2 * Copyright (C) 2007 Oracle. All rights reserved.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
19 #include <linux/kernel.h>
20 #include <linux/bio.h>
21 #include <linux/buffer_head.h>
22 #include <linux/file.h>
24 #include <linux/fsnotify.h>
25 #include <linux/pagemap.h>
26 #include <linux/highmem.h>
27 #include <linux/time.h>
28 #include <linux/init.h>
29 #include <linux/string.h>
30 #include <linux/backing-dev.h>
31 #include <linux/mount.h>
32 #include <linux/mpage.h>
33 #include <linux/namei.h>
34 #include <linux/swap.h>
35 #include <linux/writeback.h>
36 #include <linux/statfs.h>
37 #include <linux/compat.h>
38 #include <linux/bit_spinlock.h>
39 #include <linux/security.h>
40 #include <linux/xattr.h>
41 #include <linux/vmalloc.h>
42 #include <linux/slab.h>
43 #include <linux/blkdev.h>
44 #include <linux/uuid.h>
45 #include <linux/btrfs.h>
46 #include <linux/uaccess.h>
49 #include "transaction.h"
50 #include "btrfs_inode.h"
51 #include "print-tree.h"
54 #include "inode-map.h"
56 #include "rcu-string.h"
58 #include "dev-replace.h"
64 /* If we have a 32-bit userspace and 64-bit kernel, then the UAPI
65 * structures are incorrect, as the timespec structure from userspace
66 * is 4 bytes too small. We define these alternatives here to teach
67 * the kernel about the 32-bit struct packing.
69 struct btrfs_ioctl_timespec_32 {
72 } __attribute__ ((__packed__));
74 struct btrfs_ioctl_received_subvol_args_32 {
75 char uuid[BTRFS_UUID_SIZE]; /* in */
76 __u64 stransid; /* in */
77 __u64 rtransid; /* out */
78 struct btrfs_ioctl_timespec_32 stime; /* in */
79 struct btrfs_ioctl_timespec_32 rtime; /* out */
81 __u64 reserved[16]; /* in */
82 } __attribute__ ((__packed__));
84 #define BTRFS_IOC_SET_RECEIVED_SUBVOL_32 _IOWR(BTRFS_IOCTL_MAGIC, 37, \
85 struct btrfs_ioctl_received_subvol_args_32)
89 static int btrfs_clone(struct inode *src, struct inode *inode,
90 u64 off, u64 olen, u64 olen_aligned, u64 destoff);
92 /* Mask out flags that are inappropriate for the given type of inode. */
93 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
97 else if (S_ISREG(mode))
98 return flags & ~FS_DIRSYNC_FL;
100 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
104 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
106 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
108 unsigned int iflags = 0;
110 if (flags & BTRFS_INODE_SYNC)
111 iflags |= FS_SYNC_FL;
112 if (flags & BTRFS_INODE_IMMUTABLE)
113 iflags |= FS_IMMUTABLE_FL;
114 if (flags & BTRFS_INODE_APPEND)
115 iflags |= FS_APPEND_FL;
116 if (flags & BTRFS_INODE_NODUMP)
117 iflags |= FS_NODUMP_FL;
118 if (flags & BTRFS_INODE_NOATIME)
119 iflags |= FS_NOATIME_FL;
120 if (flags & BTRFS_INODE_DIRSYNC)
121 iflags |= FS_DIRSYNC_FL;
122 if (flags & BTRFS_INODE_NODATACOW)
123 iflags |= FS_NOCOW_FL;
125 if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
126 iflags |= FS_COMPR_FL;
127 else if (flags & BTRFS_INODE_NOCOMPRESS)
128 iflags |= FS_NOCOMP_FL;
134 * Update inode->i_flags based on the btrfs internal flags.
136 void btrfs_update_iflags(struct inode *inode)
138 struct btrfs_inode *ip = BTRFS_I(inode);
139 unsigned int new_fl = 0;
141 if (ip->flags & BTRFS_INODE_SYNC)
143 if (ip->flags & BTRFS_INODE_IMMUTABLE)
144 new_fl |= S_IMMUTABLE;
145 if (ip->flags & BTRFS_INODE_APPEND)
147 if (ip->flags & BTRFS_INODE_NOATIME)
149 if (ip->flags & BTRFS_INODE_DIRSYNC)
152 set_mask_bits(&inode->i_flags,
153 S_SYNC | S_APPEND | S_IMMUTABLE | S_NOATIME | S_DIRSYNC,
158 * Inherit flags from the parent inode.
160 * Currently only the compression flags and the cow flags are inherited.
162 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
169 flags = BTRFS_I(dir)->flags;
171 if (flags & BTRFS_INODE_NOCOMPRESS) {
172 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
173 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
174 } else if (flags & BTRFS_INODE_COMPRESS) {
175 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
176 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
179 if (flags & BTRFS_INODE_NODATACOW) {
180 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
181 if (S_ISREG(inode->i_mode))
182 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
185 btrfs_update_iflags(inode);
188 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
190 struct btrfs_inode *ip = BTRFS_I(file_inode(file));
191 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
193 if (copy_to_user(arg, &flags, sizeof(flags)))
198 static int check_flags(unsigned int flags)
200 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
201 FS_NOATIME_FL | FS_NODUMP_FL | \
202 FS_SYNC_FL | FS_DIRSYNC_FL | \
203 FS_NOCOMP_FL | FS_COMPR_FL |
207 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
213 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
215 struct inode *inode = file_inode(file);
216 struct btrfs_inode *ip = BTRFS_I(inode);
217 struct btrfs_root *root = ip->root;
218 struct btrfs_trans_handle *trans;
219 unsigned int flags, oldflags;
222 unsigned int i_oldflags;
225 if (!inode_owner_or_capable(inode))
228 if (btrfs_root_readonly(root))
231 if (copy_from_user(&flags, arg, sizeof(flags)))
234 ret = check_flags(flags);
238 ret = mnt_want_write_file(file);
242 mutex_lock(&inode->i_mutex);
244 ip_oldflags = ip->flags;
245 i_oldflags = inode->i_flags;
246 mode = inode->i_mode;
248 flags = btrfs_mask_flags(inode->i_mode, flags);
249 oldflags = btrfs_flags_to_ioctl(ip->flags);
250 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
251 if (!capable(CAP_LINUX_IMMUTABLE)) {
257 if (flags & FS_SYNC_FL)
258 ip->flags |= BTRFS_INODE_SYNC;
260 ip->flags &= ~BTRFS_INODE_SYNC;
261 if (flags & FS_IMMUTABLE_FL)
262 ip->flags |= BTRFS_INODE_IMMUTABLE;
264 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
265 if (flags & FS_APPEND_FL)
266 ip->flags |= BTRFS_INODE_APPEND;
268 ip->flags &= ~BTRFS_INODE_APPEND;
269 if (flags & FS_NODUMP_FL)
270 ip->flags |= BTRFS_INODE_NODUMP;
272 ip->flags &= ~BTRFS_INODE_NODUMP;
273 if (flags & FS_NOATIME_FL)
274 ip->flags |= BTRFS_INODE_NOATIME;
276 ip->flags &= ~BTRFS_INODE_NOATIME;
277 if (flags & FS_DIRSYNC_FL)
278 ip->flags |= BTRFS_INODE_DIRSYNC;
280 ip->flags &= ~BTRFS_INODE_DIRSYNC;
281 if (flags & FS_NOCOW_FL) {
284 * It's safe to turn csums off here, no extents exist.
285 * Otherwise we want the flag to reflect the real COW
286 * status of the file and will not set it.
288 if (inode->i_size == 0)
289 ip->flags |= BTRFS_INODE_NODATACOW
290 | BTRFS_INODE_NODATASUM;
292 ip->flags |= BTRFS_INODE_NODATACOW;
296 * Revert back under same assuptions as above
299 if (inode->i_size == 0)
300 ip->flags &= ~(BTRFS_INODE_NODATACOW
301 | BTRFS_INODE_NODATASUM);
303 ip->flags &= ~BTRFS_INODE_NODATACOW;
308 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
309 * flag may be changed automatically if compression code won't make
312 if (flags & FS_NOCOMP_FL) {
313 ip->flags &= ~BTRFS_INODE_COMPRESS;
314 ip->flags |= BTRFS_INODE_NOCOMPRESS;
316 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
317 if (ret && ret != -ENODATA)
319 } else if (flags & FS_COMPR_FL) {
322 ip->flags |= BTRFS_INODE_COMPRESS;
323 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
325 if (root->fs_info->compress_type == BTRFS_COMPRESS_LZO)
329 ret = btrfs_set_prop(inode, "btrfs.compression",
330 comp, strlen(comp), 0);
335 ret = btrfs_set_prop(inode, "btrfs.compression", NULL, 0, 0);
336 if (ret && ret != -ENODATA)
338 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
341 trans = btrfs_start_transaction(root, 1);
343 ret = PTR_ERR(trans);
347 btrfs_update_iflags(inode);
348 inode_inc_iversion(inode);
349 inode->i_ctime = CURRENT_TIME;
350 ret = btrfs_update_inode(trans, root, inode);
352 btrfs_end_transaction(trans, root);
355 ip->flags = ip_oldflags;
356 inode->i_flags = i_oldflags;
360 mutex_unlock(&inode->i_mutex);
361 mnt_drop_write_file(file);
365 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
367 struct inode *inode = file_inode(file);
369 return put_user(inode->i_generation, arg);
372 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
374 struct btrfs_fs_info *fs_info = btrfs_sb(file_inode(file)->i_sb);
375 struct btrfs_device *device;
376 struct request_queue *q;
377 struct fstrim_range range;
378 u64 minlen = ULLONG_MAX;
380 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
383 if (!capable(CAP_SYS_ADMIN))
387 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
391 q = bdev_get_queue(device->bdev);
392 if (blk_queue_discard(q)) {
394 minlen = min((u64)q->limits.discard_granularity,
402 if (copy_from_user(&range, arg, sizeof(range)))
404 if (range.start > total_bytes ||
405 range.len < fs_info->sb->s_blocksize)
408 range.len = min(range.len, total_bytes - range.start);
409 range.minlen = max(range.minlen, minlen);
410 ret = btrfs_trim_fs(fs_info->tree_root, &range);
414 if (copy_to_user(arg, &range, sizeof(range)))
420 int btrfs_is_empty_uuid(u8 *uuid)
424 for (i = 0; i < BTRFS_UUID_SIZE; i++) {
431 static noinline int create_subvol(struct inode *dir,
432 struct dentry *dentry,
433 char *name, int namelen,
435 struct btrfs_qgroup_inherit *inherit)
437 struct btrfs_trans_handle *trans;
438 struct btrfs_key key;
439 struct btrfs_root_item root_item;
440 struct btrfs_inode_item *inode_item;
441 struct extent_buffer *leaf;
442 struct btrfs_root *root = BTRFS_I(dir)->root;
443 struct btrfs_root *new_root;
444 struct btrfs_block_rsv block_rsv;
445 struct timespec cur_time = CURRENT_TIME;
450 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
455 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
460 * Don't create subvolume whose level is not zero. Or qgroup will be
461 * screwed up since it assume subvolme qgroup's level to be 0.
463 if (btrfs_qgroup_level(objectid))
466 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
468 * The same as the snapshot creation, please see the comment
469 * of create_snapshot().
471 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
472 8, &qgroup_reserved, false);
476 trans = btrfs_start_transaction(root, 0);
478 ret = PTR_ERR(trans);
479 btrfs_subvolume_release_metadata(root, &block_rsv,
483 trans->block_rsv = &block_rsv;
484 trans->bytes_reserved = block_rsv.size;
486 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
490 leaf = btrfs_alloc_tree_block(trans, root, 0, objectid, NULL, 0, 0, 0);
496 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
497 btrfs_set_header_bytenr(leaf, leaf->start);
498 btrfs_set_header_generation(leaf, trans->transid);
499 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
500 btrfs_set_header_owner(leaf, objectid);
502 write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(),
504 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
505 btrfs_header_chunk_tree_uuid(leaf),
507 btrfs_mark_buffer_dirty(leaf);
509 memset(&root_item, 0, sizeof(root_item));
511 inode_item = &root_item.inode;
512 btrfs_set_stack_inode_generation(inode_item, 1);
513 btrfs_set_stack_inode_size(inode_item, 3);
514 btrfs_set_stack_inode_nlink(inode_item, 1);
515 btrfs_set_stack_inode_nbytes(inode_item, root->nodesize);
516 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
518 btrfs_set_root_flags(&root_item, 0);
519 btrfs_set_root_limit(&root_item, 0);
520 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
522 btrfs_set_root_bytenr(&root_item, leaf->start);
523 btrfs_set_root_generation(&root_item, trans->transid);
524 btrfs_set_root_level(&root_item, 0);
525 btrfs_set_root_refs(&root_item, 1);
526 btrfs_set_root_used(&root_item, leaf->len);
527 btrfs_set_root_last_snapshot(&root_item, 0);
529 btrfs_set_root_generation_v2(&root_item,
530 btrfs_root_generation(&root_item));
531 uuid_le_gen(&new_uuid);
532 memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
533 btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec);
534 btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec);
535 root_item.ctime = root_item.otime;
536 btrfs_set_root_ctransid(&root_item, trans->transid);
537 btrfs_set_root_otransid(&root_item, trans->transid);
539 btrfs_tree_unlock(leaf);
540 free_extent_buffer(leaf);
543 btrfs_set_root_dirid(&root_item, new_dirid);
545 key.objectid = objectid;
547 key.type = BTRFS_ROOT_ITEM_KEY;
548 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
553 key.offset = (u64)-1;
554 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
555 if (IS_ERR(new_root)) {
556 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
557 ret = PTR_ERR(new_root);
561 btrfs_record_root_in_trans(trans, new_root);
563 ret = btrfs_create_subvol_root(trans, new_root, root, new_dirid);
565 /* We potentially lose an unused inode item here */
566 btrfs_abort_transaction(trans, root, ret);
571 * insert the directory item
573 ret = btrfs_set_inode_index(dir, &index);
575 btrfs_abort_transaction(trans, root, ret);
579 ret = btrfs_insert_dir_item(trans, root,
580 name, namelen, dir, &key,
581 BTRFS_FT_DIR, index);
583 btrfs_abort_transaction(trans, root, ret);
587 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
588 ret = btrfs_update_inode(trans, root, dir);
591 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
592 objectid, root->root_key.objectid,
593 btrfs_ino(dir), index, name, namelen);
596 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
597 root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
600 btrfs_abort_transaction(trans, root, ret);
603 trans->block_rsv = NULL;
604 trans->bytes_reserved = 0;
605 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
608 *async_transid = trans->transid;
609 err = btrfs_commit_transaction_async(trans, root, 1);
611 err = btrfs_commit_transaction(trans, root);
613 err = btrfs_commit_transaction(trans, root);
619 inode = btrfs_lookup_dentry(dir, dentry);
621 return PTR_ERR(inode);
622 d_instantiate(dentry, inode);
627 static void btrfs_wait_for_no_snapshoting_writes(struct btrfs_root *root)
633 prepare_to_wait(&root->subv_writers->wait, &wait,
634 TASK_UNINTERRUPTIBLE);
636 writers = percpu_counter_sum(&root->subv_writers->counter);
640 finish_wait(&root->subv_writers->wait, &wait);
644 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
645 struct dentry *dentry, char *name, int namelen,
646 u64 *async_transid, bool readonly,
647 struct btrfs_qgroup_inherit *inherit)
650 struct btrfs_pending_snapshot *pending_snapshot;
651 struct btrfs_trans_handle *trans;
654 if (!test_bit(BTRFS_ROOT_REF_COWS, &root->state))
657 atomic_inc(&root->will_be_snapshoted);
658 smp_mb__after_atomic();
659 btrfs_wait_for_no_snapshoting_writes(root);
661 ret = btrfs_start_delalloc_inodes(root, 0);
665 btrfs_wait_ordered_extents(root, -1);
667 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
668 if (!pending_snapshot) {
673 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
674 BTRFS_BLOCK_RSV_TEMP);
676 * 1 - parent dir inode
679 * 2 - root ref/backref
680 * 1 - root of snapshot
683 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
684 &pending_snapshot->block_rsv, 8,
685 &pending_snapshot->qgroup_reserved,
690 pending_snapshot->dentry = dentry;
691 pending_snapshot->root = root;
692 pending_snapshot->readonly = readonly;
693 pending_snapshot->dir = dir;
694 pending_snapshot->inherit = inherit;
696 trans = btrfs_start_transaction(root, 0);
698 ret = PTR_ERR(trans);
702 spin_lock(&root->fs_info->trans_lock);
703 list_add(&pending_snapshot->list,
704 &trans->transaction->pending_snapshots);
705 spin_unlock(&root->fs_info->trans_lock);
707 *async_transid = trans->transid;
708 ret = btrfs_commit_transaction_async(trans,
709 root->fs_info->extent_root, 1);
711 ret = btrfs_commit_transaction(trans, root);
713 ret = btrfs_commit_transaction(trans,
714 root->fs_info->extent_root);
719 ret = pending_snapshot->error;
723 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
727 inode = btrfs_lookup_dentry(d_inode(dentry->d_parent), dentry);
729 ret = PTR_ERR(inode);
733 d_instantiate(dentry, inode);
736 btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
737 &pending_snapshot->block_rsv,
738 pending_snapshot->qgroup_reserved);
740 kfree(pending_snapshot);
742 if (atomic_dec_and_test(&root->will_be_snapshoted))
743 wake_up_atomic_t(&root->will_be_snapshoted);
747 /* copy of may_delete in fs/namei.c()
748 * Check whether we can remove a link victim from directory dir, check
749 * whether the type of victim is right.
