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"
63 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
64 * structures are incorrect, as the timespec structure from userspace
65 * is 4 bytes too small. We define these alternatives here to teach
66 * the kernel about the 32-bit struct packing.
68 struct btrfs_ioctl_timespec_32 {
71 } __attribute__ ((__packed__));
73 struct btrfs_ioctl_received_subvol_args_32 {
74 char uuid[BTRFS_UUID_SIZE]; /* in */
75 __u64 stransid; /* in */
76 __u64 rtransid; /* out */
77 struct btrfs_ioctl_timespec_32 stime; /* in */
78 struct btrfs_ioctl_timespec_32 rtime; /* out */
80 __u64 reserved[16]; /* in */
81 } __attribute__ ((__packed__));
83 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
84 struct btrfs_ioctl_received_subvol_args_32)
88 static int btrfs_clone(struct inode *src, struct inode *inode,
89 u64 off, u64 olen, u64 olen_aligned, u64 destoff);
91 /* Mask out flags that are inappropriate for the given type of inode. */
92 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
96 else if (S_ISREG(mode))
97 return flags & ~FS_DIRSYNC_FL;
99 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
103 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
105 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
107 unsigned int iflags = 0;
109 if (flags & BTRFS_INODE_SYNC)
110 iflags |= FS_SYNC_FL;
111 if (flags & BTRFS_INODE_IMMUTABLE)
112 iflags |= FS_IMMUTABLE_FL;
113 if (flags & BTRFS_INODE_APPEND)
114 iflags |= FS_APPEND_FL;
115 if (flags & BTRFS_INODE_NODUMP)
116 iflags |= FS_NODUMP_FL;
117 if (flags & BTRFS_INODE_NOATIME)
118 iflags |= FS_NOATIME_FL;
119 if (flags & BTRFS_INODE_DIRSYNC)
120 iflags |= FS_DIRSYNC_FL;
121 if (flags & BTRFS_INODE_NODATACOW)
122 iflags |= FS_NOCOW_FL;
124 if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
125 iflags |= FS_COMPR_FL;
126 else if (flags & BTRFS_INODE_NOCOMPRESS)
127 iflags |= FS_NOCOMP_FL;
133 * Update inode->i_flags based on the btrfs internal flags.
135 void btrfs_update_iflags(struct inode *inode)
137 struct btrfs_inode *ip = BTRFS_I(inode);
139 inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
141 if (ip->flags & BTRFS_INODE_SYNC)
142 inode->i_flags |= S_SYNC;
143 if (ip->flags & BTRFS_INODE_IMMUTABLE)
144 inode->i_flags |= S_IMMUTABLE;
145 if (ip->flags & BTRFS_INODE_APPEND)
146 inode->i_flags |= S_APPEND;
147 if (ip->flags & BTRFS_INODE_NOATIME)
148 inode->i_flags |= S_NOATIME;
149 if (ip->flags & BTRFS_INODE_DIRSYNC)
150 inode->i_flags |= S_DIRSYNC;
154 * Inherit flags from the parent inode.
156 * Currently only the compression flags and the cow flags are inherited.
158 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
165 flags = BTRFS_I(dir)->flags;
167 if (flags & BTRFS_INODE_NOCOMPRESS) {
168 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
169 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
170 } else if (flags & BTRFS_INODE_COMPRESS) {
171 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
172 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
175 if (flags & BTRFS_INODE_NODATACOW) {
176 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
177 if (S_ISREG(inode->i_mode))
178 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
181 btrfs_update_iflags(inode);
184 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
186 struct btrfs_inode *ip = BTRFS_I(file_inode(file));
187 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
189 if (copy_to_user(arg, &flags, sizeof(flags)))
194 static int check_flags(unsigned int flags)
196 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
197 FS_NOATIME_FL | FS_NODUMP_FL | \
198 FS_SYNC_FL | FS_DIRSYNC_FL | \
199 FS_NOCOMP_FL | FS_COMPR_FL |
203 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
209 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
211 struct inode *inode = file_inode(file);
212 struct btrfs_inode *ip = BTRFS_I(inode);
213 struct btrfs_root *root = ip->root;
214 struct btrfs_trans_handle *trans;
215 unsigned int flags, oldflags;
218 unsigned int i_oldflags;
221 if (!inode_owner_or_capable(inode))
224 if (btrfs_root_readonly(root))
227 if (copy_from_user(&flags, arg, sizeof(flags)))
230 ret = check_flags(flags);
234 ret = mnt_want_write_file(file);
238 mutex_lock(&inode->i_mutex);
240 ip_oldflags = ip->flags;
241 i_oldflags = inode->i_flags;
242 mode = inode->i_mode;
244 flags = btrfs_mask_flags(inode->i_mode, flags);
245 oldflags = btrfs_flags_to_ioctl(ip->flags);
246 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
247 if (!capable(CAP_LINUX_IMMUTABLE)) {
253 if (flags & FS_SYNC_FL)
254 ip->flags |= BTRFS_INODE_SYNC;
256 ip->flags &= ~BTRFS_INODE_SYNC;
257 if (flags & FS_IMMUTABLE_FL)
258 ip->flags |= BTRFS_INODE_IMMUTABLE;
260 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
261 if (flags & FS_APPEND_FL)
262 ip->flags |= BTRFS_INODE_APPEND;
264 ip->flags &= ~BTRFS_INODE_APPEND;
265 if (flags & FS_NODUMP_FL)
266 ip->flags |= BTRFS_INODE_NODUMP;
268 ip->flags &= ~BTRFS_INODE_NODUMP;
269 if (flags & FS_NOATIME_FL)
270 ip->flags |= BTRFS_INODE_NOATIME;
272 ip->flags &= ~BTRFS_INODE_NOATIME;
273 if (flags & FS_DIRSYNC_FL)
274 ip->flags |= BTRFS_INODE_DIRSYNC;
276 ip->flags &= ~BTRFS_INODE_DIRSYNC;
277 if (flags & FS_NOCOW_FL) {
280 * It's safe to turn csums off here, no extents exist.
281 * Otherwise we want the flag to reflect the real COW
282 * status of the file and will not set it.
284 if (inode->i_size == 0)
285 ip->flags |= BTRFS_INODE_NODATACOW
286 | BTRFS_INODE_NODATASUM;
288 ip->flags |= BTRFS_INODE_NODATACOW;
292 * Revert back under same assuptions as above
295 if (inode->i_size == 0)
296 ip->flags &= ~(BTRFS_INODE_NODATACOW
297 | BTRFS_INODE_NODATASUM);
299 ip->flags &= ~BTRFS_INODE_NODATACOW;
304 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
305 * flag may be changed automatically if compression code won't make
308 if (flags & FS_NOCOMP_FL) {
309 ip->flags &= ~BTRFS_INODE_COMPRESS;
310 ip->flags |= BTRFS_INODE_NOCOMPRESS;
312 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
313 if (ret && ret != -ENODATA)
315 } else if (flags & FS_COMPR_FL) {
318 ip->flags |= BTRFS_INODE_COMPRESS;
319 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
321 if (root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
325 ret = btrfs_set_prop(inode, "btrfs.compression",
326 comp, strlen(comp), 0);
331 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
334 trans = btrfs_start_transaction(root, 1);
336 ret = PTR_ERR(trans);
340 btrfs_update_iflags(inode);
341 inode_inc_iversion(inode);
342 inode->i_ctime = CURRENT_TIME;
343 ret = btrfs_update_inode(trans, root, inode);
345 btrfs_end_transaction(trans, root);
348 ip->flags = ip_oldflags;
349 inode->i_flags = i_oldflags;
353 mutex_unlock(&inode->i_mutex);
354 mnt_drop_write_file(file);
358 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
360 struct inode *inode = file_inode(file);
362 return put_user(inode->i_generation, arg);
365 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
367 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
368 struct btrfs_device *device;
369 struct request_queue *q;
370 struct fstrim_range range;
371 u64 minlen = ULLONG_MAX;
373 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
376 if (!capable(CAP_SYS_ADMIN))
380 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
384 q = bdev_get_queue(device->bdev);
385 if (blk_queue_discard(q)) {
387 minlen = min((u64)q->limits.discard_granularity,
395 if (copy_from_user(&range, arg, sizeof(range)))
397 if (range.start > total_bytes ||
398 range.len < fs_info->sb->s_blocksize)
401 range.len = min(range.len, total_bytes - range.start);
402 range.minlen = max(range.minlen, minlen);
403 ret = btrfs_trim_fs(fs_info->tree_root, &range);
407 if (copy_to_user(arg, &range, sizeof(range)))
413 int btrfs_is_empty_uuid(u8 *uuid)
417 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
424 static noinline int create_subvol(struct inode *dir,
425 struct dentry *dentry,
426 char *name, int namelen,
428 struct btrfs_qgroup_inherit *inherit)
430 struct btrfs_trans_handle *trans;
431 struct btrfs_key key;
432 struct btrfs_root_item root_item;
433 struct btrfs_inode_item *inode_item;
434 struct extent_buffer *leaf;
435 struct btrfs_root *root = BTRFS_I(dir)->root;
436 struct btrfs_root *new_root;
437 struct btrfs_block_rsv block_rsv;
438 struct timespec cur_time = CURRENT_TIME;
443 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
448 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
452 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
454 * The same as the snapshot creation, please see the comment
455 * of create_snapshot().
457 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
458 8, &qgroup_reserved, false);
462 trans = btrfs_start_transaction(root, 0);
464 ret = PTR_ERR(trans);
465 btrfs_subvolume_release_metadata(root, &block_rsv,
469 trans->block_rsv = &block_rsv;
470 trans->bytes_reserved = block_rsv.size;
472 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
476 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
477 0, objectid, NULL, 0, 0, 0);
483 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
484 btrfs_set_header_bytenr(leaf, leaf->start);
485 btrfs_set_header_generation(leaf, trans->transid);
486 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
487 btrfs_set_header_owner(leaf, objectid);
489 write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(),
491 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
492 btrfs_header_chunk_tree_uuid(leaf),
494 btrfs_mark_buffer_dirty(leaf);
496 memset(&root_item, 0, sizeof(root_item));
498 inode_item = &root_item.inode;
499 btrfs_set_stack_inode_generation(inode_item, 1);
500 btrfs_set_stack_inode_size(inode_item, 3);
501 btrfs_set_stack_inode_nlink(inode_item, 1);
502 btrfs_set_stack_inode_nbytes(inode_item, root->leafsize);
503 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
505 btrfs_set_root_flags(&root_item, 0);
506 btrfs_set_root_limit(&root_item, 0);
507 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
509 btrfs_set_root_bytenr(&root_item, leaf->start);
510 btrfs_set_root_generation(&root_item, trans->transid);
511 btrfs_set_root_level(&root_item, 0);
512 btrfs_set_root_refs(&root_item, 1);
513 btrfs_set_root_used(&root_item, leaf->len);
514 btrfs_set_root_last_snapshot(&root_item, 0);
516 btrfs_set_root_generation_v2(&root_item,
517 btrfs_root_generation(&root_item));
518 uuid_le_gen(&new_uuid);
519 memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
520 btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec);
521 btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec);
522 root_item.ctime = root_item.otime;
523 btrfs_set_root_ctransid(&root_item, trans->transid);
524 btrfs_set_root_otransid(&root_item, trans->transid);
526 btrfs_tree_unlock(leaf);
527 free_extent_buffer(leaf);
530 btrfs_set_root_dirid(&root_item, new_dirid);
532 key.objectid = objectid;
534 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
535 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
540 key.offset = (u64)-1;
541 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
542 if (IS_ERR(new_root)) {
543 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
544 ret = PTR_ERR(new_root);
548 btrfs_record_root_in_trans(trans, new_root);
550 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
552 /* We potentially lose an unused inode item here */
553 btrfs_abort_transaction(trans, root, ret);
558 * insert the directory item
560 ret = btrfs_set_inode_index(dir, &index);
562 btrfs_abort_transaction(trans, root, ret);
566 ret = btrfs_insert_dir_item(trans, root,
567 name, namelen, dir, &key,
568 BTRFS_FT_DIR, index);
570 btrfs_abort_transaction(trans, root, ret);
574 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
575 ret = btrfs_update_inode(trans, root, dir);
578 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
579 objectid, root->root_key.objectid,
580 btrfs_ino(dir), index, name, namelen);
583 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
584 root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
587 btrfs_abort_transaction(trans, root, ret);
590 trans->block_rsv = NULL;
591 trans->bytes_reserved = 0;
592 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
595 *async_transid = trans->transid;
596 err = btrfs_commit_transaction_async(trans, root, 1);
598 err = btrfs_commit_transaction(trans, root);
600 err = btrfs_commit_transaction(trans, root);
606 inode = btrfs_lookup_dentry(dir, dentry);
608 return PTR_ERR(inode);
609 d_instantiate(dentry, inode);
614 static void btrfs_wait_nocow_write(struct btrfs_root *root)
620 prepare_to_wait(&root->subv_writers->wait, &wait,
621 TASK_UNINTERRUPTIBLE);
623 writers = percpu_counter_sum(&root->subv_writers->counter);
627 finish_wait(&root->subv_writers->wait, &wait);
631 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
632 struct dentry *dentry, char *name, int namelen,
633 u64 *async_transid, bool readonly,
634 struct btrfs_qgroup_inherit *inherit)
637 struct btrfs_pending_snapshot *pending_snapshot;
638 struct btrfs_trans_handle *trans;
644 atomic_inc(&root->will_be_snapshoted);
645 smp_mb__after_atomic_inc();
646 btrfs_wait_nocow_write(root);
648 ret = btrfs_start_delalloc_inodes(root, 0);
652 btrfs_wait_ordered_extents(root, -1);
654 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
655 if (!pending_snapshot) {
660 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
661 BTRFS_BLOCK_RSV_TEMP);
663 * 1 - parent dir inode
666 * 2 - root ref/backref
667 * 1 - root of snapshot
670 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
671 &pending_snapshot->block_rsv, 8,
672 &pending_snapshot->qgroup_reserved,
677 pending_snapshot->dentry = dentry;
678 pending_snapshot->root = root;
679 pending_snapshot->readonly = readonly;
680 pending_snapshot->dir = dir;
681 pending_snapshot->inherit = inherit;
683 trans = btrfs_start_transaction(root, 0);
685 ret = PTR_ERR(trans);
689 spin_lock(&root->fs_info->trans_lock);
690 list_add(&pending_snapshot->list,
691 &trans->transaction->pending_snapshots);
692 spin_unlock(&root->fs_info->trans_lock);
694 *async_transid = trans->transid;
695 ret = btrfs_commit_transaction_async(trans,
696 root->fs_info->extent_root, 1);
698 ret = btrfs_commit_transaction(trans, root);
700 ret = btrfs_commit_transaction(trans,
701 root->fs_info->extent_root);
706 ret = pending_snapshot->error;
710 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
714 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
716 ret = PTR_ERR(inode);
720 d_instantiate(dentry, inode);
723 btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
724 &pending_snapshot->block_rsv,
725 pending_snapshot->qgroup_reserved);
727 kfree(pending_snapshot);
729 atomic_dec(&root->will_be_snapshoted);
733 /* copy of check_sticky in fs/namei.c()
734 * It's inline, so penalty for filesystems that don't use sticky bit is
737 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
739 kuid_t fsuid = current_fsuid();
741 if (!(dir->i_mode & S_ISVTX))
743 if (uid_eq(inode->i_uid, fsuid))
745 if (uid_eq(dir->i_uid, fsuid))
747 return !capable(CAP_FOWNER);
750 /* copy of may_delete in fs/namei.c()
751 * Check whether we can remove a link victim from directory dir, check
752 * whether the type of victim is right.
