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>
50 #include "transaction.h"
51 #include "btrfs_inode.h"
52 #include "print-tree.h"
55 #include "inode-map.h"
57 #include "rcu-string.h"
59 #include "dev-replace.h"
61 static int btrfs_clone(struct inode *src, struct inode *inode,
62 u64 off, u64 olen, u64 olen_aligned, u64 destoff);
64 /* Mask out flags that are inappropriate for the given type of inode. */
65 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
69 else if (S_ISREG(mode))
70 return flags & ~FS_DIRSYNC_FL;
72 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
76 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
78 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
80 unsigned int iflags = 0;
82 if (flags & BTRFS_INODE_SYNC)
84 if (flags & BTRFS_INODE_IMMUTABLE)
85 iflags |= FS_IMMUTABLE_FL;
86 if (flags & BTRFS_INODE_APPEND)
87 iflags |= FS_APPEND_FL;
88 if (flags & BTRFS_INODE_NODUMP)
89 iflags |= FS_NODUMP_FL;
90 if (flags & BTRFS_INODE_NOATIME)
91 iflags |= FS_NOATIME_FL;
92 if (flags & BTRFS_INODE_DIRSYNC)
93 iflags |= FS_DIRSYNC_FL;
94 if (flags & BTRFS_INODE_NODATACOW)
95 iflags |= FS_NOCOW_FL;
97 if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
98 iflags |= FS_COMPR_FL;
99 else if (flags & BTRFS_INODE_NOCOMPRESS)
100 iflags |= FS_NOCOMP_FL;
106 * Update inode->i_flags based on the btrfs internal flags.
108 void btrfs_update_iflags(struct inode *inode)
110 struct btrfs_inode *ip = BTRFS_I(inode);
112 inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
114 if (ip->flags & BTRFS_INODE_SYNC)
115 inode->i_flags |= S_SYNC;
116 if (ip->flags & BTRFS_INODE_IMMUTABLE)
117 inode->i_flags |= S_IMMUTABLE;
118 if (ip->flags & BTRFS_INODE_APPEND)
119 inode->i_flags |= S_APPEND;
120 if (ip->flags & BTRFS_INODE_NOATIME)
121 inode->i_flags |= S_NOATIME;
122 if (ip->flags & BTRFS_INODE_DIRSYNC)
123 inode->i_flags |= S_DIRSYNC;
127 * Inherit flags from the parent inode.
129 * Currently only the compression flags and the cow flags are inherited.
131 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
138 flags = BTRFS_I(dir)->flags;
140 if (flags & BTRFS_INODE_NOCOMPRESS) {
141 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
142 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
143 } else if (flags & BTRFS_INODE_COMPRESS) {
144 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
145 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
148 if (flags & BTRFS_INODE_NODATACOW) {
149 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
150 if (S_ISREG(inode->i_mode))
151 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
154 btrfs_update_iflags(inode);
157 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
159 struct btrfs_inode *ip = BTRFS_I(file_inode(file));
160 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
162 if (copy_to_user(arg, &flags, sizeof(flags)))
167 static int check_flags(unsigned int flags)
169 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
170 FS_NOATIME_FL | FS_NODUMP_FL | \
171 FS_SYNC_FL | FS_DIRSYNC_FL | \
172 FS_NOCOMP_FL | FS_COMPR_FL |
176 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
182 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
184 struct inode *inode = file_inode(file);
185 struct btrfs_inode *ip = BTRFS_I(inode);
186 struct btrfs_root *root = ip->root;
187 struct btrfs_trans_handle *trans;
188 unsigned int flags, oldflags;
191 unsigned int i_oldflags;
194 if (btrfs_root_readonly(root))
197 if (copy_from_user(&flags, arg, sizeof(flags)))
200 ret = check_flags(flags);
204 if (!inode_owner_or_capable(inode))
207 ret = mnt_want_write_file(file);
211 mutex_lock(&inode->i_mutex);
213 ip_oldflags = ip->flags;
214 i_oldflags = inode->i_flags;
215 mode = inode->i_mode;
217 flags = btrfs_mask_flags(inode->i_mode, flags);
218 oldflags = btrfs_flags_to_ioctl(ip->flags);
219 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
220 if (!capable(CAP_LINUX_IMMUTABLE)) {
226 if (flags & FS_SYNC_FL)
227 ip->flags |= BTRFS_INODE_SYNC;
229 ip->flags &= ~BTRFS_INODE_SYNC;
230 if (flags & FS_IMMUTABLE_FL)
231 ip->flags |= BTRFS_INODE_IMMUTABLE;
233 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
234 if (flags & FS_APPEND_FL)
235 ip->flags |= BTRFS_INODE_APPEND;
237 ip->flags &= ~BTRFS_INODE_APPEND;
238 if (flags & FS_NODUMP_FL)
239 ip->flags |= BTRFS_INODE_NODUMP;
241 ip->flags &= ~BTRFS_INODE_NODUMP;
242 if (flags & FS_NOATIME_FL)
243 ip->flags |= BTRFS_INODE_NOATIME;
245 ip->flags &= ~BTRFS_INODE_NOATIME;
246 if (flags & FS_DIRSYNC_FL)
247 ip->flags |= BTRFS_INODE_DIRSYNC;
249 ip->flags &= ~BTRFS_INODE_DIRSYNC;
250 if (flags & FS_NOCOW_FL) {
253 * It's safe to turn csums off here, no extents exist.
