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);
140 inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
142 if (ip->flags & BTRFS_INODE_SYNC)
143 inode->i_flags |= S_SYNC;
144 if (ip->flags & BTRFS_INODE_IMMUTABLE)
145 inode->i_flags |= S_IMMUTABLE;
146 if (ip->flags & BTRFS_INODE_APPEND)
147 inode->i_flags |= S_APPEND;
148 if (ip->flags & BTRFS_INODE_NOATIME)
149 inode->i_flags |= S_NOATIME;
150 if (ip->flags & BTRFS_INODE_DIRSYNC)
151 inode->i_flags |= S_DIRSYNC;
155 * Inherit flags from the parent inode.
157 * Currently only the compression flags and the cow flags are inherited.
159 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
166 flags = BTRFS_I(dir)->flags;
168 if (flags & BTRFS_INODE_NOCOMPRESS) {
169 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
170 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
171 } else if (flags & BTRFS_INODE_COMPRESS) {
172 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
173 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
176 if (flags & BTRFS_INODE_NODATACOW) {
177 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
178 if (S_ISREG(inode->i_mode))
179 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
182 btrfs_update_iflags(inode);
185 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
187 struct btrfs_inode *ip = BTRFS_I(file_inode(file));
188 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
190 if (copy_to_user(arg, &flags, sizeof(flags)))
195 static int check_flags(unsigned int flags)
197 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
198 FS_NOATIME_FL | FS_NODUMP_FL | \
199 FS_SYNC_FL | FS_DIRSYNC_FL | \
200 FS_NOCOMP_FL | FS_COMPR_FL |
204 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
210 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
212 struct inode *inode = file_inode(file);
213 struct btrfs_inode *ip = BTRFS_I(inode);
214 struct btrfs_root *root = ip->root;
215 struct btrfs_trans_handle *trans;
216 unsigned int flags, oldflags;
219 unsigned int i_oldflags;
222 if (!inode_owner_or_capable(inode))
225 if (btrfs_root_readonly(root))
228 if (copy_from_user(&flags, arg, sizeof(flags)))
231 ret = check_flags(flags);
235 ret = mnt_want_write_file(file);
239 mutex_lock(&inode->i_mutex);
241 ip_oldflags = ip->flags;
242 i_oldflags = inode->i_flags;
243 mode = inode->i_mode;
245 flags = btrfs_mask_flags(inode->i_mode, flags);
246 oldflags = btrfs_flags_to_ioctl(ip->flags);
247 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
248 if (!capable(CAP_LINUX_IMMUTABLE)) {
254 if (flags & FS_SYNC_FL)
255 ip->flags |= BTRFS_INODE_SYNC;
257 ip->flags &= ~BTRFS_INODE_SYNC;
258 if (flags & FS_IMMUTABLE_FL)
259 ip->flags |= BTRFS_INODE_IMMUTABLE;
261 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
262 if (flags & FS_APPEND_FL)
263 ip->flags |= BTRFS_INODE_APPEND;
265 ip->flags &= ~BTRFS_INODE_APPEND;
266 if (flags & FS_NODUMP_FL)
267 ip->flags |= BTRFS_INODE_NODUMP;
269 ip->flags &= ~BTRFS_INODE_NODUMP;
270 if (flags & FS_NOATIME_FL)
271 ip->flags |= BTRFS_INODE_NOATIME;
273 ip->flags &= ~BTRFS_INODE_NOATIME;
274 if (flags & FS_DIRSYNC_FL)
275 ip->flags |= BTRFS_INODE_DIRSYNC;
277 ip->flags &= ~BTRFS_INODE_DIRSYNC;
278 if (flags & FS_NOCOW_FL) {
281 * It's safe to turn csums off here, no extents exist.
282 * Otherwise we want the flag to reflect the real COW
283 * status of the file and will not set it.
285 if (inode->i_size == 0)
286 ip->flags |= BTRFS_INODE_NODATACOW
287 | BTRFS_INODE_NODATASUM;
289 ip->flags |= BTRFS_INODE_NODATACOW;
293 * Revert back under same assuptions as above
296 if (inode->i_size == 0)
297 ip->flags &= ~(BTRFS_INODE_NODATACOW
298 | BTRFS_INODE_NODATASUM);
300 ip->flags &= ~BTRFS_INODE_NODATACOW;
305 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
306 * flag may be changed automatically if compression code won't make
309 if (flags & FS_NOCOMP_FL) {
310 ip->flags &= ~BTRFS_INODE_COMPRESS;
311 ip->flags |= BTRFS_INODE_NOCOMPRESS;
313 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
314 if (ret && ret != -ENODATA)
316 } else if (flags & FS_COMPR_FL) {
319 ip->flags |= BTRFS_INODE_COMPRESS;
320 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
322 if (root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
326 ret = btrfs_set_prop(inode, "btrfs.compression",
327 comp, strlen(comp), 0);
332 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
335 trans = btrfs_start_transaction(root, 1);
337 ret = PTR_ERR(trans);
341 btrfs_update_iflags(inode);
342 inode_inc_iversion(inode);
343 inode->i_ctime = CURRENT_TIME;
344 ret = btrfs_update_inode(trans, root, inode);
346 btrfs_end_transaction(trans, root);
349 ip->flags = ip_oldflags;
350 inode->i_flags = i_oldflags;
354 mutex_unlock(&inode->i_mutex);
355 mnt_drop_write_file(file);
359 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
361 struct inode *inode = file_inode(file);
363 return put_user(inode->i_generation, arg);
366 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
368 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
369 struct btrfs_device *device;
370 struct request_queue *q;
371 struct fstrim_range range;
372 u64 minlen = ULLONG_MAX;
374 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
377 if (!capable(CAP_SYS_ADMIN))
381 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
385 q = bdev_get_queue(device->bdev);
386 if (blk_queue_discard(q)) {
388 minlen = min((u64)q->limits.discard_granularity,
396 if (copy_from_user(&range, arg, sizeof(range)))
398 if (range.start > total_bytes ||
399 range.len < fs_info->sb->s_blocksize)
402 range.len = min(range.len, total_bytes - range.start);
403 range.minlen = max(range.minlen, minlen);
404 ret = btrfs_trim_fs(fs_info->tree_root, &range);
408 if (copy_to_user(arg, &range, sizeof(range)))
414 int btrfs_is_empty_uuid(u8 *uuid)
418 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
425 static noinline int create_subvol(struct inode *dir,
426 struct dentry *dentry,
427 char *name, int namelen,
429 struct btrfs_qgroup_inherit *inherit)
431 struct btrfs_trans_handle *trans;
432 struct btrfs_key key;
433 struct btrfs_root_item root_item;
434 struct btrfs_inode_item *inode_item;
435 struct extent_buffer *leaf;
436 struct btrfs_root *root = BTRFS_I(dir)->root;
437 struct btrfs_root *new_root;
438 struct btrfs_block_rsv block_rsv;
439 struct timespec cur_time = CURRENT_TIME;
444 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
449 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
453 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
455 * The same as the snapshot creation, please see the comment
456 * of create_snapshot().
458 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
459 8, &qgroup_reserved, false);
463 trans = btrfs_start_transaction(root, 0);
465 ret = PTR_ERR(trans);
466 btrfs_subvolume_release_metadata(root, &block_rsv,
470 trans->block_rsv = &block_rsv;
471 trans->bytes_reserved = block_rsv.size;
473 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
477 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
478 0, objectid, NULL, 0, 0, 0);
484 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
485 btrfs_set_header_bytenr(leaf, leaf->start);
486 btrfs_set_header_generation(leaf, trans->transid);
487 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
488 btrfs_set_header_owner(leaf, objectid);
490 write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(),
492 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
493 btrfs_header_chunk_tree_uuid(leaf),
495 btrfs_mark_buffer_dirty(leaf);
497 memset(&root_item, 0, sizeof(root_item));
499 inode_item = &root_item.inode;
500 btrfs_set_stack_inode_generation(inode_item, 1);
501 btrfs_set_stack_inode_size(inode_item, 3);
502 btrfs_set_stack_inode_nlink(inode_item, 1);
503 btrfs_set_stack_inode_nbytes(inode_item, root->leafsize);
504 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
506 btrfs_set_root_flags(&root_item, 0);
507 btrfs_set_root_limit(&root_item, 0);
508 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
510 btrfs_set_root_bytenr(&root_item, leaf->start);
511 btrfs_set_root_generation(&root_item, trans->transid);
512 btrfs_set_root_level(&root_item, 0);
513 btrfs_set_root_refs(&root_item, 1);
514 btrfs_set_root_used(&root_item, leaf->len);
515 btrfs_set_root_last_snapshot(&root_item, 0);
517 btrfs_set_root_generation_v2(&root_item,
518 btrfs_root_generation(&root_item));
519 uuid_le_gen(&new_uuid);
520 memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
521 btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec);
522 btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec);
523 root_item.ctime = root_item.otime;
524 btrfs_set_root_ctransid(&root_item, trans->transid);
525 btrfs_set_root_otransid(&root_item, trans->transid);
527 btrfs_tree_unlock(leaf);
528 free_extent_buffer(leaf);
531 btrfs_set_root_dirid(&root_item, new_dirid);
533 key.objectid = objectid;
535 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
536 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
541 key.offset = (u64)-1;
542 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
543 if (IS_ERR(new_root)) {
544 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
545 ret = PTR_ERR(new_root);
549 btrfs_record_root_in_trans(trans, new_root);
551 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
553 /* We potentially lose an unused inode item here */
554 btrfs_abort_transaction(trans, root, ret);
559 * insert the directory item
561 ret = btrfs_set_inode_index(dir, &index);
563 btrfs_abort_transaction(trans, root, ret);
567 ret = btrfs_insert_dir_item(trans, root,
568 name, namelen, dir, &key,
569 BTRFS_FT_DIR, index);
571 btrfs_abort_transaction(trans, root, ret);
575 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
576 ret = btrfs_update_inode(trans, root, dir);
579 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
580 objectid, root->root_key.objectid,
581 btrfs_ino(dir), index, name, namelen);
584 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
585 root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
588 btrfs_abort_transaction(trans, root, ret);
591 trans->block_rsv = NULL;
592 trans->bytes_reserved = 0;
593 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
596 *async_transid = trans->transid;
597 err = btrfs_commit_transaction_async(trans, root, 1);
599 err = btrfs_commit_transaction(trans, root);
601 err = btrfs_commit_transaction(trans, root);
607 inode = btrfs_lookup_dentry(dir, dentry);
609 return PTR_ERR(inode);
610 d_instantiate(dentry, inode);
615 static void btrfs_wait_nocow_write(struct btrfs_root *root)
621 prepare_to_wait(&root->subv_writers->wait, &wait,
622 TASK_UNINTERRUPTIBLE);
624 writers = percpu_counter_sum(&root->subv_writers->counter);
628 finish_wait(&root->subv_writers->wait, &wait);
632 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
633 struct dentry *dentry, char *name, int namelen,
634 u64 *async_transid, bool readonly,
635 struct btrfs_qgroup_inherit *inherit)
638 struct btrfs_pending_snapshot *pending_snapshot;
639 struct btrfs_trans_handle *trans;
642 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
645 atomic_inc(&root->will_be_snapshoted);
646 smp_mb__after_atomic();
647 btrfs_wait_nocow_write(root);
649 ret = btrfs_start_delalloc_inodes(root, 0);
653 btrfs_wait_ordered_extents(root, -1);
655 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
656 if (!pending_snapshot) {
661 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
662 BTRFS_BLOCK_RSV_TEMP);
664 * 1 - parent dir inode
667 * 2 - root ref/backref
668 * 1 - root of snapshot
671 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
672 &pending_snapshot->block_rsv, 8,
673 &pending_snapshot->qgroup_reserved,
678 pending_snapshot->dentry = dentry;
679 pending_snapshot->root = root;
680 pending_snapshot->readonly = readonly;
681 pending_snapshot->dir = dir;
682 pending_snapshot->inherit = inherit;
684 trans = btrfs_start_transaction(root, 0);
686 ret = PTR_ERR(trans);
690 spin_lock(&root->fs_info->trans_lock);
691 list_add(&pending_snapshot->list,
692 &trans->transaction->pending_snapshots);
693 spin_unlock(&root->fs_info->trans_lock);
695 *async_transid = trans->transid;
696 ret = btrfs_commit_transaction_async(trans,
697 root->fs_info->extent_root, 1);
699 ret = btrfs_commit_transaction(trans, root);
701 ret = btrfs_commit_transaction(trans,
702 root->fs_info->extent_root);
707 ret = pending_snapshot->error;
711 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
716 * If orphan cleanup did remove any orphans, it means the tree was
717 * modified and therefore the commit root is not the same as the
718 * current root anymore. This is a problem, because send uses the
719 * commit root and therefore can see inode items that don't exist
720 * in the current root anymore, and for example make calls to
721 * btrfs_iget, which will do tree lookups based on the current root
722 * and not on the commit root. Those lookups will fail, returning a
723 * -ESTALE error, and making send fail with that error. So make sure
724 * a send does not see any orphans we have just removed, and that it
725 * will see the same inodes regardless of whether a transaction
726 * commit happened before it started (meaning that the commit root
727 * will be the same as the current root) or not.
