline extents. annoying: we need to
3571 * decompress into destination's address_space (the file offset
3572 * may change, so source mapping won't do), then recompress (or
3573 * otherwise reinsert) a subrange.
3575 * - split destination inode's inline extents. The inline extents can
3576 * be either compressed or non-compressed.
3579 /* the destination must be opened for writing */
3580 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
3583 if (btrfs_root_readonly(root))
3586 ret = mnt_want_write_file(file);
3590 src_file = fdget(srcfd);
3591 if (!src_file.file) {
3593 goto out_drop_write;
3597 if (src_file.file->f_path.mnt != file->f_path.mnt)
3600 src = file_inode(src_file.file);
3606 /* the src must be open for reading */
3607 if (!(src_file.file->f_mode & FMODE_READ))
3610 /* don't make the dst file partly checksummed */
3611 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3612 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3616 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3620 if (src->i_sb != inode->i_sb)
3625 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
3626 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
3628 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
3629 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
3632 mutex_lock(&src->i_mutex);
3635 /* determine range to clone */
3637 if (off + len > src->i_size || off + len < off)
3640 olen = len = src->i_size - off;
3641 /* if we extend to eof, continue to block boundary */
3642 if (off + len == src->i_size)
3643 len = ALIGN(src->i_size, bs) - off;
3645 /* verify the end result is block aligned */
3646 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3647 !IS_ALIGNED(destoff, bs))
3650 /* verify if ranges are overlapped within the same file */
3652 if (destoff + len > off && destoff < off + len)
3656 if (destoff > inode->i_size) {
3657 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3663 * Lock the target range too. Right after we replace the file extent
3664 * items in the fs tree (which now point to the cloned data), we might
3665 * have a worker replace them with extent items relative to a write
3666 * operation that was issued before this clone operation (i.e. confront
3667 * with inode.c:btrfs_finish_ordered_io).
3670 u64 lock_start = min_t(u64, off, destoff);
3671 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
3673 lock_extent_range(src, lock_start, lock_len);
3675 lock_extent_range(src, off, len);
3676 lock_extent_range(inode, destoff, len);
3679 ret = btrfs_clone(src, inode, off, olen, len, destoff);
3682 u64 lock_start = min_t(u64, off, destoff);
3683 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3685 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3687 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
3688 unlock_extent(&BTRFS_I(inode)->io_tree, destoff,
3692 * Truncate page cache pages so that future reads will see the cloned
3693 * data immediately and not the previous data.
3695 truncate_inode_pages_range(&inode->i_data, destoff,
3696 PAGE_CACHE_ALIGN(destoff + len) - 1);
3700 mutex_unlock(&src->i_mutex);
3701 mutex_unlock(&inode->i_mutex);
3703 mutex_unlock(&inode->i_mutex);
3704 mutex_unlock(&src->i_mutex);
3707 mutex_unlock(&src->i_mutex);
3712 mnt_drop_write_file(file);
3716 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
3718 struct btrfs_ioctl_clone_range_args args;
3720 if (copy_from_user(&args, argp, sizeof(args)))
3722 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
3723 args.src_length, args.dest_offset);
3727 * there are many ways the trans_start and trans_end ioctls can lead
3728 * to deadlocks. They should only be used by applications that
3729 * basically own the machine, and have a very in depth understanding
3730 * of all the possible deadlocks and enospc problems.
3732 static long btrfs_ioctl_trans_start(struct file *file)
3734 struct inode *inode = file_inode(file);
3735 struct btrfs_root *root = BTRFS_I(inode)->root;
3736 struct btrfs_trans_handle *trans;
3740 if (!capable(CAP_SYS_ADMIN))
3744 if (file->private_data)
3748 if (btrfs_root_readonly(root))
3751 ret = mnt_want_write_file(file);
3755 atomic_inc(&root->fs_info->open_ioctl_trans);
3758 trans = btrfs_start_ioctl_transaction(root);
3762 file->private_data = trans;
3766 atomic_dec(&root->fs_info->open_ioctl_trans);
3767 mnt_drop_write_file(file);
3772 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3774 struct inode *inode = file_inode(file);
3775 struct btrfs_root *root = BTRFS_I(inode)->root;
3776 struct btrfs_root *new_root;
3777 struct btrfs_dir_item *di;
3778 struct btrfs_trans_handle *trans;
3779 struct btrfs_path *path;
3780 struct btrfs_key location;