750 * 1. We can't do it if dir is read-only (done in permission())
751 * 2. We should have write and exec permissions on dir
752 * 3. We can't remove anything from append-only dir
753 * 4. We can't do anything with immutable dir (done in permission())
754 * 5. If the sticky bit on dir is set we should either
755 * a. be owner of dir, or
756 * b. be owner of victim, or
757 * c. have CAP_FOWNER capability
758 * 6. If the victim is append-only or immutable we can't do antyhing with
759 * links pointing to it.
760 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
761 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
762 * 9. We can't remove a root or mountpoint.
763 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
764 * nfs_async_unlink().
767 static int btrfs_may_delete(struct inode *dir, struct dentry *victim, int isdir)
771 if (d_really_is_negative(victim))
774 BUG_ON(d_inode(victim->d_parent) != dir);
775 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
777 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
782 if (check_sticky(dir, d_inode(victim)) || IS_APPEND(d_inode(victim)) ||
783 IS_IMMUTABLE(d_inode(victim)) || IS_SWAPFILE(d_inode(victim)))
786 if (!d_is_dir(victim))
790 } else if (d_is_dir(victim))
794 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
799 /* copy of may_create in fs/namei.c() */
800 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
802 if (d_really_is_positive(child))
806 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
810 * Create a new subvolume below @parent. This is largely modeled after
811 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
812 * inside this filesystem so it's quite a bit simpler.
814 static noinline int btrfs_mksubvol(struct path *parent,
815 char *name, int namelen,
816 struct btrfs_root *snap_src,
817 u64 *async_transid, bool readonly,
818 struct btrfs_qgroup_inherit *inherit)
820 struct inode *dir = d_inode(parent->dentry);
821 struct dentry *dentry;
824 error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
828 dentry = lookup_one_len(name, parent->dentry, namelen);
829 error = PTR_ERR(dentry);
834 if (d_really_is_positive(dentry))
837 error = btrfs_may_create(dir, dentry);
842 * even if this name doesn't exist, we may get hash collisions.
843 * check for them now when we can safely fail
845 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
851 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
853 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
857 error = create_snapshot(snap_src, dir, dentry, name, namelen,
858 async_transid, readonly, inherit);
860 error = create_subvol(dir, dentry, name, namelen,
861 async_transid, inherit);
864 fsnotify_mkdir(dir, dentry);
866 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
870 mutex_unlock(&dir->i_mutex);
875 * When we're defragging a range, we don't want to kick it off again
876 * if it is really just waiting for delalloc to send it down.
877 * If we find a nice big extent or delalloc range for the bytes in the
878 * file you want to defrag, we return 0 to let you know to skip this
881 static int check_defrag_in_cache(struct inode *inode, u64 offset, u32 thresh)
883 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
884 struct extent_map *em = NULL;
885 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
888 read_lock(&em_tree->lock);
889 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
890 read_unlock(&em_tree->lock);
893 end = extent_map_end(em);
895 if (end - offset > thresh)
898 /* if we already have a nice delalloc here, just stop */
900 end = count_range_bits(io_tree, &offset, offset + thresh,
901 thresh, EXTENT_DELALLOC, 1);
908 * helper function to walk through a file and find extents
909 * newer than a specific transid, and smaller than thresh.
911 * This is used by the defragging code to find new and small
914 static int find_new_extents(struct btrfs_root *root,
915 struct inode *inode, u64 newer_than,
916 u64 *off, u32 thresh)
918 struct btrfs_path *path;
919 struct btrfs_key min_key;
920 struct extent_buffer *leaf;
921 struct btrfs_file_extent_item *extent;
924 u64 ino = btrfs_ino(inode);
926 path = btrfs_alloc_path();
930 min_key.objectid = ino;
931 min_key.type = BTRFS_EXTENT_DATA_KEY;
932 min_key.offset = *off;
935 ret = btrfs_search_forward(root, &min_key, path, newer_than);
939 if (min_key.objectid != ino)
941 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
944 leaf = path->nodes[0];
945 extent = btrfs_item_ptr(leaf, path->slots[0],
946 struct btrfs_file_extent_item);
948 type = btrfs_file_extent_type(leaf, extent);
949 if (type == BTRFS_FILE_EXTENT_REG &&
950 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
951 check_defrag_in_cache(inode, min_key.offset, thresh)) {
952 *off = min_key.offset;
953 btrfs_free_path(path);
958 if (path->slots[0] < btrfs_header_nritems(leaf)) {
959 btrfs_item_key_to_cpu(leaf, &min_key, path->slots[0]);
963 if (min_key.offset == (u64)-1)
967 btrfs_release_path(path);
970 btrfs_free_path(path);
974 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
976 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
977 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
978 struct extent_map *em;
979 u64 len = PAGE_CACHE_SIZE;
982 * hopefully we have this extent in the tree already, try without
983 * the full extent lock
985 read_lock(&em_tree->lock);
986 em = lookup_extent_mapping(em_tree, start, len);
987 read_unlock(&em_tree->lock);
990 struct extent_state *cached = NULL;
991 u64 end = start + len - 1;
993 /* get the big lock and read metadata off disk */
994 lock_extent_bits(io_tree, start, end, 0, &cached);
995 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
996 unlock_extent_cached(io_tree, start, end, &cached, GFP_NOFS);
1005 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
1007 struct extent_map *next;
1010 /* this is the last extent */
1011 if (em->start + em->len >= i_size_read(inode))
1014 next = defrag_lookup_extent(inode, em->start + em->len);
1015 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
1017 else if ((em->block_start + em->block_len == next->block_start) &&
1018 (em->block_len > 128 * 1024 && next->block_len > 128 * 1024))
1021 free_extent_map(next);
1025 static int should_defrag_range(struct inode *inode, u64 start, u32 thresh,
1026 u64 *last_len, u64 *skip, u64 *defrag_end,
1029 struct extent_map *em;
1031 bool next_mergeable = true;
1034 * make sure that once we start defragging an extent, we keep on
1037 if (start < *defrag_end)
1042 em = defrag_lookup_extent(inode, start);
1046 /* this will cover holes, and inline extents */
1047 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
1052 next_mergeable = defrag_check_next_extent(inode, em);
1054 * we hit a real extent, if it is big or the next extent is not a
1055 * real extent, don't bother defragging it
1057 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
1058 (em->len >= thresh || !next_mergeable))
1062 * last_len ends up being a counter of how many bytes we've defragged.
1063 * every time we choose not to defrag an extent, we reset *last_len
1064 * so that the next tiny extent will force a defrag.
1066 * The end result of this is that tiny extents before a single big
1067 * extent will force at least part of that big extent to be defragged.
1070 *defrag_end = extent_map_end(em);
1073 *skip = extent_map_end(em);
1077 free_extent_map(em);
1082 * it doesn't do much good to defrag one or two pages
1083 * at a time. This pulls in a nice chunk of pages
1084 * to COW and defrag.
1086 * It also makes sure the delalloc code has enough
1087 * dirty data to avoid making new small extents as part
1090 * It's a good idea to start RA on this range
1091 * before calling this.
1093 static int cluster_pages_for_defrag(struct inode *inode,
1094 struct page **pages,
1095 unsigned long start_index,
1096 unsigned long num_pages)
1098 unsigned long file_end;
1099 u64 isize = i_size_read(inode);
1106 struct btrfs_ordered_extent *ordered;
1107 struct extent_state *cached_state = NULL;
1108 struct extent_io_tree *tree;
1109 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1111 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1112 if (!isize || start_index > file_end)
1115 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1117 ret = btrfs_delalloc_reserve_space(inode,
1118 page_cnt << PAGE_CACHE_SHIFT);
1122 tree = &BTRFS_I(inode)->io_tree;
1124 /* step one, lock all the pages */
1125 for (i = 0; i < page_cnt; i++) {
1128 page = find_or_create_page(inode->i_mapping,
1129 start_index + i, mask);
1133 page_start = page_offset(page);
1134 page_end = page_start + PAGE_CACHE_SIZE - 1;
1136 lock_extent_bits(tree, page_start, page_end,
1138 ordered = btrfs_lookup_ordered_extent(inode,
1140 unlock_extent_cached(tree, page_start, page_end,
1141 &cached_state, GFP_NOFS);
1146 btrfs_start_ordered_extent(inode, ordered, 1);
1147 btrfs_put_ordered_extent(ordered);
1150 * we unlocked the page above, so we need check if
1151 * it was released or not.
1153 if (page->mapping != inode->i_mapping) {
1155 page_cache_release(page);
1160 if (!PageUptodate(page)) {
1161 btrfs_readpage(NULL, page);
1163 if (!PageUptodate(page)) {
1165 page_cache_release(page);
1171 if (page->mapping != inode->i_mapping) {
1173 page_cache_release(page);
1183 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1187 * so now we have a nice long stream of locked
1188 * and up to date pages, lets wait on them
1190 for (i = 0; i < i_done; i++)
1191 wait_on_page_writeback(pages[i]);
1193 page_start = page_offset(pages[0]);
1194 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1196 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1197 page_start, page_end - 1, 0, &cached_state);
1198 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1199 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1200 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1201 &cached_state, GFP_NOFS);
1203 if (i_done != page_cnt) {
1204 spin_lock(&BTRFS_I(inode)->lock);
1205 BTRFS_I(inode)->outstanding_extents++;
1206 spin_unlock(&BTRFS_I(inode)->lock);
1207 btrfs_delalloc_release_space(inode,
1208 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1212 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1213 &cached_state, GFP_NOFS);
1215 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1216 page_start, page_end - 1, &cached_state,
1219 for (i = 0; i < i_done; i++) {
1220 clear_page_dirty_for_io(pages[i]);
1221 ClearPageChecked(pages[i]);
1222 set_page_extent_mapped(pages[i]);
1223 set_page_dirty(pages[i]);
1224 unlock_page(pages[i]);
1225 page_cache_release(pages[i]);
1229 for (i = 0; i < i_done; i++) {
1230 unlock_page(pages[i]);
1231 page_cache_release(pages[i]);
1233 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1238 int btrfs_defrag_file(struct inode *inode, struct file *file,
1239 struct btrfs_ioctl_defrag_range_args *range,
1240 u64 newer_than, unsigned long max_to_defrag)
1242 struct btrfs_root *root = BTRFS_I(inode)->root;
1243 struct file_ra_state *ra = NULL;
1244 unsigned long last_index;
1245 u64 isize = i_size_read(inode);
1249 u64 newer_off = range->start;
1251 unsigned long ra_index = 0;
1253 int defrag_count = 0;
1254 int compress_type = BTRFS_COMPRESS_ZLIB;
1255 u32 extent_thresh = range->extent_thresh;
1256 unsigned long max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1257 unsigned long cluster = max_cluster;
1258 u64 new_align = ~((u64)128 * 1024 - 1);
1259 struct page **pages = NULL;
1264 if (range->start >= isize)
1267 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1268 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1270 if (range->compress_type)
1271 compress_type = range->compress_type;
1274 if (extent_thresh == 0)
1275 extent_thresh = 256 * 1024;
1278 * if we were not given a file, allocate a readahead
1282 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1285 file_ra_state_init(ra, inode->i_mapping);
1290 pages = kmalloc_array(max_cluster, sizeof(struct page *),
1297 /* find the last page to defrag */
1298 if (range->start + range->len > range->start) {
1299 last_index = min_t(u64, isize - 1,
1300 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1302 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1306 ret = find_new_extents(root, inode, newer_than,
1307 &newer_off, 64 * 1024);
1309 range->start = newer_off;
1311 * we always align our defrag to help keep
1312 * the extents in the file evenly spaced
1314 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1318 i = range->start >> PAGE_CACHE_SHIFT;
1321 max_to_defrag = last_index + 1;
1324 * make writeback starts from i, so the defrag range can be
1325 * written sequentially.
