2 * Copyright (C) 2007 Oracle. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
19 #include <linux/kernel.h>
20 #include <linux/bio.h>
21 #include <linux/buffer_head.h>
22 #include <linux/file.h>
24 #include <linux/fsnotify.h>
25 #include <linux/pagemap.h>
26 #include <linux/highmem.h>
27 #include <linux/time.h>
28 #include <linux/init.h>
29 #include <linux/string.h>
30 #include <linux/backing-dev.h>
31 #include <linux/mount.h>
32 #include <linux/mpage.h>
33 #include <linux/namei.h>
34 #include <linux/swap.h>
35 #include <linux/writeback.h>
36 #include <linux/statfs.h>
37 #include <linux/compat.h>
38 #include <linux/bit_spinlock.h>
39 #include <linux/security.h>
40 #include <linux/xattr.h>
41 #include <linux/vmalloc.h>
42 #include <linux/slab.h>
43 #include <linux/blkdev.h>
44 #include <linux/uuid.h>
45 #include <linux/btrfs.h>
46 #include <linux/uaccess.h>
49 #include "transaction.h"
50 #include "btrfs_inode.h"
51 #include "print-tree.h"
54 #include "inode-map.h"
56 #include "rcu-string.h"
58 #include "dev-replace.h"
64 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
65 * structures are incorrect, as the timespec structure from userspace
66 * is 4 bytes too small. We define these alternatives here to teach
67 * the kernel about the 32-bit struct packing.
69 struct btrfs_ioctl_timespec_32 {
72 } __attribute__ ((__packed__));
74 struct btrfs_ioctl_received_subvol_args_32 {
75 char uuid[BTRFS_UUID_SIZE]; /* in */
76 __u64 stransid; /* in */
77 __u64 rtransid; /* out */
78 struct btrfs_ioctl_timespec_32 stime; /* in */
79 struct btrfs_ioctl_timespec_32 rtime; /* out */
81 __u64 reserved[16]; /* in */
82 } __attribute__ ((__packed__));
84 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
85 struct btrfs_ioctl_received_subvol_args_32)
89 static int btrfs_clone(struct inode *src, struct inode *inode,
90 u64 off, u64 olen, u64 olen_aligned, u64 destoff);
92 /* Mask out flags that are inappropriate for the given type of inode. */
93 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
97 else if (S_ISREG(mode))
98 return flags & ~FS_DIRSYNC_FL;
100 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
104 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
106 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
108 unsigned int iflags = 0;
110 if (flags & BTRFS_INODE_SYNC)
111 iflags |= FS_SYNC_FL;
112 if (flags & BTRFS_INODE_IMMUTABLE)
113 iflags |= FS_IMMUTABLE_FL;
114 if (flags & BTRFS_INODE_APPEND)
115 iflags |= FS_APPEND_FL;
116 if (flags & BTRFS_INODE_NODUMP)
117 iflags |= FS_NODUMP_FL;
118 if (flags & BTRFS_INODE_NOATIME)
119 iflags |= FS_NOATIME_FL;
120 if (flags & BTRFS_INODE_DIRSYNC)
121 iflags |= FS_DIRSYNC_FL;
122 if (flags & BTRFS_INODE_NODATACOW)
123 iflags |= FS_NOCOW_FL;
125 if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
126 iflags |= FS_COMPR_FL;
127 else if (flags & BTRFS_INODE_NOCOMPRESS)
128 iflags |= FS_NOCOMP_FL;
134 * Update inode->i_flags based on the btrfs internal flags.
136 void btrfs_update_iflags(struct inode *inode)
138 struct btrfs_inode *ip = BTRFS_I(inode);
139 unsigned int new_fl = 0;
141 if (ip->flags & BTRFS_INODE_SYNC)
143 if (ip->flags & BTRFS_INODE_IMMUTABLE)
144 new_fl |= S_IMMUTABLE;
145 if (ip->flags & BTRFS_INODE_APPEND)
147 if (ip->flags & BTRFS_INODE_NOATIME)
149 if (ip->flags & BTRFS_INODE_DIRSYNC)
152 set_mask_bits(&inode->i_flags,
153 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC,
158 * Inherit flags from the parent inode.
160 * Currently only the compression flags and the cow flags are inherited.
162 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
169 flags = BTRFS_I(dir)->flags;
171 if (flags & BTRFS_INODE_NOCOMPRESS) {
172 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
173 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
174 } else if (flags & BTRFS_INODE_COMPRESS) {
175 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
176 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
179 if (flags & BTRFS_INODE_NODATACOW) {
180 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
181 if (S_ISREG(inode->i_mode))
182 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
185 btrfs_update_iflags(inode);
188 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
190 struct btrfs_inode *ip = BTRFS_I(file_inode(file));
191 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
193 if (copy_to_user(arg, &flags, sizeof(flags)))
198 static int check_flags(unsigned int flags)
200 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
201 FS_NOATIME_FL | FS_NODUMP_FL | \
202 FS_SYNC_FL | FS_DIRSYNC_FL | \
203 FS_NOCOMP_FL | FS_COMPR_FL |
207 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
213 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
215 struct inode *inode = file_inode(file);
216 struct btrfs_inode *ip = BTRFS_I(inode);
217 struct btrfs_root *root = ip->root;
218 struct btrfs_trans_handle *trans;
219 unsigned int flags, oldflags;
222 unsigned int i_oldflags;
225 if (!inode_owner_or_capable(inode))
228 if (btrfs_root_readonly(root))
231 if (copy_from_user(&flags, arg, sizeof(flags)))
234 ret = check_flags(flags);
238 ret = mnt_want_write_file(file);
242 mutex_lock(&inode->i_mutex);
244 ip_oldflags = ip->flags;
245 i_oldflags = inode->i_flags;
246 mode = inode->i_mode;
248 flags = btrfs_mask_flags(inode->i_mode, flags);
249 oldflags = btrfs_flags_to_ioctl(ip->flags);
250 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
251 if (!capable(CAP_LINUX_IMMUTABLE)) {
257 if (flags & FS_SYNC_FL)
258 ip->flags |= BTRFS_INODE_SYNC;
260 ip->flags &= ~BTRFS_INODE_SYNC;
261 if (flags & FS_IMMUTABLE_FL)
262 ip->flags |= BTRFS_INODE_IMMUTABLE;
264 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
265 if (flags & FS_APPEND_FL)
266 ip->flags |= BTRFS_INODE_APPEND;
268 ip->flags &= ~BTRFS_INODE_APPEND;
269 if (flags & FS_NODUMP_FL)
270 ip->flags |= BTRFS_INODE_NODUMP;
272 ip->flags &= ~BTRFS_INODE_NODUMP;
273 if (flags & FS_NOATIME_FL)
274 ip->flags |= BTRFS_INODE_NOATIME;
276 ip->flags &= ~BTRFS_INODE_NOATIME;
277 if (flags & FS_DIRSYNC_FL)
278 ip->flags |= BTRFS_INODE_DIRSYNC;
280 ip->flags &= ~BTRFS_INODE_DIRSYNC;
281 if (flags & FS_NOCOW_FL) {
284 * It's safe to turn csums off here, no extents exist.
285 * Otherwise we want the flag to reflect the real COW
286 * status of the file and will not set it.
288 if (inode->i_size == 0)
289 ip->flags |= BTRFS_INODE_NODATACOW
290 | BTRFS_INODE_NODATASUM;
292 ip->flags |= BTRFS_INODE_NODATACOW;
296 * Revert back under same assuptions as above
299 if (inode->i_size == 0)
300 ip->flags &= ~(BTRFS_INODE_NODATACOW
301 | BTRFS_INODE_NODATASUM);
303 ip->flags &= ~BTRFS_INODE_NODATACOW;
308 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
309 * flag may be changed automatically if compression code won't make
312 if (flags & FS_NOCOMP_FL) {
313 ip->flags &= ~BTRFS_INODE_COMPRESS;
314 ip->flags |= BTRFS_INODE_NOCOMPRESS;
316 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
317 if (ret && ret != -ENODATA)
319 } else if (flags & FS_COMPR_FL) {
322 ip->flags |= BTRFS_INODE_COMPRESS;
323 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
325 if (root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
329 ret = btrfs_set_prop(inode, "btrfs.compression",
330 comp, strlen(comp), 0);
335 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
338 trans = btrfs_start_transaction(root, 1);
340 ret = PTR_ERR(trans);
344 btrfs_update_iflags(inode);
345 inode_inc_iversion(inode);
346 inode->i_ctime = CURRENT_TIME;
347 ret = btrfs_update_inode(trans, root, inode);
349 btrfs_end_transaction(trans, root);
352 ip->flags = ip_oldflags;
353 inode->i_flags = i_oldflags;
357 mutex_unlock(&inode->i_mutex);
358 mnt_drop_write_file(file);
362 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
364 struct inode *inode = file_inode(file);
366 return put_user(inode->i_generation, arg);
369 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
371 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
372 struct btrfs_device *device;
373 struct request_queue *q;
374 struct fstrim_range range;
375 u64 minlen = ULLONG_MAX;
377 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
380 if (!capable(CAP_SYS_ADMIN))
384 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
388 q = bdev_get_queue(device->bdev);
389 if (blk_queue_discard(q)) {
391 minlen = min((u64)q->limits.discard_granularity,
399 if (copy_from_user(&range, arg, sizeof(range)))
401 if (range.start > total_bytes ||
402 range.len < fs_info->sb->s_blocksize)
405 range.len = min(range.len, total_bytes - range.start);
406 range.minlen = max(range.minlen, minlen);
407 ret = btrfs_trim_fs(fs_info->tree_root, &range);
411 if (copy_to_user(arg, &range, sizeof(range)))
417 int btrfs_is_empty_uuid(u8 *uuid)
421 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
428 static noinline int create_subvol(struct inode *dir,
429 struct dentry *dentry,
430 char *name, int namelen,
432 struct btrfs_qgroup_inherit *inherit)
434 struct btrfs_trans_handle *trans;
435 struct btrfs_key key;
436 struct btrfs_root_item root_item;
437 struct btrfs_inode_item *inode_item;
438 struct extent_buffer *leaf;
439 struct btrfs_root *root = BTRFS_I(dir)->root;
440 struct btrfs_root *new_root;
441 struct btrfs_block_rsv block_rsv;
442 struct timespec cur_time = CURRENT_TIME;
447 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
452 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
456 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
458 * The same as the snapshot creation, please see the comment
459 * of create_snapshot().