753 * 1. We can't do it if dir is read-only (done in permission())
754 * 2. We should have write and exec permissions on dir
755 * 3. We can't remove anything from append-only dir
756 * 4. We can't do anything with immutable dir (done in permission())
757 * 5. If the sticky bit on dir is set we should either
758 * a. be owner of dir, or
759 * b. be owner of victim, or
760 * c. have CAP_FOWNER capability
761 * 6. If the victim is append-only or immutable we can't do antyhing with
762 * links pointing to it.
763 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
764 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
765 * 9. We can't remove a root or mountpoint.
766 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
767 * nfs_async_unlink().
770 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
774 if (!victim->d_inode)
777 BUG_ON(victim->d_parent->d_inode != dir);
778 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
780 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
785 if (btrfs_check_sticky(dir, victim->d_inode)||
786 IS_APPEND(victim->d_inode)||
787 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
790 if (!S_ISDIR(victim->d_inode->i_mode))
794 } else if (S_ISDIR(victim->d_inode->i_mode))
798 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
803 /* copy of may_create in fs/namei.c() */
804 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
810 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
814 * Create a new subvolume below @parent. This is largely modeled after
815 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
816 * inside this filesystem so it's quite a bit simpler.
818 static noinline int btrfs_mksubvol(struct path *parent,
819 char *name, int namelen,
820 struct btrfs_root *snap_src,
821 u64 *async_transid, bool readonly,
822 struct btrfs_qgroup_inherit *inherit)
824 struct inode *dir = parent->dentry->d_inode;
825 struct dentry *dentry;
828 error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
832 dentry = lookup_one_len(name, parent->dentry, namelen);
833 error = PTR_ERR(dentry);
841 error = btrfs_may_create(dir, dentry);
846 * even if this name doesn't exist, we may get hash collisions.
847 * check for them now when we can safely fail
849 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
855 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
857 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
861 error = create_snapshot(snap_src, dir, dentry, name, namelen,
862 async_transid, readonly, inherit);
864 error = create_subvol(dir, dentry, name, namelen,
865 async_transid, inherit);
868 fsnotify_mkdir(dir, dentry);
870 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
874 mutex_unlock(&dir->i_mutex);
879 * When we're defragging a range, we don't want to kick it off again
880 * if it is really just waiting for delalloc to send it down.
881 * If we find a nice big extent or delalloc range for the bytes in the
882 * file you want to defrag, we return 0 to let you know to skip this
885 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
887 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
888 struct extent_map *em = NULL;
889 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
892 read_lock(&em_tree->lock);
893 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
894 read_unlock(&em_tree->lock);
897 end = extent_map_end(em);
899 if (end - offset > thresh)
902 /* if we already have a nice delalloc here, just stop */
904 end = count_range_bits(io_tree, &offset, offset + thresh,
905 thresh, EXTENT_DELALLOC, 1);
912 * helper function to walk through a file and find extents
913 * newer than a specific transid, and smaller than thresh.
915 * This is used by the defragging code to find new and small
918 static int find_new_extents(struct btrfs_root *root,
919 struct inode *inode, u64 newer_than,
920 u64 *off, int thresh)
922 struct btrfs_path *path;
923 struct btrfs_key min_key;
924 struct extent_buffer *leaf;
925 struct btrfs_file_extent_item *extent;
928 u64 ino = btrfs_ino(inode);
930 path = btrfs_alloc_path();
934 min_key.objectid = ino;
935 min_key.type = BTRFS_EXTENT_DATA_KEY;
936 min_key.offset = *off;
938 path->keep_locks = 1;
941 ret = btrfs_search_forward(root, &min_key, path, newer_than);
944 if (min_key.objectid != ino)
946 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
949 leaf = path->nodes[0];
950 extent = btrfs_item_ptr(leaf, path->slots[0],
951 struct btrfs_file_extent_item);
953 type = btrfs_file_extent_type(leaf, extent);
954 if (type == BTRFS_FILE_EXTENT_REG &&
955 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
956 check_defrag_in_cache(inode, min_key.offset, thresh)) {
957 *off = min_key.offset;
958 btrfs_free_path(path);
962 if (min_key.offset == (u64)-1)
966 btrfs_release_path(path);
969 btrfs_free_path(path);
973 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
975 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
976 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
977 struct extent_map *em;
978 u64 len = PAGE_CACHE_SIZE;
981 * hopefully we have this extent in the tree already, try without
982 * the full extent lock
984 read_lock(&em_tree->lock);
985 em = lookup_extent_mapping(em_tree, start, len);
986 read_unlock(&em_tree->lock);
989 /* get the big lock and read metadata off disk */
990 lock_extent(io_tree, start, start + len - 1);
991 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
992 unlock_extent(io_tree, start, start + len - 1);
1001 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1003 struct extent_map *next;
1006 /* this is the last extent */
1007 if (em->start + em->len >= i_size_read(inode))
1010 next = defrag_lookup_extent(inode, em->start + em->len);
1011 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE ||
1012 (em->block_start + em->block_len == next->block_start))
1015 free_extent_map(next);
1019 static int should_defrag_range(struct inode *inode, u64 start, int thresh,
1020 u64 *last_len, u64 *skip, u64 *defrag_end,
1023 struct extent_map *em;
1025 bool next_mergeable = true;
1028 * make sure that once we start defragging an extent, we keep on
1031 if (start < *defrag_end)
1036 em = defrag_lookup_extent(inode, start);
1040 /* this will cover holes, and inline extents */
1041 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1046 next_mergeable = defrag_check_next_extent(inode, em);
1049 * we hit a real extent, if it is big or the next extent is not a
1050 * real extent, don't bother defragging it
1052 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1053 (em->len >= thresh || !next_mergeable))
1057 * last_len ends up being a counter of how many bytes we've defragged.
1058 * every time we choose not to defrag an extent, we reset *last_len
1059 * so that the next tiny extent will force a defrag.
1061 * The end result of this is that tiny extents before a single big
1062 * extent will force at least part of that big extent to be defragged.
1065 *defrag_end = extent_map_end(em);
1068 *skip = extent_map_end(em);
1072 free_extent_map(em);
1077 * it doesn't do much good to defrag one or two pages
1078 * at a time. This pulls in a nice chunk of pages
1079 * to COW and defrag.
1081 * It also makes sure the delalloc code has enough
1082 * dirty data to avoid making new small extents as part
1085 * It's a good idea to start RA on this range
1086 * before calling this.
1088 static int cluster_pages_for_defrag(struct inode *inode,
1089 struct page **pages,
1090 unsigned long start_index,
1091 unsigned long num_pages)
1093 unsigned long file_end;
1094 u64 isize = i_size_read(inode);
1101 struct btrfs_ordered_extent *ordered;
1102 struct extent_state *cached_state = NULL;
1103 struct extent_io_tree *tree;
1104 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1106 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1107 if (!isize || start_index > file_end)
1110 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1112 ret = btrfs_delalloc_reserve_space(inode,
1113 page_cnt << PAGE_CACHE_SHIFT);
1117 tree = &BTRFS_I(inode)->io_tree;
1119 /* step one, lock all the pages */
1120 for (i = 0; i < page_cnt; i++) {
1123 page = find_or_create_page(inode->i_mapping,
1124 start_index + i, mask);
1128 page_start = page_offset(page);
1129 page_end = page_start + PAGE_CACHE_SIZE - 1;
1131 lock_extent(tree, page_start, page_end);
1132 ordered = btrfs_lookup_ordered_extent(inode,
1134 unlock_extent(tree, page_start, page_end);
1139 btrfs_start_ordered_extent(inode, ordered, 1);
1140 btrfs_put_ordered_extent(ordered);
1143 * we unlocked the page above, so we need check if
1144 * it was released or not.
1146 if (page->mapping != inode->i_mapping) {
1148 page_cache_release(page);
1153 if (!PageUptodate(page)) {
1154 btrfs_readpage(NULL, page);
1156 if (!PageUptodate(page)) {
1158 page_cache_release(page);
1164 if (page->mapping != inode->i_mapping) {
1166 page_cache_release(page);
1176 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1180 * so now we have a nice long stream of locked
1181 * and up to date pages, lets wait on them
1183 for (i = 0; i < i_done; i++)
1184 wait_on_page_writeback(pages[i]);
1186 page_start = page_offset(pages[0]);
1187 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1189 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1190 page_start, page_end - 1, 0, &cached_state);
1191 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1192 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1193 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1194 &cached_state, GFP_NOFS);
1196 if (i_done != page_cnt) {
1197 spin_lock(&BTRFS_I(inode)->lock);
1198 BTRFS_I(inode)->outstanding_extents++;
1199 spin_unlock(&BTRFS_I(inode)->lock);
1200 btrfs_delalloc_release_space(inode,
1201 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1205 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1206 &cached_state, GFP_NOFS);
1208 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1209 page_start, page_end - 1, &cached_state,
1212 for (i = 0; i < i_done; i++) {
1213 clear_page_dirty_for_io(pages[i]);
1214 ClearPageChecked(pages[i]);
1215 set_page_extent_mapped(pages[i]);
1216 set_page_dirty(pages[i]);
1217 unlock_page(pages[i]);
1218 page_cache_release(pages[i]);
1222 for (i = 0; i < i_done; i++) {
1223 unlock_page(pages[i]);
1224 page_cache_release(pages[i]);
1226 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1231 int btrfs_defrag_file(struct inode *inode, struct file *file,
1232 struct btrfs_ioctl_defrag_range_args *range,
1233 u64 newer_than, unsigned long max_to_defrag)
1235 struct btrfs_root *root = BTRFS_I(inode)->root;
1236 struct file_ra_state *ra = NULL;
1237 unsigned long last_index;
1238 u64 isize = i_size_read(inode);
1242 u64 newer_off = range->start;
1244 unsigned long ra_index = 0;
1246 int defrag_count = 0;
1247 int compress_type = BTRFS_COMPRESS_ZLIB;
1248 int extent_thresh = range->extent_thresh;
1249 unsigned long max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1250 unsigned long cluster = max_cluster;
1251 u64 new_align = ~((u64)128 * 1024 - 1);
1252 struct page **pages = NULL;
1257 if (range->start >= isize)
1260 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1261 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1263 if (range->compress_type)
1264 compress_type = range->compress_type;
1267 if (extent_thresh == 0)
1268 extent_thresh = 256 * 1024;
1271 * if we were not given a file, allocate a readahead
1275 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1278 file_ra_state_init(ra, inode->i_mapping);
1283 pages = kmalloc_array(max_cluster, sizeof(struct page *),
1290 /* find the last page to defrag */
1291 if (range->start + range->len > range->start) {
1292 last_index = min_t(u64, isize - 1,
1293 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1295 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1299 ret = find_new_extents(root, inode, newer_than,
1300 &newer_off, 64 * 1024);
1302 range->start = newer_off;
1304 * we always align our defrag to help keep
1305 * the extents in the file evenly spaced
1307 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1311 i = range->start >> PAGE_CACHE_SHIFT;
1314 max_to_defrag = last_index + 1;
1317 * make writeback starts from i, so the defrag range can be
1318 * written sequentially.