1327 if (i < inode->i_mapping->writeback_index)
1328 inode->i_mapping->writeback_index = i;
1330 while (i <= last_index && defrag_count < max_to_defrag &&
1331 (i < DIV_ROUND_UP(i_size_read(inode), PAGE_CACHE_SIZE))) {
1333 * make sure we stop running if someone unmounts
1336 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1339 if (btrfs_defrag_cancelled(root->fs_info)) {
1340 printk(KERN_DEBUG "BTRFS: defrag_file cancelled\n");
1345 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1346 extent_thresh, &last_len, &skip,
1347 &defrag_end, range->flags &
1348 BTRFS_DEFRAG_RANGE_COMPRESS)) {
1351 * the should_defrag function tells us how much to skip
1352 * bump our counter by the suggested amount
1354 next = DIV_ROUND_UP(skip, PAGE_CACHE_SIZE);
1355 i = max(i + 1, next);
1360 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1361 PAGE_CACHE_SHIFT) - i;
1362 cluster = min(cluster, max_cluster);
1364 cluster = max_cluster;
1367 if (i + cluster > ra_index) {
1368 ra_index = max(i, ra_index);
1369 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1371 ra_index += max_cluster;
1374 mutex_lock(&inode->i_mutex);
1375 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1376 BTRFS_I(inode)->force_compress = compress_type;
1377 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1379 mutex_unlock(&inode->i_mutex);
1383 defrag_count += ret;
1384 balance_dirty_pages_ratelimited(inode->i_mapping);
1385 mutex_unlock(&inode->i_mutex);
1388 if (newer_off == (u64)-1)
1394 newer_off = max(newer_off + 1,
1395 (u64)i << PAGE_CACHE_SHIFT);
1397 ret = find_new_extents(root, inode,
1398 newer_than, &newer_off,
1401 range->start = newer_off;
1402 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1409 last_len += ret << PAGE_CACHE_SHIFT;
1417 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO)) {
1418 filemap_flush(inode->i_mapping);
1419 if (test_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
1420 &BTRFS_I(inode)->runtime_flags))
1421 filemap_flush(inode->i_mapping);
1424 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1425 /* the filemap_flush will queue IO into the worker threads, but
1426 * we have to make sure the IO is actually started and that
1427 * ordered extents get created before we return
1429 atomic_inc(&root->fs_info->async_submit_draining);
1430 while (atomic_read(&root->fs_info->nr_async_submits) ||
1431 atomic_read(&root->fs_info->async_delalloc_pages)) {
1432 wait_event(root->fs_info->async_submit_wait,
1433 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1434 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1436 atomic_dec(&root->fs_info->async_submit_draining);
1439 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1440 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1446 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1447 mutex_lock(&inode->i_mutex);
1448 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1449 mutex_unlock(&inode->i_mutex);
1457 static noinline int btrfs_ioctl_resize(struct file *file,
1463 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
1464 struct btrfs_ioctl_vol_args *vol_args;
1465 struct btrfs_trans_handle *trans;
1466 struct btrfs_device *device = NULL;
1469 char *devstr = NULL;
1473 if (!capable(CAP_SYS_ADMIN))
1476 ret = mnt_want_write_file(file);
1480 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1482 mnt_drop_write_file(file);
1483 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
1486 mutex_lock(&root->fs_info->volume_mutex);
1487 vol_args = memdup_user(arg, sizeof(*vol_args));
1488 if (IS_ERR(vol_args)) {
1489 ret = PTR_ERR(vol_args);
1493 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1495 sizestr = vol_args->name;
1496 devstr = strchr(sizestr, ':');
1498 sizestr = devstr + 1;
1500 devstr = vol_args->name;
1501 ret = kstrtoull(devstr, 10, &devid);
1508 btrfs_info(root->fs_info, "resizing devid %llu", devid);
1511 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
1513 btrfs_info(root->fs_info, "resizer unable to find device %llu",
1519 if (!device->writeable) {
1520 btrfs_info(root->fs_info,
1521 "resizer unable to apply on readonly device %llu",
1527 if (!strcmp(sizestr, "max"))
1528 new_size = device->bdev->bd_inode->i_size;
1530 if (sizestr[0] == '-') {
1533 } else if (sizestr[0] == '+') {
1537 new_size = memparse(sizestr, &retptr);
1538 if (*retptr != '\0' || new_size == 0) {
1544 if (device->is_tgtdev_for_dev_replace) {
1549 old_size = btrfs_device_get_total_bytes(device);
1552 if (new_size > old_size) {
1556 new_size = old_size - new_size;
1557 } else if (mod > 0) {
1558 if (new_size > ULLONG_MAX - old_size) {
1562 new_size = old_size + new_size;
1565 if (new_size < 256 * 1024 * 1024) {
1569 if (new_size > device->bdev->bd_inode->i_size) {
1574 new_size = div_u64(new_size, root->sectorsize);
1575 new_size *= root->sectorsize;
1577 printk_in_rcu(KERN_INFO "BTRFS: new size for %s is %llu\n",
1578 rcu_str_deref(device->name), new_size);
1580 if (new_size > old_size) {
1581 trans = btrfs_start_transaction(root, 0);
1582 if (IS_ERR(trans)) {
1583 ret = PTR_ERR(trans);
1586 ret = btrfs_grow_device(trans, device, new_size);
1587 btrfs_commit_transaction(trans, root);
1588 } else if (new_size < old_size) {
1589 ret = btrfs_shrink_device(device, new_size);
1590 } /* equal, nothing need to do */
1595 mutex_unlock(&root->fs_info->volume_mutex);
1596 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
1597 mnt_drop_write_file(file);
1601 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1602 char *name, unsigned long fd, int subvol,
1603 u64 *transid, bool readonly,
1604 struct btrfs_qgroup_inherit *inherit)
1609 ret = mnt_want_write_file(file);
1613 namelen = strlen(name);
1614 if (strchr(name, '/')) {
1616 goto out_drop_write;
1619 if (name[0] == '.' &&
1620 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1622 goto out_drop_write;
1626 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1627 NULL, transid, readonly, inherit);
1629 struct fd src = fdget(fd);
1630 struct inode *src_inode;
1633 goto out_drop_write;
1636 src_inode = file_inode(src.file);
1637 if (src_inode->i_sb != file_inode(file)->i_sb) {
1638 btrfs_info(BTRFS_I(src_inode)->root->fs_info,
1639 "Snapshot src from another FS");
1641 } else if (!inode_owner_or_capable(src_inode)) {
1643 * Subvolume creation is not restricted, but snapshots
1644 * are limited to own subvolumes only
1648 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1649 BTRFS_I(src_inode)->root,
1650 transid, readonly, inherit);
1655 mnt_drop_write_file(file);
1660 static noinline int btrfs_ioctl_snap_create(struct file *file,
1661 void __user *arg, int subvol)
1663 struct btrfs_ioctl_vol_args *vol_args;
1666 vol_args = memdup_user(arg, sizeof(*vol_args));
1667 if (IS_ERR(vol_args))
1668 return PTR_ERR(vol_args);
1669 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1671 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1672 vol_args->fd, subvol,
1679 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1680 void __user *arg, int subvol)
1682 struct btrfs_ioctl_vol_args_v2 *vol_args;
1686 bool readonly = false;
1687 struct btrfs_qgroup_inherit *inherit = NULL;
1689 vol_args = memdup_user(arg, sizeof(*vol_args));
1690 if (IS_ERR(vol_args))
1691 return PTR_ERR(vol_args);
1692 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1694 if (vol_args->flags &
1695 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1696 BTRFS_SUBVOL_QGROUP_INHERIT)) {
1701 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1703 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1705 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1706 if (vol_args->size > PAGE_CACHE_SIZE) {
1710 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1711 if (IS_ERR(inherit)) {
1712 ret = PTR_ERR(inherit);
1717 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1718 vol_args->fd, subvol, ptr,
1723 if (ptr && copy_to_user(arg +
1724 offsetof(struct btrfs_ioctl_vol_args_v2,
1736 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1739 struct inode *inode = file_inode(file);
1740 struct btrfs_root *root = BTRFS_I(inode)->root;
1744 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1747 down_read(&root->fs_info->subvol_sem);
1748 if (btrfs_root_readonly(root))
1749 flags |= BTRFS_SUBVOL_RDONLY;
1750 up_read(&root->fs_info->subvol_sem);
1752 if (copy_to_user(arg, &flags, sizeof(flags)))
1758 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1761 struct inode *inode = file_inode(file);
1762 struct btrfs_root *root = BTRFS_I(inode)->root;
1763 struct btrfs_trans_handle *trans;
1768 if (!inode_owner_or_capable(inode))
1771 ret = mnt_want_write_file(file);
1775 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1777 goto out_drop_write;
1780 if (copy_from_user(&flags, arg, sizeof(flags))) {
1782 goto out_drop_write;
1785 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1787 goto out_drop_write;
1790 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1792 goto out_drop_write;
1795 down_write(&root->fs_info->subvol_sem);
1798 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1801 root_flags = btrfs_root_flags(&root->root_item);
1802 if (flags & BTRFS_SUBVOL_RDONLY) {
1803 btrfs_set_root_flags(&root->root_item,
1804 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1807 * Block RO -> RW transition if this subvolume is involved in
1810 spin_lock(&root->root_item_lock);
1811 if (root->send_in_progress == 0) {
1812 btrfs_set_root_flags(&root->root_item,
1813 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1814 spin_unlock(&root->root_item_lock);
1816 spin_unlock(&root->root_item_lock);
1817 btrfs_warn(root->fs_info,
1818 "Attempt to set subvolume %llu read-write during send",
1819 root->root_key.objectid);
1825 trans = btrfs_start_transaction(root, 1);
1826 if (IS_ERR(trans)) {
1827 ret = PTR_ERR(trans);
1831 ret = btrfs_update_root(trans, root->fs_info->tree_root,
1832 &root->root_key, &root->root_item);
1834 btrfs_commit_transaction(trans, root);
1837 btrfs_set_root_flags(&root->root_item, root_flags);
1839 up_write(&root->fs_info->subvol_sem);
1841 mnt_drop_write_file(file);
1847 * helper to check if the subvolume references other subvolumes
1849 static noinline int may_destroy_subvol(struct btrfs_root *root)
1851 struct btrfs_path *path;
1852 struct btrfs_dir_item *di;
1853 struct btrfs_key key;
1857 path = btrfs_alloc_path();
1861 /* Make sure this root isn't set as the default subvol */
1862 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1863 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1864 dir_id, "default", 7, 0);
1865 if (di && !IS_ERR(di)) {
1866 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1867 if (key.objectid == root->root_key.objectid) {
1869 btrfs_err(root->fs_info, "deleting default subvolume "
1870 "%llu is not allowed", key.objectid);
1873 btrfs_release_path(path);
1876 key.objectid = root->root_key.objectid;
1877 key.type = BTRFS_ROOT_REF_KEY;
1878 key.offset = (u64)-1;
1880 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1887 if (path->slots[0] > 0) {
1889 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1890 if (key.objectid == root->root_key.objectid &&
1891 key.type == BTRFS_ROOT_REF_KEY)
1895 btrfs_free_path(path);
1899 static noinline int key_in_sk(struct btrfs_key *key,
1900 struct btrfs_ioctl_search_key *sk)
1902 struct btrfs_key test;
1905 test.objectid = sk->min_objectid;
1906 test.type = sk->min_type;
1907 test.offset = sk->min_offset;
1909 ret = btrfs_comp_cpu_keys(key, &test);
1913 test.objectid = sk->max_objectid;
1914 test.type = sk->max_type;
1915 test.offset = sk->max_offset;
1917 ret = btrfs_comp_cpu_keys(key, &test);
1923 static noinline int copy_to_sk(struct btrfs_root *root,
1924 struct btrfs_path *path,
1925 struct btrfs_key *key,
1926 struct btrfs_ioctl_search_key *sk,
1929 unsigned long *sk_offset,
1933 struct extent_buffer *leaf;
1934 struct btrfs_ioctl_search_header sh;
1935 unsigned long item_off;
1936 unsigned long item_len;
1942 leaf = path->nodes[0];
1943 slot = path->slots[0];
1944 nritems = btrfs_header_nritems(leaf);
1946 if (btrfs_header_generation(leaf) > sk->max_transid) {
1950 found_transid = btrfs_header_generation(leaf);
1952 for (i = slot; i < nritems; i++) {
1953 item_off = btrfs_item_ptr_offset(leaf, i);
1954 item_len = btrfs_item_size_nr(leaf, i);
1956 btrfs_item_key_to_cpu(leaf, key, i);
1957 if (!key_in_sk(key, sk))
1960 if (sizeof(sh) + item_len > *buf_size) {
1967 * return one empty item back for v1, which does not
1971 *buf_size = sizeof(sh) + item_len;
1976 if (sizeof(sh) + item_len + *sk_offset > *buf_size) {
1981 sh.objectid = key->objectid;
1982 sh.offset = key->offset;
1983 sh.type = key->type;
1985 sh.transid = found_transid;
1987 /* copy search result header */
1988 if (copy_to_user(ubuf + *sk_offset, &sh, sizeof(sh))) {
1993 *sk_offset += sizeof(sh);
1996 char __user *up = ubuf + *sk_offset;
1998 if (read_extent_buffer_to_user(leaf, up,
1999 item_off, item_len)) {
2004 *sk_offset += item_len;
2008 if (ret) /* -EOVERFLOW from above */
2011 if (*num_found >= sk->nr_items) {
2018 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
2020 else if (key->type < (u8)-1 && key->type < sk->max_type) {
2023 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
2031 * 0: all items from this leaf copied, continue with next
2032 * 1: * more items can be copied, but unused buffer is too small
2033 * * all items were found
2034 * Either way, it will stops the loop which iterates to the next
2036 * -EOVERFLOW: item was to large for buffer
2037 * -EFAULT: could not copy extent buffer back to userspace
2042 static noinline int search_ioctl(struct inode *inode,
2043 struct btrfs_ioctl_search_key *sk,
2047 struct btrfs_root *root;
2048 struct btrfs_key key;
2049 struct btrfs_path *path;
2050 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
2053 unsigned long sk_offset = 0;
2055 if (*buf_size < sizeof(struct btrfs_ioctl_search_header)) {
2056 *buf_size = sizeof(struct btrfs_ioctl_search_header);
2060 path = btrfs_alloc_path();
2064 if (sk->tree_id == 0) {
2065 /* search the root of the inode that was passed */
2066 root = BTRFS_I(inode)->root;
2068 key.objectid = sk->tree_id;
2069 key.type = BTRFS_ROOT_ITEM_KEY;
2070 key.offset = (u64)-1;
2071 root = btrfs_read_fs_root_no_name(info, &key);
2073 printk(KERN_ERR "BTRFS: could not find root %llu\n",
2075 btrfs_free_path(path);
2080 key.objectid = sk->min_objectid;
2081 key.type = sk->min_type;
2082 key.offset = sk->min_offset;
2085 ret = btrfs_search_forward(root, &key, path, sk->min_transid);
2091 ret = copy_to_sk(root, path, &key, sk, buf_size, ubuf,
2092 &sk_offset, &num_found);
2093 btrfs_release_path(path);
2101 sk->nr_items = num_found;
2102 btrfs_free_path(path);
2106 static noinline int btrfs_ioctl_tree_search(struct file *file,
2109 struct btrfs_ioctl_search_args __user *uargs;
2110 struct btrfs_ioctl_search_key sk;
2111 struct inode *inode;
2115 if (!capable(CAP_SYS_ADMIN))
2118 uargs = (struct btrfs_ioctl_search_args __user *)argp;
2120 if (copy_from_user(&sk, &uargs->key, sizeof(sk)))
2123 buf_size = sizeof(uargs->buf);
2125 inode = file_inode(file);
2126 ret = search_ioctl(inode, &sk, &buf_size, uargs->buf);
2129 * In the origin implementation an overflow is handled by returning a
2130 * search header with a len of zero, so reset ret.