254 * Otherwise we want the flag to reflect the real COW
255 * status of the file and will not set it.
257 if (inode->i_size == 0)
258 ip->flags |= BTRFS_INODE_NODATACOW
259 | BTRFS_INODE_NODATASUM;
261 ip->flags |= BTRFS_INODE_NODATACOW;
265 * Revert back under same assuptions as above
268 if (inode->i_size == 0)
269 ip->flags &= ~(BTRFS_INODE_NODATACOW
270 | BTRFS_INODE_NODATASUM);
272 ip->flags &= ~BTRFS_INODE_NODATACOW;
277 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
278 * flag may be changed automatically if compression code won't make
281 if (flags & FS_NOCOMP_FL) {
282 ip->flags &= ~BTRFS_INODE_COMPRESS;
283 ip->flags |= BTRFS_INODE_NOCOMPRESS;
284 } else if (flags & FS_COMPR_FL) {
285 ip->flags |= BTRFS_INODE_COMPRESS;
286 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
288 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
291 trans = btrfs_start_transaction(root, 1);
293 ret = PTR_ERR(trans);
297 btrfs_update_iflags(inode);
298 inode_inc_iversion(inode);
299 inode->i_ctime = CURRENT_TIME;
300 ret = btrfs_update_inode(trans, root, inode);
302 btrfs_end_transaction(trans, root);
305 ip->flags = ip_oldflags;
306 inode->i_flags = i_oldflags;
310 mutex_unlock(&inode->i_mutex);
311 mnt_drop_write_file(file);
315 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
317 struct inode *inode = file_inode(file);
319 return put_user(inode->i_generation, arg);
322 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
324 struct btrfs_fs_info *fs_info = btrfs_sb(fdentry(file)->d_sb);
325 struct btrfs_device *device;
326 struct request_queue *q;
327 struct fstrim_range range;
328 u64 minlen = ULLONG_MAX;
330 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
333 if (!capable(CAP_SYS_ADMIN))
337 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
341 q = bdev_get_queue(device->bdev);
342 if (blk_queue_discard(q)) {
344 minlen = min((u64)q->limits.discard_granularity,
352 if (copy_from_user(&range, arg, sizeof(range)))
354 if (range.start > total_bytes ||
355 range.len < fs_info->sb->s_blocksize)
358 range.len = min(range.len, total_bytes - range.start);
359 range.minlen = max(range.minlen, minlen);
360 ret = btrfs_trim_fs(fs_info->tree_root, &range);
364 if (copy_to_user(arg, &range, sizeof(range)))
370 static noinline int create_subvol(struct inode *dir,
371 struct dentry *dentry,
372 char *name, int namelen,
374 struct btrfs_qgroup_inherit *inherit)
376 struct btrfs_trans_handle *trans;
377 struct btrfs_key key;
378 struct btrfs_root_item root_item;
379 struct btrfs_inode_item *inode_item;
380 struct extent_buffer *leaf;
381 struct btrfs_root *root = BTRFS_I(dir)->root;
382 struct btrfs_root *new_root;
383 struct btrfs_block_rsv block_rsv;
384 struct timespec cur_time = CURRENT_TIME;
388 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
393 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
397 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
399 * The same as the snapshot creation, please see the comment
400 * of create_snapshot().