1320 if (i < inode->i_mapping->writeback_index)
1321 inode->i_mapping->writeback_index = i;
1323 while (i <= last_index && defrag_count < max_to_defrag &&
1324 (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1325 PAGE_CACHE_SHIFT)) {
1327 * make sure we stop running if someone unmounts
1330 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1333 if (btrfs_defrag_cancelled(root->fs_info)) {
1334 printk(KERN_DEBUG "BTRFS: defrag_file cancelled\n");
1339 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1340 extent_thresh, &last_len, &skip,
1341 &defrag_end, range->flags &
1342 BTRFS_DEFRAG_RANGE_COMPRESS)) {
1345 * the should_defrag function tells us how much to skip
1346 * bump our counter by the suggested amount
1348 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1349 i = max(i + 1, next);
1354 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1355 PAGE_CACHE_SHIFT) - i;
1356 cluster = min(cluster, max_cluster);
1358 cluster = max_cluster;
1361 if (i + cluster > ra_index) {
1362 ra_index = max(i, ra_index);
1363 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1365 ra_index += max_cluster;
1368 mutex_lock(&inode->i_mutex);
1369 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1370 BTRFS_I(inode)->force_compress = compress_type;
1371 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1373 mutex_unlock(&inode->i_mutex);
1377 defrag_count += ret;
1378 balance_dirty_pages_ratelimited(inode->i_mapping);
1379 mutex_unlock(&inode->i_mutex);
1382 if (newer_off == (u64)-1)
1388 newer_off = max(newer_off + 1,
1389 (u64)i << PAGE_CACHE_SHIFT);
1391 ret = find_new_extents(root, inode,
1392 newer_than, &newer_off,
1395 range->start = newer_off;
1396 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1403 last_len += ret << PAGE_CACHE_SHIFT;
1411 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1412 filemap_flush(inode->i_mapping);
1413 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1414 &BTRFS_I(inode)->runtime_flags))
1415 filemap_flush(inode->i_mapping);
1418 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1419 /* the filemap_flush will queue IO into the worker threads, but
1420 * we have to make sure the IO is actually started and that
1421 * ordered extents get created before we return
1423 atomic_inc(&root->fs_info->async_submit_draining);
1424 while (atomic_read(&root->fs_info->nr_async_submits) ||
1425 atomic_read(&root->fs_info->async_delalloc_pages)) {
1426 wait_event(root->fs_info->async_submit_wait,
1427 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1428 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1430 atomic_dec(&root->fs_info->async_submit_draining);
1433 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1434 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1440 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1441 mutex_lock(&inode->i_mutex);
1442 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1443 mutex_unlock(&inode->i_mutex);
1451 static noinline int btrfs_ioctl_resize(struct file *file,
1457 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
1458 struct btrfs_ioctl_vol_args *vol_args;
1459 struct btrfs_trans_handle *trans;
1460 struct btrfs_device *device = NULL;
1462 char *devstr = NULL;
1466 if (!capable(CAP_SYS_ADMIN))
1469 ret = mnt_want_write_file(file);
1473 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1475 mnt_drop_write_file(file);
1476 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1479 mutex_lock(&root->fs_info->volume_mutex);
1480 vol_args = memdup_user(arg, sizeof(*vol_args));
1481 if (IS_ERR(vol_args)) {
1482 ret = PTR_ERR(vol_args);
1486 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1488 sizestr = vol_args->name;
1489 devstr = strchr(sizestr, ':');
1492 sizestr = devstr + 1;
1494 devstr = vol_args->name;
1495 devid = simple_strtoull(devstr, &end, 10);
1500 btrfs_info(root->fs_info, "resizing devid %llu", devid);
1503 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
1505 btrfs_info(root->fs_info, "resizer unable to find device %llu",
1511 if (!device->writeable) {
1512 btrfs_info(root->fs_info,
1513 "resizer unable to apply on readonly device %llu",
1519 if (!strcmp(sizestr, "max"))
1520 new_size = device->bdev->bd_inode->i_size;
1522 if (sizestr[0] == '-') {
1525 } else if (sizestr[0] == '+') {
1529 new_size = memparse(sizestr, NULL);
1530 if (new_size == 0) {
1536 if (device->is_tgtdev_for_dev_replace) {
1541 old_size = device->total_bytes;
1544 if (new_size > old_size) {
1548 new_size = old_size - new_size;
1549 } else if (mod > 0) {
1550 if (new_size > ULLONG_MAX - old_size) {
1554 new_size = old_size + new_size;
1557 if (new_size < 256 * 1024 * 1024) {
1561 if (new_size > device->bdev->bd_inode->i_size) {
1566 do_div(new_size, root->sectorsize);
1567 new_size *= root->sectorsize;
1569 printk_in_rcu(KERN_INFO "BTRFS: new size for %s is %llu\n",
1570 rcu_str_deref(device->name), new_size);
1572 if (new_size > old_size) {
1573 trans = btrfs_start_transaction(root, 0);
1574 if (IS_ERR(trans)) {
1575 ret = PTR_ERR(trans);
1578 ret = btrfs_grow_device(trans, device, new_size);
1579 btrfs_commit_transaction(trans, root);
1580 } else if (new_size < old_size) {
1581 ret = btrfs_shrink_device(device, new_size);
1582 } /* equal, nothing need to do */
1587 mutex_unlock(&root->fs_info->volume_mutex);
1588 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
1589 mnt_drop_write_file(file);
1593 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1594 char *name, unsigned long fd, int subvol,
1595 u64 *transid, bool readonly,
1596 struct btrfs_qgroup_inherit *inherit)
1601 ret = mnt_want_write_file(file);
1605 namelen = strlen(name);
1606 if (strchr(name, '/')) {
1608 goto out_drop_write;
1611 if (name[0] == '.' &&
1612 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1614 goto out_drop_write;
1618 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1619 NULL, transid, readonly, inherit);
1621 struct fd src = fdget(fd);
1622 struct inode *src_inode;
1625 goto out_drop_write;
1628 src_inode = file_inode(src.file);
1629 if (src_inode->i_sb != file_inode(file)->i_sb) {
1630 btrfs_info(BTRFS_I(src_inode)->root->fs_info,
1631 "Snapshot src from another FS");
1633 } else if (!inode_owner_or_capable(src_inode)) {
1635 * Subvolume creation is not restricted, but snapshots
1636 * are limited to own subvolumes only
1640 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1641 BTRFS_I(src_inode)->root,
1642 transid, readonly, inherit);
1647 mnt_drop_write_file(file);
1652 static noinline int btrfs_ioctl_snap_create(struct file *file,
1653 void __user *arg, int subvol)
1655 struct btrfs_ioctl_vol_args *vol_args;
1658 vol_args = memdup_user(arg, sizeof(*vol_args));
1659 if (IS_ERR(vol_args))
1660 return PTR_ERR(vol_args);
1661 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1663 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1664 vol_args->fd, subvol,
1671 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1672 void __user *arg, int subvol)
1674 struct btrfs_ioctl_vol_args_v2 *vol_args;
1678 bool readonly = false;
1679 struct btrfs_qgroup_inherit *inherit = NULL;
1681 vol_args = memdup_user(arg, sizeof(*vol_args));
1682 if (IS_ERR(vol_args))
1683 return PTR_ERR(vol_args);
1684 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1686 if (vol_args->flags &
1687 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1688 BTRFS_SUBVOL_QGROUP_INHERIT)) {
1693 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1695 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1697 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1698 if (vol_args->size > PAGE_CACHE_SIZE) {
1702 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1703 if (IS_ERR(inherit)) {
1704 ret = PTR_ERR(inherit);
1709 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1710 vol_args->fd, subvol, ptr,
1713 if (ret == 0 && ptr &&
1715 offsetof(struct btrfs_ioctl_vol_args_v2,
1716 transid), ptr, sizeof(*ptr)))
1724 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1727 struct inode *inode = file_inode(file);
1728 struct btrfs_root *root = BTRFS_I(inode)->root;
1732 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1735 down_read(&root->fs_info->subvol_sem);
1736 if (btrfs_root_readonly(root))
1737 flags |= BTRFS_SUBVOL_RDONLY;
1738 up_read(&root->fs_info->subvol_sem);
1740 if (copy_to_user(arg, &flags, sizeof(flags)))
1746 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1749 struct inode *inode = file_inode(file);
1750 struct btrfs_root *root = BTRFS_I(inode)->root;
1751 struct btrfs_trans_handle *trans;
1756 if (!inode_owner_or_capable(inode))
1759 ret = mnt_want_write_file(file);
1763 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1765 goto out_drop_write;
1768 if (copy_from_user(&flags, arg, sizeof(flags))) {
1770 goto out_drop_write;
1773 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1775 goto out_drop_write;
1778 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1780 goto out_drop_write;
1783 down_write(&root->fs_info->subvol_sem);
1786 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1789 root_flags = btrfs_root_flags(&root->root_item);
1790 if (flags & BTRFS_SUBVOL_RDONLY) {
1791 btrfs_set_root_flags(&root->root_item,
1792 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1795 * Block RO -> RW transition if this subvolume is involved in
1798 spin_lock(&root->root_item_lock);
1799 if (root->send_in_progress == 0) {
1800 btrfs_set_root_flags(&root->root_item,
1801 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1802 spin_unlock(&root->root_item_lock);
1804 spin_unlock(&root->root_item_lock);
1805 btrfs_warn(root->fs_info,
1806 "Attempt to set subvolume %llu read-write during send",
1807 root->root_key.objectid);
1813 trans = btrfs_start_transaction(root, 1);
1814 if (IS_ERR(trans)) {
1815 ret = PTR_ERR(trans);
1819 ret = btrfs_update_root(trans, root->fs_info->tree_root,
1820 &root->root_key, &root->root_item);
1822 btrfs_commit_transaction(trans, root);
1825 btrfs_set_root_flags(&root->root_item, root_flags);
1827 up_write(&root->fs_info->subvol_sem);
1829 mnt_drop_write_file(file);
1835 * helper to check if the subvolume references other subvolumes
1837 static noinline int may_destroy_subvol(struct btrfs_root *root)
1839 struct btrfs_path *path;
1840 struct btrfs_dir_item *di;
1841 struct btrfs_key key;
1845 path = btrfs_alloc_path();
1849 /* Make sure this root isn't set as the default subvol */
1850 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1851 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1852 dir_id, "default", 7, 0);
1853 if (di && !IS_ERR(di)) {
1854 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1855 if (key.objectid == root->root_key.objectid) {
1859 btrfs_release_path(path);
1862 key.objectid = root->root_key.objectid;
1863 key.type = BTRFS_ROOT_REF_KEY;
1864 key.offset = (u64)-1;
1866 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1873 if (path->slots[0] > 0) {
1875 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1876 if (key.objectid == root->root_key.objectid &&
1877 key.type == BTRFS_ROOT_REF_KEY)
1881 btrfs_free_path(path);
1885 static noinline int key_in_sk(struct btrfs_key *key,
1886 struct btrfs_ioctl_search_key *sk)
1888 struct btrfs_key test;
1891 test.objectid = sk->min_objectid;
1892 test.type = sk->min_type;
1893 test.offset = sk->min_offset;
1895 ret = btrfs_comp_cpu_keys(key, &test);
1899 test.objectid = sk->max_objectid;
1900 test.type = sk->max_type;
1901 test.offset = sk->max_offset;
1903 ret = btrfs_comp_cpu_keys(key, &test);
1909 static noinline int copy_to_sk(struct btrfs_root *root,
1910 struct btrfs_path *path,
1911 struct btrfs_key *key,
1912 struct btrfs_ioctl_search_key *sk,
1914 unsigned long *sk_offset,
1918 struct extent_buffer *leaf;
1919 struct btrfs_ioctl_search_header sh;
1920 unsigned long item_off;
1921 unsigned long item_len;
1927 leaf = path->nodes[0];
1928 slot = path->slots[0];
1929 nritems = btrfs_header_nritems(leaf);
1931 if (btrfs_header_generation(leaf) > sk->max_transid) {
1935 found_transid = btrfs_header_generation(leaf);
1937 for (i = slot; i < nritems; i++) {
1938 item_off = btrfs_item_ptr_offset(leaf, i);
1939 item_len = btrfs_item_size_nr(leaf, i);
1941 btrfs_item_key_to_cpu(leaf, key, i);
1942 if (!key_in_sk(key, sk))
1945 if (sizeof(sh) + item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1948 if (sizeof(sh) + item_len + *sk_offset >
1949 BTRFS_SEARCH_ARGS_BUFSIZE) {
1954 sh.objectid = key->objectid;
1955 sh.offset = key->offset;
1956 sh.type = key->type;
1958 sh.transid = found_transid;
1960 /* copy search result header */
1961 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1962 *sk_offset += sizeof(sh);
1965 char *p = buf + *sk_offset;
1967 read_extent_buffer(leaf, p,
1968 item_off, item_len);
1969 *sk_offset += item_len;
1973 if (*num_found >= sk->nr_items)
1978 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1980 else if (key->type < (u8)-1 && key->type < sk->max_type) {
1983 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1993 static noinline int search_ioctl(struct inode *inode,
1994 struct btrfs_ioctl_search_args *args)
1996 struct btrfs_root *root;
1997 struct btrfs_key key;
1998 struct btrfs_path *path;
1999 struct btrfs_ioctl_search_key *sk = &args->key;
2000 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
2003 unsigned long sk_offset = 0;
2005 path = btrfs_alloc_path();
2009 if (sk->tree_id == 0) {
2010 /* search the root of the inode that was passed */
2011 root = BTRFS_I(inode)->root;
2013 key.objectid = sk->tree_id;
2014 key.type = BTRFS_ROOT_ITEM_KEY;
2015 key.offset = (u64)-1;
2016 root = btrfs_read_fs_root_no_name(info, &key);
2018 printk(KERN_ERR "BTRFS: could not find root %llu\n",
2020 btrfs_free_path(path);
2025 key.objectid = sk->min_objectid;
2026 key.type = sk->min_type;
2027 key.offset = sk->min_offset;
2029 path->keep_locks = 1;
2032 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2038 ret = copy_to_sk(root, path, &key, sk, args->buf,
2039 &sk_offset, &num_found);
2040 btrfs_release_path(path);
2041 if (ret || num_found >= sk->nr_items)
2047 sk->nr_items = num_found;
2048 btrfs_free_path(path);
2052 static noinline int btrfs_ioctl_tree_search(struct file *file,
2055 struct btrfs_ioctl_search_args *args;
2056 struct inode *inode;
2059 if (!capable(CAP_SYS_ADMIN))
2062 args = memdup_user(argp, sizeof(*args));
2064 return PTR_ERR(args);
2066 inode = file_inode(file);
2067 ret = search_ioctl(inode, args);
2068 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2075 * Search INODE_REFs to identify path name of 'dirid' directory
2076 * in a 'tree_id' tree. and sets path name to 'name'.