729 if (readonly && pending_snapshot->snap->node !=
730 pending_snapshot->snap->commit_root) {
731 trans = btrfs_join_transaction(pending_snapshot->snap);
732 if (IS_ERR(trans) && PTR_ERR(trans) != -ENOENT) {
733 ret = PTR_ERR(trans);
736 if (!IS_ERR(trans)) {
737 ret = btrfs_commit_transaction(trans,
738 pending_snapshot->snap);
744 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
746 ret = PTR_ERR(inode);
750 d_instantiate(dentry, inode);
753 btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
754 &pending_snapshot->block_rsv,
755 pending_snapshot->qgroup_reserved);
757 kfree(pending_snapshot);
759 atomic_dec(&root->will_be_snapshoted);
763 /* copy of check_sticky in fs/namei.c()
764 * It's inline, so penalty for filesystems that don't use sticky bit is
767 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
769 kuid_t fsuid = current_fsuid();
771 if (!(dir->i_mode & S_ISVTX))
773 if (uid_eq(inode->i_uid, fsuid))
775 if (uid_eq(dir->i_uid, fsuid))
777 return !capable(CAP_FOWNER);
780 /* copy of may_delete in fs/namei.c()
781 * Check whether we can remove a link victim from directory dir, check
782 * whether the type of victim is right.
783 * 1. We can't do it if dir is read-only (done in permission())
784 * 2. We should have write and exec permissions on dir
785 * 3. We can't remove anything from append-only dir
786 * 4. We can't do anything with immutable dir (done in permission())
787 * 5. If the sticky bit on dir is set we should either
788 * a. be owner of dir, or
789 * b. be owner of victim, or
790 * c. have CAP_FOWNER capability
791 * 6. If the victim is append-only or immutable we can't do antyhing with
792 * links pointing to it.
793 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
794 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
795 * 9. We can't remove a root or mountpoint.
796 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
797 * nfs_async_unlink().
800 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
804 if (!victim->d_inode)
807 BUG_ON(victim->d_parent->d_inode != dir);
808 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
810 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
815 if (btrfs_check_sticky(dir, victim->d_inode)||
816 IS_APPEND(victim->d_inode)||
817 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
820 if (!S_ISDIR(victim->d_inode->i_mode))
824 } else if (S_ISDIR(victim->d_inode->i_mode))
828 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
833 /* copy of may_create in fs/namei.c() */
834 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
840 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
844 * Create a new subvolume below @parent. This is largely modeled after
845 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
846 * inside this filesystem so it's quite a bit simpler.
848 static noinline int btrfs_mksubvol(struct path *parent,
849 char *name, int namelen,
850 struct btrfs_root *snap_src,
851 u64 *async_transid, bool readonly,
852 struct btrfs_qgroup_inherit *inherit)
854 struct inode *dir = parent->dentry->d_inode;
855 struct dentry *dentry;
858 error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
862 dentry = lookup_one_len(name, parent->dentry, namelen);
863 error = PTR_ERR(dentry);
871 error = btrfs_may_create(dir, dentry);
876 * even if this name doesn't exist, we may get hash collisions.
877 * check for them now when we can safely fail
879 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
885 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
887 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
891 error = create_snapshot(snap_src, dir, dentry, name, namelen,
892 async_transid, readonly, inherit);
894 error = create_subvol(dir, dentry, name, namelen,
895 async_transid, inherit);
898 fsnotify_mkdir(dir, dentry);
900 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
904 mutex_unlock(&dir->i_mutex);
909 * When we're defragging a range, we don't want to kick it off again
910 * if it is really just waiting for delalloc to send it down.
911 * If we find a nice big extent or delalloc range for the bytes in the
912 * file you want to defrag, we return 0 to let you know to skip this
915 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
917 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
918 struct extent_map *em = NULL;
919 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
922 read_lock(&em_tree->lock);
923 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
924 read_unlock(&em_tree->lock);
927 end = extent_map_end(em);
929 if (end - offset > thresh)
932 /* if we already have a nice delalloc here, just stop */
934 end = count_range_bits(io_tree, &offset, offset + thresh,
935 thresh, EXTENT_DELALLOC, 1);
942 * helper function to walk through a file and find extents
943 * newer than a specific transid, and smaller than thresh.
945 * This is used by the defragging code to find new and small
948 static int find_new_extents(struct btrfs_root *root,
949 struct inode *inode, u64 newer_than,
950 u64 *off, int thresh)
952 struct btrfs_path *path;
953 struct btrfs_key min_key;
954 struct extent_buffer *leaf;
955 struct btrfs_file_extent_item *extent;
958 u64 ino = btrfs_ino(inode);
960 path = btrfs_alloc_path();
964 min_key.objectid = ino;
965 min_key.type = BTRFS_EXTENT_DATA_KEY;
966 min_key.offset = *off;
969 path->keep_locks = 1;
970 ret = btrfs_search_forward(root, &min_key, path, newer_than);
973 path->keep_locks = 0;
974 btrfs_unlock_up_safe(path, 1);
976 if (min_key.objectid != ino)
978 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
981 leaf = path->nodes[0];
982 extent = btrfs_item_ptr(leaf, path->slots[0],
983 struct btrfs_file_extent_item);
985 type = btrfs_file_extent_type(leaf, extent);
986 if (type == BTRFS_FILE_EXTENT_REG &&
987 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
988 check_defrag_in_cache(inode, min_key.offset, thresh)) {
989 *off = min_key.offset;
990 btrfs_free_path(path);
995 if (path->slots[0] < btrfs_header_nritems(leaf)) {
996 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
1000 if (min_key.offset == (u64)-1)
1004 btrfs_release_path(path);
1007 btrfs_free_path(path);
1011 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
1013 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
1014 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
1015 struct extent_map *em;
1016 u64 len = PAGE_CACHE_SIZE;
1019 * hopefully we have this extent in the tree already, try without
1020 * the full extent lock
1022 read_lock(&em_tree->lock);
1023 em = lookup_extent_mapping(em_tree, start, len);
1024 read_unlock(&em_tree->lock);
1027 struct extent_state *cached = NULL;
1028 u64 end = start + len - 1;
1030 /* get the big lock and read metadata off disk */
1031 lock_extent_bits(io_tree, start, end, 0, &cached);
1032 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
1033 unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS);
1042 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1044 struct extent_map *next;
1047 /* this is the last extent */
1048 if (em->start + em->len >= i_size_read(inode))
1051 next = defrag_lookup_extent(inode, em->start + em->len);
1052 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE ||
1053 (em->block_start + em->block_len == next->block_start))
1056 free_extent_map(next);
1060 static int should_defrag_range(struct inode *inode, u64 start, int thresh,
1061 u64 *last_len, u64 *skip, u64 *defrag_end,
1064 struct extent_map *em;
1066 bool next_mergeable = true;
1069 * make sure that once we start defragging an extent, we keep on
1072 if (start < *defrag_end)
1077 em = defrag_lookup_extent(inode, start);
1081 /* this will cover holes, and inline extents */
1082 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1087 next_mergeable = defrag_check_next_extent(inode, em);
1090 * we hit a real extent, if it is big or the next extent is not a
1091 * real extent, don't bother defragging it
1093 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1094 (em->len >= thresh || !next_mergeable))
1098 * last_len ends up being a counter of how many bytes we've defragged.
1099 * every time we choose not to defrag an extent, we reset *last_len
1100 * so that the next tiny extent will force a defrag.
1102 * The end result of this is that tiny extents before a single big
1103 * extent will force at least part of that big extent to be defragged.
1106 *defrag_end = extent_map_end(em);
1109 *skip = extent_map_end(em);
1113 free_extent_map(em);
1118 * it doesn't do much good to defrag one or two pages
1119 * at a time. This pulls in a nice chunk of pages
1120 * to COW and defrag.
1122 * It also makes sure the delalloc code has enough
1123 * dirty data to avoid making new small extents as part
1126 * It's a good idea to start RA on this range
1127 * before calling this.
1129 static int cluster_pages_for_defrag(struct inode *inode,
1130 struct page **pages,
1131 unsigned long start_index,
1132 unsigned long num_pages)
1134 unsigned long file_end;
1135 u64 isize = i_size_read(inode);
1142 struct btrfs_ordered_extent *ordered;
1143 struct extent_state *cached_state = NULL;
1144 struct extent_io_tree *tree;
1145 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1147 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1148 if (!isize || start_index > file_end)
1151 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1153 ret = btrfs_delalloc_reserve_space(inode,
1154 page_cnt << PAGE_CACHE_SHIFT);
1158 tree = &BTRFS_I(inode)->io_tree;
1160 /* step one, lock all the pages */
1161 for (i = 0; i < page_cnt; i++) {
1164 page = find_or_create_page(inode->i_mapping,
1165 start_index + i, mask);
1169 page_start = page_offset(page);
1170 page_end = page_start + PAGE_CACHE_SIZE - 1;
1172 lock_extent_bits(tree, page_start, page_end,
1174 ordered = btrfs_lookup_ordered_extent(inode,
1176 unlock_extent_cached(tree, page_start, page_end,
1177 &cached_state, GFP_NOFS);
1182 btrfs_start_ordered_extent(inode, ordered, 1);
1183 btrfs_put_ordered_extent(ordered);
1186 * we unlocked the page above, so we need check if
1187 * it was released or not.
1189 if (page->mapping != inode->i_mapping) {
1191 page_cache_release(page);
1196 if (!PageUptodate(page)) {
1197 btrfs_readpage(NULL, page);
1199 if (!PageUptodate(page)) {
1201 page_cache_release(page);
1207 if (page->mapping != inode->i_mapping) {
1209 page_cache_release(page);
1219 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1223 * so now we have a nice long stream of locked
1224 * and up to date pages, lets wait on them
1226 for (i = 0; i < i_done; i++)
1227 wait_on_page_writeback(pages[i]);
1229 page_start = page_offset(pages[0]);
1230 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1232 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1233 page_start, page_end - 1, 0, &cached_state);
1234 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1235 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1236 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1237 &cached_state, GFP_NOFS);
1239 if (i_done != page_cnt) {
1240 spin_lock(&BTRFS_I(inode)->lock);
1241 BTRFS_I(inode)->outstanding_extents++;
1242 spin_unlock(&BTRFS_I(inode)->lock);
1243 btrfs_delalloc_release_space(inode,
1244 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1248 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1249 &cached_state, GFP_NOFS);
1251 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1252 page_start, page_end - 1, &cached_state,
1255 for (i = 0; i < i_done; i++) {
1256 clear_page_dirty_for_io(pages[i]);
1257 ClearPageChecked(pages[i]);
1258 set_page_extent_mapped(pages[i]);
1259 set_page_dirty(pages[i]);
1260 unlock_page(pages[i]);
1261 page_cache_release(pages[i]);
1265 for (i = 0; i < i_done; i++) {
1266 unlock_page(pages[i]);
1267 page_cache_release(pages[i]);
1269 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1274 int btrfs_defrag_file(struct inode *inode, struct file *file,
1275 struct btrfs_ioctl_defrag_range_args *range,
1276 u64 newer_than, unsigned long max_to_defrag)
1278 struct btrfs_root *root = BTRFS_I(inode)->root;
1279 struct file_ra_state *ra = NULL;
1280 unsigned long last_index;
1281 u64 isize = i_size_read(inode);
1285 u64 newer_off = range->start;
1287 unsigned long ra_index = 0;
1289 int defrag_count = 0;
1290 int compress_type = BTRFS_COMPRESS_ZLIB;
1291 int extent_thresh = range->extent_thresh;
1292 unsigned long max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1293 unsigned long cluster = max_cluster;
1294 u64 new_align = ~((u64)128 * 1024 - 1);
1295 struct page **pages = NULL;
1300 if (range->start >= isize)
1303 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1304 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1306 if (range->compress_type)
1307 compress_type = range->compress_type;
1310 if (extent_thresh == 0)
1311 extent_thresh = 256 * 1024;
1314 * if we were not given a file, allocate a readahead
1318 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1321 file_ra_state_init(ra, inode->i_mapping);
1326 pages = kmalloc_array(max_cluster, sizeof(struct page *),
1333 /* find the last page to defrag */
1334 if (range->start + range->len > range->start) {
1335 last_index = min_t(u64, isize - 1,
1336 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1338 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1342 ret = find_new_extents(root, inode, newer_than,
1343 &newer_off, 64 * 1024);
1345 range->start = newer_off;
1347 * we always align our defrag to help keep
1348 * the extents in the file evenly spaced
1350 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1354 i = range->start >> PAGE_CACHE_SHIFT;
1357 max_to_defrag = last_index + 1;
1360 * make writeback starts from i, so the defrag range can be
1361 * written sequentially.