461 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
462 8, &qgroup_reserved, false);
466 trans = btrfs_start_transaction(root, 0);
468 ret = PTR_ERR(trans);
469 btrfs_subvolume_release_metadata(root, &block_rsv,
473 trans->block_rsv = &block_rsv;
474 trans->bytes_reserved = block_rsv.size;
476 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
480 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
481 0, objectid, NULL, 0, 0, 0);
487 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
488 btrfs_set_header_bytenr(leaf, leaf->start);
489 btrfs_set_header_generation(leaf, trans->transid);
490 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
491 btrfs_set_header_owner(leaf, objectid);
493 write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(),
495 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
496 btrfs_header_chunk_tree_uuid(leaf),
498 btrfs_mark_buffer_dirty(leaf);
500 memset(&root_item, 0, sizeof(root_item));
502 inode_item = &root_item.inode;
503 btrfs_set_stack_inode_generation(inode_item, 1);
504 btrfs_set_stack_inode_size(inode_item, 3);
505 btrfs_set_stack_inode_nlink(inode_item, 1);
506 btrfs_set_stack_inode_nbytes(inode_item, root->leafsize);
507 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
509 btrfs_set_root_flags(&root_item, 0);
510 btrfs_set_root_limit(&root_item, 0);
511 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
513 btrfs_set_root_bytenr(&root_item, leaf->start);
514 btrfs_set_root_generation(&root_item, trans->transid);
515 btrfs_set_root_level(&root_item, 0);
516 btrfs_set_root_refs(&root_item, 1);
517 btrfs_set_root_used(&root_item, leaf->len);
518 btrfs_set_root_last_snapshot(&root_item, 0);
520 btrfs_set_root_generation_v2(&root_item,
521 btrfs_root_generation(&root_item));
522 uuid_le_gen(&new_uuid);
523 memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
524 btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec);
525 btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec);
526 root_item.ctime = root_item.otime;
527 btrfs_set_root_ctransid(&root_item, trans->transid);
528 btrfs_set_root_otransid(&root_item, trans->transid);
530 btrfs_tree_unlock(leaf);
531 free_extent_buffer(leaf);
534 btrfs_set_root_dirid(&root_item, new_dirid);
536 key.objectid = objectid;
538 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
539 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
544 key.offset = (u64)-1;
545 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
546 if (IS_ERR(new_root)) {
547 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
548 ret = PTR_ERR(new_root);
552 btrfs_record_root_in_trans(trans, new_root);
554 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
556 /* We potentially lose an unused inode item here */
557 btrfs_abort_transaction(trans, root, ret);
562 * insert the directory item
564 ret = btrfs_set_inode_index(dir, &index);
566 btrfs_abort_transaction(trans, root, ret);
570 ret = btrfs_insert_dir_item(trans, root,
571 name, namelen, dir, &key,
572 BTRFS_FT_DIR, index);
574 btrfs_abort_transaction(trans, root, ret);
578 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
579 ret = btrfs_update_inode(trans, root, dir);
582 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
583 objectid, root->root_key.objectid,
584 btrfs_ino(dir), index, name, namelen);
587 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
588 root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
591 btrfs_abort_transaction(trans, root, ret);
594 trans->block_rsv = NULL;
595 trans->bytes_reserved = 0;
596 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
599 *async_transid = trans->transid;
600 err = btrfs_commit_transaction_async(trans, root, 1);
602 err = btrfs_commit_transaction(trans, root);
604 err = btrfs_commit_transaction(trans, root);
610 inode = btrfs_lookup_dentry(dir, dentry);
612 return PTR_ERR(inode);
613 d_instantiate(dentry, inode);
618 static void btrfs_wait_nocow_write(struct btrfs_root *root)
624 prepare_to_wait(&root->subv_writers->wait, &wait,
625 TASK_UNINTERRUPTIBLE);
627 writers = percpu_counter_sum(&root->subv_writers->counter);
631 finish_wait(&root->subv_writers->wait, &wait);
635 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
636 struct dentry *dentry, char *name, int namelen,
637 u64 *async_transid, bool readonly,
638 struct btrfs_qgroup_inherit *inherit)
641 struct btrfs_pending_snapshot *pending_snapshot;
642 struct btrfs_trans_handle *trans;
645 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
648 atomic_inc(&root->will_be_snapshoted);
649 smp_mb__after_atomic();
650 btrfs_wait_nocow_write(root);
652 ret = btrfs_start_delalloc_inodes(root, 0);
656 btrfs_wait_ordered_extents(root, -1);
658 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
659 if (!pending_snapshot) {
664 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
665 BTRFS_BLOCK_RSV_TEMP);
667 * 1 - parent dir inode
670 * 2 - root ref/backref
671 * 1 - root of snapshot
674 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
675 &pending_snapshot->block_rsv, 8,
676 &pending_snapshot->qgroup_reserved,
681 pending_snapshot->dentry = dentry;
682 pending_snapshot->root = root;
683 pending_snapshot->readonly = readonly;
684 pending_snapshot->dir = dir;
685 pending_snapshot->inherit = inherit;
687 trans = btrfs_start_transaction(root, 0);
689 ret = PTR_ERR(trans);
693 spin_lock(&root->fs_info->trans_lock);
694 list_add(&pending_snapshot->list,
695 &trans->transaction->pending_snapshots);
696 spin_unlock(&root->fs_info->trans_lock);
698 *async_transid = trans->transid;
699 ret = btrfs_commit_transaction_async(trans,
700 root->fs_info->extent_root, 1);
702 ret = btrfs_commit_transaction(trans, root);
704 ret = btrfs_commit_transaction(trans,
705 root->fs_info->extent_root);
710 ret = pending_snapshot->error;
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;
939 path->keep_locks = 1;
940 ret = btrfs_search_forward(root, &min_key, path, newer_than);
943 path->keep_locks = 0;
944 btrfs_unlock_up_safe(path, 1);
946 if (min_key.objectid != ino)
948 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
951 leaf = path->nodes[0];
952 extent = btrfs_item_ptr(leaf, path->slots[0],
953 struct btrfs_file_extent_item);
955 type = btrfs_file_extent_type(leaf, extent);
956 if (type == BTRFS_FILE_EXTENT_REG &&
957 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
958 check_defrag_in_cache(inode, min_key.offset, thresh)) {
959 *off = min_key.offset;
960 btrfs_free_path(path);
965 if (path->slots[0] < btrfs_header_nritems(leaf)) {
966 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
970 if (min_key.offset == (u64)-1)
974 btrfs_release_path(path);
977 btrfs_free_path(path);
981 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
983 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
984 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
985 struct extent_map *em;
986 u64 len = PAGE_CACHE_SIZE;
989 * hopefully we have this extent in the tree already, try without
990 * the full extent lock
992 read_lock(&em_tree->lock);
993 em = lookup_extent_mapping(em_tree, start, len);
994 read_unlock(&em_tree->lock);
997 struct extent_state *cached = NULL;
998 u64 end = start + len - 1;
1000 /* get the big lock and read metadata off disk */
1001 lock_extent_bits(io_tree, start, end, 0, &cached);
1002 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
1003 unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS);
1012 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1014 struct extent_map *next;
1017 /* this is the last extent */
1018 if (em->start + em->len >= i_size_read(inode))
1021 next = defrag_lookup_extent(inode, em->start + em->len);
1022 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE ||
1023 (em->block_start + em->block_len == next->block_start))
1026 free_extent_map(next);
1030 static int should_defrag_range(struct inode *inode, u64 start, int thresh,
1031 u64 *last_len, u64 *skip, u64 *defrag_end,
1034 struct extent_map *em;
1036 bool next_mergeable = true;
1039 * make sure that once we start defragging an extent, we keep on
1042 if (start < *defrag_end)
1047 em = defrag_lookup_extent(inode, start);
1051 /* this will cover holes, and inline extents */
1052 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1057 next_mergeable = defrag_check_next_extent(inode, em);
1060 * we hit a real extent, if it is big or the next extent is not a
1061 * real extent, don't bother defragging it
1063 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1064 (em->len >= thresh || !next_mergeable))
1068 * last_len ends up being a counter of how many bytes we've defragged.
1069 * every time we choose not to defrag an extent, we reset *last_len
1070 * so that the next tiny extent will force a defrag.
1072 * The end result of this is that tiny extents before a single big
1073 * extent will force at least part of that big extent to be defragged.
1076 *defrag_end = extent_map_end(em);
1079 *skip = extent_map_end(em);
1083 free_extent_map(em);
1088 * it doesn't do much good to defrag one or two pages
1089 * at a time. This pulls in a nice chunk of pages
1090 * to COW and defrag.
1092 * It also makes sure the delalloc code has enough
1093 * dirty data to avoid making new small extents as part
1096 * It's a good idea to start RA on this range
1097 * before calling this.
1099 static int cluster_pages_for_defrag(struct inode *inode,
1100 struct page **pages,
1101 unsigned long start_index,
1102 unsigned long num_pages)
1104 unsigned long file_end;
1105 u64 isize = i_size_read(inode);
1112 struct btrfs_ordered_extent *ordered;
1113 struct extent_state *cached_state = NULL;
1114 struct extent_io_tree *tree;
1115 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1117 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1118 if (!isize || start_index > file_end)
1121 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1123 ret = btrfs_delalloc_reserve_space(inode,
1124 page_cnt << PAGE_CACHE_SHIFT);
1128 tree = &BTRFS_I(inode)->io_tree;
1130 /* step one, lock all the pages */
1131 for (i = 0; i < page_cnt; i++) {
1134 page = find_or_create_page(inode->i_mapping,
1135 start_index + i, mask);
1139 page_start = page_offset(page);
1140 page_end = page_start + PAGE_CACHE_SIZE - 1;
1142 lock_extent_bits(tree, page_start, page_end,
1144 ordered = btrfs_lookup_ordered_extent(inode,
1146 unlock_extent_cached(tree, page_start, page_end,
1147 &cached_state, GFP_NOFS);
1152 btrfs_start_ordered_extent(inode, ordered, 1);
1153 btrfs_put_ordered_extent(ordered);
1156 * we unlocked the page above, so we need check if
1157 * it was released or not.
1159 if (page->mapping != inode->i_mapping) {
1161 page_cache_release(page);
1166 if (!PageUptodate(page)) {
1167 btrfs_readpage(NULL, page);
1169 if (!PageUptodate(page)) {
1171 page_cache_release(page);
1177 if (page->mapping != inode->i_mapping) {
1179 page_cache_release(page);
1189 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1193 * so now we have a nice long stream of locked
1194 * and up to date pages, lets wait on them
1196 for (i = 0; i < i_done; i++)
1197 wait_on_page_writeback(pages[i]);
1199 page_start = page_offset(pages[0]);
1200 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1202 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1203 page_start, page_end - 1, 0, &cached_state);
1204 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1205 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1206 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1207 &cached_state, GFP_NOFS);
1209 if (i_done != page_cnt) {
1210 spin_lock(&BTRFS_I(inode)->lock);
1211 BTRFS_I(inode)->outstanding_extents++;
1212 spin_unlock(&BTRFS_I(inode)->lock);
1213 btrfs_delalloc_release_space(inode,
1214 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1218 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1219 &cached_state, GFP_NOFS);
1221 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1222 page_start, page_end - 1, &cached_state,
1225 for (i = 0; i < i_done; i++) {
1226 clear_page_dirty_for_io(pages[i]);
1227 ClearPageChecked(pages[i]);
1228 set_page_extent_mapped(pages[i]);
1229 set_page_dirty(pages[i]);
1230 unlock_page(pages[i]);
1231 page_cache_release(pages[i]);
1235 for (i = 0; i < i_done; i++) {
1236 unlock_page(pages[i]);
1237 page_cache_release(pages[i]);
1239 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1244 int btrfs_defrag_file(struct inode *inode, struct file *file,
1245 struct btrfs_ioctl_defrag_range_args *range,
1246 u64 newer_than, unsigned long max_to_defrag)
1248 struct btrfs_root *root = BTRFS_I(inode)->root;
1249 struct file_ra_state *ra = NULL;
1250 unsigned long last_index;
1251 u64 isize = i_size_read(inode);
1255 u64 newer_off = range->start;
1257 unsigned long ra_index = 0;
1259 int defrag_count = 0;
1260 int compress_type = BTRFS_COMPRESS_ZLIB;
1261 int extent_thresh = range->extent_thresh;
1262 unsigned long max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1263 unsigned long cluster = max_cluster;
1264 u64 new_align = ~((u64)128 * 1024 - 1);
1265 struct page **pages = NULL;
1270 if (range->start >= isize)
1273 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1274 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1276 if (range->compress_type)
1277 compress_type = range->compress_type;
1280 if (extent_thresh == 0)
1281 extent_thresh = 256 * 1024;
1284 * if we were not given a file, allocate a readahead
1288 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1291 file_ra_state_init(ra, inode->i_mapping);
1296 pages = kmalloc_array(max_cluster, sizeof(struct page *),
1303 /* find the last page to defrag */
1304 if (range->start + range->len > range->start) {
1305 last_index = min_t(u64, isize - 1,
1306 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1308 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1312 ret = find_new_extents(root, inode, newer_than,
1313 &newer_off, 64 * 1024);
1315 range->start = newer_off;
1317 * we always align our defrag to help keep
1318 * the extents in the file evenly spaced
1320 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1324 i = range->start >> PAGE_CACHE_SHIFT;
1327 max_to_defrag = last_index + 1;
1330 * make writeback starts from i, so the defrag range can be
1331 * written sequentially.