2078 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2079 u64 tree_id, u64 dirid, char *name)
2081 struct btrfs_root *root;
2082 struct btrfs_key key;
2088 struct btrfs_inode_ref *iref;
2089 struct extent_buffer *l;
2090 struct btrfs_path *path;
2092 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2097 path = btrfs_alloc_path();
2101 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
2103 key.objectid = tree_id;
2104 key.type = BTRFS_ROOT_ITEM_KEY;
2105 key.offset = (u64)-1;
2106 root = btrfs_read_fs_root_no_name(info, &key);
2108 printk(KERN_ERR "BTRFS: could not find root %llu\n", tree_id);
2113 key.objectid = dirid;
2114 key.type = BTRFS_INODE_REF_KEY;
2115 key.offset = (u64)-1;
2118 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2122 ret = btrfs_previous_item(root, path, dirid,
2123 BTRFS_INODE_REF_KEY);
2133 slot = path->slots[0];
2134 btrfs_item_key_to_cpu(l, &key, slot);
2136 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2137 len = btrfs_inode_ref_name_len(l, iref);
2139 total_len += len + 1;
2141 ret = -ENAMETOOLONG;
2146 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2148 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2151 btrfs_release_path(path);
2152 key.objectid = key.offset;
2153 key.offset = (u64)-1;
2154 dirid = key.objectid;
2156 memmove(name, ptr, total_len);
2157 name[total_len] = '\0';
2160 btrfs_free_path(path);
2164 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2167 struct btrfs_ioctl_ino_lookup_args *args;
2168 struct inode *inode;
2171 if (!capable(CAP_SYS_ADMIN))
2174 args = memdup_user(argp, sizeof(*args));
2176 return PTR_ERR(args);
2178 inode = file_inode(file);
2180 if (args->treeid == 0)
2181 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2183 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2184 args->treeid, args->objectid,
2187 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2194 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2197 struct dentry *parent = file->f_path.dentry;
2198 struct dentry *dentry;
2199 struct inode *dir = parent->d_inode;
2200 struct inode *inode;
2201 struct btrfs_root *root = BTRFS_I(dir)->root;
2202 struct btrfs_root *dest = NULL;
2203 struct btrfs_ioctl_vol_args *vol_args;
2204 struct btrfs_trans_handle *trans;
2205 struct btrfs_block_rsv block_rsv;
2206 u64 qgroup_reserved;
2211 vol_args = memdup_user(arg, sizeof(*vol_args));
2212 if (IS_ERR(vol_args))
2213 return PTR_ERR(vol_args);
2215 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2216 namelen = strlen(vol_args->name);
2217 if (strchr(vol_args->name, '/') ||
2218 strncmp(vol_args->name, "..", namelen) == 0) {
2223 err = mnt_want_write_file(file);
2227 err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2229 goto out_drop_write;
2230 dentry = lookup_one_len(vol_args->name, parent, namelen);
2231 if (IS_ERR(dentry)) {
2232 err = PTR_ERR(dentry);
2233 goto out_unlock_dir;
2236 if (!dentry->d_inode) {
2241 inode = dentry->d_inode;
2242 dest = BTRFS_I(inode)->root;
2243 if (!capable(CAP_SYS_ADMIN)) {
2245 * Regular user. Only allow this with a special mount
2246 * option, when the user has write+exec access to the
2247 * subvol root, and when rmdir(2) would have been
2250 * Note that this is _not_ check that the subvol is
2251 * empty or doesn't contain data that we wouldn't
2252 * otherwise be able to delete.
2254 * Users who want to delete empty subvols should try
2258 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2262 * Do not allow deletion if the parent dir is the same
2263 * as the dir to be deleted. That means the ioctl
2264 * must be called on the dentry referencing the root
2265 * of the subvol, not a random directory contained
2272 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2277 /* check if subvolume may be deleted by a user */
2278 err = btrfs_may_delete(dir, dentry, 1);
2282 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2287 mutex_lock(&inode->i_mutex);
2288 err = d_invalidate(dentry);
2292 down_write(&root->fs_info->subvol_sem);
2294 err = may_destroy_subvol(dest);
2298 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2300 * One for dir inode, two for dir entries, two for root
2303 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2304 5, &qgroup_reserved, true);
2308 trans = btrfs_start_transaction(root, 0);
2309 if (IS_ERR(trans)) {
2310 err = PTR_ERR(trans);
2313 trans->block_rsv = &block_rsv;
2314 trans->bytes_reserved = block_rsv.size;
2316 ret = btrfs_unlink_subvol(trans, root, dir,
2317 dest->root_key.objectid,
2318 dentry->d_name.name,
2319 dentry->d_name.len);
2322 btrfs_abort_transaction(trans, root, ret);
2326 btrfs_record_root_in_trans(trans, dest);
2328 memset(&dest->root_item.drop_progress, 0,
2329 sizeof(dest->root_item.drop_progress));
2330 dest->root_item.drop_level = 0;
2331 btrfs_set_root_refs(&dest->root_item, 0);
2333 if (!xchg(&dest->orphan_item_inserted, 1)) {
2334 ret = btrfs_insert_orphan_item(trans,
2335 root->fs_info->tree_root,
2336 dest->root_key.objectid);
2338 btrfs_abort_transaction(trans, root, ret);
2344 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2345 dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
2346 dest->root_key.objectid);
2347 if (ret && ret != -ENOENT) {
2348 btrfs_abort_transaction(trans, root, ret);
2352 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2353 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2354 dest->root_item.received_uuid,
2355 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2356 dest->root_key.objectid);
2357 if (ret && ret != -ENOENT) {
2358 btrfs_abort_transaction(trans, root, ret);
2365 trans->block_rsv = NULL;
2366 trans->bytes_reserved = 0;
2367 ret = btrfs_end_transaction(trans, root);
2370 inode->i_flags |= S_DEAD;
2372 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
2374 up_write(&root->fs_info->subvol_sem);
2376 mutex_unlock(&inode->i_mutex);
2378 shrink_dcache_sb(root->fs_info->sb);
2379 btrfs_invalidate_inodes(dest);
2383 if (dest->cache_inode) {
2384 iput(dest->cache_inode);
2385 dest->cache_inode = NULL;
2391 mutex_unlock(&dir->i_mutex);
2393 mnt_drop_write_file(file);
2399 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2401 struct inode *inode = file_inode(file);
2402 struct btrfs_root *root = BTRFS_I(inode)->root;
2403 struct btrfs_ioctl_defrag_range_args *range;
2406 ret = mnt_want_write_file(file);
2410 if (btrfs_root_readonly(root)) {
2415 switch (inode->i_mode & S_IFMT) {
2417 if (!capable(CAP_SYS_ADMIN)) {
2421 ret = btrfs_defrag_root(root);
2424 ret = btrfs_defrag_root(root->fs_info->extent_root);
2427 if (!(file->f_mode & FMODE_WRITE)) {
2432 range = kzalloc(sizeof(*range), GFP_KERNEL);
2439 if (copy_from_user(range, argp,
2445 /* compression requires us to start the IO */
2446 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2447 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2448 range->extent_thresh = (u32)-1;
2451 /* the rest are all set to zero by kzalloc */
2452 range->len = (u64)-1;
2454 ret = btrfs_defrag_file(file_inode(file), file,
2464 mnt_drop_write_file(file);
2468 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2470 struct btrfs_ioctl_vol_args *vol_args;
2473 if (!capable(CAP_SYS_ADMIN))
2476 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2478 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2481 mutex_lock(&root->fs_info->volume_mutex);
2482 vol_args = memdup_user(arg, sizeof(*vol_args));
2483 if (IS_ERR(vol_args)) {
2484 ret = PTR_ERR(vol_args);
2488 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2489 ret = btrfs_init_new_device(root, vol_args->name);
2493 mutex_unlock(&root->fs_info->volume_mutex);
2494 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2498 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2500 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2501 struct btrfs_ioctl_vol_args *vol_args;
2504 if (!capable(CAP_SYS_ADMIN))
2507 ret = mnt_want_write_file(file);
2511 vol_args = memdup_user(arg, sizeof(*vol_args));
2512 if (IS_ERR(vol_args)) {
2513 ret = PTR_ERR(vol_args);
2517 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2519 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2521 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2525 mutex_lock(&root->fs_info->volume_mutex);
2526 ret = btrfs_rm_device(root, vol_args->name);
2527 mutex_unlock(&root->fs_info->volume_mutex);
2528 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2532 mnt_drop_write_file(file);
2536 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2538 struct btrfs_ioctl_fs_info_args *fi_args;
2539 struct btrfs_device *device;
2540 struct btrfs_device *next;
2541 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2544 if (!capable(CAP_SYS_ADMIN))
2547 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2551 mutex_lock(&fs_devices->device_list_mutex);
2552 fi_args->num_devices = fs_devices->num_devices;
2553 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2555 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2556 if (device->devid > fi_args->max_id)
2557 fi_args->max_id = device->devid;
2559 mutex_unlock(&fs_devices->device_list_mutex);
2561 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2568 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2570 struct btrfs_ioctl_dev_info_args *di_args;
2571 struct btrfs_device *dev;
2572 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2574 char *s_uuid = NULL;
2576 if (!capable(CAP_SYS_ADMIN))
2579 di_args = memdup_user(arg, sizeof(*di_args));
2580 if (IS_ERR(di_args))
2581 return PTR_ERR(di_args);
2583 if (!btrfs_is_empty_uuid(di_args->uuid))
2584 s_uuid = di_args->uuid;
2586 mutex_lock(&fs_devices->device_list_mutex);
2587 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
2594 di_args->devid = dev->devid;
2595 di_args->bytes_used = dev->bytes_used;
2596 di_args->total_bytes = dev->total_bytes;
2597 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2599 struct rcu_string *name;
2602 name = rcu_dereference(dev->name);
2603 strncpy(di_args->path, name->str, sizeof(di_args->path));
2605 di_args->path[sizeof(di_args->path) - 1] = 0;
2607 di_args->path[0] = '\0';
2611 mutex_unlock(&fs_devices->device_list_mutex);
2612 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2619 static struct page *extent_same_get_page(struct inode *inode, u64 off)
2623 struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
2625 index = off >> PAGE_CACHE_SHIFT;
2627 page = grab_cache_page(inode->i_mapping, index);
2631 if (!PageUptodate(page)) {
2632 if (extent_read_full_page_nolock(tree, page, btrfs_get_extent,
2636 if (!PageUptodate(page)) {
2638 page_cache_release(page);
2647 static inline void lock_extent_range(struct inode *inode, u64 off, u64 len)
2649 /* do any pending delalloc/csum calc on src, one way or
2650 another, and lock file content */
2652 struct btrfs_ordered_extent *ordered;
2653 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2654 ordered = btrfs_lookup_first_ordered_extent(inode,
2657 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2658 off + len - 1, EXTENT_DELALLOC, 0, NULL))
2660 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2662 btrfs_put_ordered_extent(ordered);