1363 if (i < inode->i_mapping->writeback_index)
1364 inode->i_mapping->writeback_index = i;
1366 while (i <= last_index && defrag_count < max_to_defrag &&
1367 (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1368 PAGE_CACHE_SHIFT)) {
1370 * make sure we stop running if someone unmounts
1373 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1376 if (btrfs_defrag_cancelled(root->fs_info)) {
1377 printk(KERN_DEBUG "BTRFS: defrag_file cancelled\n");
1382 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1383 extent_thresh, &last_len, &skip,
1384 &defrag_end, range->flags &
1385 BTRFS_DEFRAG_RANGE_COMPRESS)) {
1388 * the should_defrag function tells us how much to skip
1389 * bump our counter by the suggested amount
1391 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1392 i = max(i + 1, next);
1397 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1398 PAGE_CACHE_SHIFT) - i;
1399 cluster = min(cluster, max_cluster);
1401 cluster = max_cluster;
1404 if (i + cluster > ra_index) {
1405 ra_index = max(i, ra_index);
1406 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1408 ra_index += max_cluster;
1411 mutex_lock(&inode->i_mutex);
1412 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1413 BTRFS_I(inode)->force_compress = compress_type;
1414 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1416 mutex_unlock(&inode->i_mutex);
1420 defrag_count += ret;
1421 balance_dirty_pages_ratelimited(inode->i_mapping);
1422 mutex_unlock(&inode->i_mutex);
1425 if (newer_off == (u64)-1)
1431 newer_off = max(newer_off + 1,
1432 (u64)i << PAGE_CACHE_SHIFT);
1434 ret = find_new_extents(root, inode,
1435 newer_than, &newer_off,
1438 range->start = newer_off;
1439 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1446 last_len += ret << PAGE_CACHE_SHIFT;
1454 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1455 filemap_flush(inode->i_mapping);
1456 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1457 &BTRFS_I(inode)->runtime_flags))
1458 filemap_flush(inode->i_mapping);
1461 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1462 /* the filemap_flush will queue IO into the worker threads, but
1463 * we have to make sure the IO is actually started and that
1464 * ordered extents get created before we return
1466 atomic_inc(&root->fs_info->async_submit_draining);
1467 while (atomic_read(&root->fs_info->nr_async_submits) ||
1468 atomic_read(&root->fs_info->async_delalloc_pages)) {
1469 wait_event(root->fs_info->async_submit_wait,
1470 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1471 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1473 atomic_dec(&root->fs_info->async_submit_draining);
1476 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1477 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1483 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1484 mutex_lock(&inode->i_mutex);
1485 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1486 mutex_unlock(&inode->i_mutex);
1494 static noinline int btrfs_ioctl_resize(struct file *file,
1500 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
1501 struct btrfs_ioctl_vol_args *vol_args;
1502 struct btrfs_trans_handle *trans;
1503 struct btrfs_device *device = NULL;
1506 char *devstr = NULL;
1510 if (!capable(CAP_SYS_ADMIN))
1513 ret = mnt_want_write_file(file);
1517 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1519 mnt_drop_write_file(file);
1520 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1523 mutex_lock(&root->fs_info->volume_mutex);
1524 vol_args = memdup_user(arg, sizeof(*vol_args));
1525 if (IS_ERR(vol_args)) {
1526 ret = PTR_ERR(vol_args);
1530 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1532 sizestr = vol_args->name;
1533 devstr = strchr(sizestr, ':');
1535 sizestr = devstr + 1;
1537 devstr = vol_args->name;
1538 ret = kstrtoull(devstr, 10, &devid);
1545 btrfs_info(root->fs_info, "resizing devid %llu", devid);
1548 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
1550 btrfs_info(root->fs_info, "resizer unable to find device %llu",
1556 if (!device->writeable) {
1557 btrfs_info(root->fs_info,
1558 "resizer unable to apply on readonly device %llu",
1564 if (!strcmp(sizestr, "max"))
1565 new_size = device->bdev->bd_inode->i_size;
1567 if (sizestr[0] == '-') {
1570 } else if (sizestr[0] == '+') {
1574 new_size = memparse(sizestr, &retptr);
1575 if (*retptr != '\0' || new_size == 0) {
1581 if (device->is_tgtdev_for_dev_replace) {
1586 old_size = device->total_bytes;
1589 if (new_size > old_size) {
1593 new_size = old_size - new_size;
1594 } else if (mod > 0) {
1595 if (new_size > ULLONG_MAX - old_size) {
1599 new_size = old_size + new_size;
1602 if (new_size < 256 * 1024 * 1024) {
1606 if (new_size > device->bdev->bd_inode->i_size) {
1611 do_div(new_size, root->sectorsize);
1612 new_size *= root->sectorsize;
1614 printk_in_rcu(KERN_INFO "BTRFS: new size for %s is %llu\n",
1615 rcu_str_deref(device->name), new_size);
1617 if (new_size > old_size) {
1618 trans = btrfs_start_transaction(root, 0);
1619 if (IS_ERR(trans)) {
1620 ret = PTR_ERR(trans);
1623 ret = btrfs_grow_device(trans, device, new_size);
1624 btrfs_commit_transaction(trans, root);
1625 } else if (new_size < old_size) {
1626 ret = btrfs_shrink_device(device, new_size);
1627 } /* equal, nothing need to do */
1632 mutex_unlock(&root->fs_info->volume_mutex);
1633 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
1634 mnt_drop_write_file(file);
1638 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1639 char *name, unsigned long fd, int subvol,
1640 u64 *transid, bool readonly,
1641 struct btrfs_qgroup_inherit *inherit)
1646 ret = mnt_want_write_file(file);
1650 namelen = strlen(name);
1651 if (strchr(name, '/')) {
1653 goto out_drop_write;
1656 if (name[0] == '.' &&
1657 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1659 goto out_drop_write;
1663 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1664 NULL, transid, readonly, inherit);
1666 struct fd src = fdget(fd);
1667 struct inode *src_inode;
1670 goto out_drop_write;
1673 src_inode = file_inode(src.file);
1674 if (src_inode->i_sb != file_inode(file)->i_sb) {
1675 btrfs_info(BTRFS_I(src_inode)->root->fs_info,
1676 "Snapshot src from another FS");
1678 } else if (!inode_owner_or_capable(src_inode)) {
1680 * Subvolume creation is not restricted, but snapshots
1681 * are limited to own subvolumes only
1685 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1686 BTRFS_I(src_inode)->root,
1687 transid, readonly, inherit);
1692 mnt_drop_write_file(file);
1697 static noinline int btrfs_ioctl_snap_create(struct file *file,
1698 void __user *arg, int subvol)
1700 struct btrfs_ioctl_vol_args *vol_args;
1703 vol_args = memdup_user(arg, sizeof(*vol_args));
1704 if (IS_ERR(vol_args))
1705 return PTR_ERR(vol_args);
1706 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1708 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1709 vol_args->fd, subvol,
1716 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1717 void __user *arg, int subvol)
1719 struct btrfs_ioctl_vol_args_v2 *vol_args;
1723 bool readonly = false;
1724 struct btrfs_qgroup_inherit *inherit = NULL;
1726 vol_args = memdup_user(arg, sizeof(*vol_args));
1727 if (IS_ERR(vol_args))
1728 return PTR_ERR(vol_args);
1729 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1731 if (vol_args->flags &
1732 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1733 BTRFS_SUBVOL_QGROUP_INHERIT)) {
1738 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1740 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1742 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1743 if (vol_args->size > PAGE_CACHE_SIZE) {
1747 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1748 if (IS_ERR(inherit)) {
1749 ret = PTR_ERR(inherit);
1754 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1755 vol_args->fd, subvol, ptr,
1758 if (ret == 0 && ptr &&
1760 offsetof(struct btrfs_ioctl_vol_args_v2,
1761 transid), ptr, sizeof(*ptr)))
1769 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1772 struct inode *inode = file_inode(file);
1773 struct btrfs_root *root = BTRFS_I(inode)->root;
1777 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1780 down_read(&root->fs_info->subvol_sem);
1781 if (btrfs_root_readonly(root))
1782 flags |= BTRFS_SUBVOL_RDONLY;
1783 up_read(&root->fs_info->subvol_sem);
1785 if (copy_to_user(arg, &flags, sizeof(flags)))
1791 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1794 struct inode *inode = file_inode(file);
1795 struct btrfs_root *root = BTRFS_I(inode)->root;
1796 struct btrfs_trans_handle *trans;
1801 if (!inode_owner_or_capable(inode))
1804 ret = mnt_want_write_file(file);
1808 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1810 goto out_drop_write;
1813 if (copy_from_user(&flags, arg, sizeof(flags))) {
1815 goto out_drop_write;
1818 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1820 goto out_drop_write;
1823 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1825 goto out_drop_write;
1828 down_write(&root->fs_info->subvol_sem);
1831 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1834 root_flags = btrfs_root_flags(&root->root_item);
1835 if (flags & BTRFS_SUBVOL_RDONLY) {
1836 btrfs_set_root_flags(&root->root_item,
1837 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1840 * Block RO -> RW transition if this subvolume is involved in
1843 spin_lock(&root->root_item_lock);
1844 if (root->send_in_progress == 0) {
1845 btrfs_set_root_flags(&root->root_item,
1846 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1847 spin_unlock(&root->root_item_lock);
1849 spin_unlock(&root->root_item_lock);
1850 btrfs_warn(root->fs_info,
1851 "Attempt to set subvolume %llu read-write during send",
1852 root->root_key.objectid);
1858 trans = btrfs_start_transaction(root, 1);
1859 if (IS_ERR(trans)) {
1860 ret = PTR_ERR(trans);
1864 ret = btrfs_update_root(trans, root->fs_info->tree_root,
1865 &root->root_key, &root->root_item);
1867 btrfs_commit_transaction(trans, root);
1870 btrfs_set_root_flags(&root->root_item, root_flags);
1872 up_write(&root->fs_info->subvol_sem);
1874 mnt_drop_write_file(file);
1880 * helper to check if the subvolume references other subvolumes
1882 static noinline int may_destroy_subvol(struct btrfs_root *root)
1884 struct btrfs_path *path;
1885 struct btrfs_dir_item *di;
1886 struct btrfs_key key;
1890 path = btrfs_alloc_path();
1894 /* Make sure this root isn't set as the default subvol */
1895 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1896 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1897 dir_id, "default", 7, 0);
1898 if (di && !IS_ERR(di)) {
1899 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1900 if (key.objectid == root->root_key.objectid) {
1902 btrfs_err(root->fs_info, "deleting default subvolume "
1903 "%llu is not allowed", key.objectid);
1906 btrfs_release_path(path);
1909 key.objectid = root->root_key.objectid;
1910 key.type = BTRFS_ROOT_REF_KEY;
1911 key.offset = (u64)-1;
1913 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1920 if (path->slots[0] > 0) {
1922 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1923 if (key.objectid == root->root_key.objectid &&
1924 key.type == BTRFS_ROOT_REF_KEY)
1928 btrfs_free_path(path);
1932 static noinline int key_in_sk(struct btrfs_key *key,
1933 struct btrfs_ioctl_search_key *sk)
1935 struct btrfs_key test;
1938 test.objectid = sk->min_objectid;
1939 test.type = sk->min_type;
1940 test.offset = sk->min_offset;
1942 ret = btrfs_comp_cpu_keys(key, &test);
1946 test.objectid = sk->max_objectid;
1947 test.type = sk->max_type;
1948 test.offset = sk->max_offset;
1950 ret = btrfs_comp_cpu_keys(key, &test);
1956 static noinline int copy_to_sk(struct btrfs_root *root,
1957 struct btrfs_path *path,
1958 struct btrfs_key *key,
1959 struct btrfs_ioctl_search_key *sk,
1961 unsigned long *sk_offset,
1965 struct extent_buffer *leaf;
1966 struct btrfs_ioctl_search_header sh;
1967 unsigned long item_off;
1968 unsigned long item_len;
1974 leaf = path->nodes[0];
1975 slot = path->slots[0];
1976 nritems = btrfs_header_nritems(leaf);
1978 if (btrfs_header_generation(leaf) > sk->max_transid) {
1982 found_transid = btrfs_header_generation(leaf);
1984 for (i = slot; i < nritems; i++) {
1985 item_off = btrfs_item_ptr_offset(leaf, i);
1986 item_len = btrfs_item_size_nr(leaf, i);
1988 btrfs_item_key_to_cpu(leaf, key, i);
1989 if (!key_in_sk(key, sk))
1992 if (sizeof(sh) + item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1995 if (sizeof(sh) + item_len + *sk_offset >
1996 BTRFS_SEARCH_ARGS_BUFSIZE) {
2001 sh.objectid = key->objectid;
2002 sh.offset = key->offset;
2003 sh.type = key->type;
2005 sh.transid = found_transid;
2007 /* copy search result header */
2008 memcpy(buf + *sk_offset, &sh, sizeof(sh));
2009 *sk_offset += sizeof(sh);
2012 char *p = buf + *sk_offset;
2014 read_extent_buffer(leaf, p,
2015 item_off, item_len);
2016 *sk_offset += item_len;
2020 if (*num_found >= sk->nr_items)
2025 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
2027 else if (key->type < (u8)-1 && key->type < sk->max_type) {
2030 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
2040 static noinline int search_ioctl(struct inode *inode,
2041 struct btrfs_ioctl_search_args *args)
2043 struct btrfs_root *root;
2044 struct btrfs_key key;
2045 struct btrfs_path *path;
2046 struct btrfs_ioctl_search_key *sk = &args->key;
2047 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
2050 unsigned long sk_offset = 0;
2052 path = btrfs_alloc_path();
2056 if (sk->tree_id == 0) {
2057 /* search the root of the inode that was passed */
2058 root = BTRFS_I(inode)->root;
2060 key.objectid = sk->tree_id;
2061 key.type = BTRFS_ROOT_ITEM_KEY;
2062 key.offset = (u64)-1;
2063 root = btrfs_read_fs_root_no_name(info, &key);
2065 printk(KERN_ERR "BTRFS: could not find root %llu\n",
2067 btrfs_free_path(path);
2072 key.objectid = sk->min_objectid;
2073 key.type = sk->min_type;
2074 key.offset = sk->min_offset;
2076 path->keep_locks = 1;
2079 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2085 ret = copy_to_sk(root, path, &key, sk, args->buf,
2086 &sk_offset, &num_found);
2087 btrfs_release_path(path);
2088 if (ret || num_found >= sk->nr_items)
2094 sk->nr_items = num_found;
2095 btrfs_free_path(path);
2099 static noinline int btrfs_ioctl_tree_search(struct file *file,
2102 struct btrfs_ioctl_search_args *args;
2103 struct inode *inode;
2106 if (!capable(CAP_SYS_ADMIN))
2109 args = memdup_user(argp, sizeof(*args));
2111 return PTR_ERR(args);
2113 inode = file_inode(file);
2114 ret = search_ioctl(inode, args);
2115 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2122 * Search INODE_REFs to identify path name of 'dirid' directory
2123 * in a 'tree_id' tree. and sets path name to 'name'.