3781 struct btrfs_disk_key disk_key;
3786 if (!capable(CAP_SYS_ADMIN))
3789 ret = mnt_want_write_file(file);
3793 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3799 objectid = BTRFS_FS_TREE_OBJECTID;
3801 location.objectid = objectid;
3802 location.type = BTRFS_ROOT_ITEM_KEY;
3803 location.offset = (u64)-1;
3805 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
3806 if (IS_ERR(new_root)) {
3807 ret = PTR_ERR(new_root);
3811 path = btrfs_alloc_path();
3816 path->leave_spinning = 1;
3818 trans = btrfs_start_transaction(root, 1);
3819 if (IS_ERR(trans)) {
3820 btrfs_free_path(path);
3821 ret = PTR_ERR(trans);
3825 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
3826 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
3827 dir_id, "default", 7, 1);
3828 if (IS_ERR_OR_NULL(di)) {
3829 btrfs_free_path(path);
3830 btrfs_end_transaction(trans, root);
3831 btrfs_err(new_root->fs_info, "Umm, you don't have the default dir"
3832 "item, this isn't going to work");
3837 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3838 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3839 btrfs_mark_buffer_dirty(path->nodes[0]);
3840 btrfs_free_path(path);
3842 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
3843 btrfs_end_transaction(trans, root);
3845 mnt_drop_write_file(file);
3849 void btrfs_get_block_group_info(struct list_head *groups_list,
3850 struct btrfs_ioctl_space_info *space)
3852 struct btrfs_block_group_cache *block_group;
3854 space->total_bytes = 0;
3855 space->used_bytes = 0;
3857 list_for_each_entry(block_group, groups_list, list) {
3858 space->flags = block_group->flags;
3859 space->total_bytes += block_group->key.offset;
3860 space->used_bytes +=
3861 btrfs_block_group_used(&block_group->item);
3865 static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
3867 struct btrfs_ioctl_space_args space_args;
3868 struct btrfs_ioctl_space_info space;
3869 struct btrfs_ioctl_space_info *dest;
3870 struct btrfs_ioctl_space_info *dest_orig;
3871 struct btrfs_ioctl_space_info __user *user_dest;
3872 struct btrfs_space_info *info;
3873 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3874 BTRFS_BLOCK_GROUP_SYSTEM,
3875 BTRFS_BLOCK_GROUP_METADATA,
3876 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3883 if (copy_from_user(&space_args,
3884 (struct btrfs_ioctl_space_args __user *)arg,
3885 sizeof(space_args)))
3888 for (i = 0; i < num_types; i++) {
3889 struct btrfs_space_info *tmp;
3893 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3895 if (tmp->flags == types[i]) {
3905 down_read(&info->groups_sem);
3906 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3907 if (!list_empty(&info->block_groups[c]))
3910 up_read(&info->groups_sem);
3914 * Global block reserve, exported as a space_info
3918 /* space_slots == 0 means they are asking for a count */
3919 if (space_args.space_slots == 0) {
3920 space_args.total_spaces = slot_count;
3924 slot_count = min_t(u64, space_args.space_slots, slot_count);
3926 alloc_size = sizeof(*dest) * slot_count;
3928 /* we generally have at most 6 or so space infos, one for each raid
3929 * level. So, a whole page should be more than enough for everyone
3931 if (alloc_size > PAGE_CACHE_SIZE)
3934 space_args.total_spaces = 0;
3935 dest = kmalloc(alloc_size, GFP_NOFS);
3940 /* now we have a buffer to copy into */
3941 for (i = 0; i < num_types; i++) {
3942 struct btrfs_space_info *tmp;
3949 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3951 if (tmp->flags == types[i]) {
3960 down_read(&info->groups_sem);
3961 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3962 if (!list_empty(&info->block_groups[c])) {
3963 btrfs_get_block_group_info(
3964 &info->block_groups[c], &space);
3965 memcpy(dest, &space, sizeof(space));
3967 space_args.total_spaces++;
3973 up_read(&info->groups_sem);
3977 * Add global block reserve
3980 struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv;
3982 spin_lock(&block_rsv->lock);
3983 space.total_bytes = block_rsv->size;
3984 space.used_bytes = block_rsv->size - block_rsv->reserved;
3985 spin_unlock(&block_rsv->lock);
3986 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3987 memcpy(dest, &space, sizeof(space));
3988 space_args.total_spaces++;
3991 user_dest = (struct btrfs_ioctl_space_info __user *)
3992 (arg + sizeof(struct btrfs_ioctl_space_args));
3994 if (copy_to_user(user_dest, dest_orig, alloc_size))
3999 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
4006 * there are many ways the trans_start and trans_end ioctls can lead
4007 * to deadlocks. They should only be used by applications that
4008 * basically own the machine, and have a very in depth understanding
4009 * of all the possible deadlocks and enospc problems.