2663 btrfs_wait_ordered_range(inode, off, len);
2667 static void btrfs_double_unlock(struct inode *inode1, u64 loff1,
2668 struct inode *inode2, u64 loff2, u64 len)
2670 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2671 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2673 mutex_unlock(&inode1->i_mutex);
2674 mutex_unlock(&inode2->i_mutex);
2677 static void btrfs_double_lock(struct inode *inode1, u64 loff1,
2678 struct inode *inode2, u64 loff2, u64 len)
2680 if (inode1 < inode2) {
2681 swap(inode1, inode2);
2685 mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT);
2686 lock_extent_range(inode1, loff1, len);
2687 if (inode1 != inode2) {
2688 mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD);
2689 lock_extent_range(inode2, loff2, len);
2693 static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
2694 u64 dst_loff, u64 len)
2697 struct page *src_page, *dst_page;
2698 unsigned int cmp_len = PAGE_CACHE_SIZE;
2699 void *addr, *dst_addr;
2702 if (len < PAGE_CACHE_SIZE)
2705 src_page = extent_same_get_page(src, loff);
2708 dst_page = extent_same_get_page(dst, dst_loff);
2710 page_cache_release(src_page);
2713 addr = kmap_atomic(src_page);
2714 dst_addr = kmap_atomic(dst_page);
2716 flush_dcache_page(src_page);
2717 flush_dcache_page(dst_page);
2719 if (memcmp(addr, dst_addr, cmp_len))
2720 ret = BTRFS_SAME_DATA_DIFFERS;
2722 kunmap_atomic(addr);
2723 kunmap_atomic(dst_addr);
2724 page_cache_release(src_page);
2725 page_cache_release(dst_page);
2731 dst_loff += cmp_len;
2738 static int extent_same_check_offsets(struct inode *inode, u64 off, u64 len)
2740 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
2742 if (off + len > inode->i_size || off + len < off)
2744 /* Check that we are block aligned - btrfs_clone() requires this */
2745 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
2751 static int btrfs_extent_same(struct inode *src, u64 loff, u64 len,
2752 struct inode *dst, u64 dst_loff)
2757 * btrfs_clone() can't handle extents in the same file
2758 * yet. Once that works, we can drop this check and replace it
2759 * with a check for the same inode, but overlapping extents.
2764 btrfs_double_lock(src, loff, dst, dst_loff, len);
2766 ret = extent_same_check_offsets(src, loff, len);
2770 ret = extent_same_check_offsets(dst, dst_loff, len);
2774 /* don't make the dst file partly checksummed */
2775 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2776 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
2781 ret = btrfs_cmp_data(src, loff, dst, dst_loff, len);
2783 ret = btrfs_clone(src, dst, loff, len, len, dst_loff);
2786 btrfs_double_unlock(src, loff, dst, dst_loff, len);
2791 #define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024)
2793 static long btrfs_ioctl_file_extent_same(struct file *file,
2796 struct btrfs_ioctl_same_args tmp;
2797 struct btrfs_ioctl_same_args *same;
2798 struct btrfs_ioctl_same_extent_info *info;
2799 struct inode *src = file->f_dentry->d_inode;
2800 struct file *dst_file = NULL;
2807 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
2808 bool is_admin = capable(CAP_SYS_ADMIN);
2810 if (!(file->f_mode & FMODE_READ))
2813 ret = mnt_want_write_file(file);
2817 if (copy_from_user(&tmp,
2818 (struct btrfs_ioctl_same_args __user *)argp,
2824 size = sizeof(tmp) +
2825 tmp.dest_count * sizeof(struct btrfs_ioctl_same_extent_info);
2827 same = memdup_user((struct btrfs_ioctl_same_args __user *)argp, size);
2830 ret = PTR_ERR(same);
2834 off = same->logical_offset;
2838 * Limit the total length we will dedupe for each operation.
2839 * This is intended to bound the total time spent in this
2840 * ioctl to something sane.
2842 if (len > BTRFS_MAX_DEDUPE_LEN)
2843 len = BTRFS_MAX_DEDUPE_LEN;
2845 if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) {
2847 * Btrfs does not support blocksize < page_size. As a
2848 * result, btrfs_cmp_data() won't correctly handle
2849 * this situation without an update.
2856 if (S_ISDIR(src->i_mode))
2860 if (!S_ISREG(src->i_mode))
2863 /* pre-format output fields to sane values */
2864 for (i = 0; i < same->dest_count; i++) {
2865 same->info[i].bytes_deduped = 0ULL;
2866 same->info[i].status = 0;
2870 for (i = 0; i < same->dest_count; i++) {
2871 info = &same->info[i];
2873 dst_file = fget(info->fd);
2875 info->status = -EBADF;
2879 if (!(is_admin || (dst_file->f_mode & FMODE_WRITE))) {
2880 info->status = -EINVAL;
2884 info->status = -EXDEV;
2885 if (file->f_path.mnt != dst_file->f_path.mnt)
2888 dst = dst_file->f_dentry->d_inode;
2889 if (src->i_sb != dst->i_sb)
2892 if (S_ISDIR(dst->i_mode)) {
2893 info->status = -EISDIR;
2897 if (!S_ISREG(dst->i_mode)) {
2898 info->status = -EACCES;
2902 info->status = btrfs_extent_same(src, off, len, dst,
2903 info->logical_offset);
2904 if (info->status == 0)
2905 info->bytes_deduped += len;
2912 ret = copy_to_user(argp, same, size);
2917 mnt_drop_write_file(file);
2922 * btrfs_clone() - clone a range from inode file to another
2924 * @src: Inode to clone from
2925 * @inode: Inode to clone to
2926 * @off: Offset within source to start clone from
2927 * @olen: Original length, passed by user, of range to clone
2928 * @olen_aligned: Block-aligned value of olen, extent_same uses
2929 * identical values here
2930 * @destoff: Offset within @inode to start clone
2932 static int btrfs_clone(struct inode *src, struct inode *inode,
2933 u64 off, u64 olen, u64 olen_aligned, u64 destoff)
2935 struct btrfs_root *root = BTRFS_I(inode)->root;
2936 struct btrfs_path *path = NULL;
2937 struct extent_buffer *leaf;
2938 struct btrfs_trans_handle *trans;
2940 struct btrfs_key key;
2944 u64 len = olen_aligned;
2947 buf = vmalloc(btrfs_level_size(root, 0));
2951 path = btrfs_alloc_path();
2959 key.objectid = btrfs_ino(src);
2960 key.type = BTRFS_EXTENT_DATA_KEY;
2965 * note the key will change type as we walk through the
2968 path->leave_spinning = 1;
2969 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
2974 nritems = btrfs_header_nritems(path->nodes[0]);
2976 if (path->slots[0] >= nritems) {
2977 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
2982 nritems = btrfs_header_nritems(path->nodes[0]);
2984 leaf = path->nodes[0];
2985 slot = path->slots[0];
2987 btrfs_item_key_to_cpu(leaf, &key, slot);
2988 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
2989 key.objectid != btrfs_ino(src))
2992 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2993 struct btrfs_file_extent_item *extent;
2996 struct btrfs_key new_key;
2997 u64 disko = 0, diskl = 0;
2998 u64 datao = 0, datal = 0;
3002 extent = btrfs_item_ptr(leaf, slot,
3003 struct btrfs_file_extent_item);
3004 comp = btrfs_file_extent_compression(leaf, extent);
3005 type = btrfs_file_extent_type(leaf, extent);
3006 if (type == BTRFS_FILE_EXTENT_REG ||
3007 type == BTRFS_FILE_EXTENT_PREALLOC) {
3008 disko = btrfs_file_extent_disk_bytenr(leaf,
3010 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3012 datao = btrfs_file_extent_offset(leaf, extent);
3013 datal = btrfs_file_extent_num_bytes(leaf,
3015 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3016 /* take upper bound, may be compressed */
3017 datal = btrfs_file_extent_ram_bytes(leaf,
3021 if (key.offset + datal <= off ||
3022 key.offset >= off + len - 1) {
3027 size = btrfs_item_size_nr(leaf, slot);
3028 read_extent_buffer(leaf, buf,
3029 btrfs_item_ptr_offset(leaf, slot),
3032 btrfs_release_path(path);
3033 path->leave_spinning = 0;
3035 memcpy(&new_key, &key, sizeof(new_key));
3036 new_key.objectid = btrfs_ino(inode);
3037 if (off <= key.offset)
3038 new_key.offset = key.offset + destoff - off;
3040 new_key.offset = destoff;
3043 * 1 - adjusting old extent (we may have to split it)
3044 * 1 - add new extent
3047 trans = btrfs_start_transaction(root, 3);
3048 if (IS_ERR(trans)) {
3049 ret = PTR_ERR(trans);
3053 if (type == BTRFS_FILE_EXTENT_REG ||
3054 type == BTRFS_FILE_EXTENT_PREALLOC) {
3056 * a | --- range to clone ---| b
3057 * | ------------- extent ------------- |
3060 /* substract range b */
3061 if (key.offset + datal > off + len)
3062 datal = off + len - key.offset;
3064 /* substract range a */
3065 if (off > key.offset) {
3066 datao += off - key.offset;
3067 datal -= off - key.offset;
3070 ret = btrfs_drop_extents(trans, root, inode,
3072 new_key.offset + datal,
3075 btrfs_abort_transaction(trans, root,
3077 btrfs_end_transaction(trans, root);
3081 ret = btrfs_insert_empty_item(trans, root, path,
3084 btrfs_abort_transaction(trans, root,
3086 btrfs_end_transaction(trans, root);
3090 leaf = path->nodes[0];
3091 slot = path->slots[0];
3092 write_extent_buffer(leaf, buf,
3093 btrfs_item_ptr_offset(leaf, slot),
3096 extent = btrfs_item_ptr(leaf, slot,
3097 struct btrfs_file_extent_item);
3099 /* disko == 0 means it's a hole */
3103 btrfs_set_file_extent_offset(leaf, extent,
3105 btrfs_set_file_extent_num_bytes(leaf, extent,
3108 inode_add_bytes(inode, datal);
3109 ret = btrfs_inc_extent_ref(trans, root,
3111 root->root_key.objectid,
3113 new_key.offset - datao,
3116 btrfs_abort_transaction(trans,
3119 btrfs_end_transaction(trans,
3125 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3128 if (off > key.offset) {
3129 skip = off - key.offset;
3130 new_key.offset += skip;
3133 if (key.offset + datal > off + len)
3134 trim = key.offset + datal - (off + len);
3136 if (comp && (skip || trim)) {
3138 btrfs_end_transaction(trans, root);
3141 size -= skip + trim;
3142 datal -= skip + trim;
3144 ret = btrfs_drop_extents(trans, root, inode,
3146 new_key.offset + datal,
3149 btrfs_abort_transaction(trans, root,
3151 btrfs_end_transaction(trans, root);
3155 ret = btrfs_insert_empty_item(trans, root, path,
3158 btrfs_abort_transaction(trans, root,
3160 btrfs_end_transaction(trans, root);
3166 btrfs_file_extent_calc_inline_size(0);
3167 memmove(buf+start, buf+start+skip,
3171 leaf = path->nodes[0];
3172 slot = path->slots[0];
3173 write_extent_buffer(leaf, buf,
3174 btrfs_item_ptr_offset(leaf, slot),
3176 inode_add_bytes(inode, datal);
3179 btrfs_mark_buffer_dirty(leaf);
3180 btrfs_release_path(path);
3182 inode_inc_iversion(inode);
3183 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3186 * we round up to the block size at eof when
3187 * determining which extents to clone above,
3188 * but shouldn't round up the file size
3190 endoff = new_key.offset + datal;
3191 if (endoff > destoff+olen)
3192 endoff = destoff+olen;
3193 if (endoff > inode->i_size)
3194 btrfs_i_size_write(inode, endoff);
3196 ret = btrfs_update_inode(trans, root, inode);
3198 btrfs_abort_transaction(trans, root, ret);
3199 btrfs_end_transaction(trans, root);
3202 ret = btrfs_end_transaction(trans, root);
3204 btrfs_release_path(path);
3210 btrfs_release_path(path);
3211 btrfs_free_path(path);
3216 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
3217 u64 off, u64 olen, u64 destoff)
3219 struct inode *inode = file_inode(file);
3220 struct btrfs_root *root = BTRFS_I(inode)->root;
3225 u64 bs = root->fs_info->sb->s_blocksize;
3230 * - split compressed inline extents. annoying: we need to
3231 * decompress into destination's address_space (the file offset
3232 * may change, so source mapping won't do), then recompress (or
3233 * otherwise reinsert) a subrange.