2125 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2126 u64 tree_id, u64 dirid, char *name)
2128 struct btrfs_root *root;
2129 struct btrfs_key key;
2135 struct btrfs_inode_ref *iref;
2136 struct extent_buffer *l;
2137 struct btrfs_path *path;
2139 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2144 path = btrfs_alloc_path();
2148 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
2150 key.objectid = tree_id;
2151 key.type = BTRFS_ROOT_ITEM_KEY;
2152 key.offset = (u64)-1;
2153 root = btrfs_read_fs_root_no_name(info, &key);
2155 printk(KERN_ERR "BTRFS: could not find root %llu\n", tree_id);
2160 key.objectid = dirid;
2161 key.type = BTRFS_INODE_REF_KEY;
2162 key.offset = (u64)-1;
2165 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2169 ret = btrfs_previous_item(root, path, dirid,
2170 BTRFS_INODE_REF_KEY);
2180 slot = path->slots[0];
2181 btrfs_item_key_to_cpu(l, &key, slot);
2183 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2184 len = btrfs_inode_ref_name_len(l, iref);
2186 total_len += len + 1;
2188 ret = -ENAMETOOLONG;
2193 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2195 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2198 btrfs_release_path(path);
2199 key.objectid = key.offset;
2200 key.offset = (u64)-1;
2201 dirid = key.objectid;
2203 memmove(name, ptr, total_len);
2204 name[total_len] = '\0';
2207 btrfs_free_path(path);
2211 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2214 struct btrfs_ioctl_ino_lookup_args *args;
2215 struct inode *inode;
2218 if (!capable(CAP_SYS_ADMIN))
2221 args = memdup_user(argp, sizeof(*args));
2223 return PTR_ERR(args);
2225 inode = file_inode(file);
2227 if (args->treeid == 0)
2228 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2230 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2231 args->treeid, args->objectid,
2234 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2241 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2244 struct dentry *parent = file->f_path.dentry;
2245 struct dentry *dentry;
2246 struct inode *dir = parent->d_inode;
2247 struct inode *inode;
2248 struct btrfs_root *root = BTRFS_I(dir)->root;
2249 struct btrfs_root *dest = NULL;
2250 struct btrfs_ioctl_vol_args *vol_args;
2251 struct btrfs_trans_handle *trans;
2252 struct btrfs_block_rsv block_rsv;
2254 u64 qgroup_reserved;
2259 vol_args = memdup_user(arg, sizeof(*vol_args));
2260 if (IS_ERR(vol_args))
2261 return PTR_ERR(vol_args);
2263 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2264 namelen = strlen(vol_args->name);
2265 if (strchr(vol_args->name, '/') ||
2266 strncmp(vol_args->name, "..", namelen) == 0) {
2271 err = mnt_want_write_file(file);
2276 err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2278 goto out_drop_write;
2279 dentry = lookup_one_len(vol_args->name, parent, namelen);
2280 if (IS_ERR(dentry)) {
2281 err = PTR_ERR(dentry);
2282 goto out_unlock_dir;
2285 if (!dentry->d_inode) {
2290 inode = dentry->d_inode;
2291 dest = BTRFS_I(inode)->root;
2292 if (!capable(CAP_SYS_ADMIN)) {
2294 * Regular user. Only allow this with a special mount
2295 * option, when the user has write+exec access to the
2296 * subvol root, and when rmdir(2) would have been
2299 * Note that this is _not_ check that the subvol is
2300 * empty or doesn't contain data that we wouldn't
2301 * otherwise be able to delete.
2303 * Users who want to delete empty subvols should try
2307 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2311 * Do not allow deletion if the parent dir is the same
2312 * as the dir to be deleted. That means the ioctl
2313 * must be called on the dentry referencing the root
2314 * of the subvol, not a random directory contained
2321 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2326 /* check if subvolume may be deleted by a user */
2327 err = btrfs_may_delete(dir, dentry, 1);
2331 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2336 mutex_lock(&inode->i_mutex);
2339 * Don't allow to delete a subvolume with send in progress. This is
2340 * inside the i_mutex so the error handling that has to drop the bit
2341 * again is not run concurrently.
2343 spin_lock(&dest->root_item_lock);
2344 root_flags = btrfs_root_flags(&dest->root_item);
2345 if (dest->send_in_progress == 0) {
2346 btrfs_set_root_flags(&dest->root_item,
2347 root_flags | BTRFS_ROOT_SUBVOL_DEAD);
2348 spin_unlock(&dest->root_item_lock);
2350 spin_unlock(&dest->root_item_lock);
2351 btrfs_warn(root->fs_info,
2352 "Attempt to delete subvolume %llu during send",
2353 dest->root_key.objectid);
2358 err = d_invalidate(dentry);
2362 down_write(&root->fs_info->subvol_sem);
2364 err = may_destroy_subvol(dest);
2368 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2370 * One for dir inode, two for dir entries, two for root
2373 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2374 5, &qgroup_reserved, true);
2378 trans = btrfs_start_transaction(root, 0);
2379 if (IS_ERR(trans)) {
2380 err = PTR_ERR(trans);
2383 trans->block_rsv = &block_rsv;
2384 trans->bytes_reserved = block_rsv.size;
2386 ret = btrfs_unlink_subvol(trans, root, dir,
2387 dest->root_key.objectid,
2388 dentry->d_name.name,
2389 dentry->d_name.len);
2392 btrfs_abort_transaction(trans, root, ret);
2396 btrfs_record_root_in_trans(trans, dest);
2398 memset(&dest->root_item.drop_progress, 0,
2399 sizeof(dest->root_item.drop_progress));
2400 dest->root_item.drop_level = 0;
2401 btrfs_set_root_refs(&dest->root_item, 0);
2403 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
2404 ret = btrfs_insert_orphan_item(trans,
2405 root->fs_info->tree_root,
2406 dest->root_key.objectid);
2408 btrfs_abort_transaction(trans, root, ret);
2414 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2415 dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
2416 dest->root_key.objectid);
2417 if (ret && ret != -ENOENT) {
2418 btrfs_abort_transaction(trans, root, ret);
2422 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2423 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2424 dest->root_item.received_uuid,
2425 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2426 dest->root_key.objectid);
2427 if (ret && ret != -ENOENT) {
2428 btrfs_abort_transaction(trans, root, ret);
2435 trans->block_rsv = NULL;
2436 trans->bytes_reserved = 0;
2437 ret = btrfs_end_transaction(trans, root);
2440 inode->i_flags |= S_DEAD;
2442 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
2444 up_write(&root->fs_info->subvol_sem);
2447 spin_lock(&dest->root_item_lock);
2448 root_flags = btrfs_root_flags(&dest->root_item);
2449 btrfs_set_root_flags(&dest->root_item,
2450 root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
2451 spin_unlock(&dest->root_item_lock);
2453 mutex_unlock(&inode->i_mutex);
2455 shrink_dcache_sb(root->fs_info->sb);
2456 btrfs_invalidate_inodes(dest);
2458 ASSERT(dest->send_in_progress == 0);
2461 if (dest->cache_inode) {
2462 iput(dest->cache_inode);
2463 dest->cache_inode = NULL;
2469 mutex_unlock(&dir->i_mutex);
2471 mnt_drop_write_file(file);
2477 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2479 struct inode *inode = file_inode(file);
2480 struct btrfs_root *root = BTRFS_I(inode)->root;
2481 struct btrfs_ioctl_defrag_range_args *range;
2484 ret = mnt_want_write_file(file);
2488 if (btrfs_root_readonly(root)) {
2493 switch (inode->i_mode & S_IFMT) {
2495 if (!capable(CAP_SYS_ADMIN)) {
2499 ret = btrfs_defrag_root(root);
2502 ret = btrfs_defrag_root(root->fs_info->extent_root);
2505 if (!(file->f_mode & FMODE_WRITE)) {
2510 range = kzalloc(sizeof(*range), GFP_KERNEL);
2517 if (copy_from_user(range, argp,
2523 /* compression requires us to start the IO */
2524 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2525 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2526 range->extent_thresh = (u32)-1;
2529 /* the rest are all set to zero by kzalloc */
2530 range->len = (u64)-1;
2532 ret = btrfs_defrag_file(file_inode(file), file,
2542 mnt_drop_write_file(file);
2546 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2548 struct btrfs_ioctl_vol_args *vol_args;
2551 if (!capable(CAP_SYS_ADMIN))
2554 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2556 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2559 mutex_lock(&root->fs_info->volume_mutex);
2560 vol_args = memdup_user(arg, sizeof(*vol_args));
2561 if (IS_ERR(vol_args)) {
2562 ret = PTR_ERR(vol_args);
2566 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2567 ret = btrfs_init_new_device(root, vol_args->name);
2571 mutex_unlock(&root->fs_info->volume_mutex);
2572 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2576 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2578 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2579 struct btrfs_ioctl_vol_args *vol_args;
2582 if (!capable(CAP_SYS_ADMIN))
2585 ret = mnt_want_write_file(file);
2589 vol_args = memdup_user(arg, sizeof(*vol_args));
2590 if (IS_ERR(vol_args)) {
2591 ret = PTR_ERR(vol_args);
2595 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2597 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2599 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2603 mutex_lock(&root->fs_info->volume_mutex);
2604 ret = btrfs_rm_device(root, vol_args->name);
2605 mutex_unlock(&root->fs_info->volume_mutex);
2606 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2610 mnt_drop_write_file(file);
2614 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2616 struct btrfs_ioctl_fs_info_args *fi_args;
2617 struct btrfs_device *device;
2618 struct btrfs_device *next;
2619 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2622 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2626 mutex_lock(&fs_devices->device_list_mutex);
2627 fi_args->num_devices = fs_devices->num_devices;
2628 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2630 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2631 if (device->devid > fi_args->max_id)
2632 fi_args->max_id = device->devid;
2634 mutex_unlock(&fs_devices->device_list_mutex);
2636 fi_args->nodesize = root->fs_info->super_copy->nodesize;
2637 fi_args->sectorsize = root->fs_info->super_copy->sectorsize;
2638 fi_args->clone_alignment = root->fs_info->super_copy->sectorsize;
2640 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2647 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2649 struct btrfs_ioctl_dev_info_args *di_args;
2650 struct btrfs_device *dev;
2651 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2653 char *s_uuid = NULL;
2655 di_args = memdup_user(arg, sizeof(*di_args));
2656 if (IS_ERR(di_args))
2657 return PTR_ERR(di_args);
2659 if (!btrfs_is_empty_uuid(di_args->uuid))
2660 s_uuid = di_args->uuid;
2662 mutex_lock(&fs_devices->device_list_mutex);
2663 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
2670 di_args->devid = dev->devid;
2671 di_args->bytes_used = dev->bytes_used;
2672 di_args->total_bytes = dev->total_bytes;
2673 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2675 struct rcu_string *name;
2678 name = rcu_dereference(dev->name);
2679 strncpy(di_args->path, name->str, sizeof(di_args->path));
2681 di_args->path[sizeof(di_args->path) - 1] = 0;
2683 di_args->path[0] = '\0';
2687 mutex_unlock(&fs_devices->device_list_mutex);
2688 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2695 static struct page *extent_same_get_page(struct inode *inode, u64 off)
2699 struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
2701 index = off >> PAGE_CACHE_SHIFT;
2703 page = grab_cache_page(inode->i_mapping, index);
2707 if (!PageUptodate(page)) {
2708 if (extent_read_full_page_nolock(tree, page, btrfs_get_extent,
2712 if (!PageUptodate(page)) {
2714 page_cache_release(page);
2723 static inline void lock_extent_range(struct inode *inode, u64 off, u64 len)
2725 /* do any pending delalloc/csum calc on src, one way or
2726 another, and lock file content */
2728 struct btrfs_ordered_extent *ordered;
2729 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2730 ordered = btrfs_lookup_first_ordered_extent(inode,
2733 ordered->file_offset + ordered->len <= off ||
2734 ordered->file_offset >= off + len) &&
2735 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2736 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2738 btrfs_put_ordered_extent(ordered);
2741 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2743 btrfs_put_ordered_extent(ordered);
2744 btrfs_wait_ordered_range(inode, off, len);
2748 static void btrfs_double_unlock(struct inode *inode1, u64 loff1,
2749 struct inode *inode2, u64 loff2, u64 len)
2751 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2752 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2754 mutex_unlock(&inode1->i_mutex);
2755 mutex_unlock(&inode2->i_mutex);
2758 static void btrfs_double_lock(struct inode *inode1, u64 loff1,
2759 struct inode *inode2, u64 loff2, u64 len)
2761 if (inode1 < inode2) {
2762 swap(inode1, inode2);
2766 mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT);
2767 lock_extent_range(inode1, loff1, len);
2768 if (inode1 != inode2) {
2769 mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD);
2770 lock_extent_range(inode2, loff2, len);
2774 static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
2775 u64 dst_loff, u64 len)
2778 struct page *src_page, *dst_page;
2779 unsigned int cmp_len = PAGE_CACHE_SIZE;
2780 void *addr, *dst_addr;
2783 if (len < PAGE_CACHE_SIZE)
2786 src_page = extent_same_get_page(src, loff);
2789 dst_page = extent_same_get_page(dst, dst_loff);
2791 page_cache_release(src_page);
2794 addr = kmap_atomic(src_page);
2795 dst_addr = kmap_atomic(dst_page);
2797 flush_dcache_page(src_page);
2798 flush_dcache_page(dst_page);
2800 if (memcmp(addr, dst_addr, cmp_len))
2801 ret = BTRFS_SAME_DATA_DIFFERS;
2803 kunmap_atomic(addr);
2804 kunmap_atomic(dst_addr);
2805 page_cache_release(src_page);
2806 page_cache_release(dst_page);
2812 dst_loff += cmp_len;
2819 static int extent_same_check_offsets(struct inode *inode, u64 off, u64 len)
2821 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
2823 if (off + len > inode->i_size || off + len < off)
2825 /* Check that we are block aligned - btrfs_clone() requires this */
2826 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
2832 static int btrfs_extent_same(struct inode *src, u64 loff, u64 len,
2833 struct inode *dst, u64 dst_loff)
2838 * btrfs_clone() can't handle extents in the same file
2839 * yet. Once that works, we can drop this check and replace it
2840 * with a check for the same inode, but overlapping extents.