4011 long btrfs_ioctl_trans_end(struct file *file)
4013 struct inode *inode = file_inode(file);
4014 struct btrfs_root *root = BTRFS_I(inode)->root;
4015 struct btrfs_trans_handle *trans;
4017 trans = file->private_data;
4020 file->private_data = NULL;
4022 btrfs_end_transaction(trans, root);
4024 atomic_dec(&root->fs_info->open_ioctl_trans);
4026 mnt_drop_write_file(file);
4030 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4033 struct btrfs_trans_handle *trans;
4037 trans = btrfs_attach_transaction_barrier(root);
4038 if (IS_ERR(trans)) {
4039 if (PTR_ERR(trans) != -ENOENT)
4040 return PTR_ERR(trans);
4042 /* No running transaction, don't bother */
4043 transid = root->fs_info->last_trans_committed;
4046 transid = trans->transid;
4047 ret = btrfs_commit_transaction_async(trans, root, 0);
4049 btrfs_end_transaction(trans, root);
4054 if (copy_to_user(argp, &transid, sizeof(transid)))
4059 static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
4065 if (copy_from_user(&transid, argp, sizeof(transid)))
4068 transid = 0; /* current trans */
4070 return btrfs_wait_for_commit(root, transid);
4073 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
4075 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4076 struct btrfs_ioctl_scrub_args *sa;
4079 if (!capable(CAP_SYS_ADMIN))
4082 sa = memdup_user(arg, sizeof(*sa));
4086 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4087 ret = mnt_want_write_file(file);
4092 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
4093 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4096 if (copy_to_user(arg, sa, sizeof(*sa)))
4099 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4100 mnt_drop_write_file(file);
4106 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
4108 if (!capable(CAP_SYS_ADMIN))
4111 return btrfs_scrub_cancel(root->fs_info);
4114 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
4117 struct btrfs_ioctl_scrub_args *sa;
4120 if (!capable(CAP_SYS_ADMIN))
4123 sa = memdup_user(arg, sizeof(*sa));
4127 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
4129 if (copy_to_user(arg, sa, sizeof(*sa)))
4136 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
4139 struct btrfs_ioctl_get_dev_stats *sa;
4142 sa = memdup_user(arg, sizeof(*sa));
4146 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4151 ret = btrfs_get_dev_stats(root, sa);
4153 if (copy_to_user(arg, sa, sizeof(*sa)))
4160 static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
4162 struct btrfs_ioctl_dev_replace_args *p;
4165 if (!capable(CAP_SYS_ADMIN))
4168 p = memdup_user(arg, sizeof(*p));
4173 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
4174 if (root->fs_info->sb->s_flags & MS_RDONLY) {
4179 &root->fs_info->mutually_exclusive_operation_running,
4181 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4183 ret = btrfs_dev_replace_start(root, p);
4185 &root->fs_info->mutually_exclusive_operation_running,
4189 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4190 btrfs_dev_replace_status(root->fs_info, p);
4193 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4194 ret = btrfs_dev_replace_cancel(root->fs_info, p);
4201 if (copy_to_user(arg, p, sizeof(*p)))
4208 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4214 struct btrfs_ioctl_ino_path_args *ipa = NULL;
4215 struct inode_fs_paths *ipath = NULL;
4216 struct btrfs_path *path;
4218 if (!capable(CAP_DAC_READ_SEARCH))
4221 path = btrfs_alloc_path();
4227 ipa = memdup_user(arg, sizeof(*ipa));
4234 size = min_t(u32, ipa->size, 4096);
4235 ipath = init_ipath(size, root, path);
4236 if (IS_ERR(ipath)) {
4237 ret = PTR_ERR(ipath);
4242 ret = paths_from_inode(ipa->inum, ipath);
4246 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
4247 rel_ptr = ipath->fspath->val[i] -
4248 (u64)(unsigned long)ipath->fspath->val;
4249 ipath->fspath->val[i] = rel_ptr;
4252 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
4253 (void *)(unsigned long)ipath->fspath, size);
4260 btrfs_free_path(path);
4267 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4269 struct btrfs_data_container *inodes = ctx;
4270 const size_t c = 3 * sizeof(u64);
4272 if (inodes->bytes_left >= c) {
4273 inodes->bytes_left -= c;
4274 inodes->val[inodes->elem_cnt] = inum;
4275 inodes->val[inodes->elem_cnt + 1] = offset;
4276 inodes->val[inodes->elem_cnt + 2] = root;
4277 inodes->elem_cnt += 3;
4279 inodes->bytes_missing += c - inodes->bytes_left;
4280 inodes->bytes_left = 0;
4281 inodes->elem_missed += 3;
4287 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
4292 struct btrfs_ioctl_logical_ino_args *loi;
4293 struct btrfs_data_container *inodes = NULL;