3234 * - allow ranges within the same file to be cloned (provided
3235 * they don't overlap)?
3238 /* the destination must be opened for writing */
3239 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
3242 if (btrfs_root_readonly(root))
3245 ret = mnt_want_write_file(file);
3249 src_file = fdget(srcfd);
3250 if (!src_file.file) {
3252 goto out_drop_write;
3256 if (src_file.file->f_path.mnt != file->f_path.mnt)
3259 src = file_inode(src_file.file);
3265 /* the src must be open for reading */
3266 if (!(src_file.file->f_mode & FMODE_READ))
3269 /* don't make the dst file partly checksummed */
3270 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3271 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3275 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3279 if (src->i_sb != inode->i_sb)
3284 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
3285 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
3287 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
3288 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
3291 mutex_lock(&src->i_mutex);
3294 /* determine range to clone */
3296 if (off + len > src->i_size || off + len < off)
3299 olen = len = src->i_size - off;
3300 /* if we extend to eof, continue to block boundary */
3301 if (off + len == src->i_size)
3302 len = ALIGN(src->i_size, bs) - off;
3304 /* verify the end result is block aligned */
3305 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3306 !IS_ALIGNED(destoff, bs))
3309 /* verify if ranges are overlapped within the same file */
3311 if (destoff + len > off && destoff < off + len)
3315 if (destoff > inode->i_size) {
3316 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3321 /* truncate page cache pages from target inode range */
3322 truncate_inode_pages_range(&inode->i_data, destoff,
3323 PAGE_CACHE_ALIGN(destoff + len) - 1);
3325 lock_extent_range(src, off, len);
3327 ret = btrfs_clone(src, inode, off, olen, len, destoff);
3329 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
3333 mutex_unlock(&src->i_mutex);
3334 mutex_unlock(&inode->i_mutex);
3336 mutex_unlock(&inode->i_mutex);
3337 mutex_unlock(&src->i_mutex);
3340 mutex_unlock(&src->i_mutex);
3345 mnt_drop_write_file(file);
3349 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
3351 struct btrfs_ioctl_clone_range_args args;
3353 if (copy_from_user(&args, argp, sizeof(args)))
3355 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
3356 args.src_length, args.dest_offset);
3360 * there are many ways the trans_start and trans_end ioctls can lead
3361 * to deadlocks. They should only be used by applications that
3362 * basically own the machine, and have a very in depth understanding
3363 * of all the possible deadlocks and enospc problems.
3365 static long btrfs_ioctl_trans_start(struct file *file)
3367 struct inode *inode = file_inode(file);
3368 struct btrfs_root *root = BTRFS_I(inode)->root;
3369 struct btrfs_trans_handle *trans;
3373 if (!capable(CAP_SYS_ADMIN))
3377 if (file->private_data)
3381 if (btrfs_root_readonly(root))
3384 ret = mnt_want_write_file(file);
3388 atomic_inc(&root->fs_info->open_ioctl_trans);
3391 trans = btrfs_start_ioctl_transaction(root);
3395 file->private_data = trans;
3399 atomic_dec(&root->fs_info->open_ioctl_trans);
3400 mnt_drop_write_file(file);
3405 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3407 struct inode *inode = file_inode(file);
3408 struct btrfs_root *root = BTRFS_I(inode)->root;
3409 struct btrfs_root *new_root;
3410 struct btrfs_dir_item *di;
3411 struct btrfs_trans_handle *trans;
3412 struct btrfs_path *path;
3413 struct btrfs_key location;
3414 struct btrfs_disk_key disk_key;
3419 if (!capable(CAP_SYS_ADMIN))
3422 ret = mnt_want_write_file(file);
3426 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3432 objectid = BTRFS_FS_TREE_OBJECTID;
3434 location.objectid = objectid;
3435 location.type = BTRFS_ROOT_ITEM_KEY;
3436 location.offset = (u64)-1;
3438 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
3439 if (IS_ERR(new_root)) {
3440 ret = PTR_ERR(new_root);
3444 path = btrfs_alloc_path();
3449 path->leave_spinning = 1;
3451 trans = btrfs_start_transaction(root, 1);
3452 if (IS_ERR(trans)) {
3453 btrfs_free_path(path);
3454 ret = PTR_ERR(trans);
3458 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
3459 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
3460 dir_id, "default", 7, 1);
3461 if (IS_ERR_OR_NULL(di)) {
3462 btrfs_free_path(path);
3463 btrfs_end_transaction(trans, root);
3464 btrfs_err(new_root->fs_info, "Umm, you don't have the default dir"
3465 "item, this isn't going to work");
3470 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3471 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3472 btrfs_mark_buffer_dirty(path->nodes[0]);
3473 btrfs_free_path(path);
3475 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
3476 btrfs_end_transaction(trans, root);
3478 mnt_drop_write_file(file);
3482 void btrfs_get_block_group_info(struct list_head *groups_list,
3483 struct btrfs_ioctl_space_info *space)
3485 struct btrfs_block_group_cache *block_group;
3487 space->total_bytes = 0;
3488 space->used_bytes = 0;
3490 list_for_each_entry(block_group, groups_list, list) {
3491 space->flags = block_group->flags;
3492 space->total_bytes += block_group->key.offset;
3493 space->used_bytes +=
3494 btrfs_block_group_used(&block_group->item);
3498 static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
3500 struct btrfs_ioctl_space_args space_args;
3501 struct btrfs_ioctl_space_info space;
3502 struct btrfs_ioctl_space_info *dest;
3503 struct btrfs_ioctl_space_info *dest_orig;
3504 struct btrfs_ioctl_space_info __user *user_dest;
3505 struct btrfs_space_info *info;
3506 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3507 BTRFS_BLOCK_GROUP_SYSTEM,
3508 BTRFS_BLOCK_GROUP_METADATA,
3509 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3516 if (copy_from_user(&space_args,
3517 (struct btrfs_ioctl_space_args __user *)arg,
3518 sizeof(space_args)))
3521 for (i = 0; i < num_types; i++) {
3522 struct btrfs_space_info *tmp;
3526 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3528 if (tmp->flags == types[i]) {
3538 down_read(&info->groups_sem);
3539 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3540 if (!list_empty(&info->block_groups[c]))
3543 up_read(&info->groups_sem);
3546 /* space_slots == 0 means they are asking for a count */
3547 if (space_args.space_slots == 0) {
3548 space_args.total_spaces = slot_count;
3552 slot_count = min_t(u64, space_args.space_slots, slot_count);
3554 alloc_size = sizeof(*dest) * slot_count;
3556 /* we generally have at most 6 or so space infos, one for each raid
3557 * level. So, a whole page should be more than enough for everyone
3559 if (alloc_size > PAGE_CACHE_SIZE)
3562 space_args.total_spaces = 0;
3563 dest = kmalloc(alloc_size, GFP_NOFS);
3568 /* now we have a buffer to copy into */
3569 for (i = 0; i < num_types; i++) {
3570 struct btrfs_space_info *tmp;
3577 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3579 if (tmp->flags == types[i]) {
3588 down_read(&info->groups_sem);
3589 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3590 if (!list_empty(&info->block_groups[c])) {
3591 btrfs_get_block_group_info(
3592 &info->block_groups[c], &space);
3593 memcpy(dest, &space, sizeof(space));
3595 space_args.total_spaces++;
3601 up_read(&info->groups_sem);
3604 user_dest = (struct btrfs_ioctl_space_info __user *)
3605 (arg + sizeof(struct btrfs_ioctl_space_args));
3607 if (copy_to_user(user_dest, dest_orig, alloc_size))
3612 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3619 * there are many ways the trans_start and trans_end ioctls can lead
3620 * to deadlocks. They should only be used by applications that
3621 * basically own the machine, and have a very in depth understanding
3622 * of all the possible deadlocks and enospc problems.