2845 btrfs_double_lock(src, loff, dst, dst_loff, len);
2847 ret = extent_same_check_offsets(src, loff, len);
2851 ret = extent_same_check_offsets(dst, dst_loff, len);
2855 /* don't make the dst file partly checksummed */
2856 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2857 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
2862 ret = btrfs_cmp_data(src, loff, dst, dst_loff, len);
2864 ret = btrfs_clone(src, dst, loff, len, len, dst_loff);
2867 btrfs_double_unlock(src, loff, dst, dst_loff, len);
2872 #define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024)
2874 static long btrfs_ioctl_file_extent_same(struct file *file,
2875 struct btrfs_ioctl_same_args __user *argp)
2877 struct btrfs_ioctl_same_args *same;
2878 struct btrfs_ioctl_same_extent_info *info;
2879 struct inode *src = file_inode(file);
2885 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
2886 bool is_admin = capable(CAP_SYS_ADMIN);
2889 if (!(file->f_mode & FMODE_READ))
2892 ret = mnt_want_write_file(file);
2896 if (get_user(count, &argp->dest_count)) {
2901 size = offsetof(struct btrfs_ioctl_same_args __user, info[count]);
2903 same = memdup_user(argp, size);
2906 ret = PTR_ERR(same);
2910 off = same->logical_offset;
2914 * Limit the total length we will dedupe for each operation.
2915 * This is intended to bound the total time spent in this
2916 * ioctl to something sane.
2918 if (len > BTRFS_MAX_DEDUPE_LEN)
2919 len = BTRFS_MAX_DEDUPE_LEN;
2921 if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) {
2923 * Btrfs does not support blocksize < page_size. As a
2924 * result, btrfs_cmp_data() won't correctly handle
2925 * this situation without an update.
2932 if (S_ISDIR(src->i_mode))
2936 if (!S_ISREG(src->i_mode))
2939 /* pre-format output fields to sane values */
2940 for (i = 0; i < count; i++) {
2941 same->info[i].bytes_deduped = 0ULL;
2942 same->info[i].status = 0;
2945 for (i = 0, info = same->info; i < count; i++, info++) {
2947 struct fd dst_file = fdget(info->fd);
2948 if (!dst_file.file) {
2949 info->status = -EBADF;
2952 dst = file_inode(dst_file.file);
2954 if (!(is_admin || (dst_file.file->f_mode & FMODE_WRITE))) {
2955 info->status = -EINVAL;
2956 } else if (file->f_path.mnt != dst_file.file->f_path.mnt) {
2957 info->status = -EXDEV;
2958 } else if (S_ISDIR(dst->i_mode)) {
2959 info->status = -EISDIR;
2960 } else if (!S_ISREG(dst->i_mode)) {
2961 info->status = -EACCES;
2963 info->status = btrfs_extent_same(src, off, len, dst,
2964 info->logical_offset);
2965 if (info->status == 0)
2966 info->bytes_deduped += len;
2971 ret = copy_to_user(argp, same, size);
2976 mnt_drop_write_file(file);
2980 /* Helper to check and see if this root currently has a ref on the given disk
2981 * bytenr. If it does then we need to update the quota for this root. This
2982 * doesn't do anything if quotas aren't enabled.
2984 static int check_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
2987 struct seq_list tree_mod_seq_elem = {};
2988 struct ulist *roots;
2989 struct ulist_iterator uiter;
2990 struct ulist_node *root_node = NULL;
2993 if (!root->fs_info->quota_enabled)
2996 btrfs_get_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
2997 ret = btrfs_find_all_roots(trans, root->fs_info, disko,
2998 tree_mod_seq_elem.seq, &roots);
3002 ULIST_ITER_INIT(&uiter);
3003 while ((root_node = ulist_next(roots, &uiter))) {
3004 if (root_node->val == root->objectid) {
3011 btrfs_put_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3015 static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3016 struct inode *inode,
3021 struct btrfs_root *root = BTRFS_I(inode)->root;
3024 inode_inc_iversion(inode);
3025 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3027 * We round up to the block size at eof when determining which
3028 * extents to clone above, but shouldn't round up the file size.
3030 if (endoff > destoff + olen)
3031 endoff = destoff + olen;
3032 if (endoff > inode->i_size)
3033 btrfs_i_size_write(inode, endoff);
3035 ret = btrfs_update_inode(trans, root, inode);
3037 btrfs_abort_transaction(trans, root, ret);
3038 btrfs_end_transaction(trans, root);
3041 ret = btrfs_end_transaction(trans, root);
3046 static void clone_update_extent_map(struct inode *inode,
3047 const struct btrfs_trans_handle *trans,
3048 const struct btrfs_path *path,
3049 struct btrfs_file_extent_item *fi,
3050 const u64 hole_offset,
3053 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
3054 struct extent_map *em;
3057 em = alloc_extent_map();
3059 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3060 &BTRFS_I(inode)->runtime_flags);
3065 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3066 em->generation = -1;
3067 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3068 BTRFS_FILE_EXTENT_INLINE)
3069 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3070 &BTRFS_I(inode)->runtime_flags);
3072 em->start = hole_offset;
3074 em->ram_bytes = em->len;
3075 em->orig_start = hole_offset;
3076 em->block_start = EXTENT_MAP_HOLE;
3078 em->orig_block_len = 0;
3079 em->compress_type = BTRFS_COMPRESS_NONE;
3080 em->generation = trans->transid;
3084 write_lock(&em_tree->lock);
3085 ret = add_extent_mapping(em_tree, em, 1);
3086 write_unlock(&em_tree->lock);
3087 if (ret != -EEXIST) {
3088 free_extent_map(em);
3091 btrfs_drop_extent_cache(inode, em->start,
3092 em->start + em->len - 1, 0);
3096 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3097 &BTRFS_I(inode)->runtime_flags);
3101 * btrfs_clone() - clone a range from inode file to another
3103 * @src: Inode to clone from
3104 * @inode: Inode to clone to
3105 * @off: Offset within source to start clone from
3106 * @olen: Original length, passed by user, of range to clone
3107 * @olen_aligned: Block-aligned value of olen, extent_same uses
3108 * identical values here
3109 * @destoff: Offset within @inode to start clone
3111 static int btrfs_clone(struct inode *src, struct inode *inode,
3112 const u64 off, const u64 olen, const u64 olen_aligned,
3115 struct btrfs_root *root = BTRFS_I(inode)->root;
3116 struct btrfs_path *path = NULL;
3117 struct extent_buffer *leaf;
3118 struct btrfs_trans_handle *trans;
3120 struct btrfs_key key;
3125 const u64 len = olen_aligned;
3127 u64 last_dest_end = destoff;
3130 buf = vmalloc(btrfs_level_size(root, 0));
3134 path = btrfs_alloc_path();
3142 key.objectid = btrfs_ino(src);
3143 key.type = BTRFS_EXTENT_DATA_KEY;
3148 * note the key will change type as we walk through the
3151 path->leave_spinning = 1;
3152 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3157 * First search, if no extent item that starts at offset off was
3158 * found but the previous item is an extent item, it's possible
3159 * it might overlap our target range, therefore process it.
3161 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3162 btrfs_item_key_to_cpu(path->nodes[0], &key,
3163 path->slots[0] - 1);
3164 if (key.type == BTRFS_EXTENT_DATA_KEY)
3168 nritems = btrfs_header_nritems(path->nodes[0]);
3171 if (path->slots[0] >= nritems) {
3172 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
3177 nritems = btrfs_header_nritems(path->nodes[0]);
3179 leaf = path->nodes[0];
3180 slot = path->slots[0];
3182 btrfs_item_key_to_cpu(leaf, &key, slot);
3183 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
3184 key.objectid != btrfs_ino(src))
3187 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
3188 struct btrfs_file_extent_item *extent;
3191 struct btrfs_key new_key;
3192 u64 disko = 0, diskl = 0;
3193 u64 datao = 0, datal = 0;
3197 extent = btrfs_item_ptr(leaf, slot,
3198 struct btrfs_file_extent_item);
3199 comp = btrfs_file_extent_compression(leaf, extent);
3200 type = btrfs_file_extent_type(leaf, extent);
3201 if (type == BTRFS_FILE_EXTENT_REG ||
3202 type == BTRFS_FILE_EXTENT_PREALLOC) {
3203 disko = btrfs_file_extent_disk_bytenr(leaf,
3205 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3207 datao = btrfs_file_extent_offset(leaf, extent);
3208 datal = btrfs_file_extent_num_bytes(leaf,
3210 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3211 /* take upper bound, may be compressed */
3212 datal = btrfs_file_extent_ram_bytes(leaf,
3217 * The first search might have left us at an extent
3218 * item that ends before our target range's start, can
3219 * happen if we have holes and NO_HOLES feature enabled.
3221 if (key.offset + datal <= off) {
3224 } else if (key.offset >= off + len) {
3228 size = btrfs_item_size_nr(leaf, slot);
3229 read_extent_buffer(leaf, buf,
3230 btrfs_item_ptr_offset(leaf, slot),
3233 btrfs_release_path(path);
3234 path->leave_spinning = 0;
3236 memcpy(&new_key, &key, sizeof(new_key));
3237 new_key.objectid = btrfs_ino(inode);
3238 if (off <= key.offset)
3239 new_key.offset = key.offset + destoff - off;
3241 new_key.offset = destoff;
3244 * Deal with a hole that doesn't have an extent item
3245 * that represents it (NO_HOLES feature enabled).
3246 * This hole is either in the middle of the cloning
3247 * range or at the beginning (fully overlaps it or
3248 * partially overlaps it).
3250 if (new_key.offset != last_dest_end)
3251 drop_start = last_dest_end;
3253 drop_start = new_key.offset;
3256 * 1 - adjusting old extent (we may have to split it)
3257 * 1 - add new extent
3260 trans = btrfs_start_transaction(root, 3);
3261 if (IS_ERR(trans)) {
3262 ret = PTR_ERR(trans);
3266 if (type == BTRFS_FILE_EXTENT_REG ||
3267 type == BTRFS_FILE_EXTENT_PREALLOC) {
3269 * a | --- range to clone ---| b
3270 * | ------------- extent ------------- |
3273 /* subtract range b */
3274 if (key.offset + datal > off + len)
3275 datal = off + len - key.offset;
3277 /* subtract range a */
3278 if (off > key.offset) {
3279 datao += off - key.offset;
3280 datal -= off - key.offset;
3283 ret = btrfs_drop_extents(trans, root, inode,
3285 new_key.offset + datal,
3288 if (ret != -EOPNOTSUPP)
3289 btrfs_abort_transaction(trans,
3291 btrfs_end_transaction(trans, root);
3295 ret = btrfs_insert_empty_item(trans, root, path,
3298 btrfs_abort_transaction(trans, root,
3300 btrfs_end_transaction(trans, root);
3304 leaf = path->nodes[0];
3305 slot = path->slots[0];
3306 write_extent_buffer(leaf, buf,
3307 btrfs_item_ptr_offset(leaf, slot),
3310 extent = btrfs_item_ptr(leaf, slot,
3311 struct btrfs_file_extent_item);
3313 /* disko == 0 means it's a hole */
3317 btrfs_set_file_extent_offset(leaf, extent,
3319 btrfs_set_file_extent_num_bytes(leaf, extent,
3323 * We need to look up the roots that point at
3324 * this bytenr and see if the new root does. If
3325 * it does not we need to make sure we update
3326 * quotas appropriately.