402 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
403 7, &qgroup_reserved, false);
407 trans = btrfs_start_transaction(root, 0);
409 ret = PTR_ERR(trans);
412 trans->block_rsv = &block_rsv;
413 trans->bytes_reserved = block_rsv.size;
415 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
419 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
420 0, objectid, NULL, 0, 0, 0);
426 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
427 btrfs_set_header_bytenr(leaf, leaf->start);
428 btrfs_set_header_generation(leaf, trans->transid);
429 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
430 btrfs_set_header_owner(leaf, objectid);
432 write_extent_buffer(leaf, root->fs_info->fsid,
433 (unsigned long)btrfs_header_fsid(leaf),
435 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
436 (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
438 btrfs_mark_buffer_dirty(leaf);
440 memset(&root_item, 0, sizeof(root_item));
442 inode_item = &root_item.inode;
443 btrfs_set_stack_inode_generation(inode_item, 1);
444 btrfs_set_stack_inode_size(inode_item, 3);
445 btrfs_set_stack_inode_nlink(inode_item, 1);
446 btrfs_set_stack_inode_nbytes(inode_item, root->leafsize);
447 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
449 btrfs_set_root_flags(&root_item, 0);
450 btrfs_set_root_limit(&root_item, 0);
451 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
453 btrfs_set_root_bytenr(&root_item, leaf->start);
454 btrfs_set_root_generation(&root_item, trans->transid);
455 btrfs_set_root_level(&root_item, 0);
456 btrfs_set_root_refs(&root_item, 1);
457 btrfs_set_root_used(&root_item, leaf->len);
458 btrfs_set_root_last_snapshot(&root_item, 0);
460 btrfs_set_root_generation_v2(&root_item,
461 btrfs_root_generation(&root_item));
462 uuid_le_gen(&new_uuid);
463 memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
464 btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec);
465 btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec);
466 root_item.ctime = root_item.otime;
467 btrfs_set_root_ctransid(&root_item, trans->transid);
468 btrfs_set_root_otransid(&root_item, trans->transid);
470 btrfs_tree_unlock(leaf);
471 free_extent_buffer(leaf);
474 btrfs_set_root_dirid(&root_item, new_dirid);
476 key.objectid = objectid;
478 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
479 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
484 key.offset = (u64)-1;
485 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
486 if (IS_ERR(new_root)) {
487 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
488 ret = PTR_ERR(new_root);
492 btrfs_record_root_in_trans(trans, new_root);
494 ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
496 /* We potentially lose an unused inode item here */
497 btrfs_abort_transaction(trans, root, ret);
502 * insert the directory item
504 ret = btrfs_set_inode_index(dir, &index);
506 btrfs_abort_transaction(trans, root, ret);
510 ret = btrfs_insert_dir_item(trans, root,
511 name, namelen, dir, &key,
512 BTRFS_FT_DIR, index);
514 btrfs_abort_transaction(trans, root, ret);
518 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
519 ret = btrfs_update_inode(trans, root, dir);
522 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
523 objectid, root->root_key.objectid,
524 btrfs_ino(dir), index, name, namelen);
529 trans->block_rsv = NULL;
530 trans->bytes_reserved = 0;
532 *async_transid = trans->transid;
533 err = btrfs_commit_transaction_async(trans, root, 1);
535 err = btrfs_commit_transaction(trans, root);
537 err = btrfs_commit_transaction(trans, root);
543 d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
545 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
549 static int create_snapshot(struct btrfs_root *root, struct inode *dir,
550 struct dentry *dentry, char *name, int namelen,
551 u64 *async_transid, bool readonly,
552 struct btrfs_qgroup_inherit *inherit)
555 struct btrfs_pending_snapshot *pending_snapshot;
556 struct btrfs_trans_handle *trans;
562 ret = btrfs_start_delalloc_inodes(root, 0);
566 btrfs_wait_ordered_extents(root, 0);
568 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
569 if (!pending_snapshot)
572 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
573 BTRFS_BLOCK_RSV_TEMP);
575 * 1 - parent dir inode
578 * 2 - root ref/backref
579 * 1 - root of snapshot
581 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
582 &pending_snapshot->block_rsv, 7,
583 &pending_snapshot->qgroup_reserved,