1333 if (i < inode->i_mapping->writeback_index)
1334 inode->i_mapping->writeback_index = i;
1336 while (i <= last_index && defrag_count < max_to_defrag &&
1337 (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1338 PAGE_CACHE_SHIFT)) {
1340 * make sure we stop running if someone unmounts
1343 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1346 if (btrfs_defrag_cancelled(root->fs_info)) {
1347 printk(KERN_DEBUG "BTRFS: defrag_file cancelled\n");
1352 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1353 extent_thresh, &last_len, &skip,
1354 &defrag_end, range->flags &
1355 BTRFS_DEFRAG_RANGE_COMPRESS)) {
1358 * the should_defrag function tells us how much to skip
1359 * bump our counter by the suggested amount
1361 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1362 i = max(i + 1, next);
1367 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1368 PAGE_CACHE_SHIFT) - i;
1369 cluster = min(cluster, max_cluster);
1371 cluster = max_cluster;
1374 if (i + cluster > ra_index) {
1375 ra_index = max(i, ra_index);
1376 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1378 ra_index += max_cluster;
1381 mutex_lock(&inode->i_mutex);
1382 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1383 BTRFS_I(inode)->force_compress = compress_type;
1384 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1386 mutex_unlock(&inode->i_mutex);
1390 defrag_count += ret;
1391 balance_dirty_pages_ratelimited(inode->i_mapping);
1392 mutex_unlock(&inode->i_mutex);
1395 if (newer_off == (u64)-1)
1401 newer_off = max(newer_off + 1,
1402 (u64)i << PAGE_CACHE_SHIFT);
1404 ret = find_new_extents(root, inode,
1405 newer_than, &newer_off,
1408 range->start = newer_off;
1409 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1416 last_len += ret << PAGE_CACHE_SHIFT;
1424 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1425 filemap_flush(inode->i_mapping);
1426 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1427 &BTRFS_I(inode)->runtime_flags))
1428 filemap_flush(inode->i_mapping);
1431 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1432 /* the filemap_flush will queue IO into the worker threads, but
1433 * we have to make sure the IO is actually started and that
1434 * ordered extents get created before we return
1436 atomic_inc(&root->fs_info->async_submit_draining);
1437 while (atomic_read(&root->fs_info->nr_async_submits) ||
1438 atomic_read(&root->fs_info->async_delalloc_pages)) {
1439 wait_event(root->fs_info->async_submit_wait,
1440 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1441 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1443 atomic_dec(&root->fs_info->async_submit_draining);
1446 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1447 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1453 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1454 mutex_lock(&inode->i_mutex);
1455 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1456 mutex_unlock(&inode->i_mutex);
1464 static noinline int btrfs_ioctl_resize(struct file *file,
1470 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
1471 struct btrfs_ioctl_vol_args *vol_args;
1472 struct btrfs_trans_handle *trans;
1473 struct btrfs_device *device = NULL;
1476 char *devstr = NULL;
1480 if (!capable(CAP_SYS_ADMIN))
1483 ret = mnt_want_write_file(file);
1487 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1489 mnt_drop_write_file(file);
1490 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1493 mutex_lock(&root->fs_info->volume_mutex);
1494 vol_args = memdup_user(arg, sizeof(*vol_args));
1495 if (IS_ERR(vol_args)) {
1496 ret = PTR_ERR(vol_args);
1500 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1502 sizestr = vol_args->name;
1503 devstr = strchr(sizestr, ':');
1505 sizestr = devstr + 1;
1507 devstr = vol_args->name;
1508 ret = kstrtoull(devstr, 10, &devid);
1515 btrfs_info(root->fs_info, "resizing devid %llu", devid);
1518 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
1520 btrfs_info(root->fs_info, "resizer unable to find device %llu",
1526 if (!device->writeable) {
1527 btrfs_info(root->fs_info,
1528 "resizer unable to apply on readonly device %llu",
1534 if (!strcmp(sizestr, "max"))
1535 new_size = device->bdev->bd_inode->i_size;
1537 if (sizestr[0] == '-') {
1540 } else if (sizestr[0] == '+') {
1544 new_size = memparse(sizestr, &retptr);
1545 if (*retptr != '\0' || new_size == 0) {
1551 if (device->is_tgtdev_for_dev_replace) {
1556 old_size = device->total_bytes;
1559 if (new_size > old_size) {
1563 new_size = old_size - new_size;
1564 } else if (mod > 0) {
1565 if (new_size > ULLONG_MAX - old_size) {
1569 new_size = old_size + new_size;
1572 if (new_size < 256 * 1024 * 1024) {
1576 if (new_size > device->bdev->bd_inode->i_size) {
1581 do_div(new_size, root->sectorsize);
1582 new_size *= root->sectorsize;
1584 printk_in_rcu(KERN_INFO "BTRFS: new size for %s is %llu\n",
1585 rcu_str_deref(device->name), new_size);
1587 if (new_size > old_size) {
1588 trans = btrfs_start_transaction(root, 0);
1589 if (IS_ERR(trans)) {
1590 ret = PTR_ERR(trans);
1593 ret = btrfs_grow_device(trans, device, new_size);
1594 btrfs_commit_transaction(trans, root);
1595 } else if (new_size < old_size) {
1596 ret = btrfs_shrink_device(device, new_size);
1597 } /* equal, nothing need to do */
1602 mutex_unlock(&root->fs_info->volume_mutex);
1603 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
1604 mnt_drop_write_file(file);
1608 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1609 char *name, unsigned long fd, int subvol,
1610 u64 *transid, bool readonly,
1611 struct btrfs_qgroup_inherit *inherit)
1616 ret = mnt_want_write_file(file);
1620 namelen = strlen(name);
1621 if (strchr(name, '/')) {
1623 goto out_drop_write;
1626 if (name[0] == '.' &&
1627 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1629 goto out_drop_write;
1633 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1634 NULL, transid, readonly, inherit);
1636 struct fd src = fdget(fd);
1637 struct inode *src_inode;
1640 goto out_drop_write;
1643 src_inode = file_inode(src.file);
1644 if (src_inode->i_sb != file_inode(file)->i_sb) {
1645 btrfs_info(BTRFS_I(src_inode)->root->fs_info,
1646 "Snapshot src from another FS");
1648 } else if (!inode_owner_or_capable(src_inode)) {
1650 * Subvolume creation is not restricted, but snapshots
1651 * are limited to own subvolumes only
1655 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1656 BTRFS_I(src_inode)->root,
1657 transid, readonly, inherit);
1662 mnt_drop_write_file(file);
1667 static noinline int btrfs_ioctl_snap_create(struct file *file,
1668 void __user *arg, int subvol)
1670 struct btrfs_ioctl_vol_args *vol_args;
1673 vol_args = memdup_user(arg, sizeof(*vol_args));
1674 if (IS_ERR(vol_args))
1675 return PTR_ERR(vol_args);
1676 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1678 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1679 vol_args->fd, subvol,
1686 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1687 void __user *arg, int subvol)
1689 struct btrfs_ioctl_vol_args_v2 *vol_args;
1693 bool readonly = false;
1694 struct btrfs_qgroup_inherit *inherit = NULL;
1696 vol_args = memdup_user(arg, sizeof(*vol_args));
1697 if (IS_ERR(vol_args))
1698 return PTR_ERR(vol_args);
1699 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1701 if (vol_args->flags &
1702 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1703 BTRFS_SUBVOL_QGROUP_INHERIT)) {
1708 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1710 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1712 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1713 if (vol_args->size > PAGE_CACHE_SIZE) {
1717 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1718 if (IS_ERR(inherit)) {
1719 ret = PTR_ERR(inherit);
1724 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1725 vol_args->fd, subvol, ptr,
1728 if (ret == 0 && ptr &&
1730 offsetof(struct btrfs_ioctl_vol_args_v2,
1731 transid), ptr, sizeof(*ptr)))
1739 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1742 struct inode *inode = file_inode(file);
1743 struct btrfs_root *root = BTRFS_I(inode)->root;
1747 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1750 down_read(&root->fs_info->subvol_sem);
1751 if (btrfs_root_readonly(root))
1752 flags |= BTRFS_SUBVOL_RDONLY;
1753 up_read(&root->fs_info->subvol_sem);
1755 if (copy_to_user(arg, &flags, sizeof(flags)))
1761 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1764 struct inode *inode = file_inode(file);
1765 struct btrfs_root *root = BTRFS_I(inode)->root;
1766 struct btrfs_trans_handle *trans;
1771 if (!inode_owner_or_capable(inode))
1774 ret = mnt_want_write_file(file);
1778 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1780 goto out_drop_write;
1783 if (copy_from_user(&flags, arg, sizeof(flags))) {
1785 goto out_drop_write;
1788 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1790 goto out_drop_write;
1793 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1795 goto out_drop_write;
1798 down_write(&root->fs_info->subvol_sem);
1801 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1804 root_flags = btrfs_root_flags(&root->root_item);
1805 if (flags & BTRFS_SUBVOL_RDONLY) {
1806 btrfs_set_root_flags(&root->root_item,
1807 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1810 * Block RO -> RW transition if this subvolume is involved in
1813 spin_lock(&root->root_item_lock);
1814 if (root->send_in_progress == 0) {
1815 btrfs_set_root_flags(&root->root_item,
1816 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1817 spin_unlock(&root->root_item_lock);
1819 spin_unlock(&root->root_item_lock);
1820 btrfs_warn(root->fs_info,
1821 "Attempt to set subvolume %llu read-write during send",
1822 root->root_key.objectid);
1828 trans = btrfs_start_transaction(root, 1);
1829 if (IS_ERR(trans)) {
1830 ret = PTR_ERR(trans);
1834 ret = btrfs_update_root(trans, root->fs_info->tree_root,
1835 &root->root_key, &root->root_item);
1837 btrfs_commit_transaction(trans, root);
1840 btrfs_set_root_flags(&root->root_item, root_flags);
1842 up_write(&root->fs_info->subvol_sem);
1844 mnt_drop_write_file(file);
1850 * helper to check if the subvolume references other subvolumes
1852 static noinline int may_destroy_subvol(struct btrfs_root *root)
1854 struct btrfs_path *path;
1855 struct btrfs_dir_item *di;
1856 struct btrfs_key key;
1860 path = btrfs_alloc_path();
1864 /* Make sure this root isn't set as the default subvol */
1865 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1866 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1867 dir_id, "default", 7, 0);
1868 if (di && !IS_ERR(di)) {
1869 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1870 if (key.objectid == root->root_key.objectid) {
1872 btrfs_err(root->fs_info, "deleting default subvolume "
1873 "%llu is not allowed", key.objectid);
1876 btrfs_release_path(path);
1879 key.objectid = root->root_key.objectid;
1880 key.type = BTRFS_ROOT_REF_KEY;
1881 key.offset = (u64)-1;
1883 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1890 if (path->slots[0] > 0) {
1892 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1893 if (key.objectid == root->root_key.objectid &&
1894 key.type == BTRFS_ROOT_REF_KEY)
1898 btrfs_free_path(path);
1902 static noinline int key_in_sk(struct btrfs_key *key,
1903 struct btrfs_ioctl_search_key *sk)
1905 struct btrfs_key test;
1908 test.objectid = sk->min_objectid;
1909 test.type = sk->min_type;
1910 test.offset = sk->min_offset;
1912 ret = btrfs_comp_cpu_keys(key, &test);
1916 test.objectid = sk->max_objectid;
1917 test.type = sk->max_type;
1918 test.offset = sk->max_offset;
1920 ret = btrfs_comp_cpu_keys(key, &test);
1926 static noinline int copy_to_sk(struct btrfs_root *root,
1927 struct btrfs_path *path,
1928 struct btrfs_key *key,
1929 struct btrfs_ioctl_search_key *sk,
1932 unsigned long *sk_offset,
1936 struct extent_buffer *leaf;
1937 struct btrfs_ioctl_search_header sh;
1938 unsigned long item_off;
1939 unsigned long item_len;
1945 leaf = path->nodes[0];
1946 slot = path->slots[0];
1947 nritems = btrfs_header_nritems(leaf);
1949 if (btrfs_header_generation(leaf) > sk->max_transid) {
1953 found_transid = btrfs_header_generation(leaf);
1955 for (i = slot; i < nritems; i++) {
1956 item_off = btrfs_item_ptr_offset(leaf, i);
1957 item_len = btrfs_item_size_nr(leaf, i);
1959 btrfs_item_key_to_cpu(leaf, key, i);
1960 if (!key_in_sk(key, sk))
1963 if (sizeof(sh) + item_len > *buf_size) {
1970 * return one empty item back for v1, which does not
1974 *buf_size = sizeof(sh) + item_len;
1979 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
1984 sh.objectid = key->objectid;
1985 sh.offset = key->offset;
1986 sh.type = key->type;
1988 sh.transid = found_transid;
1990 /* copy search result header */
1991 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
1996 *sk_offset += sizeof(sh);
1999 char __user *up = ubuf + *sk_offset;
2001 if (read_extent_buffer_to_user(leaf, up,
2002 item_off, item_len)) {
2007 *sk_offset += item_len;
2011 if (ret) /* -EOVERFLOW from above */
2014 if (*num_found >= sk->nr_items) {
2021 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
2023 else if (key->type < (u8)-1 && key->type < sk->max_type) {
2026 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
2034 * 0: all items from this leaf copied, continue with next
2035 * 1: * more items can be copied, but unused buffer is too small
2036 * * all items were found
2037 * Either way, it will stops the loop which iterates to the next
2039 * -EOVERFLOW: item was to large for buffer
2040 * -EFAULT: could not copy extent buffer back to userspace
2045 static noinline int search_ioctl(struct inode *inode,
2046 struct btrfs_ioctl_search_key *sk,
2050 struct btrfs_root *root;
2051 struct btrfs_key key;
2052 struct btrfs_path *path;
2053 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
2056 unsigned long sk_offset = 0;
2058 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2059 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2063 path = btrfs_alloc_path();
2067 if (sk->tree_id == 0) {
2068 /* search the root of the inode that was passed */
2069 root = BTRFS_I(inode)->root;
2071 key.objectid = sk->tree_id;
2072 key.type = BTRFS_ROOT_ITEM_KEY;
2073 key.offset = (u64)-1;
2074 root = btrfs_read_fs_root_no_name(info, &key);
2076 printk(KERN_ERR "BTRFS: could not find root %llu\n",
2078 btrfs_free_path(path);
2083 key.objectid = sk->min_objectid;
2084 key.type = sk->min_type;
2085 key.offset = sk->min_offset;
2087 path->keep_locks = 1;
2090 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2096 ret = copy_to_sk(root, path, &key, sk, buf_size, ubuf,
2097 &sk_offset, &num_found);
2098 btrfs_release_path(path);
2106 sk->nr_items = num_found;
2107 btrfs_free_path(path);
2111 static noinline int btrfs_ioctl_tree_search(struct file *file,
2114 struct btrfs_ioctl_search_args __user *uargs;
2115 struct btrfs_ioctl_search_key sk;
2116 struct inode *inode;
2120 if (!capable(CAP_SYS_ADMIN))
2123 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2125 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2128 buf_size = sizeof(uargs->buf);
2130 inode = file_inode(file);
2131 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2134 * In the origin implementation an overflow is handled by returning a
2135 * search header with a len of zero, so reset ret.
2137 if (ret == -EOVERFLOW)
2140 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2145 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2148 struct btrfs_ioctl_search_args_v2 __user *uarg;
2149 struct btrfs_ioctl_search_args_v2 args;
2150 struct inode *inode;
2153 const size_t buf_limit = 16 * 1024 * 1024;
2155 if (!capable(CAP_SYS_ADMIN))
2158 /* copy search header and buffer size */
2159 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2160 if (copy_from_user(&args, uarg, sizeof(args)))
2163 buf_size = args.buf_size;
2165 if (buf_size < sizeof(struct btrfs_ioctl_search_header))
2168 /* limit result size to 16MB */
2169 if (buf_size > buf_limit)
2170 buf_size = buf_limit;
2172 inode = file_inode(file);
2173 ret = search_ioctl(inode, &args.key, &buf_size,
2174 (char *)(&uarg->buf[0]));
2175 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2177 else if (ret == -EOVERFLOW &&
2178 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2185 * Search INODE_REFs to identify path name of 'dirid' directory
2186 * in a 'tree_id' tree. and sets path name to 'name'.