2132 if (ret == -EOVERFLOW)
2135 if (ret == 0 && copy_to_user(&uargs->key, &sk, sizeof(sk)))
2140 static noinline int btrfs_ioctl_tree_search_v2(struct file *file,
2143 struct btrfs_ioctl_search_args_v2 __user *uarg;
2144 struct btrfs_ioctl_search_args_v2 args;
2145 struct inode *inode;
2148 const size_t buf_limit = 16 * 1024 * 1024;
2150 if (!capable(CAP_SYS_ADMIN))
2153 /* copy search header and buffer size */
2154 uarg = (struct btrfs_ioctl_search_args_v2 __user *)argp;
2155 if (copy_from_user(&args, uarg, sizeof(args)))
2158 buf_size = args.buf_size;
2160 if (buf_size < sizeof(struct btrfs_ioctl_search_header))
2163 /* limit result size to 16MB */
2164 if (buf_size > buf_limit)
2165 buf_size = buf_limit;
2167 inode = file_inode(file);
2168 ret = search_ioctl(inode, &args.key, &buf_size,
2169 (char *)(&uarg->buf[0]));
2170 if (ret == 0 && copy_to_user(&uarg->key, &args.key, sizeof(args.key)))
2172 else if (ret == -EOVERFLOW &&
2173 copy_to_user(&uarg->buf_size, &buf_size, sizeof(buf_size)))
2180 * Search INODE_REFs to identify path name of 'dirid' directory
2181 * in a 'tree_id' tree. and sets path name to 'name'.
2183 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
2184 u64 tree_id, u64 dirid, char *name)
2186 struct btrfs_root *root;
2187 struct btrfs_key key;
2193 struct btrfs_inode_ref *iref;
2194 struct extent_buffer *l;
2195 struct btrfs_path *path;
2197 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2202 path = btrfs_alloc_path();
2206 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
2208 key.objectid = tree_id;
2209 key.type = BTRFS_ROOT_ITEM_KEY;
2210 key.offset = (u64)-1;
2211 root = btrfs_read_fs_root_no_name(info, &key);
2213 printk(KERN_ERR "BTRFS: could not find root %llu\n", tree_id);
2218 key.objectid = dirid;
2219 key.type = BTRFS_INODE_REF_KEY;
2220 key.offset = (u64)-1;
2223 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2227 ret = btrfs_previous_item(root, path, dirid,
2228 BTRFS_INODE_REF_KEY);
2238 slot = path->slots[0];
2239 btrfs_item_key_to_cpu(l, &key, slot);
2241 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2242 len = btrfs_inode_ref_name_len(l, iref);
2244 total_len += len + 1;
2246 ret = -ENAMETOOLONG;
2251 read_extent_buffer(l, ptr, (unsigned long)(iref + 1), len);
2253 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2256 btrfs_release_path(path);
2257 key.objectid = key.offset;
2258 key.offset = (u64)-1;
2259 dirid = key.objectid;
2261 memmove(name, ptr, total_len);
2262 name[total_len] = '\0';
2265 btrfs_free_path(path);
2269 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2272 struct btrfs_ioctl_ino_lookup_args *args;
2273 struct inode *inode;
2276 if (!capable(CAP_SYS_ADMIN))
2279 args = memdup_user(argp, sizeof(*args));
2281 return PTR_ERR(args);
2283 inode = file_inode(file);
2285 if (args->treeid == 0)
2286 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2288 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2289 args->treeid, args->objectid,
2292 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2299 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2302 struct dentry *parent = file->f_path.dentry;
2303 struct dentry *dentry;
2304 struct inode *dir = d_inode(parent);
2305 struct inode *inode;
2306 struct btrfs_root *root = BTRFS_I(dir)->root;
2307 struct btrfs_root *dest = NULL;
2308 struct btrfs_ioctl_vol_args *vol_args;
2309 struct btrfs_trans_handle *trans;
2310 struct btrfs_block_rsv block_rsv;
2312 u64 qgroup_reserved;
2317 vol_args = memdup_user(arg, sizeof(*vol_args));
2318 if (IS_ERR(vol_args))
2319 return PTR_ERR(vol_args);
2321 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2322 namelen = strlen(vol_args->name);
2323 if (strchr(vol_args->name, '/') ||
2324 strncmp(vol_args->name, "..", namelen) == 0) {
2329 err = mnt_want_write_file(file);
2334 err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2336 goto out_drop_write;
2337 dentry = lookup_one_len(vol_args->name, parent, namelen);
2338 if (IS_ERR(dentry)) {
2339 err = PTR_ERR(dentry);
2340 goto out_unlock_dir;
2343 if (d_really_is_negative(dentry)) {
2348 inode = d_inode(dentry);
2349 dest = BTRFS_I(inode)->root;
2350 if (!capable(CAP_SYS_ADMIN)) {
2352 * Regular user. Only allow this with a special mount
2353 * option, when the user has write+exec access to the
2354 * subvol root, and when rmdir(2) would have been
2357 * Note that this is _not_ check that the subvol is
2358 * empty or doesn't contain data that we wouldn't
2359 * otherwise be able to delete.
2361 * Users who want to delete empty subvols should try
2365 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2369 * Do not allow deletion if the parent dir is the same
2370 * as the dir to be deleted. That means the ioctl
2371 * must be called on the dentry referencing the root
2372 * of the subvol, not a random directory contained
2379 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2384 /* check if subvolume may be deleted by a user */
2385 err = btrfs_may_delete(dir, dentry, 1);
2389 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2394 mutex_lock(&inode->i_mutex);
2397 * Don't allow to delete a subvolume with send in progress. This is
2398 * inside the i_mutex so the error handling that has to drop the bit
2399 * again is not run concurrently.
2401 spin_lock(&dest->root_item_lock);
2402 root_flags = btrfs_root_flags(&dest->root_item);
2403 if (dest->send_in_progress == 0) {
2404 btrfs_set_root_flags(&dest->root_item,
2405 root_flags | BTRFS_ROOT_SUBVOL_DEAD);
2406 spin_unlock(&dest->root_item_lock);
2408 spin_unlock(&dest->root_item_lock);
2409 btrfs_warn(root->fs_info,
2410 "Attempt to delete subvolume %llu during send",
2411 dest->root_key.objectid);
2416 d_invalidate(dentry);
2418 down_write(&root->fs_info->subvol_sem);
2420 err = may_destroy_subvol(dest);
2424 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2426 * One for dir inode, two for dir entries, two for root
2429 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2430 5, &qgroup_reserved, true);
2434 trans = btrfs_start_transaction(root, 0);
2435 if (IS_ERR(trans)) {
2436 err = PTR_ERR(trans);
2439 trans->block_rsv = &block_rsv;
2440 trans->bytes_reserved = block_rsv.size;
2442 ret = btrfs_unlink_subvol(trans, root, dir,
2443 dest->root_key.objectid,
2444 dentry->d_name.name,
2445 dentry->d_name.len);
2448 btrfs_abort_transaction(trans, root, ret);
2452 btrfs_record_root_in_trans(trans, dest);
2454 memset(&dest->root_item.drop_progress, 0,
2455 sizeof(dest->root_item.drop_progress));
2456 dest->root_item.drop_level = 0;
2457 btrfs_set_root_refs(&dest->root_item, 0);
2459 if (!test_and_set_bit(BTRFS_ROOT_ORPHAN_ITEM_INSERTED, &dest->state)) {
2460 ret = btrfs_insert_orphan_item(trans,
2461 root->fs_info->tree_root,
2462 dest->root_key.objectid);
2464 btrfs_abort_transaction(trans, root, ret);
2470 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2471 dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
2472 dest->root_key.objectid);
2473 if (ret && ret != -ENOENT) {
2474 btrfs_abort_transaction(trans, root, ret);
2478 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2479 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2480 dest->root_item.received_uuid,
2481 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2482 dest->root_key.objectid);
2483 if (ret && ret != -ENOENT) {
2484 btrfs_abort_transaction(trans, root, ret);
2491 trans->block_rsv = NULL;
2492 trans->bytes_reserved = 0;
2493 ret = btrfs_end_transaction(trans, root);
2496 inode->i_flags |= S_DEAD;
2498 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
2500 up_write(&root->fs_info->subvol_sem);
2502 spin_lock(&dest->root_item_lock);
2503 root_flags = btrfs_root_flags(&dest->root_item);
2504 btrfs_set_root_flags(&dest->root_item,
2505 root_flags & ~BTRFS_ROOT_SUBVOL_DEAD);
2506 spin_unlock(&dest->root_item_lock);
2508 mutex_unlock(&inode->i_mutex);
2510 shrink_dcache_sb(root->fs_info->sb);
2511 btrfs_invalidate_inodes(dest);
2513 ASSERT(dest->send_in_progress == 0);
2516 if (dest->ino_cache_inode) {
2517 iput(dest->ino_cache_inode);
2518 dest->ino_cache_inode = NULL;
2524 mutex_unlock(&dir->i_mutex);
2526 mnt_drop_write_file(file);
2532 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2534 struct inode *inode = file_inode(file);
2535 struct btrfs_root *root = BTRFS_I(inode)->root;
2536 struct btrfs_ioctl_defrag_range_args *range;
2539 ret = mnt_want_write_file(file);
2543 if (btrfs_root_readonly(root)) {
2548 switch (inode->i_mode & S_IFMT) {
2550 if (!capable(CAP_SYS_ADMIN)) {
2554 ret = btrfs_defrag_root(root);
2557 ret = btrfs_defrag_root(root->fs_info->extent_root);
2560 if (!(file->f_mode & FMODE_WRITE)) {
2565 range = kzalloc(sizeof(*range), GFP_KERNEL);
2572 if (copy_from_user(range, argp,
2578 /* compression requires us to start the IO */
2579 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2580 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2581 range->extent_thresh = (u32)-1;
2584 /* the rest are all set to zero by kzalloc */
2585 range->len = (u64)-1;
2587 ret = btrfs_defrag_file(file_inode(file), file,
2597 mnt_drop_write_file(file);
2601 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2603 struct btrfs_ioctl_vol_args *vol_args;
2606 if (!capable(CAP_SYS_ADMIN))
2609 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2611 return BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2614 mutex_lock(&root->fs_info->volume_mutex);
2615 vol_args = memdup_user(arg, sizeof(*vol_args));
2616 if (IS_ERR(vol_args)) {
2617 ret = PTR_ERR(vol_args);
2621 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2622 ret = btrfs_init_new_device(root, vol_args->name);
2625 btrfs_info(root->fs_info, "disk added %s",vol_args->name);
2629 mutex_unlock(&root->fs_info->volume_mutex);
2630 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2634 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2636 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2637 struct btrfs_ioctl_vol_args *vol_args;
2640 if (!capable(CAP_SYS_ADMIN))
2643 ret = mnt_want_write_file(file);
2647 vol_args = memdup_user(arg, sizeof(*vol_args));
2648 if (IS_ERR(vol_args)) {
2649 ret = PTR_ERR(vol_args);
2653 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2655 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2657 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2661 mutex_lock(&root->fs_info->volume_mutex);
2662 ret = btrfs_rm_device(root, vol_args->name);
2663 mutex_unlock(&root->fs_info->volume_mutex);
2664 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2667 btrfs_info(root->fs_info, "disk deleted %s",vol_args->name);
2672 mnt_drop_write_file(file);
2676 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2678 struct btrfs_ioctl_fs_info_args *fi_args;
2679 struct btrfs_device *device;
2680 struct btrfs_device *next;
2681 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2684 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2688 mutex_lock(&fs_devices->device_list_mutex);
2689 fi_args->num_devices = fs_devices->num_devices;
2690 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2692 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2693 if (device->devid > fi_args->max_id)
2694 fi_args->max_id = device->devid;
2696 mutex_unlock(&fs_devices->device_list_mutex);
2698 fi_args->nodesize = root->fs_info->super_copy->nodesize;
2699 fi_args->sectorsize = root->fs_info->super_copy->sectorsize;
2700 fi_args->clone_alignment = root->fs_info->super_copy->sectorsize;
2702 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2709 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2711 struct btrfs_ioctl_dev_info_args *di_args;
2712 struct btrfs_device *dev;
2713 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2715 char *s_uuid = NULL;
2717 di_args = memdup_user(arg, sizeof(*di_args));
2718 if (IS_ERR(di_args))
2719 return PTR_ERR(di_args);
2721 if (!btrfs_is_empty_uuid(di_args->uuid))
2722 s_uuid = di_args->uuid;
2724 mutex_lock(&fs_devices->device_list_mutex);
2725 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
2732 di_args->devid = dev->devid;
2733 di_args->bytes_used = btrfs_device_get_bytes_used(dev);
2734 di_args->total_bytes = btrfs_device_get_total_bytes(dev);
2735 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2737 struct rcu_string *name;
2740 name = rcu_dereference(dev->name);
2741 strncpy(di_args->path, name->str, sizeof(di_args->path));
2743 di_args->path[sizeof(di_args->path) - 1] = 0;
2745 di_args->path[0] = '\0';
2749 mutex_unlock(&fs_devices->device_list_mutex);
2750 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2757 static struct page *extent_same_get_page(struct inode *inode, u64 off)
2761 struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
2763 index = off >> PAGE_CACHE_SHIFT;
2765 page = grab_cache_page(inode->i_mapping, index);
2769 if (!PageUptodate(page)) {
2770 if (extent_read_full_page_nolock(tree, page, btrfs_get_extent,
2774 if (!PageUptodate(page)) {
2776 page_cache_release(page);
2785 static inline void lock_extent_range(struct inode *inode, u64 off, u64 len)
2787 /* do any pending delalloc/csum calc on src, one way or
2788 another, and lock file content */
2790 struct btrfs_ordered_extent *ordered;
2791 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2792 ordered = btrfs_lookup_first_ordered_extent(inode,
2795 ordered->file_offset + ordered->len <= off ||
2796 ordered->file_offset >= off + len) &&
2797 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2798 off + len - 1, EXTENT_DELALLOC, 0, NULL)) {
2800 btrfs_put_ordered_extent(ordered);
2803 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2805 btrfs_put_ordered_extent(ordered);
2806 btrfs_wait_ordered_range(inode, off, len);
2810 static void btrfs_double_unlock(struct inode *inode1, u64 loff1,
2811 struct inode *inode2, u64 loff2, u64 len)
2813 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2814 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2816 mutex_unlock(&inode1->i_mutex);
2817 mutex_unlock(&inode2->i_mutex);
2820 static void btrfs_double_lock(struct inode *inode1, u64 loff1,
2821 struct inode *inode2, u64 loff2, u64 len)
2823 if (inode1 < inode2) {
2824 swap(inode1, inode2);
2828 mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT);
2829 lock_extent_range(inode1, loff1, len);
2830 if (inode1 != inode2) {
2831 mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD);
2832 lock_extent_range(inode2, loff2, len);
2836 static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
2837 u64 dst_loff, u64 len)
2840 struct page *src_page, *dst_page;
2841 unsigned int cmp_len = PAGE_CACHE_SIZE;
2842 void *addr, *dst_addr;
2845 if (len < PAGE_CACHE_SIZE)
2848 src_page = extent_same_get_page(src, loff);
2851 dst_page = extent_same_get_page(dst, dst_loff);
2853 page_cache_release(src_page);
2856 addr = kmap_atomic(src_page);
2857 dst_addr = kmap_atomic(dst_page);
2859 flush_dcache_page(src_page);
2860 flush_dcache_page(dst_page);
2862 if (memcmp(addr, dst_addr, cmp_len))
2863 ret = BTRFS_SAME_DATA_DIFFERS;
2865 kunmap_atomic(addr);
2866 kunmap_atomic(dst_addr);
2867 page_cache_release(src_page);
2868 page_cache_release(dst_page);
2874 dst_loff += cmp_len;
2881 static int extent_same_check_offsets(struct inode *inode, u64 off, u64 len)
2883 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
2885 if (off + len > inode->i_size || off + len < off)
2887 /* Check that we are block aligned - btrfs_clone() requires this */
2888 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
2894 static int btrfs_extent_same(struct inode *src, u64 loff, u64 len,
2895 struct inode *dst, u64 dst_loff)
2900 * btrfs_clone() can't handle extents in the same file
2901 * yet. Once that works, we can drop this check and replace it
2902 * with a check for the same inode, but overlapping extents.