4294 struct btrfs_path *path = NULL;
4296 if (!capable(CAP_SYS_ADMIN))
4299 loi = memdup_user(arg, sizeof(*loi));
4306 path = btrfs_alloc_path();
4312 size = min_t(u32, loi->size, 64 * 1024);
4313 inodes = init_data_container(size);
4314 if (IS_ERR(inodes)) {
4315 ret = PTR_ERR(inodes);
4320 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
4321 build_ino_list, inodes);
4327 ret = copy_to_user((void *)(unsigned long)loi->inodes,
4328 (void *)(unsigned long)inodes, size);
4333 btrfs_free_path(path);
4340 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
4341 struct btrfs_ioctl_balance_args *bargs)
4343 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4345 bargs->flags = bctl->flags;
4347 if (atomic_read(&fs_info->balance_running))
4348 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4349 if (atomic_read(&fs_info->balance_pause_req))
4350 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
4351 if (atomic_read(&fs_info->balance_cancel_req))
4352 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
4354 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4355 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4356 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4359 spin_lock(&fs_info->balance_lock);
4360 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4361 spin_unlock(&fs_info->balance_lock);
4363 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4367 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4369 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4370 struct btrfs_fs_info *fs_info = root->fs_info;
4371 struct btrfs_ioctl_balance_args *bargs;
4372 struct btrfs_balance_control *bctl;
4373 bool need_unlock; /* for mut. excl. ops lock */
4376 if (!capable(CAP_SYS_ADMIN))
4379 ret = mnt_want_write_file(file);
4384 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
4385 mutex_lock(&fs_info->volume_mutex);
4386 mutex_lock(&fs_info->balance_mutex);
4392 * mut. excl. ops lock is locked. Three possibilites:
4393 * (1) some other op is running
4394 * (2) balance is running
4395 * (3) balance is paused -- special case (think resume)
4397 mutex_lock(&fs_info->balance_mutex);
4398 if (fs_info->balance_ctl) {
4399 /* this is either (2) or (3) */
4400 if (!atomic_read(&fs_info->balance_running)) {
4401 mutex_unlock(&fs_info->balance_mutex);
4402 if (!mutex_trylock(&fs_info->volume_mutex))
4404 mutex_lock(&fs_info->balance_mutex);
4406 if (fs_info->balance_ctl &&
4407 !atomic_read(&fs_info->balance_running)) {
4409 need_unlock = false;
4413 mutex_unlock(&fs_info->balance_mutex);
4414 mutex_unlock(&fs_info->volume_mutex);
4418 mutex_unlock(&fs_info->balance_mutex);
4424 mutex_unlock(&fs_info->balance_mutex);
4425 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4430 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
4433 bargs = memdup_user(arg, sizeof(*bargs));
4434 if (IS_ERR(bargs)) {
4435 ret = PTR_ERR(bargs);
4439 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4440 if (!fs_info->balance_ctl) {
4445 bctl = fs_info->balance_ctl;
4446 spin_lock(&fs_info->balance_lock);
4447 bctl->flags |= BTRFS_BALANCE_RESUME;
4448 spin_unlock(&fs_info->balance_lock);
4456 if (fs_info->balance_ctl) {
4461 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
4467 bctl->fs_info = fs_info;
4469 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4470 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4471 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4473 bctl->flags = bargs->flags;
4475 /* balance everything - no filters */
4476 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4481 * Ownership of bctl and mutually_exclusive_operation_running
4482 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4483 * or, if restriper was paused all the way until unmount, in
4484 * free_fs_info. mutually_exclusive_operation_running is
4485 * cleared in __cancel_balance.
4487 need_unlock = false;
4489 ret = btrfs_balance(bctl, bargs);
4492 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4499 mutex_unlock(&fs_info->balance_mutex);
4500 mutex_unlock(&fs_info->volume_mutex);
4502 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4504 mnt_drop_write_file(file);
4508 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4510 if (!capable(CAP_SYS_ADMIN))
4514 case BTRFS_BALANCE_CTL_PAUSE:
4515 return btrfs_pause_balance(root->fs_info);
4516 case BTRFS_BALANCE_CTL_CANCEL:
4517 return btrfs_cancel_balance(root->fs_info);
4523 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4526 struct btrfs_fs_info *fs_info = root->fs_info;
4527 struct btrfs_ioctl_balance_args *bargs;
4530 if (!capable(CAP_SYS_ADMIN))
4533 mutex_lock(&fs_info->balance_mutex);