2188 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2189 u64 tree_id, u64 dirid, char *name)
2191 struct btrfs_root *root;
2192 struct btrfs_key key;
2198 struct btrfs_inode_ref *iref;
2199 struct extent_buffer *l;
2200 struct btrfs_path *path;
2202 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2207 path = btrfs_alloc_path();
2211 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
2213 key.objectid = tree_id;
2214 key.type = BTRFS_ROOT_ITEM_KEY;
2215 key.offset = (u64)-1;
2216 root = btrfs_read_fs_root_no_name(info, &key);
2218 printk(KERN_ERR "BTRFS: could not find root %llu\n", tree_id);
2223 key.objectid = dirid;
2224 key.type = BTRFS_INODE_REF_KEY;
2225 key.offset = (u64)-1;
2228 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2232 ret = btrfs_previous_item(root, path, dirid,
2233 BTRFS_INODE_REF_KEY);
2243 slot = path->slots[0];
2244 btrfs_item_key_to_cpu(l, &key, slot);
2246 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2247 len = btrfs_inode_ref_name_len(l, iref);
2249 total_len += len + 1;
2251 ret = -ENAMETOOLONG;
2256 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2258 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2261 btrfs_release_path(path);
2262 key.objectid = key.offset;
2263 key.offset = (u64)-1;
2264 dirid = key.objectid;
2266 memmove(name, ptr, total_len);
2267 name[total_len] = '\0';
2270 btrfs_free_path(path);
2274 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2277 struct btrfs_ioctl_ino_lookup_args *args;
2278 struct inode *inode;
2281 if (!capable(CAP_SYS_ADMIN))
2284 args = memdup_user(argp, sizeof(*args));
2286 return PTR_ERR(args);
2288 inode = file_inode(file);
2290 if (args->treeid == 0)
2291 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2293 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2294 args->treeid, args->objectid,
2297 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2304 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2307 struct dentry *parent = file->f_path.dentry;
2308 struct dentry *dentry;
2309 struct inode *dir = parent->d_inode;
2310 struct inode *inode;
2311 struct btrfs_root *root = BTRFS_I(dir)->root;
2312 struct btrfs_root *dest = NULL;
2313 struct btrfs_ioctl_vol_args *vol_args;
2314 struct btrfs_trans_handle *trans;
2315 struct btrfs_block_rsv block_rsv;
2317 u64 qgroup_reserved;
2322 vol_args = memdup_user(arg, sizeof(*vol_args));
2323 if (IS_ERR(vol_args))
2324 return PTR_ERR(vol_args);
2326 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2327 namelen = strlen(vol_args->name);
2328 if (strchr(vol_args->name, '/') ||
2329 strncmp(vol_args->name, "..", namelen) == 0) {
2334 err = mnt_want_write_file(file);
2339 err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2341 goto out_drop_write;
2342 dentry = lookup_one_len(vol_args->name, parent, namelen);
2343 if (IS_ERR(dentry)) {
2344 err = PTR_ERR(dentry);
2345 goto out_unlock_dir;
2348 if (!dentry->d_inode) {
2353 inode = dentry->d_inode;
2354 dest = BTRFS_I(inode)->root;
2355 if (!capable(CAP_SYS_ADMIN)) {
2357 * Regular user. Only allow this with a special mount
2358 * option, when the user has write+exec access to the
2359 * subvol root, and when rmdir(2) would have been
2362 * Note that this is _not_ check that the subvol is
2363 * empty or doesn't contain data that we wouldn't
2364 * otherwise be able to delete.
2366 * Users who want to delete empty subvols should try
2370 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2374 * Do not allow deletion if the parent dir is the same
2375 * as the dir to be deleted. That means the ioctl
2376 * must be called on the dentry referencing the root
2377 * of the subvol, not a random directory contained
2384 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2389 /* check if subvolume may be deleted by a user */
2390 err = btrfs_may_delete(dir, dentry, 1);
2394 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2399 mutex_lock(&inode->i_mutex);
2402 * Don't allow to delete a subvolume with send in progress. This is
2403 * inside the i_mutex so the error handling that has to drop the bit
2404 * again is not run concurrently.
2406 spin_lock(&dest->root_item_lock);
2407 root_flags = btrfs_root_flags(&dest->root_item);
2408 if (dest->send_in_progress == 0) {
2409 btrfs_set_root_flags(&dest->root_item,
2410 root_flags | BTRFS_ROOT_SUBVOL_DEAD);
2411 spin_unlock(&dest->root_item_lock);
2413 spin_unlock(&dest->root_item_lock);
2414 btrfs_warn(root->fs_info,
2415 "Attempt to delete subvolume %llu during send",
2416 dest->root_key.objectid);
2421 err = d_invalidate(dentry);
2425 down_write(&root->fs_info->subvol_sem);
2427 err = may_destroy_subvol(dest);
2431 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2433 * One for dir inode, two for dir entries, two for root
2436 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2437 5, &qgroup_reserved, true);
2441 trans = btrfs_start_transaction(root, 0);
2442 if (IS_ERR(trans)) {
2443 err = PTR_ERR(trans);
2446 trans->block_rsv = &block_rsv;
2447 trans->bytes_reserved = block_rsv.size;
2449 ret = btrfs_unlink_subvol(trans, root, dir,
2450 dest->root_key.objectid,
2451 dentry->d_name.name,
2452 dentry->d_name.len);
2455 btrfs_abort_transaction(trans, root, ret);
2459 btrfs_record_root_in_trans(trans, dest);
2461 memset(&dest->root_item.drop_progress, 0,
2462 sizeof(dest->root_item.drop_progress));
2463 dest->root_item.drop_level = 0;
2464 btrfs_set_root_refs(&dest->root_item, 0);
2466 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
2467 ret = btrfs_insert_orphan_item(trans,
2468 root->fs_info->tree_root,
2469 dest->root_key.objectid);
2471 btrfs_abort_transaction(trans, root, ret);
2477 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2478 dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
2479 dest->root_key.objectid);
2480 if (ret && ret != -ENOENT) {
2481 btrfs_abort_transaction(trans, root, ret);
2485 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2486 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2487 dest->root_item.received_uuid,
2488 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2489 dest->root_key.objectid);
2490 if (ret && ret != -ENOENT) {
2491 btrfs_abort_transaction(trans, root, ret);
2498 trans->block_rsv = NULL;
2499 trans->bytes_reserved = 0;
2500 ret = btrfs_end_transaction(trans, root);
2503 inode->i_flags |= S_DEAD;
2505 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
2507 up_write(&root->fs_info->subvol_sem);
2510 spin_lock(&dest->root_item_lock);
2511 root_flags = btrfs_root_flags(&dest->root_item);
2512 btrfs_set_root_flags(&dest->root_item,
2513 root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
2514 spin_unlock(&dest->root_item_lock);
2516 mutex_unlock(&inode->i_mutex);
2518 shrink_dcache_sb(root->fs_info->sb);
2519 btrfs_invalidate_inodes(dest);
2521 ASSERT(dest->send_in_progress == 0);
2524 if (dest->cache_inode) {
2525 iput(dest->cache_inode);
2526 dest->cache_inode = NULL;
2532 mutex_unlock(&dir->i_mutex);
2534 mnt_drop_write_file(file);
2540 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2542 struct inode *inode = file_inode(file);
2543 struct btrfs_root *root = BTRFS_I(inode)->root;
2544 struct btrfs_ioctl_defrag_range_args *range;
2547 ret = mnt_want_write_file(file);
2551 if (btrfs_root_readonly(root)) {
2556 switch (inode->i_mode & S_IFMT) {
2558 if (!capable(CAP_SYS_ADMIN)) {
2562 ret = btrfs_defrag_root(root);
2565 ret = btrfs_defrag_root(root->fs_info->extent_root);
2568 if (!(file->f_mode & FMODE_WRITE)) {
2573 range = kzalloc(sizeof(*range), GFP_KERNEL);
2580 if (copy_from_user(range, argp,
2586 /* compression requires us to start the IO */
2587 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2588 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2589 range->extent_thresh = (u32)-1;
2592 /* the rest are all set to zero by kzalloc */
2593 range->len = (u64)-1;
2595 ret = btrfs_defrag_file(file_inode(file), file,
2605 mnt_drop_write_file(file);
2609 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2611 struct btrfs_ioctl_vol_args *vol_args;
2614 if (!capable(CAP_SYS_ADMIN))
2617 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2619 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2622 mutex_lock(&root->fs_info->volume_mutex);
2623 vol_args = memdup_user(arg, sizeof(*vol_args));
2624 if (IS_ERR(vol_args)) {
2625 ret = PTR_ERR(vol_args);
2629 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2630 ret = btrfs_init_new_device(root, vol_args->name);
2634 mutex_unlock(&root->fs_info->volume_mutex);
2635 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2639 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2641 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2642 struct btrfs_ioctl_vol_args *vol_args;
2645 if (!capable(CAP_SYS_ADMIN))
2648 ret = mnt_want_write_file(file);
2652 vol_args = memdup_user(arg, sizeof(*vol_args));
2653 if (IS_ERR(vol_args)) {
2654 ret = PTR_ERR(vol_args);
2658 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2660 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2662 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2666 mutex_lock(&root->fs_info->volume_mutex);
2667 ret = btrfs_rm_device(root, vol_args->name);
2668 mutex_unlock(&root->fs_info->volume_mutex);
2669 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2673 mnt_drop_write_file(file);
2677 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2679 struct btrfs_ioctl_fs_info_args *fi_args;
2680 struct btrfs_device *device;
2681 struct btrfs_device *next;
2682 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2685 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2689 mutex_lock(&fs_devices->device_list_mutex);
2690 fi_args->num_devices = fs_devices->num_devices;
2691 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2693 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2694 if (device->devid > fi_args->max_id)
2695 fi_args->max_id = device->devid;
2697 mutex_unlock(&fs_devices->device_list_mutex);
2699 fi_args->nodesize = root->fs_info->super_copy->nodesize;
2700 fi_args->sectorsize = root->fs_info->super_copy->sectorsize;
2701 fi_args->clone_alignment = root->fs_info->super_copy->sectorsize;
2703 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2710 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2712 struct btrfs_ioctl_dev_info_args *di_args;
2713 struct btrfs_device *dev;
2714 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2716 char *s_uuid = NULL;
2718 di_args = memdup_user(arg, sizeof(*di_args));
2719 if (IS_ERR(di_args))
2720 return PTR_ERR(di_args);
2722 if (!btrfs_is_empty_uuid(di_args->uuid))
2723 s_uuid = di_args->uuid;
2725 mutex_lock(&fs_devices->device_list_mutex);
2726 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
2733 di_args->devid = dev->devid;
2734 di_args->bytes_used = dev->bytes_used;
2735 di_args->total_bytes = dev->total_bytes;
2736 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2738 struct rcu_string *name;
2741 name = rcu_dereference(dev->name);
2742 strncpy(di_args->path, name->str, sizeof(di_args->path));
2744 di_args->path[sizeof(di_args->path) - 1] = 0;
2746 di_args->path[0] = '\0';
2750 mutex_unlock(&fs_devices->device_list_mutex);
2751 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2758 static struct page *extent_same_get_page(struct inode *inode, u64 off)
2762 struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
2764 index = off >> PAGE_CACHE_SHIFT;
2766 page = grab_cache_page(inode->i_mapping, index);
2770 if (!PageUptodate(page)) {
2771 if (extent_read_full_page_nolock(tree, page, btrfs_get_extent,
2775 if (!PageUptodate(page)) {
2777 page_cache_release(page);
2786 static inline void lock_extent_range(struct inode *inode, u64 off, u64 len)
2788 /* do any pending delalloc/csum calc on src, one way or
2789 another, and lock file content */
2791 struct btrfs_ordered_extent *ordered;
2792 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2793 ordered = btrfs_lookup_first_ordered_extent(inode,
2796 ordered->file_offset + ordered->len <= off ||
2797 ordered->file_offset >= off + len) &&
2798 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2799 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2801 btrfs_put_ordered_extent(ordered);
2804 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2806 btrfs_put_ordered_extent(ordered);
2807 btrfs_wait_ordered_range(inode, off, len);
2811 static void btrfs_double_unlock(struct inode *inode1, u64 loff1,
2812 struct inode *inode2, u64 loff2, u64 len)
2814 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2815 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2817 mutex_unlock(&inode1->i_mutex);
2818 mutex_unlock(&inode2->i_mutex);
2821 static void btrfs_double_lock(struct inode *inode1, u64 loff1,
2822 struct inode *inode2, u64 loff2, u64 len)
2824 if (inode1 < inode2) {
2825 swap(inode1, inode2);
2829 mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT);
2830 lock_extent_range(inode1, loff1, len);
2831 if (inode1 != inode2) {
2832 mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD);
2833 lock_extent_range(inode2, loff2, len);
2837 static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
2838 u64 dst_loff, u64 len)
2841 struct page *src_page, *dst_page;
2842 unsigned int cmp_len = PAGE_CACHE_SIZE;
2843 void *addr, *dst_addr;
2846 if (len < PAGE_CACHE_SIZE)
2849 src_page = extent_same_get_page(src, loff);
2852 dst_page = extent_same_get_page(dst, dst_loff);
2854 page_cache_release(src_page);
2857 addr = kmap_atomic(src_page);
2858 dst_addr = kmap_atomic(dst_page);
2860 flush_dcache_page(src_page);
2861 flush_dcache_page(dst_page);
2863 if (memcmp(addr, dst_addr, cmp_len))
2864 ret = BTRFS_SAME_DATA_DIFFERS;
2866 kunmap_atomic(addr);
2867 kunmap_atomic(dst_addr);
2868 page_cache_release(src_page);
2869 page_cache_release(dst_page);
2875 dst_loff += cmp_len;
2882 static int extent_same_check_offsets(struct inode *inode, u64 off, u64 len)
2884 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
2886 if (off + len > inode->i_size || off + len < off)
2888 /* Check that we are block aligned - btrfs_clone() requires this */
2889 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
2895 static int btrfs_extent_same(struct inode *src, u64 loff, u64 len,
2896 struct inode *dst, u64 dst_loff)
2901 * btrfs_clone() can't handle extents in the same file
2902 * yet. Once that works, we can drop this check and replace it
2903 * with a check for the same inode, but overlapping extents.