588 pending_snapshot->dentry = dentry;
589 pending_snapshot->root = root;
590 pending_snapshot->readonly = readonly;
591 pending_snapshot->dir = dir;
592 pending_snapshot->inherit = inherit;
594 trans = btrfs_start_transaction(root, 0);
596 ret = PTR_ERR(trans);
600 spin_lock(&root->fs_info->trans_lock);
601 list_add(&pending_snapshot->list,
602 &trans->transaction->pending_snapshots);
603 spin_unlock(&root->fs_info->trans_lock);
605 *async_transid = trans->transid;
606 ret = btrfs_commit_transaction_async(trans,
607 root->fs_info->extent_root, 1);
609 ret = btrfs_commit_transaction(trans, root);
611 ret = btrfs_commit_transaction(trans,
612 root->fs_info->extent_root);
617 ret = pending_snapshot->error;
621 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
625 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
627 ret = PTR_ERR(inode);
631 d_instantiate(dentry, inode);
634 btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
635 &pending_snapshot->block_rsv,
636 pending_snapshot->qgroup_reserved);
638 kfree(pending_snapshot);
642 /* copy of check_sticky in fs/namei.c()
643 * It's inline, so penalty for filesystems that don't use sticky bit is
646 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
648 kuid_t fsuid = current_fsuid();
650 if (!(dir->i_mode & S_ISVTX))
652 if (uid_eq(inode->i_uid, fsuid))
654 if (uid_eq(dir->i_uid, fsuid))
656 return !capable(CAP_FOWNER);
659 /* copy of may_delete in fs/namei.c()
660 * Check whether we can remove a link victim from directory dir, check
661 * whether the type of victim is right.
662 * 1. We can't do it if dir is read-only (done in permission())
663 * 2. We should have write and exec permissions on dir
664 * 3. We can't remove anything from append-only dir
665 * 4. We can't do anything with immutable dir (done in permission())
666 * 5. If the sticky bit on dir is set we should either
667 * a. be owner of dir, or
668 * b. be owner of victim, or
669 * c. have CAP_FOWNER capability
670 * 6. If the victim is append-only or immutable we can't do antyhing with
671 * links pointing to it.
672 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
673 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
674 * 9. We can't remove a root or mountpoint.
675 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
676 * nfs_async_unlink().
679 static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
683 if (!victim->d_inode)
686 BUG_ON(victim->d_parent->d_inode != dir);
687 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
689 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
694 if (btrfs_check_sticky(dir, victim->d_inode)||
695 IS_APPEND(victim->d_inode)||
696 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
699 if (!S_ISDIR(victim->d_inode->i_mode))
703 } else if (S_ISDIR(victim->d_inode->i_mode))
707 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
712 /* copy of may_create in fs/namei.c() */
713 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
719 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
723 * Create a new subvolume below @parent. This is largely modeled after
724 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
725 * inside this filesystem so it's quite a bit simpler.
727 static noinline int btrfs_mksubvol(struct path *parent,
728 char *name, int namelen,
729 struct btrfs_root *snap_src,
730 u64 *async_transid, bool readonly,
731 struct btrfs_qgroup_inherit *inherit)
733 struct inode *dir = parent->dentry->d_inode;
734 struct dentry *dentry;
737 error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
741 dentry = lookup_one_len(name, parent->dentry, namelen);
742 error = PTR_ERR(dentry);
750 error = btrfs_may_create(dir, dentry);
755 * even if this name doesn't exist, we may get hash collisions.
756 * check for them now when we can safely fail
758 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
764 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
766 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
770 error = create_snapshot(snap_src, dir, dentry, name, namelen,
771 async_transid, readonly, inherit);
773 error = create_subvol(dir, dentry, name, namelen,
774 async_transid, inherit);
777 fsnotify_mkdir(dir, dentry);
779 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
783 mutex_unlock(&dir->i_mutex);
788 * When we're defragging a range, we don't want to kick it off again
789 * if it is really just waiting for delalloc to send it down.