4534 if (!fs_info->balance_ctl) {
4539 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
4545 update_ioctl_balance_args(fs_info, 1, bargs);
4547 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4552 mutex_unlock(&fs_info->balance_mutex);
4556 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4558 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4559 struct btrfs_ioctl_quota_ctl_args *sa;
4560 struct btrfs_trans_handle *trans = NULL;
4564 if (!capable(CAP_SYS_ADMIN))
4567 ret = mnt_want_write_file(file);
4571 sa = memdup_user(arg, sizeof(*sa));
4577 down_write(&root->fs_info->subvol_sem);
4578 trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4579 if (IS_ERR(trans)) {
4580 ret = PTR_ERR(trans);
4585 case BTRFS_QUOTA_CTL_ENABLE:
4586 ret = btrfs_quota_enable(trans, root->fs_info);
4588 case BTRFS_QUOTA_CTL_DISABLE:
4589 ret = btrfs_quota_disable(trans, root->fs_info);
4596 err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4601 up_write(&root->fs_info->subvol_sem);
4603 mnt_drop_write_file(file);
4607 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4609 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4610 struct btrfs_ioctl_qgroup_assign_args *sa;
4611 struct btrfs_trans_handle *trans;
4615 if (!capable(CAP_SYS_ADMIN))
4618 ret = mnt_want_write_file(file);
4622 sa = memdup_user(arg, sizeof(*sa));
4628 trans = btrfs_join_transaction(root);
4629 if (IS_ERR(trans)) {
4630 ret = PTR_ERR(trans);
4634 /* FIXME: check if the IDs really exist */
4636 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4639 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4643 err = btrfs_end_transaction(trans, root);
4650 mnt_drop_write_file(file);
4654 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4656 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4657 struct btrfs_ioctl_qgroup_create_args *sa;
4658 struct btrfs_trans_handle *trans;
4662 if (!capable(CAP_SYS_ADMIN))
4665 ret = mnt_want_write_file(file);
4669 sa = memdup_user(arg, sizeof(*sa));
4675 if (!sa->qgroupid) {
4680 trans = btrfs_join_transaction(root);
4681 if (IS_ERR(trans)) {
4682 ret = PTR_ERR(trans);
4686 /* FIXME: check if the IDs really exist */
4688 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
4691 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4694 err = btrfs_end_transaction(trans, root);
4701 mnt_drop_write_file(file);
4705 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4707 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4708 struct btrfs_ioctl_qgroup_limit_args *sa;
4709 struct btrfs_trans_handle *trans;
4714 if (!capable(CAP_SYS_ADMIN))
4717 ret = mnt_want_write_file(file);
4721 sa = memdup_user(arg, sizeof(*sa));
4727 trans = btrfs_join_transaction(root);
4728 if (IS_ERR(trans)) {
4729 ret = PTR_ERR(trans);
4733 qgroupid = sa->qgroupid;
4735 /* take the current subvol as qgroup */
4736 qgroupid = root->root_key.objectid;
4739 /* FIXME: check if the IDs really exist */
4740 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
4742 err = btrfs_end_transaction(trans, root);
4749 mnt_drop_write_file(file);
4753 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4755 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4756 struct btrfs_ioctl_quota_rescan_args *qsa;
4759 if (!capable(CAP_SYS_ADMIN))
4762 ret = mnt_want_write_file(file);
4766 qsa = memdup_user(arg, sizeof(*qsa));
4777 ret = btrfs_qgroup_rescan(root->fs_info);
4782 mnt_drop_write_file(file);
4786 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
4788 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4789 struct btrfs_ioctl_quota_rescan_args *qsa;
4792 if (!capable(CAP_SYS_ADMIN))
4795 qsa = kzalloc(sizeof(*qsa), GFP_NOFS);
4799 if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4801 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
4804 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4811 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
4813 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4815 if (!capable(CAP_SYS_ADMIN))
4818 return btrfs_qgroup_wait_for_completion(root->fs_info);
4821 static long _btrfs_ioctl_set_received_subvol(struct file *file,
4822 struct btrfs_ioctl_received_subvol_args *sa)
4824 struct inode *inode = file_inode(file);
4825 struct btrfs_root *root = BTRFS_I(inode)->root;
4826 struct btrfs_root_item *root_item = &root->root_item;
4827 struct btrfs_trans_handle *trans;
4828 struct timespec ct = CURRENT_TIME;
4830 int received_uuid_changed;
4832 if (!inode_owner_or_capable(inode))
4835 ret = mnt_want_write_file(file);
4839 down_write(&root->fs_info->subvol_sem);
4841 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