3624 long btrfs_ioctl_trans_end(struct file *file)
3626 struct inode *inode = file_inode(file);
3627 struct btrfs_root *root = BTRFS_I(inode)->root;
3628 struct btrfs_trans_handle *trans;
3630 trans = file->private_data;
3633 file->private_data = NULL;
3635 btrfs_end_transaction(trans, root);
3637 atomic_dec(&root->fs_info->open_ioctl_trans);
3639 mnt_drop_write_file(file);
3643 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3646 struct btrfs_trans_handle *trans;
3650 trans = btrfs_attach_transaction_barrier(root);
3651 if (IS_ERR(trans)) {
3652 if (PTR_ERR(trans) != -ENOENT)
3653 return PTR_ERR(trans);
3655 /* No running transaction, don't bother */
3656 transid = root->fs_info->last_trans_committed;
3659 transid = trans->transid;
3660 ret = btrfs_commit_transaction_async(trans, root, 0);
3662 btrfs_end_transaction(trans, root);
3667 if (copy_to_user(argp, &transid, sizeof(transid)))
3672 static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
3678 if (copy_from_user(&transid, argp, sizeof(transid)))
3681 transid = 0; /* current trans */
3683 return btrfs_wait_for_commit(root, transid);
3686 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
3688 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3689 struct btrfs_ioctl_scrub_args *sa;
3692 if (!capable(CAP_SYS_ADMIN))
3695 sa = memdup_user(arg, sizeof(*sa));
3699 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3700 ret = mnt_want_write_file(file);
3705 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
3706 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3709 if (copy_to_user(arg, sa, sizeof(*sa)))
3712 if (!(sa->flags & BTRFS_SCRUB_READONLY))
3713 mnt_drop_write_file(file);
3719 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3721 if (!capable(CAP_SYS_ADMIN))
3724 return btrfs_scrub_cancel(root->fs_info);
3727 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3730 struct btrfs_ioctl_scrub_args *sa;
3733 if (!capable(CAP_SYS_ADMIN))
3736 sa = memdup_user(arg, sizeof(*sa));
3740 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3742 if (copy_to_user(arg, sa, sizeof(*sa)))
3749 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
3752 struct btrfs_ioctl_get_dev_stats *sa;
3755 sa = memdup_user(arg, sizeof(*sa));
3759 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3764 ret = btrfs_get_dev_stats(root, sa);
3766 if (copy_to_user(arg, sa, sizeof(*sa)))
3773 static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
3775 struct btrfs_ioctl_dev_replace_args *p;
3778 if (!capable(CAP_SYS_ADMIN))
3781 p = memdup_user(arg, sizeof(*p));
3786 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
3787 if (root->fs_info->sb->s_flags & MS_RDONLY) {
3792 &root->fs_info->mutually_exclusive_operation_running,
3794 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
3796 ret = btrfs_dev_replace_start(root, p);
3798 &root->fs_info->mutually_exclusive_operation_running,
3802 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3803 btrfs_dev_replace_status(root->fs_info, p);
3806 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3807 ret = btrfs_dev_replace_cancel(root->fs_info, p);
3814 if (copy_to_user(arg, p, sizeof(*p)))
3821 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3827 struct btrfs_ioctl_ino_path_args *ipa = NULL;
3828 struct inode_fs_paths *ipath = NULL;
3829 struct btrfs_path *path;
3831 if (!capable(CAP_DAC_READ_SEARCH))
3834 path = btrfs_alloc_path();
3840 ipa = memdup_user(arg, sizeof(*ipa));
3847 size = min_t(u32, ipa->size, 4096);
3848 ipath = init_ipath(size, root, path);
3849 if (IS_ERR(ipath)) {
3850 ret = PTR_ERR(ipath);
3855 ret = paths_from_inode(ipa->inum, ipath);
3859 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3860 rel_ptr = ipath->fspath->val[i] -
3861 (u64)(unsigned long)ipath->fspath->val;
3862 ipath->fspath->val[i] = rel_ptr;
3865 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3866 (void *)(unsigned long)ipath->fspath, size);
3873 btrfs_free_path(path);
3880 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3882 struct btrfs_data_container *inodes = ctx;
3883 const size_t c = 3 * sizeof(u64);
3885 if (inodes->bytes_left >= c) {
3886 inodes->bytes_left -= c;
3887 inodes->val[inodes->elem_cnt] = inum;
3888 inodes->val[inodes->elem_cnt + 1] = offset;
3889 inodes->val[inodes->elem_cnt + 2] = root;
3890 inodes->elem_cnt += 3;
3892 inodes->bytes_missing += c - inodes->bytes_left;
3893 inodes->bytes_left = 0;
3894 inodes->elem_missed += 3;
3900 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3905 struct btrfs_ioctl_logical_ino_args *loi;
3906 struct btrfs_data_container *inodes = NULL;
3907 struct btrfs_path *path = NULL;
3909 if (!capable(CAP_SYS_ADMIN))
3912 loi = memdup_user(arg, sizeof(*loi));
3919 path = btrfs_alloc_path();
3925 size = min_t(u32, loi->size, 64 * 1024);
3926 inodes = init_data_container(size);
3927 if (IS_ERR(inodes)) {
3928 ret = PTR_ERR(inodes);
3933 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
3934 build_ino_list, inodes);
3940 ret = copy_to_user((void *)(unsigned long)loi->inodes,
3941 (void *)(unsigned long)inodes, size);
3946 btrfs_free_path(path);
3953 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
3954 struct btrfs_ioctl_balance_args *bargs)
3956 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3958 bargs->flags = bctl->flags;
3960 if (atomic_read(&fs_info->balance_running))
3961 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3962 if (atomic_read(&fs_info->balance_pause_req))
3963 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3964 if (atomic_read(&fs_info->balance_cancel_req))
3965 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3967 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3968 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3969 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3972 spin_lock(&fs_info->balance_lock);
3973 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3974 spin_unlock(&fs_info->balance_lock);
3976 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3980 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3982 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3983 struct btrfs_fs_info *fs_info = root->fs_info;
3984 struct btrfs_ioctl_balance_args *bargs;
3985 struct btrfs_balance_control *bctl;
3986 bool need_unlock; /* for mut. excl. ops lock */
3989 if (!capable(CAP_SYS_ADMIN))
3992 ret = mnt_want_write_file(file);
3997 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
3998 mutex_lock(&fs_info->volume_mutex);
3999 mutex_lock(&fs_info->balance_mutex);
4005 * mut. excl. ops lock is locked. Three possibilites:
4006 * (1) some other op is running
4007 * (2) balance is running
4008 * (3) balance is paused -- special case (think resume)
4010 mutex_lock(&fs_info->balance_mutex);
4011 if (fs_info->balance_ctl) {
4012 /* this is either (2) or (3) */
4013 if (!atomic_read(&fs_info->balance_running)) {
4014 mutex_unlock(&fs_info->balance_mutex);
4015 if (!mutex_trylock(&fs_info->volume_mutex))
4017 mutex_lock(&fs_info->balance_mutex);
4019 if (fs_info->balance_ctl &&
4020 !atomic_read(&fs_info->balance_running)) {
4022 need_unlock = false;
4026 mutex_unlock(&fs_info->balance_mutex);
4027 mutex_unlock(&fs_info->volume_mutex);
4031 mutex_unlock(&fs_info->balance_mutex);
4037 mutex_unlock(&fs_info->balance_mutex);
4038 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4043 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
4046 bargs = memdup_user(arg, sizeof(*bargs));
4047 if (IS_ERR(bargs)) {
4048 ret = PTR_ERR(bargs);
4052 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4053 if (!fs_info->balance_ctl) {
4058 bctl = fs_info->balance_ctl;
4059 spin_lock(&fs_info->balance_lock);
4060 bctl->flags |= BTRFS_BALANCE_RESUME;
4061 spin_unlock(&fs_info->balance_lock);
4069 if (fs_info->balance_ctl) {
4074 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
4080 bctl->fs_info = fs_info;
4082 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4083 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4084 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4086 bctl->flags = bargs->flags;
4088 /* balance everything - no filters */
4089 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4094 * Ownership of bctl and mutually_exclusive_operation_running
4095 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4096 * or, if restriper was paused all the way until unmount, in
4097 * free_fs_info. mutually_exclusive_operation_running is
4098 * cleared in __cancel_balance.
4100 need_unlock = false;
4102 ret = btrfs_balance(bctl, bargs);
4105 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4112 mutex_unlock(&fs_info->balance_mutex);
4113 mutex_unlock(&fs_info->volume_mutex);
4115 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4117 mnt_drop_write_file(file);
4121 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4123 if (!capable(CAP_SYS_ADMIN))
4127 case BTRFS_BALANCE_CTL_PAUSE:
4128 return btrfs_pause_balance(root->fs_info);
4129 case BTRFS_BALANCE_CTL_CANCEL:
4130 return btrfs_cancel_balance(root->fs_info);
4136 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4139 struct btrfs_fs_info *fs_info = root->fs_info;
4140 struct btrfs_ioctl_balance_args *bargs;
4143 if (!capable(CAP_SYS_ADMIN))
4146 mutex_lock(&fs_info->balance_mutex);
4147 if (!fs_info->balance_ctl) {
4152 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
4158 update_ioctl_balance_args(fs_info, 1, bargs);
4160 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4165 mutex_unlock(&fs_info->balance_mutex);
4169 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4171 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4172 struct btrfs_ioctl_quota_ctl_args *sa;
4173 struct btrfs_trans_handle *trans = NULL;
4177 if (!capable(CAP_SYS_ADMIN))
4180 ret = mnt_want_write_file(file);
4184 sa = memdup_user(arg, sizeof(*sa));
4190 down_write(&root->fs_info->subvol_sem);
4191 trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4192 if (IS_ERR(trans)) {
4193 ret = PTR_ERR(trans);
4198 case BTRFS_QUOTA_CTL_ENABLE:
4199 ret = btrfs_quota_enable(trans, root->fs_info);
4201 case BTRFS_QUOTA_CTL_DISABLE:
4202 ret = btrfs_quota_disable(trans, root->fs_info);
4209 err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4214 up_write(&root->fs_info->subvol_sem);
4216 mnt_drop_write_file(file);
4220 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4222 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4223 struct btrfs_ioctl_qgroup_assign_args *sa;
4224 struct btrfs_trans_handle *trans;
4228 if (!capable(CAP_SYS_ADMIN))
4231 ret = mnt_want_write_file(file);
4235 sa = memdup_user(arg, sizeof(*sa));
4241 trans = btrfs_join_transaction(root);
4242 if (IS_ERR(trans)) {
4243 ret = PTR_ERR(trans);
4247 /* FIXME: check if the IDs really exist */
4249 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4252 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4256 err = btrfs_end_transaction(trans, root);
4263 mnt_drop_write_file(file);
4267 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4269 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4270 struct btrfs_ioctl_qgroup_create_args *sa;
4271 struct btrfs_trans_handle *trans;
4275 if (!capable(CAP_SYS_ADMIN))
4278 ret = mnt_want_write_file(file);
4282 sa = memdup_user(arg, sizeof(*sa));
4288 if (!sa->qgroupid) {
4293 trans = btrfs_join_transaction(root);
4294 if (IS_ERR(trans)) {
4295 ret = PTR_ERR(trans);
4299 /* FIXME: check if the IDs really exist */
4301 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
4304 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4307 err = btrfs_end_transaction(trans, root);
4314 mnt_drop_write_file(file);
4318 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4320 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4321 struct btrfs_ioctl_qgroup_limit_args *sa;
4322 struct btrfs_trans_handle *trans;
4327 if (!capable(CAP_SYS_ADMIN))
4330 ret = mnt_want_write_file(file);
4334 sa = memdup_user(arg, sizeof(*sa));
4340 trans = btrfs_join_transaction(root);
4341 if (IS_ERR(trans)) {
4342 ret = PTR_ERR(trans);
4346 qgroupid = sa->qgroupid;
4348 /* take the current subvol as qgroup */
4349 qgroupid = root->root_key.objectid;
4352 /* FIXME: check if the IDs really exist */
4353 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
4355 err = btrfs_end_transaction(trans, root);
4362 mnt_drop_write_file(file);
4366 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4368 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4369 struct btrfs_ioctl_quota_rescan_args *qsa;
4372 if (!capable(CAP_SYS_ADMIN))
4375 ret = mnt_want_write_file(file);
4379 qsa = memdup_user(arg, sizeof(*qsa));
4390 ret = btrfs_qgroup_rescan(root->fs_info);
4395 mnt_drop_write_file(file);
4399 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
4401 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4402 struct btrfs_ioctl_quota_rescan_args *qsa;
4405 if (!capable(CAP_SYS_ADMIN))
4408 qsa = kzalloc(sizeof(*qsa), GFP_NOFS);
4412 if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4414 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
4417 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4424 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
4426 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4428 if (!capable(CAP_SYS_ADMIN))
4431 return btrfs_qgroup_wait_for_completion(root->fs_info);
4434 static long _btrfs_ioctl_set_received_subvol(struct file *file,
4435 struct btrfs_ioctl_received_subvol_args *sa)
4437 struct inode *inode = file_inode(file);
4438 struct btrfs_root *root = BTRFS_I(inode)->root;
4439 struct btrfs_root_item *root_item = &root->root_item;
4440 struct btrfs_trans_handle *trans;
4441 struct timespec ct = CURRENT_TIME;
4443 int received_uuid_changed;
4445 if (!inode_owner_or_capable(inode))
4448 ret = mnt_want_write_file(file);