3328 if (disko && root != BTRFS_I(src)->root &&
3329 disko != last_disko) {
3330 no_quota = check_ref(trans, root,
3333 btrfs_abort_transaction(trans,
3336 btrfs_end_transaction(trans,
3344 inode_add_bytes(inode, datal);
3345 ret = btrfs_inc_extent_ref(trans, root,
3347 root->root_key.objectid,
3349 new_key.offset - datao,
3352 btrfs_abort_transaction(trans,
3355 btrfs_end_transaction(trans,
3361 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3364 u64 aligned_end = 0;
3366 if (off > key.offset) {
3367 skip = off - key.offset;
3368 new_key.offset += skip;
3371 if (key.offset + datal > off + len)
3372 trim = key.offset + datal - (off + len);
3374 if (comp && (skip || trim)) {
3376 btrfs_end_transaction(trans, root);
3379 size -= skip + trim;
3380 datal -= skip + trim;
3382 aligned_end = ALIGN(new_key.offset + datal,
3384 ret = btrfs_drop_extents(trans, root, inode,
3389 if (ret != -EOPNOTSUPP)
3390 btrfs_abort_transaction(trans,
3392 btrfs_end_transaction(trans, root);
3396 ret = btrfs_insert_empty_item(trans, root, path,
3399 btrfs_abort_transaction(trans, root,
3401 btrfs_end_transaction(trans, root);
3407 btrfs_file_extent_calc_inline_size(0);
3408 memmove(buf+start, buf+start+skip,
3412 leaf = path->nodes[0];
3413 slot = path->slots[0];
3414 write_extent_buffer(leaf, buf,
3415 btrfs_item_ptr_offset(leaf, slot),
3417 inode_add_bytes(inode, datal);
3418 extent = btrfs_item_ptr(leaf, slot,
3419 struct btrfs_file_extent_item);
3422 /* If we have an implicit hole (NO_HOLES feature). */
3423 if (drop_start < new_key.offset)
3424 clone_update_extent_map(inode, trans,
3425 path, NULL, drop_start,
3426 new_key.offset - drop_start);
3428 clone_update_extent_map(inode, trans, path,
3431 btrfs_mark_buffer_dirty(leaf);
3432 btrfs_release_path(path);
3434 last_dest_end = new_key.offset + datal;
3435 ret = clone_finish_inode_update(trans, inode,
3440 if (new_key.offset + datal >= destoff + len)
3443 btrfs_release_path(path);
3448 if (last_dest_end < destoff + len) {
3450 * We have an implicit hole (NO_HOLES feature is enabled) that
3451 * fully or partially overlaps our cloning range at its end.
3453 btrfs_release_path(path);
3456 * 1 - remove extent(s)
3459 trans = btrfs_start_transaction(root, 2);
3460 if (IS_ERR(trans)) {
3461 ret = PTR_ERR(trans);
3464 ret = btrfs_drop_extents(trans, root, inode,
3465 last_dest_end, destoff + len, 1);
3467 if (ret != -EOPNOTSUPP)
3468 btrfs_abort_transaction(trans, root, ret);
3469 btrfs_end_transaction(trans, root);
3472 ret = clone_finish_inode_update(trans, inode, destoff + len,
3476 clone_update_extent_map(inode, trans, path, NULL, last_dest_end,
3477 destoff + len - last_dest_end);
3481 btrfs_free_path(path);
3486 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
3487 u64 off, u64 olen, u64 destoff)
3489 struct inode *inode = file_inode(file);
3490 struct btrfs_root *root = BTRFS_I(inode)->root;
3495 u64 bs = root->fs_info->sb->s_blocksize;
3500 * - split compressed inline extents. annoying: we need to
3501 * decompress into destination's address_space (the file offset
3502 * may change, so source mapping won't do), then recompress (or
3503 * otherwise reinsert) a subrange.
3505 * - split destination inode's inline extents. The inline extents can
3506 * be either compressed or non-compressed.
3509 /* the destination must be opened for writing */
3510 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
3513 if (btrfs_root_readonly(root))
3516 ret = mnt_want_write_file(file);
3520 src_file = fdget(srcfd);
3521 if (!src_file.file) {
3523 goto out_drop_write;
3527 if (src_file.file->f_path.mnt != file->f_path.mnt)
3530 src = file_inode(src_file.file);
3536 /* the src must be open for reading */
3537 if (!(src_file.file->f_mode & FMODE_READ))
3540 /* don't make the dst file partly checksummed */
3541 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3542 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3546 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3550 if (src->i_sb != inode->i_sb)
3555 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
3556 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
3558 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
3559 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
3562 mutex_lock(&src->i_mutex);
3565 /* determine range to clone */
3567 if (off + len > src->i_size || off + len < off)
3570 olen = len = src->i_size - off;
3571 /* if we extend to eof, continue to block boundary */
3572 if (off + len == src->i_size)
3573 len = ALIGN(src->i_size, bs) - off;
3575 /* verify the end result is block aligned */
3576 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3577 !IS_ALIGNED(destoff, bs))
3580 /* verify if ranges are overlapped within the same file */
3582 if (destoff + len > off && destoff < off + len)
3586 if (destoff > inode->i_size) {
3587 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3593 * Lock the target range too. Right after we replace the file extent
3594 * items in the fs tree (which now point to the cloned data), we might
3595 * have a worker replace them with extent items relative to a write
3596 * operation that was issued before this clone operation (i.e. confront
3597 * with inode.c:btrfs_finish_ordered_io).
3600 u64 lock_start = min_t(u64, off, destoff);
3601 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
3603 lock_extent_range(src, lock_start, lock_len);
3605 lock_extent_range(src, off, len);
3606 lock_extent_range(inode, destoff, len);
3609 ret = btrfs_clone(src, inode, off, olen, len, destoff);
3612 u64 lock_start = min_t(u64, off, destoff);
3613 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3615 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3617 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
3618 unlock_extent(&BTRFS_I(inode)->io_tree, destoff,
3622 * Truncate page cache pages so that future reads will see the cloned
3623 * data immediately and not the previous data.
3625 truncate_inode_pages_range(&inode->i_data, destoff,
3626 PAGE_CACHE_ALIGN(destoff + len) - 1);
3630 mutex_unlock(&src->i_mutex);
3631 mutex_unlock(&inode->i_mutex);
3633 mutex_unlock(&inode->i_mutex);
3634 mutex_unlock(&src->i_mutex);
3637 mutex_unlock(&src->i_mutex);
3642 mnt_drop_write_file(file);
3646 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
3648 struct btrfs_ioctl_clone_range_args args;
3650 if (copy_from_user(&args, argp, sizeof(args)))
3652 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
3653 args.src_length, args.dest_offset);
3657 * there are many ways the trans_start and trans_end ioctls can lead
3658 * to deadlocks. They should only be used by applications that
3659 * basically own the machine, and have a very in depth understanding
3660 * of all the possible deadlocks and enospc problems.
3662 static long btrfs_ioctl_trans_start(struct file *file)
3664 struct inode *inode = file_inode(file);
3665 struct btrfs_root *root = BTRFS_I(inode)->root;
3666 struct btrfs_trans_handle *trans;
3670 if (!capable(CAP_SYS_ADMIN))
3674 if (file->private_data)
3678 if (btrfs_root_readonly(root))
3681 ret = mnt_want_write_file(file);
3685 atomic_inc(&root->fs_info->open_ioctl_trans);
3688 trans = btrfs_start_ioctl_transaction(root);
3692 file->private_data = trans;
3696 atomic_dec(&root->fs_info->open_ioctl_trans);
3697 mnt_drop_write_file(file);
3702 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3704 struct inode *inode = file_inode(file);
3705 struct btrfs_root *root = BTRFS_I(inode)->root;
3706 struct btrfs_root *new_root;
3707 struct btrfs_dir_item *di;
3708 struct btrfs_trans_handle *trans;
3709 struct btrfs_path *path;
3710 struct btrfs_key location;
3711 struct btrfs_disk_key disk_key;
3716 if (!capable(CAP_SYS_ADMIN))
3719 ret = mnt_want_write_file(file);
3723 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3729 objectid = BTRFS_FS_TREE_OBJECTID;
3731 location.objectid = objectid;
3732 location.type = BTRFS_ROOT_ITEM_KEY;
3733 location.offset = (u64)-1;
3735 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
3736 if (IS_ERR(new_root)) {
3737 ret = PTR_ERR(new_root);
3741 path = btrfs_alloc_path();
3746 path->leave_spinning = 1;
3748 trans = btrfs_start_transaction(root, 1);
3749 if (IS_ERR(trans)) {
3750 btrfs_free_path(path);
3751 ret = PTR_ERR(trans);
3755 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
3756 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
3757 dir_id, "default", 7, 1);
3758 if (IS_ERR_OR_NULL(di)) {
3759 btrfs_free_path(path);
3760 btrfs_end_transaction(trans, root);
3761 btrfs_err(new_root->fs_info, "Umm, you don't have the default dir"
3762 "item, this isn't going to work");
3767 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3768 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3769 btrfs_mark_buffer_dirty(path->nodes[0]);
3770 btrfs_free_path(path);
3772 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
3773 btrfs_end_transaction(trans, root);
3775 mnt_drop_write_file(file);
3779 void btrfs_get_block_group_info(struct list_head *groups_list,
3780 struct btrfs_ioctl_space_info *space)
3782 struct btrfs_block_group_cache *block_group;
3784 space->total_bytes = 0;
3785 space->used_bytes = 0;
3787 list_for_each_entry(block_group, groups_list, list) {
3788 space->flags = block_group->flags;
3789 space->total_bytes += block_group->key.offset;
3790 space->used_bytes +=
3791 btrfs_block_group_used(&block_group->item);
3795 static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
3797 struct btrfs_ioctl_space_args space_args;
3798 struct btrfs_ioctl_space_info space;
3799 struct btrfs_ioctl_space_info *dest;
3800 struct btrfs_ioctl_space_info *dest_orig;
3801 struct btrfs_ioctl_space_info __user *user_dest;
3802 struct btrfs_space_info *info;
3803 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3804 BTRFS_BLOCK_GROUP_SYSTEM,
3805 BTRFS_BLOCK_GROUP_METADATA,
3806 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3813 if (copy_from_user(&space_args,
3814 (struct btrfs_ioctl_space_args __user *)arg,
3815 sizeof(space_args)))
3818 for (i = 0; i < num_types; i++) {
3819 struct btrfs_space_info *tmp;
3823 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3825 if (tmp->flags == types[i]) {
3835 down_read(&info->groups_sem);
3836 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3837 if (!list_empty(&info->block_groups[c]))
3840 up_read(&info->groups_sem);
3844 * Global block reserve, exported as a space_info
3848 /* space_slots == 0 means they are asking for a count */
3849 if (space_args.space_slots == 0) {
3850 space_args.total_spaces = slot_count;
3854 slot_count = min_t(u64, space_args.space_slots, slot_count);
3856 alloc_size = sizeof(*dest) * slot_count;
3858 /* we generally have at most 6 or so space infos, one for each raid
3859 * level. So, a whole page should be more than enough for everyone
3861 if (alloc_size > PAGE_CACHE_SIZE)
3864 space_args.total_spaces = 0;
3865 dest = kmalloc(alloc_size, GFP_NOFS);
3870 /* now we have a buffer to copy into */
3871 for (i = 0; i < num_types; i++) {
3872 struct btrfs_space_info *tmp;
3879 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3881 if (tmp->flags == types[i]) {
3890 down_read(&info->groups_sem);
3891 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3892 if (!list_empty(&info->block_groups[c])) {
3893 btrfs_get_block_group_info(
3894 &info->block_groups[c], &space);
3895 memcpy(dest, &space, sizeof(space));
3897 space_args.total_spaces++;
3903 up_read(&info->groups_sem);
3907 * Add global block reserve
3910 struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv;
3912 spin_lock(&block_rsv->lock);
3913 space.total_bytes = block_rsv->size;
3914 space.used_bytes = block_rsv->size - block_rsv->reserved;
3915 spin_unlock(&block_rsv->lock);
3916 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3917 memcpy(dest, &space, sizeof(space));
3918 space_args.total_spaces++;
3921 user_dest = (struct btrfs_ioctl_space_info __user *)
3922 (arg + sizeof(struct btrfs_ioctl_space_args));
3924 if (copy_to_user(user_dest, dest_orig, alloc_size))
3929 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3936 * there are many ways the trans_start and trans_end ioctls can lead
3937 * to deadlocks. They should only be used by applications that
3938 * basically own the machine, and have a very in depth understanding
3939 * of all the possible deadlocks and enospc problems.