2908 btrfs_double_lock(src, loff, dst, dst_loff, len);
2910 ret = extent_same_check_offsets(src, loff, len);
2914 ret = extent_same_check_offsets(dst, dst_loff, len);
2918 /* don't make the dst file partly checksummed */
2919 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2920 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
2925 ret = btrfs_cmp_data(src, loff, dst, dst_loff, len);
2927 ret = btrfs_clone(src, dst, loff, len, len, dst_loff);
2930 btrfs_double_unlock(src, loff, dst, dst_loff, len);
2935 #define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024)
2937 static long btrfs_ioctl_file_extent_same(struct file *file,
2938 struct btrfs_ioctl_same_args __user *argp)
2940 struct btrfs_ioctl_same_args *same;
2941 struct btrfs_ioctl_same_extent_info *info;
2942 struct inode *src = file_inode(file);
2948 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
2949 bool is_admin = capable(CAP_SYS_ADMIN);
2952 if (!(file->f_mode & FMODE_READ))
2955 ret = mnt_want_write_file(file);
2959 if (get_user(count, &argp->dest_count)) {
2964 size = offsetof(struct btrfs_ioctl_same_args __user, info[count]);
2966 same = memdup_user(argp, size);
2969 ret = PTR_ERR(same);
2973 off = same->logical_offset;
2977 * Limit the total length we will dedupe for each operation.
2978 * This is intended to bound the total time spent in this
2979 * ioctl to something sane.
2981 if (len > BTRFS_MAX_DEDUPE_LEN)
2982 len = BTRFS_MAX_DEDUPE_LEN;
2984 if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) {
2986 * Btrfs does not support blocksize < page_size. As a
2987 * result, btrfs_cmp_data() won't correctly handle
2988 * this situation without an update.
2995 if (S_ISDIR(src->i_mode))
2999 if (!S_ISREG(src->i_mode))
3002 /* pre-format output fields to sane values */
3003 for (i = 0; i < count; i++) {
3004 same->info[i].bytes_deduped = 0ULL;
3005 same->info[i].status = 0;
3008 for (i = 0, info = same->info; i < count; i++, info++) {
3010 struct fd dst_file = fdget(info->fd);
3011 if (!dst_file.file) {
3012 info->status = -EBADF;
3015 dst = file_inode(dst_file.file);
3017 if (!(is_admin || (dst_file.file->f_mode & FMODE_WRITE))) {
3018 info->status = -EINVAL;
3019 } else if (file->f_path.mnt != dst_file.file->f_path.mnt) {
3020 info->status = -EXDEV;
3021 } else if (S_ISDIR(dst->i_mode)) {
3022 info->status = -EISDIR;
3023 } else if (!S_ISREG(dst->i_mode)) {
3024 info->status = -EACCES;
3026 info->status = btrfs_extent_same(src, off, len, dst,
3027 info->logical_offset);
3028 if (info->status == 0)
3029 info->bytes_deduped += len;
3034 ret = copy_to_user(argp, same, size);
3039 mnt_drop_write_file(file);
3043 /* Helper to check and see if this root currently has a ref on the given disk
3044 * bytenr. If it does then we need to update the quota for this root. This
3045 * doesn't do anything if quotas aren't enabled.
3047 static int check_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
3050 struct seq_list tree_mod_seq_elem = {};
3051 struct ulist *roots;
3052 struct ulist_iterator uiter;
3053 struct ulist_node *root_node = NULL;
3056 if (!root->fs_info->quota_enabled)
3059 btrfs_get_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3060 ret = btrfs_find_all_roots(trans, root->fs_info, disko,
3061 tree_mod_seq_elem.seq, &roots);
3065 ULIST_ITER_INIT(&uiter);
3066 while ((root_node = ulist_next(roots, &uiter))) {
3067 if (root_node->val == root->objectid) {
3074 btrfs_put_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3078 static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3079 struct inode *inode,
3084 struct btrfs_root *root = BTRFS_I(inode)->root;
3087 inode_inc_iversion(inode);
3088 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3090 * We round up to the block size at eof when determining which
3091 * extents to clone above, but shouldn't round up the file size.
3093 if (endoff > destoff + olen)
3094 endoff = destoff + olen;
3095 if (endoff > inode->i_size)
3096 btrfs_i_size_write(inode, endoff);
3098 ret = btrfs_update_inode(trans, root, inode);
3100 btrfs_abort_transaction(trans, root, ret);
3101 btrfs_end_transaction(trans, root);
3104 ret = btrfs_end_transaction(trans, root);
3109 static void clone_update_extent_map(struct inode *inode,
3110 const struct btrfs_trans_handle *trans,
3111 const struct btrfs_path *path,
3112 const u64 hole_offset,
3115 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
3116 struct extent_map *em;
3119 em = alloc_extent_map();
3121 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3122 &BTRFS_I(inode)->runtime_flags);
3127 struct btrfs_file_extent_item *fi;
3129 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3130 struct btrfs_file_extent_item);
3131 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3132 em->generation = -1;
3133 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3134 BTRFS_FILE_EXTENT_INLINE)
3135 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3136 &BTRFS_I(inode)->runtime_flags);
3138 em->start = hole_offset;
3140 em->ram_bytes = em->len;
3141 em->orig_start = hole_offset;
3142 em->block_start = EXTENT_MAP_HOLE;
3144 em->orig_block_len = 0;
3145 em->compress_type = BTRFS_COMPRESS_NONE;
3146 em->generation = trans->transid;
3150 write_lock(&em_tree->lock);
3151 ret = add_extent_mapping(em_tree, em, 1);
3152 write_unlock(&em_tree->lock);
3153 if (ret != -EEXIST) {
3154 free_extent_map(em);
3157 btrfs_drop_extent_cache(inode, em->start,
3158 em->start + em->len - 1, 0);
3162 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3163 &BTRFS_I(inode)->runtime_flags);
3167 * btrfs_clone() - clone a range from inode file to another
3169 * @src: Inode to clone from
3170 * @inode: Inode to clone to
3171 * @off: Offset within source to start clone from
3172 * @olen: Original length, passed by user, of range to clone
3173 * @olen_aligned: Block-aligned value of olen, extent_same uses
3174 * identical values here
3175 * @destoff: Offset within @inode to start clone
3177 static int btrfs_clone(struct inode *src, struct inode *inode,
3178 const u64 off, const u64 olen, const u64 olen_aligned,
3181 struct btrfs_root *root = BTRFS_I(inode)->root;
3182 struct btrfs_path *path = NULL;
3183 struct extent_buffer *leaf;
3184 struct btrfs_trans_handle *trans;
3186 struct btrfs_key key;
3191 const u64 len = olen_aligned;
3193 u64 last_dest_end = destoff;
3196 buf = vmalloc(btrfs_level_size(root, 0));
3200 path = btrfs_alloc_path();
3208 key.objectid = btrfs_ino(src);
3209 key.type = BTRFS_EXTENT_DATA_KEY;
3214 * note the key will change type as we walk through the
3217 path->leave_spinning = 1;
3218 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3223 * First search, if no extent item that starts at offset off was
3224 * found but the previous item is an extent item, it's possible
3225 * it might overlap our target range, therefore process it.
3227 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3228 btrfs_item_key_to_cpu(path->nodes[0], &key,
3229 path->slots[0] - 1);
3230 if (key.type == BTRFS_EXTENT_DATA_KEY)
3234 nritems = btrfs_header_nritems(path->nodes[0]);
3237 if (path->slots[0] >= nritems) {
3238 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
3243 nritems = btrfs_header_nritems(path->nodes[0]);
3245 leaf = path->nodes[0];
3246 slot = path->slots[0];
3248 btrfs_item_key_to_cpu(leaf, &key, slot);
3249 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
3250 key.objectid != btrfs_ino(src))
3253 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
3254 struct btrfs_file_extent_item *extent;
3257 struct btrfs_key new_key;
3258 u64 disko = 0, diskl = 0;
3259 u64 datao = 0, datal = 0;
3263 extent = btrfs_item_ptr(leaf, slot,
3264 struct btrfs_file_extent_item);
3265 comp = btrfs_file_extent_compression(leaf, extent);
3266 type = btrfs_file_extent_type(leaf, extent);
3267 if (type == BTRFS_FILE_EXTENT_REG ||
3268 type == BTRFS_FILE_EXTENT_PREALLOC) {
3269 disko = btrfs_file_extent_disk_bytenr(leaf,
3271 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3273 datao = btrfs_file_extent_offset(leaf, extent);
3274 datal = btrfs_file_extent_num_bytes(leaf,
3276 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3277 /* take upper bound, may be compressed */
3278 datal = btrfs_file_extent_ram_bytes(leaf,
3283 * The first search might have left us at an extent
3284 * item that ends before our target range's start, can
3285 * happen if we have holes and NO_HOLES feature enabled.
3287 if (key.offset + datal <= off) {
3290 } else if (key.offset >= off + len) {
3294 size = btrfs_item_size_nr(leaf, slot);
3295 read_extent_buffer(leaf, buf,
3296 btrfs_item_ptr_offset(leaf, slot),
3299 btrfs_release_path(path);
3300 path->leave_spinning = 0;
3302 memcpy(&new_key, &key, sizeof(new_key));
3303 new_key.objectid = btrfs_ino(inode);
3304 if (off <= key.offset)
3305 new_key.offset = key.offset + destoff - off;
3307 new_key.offset = destoff;
3310 * Deal with a hole that doesn't have an extent item
3311 * that represents it (NO_HOLES feature enabled).
3312 * This hole is either in the middle of the cloning
3313 * range or at the beginning (fully overlaps it or
3314 * partially overlaps it).
3316 if (new_key.offset != last_dest_end)
3317 drop_start = last_dest_end;
3319 drop_start = new_key.offset;
3322 * 1 - adjusting old extent (we may have to split it)
3323 * 1 - add new extent
3326 trans = btrfs_start_transaction(root, 3);
3327 if (IS_ERR(trans)) {
3328 ret = PTR_ERR(trans);
3332 if (type == BTRFS_FILE_EXTENT_REG ||
3333 type == BTRFS_FILE_EXTENT_PREALLOC) {
3335 * a | --- range to clone ---| b
3336 * | ------------- extent ------------- |
3339 /* subtract range b */
3340 if (key.offset + datal > off + len)
3341 datal = off + len - key.offset;
3343 /* subtract range a */
3344 if (off > key.offset) {
3345 datao += off - key.offset;
3346 datal -= off - key.offset;
3349 ret = btrfs_drop_extents(trans, root, inode,
3351 new_key.offset + datal,
3354 if (ret != -EOPNOTSUPP)
3355 btrfs_abort_transaction(trans,
3357 btrfs_end_transaction(trans, root);
3361 ret = btrfs_insert_empty_item(trans, root, path,
3364 btrfs_abort_transaction(trans, root,
3366 btrfs_end_transaction(trans, root);
3370 leaf = path->nodes[0];
3371 slot = path->slots[0];
3372 write_extent_buffer(leaf, buf,
3373 btrfs_item_ptr_offset(leaf, slot),
3376 extent = btrfs_item_ptr(leaf, slot,
3377 struct btrfs_file_extent_item);
3379 /* disko == 0 means it's a hole */
3383 btrfs_set_file_extent_offset(leaf, extent,
3385 btrfs_set_file_extent_num_bytes(leaf, extent,
3389 * We need to look up the roots that point at
3390 * this bytenr and see if the new root does. If
3391 * it does not we need to make sure we update
3392 * quotas appropriately.