4846 if (btrfs_root_readonly(root)) {
4853 * 2 - uuid items (received uuid + subvol uuid)
4855 trans = btrfs_start_transaction(root, 3);
4856 if (IS_ERR(trans)) {
4857 ret = PTR_ERR(trans);
4862 sa->rtransid = trans->transid;
4863 sa->rtime.sec = ct.tv_sec;
4864 sa->rtime.nsec = ct.tv_nsec;
4866 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4868 if (received_uuid_changed &&
4869 !btrfs_is_empty_uuid(root_item->received_uuid))
4870 btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
4871 root_item->received_uuid,
4872 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4873 root->root_key.objectid);
4874 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4875 btrfs_set_root_stransid(root_item, sa->stransid);
4876 btrfs_set_root_rtransid(root_item, sa->rtransid);
4877 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4878 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4879 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4880 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4882 ret = btrfs_update_root(trans, root->fs_info->tree_root,
4883 &root->root_key, &root->root_item);
4885 btrfs_end_transaction(trans, root);
4888 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4889 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
4891 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4892 root->root_key.objectid);
4893 if (ret < 0 && ret != -EEXIST) {
4894 btrfs_abort_transaction(trans, root, ret);
4898 ret = btrfs_commit_transaction(trans, root);
4900 btrfs_abort_transaction(trans, root, ret);
4905 up_write(&root->fs_info->subvol_sem);
4906 mnt_drop_write_file(file);
4911 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4914 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4915 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4918 args32 = memdup_user(arg, sizeof(*args32));
4919 if (IS_ERR(args32)) {
4920 ret = PTR_ERR(args32);
4925 args64 = kmalloc(sizeof(*args64), GFP_NOFS);
4931 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4932 args64->stransid = args32->stransid;
4933 args64->rtransid = args32->rtransid;
4934 args64->stime.sec = args32->stime.sec;
4935 args64->stime.nsec = args32->stime.nsec;
4936 args64->rtime.sec = args32->rtime.sec;
4937 args64->rtime.nsec = args32->rtime.nsec;
4938 args64->flags = args32->flags;
4940 ret = _btrfs_ioctl_set_received_subvol(file, args64);
4944 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4945 args32->stransid = args64->stransid;
4946 args32->rtransid = args64->rtransid;
4947 args32->stime.sec = args64->stime.sec;
4948 args32->stime.nsec = args64->stime.nsec;
4949 args32->rtime.sec = args64->rtime.sec;
4950 args32->rtime.nsec = args64->rtime.nsec;
4951 args32->flags = args64->flags;
4953 ret = copy_to_user(arg, args32, sizeof(*args32));
4964 static long btrfs_ioctl_set_received_subvol(struct file *file,
4967 struct btrfs_ioctl_received_subvol_args *sa = NULL;
4970 sa = memdup_user(arg, sizeof(*sa));
4977 ret = _btrfs_ioctl_set_received_subvol(file, sa);
4982 ret = copy_to_user(arg, sa, sizeof(*sa));
4991 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
4993 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4996 char label[BTRFS_LABEL_SIZE];
4998 spin_lock(&root->fs_info->super_lock);
4999 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
5000 spin_unlock(&root->fs_info->super_lock);
5002 len = strnlen(label, BTRFS_LABEL_SIZE);
5004 if (len == BTRFS_LABEL_SIZE) {
5005 btrfs_warn(root->fs_info,
5006 "label is too long, return the first %zu bytes", --len);
5009 ret = copy_to_user(arg, label, len);
5011 return ret ? -EFAULT : 0;
5014 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5016 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5017 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5018 struct btrfs_trans_handle *trans;
5019 char label[BTRFS_LABEL_SIZE];
5022 if (!capable(CAP_SYS_ADMIN))
5025 if (copy_from_user(label, arg, sizeof(label)))
5028 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
5029 btrfs_err(root->fs_info, "unable to set label with more than %d bytes",
5030 BTRFS_LABEL_SIZE - 1);
5034 ret = mnt_want_write_file(file);
5038 trans = btrfs_start_transaction(root, 0);
5039 if (IS_ERR(trans)) {
5040 ret = PTR_ERR(trans);
5044 spin_lock(&root->fs_info->super_lock);
5045 strcpy(super_block->label, label);
5046 spin_unlock(&root->fs_info->super_lock);
5047 ret = btrfs_commit_transaction(trans, root);
5050 mnt_drop_write_file(file);
5054 #define INIT_FEATURE_FLAGS(suffix) \
5055 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5056 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5057 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5059 static int btrfs_ioctl_get_supported_features(struct file *file,