4452 down_write(&root->fs_info->subvol_sem);
4454 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
4459 if (btrfs_root_readonly(root)) {
4466 * 2 - uuid items (received uuid + subvol uuid)
4468 trans = btrfs_start_transaction(root, 3);
4469 if (IS_ERR(trans)) {
4470 ret = PTR_ERR(trans);
4475 sa->rtransid = trans->transid;
4476 sa->rtime.sec = ct.tv_sec;
4477 sa->rtime.nsec = ct.tv_nsec;
4479 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4481 if (received_uuid_changed &&
4482 !btrfs_is_empty_uuid(root_item->received_uuid))
4483 btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
4484 root_item->received_uuid,
4485 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4486 root->root_key.objectid);
4487 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4488 btrfs_set_root_stransid(root_item, sa->stransid);
4489 btrfs_set_root_rtransid(root_item, sa->rtransid);
4490 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4491 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4492 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4493 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4495 ret = btrfs_update_root(trans, root->fs_info->tree_root,
4496 &root->root_key, &root->root_item);
4498 btrfs_end_transaction(trans, root);
4501 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4502 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
4504 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4505 root->root_key.objectid);
4506 if (ret < 0 && ret != -EEXIST) {
4507 btrfs_abort_transaction(trans, root, ret);
4511 ret = btrfs_commit_transaction(trans, root);
4513 btrfs_abort_transaction(trans, root, ret);
4518 up_write(&root->fs_info->subvol_sem);
4519 mnt_drop_write_file(file);
4524 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4527 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4528 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4531 args32 = memdup_user(arg, sizeof(*args32));
4532 if (IS_ERR(args32)) {
4533 ret = PTR_ERR(args32);
4538 args64 = kmalloc(sizeof(*args64), GFP_NOFS);
4539 if (IS_ERR(args64)) {
4540 ret = PTR_ERR(args64);
4545 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4546 args64->stransid = args32->stransid;
4547 args64->rtransid = args32->rtransid;
4548 args64->stime.sec = args32->stime.sec;
4549 args64->stime.nsec = args32->stime.nsec;
4550 args64->rtime.sec = args32->rtime.sec;
4551 args64->rtime.nsec = args32->rtime.nsec;
4552 args64->flags = args32->flags;
4554 ret = _btrfs_ioctl_set_received_subvol(file, args64);
4558 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4559 args32->stransid = args64->stransid;
4560 args32->rtransid = args64->rtransid;
4561 args32->stime.sec = args64->stime.sec;
4562 args32->stime.nsec = args64->stime.nsec;
4563 args32->rtime.sec = args64->rtime.sec;
4564 args32->rtime.nsec = args64->rtime.nsec;
4565 args32->flags = args64->flags;
4567 ret = copy_to_user(arg, args32, sizeof(*args32));
4578 static long btrfs_ioctl_set_received_subvol(struct file *file,
4581 struct btrfs_ioctl_received_subvol_args *sa = NULL;
4584 sa = memdup_user(arg, sizeof(*sa));
4591 ret = _btrfs_ioctl_set_received_subvol(file, sa);
4596 ret = copy_to_user(arg, sa, sizeof(*sa));
4605 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
4607 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4610 char label[BTRFS_LABEL_SIZE];
4612 spin_lock(&root->fs_info->super_lock);
4613 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4614 spin_unlock(&root->fs_info->super_lock);
4616 len = strnlen(label, BTRFS_LABEL_SIZE);
4618 if (len == BTRFS_LABEL_SIZE) {
4619 btrfs_warn(root->fs_info,
4620 "label is too long, return the first %zu bytes", --len);
4623 ret = copy_to_user(arg, label, len);
4625 return ret ? -EFAULT : 0;
4628 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4630 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4631 struct btrfs_super_block *super_block = root->fs_info->super_copy;
4632 struct btrfs_trans_handle *trans;
4633 char label[BTRFS_LABEL_SIZE];
4636 if (!capable(CAP_SYS_ADMIN))
4639 if (copy_from_user(label, arg, sizeof(label)))
4642 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
4643 btrfs_err(root->fs_info, "unable to set label with more than %d bytes",
4644 BTRFS_LABEL_SIZE - 1);
4648 ret = mnt_want_write_file(file);
4652 trans = btrfs_start_transaction(root, 0);
4653 if (IS_ERR(trans)) {
4654 ret = PTR_ERR(trans);
4658 spin_lock(&root->fs_info->super_lock);
4659 strcpy(super_block->label, label);
4660 spin_unlock(&root->fs_info->super_lock);
4661 ret = btrfs_commit_transaction(trans, root);
4664 mnt_drop_write_file(file);
4668 #define INIT_FEATURE_FLAGS(suffix) \
4669 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
4670 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
4671 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
4673 static int btrfs_ioctl_get_supported_features(struct file *file,
4676 static struct btrfs_ioctl_feature_flags features[3] = {
4677 INIT_FEATURE_FLAGS(SUPP),
4678 INIT_FEATURE_FLAGS(SAFE_SET),
4679 INIT_FEATURE_FLAGS(SAFE_CLEAR)
4682 if (copy_to_user(arg, &features, sizeof(features)))
4688 static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
4690 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4691 struct btrfs_super_block *super_block = root->fs_info->super_copy;
4692 struct btrfs_ioctl_feature_flags features;
4694 features.compat_flags = btrfs_super_compat_flags(super_block);
4695 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
4696 features.incompat_flags = btrfs_super_incompat_flags(super_block);
4698 if (copy_to_user(arg, &features, sizeof(features)))
4704 static int check_feature_bits(struct btrfs_root *root,
4705 enum btrfs_feature_set set,
4706 u64 change_mask, u64 flags, u64 supported_flags,
4707 u64 safe_set, u64 safe_clear)
4709 const char *type = btrfs_feature_set_names[set];
4711 u64 disallowed, unsupported;
4712 u64 set_mask = flags & change_mask;
4713 u64 clear_mask = ~flags & change_mask;
4715 unsupported = set_mask & ~supported_flags;
4717 names = btrfs_printable_features(set, unsupported);
4719 btrfs_warn(root->fs_info,
4720 "this kernel does not support the %s feature bit%s",
4721 names, strchr(names, ',') ? "s" : "");
4724 btrfs_warn(root->fs_info,
4725 "this kernel does not support %s bits 0x%llx",
4730 disallowed = set_mask & ~safe_set;
4732 names = btrfs_printable_features(set, disallowed);
4734 btrfs_warn(root->fs_info,
4735 "can't set the %s feature bit%s while mounted",
4736 names, strchr(names, ',') ? "s" : "");
4739 btrfs_warn(root->fs_info,
4740 "can't set %s bits 0x%llx while mounted",
4745 disallowed = clear_mask & ~safe_clear;
4747 names = btrfs_printable_features(set, disallowed);
4749 btrfs_warn(root->fs_info,
4750 "can't clear the %s feature bit%s while mounted",
4751 names, strchr(names, ',') ? "s" : "");
4754 btrfs_warn(root->fs_info,
4755 "can't clear %s bits 0x%llx while mounted",
4763 #define check_feature(root, change_mask, flags, mask_base) \
4764 check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
4765 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
4766 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
4767 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
4769 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
4771 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4772 struct btrfs_super_block *super_block = root->fs_info->super_copy;
4773 struct btrfs_ioctl_feature_flags flags[2];
4774 struct btrfs_trans_handle *trans;
4778 if (!capable(CAP_SYS_ADMIN))
4781 if (copy_from_user(flags, arg, sizeof(flags)))
4785 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
4786 !flags[0].incompat_flags)
4789 ret = check_feature(root, flags[0].compat_flags,
4790 flags[1].compat_flags, COMPAT);
4794 ret = check_feature(root, flags[0].compat_ro_flags,
4795 flags[1].compat_ro_flags, COMPAT_RO);
4799 ret = check_feature(root, flags[0].incompat_flags,
4800 flags[1].incompat_flags, INCOMPAT);
4804 trans = btrfs_start_transaction(root, 0);
4806 return PTR_ERR(trans);
4808 spin_lock(&root->fs_info->super_lock);
4809 newflags = btrfs_super_compat_flags(super_block);
4810 newflags |= flags[0].compat_flags & flags[1].compat_flags;
4811 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
4812 btrfs_set_super_compat_flags(super_block, newflags);
4814 newflags = btrfs_super_compat_ro_flags(super_block);
4815 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
4816 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
4817 btrfs_set_super_compat_ro_flags(super_block, newflags);
4819 newflags = btrfs_super_incompat_flags(super_block);
4820 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
4821 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
4822 btrfs_set_super_incompat_flags(super_block, newflags);
4823 spin_unlock(&root->fs_info->super_lock);
4825 return btrfs_commit_transaction(trans, root);
4828 long btrfs_ioctl(struct file *file, unsigned int
4829 cmd, unsigned long arg)
4831 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4832 void __user *argp = (void __user *)arg;
4835 case FS_IOC_GETFLAGS:
4836 return btrfs_ioctl_getflags(file, argp);
4837 case FS_IOC_SETFLAGS:
4838 return btrfs_ioctl_setflags(file, argp);
4839 case FS_IOC_GETVERSION:
4840 return btrfs_ioctl_getversion(file, argp);
4842 return btrfs_ioctl_fitrim(file, argp);
4843 case BTRFS_IOC_SNAP_CREATE:
4844 return btrfs_ioctl_snap_create(file, argp, 0);
4845 case BTRFS_IOC_SNAP_CREATE_V2:
4846 return btrfs_ioctl_snap_create_v2(file, argp, 0);
4847 case BTRFS_IOC_SUBVOL_CREATE:
4848 return btrfs_ioctl_snap_create(file, argp, 1);
4849 case BTRFS_IOC_SUBVOL_CREATE_V2:
4850 return btrfs_ioctl_snap_create_v2(file, argp, 1);
4851 case BTRFS_IOC_SNAP_DESTROY:
4852 return btrfs_ioctl_snap_destroy(file, argp);
4853 case BTRFS_IOC_SUBVOL_GETFLAGS:
4854 return btrfs_ioctl_subvol_getflags(file, argp);
4855 case BTRFS_IOC_SUBVOL_SETFLAGS:
4856 return btrfs_ioctl_subvol_setflags(file, argp);
4857 case BTRFS_IOC_DEFAULT_SUBVOL:
4858 return btrfs_ioctl_default_subvol(file, argp);
4859 case BTRFS_IOC_DEFRAG:
4860 return btrfs_ioctl_defrag(file, NULL);
4861 case BTRFS_IOC_DEFRAG_RANGE:
4862 return btrfs_ioctl_defrag(file, argp);
4863 case BTRFS_IOC_RESIZE:
4864 return btrfs_ioctl_resize(file, argp);
4865 case BTRFS_IOC_ADD_DEV:
4866 return btrfs_ioctl_add_dev(root, argp);
4867 case BTRFS_IOC_RM_DEV:
4868 return btrfs_ioctl_rm_dev(file, argp);
4869 case BTRFS_IOC_FS_INFO:
4870 return btrfs_ioctl_fs_info(root, argp);
4871 case BTRFS_IOC_DEV_INFO:
4872 return btrfs_ioctl_dev_info(root, argp);
4873 case BTRFS_IOC_BALANCE:
4874 return btrfs_ioctl_balance(file, NULL);
4875 case BTRFS_IOC_CLONE:
4876 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
4877 case BTRFS_IOC_CLONE_RANGE:
4878 return btrfs_ioctl_clone_range(file, argp);
4879 case BTRFS_IOC_TRANS_START:
4880 return btrfs_ioctl_trans_start(file);
4881 case BTRFS_IOC_TRANS_END:
4882 return btrfs_ioctl_trans_end(file);
4883 case BTRFS_IOC_TREE_SEARCH:
4884 return btrfs_ioctl_tree_search(file, argp);
4885 case BTRFS_IOC_INO_LOOKUP:
4886 return btrfs_ioctl_ino_lookup(file, argp);
4887 case BTRFS_IOC_INO_PATHS:
4888 return btrfs_ioctl_ino_to_path(root, argp);
4889 case BTRFS_IOC_LOGICAL_INO:
4890 return btrfs_ioctl_logical_to_ino(root, argp);
4891 case BTRFS_IOC_SPACE_INFO:
4892 return btrfs_ioctl_space_info(root, argp);
4893 case BTRFS_IOC_SYNC: {
4896 ret = btrfs_start_delalloc_roots(root->fs_info, 0);
4899 ret = btrfs_sync_fs(file->f_dentry->d_sb, 1);
4902 case BTRFS_IOC_START_SYNC:
4903 return btrfs_ioctl_start_sync(root, argp);
4904 case BTRFS_IOC_WAIT_SYNC:
4905 return btrfs_ioctl_wait_sync(root, argp);
4906 case BTRFS_IOC_SCRUB:
4907 return btrfs_ioctl_scrub(file, argp);
4908 case BTRFS_IOC_SCRUB_CANCEL:
4909 return btrfs_ioctl_scrub_cancel(root, argp);
4910 case BTRFS_IOC_SCRUB_PROGRESS:
4911 return btrfs_ioctl_scrub_progress(root, argp);
4912 case BTRFS_IOC_BALANCE_V2:
4913 return btrfs_ioctl_balance(file, argp);
4914 case BTRFS_IOC_BALANCE_CTL:
4915 return btrfs_ioctl_balance_ctl(root, arg);
4916 case BTRFS_IOC_BALANCE_PROGRESS:
4917 return btrfs_ioctl_balance_progress(root, argp);
4918 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
4919 return btrfs_ioctl_set_received_subvol(file, argp);
4921 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
4922 return btrfs_ioctl_set_received_subvol_32(file, argp);
4924 case BTRFS_IOC_SEND:
4925 return btrfs_ioctl_send(file, argp);
4926 case BTRFS_IOC_GET_DEV_STATS:
4927 return btrfs_ioctl_get_dev_stats(root, argp);
4928 case BTRFS_IOC_QUOTA_CTL:
4929 return btrfs_ioctl_quota_ctl(file, argp);
4930 case BTRFS_IOC_QGROUP_ASSIGN:
4931 return btrfs_ioctl_qgroup_assign(file, argp);
4932 case BTRFS_IOC_QGROUP_CREATE:
4933 return btrfs_ioctl_qgroup_create(file, argp);
4934 case BTRFS_IOC_QGROUP_LIMIT:
4935 return btrfs_ioctl_qgroup_limit(file, argp);
4936 case BTRFS_IOC_QUOTA_RESCAN:
4937 return btrfs_ioctl_quota_rescan(file, argp);
4938 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
4939 return btrfs_ioctl_quota_rescan_status(file, argp);
4940 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
4941 return btrfs_ioctl_quota_rescan_wait(file, argp);
4942 case BTRFS_IOC_DEV_REPLACE:
4943 return btrfs_ioctl_dev_replace(root, argp);
4944 case BTRFS_IOC_GET_FSLABEL:
4945 return btrfs_ioctl_get_fslabel(file, argp);
4946 case BTRFS_IOC_SET_FSLABEL:
4947 return btrfs_ioctl_set_fslabel(file, argp);
4948 case BTRFS_IOC_FILE_EXTENT_SAME:
4949 return btrfs_ioctl_file_extent_same(file, argp);
4950 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
4951 return btrfs_ioctl_get_supported_features(file, argp);
4952 case BTRFS_IOC_GET_FEATURES:
4953 return btrfs_ioctl_get_features(file, argp);
4954 case BTRFS_IOC_SET_FEATURES:
4955 return btrfs_ioctl_set_features(file, argp);