3941 long btrfs_ioctl_trans_end(struct file *file)
3943 struct inode *inode = file_inode(file);
3944 struct btrfs_root *root = BTRFS_I(inode)->root;
3945 struct btrfs_trans_handle *trans;
3947 trans = file->private_data;
3950 file->private_data = NULL;
3952 btrfs_end_transaction(trans, root);
3954 atomic_dec(&root->fs_info->open_ioctl_trans);
3956 mnt_drop_write_file(file);
3960 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3963 struct btrfs_trans_handle *trans;
3967 trans = btrfs_attach_transaction_barrier(root);
3968 if (IS_ERR(trans)) {
3969 if (PTR_ERR(trans) != -ENOENT)
3970 return PTR_ERR(trans);
3972 /* No running transaction, don't bother */
3973 transid = root->fs_info->last_trans_committed;
3976 transid = trans->transid;
3977 ret = btrfs_commit_transaction_async(trans, root, 0);
3979 btrfs_end_transaction(trans, root);
3984 if (copy_to_user(argp, &transid, sizeof(transid)))
3989 static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
3995 if (copy_from_user(&transid, argp, sizeof(transid)))
3998 transid = 0; /* current trans */
4000 return btrfs_wait_for_commit(root, transid);
4003 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
4005 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4006 struct btrfs_ioctl_scrub_args *sa;
4009 if (!capable(CAP_SYS_ADMIN))
4012 sa = memdup_user(arg, sizeof(*sa));
4016 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4017 ret = mnt_want_write_file(file);
4022 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
4023 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4026 if (copy_to_user(arg, sa, sizeof(*sa)))
4029 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4030 mnt_drop_write_file(file);
4036 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
4038 if (!capable(CAP_SYS_ADMIN))
4041 return btrfs_scrub_cancel(root->fs_info);
4044 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
4047 struct btrfs_ioctl_scrub_args *sa;
4050 if (!capable(CAP_SYS_ADMIN))
4053 sa = memdup_user(arg, sizeof(*sa));
4057 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
4059 if (copy_to_user(arg, sa, sizeof(*sa)))
4066 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
4069 struct btrfs_ioctl_get_dev_stats *sa;
4072 sa = memdup_user(arg, sizeof(*sa));
4076 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4081 ret = btrfs_get_dev_stats(root, sa);
4083 if (copy_to_user(arg, sa, sizeof(*sa)))
4090 static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
4092 struct btrfs_ioctl_dev_replace_args *p;
4095 if (!capable(CAP_SYS_ADMIN))
4098 p = memdup_user(arg, sizeof(*p));
4103 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
4104 if (root->fs_info->sb->s_flags & MS_RDONLY) {
4109 &root->fs_info->mutually_exclusive_operation_running,
4111 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4113 ret = btrfs_dev_replace_start(root, p);
4115 &root->fs_info->mutually_exclusive_operation_running,
4119 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4120 btrfs_dev_replace_status(root->fs_info, p);
4123 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4124 ret = btrfs_dev_replace_cancel(root->fs_info, p);
4131 if (copy_to_user(arg, p, sizeof(*p)))
4138 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4144 struct btrfs_ioctl_ino_path_args *ipa = NULL;
4145 struct inode_fs_paths *ipath = NULL;
4146 struct btrfs_path *path;
4148 if (!capable(CAP_DAC_READ_SEARCH))
4151 path = btrfs_alloc_path();
4157 ipa = memdup_user(arg, sizeof(*ipa));
4164 size = min_t(u32, ipa->size, 4096);
4165 ipath = init_ipath(size, root, path);
4166 if (IS_ERR(ipath)) {
4167 ret = PTR_ERR(ipath);
4172 ret = paths_from_inode(ipa->inum, ipath);
4176 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
4177 rel_ptr = ipath->fspath->val[i] -
4178 (u64)(unsigned long)ipath->fspath->val;
4179 ipath->fspath->val[i] = rel_ptr;
4182 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
4183 (void *)(unsigned long)ipath->fspath, size);
4190 btrfs_free_path(path);
4197 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4199 struct btrfs_data_container *inodes = ctx;
4200 const size_t c = 3 * sizeof(u64);
4202 if (inodes->bytes_left >= c) {
4203 inodes->bytes_left -= c;
4204 inodes->val[inodes->elem_cnt] = inum;
4205 inodes->val[inodes->elem_cnt + 1] = offset;
4206 inodes->val[inodes->elem_cnt + 2] = root;
4207 inodes->elem_cnt += 3;
4209 inodes->bytes_missing += c - inodes->bytes_left;
4210 inodes->bytes_left = 0;
4211 inodes->elem_missed += 3;
4217 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
4222 struct btrfs_ioctl_logical_ino_args *loi;
4223 struct btrfs_data_container *inodes = NULL;
4224 struct btrfs_path *path = NULL;
4226 if (!capable(CAP_SYS_ADMIN))
4229 loi = memdup_user(arg, sizeof(*loi));
4236 path = btrfs_alloc_path();
4242 size = min_t(u32, loi->size, 64 * 1024);
4243 inodes = init_data_container(size);
4244 if (IS_ERR(inodes)) {
4245 ret = PTR_ERR(inodes);
4250 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
4251 build_ino_list, inodes);
4257 ret = copy_to_user((void *)(unsigned long)loi->inodes,
4258 (void *)(unsigned long)inodes, size);
4263 btrfs_free_path(path);
4270 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
4271 struct btrfs_ioctl_balance_args *bargs)
4273 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4275 bargs->flags = bctl->flags;
4277 if (atomic_read(&fs_info->balance_running))
4278 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4279 if (atomic_read(&fs_info->balance_pause_req))
4280 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
4281 if (atomic_read(&fs_info->balance_cancel_req))
4282 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
4284 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4285 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4286 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4289 spin_lock(&fs_info->balance_lock);
4290 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4291 spin_unlock(&fs_info->balance_lock);
4293 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4297 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4299 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4300 struct btrfs_fs_info *fs_info = root->fs_info;
4301 struct btrfs_ioctl_balance_args *bargs;
4302 struct btrfs_balance_control *bctl;
4303 bool need_unlock; /* for mut. excl. ops lock */
4306 if (!capable(CAP_SYS_ADMIN))
4309 ret = mnt_want_write_file(file);
4314 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
4315 mutex_lock(&fs_info->volume_mutex);
4316 mutex_lock(&fs_info->balance_mutex);
4322 * mut. excl. ops lock is locked. Three possibilites:
4323 * (1) some other op is running
4324 * (2) balance is running
4325 * (3) balance is paused -- special case (think resume)
4327 mutex_lock(&fs_info->balance_mutex);
4328 if (fs_info->balance_ctl) {
4329 /* this is either (2) or (3) */
4330 if (!atomic_read(&fs_info->balance_running)) {
4331 mutex_unlock(&fs_info->balance_mutex);
4332 if (!mutex_trylock(&fs_info->volume_mutex))
4334 mutex_lock(&fs_info->balance_mutex);
4336 if (fs_info->balance_ctl &&
4337 !atomic_read(&fs_info->balance_running)) {
4339 need_unlock = false;
4343 mutex_unlock(&fs_info->balance_mutex);
4344 mutex_unlock(&fs_info->volume_mutex);
4348 mutex_unlock(&fs_info->balance_mutex);
4354 mutex_unlock(&fs_info->balance_mutex);
4355 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4360 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
4363 bargs = memdup_user(arg, sizeof(*bargs));
4364 if (IS_ERR(bargs)) {
4365 ret = PTR_ERR(bargs);
4369 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4370 if (!fs_info->balance_ctl) {
4375 bctl = fs_info->balance_ctl;
4376 spin_lock(&fs_info->balance_lock);
4377 bctl->flags |= BTRFS_BALANCE_RESUME;
4378 spin_unlock(&fs_info->balance_lock);
4386 if (fs_info->balance_ctl) {
4391 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
4397 bctl->fs_info = fs_info;
4399 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4400 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4401 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4403 bctl->flags = bargs->flags;
4405 /* balance everything - no filters */
4406 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4411 * Ownership of bctl and mutually_exclusive_operation_running
4412 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4413 * or, if restriper was paused all the way until unmount, in
4414 * free_fs_info. mutually_exclusive_operation_running is
4415 * cleared in __cancel_balance.
4417 need_unlock = false;
4419 ret = btrfs_balance(bctl, bargs);
4422 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4429 mutex_unlock(&fs_info->balance_mutex);
4430 mutex_unlock(&fs_info->volume_mutex);
4432 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4434 mnt_drop_write_file(file);
4438 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4440 if (!capable(CAP_SYS_ADMIN))
4444 case BTRFS_BALANCE_CTL_PAUSE:
4445 return btrfs_pause_balance(root->fs_info);
4446 case BTRFS_BALANCE_CTL_CANCEL:
4447 return btrfs_cancel_balance(root->fs_info);
4453 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4456 struct btrfs_fs_info *fs_info = root->fs_info;
4457 struct btrfs_ioctl_balance_args *bargs;
4460 if (!capable(CAP_SYS_ADMIN))
4463 mutex_lock(&fs_info->balance_mutex);
4464 if (!fs_info->balance_ctl) {
4469 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
4475 update_ioctl_balance_args(fs_info, 1, bargs);
4477 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4482 mutex_unlock(&fs_info->balance_mutex);
4486 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4488 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4489 struct btrfs_ioctl_quota_ctl_args *sa;
4490 struct btrfs_trans_handle *trans = NULL;
4494 if (!capable(CAP_SYS_ADMIN))
4497 ret = mnt_want_write_file(file);
4501 sa = memdup_user(arg, sizeof(*sa));
4507 down_write(&root->fs_info->subvol_sem);
4508 trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4509 if (IS_ERR(trans)) {
4510 ret = PTR_ERR(trans);
4515 case BTRFS_QUOTA_CTL_ENABLE:
4516 ret = btrfs_quota_enable(trans, root->fs_info);
4518 case BTRFS_QUOTA_CTL_DISABLE:
4519 ret = btrfs_quota_disable(trans, root->fs_info);
4526 err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4531 up_write(&root->fs_info->subvol_sem);
4533 mnt_drop_write_file(file);
4537 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4539 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4540 struct btrfs_ioctl_qgroup_assign_args *sa;
4541 struct btrfs_trans_handle *trans;
4545 if (!capable(CAP_SYS_ADMIN))
4548 ret = mnt_want_write_file(file);
4552 sa = memdup_user(arg, sizeof(*sa));
4558 trans = btrfs_join_transaction(root);
4559 if (IS_ERR(trans)) {
4560 ret = PTR_ERR(trans);
4564 /* FIXME: check if the IDs really exist */
4566 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4569 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4573 err = btrfs_end_transaction(trans, root);
4580 mnt_drop_write_file(file);
4584 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4586 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4587 struct btrfs_ioctl_qgroup_create_args *sa;
4588 struct btrfs_trans_handle *trans;
4592 if (!capable(CAP_SYS_ADMIN))
4595 ret = mnt_want_write_file(file);
4599 sa = memdup_user(arg, sizeof(*sa));
4605 if (!sa->qgroupid) {
4610 trans = btrfs_join_transaction(root);
4611 if (IS_ERR(trans)) {
4612 ret = PTR_ERR(trans);
4616 /* FIXME: check if the IDs really exist */
4618 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
4621 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4624 err = btrfs_end_transaction(trans, root);
4631 mnt_drop_write_file(file);
4635 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4637 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4638 struct btrfs_ioctl_qgroup_limit_args *sa;
4639 struct btrfs_trans_handle *trans;
4644 if (!capable(CAP_SYS_ADMIN))
4647 ret = mnt_want_write_file(file);
4651 sa = memdup_user(arg, sizeof(*sa));
4657 trans = btrfs_join_transaction(root);
4658 if (IS_ERR(trans)) {
4659 ret = PTR_ERR(trans);
4663 qgroupid = sa->qgroupid;
4665 /* take the current subvol as qgroup */
4666 qgroupid = root->root_key.objectid;
4669 /* FIXME: check if the IDs really exist */
4670 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
4672 err = btrfs_end_transaction(trans, root);
4679 mnt_drop_write_file(file);
4683 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4685 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4686 struct btrfs_ioctl_quota_rescan_args *qsa;
4689 if (!capable(CAP_SYS_ADMIN))
4692 ret = mnt_want_write_file(file);
4696 qsa = memdup_user(arg, sizeof(*qsa));
4707 ret = btrfs_qgroup_rescan(root->fs_info);
4712 mnt_drop_write_file(file);
4716 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
4718 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4719 struct btrfs_ioctl_quota_rescan_args *qsa;
4722 if (!capable(CAP_SYS_ADMIN))
4725 qsa = kzalloc(sizeof(*qsa), GFP_NOFS);
4729 if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4731 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
4734 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4741 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
4743 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4745 if (!capable(CAP_SYS_ADMIN))
4748 return btrfs_qgroup_wait_for_completion(root->fs_info);
4751 static long _btrfs_ioctl_set_received_subvol(struct file *file,
4752 struct btrfs_ioctl_received_subvol_args *sa)
4754 struct inode *inode = file_inode(file);
4755 struct btrfs_root *root = BTRFS_I(inode)->root;
4756 struct btrfs_root_item *root_item = &root->root_item;
4757 struct btrfs_trans_handle *trans;
4758 struct timespec ct = CURRENT_TIME;
4760 int received_uuid_changed;