2910 btrfs_double_lock(src, loff, dst, dst_loff, len);
2912 ret = extent_same_check_offsets(src, loff, len);
2916 ret = extent_same_check_offsets(dst, dst_loff, len);
2920 /* don't make the dst file partly checksummed */
2921 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2922 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
2927 ret = btrfs_cmp_data(src, loff, dst, dst_loff, len);
2929 ret = btrfs_clone(src, dst, loff, len, len, dst_loff);
2932 btrfs_double_unlock(src, loff, dst, dst_loff, len);
2937 #define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024)
2939 static long btrfs_ioctl_file_extent_same(struct file *file,
2940 struct btrfs_ioctl_same_args __user *argp)
2942 struct btrfs_ioctl_same_args *same;
2943 struct btrfs_ioctl_same_extent_info *info;
2944 struct inode *src = file_inode(file);
2950 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
2951 bool is_admin = capable(CAP_SYS_ADMIN);
2954 if (!(file->f_mode & FMODE_READ))
2957 ret = mnt_want_write_file(file);
2961 if (get_user(count, &argp->dest_count)) {
2966 size = offsetof(struct btrfs_ioctl_same_args __user, info[count]);
2968 same = memdup_user(argp, size);
2971 ret = PTR_ERR(same);
2975 off = same->logical_offset;
2979 * Limit the total length we will dedupe for each operation.
2980 * This is intended to bound the total time spent in this
2981 * ioctl to something sane.
2983 if (len > BTRFS_MAX_DEDUPE_LEN)
2984 len = BTRFS_MAX_DEDUPE_LEN;
2986 if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) {
2988 * Btrfs does not support blocksize < page_size. As a
2989 * result, btrfs_cmp_data() won't correctly handle
2990 * this situation without an update.
2997 if (S_ISDIR(src->i_mode))
3001 if (!S_ISREG(src->i_mode))
3004 /* pre-format output fields to sane values */
3005 for (i = 0; i < count; i++) {
3006 same->info[i].bytes_deduped = 0ULL;
3007 same->info[i].status = 0;
3010 for (i = 0, info = same->info; i < count; i++, info++) {
3012 struct fd dst_file = fdget(info->fd);
3013 if (!dst_file.file) {
3014 info->status = -EBADF;
3017 dst = file_inode(dst_file.file);
3019 if (!(is_admin || (dst_file.file->f_mode & FMODE_WRITE))) {
3020 info->status = -EINVAL;
3021 } else if (file->f_path.mnt != dst_file.file->f_path.mnt) {
3022 info->status = -EXDEV;
3023 } else if (S_ISDIR(dst->i_mode)) {
3024 info->status = -EISDIR;
3025 } else if (!S_ISREG(dst->i_mode)) {
3026 info->status = -EACCES;
3028 info->status = btrfs_extent_same(src, off, len, dst,
3029 info->logical_offset);
3030 if (info->status == 0)
3031 info->bytes_deduped += len;
3036 ret = copy_to_user(argp, same, size);
3041 mnt_drop_write_file(file);
3045 /* Helper to check and see if this root currently has a ref on the given disk
3046 * bytenr. If it does then we need to update the quota for this root. This
3047 * doesn't do anything if quotas aren't enabled.
3049 static int check_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
3052 struct seq_list tree_mod_seq_elem = SEQ_LIST_INIT(tree_mod_seq_elem);
3053 struct ulist *roots;
3054 struct ulist_iterator uiter;
3055 struct ulist_node *root_node = NULL;
3058 if (!root->fs_info->quota_enabled)
3061 btrfs_get_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3062 ret = btrfs_find_all_roots(trans, root->fs_info, disko,
3063 tree_mod_seq_elem.seq, &roots);
3067 ULIST_ITER_INIT(&uiter);
3068 while ((root_node = ulist_next(roots, &uiter))) {
3069 if (root_node->val == root->objectid) {
3076 btrfs_put_tree_mod_seq(root->fs_info, &tree_mod_seq_elem);
3080 static int clone_finish_inode_update(struct btrfs_trans_handle *trans,
3081 struct inode *inode,
3086 struct btrfs_root *root = BTRFS_I(inode)->root;
3089 inode_inc_iversion(inode);
3090 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3092 * We round up to the block size at eof when determining which
3093 * extents to clone above, but shouldn't round up the file size.
3095 if (endoff > destoff + olen)
3096 endoff = destoff + olen;
3097 if (endoff > inode->i_size)
3098 btrfs_i_size_write(inode, endoff);
3100 ret = btrfs_update_inode(trans, root, inode);
3102 btrfs_abort_transaction(trans, root, ret);
3103 btrfs_end_transaction(trans, root);
3106 ret = btrfs_end_transaction(trans, root);
3111 static void clone_update_extent_map(struct inode *inode,
3112 const struct btrfs_trans_handle *trans,
3113 const struct btrfs_path *path,
3114 const u64 hole_offset,
3117 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
3118 struct extent_map *em;
3121 em = alloc_extent_map();
3123 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3124 &BTRFS_I(inode)->runtime_flags);
3129 struct btrfs_file_extent_item *fi;
3131 fi = btrfs_item_ptr(path->nodes[0], path->slots[0],
3132 struct btrfs_file_extent_item);
3133 btrfs_extent_item_to_extent_map(inode, path, fi, false, em);
3134 em->generation = -1;
3135 if (btrfs_file_extent_type(path->nodes[0], fi) ==
3136 BTRFS_FILE_EXTENT_INLINE)
3137 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3138 &BTRFS_I(inode)->runtime_flags);
3140 em->start = hole_offset;
3142 em->ram_bytes = em->len;
3143 em->orig_start = hole_offset;
3144 em->block_start = EXTENT_MAP_HOLE;
3146 em->orig_block_len = 0;
3147 em->compress_type = BTRFS_COMPRESS_NONE;
3148 em->generation = trans->transid;
3152 write_lock(&em_tree->lock);
3153 ret = add_extent_mapping(em_tree, em, 1);
3154 write_unlock(&em_tree->lock);
3155 if (ret != -EEXIST) {
3156 free_extent_map(em);
3159 btrfs_drop_extent_cache(inode, em->start,
3160 em->start + em->len - 1, 0);
3164 set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
3165 &BTRFS_I(inode)->runtime_flags);
3169 * btrfs_clone() - clone a range from inode file to another
3171 * @src: Inode to clone from
3172 * @inode: Inode to clone to
3173 * @off: Offset within source to start clone from
3174 * @olen: Original length, passed by user, of range to clone
3175 * @olen_aligned: Block-aligned value of olen, extent_same uses
3176 * identical values here
3177 * @destoff: Offset within @inode to start clone
3179 static int btrfs_clone(struct inode *src, struct inode *inode,
3180 const u64 off, const u64 olen, const u64 olen_aligned,
3183 struct btrfs_root *root = BTRFS_I(inode)->root;
3184 struct btrfs_path *path = NULL;
3185 struct extent_buffer *leaf;
3186 struct btrfs_trans_handle *trans;
3188 struct btrfs_key key;
3193 const u64 len = olen_aligned;
3195 u64 last_dest_end = destoff;
3198 buf = vmalloc(root->nodesize);
3202 path = btrfs_alloc_path();
3210 key.objectid = btrfs_ino(src);
3211 key.type = BTRFS_EXTENT_DATA_KEY;
3215 u64 next_key_min_offset = key.offset + 1;
3218 * note the key will change type as we walk through the
3221 path->leave_spinning = 1;
3222 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
3227 * First search, if no extent item that starts at offset off was
3228 * found but the previous item is an extent item, it's possible
3229 * it might overlap our target range, therefore process it.
3231 if (key.offset == off && ret > 0 && path->slots[0] > 0) {
3232 btrfs_item_key_to_cpu(path->nodes[0], &key,
3233 path->slots[0] - 1);
3234 if (key.type == BTRFS_EXTENT_DATA_KEY)
3238 nritems = btrfs_header_nritems(path->nodes[0]);
3241 if (path->slots[0] >= nritems) {
3242 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
3247 nritems = btrfs_header_nritems(path->nodes[0]);
3249 leaf = path->nodes[0];
3250 slot = path->slots[0];
3252 btrfs_item_key_to_cpu(leaf, &key, slot);
3253 if (key.type > BTRFS_EXTENT_DATA_KEY ||
3254 key.objectid != btrfs_ino(src))
3257 if (key.type == BTRFS_EXTENT_DATA_KEY) {
3258 struct btrfs_file_extent_item *extent;
3261 struct btrfs_key new_key;
3262 u64 disko = 0, diskl = 0;
3263 u64 datao = 0, datal = 0;
3267 extent = btrfs_item_ptr(leaf, slot,
3268 struct btrfs_file_extent_item);
3269 comp = btrfs_file_extent_compression(leaf, extent);
3270 type = btrfs_file_extent_type(leaf, extent);
3271 if (type == BTRFS_FILE_EXTENT_REG ||
3272 type == BTRFS_FILE_EXTENT_PREALLOC) {
3273 disko = btrfs_file_extent_disk_bytenr(leaf,
3275 diskl = btrfs_file_extent_disk_num_bytes(leaf,
3277 datao = btrfs_file_extent_offset(leaf, extent);
3278 datal = btrfs_file_extent_num_bytes(leaf,
3280 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3281 /* take upper bound, may be compressed */
3282 datal = btrfs_file_extent_ram_bytes(leaf,
3287 * The first search might have left us at an extent
3288 * item that ends before our target range's start, can
3289 * happen if we have holes and NO_HOLES feature enabled.
3291 if (key.offset + datal <= off) {
3294 } else if (key.offset >= off + len) {
3297 next_key_min_offset = key.offset + datal;
3298 size = btrfs_item_size_nr(leaf, slot);
3299 read_extent_buffer(leaf, buf,
3300 btrfs_item_ptr_offset(leaf, slot),
3303 btrfs_release_path(path);
3304 path->leave_spinning = 0;
3306 memcpy(&new_key, &key, sizeof(new_key));
3307 new_key.objectid = btrfs_ino(inode);
3308 if (off <= key.offset)
3309 new_key.offset = key.offset + destoff - off;
3311 new_key.offset = destoff;
3314 * Deal with a hole that doesn't have an extent item
3315 * that represents it (NO_HOLES feature enabled).
3316 * This hole is either in the middle of the cloning
3317 * range or at the beginning (fully overlaps it or
3318 * partially overlaps it).
3320 if (new_key.offset != last_dest_end)
3321 drop_start = last_dest_end;
3323 drop_start = new_key.offset;
3326 * 1 - adjusting old extent (we may have to split it)
3327 * 1 - add new extent
3330 trans = btrfs_start_transaction(root, 3);
3331 if (IS_ERR(trans)) {
3332 ret = PTR_ERR(trans);
3336 if (type == BTRFS_FILE_EXTENT_REG ||
3337 type == BTRFS_FILE_EXTENT_PREALLOC) {
3339 * a | --- range to clone ---| b
3340 * | ------------- extent ------------- |
3343 /* subtract range b */
3344 if (key.offset + datal > off + len)
3345 datal = off + len - key.offset;
3347 /* subtract range a */
3348 if (off > key.offset) {
3349 datao += off - key.offset;
3350 datal -= off - key.offset;
3353 ret = btrfs_drop_extents(trans, root, inode,
3355 new_key.offset + datal,
3358 if (ret != -EOPNOTSUPP)
3359 btrfs_abort_transaction(trans,
3361 btrfs_end_transaction(trans, root);
3365 ret = btrfs_insert_empty_item(trans, root, path,
3368 btrfs_abort_transaction(trans, root,
3370 btrfs_end_transaction(trans, root);
3374 leaf = path->nodes[0];
3375 slot = path->slots[0];
3376 write_extent_buffer(leaf, buf,
3377 btrfs_item_ptr_offset(leaf, slot),
3380 extent = btrfs_item_ptr(leaf, slot,
3381 struct btrfs_file_extent_item);
3383 /* disko == 0 means it's a hole */
3387 btrfs_set_file_extent_offset(leaf, extent,
3389 btrfs_set_file_extent_num_bytes(leaf, extent,
3393 * We need to look up the roots that point at
3394 * this bytenr and see if the new root does. If
3395 * it does not we need to make sure we update
3396 * quotas appropriately.