3394 if (disko && root != BTRFS_I(src)->root &&
3395 disko != last_disko) {
3396 no_quota = check_ref(trans, root,
3399 btrfs_abort_transaction(trans,
3402 btrfs_end_transaction(trans,
3410 inode_add_bytes(inode, datal);
3411 ret = btrfs_inc_extent_ref(trans, root,
3413 root->root_key.objectid,
3415 new_key.offset - datao,
3418 btrfs_abort_transaction(trans,
3421 btrfs_end_transaction(trans,
3427 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3430 u64 aligned_end = 0;
3432 if (off > key.offset) {
3433 skip = off - key.offset;
3434 new_key.offset += skip;
3437 if (key.offset + datal > off + len)
3438 trim = key.offset + datal - (off + len);
3440 if (comp && (skip || trim)) {
3442 btrfs_end_transaction(trans, root);
3445 size -= skip + trim;
3446 datal -= skip + trim;
3448 aligned_end = ALIGN(new_key.offset + datal,
3450 ret = btrfs_drop_extents(trans, root, inode,
3455 if (ret != -EOPNOTSUPP)
3456 btrfs_abort_transaction(trans,
3458 btrfs_end_transaction(trans, root);
3462 ret = btrfs_insert_empty_item(trans, root, path,
3465 btrfs_abort_transaction(trans, root,
3467 btrfs_end_transaction(trans, root);
3473 btrfs_file_extent_calc_inline_size(0);
3474 memmove(buf+start, buf+start+skip,
3478 leaf = path->nodes[0];
3479 slot = path->slots[0];
3480 write_extent_buffer(leaf, buf,
3481 btrfs_item_ptr_offset(leaf, slot),
3483 inode_add_bytes(inode, datal);
3486 /* If we have an implicit hole (NO_HOLES feature). */
3487 if (drop_start < new_key.offset)
3488 clone_update_extent_map(inode, trans,
3490 new_key.offset - drop_start);
3492 clone_update_extent_map(inode, trans, path, 0, 0);
3494 btrfs_mark_buffer_dirty(leaf);
3495 btrfs_release_path(path);
3497 last_dest_end = new_key.offset + datal;
3498 ret = clone_finish_inode_update(trans, inode,
3503 if (new_key.offset + datal >= destoff + len)
3506 btrfs_release_path(path);
3511 if (last_dest_end < destoff + len) {
3513 * We have an implicit hole (NO_HOLES feature is enabled) that
3514 * fully or partially overlaps our cloning range at its end.
3516 btrfs_release_path(path);
3519 * 1 - remove extent(s)
3522 trans = btrfs_start_transaction(root, 2);
3523 if (IS_ERR(trans)) {
3524 ret = PTR_ERR(trans);
3527 ret = btrfs_drop_extents(trans, root, inode,
3528 last_dest_end, destoff + len, 1);
3530 if (ret != -EOPNOTSUPP)
3531 btrfs_abort_transaction(trans, root, ret);
3532 btrfs_end_transaction(trans, root);
3535 clone_update_extent_map(inode, trans, NULL, last_dest_end,
3536 destoff + len - last_dest_end);
3537 ret = clone_finish_inode_update(trans, inode, destoff + len,
3542 btrfs_free_path(path);
3547 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
3548 u64 off, u64 olen, u64 destoff)
3550 struct inode *inode = file_inode(file);
3551 struct btrfs_root *root = BTRFS_I(inode)->root;
3556 u64 bs = root->fs_info->sb->s_blocksize;
3561 * - split compressed inline extents. annoying: we need to
3562 * decompress into destination's address_space (the file offset
3563 * may change, so source mapping won't do), then recompress (or
3564 * otherwise reinsert) a subrange.
3566 * - split destination inode's inline extents. The inline extents can
3567 * be either compressed or non-compressed.
3570 /* the destination must be opened for writing */
3571 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
3574 if (btrfs_root_readonly(root))
3577 ret = mnt_want_write_file(file);
3581 src_file = fdget(srcfd);
3582 if (!src_file.file) {
3584 goto out_drop_write;
3588 if (src_file.file->f_path.mnt != file->f_path.mnt)
3591 src = file_inode(src_file.file);
3597 /* the src must be open for reading */
3598 if (!(src_file.file->f_mode & FMODE_READ))
3601 /* don't make the dst file partly checksummed */
3602 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3603 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3607 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3611 if (src->i_sb != inode->i_sb)
3616 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
3617 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
3619 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
3620 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
3623 mutex_lock(&src->i_mutex);
3626 /* determine range to clone */
3628 if (off + len > src->i_size || off + len < off)
3631 olen = len = src->i_size - off;
3632 /* if we extend to eof, continue to block boundary */
3633 if (off + len == src->i_size)
3634 len = ALIGN(src->i_size, bs) - off;
3636 /* verify the end result is block aligned */
3637 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3638 !IS_ALIGNED(destoff, bs))
3641 /* verify if ranges are overlapped within the same file */
3643 if (destoff + len > off && destoff < off + len)
3647 if (destoff > inode->i_size) {
3648 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3654 * Lock the target range too. Right after we replace the file extent
3655 * items in the fs tree (which now point to the cloned data), we might
3656 * have a worker replace them with extent items relative to a write
3657 * operation that was issued before this clone operation (i.e. confront
3658 * with inode.c:btrfs_finish_ordered_io).
3661 u64 lock_start = min_t(u64, off, destoff);
3662 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
3664 lock_extent_range(src, lock_start, lock_len);
3666 lock_extent_range(src, off, len);
3667 lock_extent_range(inode, destoff, len);
3670 ret = btrfs_clone(src, inode, off, olen, len, destoff);
3673 u64 lock_start = min_t(u64, off, destoff);
3674 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3676 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3678 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
3679 unlock_extent(&BTRFS_I(inode)->io_tree, destoff,
3683 * Truncate page cache pages so that future reads will see the cloned
3684 * data immediately and not the previous data.
3686 truncate_inode_pages_range(&inode->i_data, destoff,
3687 PAGE_CACHE_ALIGN(destoff + len) - 1);
3691 mutex_unlock(&src->i_mutex);
3692 mutex_unlock(&inode->i_mutex);
3694 mutex_unlock(&inode->i_mutex);
3695 mutex_unlock(&src->i_mutex);
3698 mutex_unlock(&src->i_mutex);
3703 mnt_drop_write_file(file);
3707 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
3709 struct btrfs_ioctl_clone_range_args args;
3711 if (copy_from_user(&args, argp, sizeof(args)))
3713 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
3714 args.src_length, args.dest_offset);
3718 * there are many ways the trans_start and trans_end ioctls can lead
3719 * to deadlocks. They should only be used by applications that
3720 * basically own the machine, and have a very in depth understanding
3721 * of all the possible deadlocks and enospc problems.
3723 static long btrfs_ioctl_trans_start(struct file *file)
3725 struct inode *inode = file_inode(file);
3726 struct btrfs_root *root = BTRFS_I(inode)->root;
3727 struct btrfs_trans_handle *trans;
3731 if (!capable(CAP_SYS_ADMIN))
3735 if (file->private_data)
3739 if (btrfs_root_readonly(root))
3742 ret = mnt_want_write_file(file);
3746 atomic_inc(&root->fs_info->open_ioctl_trans);
3749 trans = btrfs_start_ioctl_transaction(root);
3753 file->private_data = trans;
3757 atomic_dec(&root->fs_info->open_ioctl_trans);
3758 mnt_drop_write_file(file);
3763 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3765 struct inode *inode = file_inode(file);
3766 struct btrfs_root *root = BTRFS_I(inode)->root;
3767 struct btrfs_root *new_root;
3768 struct btrfs_dir_item *di;
3769 struct btrfs_trans_handle *trans;
3770 struct btrfs_path *path;
3771 struct btrfs_key location;
3772 struct btrfs_disk_key disk_key;
3777 if (!capable(CAP_SYS_ADMIN))
3780 ret = mnt_want_write_file(file);
3784 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3790 objectid = BTRFS_FS_TREE_OBJECTID;
3792 location.objectid = objectid;
3793 location.type = BTRFS_ROOT_ITEM_KEY;
3794 location.offset = (u64)-1;
3796 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
3797 if (IS_ERR(new_root)) {
3798 ret = PTR_ERR(new_root);
3802 path = btrfs_alloc_path();
3807 path->leave_spinning = 1;
3809 trans = btrfs_start_transaction(root, 1);
3810 if (IS_ERR(trans)) {
3811 btrfs_free_path(path);
3812 ret = PTR_ERR(trans);
3816 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
3817 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
3818 dir_id, "default", 7, 1);
3819 if (IS_ERR_OR_NULL(di)) {
3820 btrfs_free_path(path);
3821 btrfs_end_transaction(trans, root);
3822 btrfs_err(new_root->fs_info, "Umm, you don't have the default dir"
3823 "item, this isn't going to work");
3828 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3829 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3830 btrfs_mark_buffer_dirty(path->nodes[0]);
3831 btrfs_free_path(path);
3833 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
3834 btrfs_end_transaction(trans, root);
3836 mnt_drop_write_file(file);
3840 void btrfs_get_block_group_info(struct list_head *groups_list,
3841 struct btrfs_ioctl_space_info *space)
3843 struct btrfs_block_group_cache *block_group;
3845 space->total_bytes = 0;
3846 space->used_bytes = 0;
3848 list_for_each_entry(block_group, groups_list, list) {
3849 space->flags = block_group->flags;
3850 space->total_bytes += block_group->key.offset;
3851 space->used_bytes +=
3852 btrfs_block_group_used(&block_group->item);
3856 static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
3858 struct btrfs_ioctl_space_args space_args;
3859 struct btrfs_ioctl_space_info space;
3860 struct btrfs_ioctl_space_info *dest;
3861 struct btrfs_ioctl_space_info *dest_orig;
3862 struct btrfs_ioctl_space_info __user *user_dest;
3863 struct btrfs_space_info *info;
3864 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3865 BTRFS_BLOCK_GROUP_SYSTEM,
3866 BTRFS_BLOCK_GROUP_METADATA,
3867 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3874 if (copy_from_user(&space_args,
3875 (struct btrfs_ioctl_space_args __user *)arg,
3876 sizeof(space_args)))
3879 for (i = 0; i < num_types; i++) {
3880 struct btrfs_space_info *tmp;
3884 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3886 if (tmp->flags == types[i]) {
3896 down_read(&info->groups_sem);
3897 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3898 if (!list_empty(&info->block_groups[c]))
3901 up_read(&info->groups_sem);
3905 * Global block reserve, exported as a space_info
3909 /* space_slots == 0 means they are asking for a count */
3910 if (space_args.space_slots == 0) {
3911 space_args.total_spaces = slot_count;
3915 slot_count = min_t(u64, space_args.space_slots, slot_count);
3917 alloc_size = sizeof(*dest) * slot_count;
3919 /* we generally have at most 6 or so space infos, one for each raid
3920 * level. So, a whole page should be more than enough for everyone
3922 if (alloc_size > PAGE_CACHE_SIZE)
3925 space_args.total_spaces = 0;
3926 dest = kmalloc(alloc_size, GFP_NOFS);
3931 /* now we have a buffer to copy into */
3932 for (i = 0; i < num_types; i++) {
3933 struct btrfs_space_info *tmp;
3940 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3942 if (tmp->flags == types[i]) {
3951 down_read(&info->groups_sem);
3952 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3953 if (!list_empty(&info->block_groups[c])) {
3954 btrfs_get_block_group_info(
3955 &info->block_groups[c], &space);
3956 memcpy(dest, &space, sizeof(space));
3958 space_args.total_spaces++;
3964 up_read(&info->groups_sem);
3968 * Add global block reserve
3971 struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv;
3973 spin_lock(&block_rsv->lock);
3974 space.total_bytes = block_rsv->size;
3975 space.used_bytes = block_rsv->size - block_rsv->reserved;
3976 spin_unlock(&block_rsv->lock);
3977 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3978 memcpy(dest, &space, sizeof(space));
3979 space_args.total_spaces++;
3982 user_dest = (struct btrfs_ioctl_space_info __user *)
3983 (arg + sizeof(struct btrfs_ioctl_space_args));
3985 if (copy_to_user(user_dest, dest_orig, alloc_size))
3990 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3997 * there are many ways the trans_start and trans_end ioctls can lead
3998 * to deadlocks. They should only be used by applications that
3999 * basically own the machine, and have a very in depth understanding
4000 * of all the possible deadlocks and enospc problems.