5062 static struct btrfs_ioctl_feature_flags features[3] = {
5063 INIT_FEATURE_FLAGS(SUPP),
5064 INIT_FEATURE_FLAGS(SAFE_SET),
5065 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5068 if (copy_to_user(arg, &features, sizeof(features)))
5074 static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5076 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5077 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5078 struct btrfs_ioctl_feature_flags features;
5080 features.compat_flags = btrfs_super_compat_flags(super_block);
5081 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5082 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5084 if (copy_to_user(arg, &features, sizeof(features)))
5090 static int check_feature_bits(struct btrfs_root *root,
5091 enum btrfs_feature_set set,
5092 u64 change_mask, u64 flags, u64 supported_flags,
5093 u64 safe_set, u64 safe_clear)
5095 const char *type = btrfs_feature_set_names[set];
5097 u64 disallowed, unsupported;
5098 u64 set_mask = flags & change_mask;
5099 u64 clear_mask = ~flags & change_mask;
5101 unsupported = set_mask & ~supported_flags;
5103 names = btrfs_printable_features(set, unsupported);
5105 btrfs_warn(root->fs_info,
5106 "this kernel does not support the %s feature bit%s",
5107 names, strchr(names, ',') ? "s" : "");
5110 btrfs_warn(root->fs_info,
5111 "this kernel does not support %s bits 0x%llx",
5116 disallowed = set_mask & ~safe_set;
5118 names = btrfs_printable_features(set, disallowed);
5120 btrfs_warn(root->fs_info,
5121 "can't set the %s feature bit%s while mounted",
5122 names, strchr(names, ',') ? "s" : "");
5125 btrfs_warn(root->fs_info,
5126 "can't set %s bits 0x%llx while mounted",
5131 disallowed = clear_mask & ~safe_clear;
5133 names = btrfs_printable_features(set, disallowed);
5135 btrfs_warn(root->fs_info,
5136 "can't clear the %s feature bit%s while mounted",
5137 names, strchr(names, ',') ? "s" : "");
5140 btrfs_warn(root->fs_info,
5141 "can't clear %s bits 0x%llx while mounted",
5149 #define check_feature(root, change_mask, flags, mask_base) \
5150 check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
5151 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5152 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5153 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5155 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5157 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5158 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5159 struct btrfs_ioctl_feature_flags flags[2];
5160 struct btrfs_trans_handle *trans;
5164 if (!capable(CAP_SYS_ADMIN))
5167 if (copy_from_user(flags, arg, sizeof(flags)))
5171 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5172 !flags[0].incompat_flags)
5175 ret = check_feature(root, flags[0].compat_flags,
5176 flags[1].compat_flags, COMPAT);
5180 ret = check_feature(root, flags[0].compat_ro_flags,
5181 flags[1].compat_ro_flags, COMPAT_RO);
5185 ret = check_feature(root, flags[0].incompat_flags,
5186 flags[1].incompat_flags, INCOMPAT);
5190 trans = btrfs_start_transaction(root, 0);
5192 return PTR_ERR(trans);
5194 spin_lock(&root->fs_info->super_lock);
5195 newflags = btrfs_super_compat_flags(super_block);
5196 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5197 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5198 btrfs_set_super_compat_flags(super_block, newflags);
5200 newflags = btrfs_super_compat_ro_flags(super_block);
5201 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5202 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5203 btrfs_set_super_compat_ro_flags(super_block, newflags);
5205 newflags = btrfs_super_incompat_flags(super_block);
5206 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5207 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5208 btrfs_set_super_incompat_flags(super_block, newflags);
5209 spin_unlock(&root->fs_info->super_lock);
5211 return btrfs_commit_transaction(trans, root);
5214 long btrfs_ioctl(struct file *file, unsigned int
5215 cmd, unsigned long arg)
5217 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5218 void __user *argp = (void __user *)arg;
5221 case FS_IOC_GETFLAGS:
5222 return btrfs_ioctl_getflags(file, argp);
5223 case FS_IOC_SETFLAGS:
5224 return btrfs_ioctl_setflags(file, argp);
5225 case FS_IOC_GETVERSION:
5226 return btrfs_ioctl_getversion(file, argp);
5228 return btrfs_ioctl_fitrim(file, argp);
5229 case BTRFS_IOC_SNAP_CREATE:
5230 return btrfs_ioctl_snap_create(file, argp, 0);
5231 case BTRFS_IOC_SNAP_CREATE_V2:
5232 return btrfs_ioctl_snap_create_v2(file, argp, 0);
5233 case BTRFS_IOC_SUBVOL_CREATE:
5234 return btrfs_ioctl_snap_create(file, argp, 1);
5235 case BTRFS_IOC_SUBVOL_CREATE_V2:
5236 return btrfs_ioctl_snap_create_v2(file, argp, 1);
5237 case BTRFS_IOC_SNAP_DESTROY:
5238 return btrfs_ioctl_snap_destroy(file, argp);
5239 case BTRFS_IOC_SUBVOL_GETFLAGS:
5240 return btrfs_ioctl_subvol_getflags(file, argp);
5241 case BTRFS_IOC_SUBVOL_SETFLAGS:
5242 return btrfs_ioctl_subvol_setflags(file, argp);
5243 case BTRFS_IOC_DEFAULT_SUBVOL:
5244 return btrfs_ioctl_default_subvol(file, argp);
5245 case BTRFS_IOC_DEFRAG:
5246 return btrfs_ioctl_defrag(file, NULL);
5247 case BTRFS_IOC_DEFRAG_RANGE:
5248 return btrfs_ioctl_defrag(file, argp);
5249 case BTRFS_IOC_RESIZE:
5250 return btrfs_ioctl_resize(file, argp);
5251 case BTRFS_IOC_ADD_DEV:
5252 return btrfs_ioctl_add_dev(root, argp);
5253 case BTRFS_IOC_RM_DEV:
5254 return btrfs_ioctl_rm_dev(file, argp);
5255 case BTRFS_IOC_FS_INFO:
5256 return btrfs_ioctl_fs_info(root, argp);
5257 case BTRFS_IOC_DEV_INFO:
5258 return btrfs_ioctl_dev_info(root, argp);
5259 case BTRFS_IOC_BALANCE:
5260 return btrfs_ioctl_balance(file, NULL);
5261 case BTRFS_IOC_CLONE:
5262 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
5263 case BTRFS_IOC_CLONE_RANGE:
5264 return btrfs_ioctl_clone_range(file, argp);
5265 case BTRFS_IOC_TRANS_START:
5266 return btrfs_ioctl_trans_start(file);
5267 case BTRFS_IOC_TRANS_END:
5268 return btrfs_ioctl_trans_end(file);
5269 case BTRFS_IOC_TREE_SEARCH:
5270 return btrfs_ioctl_tree_search(file, argp);
5271 case BTRFS_IOC_TREE_SEARCH_V2:
5272 return btrfs_ioctl_tree_search_v2(file, argp);
5273 case BTRFS_IOC_INO_LOOKUP:
5274 return btrfs_ioctl_ino_lookup(file, argp);
5275 case BTRFS_IOC_INO_PATHS:
5276 return btrfs_ioctl_ino_to_path(root, argp);
5277 case BTRFS_IOC_LOGICAL_INO:
5278 return btrfs_ioctl_logical_to_ino(root, argp);
5279 case BTRFS_IOC_SPACE_INFO:
5280 return btrfs_ioctl_space_info(root, argp);
5281 case BTRFS_IOC_SYNC: {
5284 ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1);
5287 ret = btrfs_sync_fs(file->f_dentry->d_sb, 1);
5289 * The transaction thread may want to do more work,
5290 * namely it pokes the cleaner ktread that will start
5291 * processing uncleaned subvols.
5293 wake_up_process(root->fs_info->transaction_kthread);
5296 case BTRFS_IOC_START_SYNC:
5297 return btrfs_ioctl_start_sync(root, argp);
5298 case BTRFS_IOC_WAIT_SYNC:
5299 return btrfs_ioctl_wait_sync(root, argp);
5300 case BTRFS_IOC_SCRUB:
5301 return btrfs_ioctl_scrub(file, argp);
5302 case BTRFS_IOC_SCRUB_CANCEL:
5303 return btrfs_ioctl_scrub_cancel(root, argp);
5304 case BTRFS_IOC_SCRUB_PROGRESS:
5305 return btrfs_ioctl_scrub_progress(root, argp);
5306 case BTRFS_IOC_BALANCE_V2:
5307 return btrfs_ioctl_balance(file, argp);
5308 case BTRFS_IOC_BALANCE_CTL:
5309 return btrfs_ioctl_balance_ctl(root, arg);
5310 case BTRFS_IOC_BALANCE_PROGRESS:
5311 return btrfs_ioctl_balance_progress(root, argp);
5312 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5313 return btrfs_ioctl_set_received_subvol(file, argp);
5315 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5316 return btrfs_ioctl_set_received_subvol_32(file, argp);
5318 case BTRFS_IOC_SEND:
5319 return btrfs_ioctl_send(file, argp);
5320 case BTRFS_IOC_GET_DEV_STATS:
5321 return btrfs_ioctl_get_dev_stats(root, argp);
5322 case BTRFS_IOC_QUOTA_CTL:
5323 return btrfs_ioctl_quota_ctl(file, argp);
5324 case BTRFS_IOC_QGROUP_ASSIGN:
5325 return btrfs_ioctl_qgroup_assign(file, argp);
5326 case BTRFS_IOC_QGROUP_CREATE:
5327 return btrfs_ioctl_qgroup_create(file, argp);
5328 case BTRFS_IOC_QGROUP_LIMIT:
5329 return btrfs_ioctl_qgroup_limit(file, argp);
5330 case BTRFS_IOC_QUOTA_RESCAN:
5331 return btrfs_ioctl_quota_rescan(file, argp);
5332 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5333 return btrfs_ioctl_quota_rescan_status(file, argp);
5334 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5335 return btrfs_ioctl_quota_rescan_wait(file, argp);
5336 case BTRFS_IOC_DEV_REPLACE:
5337 return btrfs_ioctl_dev_replace(root, argp);
5338 case BTRFS_IOC_GET_FSLABEL:
5339 return btrfs_ioctl_get_fslabel(file, argp);
5340 case BTRFS_IOC_SET_FSLABEL:
5341 return btrfs_ioctl_set_fslabel(file, argp);
5342 case BTRFS_IOC_FILE_EXTENT_SAME:
5343 return btrfs_ioctl_file_extent_same(file, argp);
5344 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5345 return btrfs_ioctl_get_supported_features(file, argp);
5346 case BTRFS_IOC_GET_FEATURES:
5347 return btrfs_ioctl_get_features(file, argp);
5348 case BTRFS_IOC_SET_FEATURES:
5349 return btrfs_ioctl_set_features(file, argp);