790 * If we find a nice big extent or delalloc range for the bytes in the
791 * file you want to defrag, we return 0 to let you know to skip this
794 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
796 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
797 struct extent_map *em = NULL;
798 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
801 read_lock(&em_tree->lock);
802 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
803 read_unlock(&em_tree->lock);
806 end = extent_map_end(em);
808 if (end - offset > thresh)
811 /* if we already have a nice delalloc here, just stop */
813 end = count_range_bits(io_tree, &offset, offset + thresh,
814 thresh, EXTENT_DELALLOC, 1);
821 * helper function to walk through a file and find extents
822 * newer than a specific transid, and smaller than thresh.
824 * This is used by the defragging code to find new and small
827 static int find_new_extents(struct btrfs_root *root,
828 struct inode *inode, u64 newer_than,
829 u64 *off, int thresh)
831 struct btrfs_path *path;
832 struct btrfs_key min_key;
833 struct btrfs_key max_key;
834 struct extent_buffer *leaf;
835 struct btrfs_file_extent_item *extent;
838 u64 ino = btrfs_ino(inode);
840 path = btrfs_alloc_path();
844 min_key.objectid = ino;
845 min_key.type = BTRFS_EXTENT_DATA_KEY;
846 min_key.offset = *off;
848 max_key.objectid = ino;
849 max_key.type = (u8)-1;
850 max_key.offset = (u64)-1;
852 path->keep_locks = 1;
855 ret = btrfs_search_forward(root, &min_key, &max_key,
859 if (min_key.objectid != ino)
861 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
864 leaf = path->nodes[0];
865 extent = btrfs_item_ptr(leaf, path->slots[0],
866 struct btrfs_file_extent_item);
868 type = btrfs_file_extent_type(leaf, extent);
869 if (type == BTRFS_FILE_EXTENT_REG &&
870 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
871 check_defrag_in_cache(inode, min_key.offset, thresh)) {
872 *off = min_key.offset;
873 btrfs_free_path(path);
877 if (min_key.offset == (u64)-1)
881 btrfs_release_path(path);
884 btrfs_free_path(path);
888 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
890 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
891 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
892 struct extent_map *em;
893 u64 len = PAGE_CACHE_SIZE;
896 * hopefully we have this extent in the tree already, try without
897 * the full extent lock
899 read_lock(&em_tree->lock);
900 em = lookup_extent_mapping(em_tree, start, len);
901 read_unlock(&em_tree->lock);
904 /* get the big lock and read metadata off disk */
905 lock_extent(io_tree, start, start + len - 1);
906 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
907 unlock_extent(io_tree, start, start + len - 1);
916 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
918 struct extent_map *next;
921 /* this is the last extent */
922 if (em->start + em->len >= i_size_read(inode))
925 next = defrag_lookup_extent(inode, em->start + em->len);
926 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
929 free_extent_map(next);
933 static int should_defrag_range(struct inode *inode, u64 start, int thresh,
934 u64 *last_len, u64 *skip, u64 *defrag_end,
937 struct extent_map *em;
939 bool next_mergeable = true;
942 * make sure that once we start defragging an extent, we keep on
945 if (start < *defrag_end)
950 em = defrag_lookup_extent(inode, start);
954 /* this will cover holes, and inline extents */
955 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
960 next_mergeable = defrag_check_next_extent(inode, em);
963 * we hit a real extent, if it is big or the next extent is not a
964 * real extent, don't bother defragging it
966 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
967 (em->len >= thresh || !next_mergeable))
971 * last_len ends up being a counter of how many bytes we've defragged.
972 * every time we choose not to defrag an extent, we reset *last_len
973 * so that the next tiny extent will force a defrag.
975 * The end result of this is that tiny extents before a single big
976 * extent will force at least part of that big extent to be defragged.
979 *defrag_end = extent_map_end(em);
982 *skip = extent_map_end(em);
991 * it doesn't do much good to defrag one or two pages
992 * at a time. This pulls in a nice chunk of pages
995 * It also makes sure the delalloc code has enough
996 * dirty data to avoid making new small extents as part
999 * It's a good idea to start RA on this range
1000 * before calling this.