4002 long btrfs_ioctl_trans_end(struct file *file)
4004 struct inode *inode = file_inode(file);
4005 struct btrfs_root *root = BTRFS_I(inode)->root;
4006 struct btrfs_trans_handle *trans;
4008 trans = file->private_data;
4011 file->private_data = NULL;
4013 btrfs_end_transaction(trans, root);
4015 atomic_dec(&root->fs_info->open_ioctl_trans);
4017 mnt_drop_write_file(file);
4021 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4024 struct btrfs_trans_handle *trans;
4028 trans = btrfs_attach_transaction_barrier(root);
4029 if (IS_ERR(trans)) {
4030 if (PTR_ERR(trans) != -ENOENT)
4031 return PTR_ERR(trans);
4033 /* No running transaction, don't bother */
4034 transid = root->fs_info->last_trans_committed;
4037 transid = trans->transid;
4038 ret = btrfs_commit_transaction_async(trans, root, 0);
4040 btrfs_end_transaction(trans, root);
4045 if (copy_to_user(argp, &transid, sizeof(transid)))
4050 static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
4056 if (copy_from_user(&transid, argp, sizeof(transid)))
4059 transid = 0; /* current trans */
4061 return btrfs_wait_for_commit(root, transid);
4064 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
4066 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4067 struct btrfs_ioctl_scrub_args *sa;
4070 if (!capable(CAP_SYS_ADMIN))
4073 sa = memdup_user(arg, sizeof(*sa));
4077 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4078 ret = mnt_want_write_file(file);
4083 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
4084 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4087 if (copy_to_user(arg, sa, sizeof(*sa)))
4090 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4091 mnt_drop_write_file(file);
4097 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
4099 if (!capable(CAP_SYS_ADMIN))
4102 return btrfs_scrub_cancel(root->fs_info);
4105 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
4108 struct btrfs_ioctl_scrub_args *sa;
4111 if (!capable(CAP_SYS_ADMIN))
4114 sa = memdup_user(arg, sizeof(*sa));
4118 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
4120 if (copy_to_user(arg, sa, sizeof(*sa)))
4127 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
4130 struct btrfs_ioctl_get_dev_stats *sa;
4133 sa = memdup_user(arg, sizeof(*sa));
4137 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4142 ret = btrfs_get_dev_stats(root, sa);
4144 if (copy_to_user(arg, sa, sizeof(*sa)))
4151 static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
4153 struct btrfs_ioctl_dev_replace_args *p;
4156 if (!capable(CAP_SYS_ADMIN))
4159 p = memdup_user(arg, sizeof(*p));
4164 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
4165 if (root->fs_info->sb->s_flags & MS_RDONLY) {
4170 &root->fs_info->mutually_exclusive_operation_running,
4172 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4174 ret = btrfs_dev_replace_start(root, p);
4176 &root->fs_info->mutually_exclusive_operation_running,
4180 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4181 btrfs_dev_replace_status(root->fs_info, p);
4184 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4185 ret = btrfs_dev_replace_cancel(root->fs_info, p);
4192 if (copy_to_user(arg, p, sizeof(*p)))
4199 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4205 struct btrfs_ioctl_ino_path_args *ipa = NULL;
4206 struct inode_fs_paths *ipath = NULL;
4207 struct btrfs_path *path;
4209 if (!capable(CAP_DAC_READ_SEARCH))
4212 path = btrfs_alloc_path();
4218 ipa = memdup_user(arg, sizeof(*ipa));
4225 size = min_t(u32, ipa->size, 4096);
4226 ipath = init_ipath(size, root, path);
4227 if (IS_ERR(ipath)) {
4228 ret = PTR_ERR(ipath);
4233 ret = paths_from_inode(ipa->inum, ipath);
4237 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
4238 rel_ptr = ipath->fspath->val[i] -
4239 (u64)(unsigned long)ipath->fspath->val;
4240 ipath->fspath->val[i] = rel_ptr;
4243 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
4244 (void *)(unsigned long)ipath->fspath, size);
4251 btrfs_free_path(path);
4258 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4260 struct btrfs_data_container *inodes = ctx;
4261 const size_t c = 3 * sizeof(u64);
4263 if (inodes->bytes_left >= c) {
4264 inodes->bytes_left -= c;
4265 inodes->val[inodes->elem_cnt] = inum;
4266 inodes->val[inodes->elem_cnt + 1] = offset;
4267 inodes->val[inodes->elem_cnt + 2] = root;
4268 inodes->elem_cnt += 3;
4270 inodes->bytes_missing += c - inodes->bytes_left;
4271 inodes->bytes_left = 0;
4272 inodes->elem_missed += 3;
4278 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
4283 struct btrfs_ioctl_logical_ino_args *loi;
4284 struct btrfs_data_container *inodes = NULL;
4285 struct btrfs_path *path = NULL;
4287 if (!capable(CAP_SYS_ADMIN))
4290 loi = memdup_user(arg, sizeof(*loi));
4297 path = btrfs_alloc_path();
4303 size = min_t(u32, loi->size, 64 * 1024);
4304 inodes = init_data_container(size);
4305 if (IS_ERR(inodes)) {
4306 ret = PTR_ERR(inodes);
4311 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
4312 build_ino_list, inodes);
4318 ret = copy_to_user((void *)(unsigned long)loi->inodes,
4319 (void *)(unsigned long)inodes, size);
4324 btrfs_free_path(path);
4331 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
4332 struct btrfs_ioctl_balance_args *bargs)
4334 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4336 bargs->flags = bctl->flags;
4338 if (atomic_read(&fs_info->balance_running))
4339 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4340 if (atomic_read(&fs_info->balance_pause_req))
4341 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
4342 if (atomic_read(&fs_info->balance_cancel_req))
4343 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
4345 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4346 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4347 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4350 spin_lock(&fs_info->balance_lock);
4351 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4352 spin_unlock(&fs_info->balance_lock);
4354 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4358 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4360 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4361 struct btrfs_fs_info *fs_info = root->fs_info;
4362 struct btrfs_ioctl_balance_args *bargs;
4363 struct btrfs_balance_control *bctl;
4364 bool need_unlock; /* for mut. excl. ops lock */
4367 if (!capable(CAP_SYS_ADMIN))
4370 ret = mnt_want_write_file(file);
4375 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
4376 mutex_lock(&fs_info->volume_mutex);
4377 mutex_lock(&fs_info->balance_mutex);
4383 * mut. excl. ops lock is locked. Three possibilites:
4384 * (1) some other op is running
4385 * (2) balance is running
4386 * (3) balance is paused -- special case (think resume)
4388 mutex_lock(&fs_info->balance_mutex);
4389 if (fs_info->balance_ctl) {
4390 /* this is either (2) or (3) */
4391 if (!atomic_read(&fs_info->balance_running)) {
4392 mutex_unlock(&fs_info->balance_mutex);
4393 if (!mutex_trylock(&fs_info->volume_mutex))
4395 mutex_lock(&fs_info->balance_mutex);
4397 if (fs_info->balance_ctl &&
4398 !atomic_read(&fs_info->balance_running)) {
4400 need_unlock = false;
4404 mutex_unlock(&fs_info->balance_mutex);
4405 mutex_unlock(&fs_info->volume_mutex);
4409 mutex_unlock(&fs_info->balance_mutex);
4415 mutex_unlock(&fs_info->balance_mutex);
4416 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4421 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
4424 bargs = memdup_user(arg, sizeof(*bargs));
4425 if (IS_ERR(bargs)) {
4426 ret = PTR_ERR(bargs);
4430 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4431 if (!fs_info->balance_ctl) {
4436 bctl = fs_info->balance_ctl;
4437 spin_lock(&fs_info->balance_lock);
4438 bctl->flags |= BTRFS_BALANCE_RESUME;
4439 spin_unlock(&fs_info->balance_lock);
4447 if (fs_info->balance_ctl) {
4452 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
4458 bctl->fs_info = fs_info;
4460 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4461 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4462 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4464 bctl->flags = bargs->flags;
4466 /* balance everything - no filters */
4467 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4472 * Ownership of bctl and mutually_exclusive_operation_running
4473 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4474 * or, if restriper was paused all the way until unmount, in
4475 * free_fs_info. mutually_exclusive_operation_running is
4476 * cleared in __cancel_balance.
4478 need_unlock = false;
4480 ret = btrfs_balance(bctl, bargs);
4483 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4490 mutex_unlock(&fs_info->balance_mutex);
4491 mutex_unlock(&fs_info->volume_mutex);
4493 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4495 mnt_drop_write_file(file);
4499 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4501 if (!capable(CAP_SYS_ADMIN))
4505 case BTRFS_BALANCE_CTL_PAUSE:
4506 return btrfs_pause_balance(root->fs_info);
4507 case BTRFS_BALANCE_CTL_CANCEL:
4508 return btrfs_cancel_balance(root->fs_info);
4514 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4517 struct btrfs_fs_info *fs_info = root->fs_info;
4518 struct btrfs_ioctl_balance_args *bargs;
4521 if (!capable(CAP_SYS_ADMIN))
4524 mutex_lock(&fs_info->balance_mutex);
4525 if (!fs_info->balance_ctl) {
4530 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
4536 update_ioctl_balance_args(fs_info, 1, bargs);
4538 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4543 mutex_unlock(&fs_info->balance_mutex);
4547 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4549 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4550 struct btrfs_ioctl_quota_ctl_args *sa;
4551 struct btrfs_trans_handle *trans = NULL;
4555 if (!capable(CAP_SYS_ADMIN))
4558 ret = mnt_want_write_file(file);
4562 sa = memdup_user(arg, sizeof(*sa));
4568 down_write(&root->fs_info->subvol_sem);
4569 trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4570 if (IS_ERR(trans)) {
4571 ret = PTR_ERR(trans);
4576 case BTRFS_QUOTA_CTL_ENABLE:
4577 ret = btrfs_quota_enable(trans, root->fs_info);
4579 case BTRFS_QUOTA_CTL_DISABLE:
4580 ret = btrfs_quota_disable(trans, root->fs_info);
4587 err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4592 up_write(&root->fs_info->subvol_sem);
4594 mnt_drop_write_file(file);
4598 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4600 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4601 struct btrfs_ioctl_qgroup_assign_args *sa;
4602 struct btrfs_trans_handle *trans;
4606 if (!capable(CAP_SYS_ADMIN))
4609 ret = mnt_want_write_file(file);
4613 sa = memdup_user(arg, sizeof(*sa));
4619 trans = btrfs_join_transaction(root);
4620 if (IS_ERR(trans)) {
4621 ret = PTR_ERR(trans);
4625 /* FIXME: check if the IDs really exist */
4627 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4630 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4634 err = btrfs_end_transaction(trans, root);
4641 mnt_drop_write_file(file);
4645 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4647 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4648 struct btrfs_ioctl_qgroup_create_args *sa;
4649 struct btrfs_trans_handle *trans;
4653 if (!capable(CAP_SYS_ADMIN))
4656 ret = mnt_want_write_file(file);
4660 sa = memdup_user(arg, sizeof(*sa));
4666 if (!sa->qgroupid) {
4671 trans = btrfs_join_transaction(root);
4672 if (IS_ERR(trans)) {
4673 ret = PTR_ERR(trans);
4677 /* FIXME: check if the IDs really exist */
4679 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
4682 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4685 err = btrfs_end_transaction(trans, root);
4692 mnt_drop_write_file(file);
4696 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4698 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4699 struct btrfs_ioctl_qgroup_limit_args *sa;
4700 struct btrfs_trans_handle *trans;
4705 if (!capable(CAP_SYS_ADMIN))
4708 ret = mnt_want_write_file(file);
4712 sa = memdup_user(arg, sizeof(*sa));
4718 trans = btrfs_join_transaction(root);
4719 if (IS_ERR(trans)) {
4720 ret = PTR_ERR(trans);
4724 qgroupid = sa->qgroupid;
4726 /* take the current subvol as qgroup */
4727 qgroupid = root->root_key.objectid;
4730 /* FIXME: check if the IDs really exist */
4731 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
4733 err = btrfs_end_transaction(trans, root);
4740 mnt_drop_write_file(file);
4744 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4746 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4747 struct btrfs_ioctl_quota_rescan_args *qsa;
4750 if (!capable(CAP_SYS_ADMIN))
4753 ret = mnt_want_write_file(file);
4757 qsa = memdup_user(arg, sizeof(*qsa));
4768 ret = btrfs_qgroup_rescan(root->fs_info);
4773 mnt_drop_write_file(file);
4777 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
4779 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4780 struct btrfs_ioctl_quota_rescan_args *qsa;
4783 if (!capable(CAP_SYS_ADMIN))
4786 qsa = kzalloc(sizeof(*qsa), GFP_NOFS);
4790 if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4792 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
4795 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4802 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
4804 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4806 if (!capable(CAP_SYS_ADMIN))
4809 return btrfs_qgroup_wait_for_completion(root->fs_info);
4812 static long _btrfs_ioctl_set_received_subvol(struct file *file,
4813 struct btrfs_ioctl_received_subvol_args *sa)
4815 struct inode *inode = file_inode(file);
4816 struct btrfs_root *root = BTRFS_I(inode)->root;
4817 struct btrfs_root_item *root_item = &root->root_item;
4818 struct btrfs_trans_handle *trans;
4819 struct timespec ct = CURRENT_TIME;
4821 int received_uuid_changed;
4823 if (!inode_owner_or_capable(inode))