3398 if (disko && root != BTRFS_I(src)->root &&
3399 disko != last_disko) {
3400 no_quota = check_ref(trans, root,
3403 btrfs_abort_transaction(trans,
3406 btrfs_end_transaction(trans,
3414 inode_add_bytes(inode, datal);
3415 ret = btrfs_inc_extent_ref(trans, root,
3417 root->root_key.objectid,
3419 new_key.offset - datao,
3422 btrfs_abort_transaction(trans,
3425 btrfs_end_transaction(trans,
3431 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3434 u64 aligned_end = 0;
3436 if (off > key.offset) {
3437 skip = off - key.offset;
3438 new_key.offset += skip;
3441 if (key.offset + datal > off + len)
3442 trim = key.offset + datal - (off + len);
3444 if (comp && (skip || trim)) {
3446 btrfs_end_transaction(trans, root);
3449 size -= skip + trim;
3450 datal -= skip + trim;
3452 aligned_end = ALIGN(new_key.offset + datal,
3454 ret = btrfs_drop_extents(trans, root, inode,
3459 if (ret != -EOPNOTSUPP)
3460 btrfs_abort_transaction(trans,
3462 btrfs_end_transaction(trans, root);
3466 ret = btrfs_insert_empty_item(trans, root, path,
3469 btrfs_abort_transaction(trans, root,
3471 btrfs_end_transaction(trans, root);
3477 btrfs_file_extent_calc_inline_size(0);
3478 memmove(buf+start, buf+start+skip,
3482 leaf = path->nodes[0];
3483 slot = path->slots[0];
3484 write_extent_buffer(leaf, buf,
3485 btrfs_item_ptr_offset(leaf, slot),
3487 inode_add_bytes(inode, datal);
3490 /* If we have an implicit hole (NO_HOLES feature). */
3491 if (drop_start < new_key.offset)
3492 clone_update_extent_map(inode, trans,
3494 new_key.offset - drop_start);
3496 clone_update_extent_map(inode, trans, path, 0, 0);
3498 btrfs_mark_buffer_dirty(leaf);
3499 btrfs_release_path(path);
3501 last_dest_end = ALIGN(new_key.offset + datal,
3503 ret = clone_finish_inode_update(trans, inode,
3508 if (new_key.offset + datal >= destoff + len)
3511 btrfs_release_path(path);
3512 key.offset = next_key_min_offset;
3516 if (last_dest_end < destoff + len) {
3518 * We have an implicit hole (NO_HOLES feature is enabled) that
3519 * fully or partially overlaps our cloning range at its end.
3521 btrfs_release_path(path);
3524 * 1 - remove extent(s)
3527 trans = btrfs_start_transaction(root, 2);
3528 if (IS_ERR(trans)) {
3529 ret = PTR_ERR(trans);
3532 ret = btrfs_drop_extents(trans, root, inode,
3533 last_dest_end, destoff + len, 1);
3535 if (ret != -EOPNOTSUPP)
3536 btrfs_abort_transaction(trans, root, ret);
3537 btrfs_end_transaction(trans, root);
3540 clone_update_extent_map(inode, trans, NULL, last_dest_end,
3541 destoff + len - last_dest_end);
3542 ret = clone_finish_inode_update(trans, inode, destoff + len,
3547 btrfs_free_path(path);
3552 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
3553 u64 off, u64 olen, u64 destoff)
3555 struct inode *inode = file_inode(file);
3556 struct btrfs_root *root = BTRFS_I(inode)->root;
3561 u64 bs = root->fs_info->sb->s_blocksize;
3566 * - split compressed inline extents. annoying: we need to
3567 * decompress into destination's address_space (the file offset
3568 * may change, so source mapping won't do), then recompress (or
3569 * otherwise reinsert) a subrange.
3571 * - split destination inode's inline extents. The inline extents can
3572 * be either compressed or non-compressed.
3575 /* the destination must be opened for writing */
3576 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
3579 if (btrfs_root_readonly(root))
3582 ret = mnt_want_write_file(file);
3586 src_file = fdget(srcfd);
3587 if (!src_file.file) {
3589 goto out_drop_write;
3593 if (src_file.file->f_path.mnt != file->f_path.mnt)
3596 src = file_inode(src_file.file);
3602 /* the src must be open for reading */
3603 if (!(src_file.file->f_mode & FMODE_READ))
3606 /* don't make the dst file partly checksummed */
3607 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3608 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3612 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3616 if (src->i_sb != inode->i_sb)
3621 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
3622 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
3624 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
3625 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
3628 mutex_lock(&src->i_mutex);
3631 /* determine range to clone */
3633 if (off + len > src->i_size || off + len < off)
3636 olen = len = src->i_size - off;
3637 /* if we extend to eof, continue to block boundary */
3638 if (off + len == src->i_size)
3639 len = ALIGN(src->i_size, bs) - off;
3646 /* verify the end result is block aligned */
3647 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3648 !IS_ALIGNED(destoff, bs))
3651 /* verify if ranges are overlapped within the same file */
3653 if (destoff + len > off && destoff < off + len)
3657 if (destoff > inode->i_size) {
3658 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3664 * Lock the target range too. Right after we replace the file extent
3665 * items in the fs tree (which now point to the cloned data), we might
3666 * have a worker replace them with extent items relative to a write
3667 * operation that was issued before this clone operation (i.e. confront
3668 * with inode.c:btrfs_finish_ordered_io).
3671 u64 lock_start = min_t(u64, off, destoff);
3672 u64 lock_len = max_t(u64, off, destoff) + len - lock_start;
3674 lock_extent_range(src, lock_start, lock_len);
3676 lock_extent_range(src, off, len);
3677 lock_extent_range(inode, destoff, len);
3680 ret = btrfs_clone(src, inode, off, olen, len, destoff);
3683 u64 lock_start = min_t(u64, off, destoff);
3684 u64 lock_end = max_t(u64, off, destoff) + len - 1;
3686 unlock_extent(&BTRFS_I(src)->io_tree, lock_start, lock_end);
3688 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
3689 unlock_extent(&BTRFS_I(inode)->io_tree, destoff,
3693 * Truncate page cache pages so that future reads will see the cloned
3694 * data immediately and not the previous data.
3696 truncate_inode_pages_range(&inode->i_data, destoff,
3697 PAGE_CACHE_ALIGN(destoff + len) - 1);
3701 mutex_unlock(&src->i_mutex);
3702 mutex_unlock(&inode->i_mutex);
3704 mutex_unlock(&inode->i_mutex);
3705 mutex_unlock(&src->i_mutex);
3708 mutex_unlock(&src->i_mutex);
3713 mnt_drop_write_file(file);
3717 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
3719 struct btrfs_ioctl_clone_range_args args;
3721 if (copy_from_user(&args, argp, sizeof(args)))
3723 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
3724 args.src_length, args.dest_offset);
3728 * there are many ways the trans_start and trans_end ioctls can lead
3729 * to deadlocks. They should only be used by applications that
3730 * basically own the machine, and have a very in depth understanding
3731 * of all the possible deadlocks and enospc problems.
3733 static long btrfs_ioctl_trans_start(struct file *file)
3735 struct inode *inode = file_inode(file);
3736 struct btrfs_root *root = BTRFS_I(inode)->root;
3737 struct btrfs_trans_handle *trans;
3741 if (!capable(CAP_SYS_ADMIN))
3745 if (file->private_data)
3749 if (btrfs_root_readonly(root))
3752 ret = mnt_want_write_file(file);
3756 atomic_inc(&root->fs_info->open_ioctl_trans);
3759 trans = btrfs_start_ioctl_transaction(root);
3763 file->private_data = trans;
3767 atomic_dec(&root->fs_info->open_ioctl_trans);
3768 mnt_drop_write_file(file);
3773 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3775 struct inode *inode = file_inode(file);
3776 struct btrfs_root *root = BTRFS_I(inode)->root;
3777 struct btrfs_root *new_root;
3778 struct btrfs_dir_item *di;
3779 struct btrfs_trans_handle *trans;
3780 struct btrfs_path *path;
3781 struct btrfs_key location;
3782 struct btrfs_disk_key disk_key;
3787 if (!capable(CAP_SYS_ADMIN))
3790 ret = mnt_want_write_file(file);
3794 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3800 objectid = BTRFS_FS_TREE_OBJECTID;
3802 location.objectid = objectid;
3803 location.type = BTRFS_ROOT_ITEM_KEY;
3804 location.offset = (u64)-1;
3806 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
3807 if (IS_ERR(new_root)) {
3808 ret = PTR_ERR(new_root);
3812 path = btrfs_alloc_path();
3817 path->leave_spinning = 1;
3819 trans = btrfs_start_transaction(root, 1);
3820 if (IS_ERR(trans)) {
3821 btrfs_free_path(path);
3822 ret = PTR_ERR(trans);
3826 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
3827 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
3828 dir_id, "default", 7, 1);
3829 if (IS_ERR_OR_NULL(di)) {
3830 btrfs_free_path(path);
3831 btrfs_end_transaction(trans, root);
3832 btrfs_err(new_root->fs_info, "Umm, you don't have the default dir"
3833 "item, this isn't going to work");
3838 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3839 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3840 btrfs_mark_buffer_dirty(path->nodes[0]);
3841 btrfs_free_path(path);
3843 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
3844 btrfs_end_transaction(trans, root);
3846 mnt_drop_write_file(file);
3850 void btrfs_get_block_group_info(struct list_head *groups_list,
3851 struct btrfs_ioctl_space_info *space)
3853 struct btrfs_block_group_cache *block_group;
3855 space->total_bytes = 0;
3856 space->used_bytes = 0;
3858 list_for_each_entry(block_group, groups_list, list) {
3859 space->flags = block_group->flags;
3860 space->total_bytes += block_group->key.offset;
3861 space->used_bytes +=
3862 btrfs_block_group_used(&block_group->item);
3866 static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
3868 struct btrfs_ioctl_space_args space_args;
3869 struct btrfs_ioctl_space_info space;
3870 struct btrfs_ioctl_space_info *dest;
3871 struct btrfs_ioctl_space_info *dest_orig;
3872 struct btrfs_ioctl_space_info __user *user_dest;
3873 struct btrfs_space_info *info;
3874 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3875 BTRFS_BLOCK_GROUP_SYSTEM,
3876 BTRFS_BLOCK_GROUP_METADATA,
3877 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3884 if (copy_from_user(&space_args,
3885 (struct btrfs_ioctl_space_args __user *)arg,
3886 sizeof(space_args)))
3889 for (i = 0; i < num_types; i++) {
3890 struct btrfs_space_info *tmp;
3894 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3896 if (tmp->flags == types[i]) {
3906 down_read(&info->groups_sem);
3907 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3908 if (!list_empty(&info->block_groups[c]))
3911 up_read(&info->groups_sem);
3915 * Global block reserve, exported as a space_info
3919 /* space_slots == 0 means they are asking for a count */
3920 if (space_args.space_slots == 0) {
3921 space_args.total_spaces = slot_count;
3925 slot_count = min_t(u64, space_args.space_slots, slot_count);
3927 alloc_size = sizeof(*dest) * slot_count;
3929 /* we generally have at most 6 or so space infos, one for each raid
3930 * level. So, a whole page should be more than enough for everyone
3932 if (alloc_size > PAGE_CACHE_SIZE)
3935 space_args.total_spaces = 0;
3936 dest = kmalloc(alloc_size, GFP_NOFS);
3941 /* now we have a buffer to copy into */
3942 for (i = 0; i < num_types; i++) {
3943 struct btrfs_space_info *tmp;
3950 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3952 if (tmp->flags == types[i]) {
3961 down_read(&info->groups_sem);
3962 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3963 if (!list_empty(&info->block_groups[c])) {
3964 btrfs_get_block_group_info(
3965 &info->block_groups[c], &space);
3966 memcpy(dest, &space, sizeof(space));
3968 space_args.total_spaces++;
3974 up_read(&info->groups_sem);
3978 * Add global block reserve
3981 struct btrfs_block_rsv *block_rsv = &root->fs_info->global_block_rsv;
3983 spin_lock(&block_rsv->lock);
3984 space.total_bytes = block_rsv->size;
3985 space.used_bytes = block_rsv->size - block_rsv->reserved;
3986 spin_unlock(&block_rsv->lock);
3987 space.flags = BTRFS_SPACE_INFO_GLOBAL_RSV;
3988 memcpy(dest, &space, sizeof(space));
3989 space_args.total_spaces++;
3992 user_dest = (struct btrfs_ioctl_space_info __user *)
3993 (arg + sizeof(struct btrfs_ioctl_space_args));
3995 if (copy_to_user(user_dest, dest_orig, alloc_size))
4000 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
4007 * there are many ways the trans_start and trans_end ioctls can lead
4008 * to deadlocks. They should only be used by applications that
4009 * basically own the machine, and have a very in depth understanding
4010 * of all the possible deadlocks and enospc problems.