4762 if (!inode_owner_or_capable(inode))
4765 ret = mnt_want_write_file(file);
4769 down_write(&root->fs_info->subvol_sem);
4771 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
4776 if (btrfs_root_readonly(root)) {
4783 * 2 - uuid items (received uuid + subvol uuid)
4785 trans = btrfs_start_transaction(root, 3);
4786 if (IS_ERR(trans)) {
4787 ret = PTR_ERR(trans);
4792 sa->rtransid = trans->transid;
4793 sa->rtime.sec = ct.tv_sec;
4794 sa->rtime.nsec = ct.tv_nsec;
4796 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4798 if (received_uuid_changed &&
4799 !btrfs_is_empty_uuid(root_item->received_uuid))
4800 btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
4801 root_item->received_uuid,
4802 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4803 root->root_key.objectid);
4804 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4805 btrfs_set_root_stransid(root_item, sa->stransid);
4806 btrfs_set_root_rtransid(root_item, sa->rtransid);
4807 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4808 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4809 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4810 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4812 ret = btrfs_update_root(trans, root->fs_info->tree_root,
4813 &root->root_key, &root->root_item);
4815 btrfs_end_transaction(trans, root);
4818 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4819 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
4821 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4822 root->root_key.objectid);
4823 if (ret < 0 && ret != -EEXIST) {
4824 btrfs_abort_transaction(trans, root, ret);
4828 ret = btrfs_commit_transaction(trans, root);
4830 btrfs_abort_transaction(trans, root, ret);
4835 up_write(&root->fs_info->subvol_sem);
4836 mnt_drop_write_file(file);
4841 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4844 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4845 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4848 args32 = memdup_user(arg, sizeof(*args32));
4849 if (IS_ERR(args32)) {
4850 ret = PTR_ERR(args32);
4855 args64 = kmalloc(sizeof(*args64), GFP_NOFS);
4861 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4862 args64->stransid = args32->stransid;
4863 args64->rtransid = args32->rtransid;
4864 args64->stime.sec = args32->stime.sec;
4865 args64->stime.nsec = args32->stime.nsec;
4866 args64->rtime.sec = args32->rtime.sec;
4867 args64->rtime.nsec = args32->rtime.nsec;
4868 args64->flags = args32->flags;
4870 ret = _btrfs_ioctl_set_received_subvol(file, args64);
4874 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4875 args32->stransid = args64->stransid;
4876 args32->rtransid = args64->rtransid;
4877 args32->stime.sec = args64->stime.sec;
4878 args32->stime.nsec = args64->stime.nsec;
4879 args32->rtime.sec = args64->rtime.sec;
4880 args32->rtime.nsec = args64->rtime.nsec;
4881 args32->flags = args64->flags;
4883 ret = copy_to_user(arg, args32, sizeof(*args32));
4894 static long btrfs_ioctl_set_received_subvol(struct file *file,
4897 struct btrfs_ioctl_received_subvol_args *sa = NULL;
4900 sa = memdup_user(arg, sizeof(*sa));
4907 ret = _btrfs_ioctl_set_received_subvol(file, sa);
4912 ret = copy_to_user(arg, sa, sizeof(*sa));
4921 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
4923 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4926 char label[BTRFS_LABEL_SIZE];
4928 spin_lock(&root->fs_info->super_lock);
4929 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4930 spin_unlock(&root->fs_info->super_lock);
4932 len = strnlen(label, BTRFS_LABEL_SIZE);
4934 if (len == BTRFS_LABEL_SIZE) {
4935 btrfs_warn(root->fs_info,
4936 "label is too long, return the first %zu bytes", --len);
4939 ret = copy_to_user(arg, label, len);
4941 return ret ? -EFAULT : 0;
4944 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4946 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4947 struct btrfs_super_block *super_block = root->fs_info->super_copy;
4948 struct btrfs_trans_handle *trans;
4949 char label[BTRFS_LABEL_SIZE];
4952 if (!capable(CAP_SYS_ADMIN))
4955 if (copy_from_user(label, arg, sizeof(label)))
4958 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
4959 btrfs_err(root->fs_info, "unable to set label with more than %d bytes",
4960 BTRFS_LABEL_SIZE - 1);
4964 ret = mnt_want_write_file(file);
4968 trans = btrfs_start_transaction(root, 0);
4969 if (IS_ERR(trans)) {
4970 ret = PTR_ERR(trans);
4974 spin_lock(&root->fs_info->super_lock);
4975 strcpy(super_block->label, label);
4976 spin_unlock(&root->fs_info->super_lock);
4977 ret = btrfs_commit_transaction(trans, root);
4980 mnt_drop_write_file(file);
4984 #define INIT_FEATURE_FLAGS(suffix) \
4985 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
4986 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
4987 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
4989 static int btrfs_ioctl_get_supported_features(struct file *file,
4992 static struct btrfs_ioctl_feature_flags features[3] = {
4993 INIT_FEATURE_FLAGS(SUPP),
4994 INIT_FEATURE_FLAGS(SAFE_SET),
4995 INIT_FEATURE_FLAGS(SAFE_CLEAR)
4998 if (copy_to_user(arg, &features, sizeof(features)))
5004 static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5006 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5007 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5008 struct btrfs_ioctl_feature_flags features;
5010 features.compat_flags = btrfs_super_compat_flags(super_block);
5011 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5012 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5014 if (copy_to_user(arg, &features, sizeof(features)))
5020 static int check_feature_bits(struct btrfs_root *root,
5021 enum btrfs_feature_set set,
5022 u64 change_mask, u64 flags, u64 supported_flags,
5023 u64 safe_set, u64 safe_clear)
5025 const char *type = btrfs_feature_set_names[set];
5027 u64 disallowed, unsupported;
5028 u64 set_mask = flags & change_mask;
5029 u64 clear_mask = ~flags & change_mask;
5031 unsupported = set_mask & ~supported_flags;
5033 names = btrfs_printable_features(set, unsupported);
5035 btrfs_warn(root->fs_info,
5036 "this kernel does not support the %s feature bit%s",
5037 names, strchr(names, ',') ? "s" : "");
5040 btrfs_warn(root->fs_info,
5041 "this kernel does not support %s bits 0x%llx",
5046 disallowed = set_mask & ~safe_set;
5048 names = btrfs_printable_features(set, disallowed);
5050 btrfs_warn(root->fs_info,
5051 "can't set the %s feature bit%s while mounted",
5052 names, strchr(names, ',') ? "s" : "");
5055 btrfs_warn(root->fs_info,
5056 "can't set %s bits 0x%llx while mounted",
5061 disallowed = clear_mask & ~safe_clear;
5063 names = btrfs_printable_features(set, disallowed);
5065 btrfs_warn(root->fs_info,
5066 "can't clear the %s feature bit%s while mounted",
5067 names, strchr(names, ',') ? "s" : "");
5070 btrfs_warn(root->fs_info,
5071 "can't clear %s bits 0x%llx while mounted",
5079 #define check_feature(root, change_mask, flags, mask_base) \
5080 check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
5081 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5082 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5083 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5085 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5087 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5088 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5089 struct btrfs_ioctl_feature_flags flags[2];
5090 struct btrfs_trans_handle *trans;
5094 if (!capable(CAP_SYS_ADMIN))
5097 if (copy_from_user(flags, arg, sizeof(flags)))
5101 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5102 !flags[0].incompat_flags)
5105 ret = check_feature(root, flags[0].compat_flags,
5106 flags[1].compat_flags, COMPAT);
5110 ret = check_feature(root, flags[0].compat_ro_flags,
5111 flags[1].compat_ro_flags, COMPAT_RO);
5115 ret = check_feature(root, flags[0].incompat_flags,
5116 flags[1].incompat_flags, INCOMPAT);
5120 trans = btrfs_start_transaction(root, 0);
5122 return PTR_ERR(trans);
5124 spin_lock(&root->fs_info->super_lock);
5125 newflags = btrfs_super_compat_flags(super_block);
5126 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5127 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5128 btrfs_set_super_compat_flags(super_block, newflags);
5130 newflags = btrfs_super_compat_ro_flags(super_block);
5131 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5132 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5133 btrfs_set_super_compat_ro_flags(super_block, newflags);
5135 newflags = btrfs_super_incompat_flags(super_block);
5136 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5137 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5138 btrfs_set_super_incompat_flags(super_block, newflags);
5139 spin_unlock(&root->fs_info->super_lock);
5141 return btrfs_commit_transaction(trans, root);
5144 long btrfs_ioctl(struct file *file, unsigned int
5145 cmd, unsigned long arg)
5147 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5148 void __user *argp = (void __user *)arg;
5151 case FS_IOC_GETFLAGS:
5152 return btrfs_ioctl_getflags(file, argp);
5153 case FS_IOC_SETFLAGS:
5154 return btrfs_ioctl_setflags(file, argp);
5155 case FS_IOC_GETVERSION:
5156 return btrfs_ioctl_getversion(file, argp);
5158 return btrfs_ioctl_fitrim(file, argp);
5159 case BTRFS_IOC_SNAP_CREATE:
5160 return btrfs_ioctl_snap_create(file, argp, 0);
5161 case BTRFS_IOC_SNAP_CREATE_V2:
5162 return btrfs_ioctl_snap_create_v2(file, argp, 0);
5163 case BTRFS_IOC_SUBVOL_CREATE:
5164 return btrfs_ioctl_snap_create(file, argp, 1);
5165 case BTRFS_IOC_SUBVOL_CREATE_V2:
5166 return btrfs_ioctl_snap_create_v2(file, argp, 1);
5167 case BTRFS_IOC_SNAP_DESTROY:
5168 return btrfs_ioctl_snap_destroy(file, argp);
5169 case BTRFS_IOC_SUBVOL_GETFLAGS:
5170 return btrfs_ioctl_subvol_getflags(file, argp);
5171 case BTRFS_IOC_SUBVOL_SETFLAGS:
5172 return btrfs_ioctl_subvol_setflags(file, argp);
5173 case BTRFS_IOC_DEFAULT_SUBVOL:
5174 return btrfs_ioctl_default_subvol(file, argp);
5175 case BTRFS_IOC_DEFRAG:
5176 return btrfs_ioctl_defrag(file, NULL);
5177 case BTRFS_IOC_DEFRAG_RANGE:
5178 return btrfs_ioctl_defrag(file, argp);
5179 case BTRFS_IOC_RESIZE:
5180 return btrfs_ioctl_resize(file, argp);
5181 case BTRFS_IOC_ADD_DEV:
5182 return btrfs_ioctl_add_dev(root, argp);
5183 case BTRFS_IOC_RM_DEV:
5184 return btrfs_ioctl_rm_dev(file, argp);
5185 case BTRFS_IOC_FS_INFO:
5186 return btrfs_ioctl_fs_info(root, argp);
5187 case BTRFS_IOC_DEV_INFO:
5188 return btrfs_ioctl_dev_info(root, argp);
5189 case BTRFS_IOC_BALANCE:
5190 return btrfs_ioctl_balance(file, NULL);
5191 case BTRFS_IOC_CLONE:
5192 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
5193 case BTRFS_IOC_CLONE_RANGE:
5194 return btrfs_ioctl_clone_range(file, argp);
5195 case BTRFS_IOC_TRANS_START:
5196 return btrfs_ioctl_trans_start(file);
5197 case BTRFS_IOC_TRANS_END:
5198 return btrfs_ioctl_trans_end(file);
5199 case BTRFS_IOC_TREE_SEARCH:
5200 return btrfs_ioctl_tree_search(file, argp);
5201 case BTRFS_IOC_INO_LOOKUP:
5202 return btrfs_ioctl_ino_lookup(file, argp);
5203 case BTRFS_IOC_INO_PATHS:
5204 return btrfs_ioctl_ino_to_path(root, argp);
5205 case BTRFS_IOC_LOGICAL_INO:
5206 return btrfs_ioctl_logical_to_ino(root, argp);
5207 case BTRFS_IOC_SPACE_INFO:
5208 return btrfs_ioctl_space_info(root, argp);
5209 case BTRFS_IOC_SYNC: {
5212 ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1);
5215 ret = btrfs_sync_fs(file->f_dentry->d_sb, 1);
5218 case BTRFS_IOC_START_SYNC:
5219 return btrfs_ioctl_start_sync(root, argp);
5220 case BTRFS_IOC_WAIT_SYNC:
5221 return btrfs_ioctl_wait_sync(root, argp);
5222 case BTRFS_IOC_SCRUB:
5223 return btrfs_ioctl_scrub(file, argp);
5224 case BTRFS_IOC_SCRUB_CANCEL:
5225 return btrfs_ioctl_scrub_cancel(root, argp);
5226 case BTRFS_IOC_SCRUB_PROGRESS:
5227 return btrfs_ioctl_scrub_progress(root, argp);
5228 case BTRFS_IOC_BALANCE_V2:
5229 return btrfs_ioctl_balance(file, argp);
5230 case BTRFS_IOC_BALANCE_CTL:
5231 return btrfs_ioctl_balance_ctl(root, arg);
5232 case BTRFS_IOC_BALANCE_PROGRESS:
5233 return btrfs_ioctl_balance_progress(root, argp);
5234 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5235 return btrfs_ioctl_set_received_subvol(file, argp);
5237 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5238 return btrfs_ioctl_set_received_subvol_32(file, argp);
5240 case BTRFS_IOC_SEND:
5241 return btrfs_ioctl_send(file, argp);
5242 case BTRFS_IOC_GET_DEV_STATS:
5243 return btrfs_ioctl_get_dev_stats(root, argp);
5244 case BTRFS_IOC_QUOTA_CTL:
5245 return btrfs_ioctl_quota_ctl(file, argp);
5246 case BTRFS_IOC_QGROUP_ASSIGN:
5247 return btrfs_ioctl_qgroup_assign(file, argp);
5248 case BTRFS_IOC_QGROUP_CREATE:
5249 return btrfs_ioctl_qgroup_create(file, argp);
5250 case BTRFS_IOC_QGROUP_LIMIT:
5251 return btrfs_ioctl_qgroup_limit(file, argp);
5252 case BTRFS_IOC_QUOTA_RESCAN:
5253 return btrfs_ioctl_quota_rescan(file, argp);
5254 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5255 return btrfs_ioctl_quota_rescan_status(file, argp);
5256 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5257 return btrfs_ioctl_quota_rescan_wait(file, argp);
5258 case BTRFS_IOC_DEV_REPLACE:
5259 return btrfs_ioctl_dev_replace(root, argp);
5260 case BTRFS_IOC_GET_FSLABEL:
5261 return btrfs_ioctl_get_fslabel(file, argp);
5262 case BTRFS_IOC_SET_FSLABEL:
5263 return btrfs_ioctl_set_fslabel(file, argp);
5264 case BTRFS_IOC_FILE_EXTENT_SAME:
5265 return btrfs_ioctl_file_extent_same(file, argp);
5266 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5267 return btrfs_ioctl_get_supported_features(file, argp);
5268 case BTRFS_IOC_GET_FEATURES:
5269 return btrfs_ioctl_get_features(file, argp);
5270 case BTRFS_IOC_SET_FEATURES:
5271 return btrfs_ioctl_set_features(file, argp);