4826 ret = mnt_want_write_file(file);
4830 down_write(&root->fs_info->subvol_sem);
4832 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
4837 if (btrfs_root_readonly(root)) {
4844 * 2 - uuid items (received uuid + subvol uuid)
4846 trans = btrfs_start_transaction(root, 3);
4847 if (IS_ERR(trans)) {
4848 ret = PTR_ERR(trans);
4853 sa->rtransid = trans->transid;
4854 sa->rtime.sec = ct.tv_sec;
4855 sa->rtime.nsec = ct.tv_nsec;
4857 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4859 if (received_uuid_changed &&
4860 !btrfs_is_empty_uuid(root_item->received_uuid))
4861 btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
4862 root_item->received_uuid,
4863 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4864 root->root_key.objectid);
4865 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4866 btrfs_set_root_stransid(root_item, sa->stransid);
4867 btrfs_set_root_rtransid(root_item, sa->rtransid);
4868 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4869 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4870 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4871 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4873 ret = btrfs_update_root(trans, root->fs_info->tree_root,
4874 &root->root_key, &root->root_item);
4876 btrfs_end_transaction(trans, root);
4879 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4880 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
4882 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4883 root->root_key.objectid);
4884 if (ret < 0 && ret != -EEXIST) {
4885 btrfs_abort_transaction(trans, root, ret);
4889 ret = btrfs_commit_transaction(trans, root);
4891 btrfs_abort_transaction(trans, root, ret);
4896 up_write(&root->fs_info->subvol_sem);
4897 mnt_drop_write_file(file);
4902 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4905 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4906 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4909 args32 = memdup_user(arg, sizeof(*args32));
4910 if (IS_ERR(args32)) {
4911 ret = PTR_ERR(args32);
4916 args64 = kmalloc(sizeof(*args64), GFP_NOFS);
4922 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4923 args64->stransid = args32->stransid;
4924 args64->rtransid = args32->rtransid;
4925 args64->stime.sec = args32->stime.sec;
4926 args64->stime.nsec = args32->stime.nsec;
4927 args64->rtime.sec = args32->rtime.sec;
4928 args64->rtime.nsec = args32->rtime.nsec;
4929 args64->flags = args32->flags;
4931 ret = _btrfs_ioctl_set_received_subvol(file, args64);
4935 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4936 args32->stransid = args64->stransid;
4937 args32->rtransid = args64->rtransid;
4938 args32->stime.sec = args64->stime.sec;
4939 args32->stime.nsec = args64->stime.nsec;
4940 args32->rtime.sec = args64->rtime.sec;
4941 args32->rtime.nsec = args64->rtime.nsec;
4942 args32->flags = args64->flags;
4944 ret = copy_to_user(arg, args32, sizeof(*args32));
4955 static long btrfs_ioctl_set_received_subvol(struct file *file,
4958 struct btrfs_ioctl_received_subvol_args *sa = NULL;
4961 sa = memdup_user(arg, sizeof(*sa));
4968 ret = _btrfs_ioctl_set_received_subvol(file, sa);
4973 ret = copy_to_user(arg, sa, sizeof(*sa));
4982 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
4984 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4987 char label[BTRFS_LABEL_SIZE];
4989 spin_lock(&root->fs_info->super_lock);
4990 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4991 spin_unlock(&root->fs_info->super_lock);
4993 len = strnlen(label, BTRFS_LABEL_SIZE);
4995 if (len == BTRFS_LABEL_SIZE) {
4996 btrfs_warn(root->fs_info,
4997 "label is too long, return the first %zu bytes", --len);
5000 ret = copy_to_user(arg, label, len);
5002 return ret ? -EFAULT : 0;
5005 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5007 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5008 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5009 struct btrfs_trans_handle *trans;
5010 char label[BTRFS_LABEL_SIZE];
5013 if (!capable(CAP_SYS_ADMIN))
5016 if (copy_from_user(label, arg, sizeof(label)))
5019 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
5020 btrfs_err(root->fs_info, "unable to set label with more than %d bytes",
5021 BTRFS_LABEL_SIZE - 1);
5025 ret = mnt_want_write_file(file);
5029 trans = btrfs_start_transaction(root, 0);
5030 if (IS_ERR(trans)) {
5031 ret = PTR_ERR(trans);
5035 spin_lock(&root->fs_info->super_lock);
5036 strcpy(super_block->label, label);
5037 spin_unlock(&root->fs_info->super_lock);
5038 ret = btrfs_commit_transaction(trans, root);
5041 mnt_drop_write_file(file);
5045 #define INIT_FEATURE_FLAGS(suffix) \
5046 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5047 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5048 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5050 static int btrfs_ioctl_get_supported_features(struct file *file,
5053 static struct btrfs_ioctl_feature_flags features[3] = {
5054 INIT_FEATURE_FLAGS(SUPP),
5055 INIT_FEATURE_FLAGS(SAFE_SET),
5056 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5059 if (copy_to_user(arg, &features, sizeof(features)))
5065 static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5067 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5068 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5069 struct btrfs_ioctl_feature_flags features;
5071 features.compat_flags = btrfs_super_compat_flags(super_block);
5072 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5073 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5075 if (copy_to_user(arg, &features, sizeof(features)))
5081 static int check_feature_bits(struct btrfs_root *root,
5082 enum btrfs_feature_set set,
5083 u64 change_mask, u64 flags, u64 supported_flags,
5084 u64 safe_set, u64 safe_clear)
5086 const char *type = btrfs_feature_set_names[set];
5088 u64 disallowed, unsupported;
5089 u64 set_mask = flags & change_mask;
5090 u64 clear_mask = ~flags & change_mask;
5092 unsupported = set_mask & ~supported_flags;
5094 names = btrfs_printable_features(set, unsupported);
5096 btrfs_warn(root->fs_info,
5097 "this kernel does not support the %s feature bit%s",
5098 names, strchr(names, ',') ? "s" : "");
5101 btrfs_warn(root->fs_info,
5102 "this kernel does not support %s bits 0x%llx",
5107 disallowed = set_mask & ~safe_set;
5109 names = btrfs_printable_features(set, disallowed);
5111 btrfs_warn(root->fs_info,
5112 "can't set the %s feature bit%s while mounted",
5113 names, strchr(names, ',') ? "s" : "");
5116 btrfs_warn(root->fs_info,
5117 "can't set %s bits 0x%llx while mounted",
5122 disallowed = clear_mask & ~safe_clear;
5124 names = btrfs_printable_features(set, disallowed);
5126 btrfs_warn(root->fs_info,
5127 "can't clear the %s feature bit%s while mounted",
5128 names, strchr(names, ',') ? "s" : "");
5131 btrfs_warn(root->fs_info,
5132 "can't clear %s bits 0x%llx while mounted",
5140 #define check_feature(root, change_mask, flags, mask_base) \
5141 check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
5142 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5143 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5144 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5146 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5148 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5149 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5150 struct btrfs_ioctl_feature_flags flags[2];
5151 struct btrfs_trans_handle *trans;
5155 if (!capable(CAP_SYS_ADMIN))
5158 if (copy_from_user(flags, arg, sizeof(flags)))
5162 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5163 !flags[0].incompat_flags)
5166 ret = check_feature(root, flags[0].compat_flags,
5167 flags[1].compat_flags, COMPAT);
5171 ret = check_feature(root, flags[0].compat_ro_flags,
5172 flags[1].compat_ro_flags, COMPAT_RO);
5176 ret = check_feature(root, flags[0].incompat_flags,
5177 flags[1].incompat_flags, INCOMPAT);
5181 trans = btrfs_start_transaction(root, 0);
5183 return PTR_ERR(trans);
5185 spin_lock(&root->fs_info->super_lock);
5186 newflags = btrfs_super_compat_flags(super_block);
5187 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5188 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5189 btrfs_set_super_compat_flags(super_block, newflags);
5191 newflags = btrfs_super_compat_ro_flags(super_block);
5192 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5193 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5194 btrfs_set_super_compat_ro_flags(super_block, newflags);
5196 newflags = btrfs_super_incompat_flags(super_block);
5197 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5198 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5199 btrfs_set_super_incompat_flags(super_block, newflags);
5200 spin_unlock(&root->fs_info->super_lock);
5202 return btrfs_commit_transaction(trans, root);
5205 long btrfs_ioctl(struct file *file, unsigned int
5206 cmd, unsigned long arg)
5208 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5209 void __user *argp = (void __user *)arg;
5212 case FS_IOC_GETFLAGS:
5213 return btrfs_ioctl_getflags(file, argp);
5214 case FS_IOC_SETFLAGS:
5215 return btrfs_ioctl_setflags(file, argp);
5216 case FS_IOC_GETVERSION:
5217 return btrfs_ioctl_getversion(file, argp);
5219 return btrfs_ioctl_fitrim(file, argp);
5220 case BTRFS_IOC_SNAP_CREATE:
5221 return btrfs_ioctl_snap_create(file, argp, 0);
5222 case BTRFS_IOC_SNAP_CREATE_V2:
5223 return btrfs_ioctl_snap_create_v2(file, argp, 0);
5224 case BTRFS_IOC_SUBVOL_CREATE:
5225 return btrfs_ioctl_snap_create(file, argp, 1);
5226 case BTRFS_IOC_SUBVOL_CREATE_V2:
5227 return btrfs_ioctl_snap_create_v2(file, argp, 1);
5228 case BTRFS_IOC_SNAP_DESTROY:
5229 return btrfs_ioctl_snap_destroy(file, argp);
5230 case BTRFS_IOC_SUBVOL_GETFLAGS:
5231 return btrfs_ioctl_subvol_getflags(file, argp);
5232 case BTRFS_IOC_SUBVOL_SETFLAGS:
5233 return btrfs_ioctl_subvol_setflags(file, argp);
5234 case BTRFS_IOC_DEFAULT_SUBVOL:
5235 return btrfs_ioctl_default_subvol(file, argp);
5236 case BTRFS_IOC_DEFRAG:
5237 return btrfs_ioctl_defrag(file, NULL);
5238 case BTRFS_IOC_DEFRAG_RANGE:
5239 return btrfs_ioctl_defrag(file, argp);
5240 case BTRFS_IOC_RESIZE:
5241 return btrfs_ioctl_resize(file, argp);
5242 case BTRFS_IOC_ADD_DEV:
5243 return btrfs_ioctl_add_dev(root, argp);
5244 case BTRFS_IOC_RM_DEV:
5245 return btrfs_ioctl_rm_dev(file, argp);
5246 case BTRFS_IOC_FS_INFO:
5247 return btrfs_ioctl_fs_info(root, argp);
5248 case BTRFS_IOC_DEV_INFO:
5249 return btrfs_ioctl_dev_info(root, argp);
5250 case BTRFS_IOC_BALANCE:
5251 return btrfs_ioctl_balance(file, NULL);
5252 case BTRFS_IOC_CLONE:
5253 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
5254 case BTRFS_IOC_CLONE_RANGE:
5255 return btrfs_ioctl_clone_range(file, argp);
5256 case BTRFS_IOC_TRANS_START:
5257 return btrfs_ioctl_trans_start(file);
5258 case BTRFS_IOC_TRANS_END:
5259 return btrfs_ioctl_trans_end(file);
5260 case BTRFS_IOC_TREE_SEARCH:
5261 return btrfs_ioctl_tree_search(file, argp);
5262 case BTRFS_IOC_TREE_SEARCH_V2:
5263 return btrfs_ioctl_tree_search_v2(file, argp);
5264 case BTRFS_IOC_INO_LOOKUP:
5265 return btrfs_ioctl_ino_lookup(file, argp);
5266 case BTRFS_IOC_INO_PATHS:
5267 return btrfs_ioctl_ino_to_path(root, argp);
5268 case BTRFS_IOC_LOGICAL_INO:
5269 return btrfs_ioctl_logical_to_ino(root, argp);
5270 case BTRFS_IOC_SPACE_INFO:
5271 return btrfs_ioctl_space_info(root, argp);
5272 case BTRFS_IOC_SYNC: {
5275 ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1);
5278 ret = btrfs_sync_fs(file->f_dentry->d_sb, 1);
5281 case BTRFS_IOC_START_SYNC:
5282 return btrfs_ioctl_start_sync(root, argp);
5283 case BTRFS_IOC_WAIT_SYNC:
5284 return btrfs_ioctl_wait_sync(root, argp);
5285 case BTRFS_IOC_SCRUB:
5286 return btrfs_ioctl_scrub(file, argp);
5287 case BTRFS_IOC_SCRUB_CANCEL:
5288 return btrfs_ioctl_scrub_cancel(root, argp);
5289 case BTRFS_IOC_SCRUB_PROGRESS:
5290 return btrfs_ioctl_scrub_progress(root, argp);
5291 case BTRFS_IOC_BALANCE_V2:
5292 return btrfs_ioctl_balance(file, argp);
5293 case BTRFS_IOC_BALANCE_CTL:
5294 return btrfs_ioctl_balance_ctl(root, arg);
5295 case BTRFS_IOC_BALANCE_PROGRESS:
5296 return btrfs_ioctl_balance_progress(root, argp);
5297 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5298 return btrfs_ioctl_set_received_subvol(file, argp);
5300 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5301 return btrfs_ioctl_set_received_subvol_32(file, argp);
5303 case BTRFS_IOC_SEND:
5304 return btrfs_ioctl_send(file, argp);
5305 case BTRFS_IOC_GET_DEV_STATS:
5306 return btrfs_ioctl_get_dev_stats(root, argp);
5307 case BTRFS_IOC_QUOTA_CTL:
5308 return btrfs_ioctl_quota_ctl(file, argp);
5309 case BTRFS_IOC_QGROUP_ASSIGN:
5310 return btrfs_ioctl_qgroup_assign(file, argp);
5311 case BTRFS_IOC_QGROUP_CREATE:
5312 return btrfs_ioctl_qgroup_create(file, argp);
5313 case BTRFS_IOC_QGROUP_LIMIT:
5314 return btrfs_ioctl_qgroup_limit(file, argp);
5315 case BTRFS_IOC_QUOTA_RESCAN:
5316 return btrfs_ioctl_quota_rescan(file, argp);
5317 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5318 return btrfs_ioctl_quota_rescan_status(file, argp);
5319 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5320 return btrfs_ioctl_quota_rescan_wait(file, argp);
5321 case BTRFS_IOC_DEV_REPLACE:
5322 return btrfs_ioctl_dev_replace(root, argp);
5323 case BTRFS_IOC_GET_FSLABEL:
5324 return btrfs_ioctl_get_fslabel(file, argp);
5325 case BTRFS_IOC_SET_FSLABEL:
5326 return btrfs_ioctl_set_fslabel(file, argp);
5327 case BTRFS_IOC_FILE_EXTENT_SAME:
5328 return btrfs_ioctl_file_extent_same(file, argp);
5329 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5330 return btrfs_ioctl_get_supported_features(file, argp);
5331 case BTRFS_IOC_GET_FEATURES:
5332 return btrfs_ioctl_get_features(file, argp);
5333 case BTRFS_IOC_SET_FEATURES:
5334 return btrfs_ioctl_set_features(file, argp);