4012 long btrfs_ioctl_trans_end(struct file *file)
4014 struct inode *inode = file_inode(file);
4015 struct btrfs_root *root = BTRFS_I(inode)->root;
4016 struct btrfs_trans_handle *trans;
4018 trans = file->private_data;
4021 file->private_data = NULL;
4023 btrfs_end_transaction(trans, root);
4025 atomic_dec(&root->fs_info->open_ioctl_trans);
4027 mnt_drop_write_file(file);
4031 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
4034 struct btrfs_trans_handle *trans;
4038 trans = btrfs_attach_transaction_barrier(root);
4039 if (IS_ERR(trans)) {
4040 if (PTR_ERR(trans) != -ENOENT)
4041 return PTR_ERR(trans);
4043 /* No running transaction, don't bother */
4044 transid = root->fs_info->last_trans_committed;
4047 transid = trans->transid;
4048 ret = btrfs_commit_transaction_async(trans, root, 0);
4050 btrfs_end_transaction(trans, root);
4055 if (copy_to_user(argp, &transid, sizeof(transid)))
4060 static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
4066 if (copy_from_user(&transid, argp, sizeof(transid)))
4069 transid = 0; /* current trans */
4071 return btrfs_wait_for_commit(root, transid);
4074 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
4076 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4077 struct btrfs_ioctl_scrub_args *sa;
4080 if (!capable(CAP_SYS_ADMIN))
4083 sa = memdup_user(arg, sizeof(*sa));
4087 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
4088 ret = mnt_want_write_file(file);
4093 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
4094 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
4097 if (copy_to_user(arg, sa, sizeof(*sa)))
4100 if (!(sa->flags & BTRFS_SCRUB_READONLY))
4101 mnt_drop_write_file(file);
4107 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
4109 if (!capable(CAP_SYS_ADMIN))
4112 return btrfs_scrub_cancel(root->fs_info);
4115 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
4118 struct btrfs_ioctl_scrub_args *sa;
4121 if (!capable(CAP_SYS_ADMIN))
4124 sa = memdup_user(arg, sizeof(*sa));
4128 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
4130 if (copy_to_user(arg, sa, sizeof(*sa)))
4137 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
4140 struct btrfs_ioctl_get_dev_stats *sa;
4143 sa = memdup_user(arg, sizeof(*sa));
4147 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
4152 ret = btrfs_get_dev_stats(root, sa);
4154 if (copy_to_user(arg, sa, sizeof(*sa)))
4161 static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
4163 struct btrfs_ioctl_dev_replace_args *p;
4166 if (!capable(CAP_SYS_ADMIN))
4169 p = memdup_user(arg, sizeof(*p));
4174 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
4175 if (root->fs_info->sb->s_flags & MS_RDONLY) {
4180 &root->fs_info->mutually_exclusive_operation_running,
4182 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4184 ret = btrfs_dev_replace_start(root, p);
4186 &root->fs_info->mutually_exclusive_operation_running,
4190 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
4191 btrfs_dev_replace_status(root->fs_info, p);
4194 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
4195 ret = btrfs_dev_replace_cancel(root->fs_info, p);
4202 if (copy_to_user(arg, p, sizeof(*p)))
4209 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
4215 struct btrfs_ioctl_ino_path_args *ipa = NULL;
4216 struct inode_fs_paths *ipath = NULL;
4217 struct btrfs_path *path;
4219 if (!capable(CAP_DAC_READ_SEARCH))
4222 path = btrfs_alloc_path();
4228 ipa = memdup_user(arg, sizeof(*ipa));
4235 size = min_t(u32, ipa->size, 4096);
4236 ipath = init_ipath(size, root, path);
4237 if (IS_ERR(ipath)) {
4238 ret = PTR_ERR(ipath);
4243 ret = paths_from_inode(ipa->inum, ipath);
4247 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
4248 rel_ptr = ipath->fspath->val[i] -
4249 (u64)(unsigned long)ipath->fspath->val;
4250 ipath->fspath->val[i] = rel_ptr;
4253 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
4254 (void *)(unsigned long)ipath->fspath, size);
4261 btrfs_free_path(path);
4268 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
4270 struct btrfs_data_container *inodes = ctx;
4271 const size_t c = 3 * sizeof(u64);
4273 if (inodes->bytes_left >= c) {
4274 inodes->bytes_left -= c;
4275 inodes->val[inodes->elem_cnt] = inum;
4276 inodes->val[inodes->elem_cnt + 1] = offset;
4277 inodes->val[inodes->elem_cnt + 2] = root;
4278 inodes->elem_cnt += 3;
4280 inodes->bytes_missing += c - inodes->bytes_left;
4281 inodes->bytes_left = 0;
4282 inodes->elem_missed += 3;
4288 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
4293 struct btrfs_ioctl_logical_ino_args *loi;
4294 struct btrfs_data_container *inodes = NULL;
4295 struct btrfs_path *path = NULL;
4297 if (!capable(CAP_SYS_ADMIN))
4300 loi = memdup_user(arg, sizeof(*loi));
4307 path = btrfs_alloc_path();
4313 size = min_t(u32, loi->size, 64 * 1024);
4314 inodes = init_data_container(size);
4315 if (IS_ERR(inodes)) {
4316 ret = PTR_ERR(inodes);
4321 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
4322 build_ino_list, inodes);
4328 ret = copy_to_user((void *)(unsigned long)loi->inodes,
4329 (void *)(unsigned long)inodes, size);
4334 btrfs_free_path(path);
4341 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
4342 struct btrfs_ioctl_balance_args *bargs)
4344 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
4346 bargs->flags = bctl->flags;
4348 if (atomic_read(&fs_info->balance_running))
4349 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
4350 if (atomic_read(&fs_info->balance_pause_req))
4351 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
4352 if (atomic_read(&fs_info->balance_cancel_req))
4353 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
4355 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
4356 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
4357 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
4360 spin_lock(&fs_info->balance_lock);
4361 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4362 spin_unlock(&fs_info->balance_lock);
4364 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
4368 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
4370 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4371 struct btrfs_fs_info *fs_info = root->fs_info;
4372 struct btrfs_ioctl_balance_args *bargs;
4373 struct btrfs_balance_control *bctl;
4374 bool need_unlock; /* for mut. excl. ops lock */
4377 if (!capable(CAP_SYS_ADMIN))
4380 ret = mnt_want_write_file(file);
4385 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
4386 mutex_lock(&fs_info->volume_mutex);
4387 mutex_lock(&fs_info->balance_mutex);
4393 * mut. excl. ops lock is locked. Three possibilites:
4394 * (1) some other op is running
4395 * (2) balance is running
4396 * (3) balance is paused -- special case (think resume)
4398 mutex_lock(&fs_info->balance_mutex);
4399 if (fs_info->balance_ctl) {
4400 /* this is either (2) or (3) */
4401 if (!atomic_read(&fs_info->balance_running)) {
4402 mutex_unlock(&fs_info->balance_mutex);
4403 if (!mutex_trylock(&fs_info->volume_mutex))
4405 mutex_lock(&fs_info->balance_mutex);
4407 if (fs_info->balance_ctl &&
4408 !atomic_read(&fs_info->balance_running)) {
4410 need_unlock = false;
4414 mutex_unlock(&fs_info->balance_mutex);
4415 mutex_unlock(&fs_info->volume_mutex);
4419 mutex_unlock(&fs_info->balance_mutex);
4425 mutex_unlock(&fs_info->balance_mutex);
4426 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
4431 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
4434 bargs = memdup_user(arg, sizeof(*bargs));
4435 if (IS_ERR(bargs)) {
4436 ret = PTR_ERR(bargs);
4440 if (bargs->flags & BTRFS_BALANCE_RESUME) {
4441 if (!fs_info->balance_ctl) {
4446 bctl = fs_info->balance_ctl;
4447 spin_lock(&fs_info->balance_lock);
4448 bctl->flags |= BTRFS_BALANCE_RESUME;
4449 spin_unlock(&fs_info->balance_lock);
4457 if (fs_info->balance_ctl) {
4462 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
4468 bctl->fs_info = fs_info;
4470 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
4471 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
4472 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
4474 bctl->flags = bargs->flags;
4476 /* balance everything - no filters */
4477 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
4482 * Ownership of bctl and mutually_exclusive_operation_running
4483 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
4484 * or, if restriper was paused all the way until unmount, in
4485 * free_fs_info. mutually_exclusive_operation_running is
4486 * cleared in __cancel_balance.
4488 need_unlock = false;
4490 ret = btrfs_balance(bctl, bargs);
4493 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4500 mutex_unlock(&fs_info->balance_mutex);
4501 mutex_unlock(&fs_info->volume_mutex);
4503 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4505 mnt_drop_write_file(file);
4509 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4511 if (!capable(CAP_SYS_ADMIN))
4515 case BTRFS_BALANCE_CTL_PAUSE:
4516 return btrfs_pause_balance(root->fs_info);
4517 case BTRFS_BALANCE_CTL_CANCEL:
4518 return btrfs_cancel_balance(root->fs_info);
4524 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4527 struct btrfs_fs_info *fs_info = root->fs_info;
4528 struct btrfs_ioctl_balance_args *bargs;
4531 if (!capable(CAP_SYS_ADMIN))
4534 mutex_lock(&fs_info->balance_mutex);
4535 if (!fs_info->balance_ctl) {
4540 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
4546 update_ioctl_balance_args(fs_info, 1, bargs);
4548 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4553 mutex_unlock(&fs_info->balance_mutex);
4557 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4559 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4560 struct btrfs_ioctl_quota_ctl_args *sa;
4561 struct btrfs_trans_handle *trans = NULL;
4565 if (!capable(CAP_SYS_ADMIN))
4568 ret = mnt_want_write_file(file);
4572 sa = memdup_user(arg, sizeof(*sa));
4578 down_write(&root->fs_info->subvol_sem);
4579 trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4580 if (IS_ERR(trans)) {
4581 ret = PTR_ERR(trans);
4586 case BTRFS_QUOTA_CTL_ENABLE:
4587 ret = btrfs_quota_enable(trans, root->fs_info);
4589 case BTRFS_QUOTA_CTL_DISABLE:
4590 ret = btrfs_quota_disable(trans, root->fs_info);
4597 err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4602 up_write(&root->fs_info->subvol_sem);
4604 mnt_drop_write_file(file);
4608 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4610 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4611 struct btrfs_ioctl_qgroup_assign_args *sa;
4612 struct btrfs_trans_handle *trans;
4616 if (!capable(CAP_SYS_ADMIN))
4619 ret = mnt_want_write_file(file);
4623 sa = memdup_user(arg, sizeof(*sa));
4629 trans = btrfs_join_transaction(root);
4630 if (IS_ERR(trans)) {
4631 ret = PTR_ERR(trans);
4635 /* FIXME: check if the IDs really exist */
4637 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4640 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4644 /* update qgroup status and info */
4645 err = btrfs_run_qgroups(trans, root->fs_info);
4647 btrfs_error(root->fs_info, ret,
4648 "failed to update qgroup status and info\n");
4649 err = btrfs_end_transaction(trans, root);
4656 mnt_drop_write_file(file);
4660 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4662 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4663 struct btrfs_ioctl_qgroup_create_args *sa;
4664 struct btrfs_trans_handle *trans;
4668 if (!capable(CAP_SYS_ADMIN))
4671 ret = mnt_want_write_file(file);
4675 sa = memdup_user(arg, sizeof(*sa));
4681 if (!sa->qgroupid) {
4686 trans = btrfs_join_transaction(root);
4687 if (IS_ERR(trans)) {
4688 ret = PTR_ERR(trans);
4692 /* FIXME: check if the IDs really exist */
4694 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid);
4696 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4699 err = btrfs_end_transaction(trans, root);
4706 mnt_drop_write_file(file);
4710 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4712 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4713 struct btrfs_ioctl_qgroup_limit_args *sa;
4714 struct btrfs_trans_handle *trans;
4719 if (!capable(CAP_SYS_ADMIN))
4722 ret = mnt_want_write_file(file);
4726 sa = memdup_user(arg, sizeof(*sa));
4732 trans = btrfs_join_transaction(root);
4733 if (IS_ERR(trans)) {
4734 ret = PTR_ERR(trans);
4738 qgroupid = sa->qgroupid;
4740 /* take the current subvol as qgroup */
4741 qgroupid = root->root_key.objectid;
4744 /* FIXME: check if the IDs really exist */
4745 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
4747 err = btrfs_end_transaction(trans, root);
4754 mnt_drop_write_file(file);
4758 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4760 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4761 struct btrfs_ioctl_quota_rescan_args *qsa;
4764 if (!capable(CAP_SYS_ADMIN))
4767 ret = mnt_want_write_file(file);
4771 qsa = memdup_user(arg, sizeof(*qsa));
4782 ret = btrfs_qgroup_rescan(root->fs_info);
4787 mnt_drop_write_file(file);
4791 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
4793 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4794 struct btrfs_ioctl_quota_rescan_args *qsa;
4797 if (!capable(CAP_SYS_ADMIN))
4800 qsa = kzalloc(sizeof(*qsa), GFP_NOFS);
4804 if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4806 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
4809 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4816 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
4818 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4820 if (!capable(CAP_SYS_ADMIN))
4823 return btrfs_qgroup_wait_for_completion(root->fs_info);
4826 static long _btrfs_ioctl_set_received_subvol(struct file *file,
4827 struct btrfs_ioctl_received_subvol_args *sa)
4829 struct inode *inode = file_inode(file);
4830 struct btrfs_root *root = BTRFS_I(inode)->root;
4831 struct btrfs_root_item *root_item = &root->root_item;
4832 struct btrfs_trans_handle *trans;
4833 struct timespec ct = CURRENT_TIME;
4835 int received_uuid_changed;
4837 if (!inode_owner_or_capable(inode))
4840 ret = mnt_want_write_file(file);
4844 down_write(&root->fs_info->subvol_sem);
4846 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
4851 if (btrfs_root_readonly(root)) {
4858 * 2 - uuid items (received uuid + subvol uuid)
4860 trans = btrfs_start_transaction(root, 3);
4861 if (IS_ERR(trans)) {
4862 ret = PTR_ERR(trans);
4867 sa->rtransid = trans->transid;
4868 sa->rtime.sec = ct.tv_sec;
4869 sa->rtime.nsec = ct.tv_nsec;
4871 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4873 if (received_uuid_changed &&
4874 !btrfs_is_empty_uuid(root_item->received_uuid))
4875 btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
4876 root_item->received_uuid,
4877 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4878 root->root_key.objectid);
4879 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4880 btrfs_set_root_stransid(root_item, sa->stransid);
4881 btrfs_set_root_rtransid(root_item, sa->rtransid);
4882 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4883 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4884 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4885 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4887 ret = btrfs_update_root(trans, root->fs_info->tree_root,
4888 &root->root_key, &root->root_item);
4890 btrfs_end_transaction(trans, root);
4893 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4894 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
4896 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4897 root->root_key.objectid);
4898 if (ret < 0 && ret != -EEXIST) {
4899 btrfs_abort_transaction(trans, root, ret);
4903 ret = btrfs_commit_transaction(trans, root);
4905 btrfs_abort_transaction(trans, root, ret);
4910 up_write(&root->fs_info->subvol_sem);
4911 mnt_drop_write_file(file);
4916 static long btrfs_ioctl_set_received_subvol_32(struct file *file,
4919 struct btrfs_ioctl_received_subvol_args_32 *args32 = NULL;
4920 struct btrfs_ioctl_received_subvol_args *args64 = NULL;
4923 args32 = memdup_user(arg, sizeof(*args32));
4924 if (IS_ERR(args32)) {
4925 ret = PTR_ERR(args32);
4930 args64 = kmalloc(sizeof(*args64), GFP_NOFS);
4936 memcpy(args64->uuid, args32->uuid, BTRFS_UUID_SIZE);
4937 args64->stransid = args32->stransid;
4938 args64->rtransid = args32->rtransid;
4939 args64->stime.sec = args32->stime.sec;
4940 args64->stime.nsec = args32->stime.nsec;
4941 args64->rtime.sec = args32->rtime.sec;
4942 args64->rtime.nsec = args32->rtime.nsec;
4943 args64->flags = args32->flags;
4945 ret = _btrfs_ioctl_set_received_subvol(file, args64);
4949 memcpy(args32->uuid, args64->uuid, BTRFS_UUID_SIZE);
4950 args32->stransid = args64->stransid;
4951 args32->rtransid = args64->rtransid;
4952 args32->stime.sec = args64->stime.sec;
4953 args32->stime.nsec = args64->stime.nsec;
4954 args32->rtime.sec = args64->rtime.sec;
4955 args32->rtime.nsec = args64->rtime.nsec;
4956 args32->flags = args64->flags;
4958 ret = copy_to_user(arg, args32, sizeof(*args32));
4969 static long btrfs_ioctl_set_received_subvol(struct file *file,
4972 struct btrfs_ioctl_received_subvol_args *sa = NULL;
4975 sa = memdup_user(arg, sizeof(*sa));
4982 ret = _btrfs_ioctl_set_received_subvol(file, sa);
4987 ret = copy_to_user(arg, sa, sizeof(*sa));
4996 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
4998 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5001 char label[BTRFS_LABEL_SIZE];
5003 spin_lock(&root->fs_info->super_lock);
5004 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
5005 spin_unlock(&root->fs_info->super_lock);
5007 len = strnlen(label, BTRFS_LABEL_SIZE);
5009 if (len == BTRFS_LABEL_SIZE) {
5010 btrfs_warn(root->fs_info,
5011 "label is too long, return the first %zu bytes", --len);
5014 ret = copy_to_user(arg, label, len);
5016 return ret ? -EFAULT : 0;
5019 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
5021 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5022 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5023 struct btrfs_trans_handle *trans;
5024 char label[BTRFS_LABEL_SIZE];
5027 if (!capable(CAP_SYS_ADMIN))
5030 if (copy_from_user(label, arg, sizeof(label)))
5033 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
5034 btrfs_err(root->fs_info, "unable to set label with more than %d bytes",
5035 BTRFS_LABEL_SIZE - 1);
5039 ret = mnt_want_write_file(file);
5043 trans = btrfs_start_transaction(root, 0);
5044 if (IS_ERR(trans)) {
5045 ret = PTR_ERR(trans);
5049 spin_lock(&root->fs_info->super_lock);
5050 strcpy(super_block->label, label);
5051 spin_unlock(&root->fs_info->super_lock);
5052 ret = btrfs_commit_transaction(trans, root);
5055 mnt_drop_write_file(file);
5059 #define INIT_FEATURE_FLAGS(suffix) \
5060 { .compat_flags = BTRFS_FEATURE_COMPAT_##suffix, \
5061 .compat_ro_flags = BTRFS_FEATURE_COMPAT_RO_##suffix, \
5062 .incompat_flags = BTRFS_FEATURE_INCOMPAT_##suffix }
5064 static int btrfs_ioctl_get_supported_features(struct file *file,
5067 static struct btrfs_ioctl_feature_flags features[3] = {
5068 INIT_FEATURE_FLAGS(SUPP),
5069 INIT_FEATURE_FLAGS(SAFE_SET),
5070 INIT_FEATURE_FLAGS(SAFE_CLEAR)
5073 if (copy_to_user(arg, &features, sizeof(features)))
5079 static int btrfs_ioctl_get_features(struct file *file, void __user *arg)
5081 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5082 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5083 struct btrfs_ioctl_feature_flags features;
5085 features.compat_flags = btrfs_super_compat_flags(super_block);
5086 features.compat_ro_flags = btrfs_super_compat_ro_flags(super_block);
5087 features.incompat_flags = btrfs_super_incompat_flags(super_block);
5089 if (copy_to_user(arg, &features, sizeof(features)))
5095 static int check_feature_bits(struct btrfs_root *root,
5096 enum btrfs_feature_set set,
5097 u64 change_mask, u64 flags, u64 supported_flags,
5098 u64 safe_set, u64 safe_clear)
5100 const char *type = btrfs_feature_set_names[set];
5102 u64 disallowed, unsupported;
5103 u64 set_mask = flags & change_mask;
5104 u64 clear_mask = ~flags & change_mask;
5106 unsupported = set_mask & ~supported_flags;
5108 names = btrfs_printable_features(set, unsupported);
5110 btrfs_warn(root->fs_info,
5111 "this kernel does not support the %s feature bit%s",
5112 names, strchr(names, ',') ? "s" : "");
5115 btrfs_warn(root->fs_info,
5116 "this kernel does not support %s bits 0x%llx",
5121 disallowed = set_mask & ~safe_set;
5123 names = btrfs_printable_features(set, disallowed);
5125 btrfs_warn(root->fs_info,
5126 "can't set the %s feature bit%s while mounted",
5127 names, strchr(names, ',') ? "s" : "");
5130 btrfs_warn(root->fs_info,
5131 "can't set %s bits 0x%llx while mounted",
5136 disallowed = clear_mask & ~safe_clear;
5138 names = btrfs_printable_features(set, disallowed);
5140 btrfs_warn(root->fs_info,
5141 "can't clear the %s feature bit%s while mounted",
5142 names, strchr(names, ',') ? "s" : "");
5145 btrfs_warn(root->fs_info,
5146 "can't clear %s bits 0x%llx while mounted",
5154 #define check_feature(root, change_mask, flags, mask_base) \
5155 check_feature_bits(root, FEAT_##mask_base, change_mask, flags, \
5156 BTRFS_FEATURE_ ## mask_base ## _SUPP, \
5157 BTRFS_FEATURE_ ## mask_base ## _SAFE_SET, \
5158 BTRFS_FEATURE_ ## mask_base ## _SAFE_CLEAR)
5160 static int btrfs_ioctl_set_features(struct file *file, void __user *arg)
5162 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5163 struct btrfs_super_block *super_block = root->fs_info->super_copy;
5164 struct btrfs_ioctl_feature_flags flags[2];
5165 struct btrfs_trans_handle *trans;
5169 if (!capable(CAP_SYS_ADMIN))
5172 if (copy_from_user(flags, arg, sizeof(flags)))
5176 if (!flags[0].compat_flags && !flags[0].compat_ro_flags &&
5177 !flags[0].incompat_flags)
5180 ret = check_feature(root, flags[0].compat_flags,
5181 flags[1].compat_flags, COMPAT);
5185 ret = check_feature(root, flags[0].compat_ro_flags,
5186 flags[1].compat_ro_flags, COMPAT_RO);
5190 ret = check_feature(root, flags[0].incompat_flags,
5191 flags[1].incompat_flags, INCOMPAT);
5195 trans = btrfs_start_transaction(root, 0);
5197 return PTR_ERR(trans);
5199 spin_lock(&root->fs_info->super_lock);
5200 newflags = btrfs_super_compat_flags(super_block);
5201 newflags |= flags[0].compat_flags & flags[1].compat_flags;
5202 newflags &= ~(flags[0].compat_flags & ~flags[1].compat_flags);
5203 btrfs_set_super_compat_flags(super_block, newflags);
5205 newflags = btrfs_super_compat_ro_flags(super_block);
5206 newflags |= flags[0].compat_ro_flags & flags[1].compat_ro_flags;
5207 newflags &= ~(flags[0].compat_ro_flags & ~flags[1].compat_ro_flags);
5208 btrfs_set_super_compat_ro_flags(super_block, newflags);
5210 newflags = btrfs_super_incompat_flags(super_block);
5211 newflags |= flags[0].incompat_flags & flags[1].incompat_flags;
5212 newflags &= ~(flags[0].incompat_flags & ~flags[1].incompat_flags);
5213 btrfs_set_super_incompat_flags(super_block, newflags);
5214 spin_unlock(&root->fs_info->super_lock);
5216 return btrfs_commit_transaction(trans, root);
5219 long btrfs_ioctl(struct file *file, unsigned int
5220 cmd, unsigned long arg)
5222 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5223 void __user *argp = (void __user *)arg;
5226 case FS_IOC_GETFLAGS:
5227 return btrfs_ioctl_getflags(file, argp);
5228 case FS_IOC_SETFLAGS:
5229 return btrfs_ioctl_setflags(file, argp);
5230 case FS_IOC_GETVERSION:
5231 return btrfs_ioctl_getversion(file, argp);
5233 return btrfs_ioctl_fitrim(file, argp);
5234 case BTRFS_IOC_SNAP_CREATE:
5235 return btrfs_ioctl_snap_create(file, argp, 0);
5236 case BTRFS_IOC_SNAP_CREATE_V2:
5237 return btrfs_ioctl_snap_create_v2(file, argp, 0);
5238 case BTRFS_IOC_SUBVOL_CREATE:
5239 return btrfs_ioctl_snap_create(file, argp, 1);
5240 case BTRFS_IOC_SUBVOL_CREATE_V2:
5241 return btrfs_ioctl_snap_create_v2(file, argp, 1);
5242 case BTRFS_IOC_SNAP_DESTROY:
5243 return btrfs_ioctl_snap_destroy(file, argp);
5244 case BTRFS_IOC_SUBVOL_GETFLAGS:
5245 return btrfs_ioctl_subvol_getflags(file, argp);
5246 case BTRFS_IOC_SUBVOL_SETFLAGS:
5247 return btrfs_ioctl_subvol_setflags(file, argp);
5248 case BTRFS_IOC_DEFAULT_SUBVOL:
5249 return btrfs_ioctl_default_subvol(file, argp);
5250 case BTRFS_IOC_DEFRAG:
5251 return btrfs_ioctl_defrag(file, NULL);
5252 case BTRFS_IOC_DEFRAG_RANGE:
5253 return btrfs_ioctl_defrag(file, argp);
5254 case BTRFS_IOC_RESIZE:
5255 return btrfs_ioctl_resize(file, argp);
5256 case BTRFS_IOC_ADD_DEV:
5257 return btrfs_ioctl_add_dev(root, argp);
5258 case BTRFS_IOC_RM_DEV:
5259 return btrfs_ioctl_rm_dev(file, argp);
5260 case BTRFS_IOC_FS_INFO:
5261 return btrfs_ioctl_fs_info(root, argp);
5262 case BTRFS_IOC_DEV_INFO:
5263 return btrfs_ioctl_dev_info(root, argp);
5264 case BTRFS_IOC_BALANCE:
5265 return btrfs_ioctl_balance(file, NULL);
5266 case BTRFS_IOC_CLONE:
5267 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
5268 case BTRFS_IOC_CLONE_RANGE:
5269 return btrfs_ioctl_clone_range(file, argp);
5270 case BTRFS_IOC_TRANS_START:
5271 return btrfs_ioctl_trans_start(file);
5272 case BTRFS_IOC_TRANS_END:
5273 return btrfs_ioctl_trans_end(file);
5274 case BTRFS_IOC_TREE_SEARCH:
5275 return btrfs_ioctl_tree_search(file, argp);
5276 case BTRFS_IOC_TREE_SEARCH_V2:
5277 return btrfs_ioctl_tree_search_v2(file, argp);
5278 case BTRFS_IOC_INO_LOOKUP:
5279 return btrfs_ioctl_ino_lookup(file, argp);
5280 case BTRFS_IOC_INO_PATHS:
5281 return btrfs_ioctl_ino_to_path(root, argp);
5282 case BTRFS_IOC_LOGICAL_INO:
5283 return btrfs_ioctl_logical_to_ino(root, argp);
5284 case BTRFS_IOC_SPACE_INFO:
5285 return btrfs_ioctl_space_info(root, argp);
5286 case BTRFS_IOC_SYNC: {
5289 ret = btrfs_start_delalloc_roots(root->fs_info, 0, -1);
5292 ret = btrfs_sync_fs(file_inode(file)->i_sb, 1);
5294 * The transaction thread may want to do more work,
5295 * namely it pokes the cleaner ktread that will start
5296 * processing uncleaned subvols.
5298 wake_up_process(root->fs_info->transaction_kthread);
5301 case BTRFS_IOC_START_SYNC:
5302 return btrfs_ioctl_start_sync(root, argp);
5303 case BTRFS_IOC_WAIT_SYNC:
5304 return btrfs_ioctl_wait_sync(root, argp);
5305 case BTRFS_IOC_SCRUB:
5306 return btrfs_ioctl_scrub(file, argp);
5307 case BTRFS_IOC_SCRUB_CANCEL:
5308 return btrfs_ioctl_scrub_cancel(root, argp);
5309 case BTRFS_IOC_SCRUB_PROGRESS:
5310 return btrfs_ioctl_scrub_progress(root, argp);
5311 case BTRFS_IOC_BALANCE_V2:
5312 return btrfs_ioctl_balance(file, argp);
5313 case BTRFS_IOC_BALANCE_CTL:
5314 return btrfs_ioctl_balance_ctl(root, arg);
5315 case BTRFS_IOC_BALANCE_PROGRESS:
5316 return btrfs_ioctl_balance_progress(root, argp);
5317 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
5318 return btrfs_ioctl_set_received_subvol(file, argp);
5320 case BTRFS_IOC_SET_RECEIVED_SUBVOL_32:
5321 return btrfs_ioctl_set_received_subvol_32(file, argp);
5323 case BTRFS_IOC_SEND:
5324 return btrfs_ioctl_send(file, argp);
5325 case BTRFS_IOC_GET_DEV_STATS:
5326 return btrfs_ioctl_get_dev_stats(root, argp);
5327 case BTRFS_IOC_QUOTA_CTL:
5328 return btrfs_ioctl_quota_ctl(file, argp);
5329 case BTRFS_IOC_QGROUP_ASSIGN:
5330 return btrfs_ioctl_qgroup_assign(file, argp);
5331 case BTRFS_IOC_QGROUP_CREATE:
5332 return btrfs_ioctl_qgroup_create(file, argp);
5333 case BTRFS_IOC_QGROUP_LIMIT:
5334 return btrfs_ioctl_qgroup_limit(file, argp);
5335 case BTRFS_IOC_QUOTA_RESCAN:
5336 return btrfs_ioctl_quota_rescan(file, argp);
5337 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
5338 return btrfs_ioctl_quota_rescan_status(file, argp);
5339 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
5340 return btrfs_ioctl_quota_rescan_wait(file, argp);
5341 case BTRFS_IOC_DEV_REPLACE:
5342 return btrfs_ioctl_dev_replace(root, argp);
5343 case BTRFS_IOC_GET_FSLABEL:
5344 return btrfs_ioctl_get_fslabel(file, argp);
5345 case BTRFS_IOC_SET_FSLABEL:
5346 return btrfs_ioctl_set_fslabel(file, argp);
5347 case BTRFS_IOC_FILE_EXTENT_SAME:
5348 return btrfs_ioctl_file_extent_same(file, argp);
5349 case BTRFS_IOC_GET_SUPPORTED_FEATURES:
5350 return btrfs_ioctl_get_supported_features(file, argp);
5351 case BTRFS_IOC_GET_FEATURES:
5352 return btrfs_ioctl_get_features(file, argp);
5353 case BTRFS_IOC_SET_FEATURES:
5354 return btrfs_ioctl_set_features(file, argp);