1002 static int cluster_pages_for_defrag(struct inode *inode,
1003 struct page **pages,
1004 unsigned long start_index,
1007 unsigned long file_end;
1008 u64 isize = i_size_read(inode);
1015 struct btrfs_ordered_extent *ordered;
1016 struct extent_state *cached_state = NULL;
1017 struct extent_io_tree *tree;
1018 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
1020 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1021 if (!isize || start_index > file_end)
1024 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
1026 ret = btrfs_delalloc_reserve_space(inode,
1027 page_cnt << PAGE_CACHE_SHIFT);
1031 tree = &BTRFS_I(inode)->io_tree;
1033 /* step one, lock all the pages */
1034 for (i = 0; i < page_cnt; i++) {
1037 page = find_or_create_page(inode->i_mapping,
1038 start_index + i, mask);
1042 page_start = page_offset(page);
1043 page_end = page_start + PAGE_CACHE_SIZE - 1;
1045 lock_extent(tree, page_start, page_end);
1046 ordered = btrfs_lookup_ordered_extent(inode,
1048 unlock_extent(tree, page_start, page_end);
1053 btrfs_start_ordered_extent(inode, ordered, 1);
1054 btrfs_put_ordered_extent(ordered);
1057 * we unlocked the page above, so we need check if
1058 * it was released or not.
1060 if (page->mapping != inode->i_mapping) {
1062 page_cache_release(page);
1067 if (!PageUptodate(page)) {
1068 btrfs_readpage(NULL, page);
1070 if (!PageUptodate(page)) {
1072 page_cache_release(page);
1078 if (page->mapping != inode->i_mapping) {
1080 page_cache_release(page);
1090 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1094 * so now we have a nice long stream of locked
1095 * and up to date pages, lets wait on them
1097 for (i = 0; i < i_done; i++)
1098 wait_on_page_writeback(pages[i]);
1100 page_start = page_offset(pages[0]);
1101 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1103 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1104 page_start, page_end - 1, 0, &cached_state);
1105 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1106 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1107 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1108 &cached_state, GFP_NOFS);
1110 if (i_done != page_cnt) {
1111 spin_lock(&BTRFS_I(inode)->lock);
1112 BTRFS_I(inode)->outstanding_extents++;
1113 spin_unlock(&BTRFS_I(inode)->lock);
1114 btrfs_delalloc_release_space(inode,
1115 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1119 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1120 &cached_state, GFP_NOFS);
1122 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1123 page_start, page_end - 1, &cached_state,
1126 for (i = 0; i < i_done; i++) {
1127 clear_page_dirty_for_io(pages[i]);
1128 ClearPageChecked(pages[i]);
1129 set_page_extent_mapped(pages[i]);
1130 set_page_dirty(pages[i]);
1131 unlock_page(pages[i]);
1132 page_cache_release(pages[i]);
1136 for (i = 0; i < i_done; i++) {
1137 unlock_page(pages[i]);
1138 page_cache_release(pages[i]);
1140 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1145 int btrfs_defrag_file(struct inode *inode, struct file *file,
1146 struct btrfs_ioctl_defrag_range_args *range,
1147 u64 newer_than, unsigned long max_to_defrag)
1149 struct btrfs_root *root = BTRFS_I(inode)->root;
1150 struct file_ra_state *ra = NULL;
1151 unsigned long last_index;
1152 u64 isize = i_size_read(inode);
1156 u64 newer_off = range->start;
1158 unsigned long ra_index = 0;
1160 int defrag_count = 0;
1161 int compress_type = BTRFS_COMPRESS_ZLIB;
1162 int extent_thresh = range->extent_thresh;
1163 int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1164 int cluster = max_cluster;
1165 u64 new_align = ~((u64)128 * 1024 - 1);
1166 struct page **pages = NULL;
1171 if (range->start >= isize)
1174 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1175 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1177 if (range->compress_type)
1178 compress_type = range->compress_type;
1181 if (extent_thresh == 0)
1182 extent_thresh = 256 * 1024;
1185 * if we were not given a file, allocate a readahead
1189 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1192 file_ra_state_init(ra, inode->i_mapping);
1197 pages = kmalloc(sizeof(struct page *) * max_cluster,
1204 /* find the last page to defrag */
1205 if (range->start + range->len > range->start) {
1206 last_index = min_t(u64, isize - 1,
1207 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1209 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1213 ret = find_new_extents(root, inode, newer_than,
1214 &newer_off, 64 * 1024);
1216 range->start = newer_off;
1218 * we always align our defrag to help keep
1219 * the extents in the file evenly spaced
1221 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1225 i = range->start >> PAGE_CACHE_SHIFT;
1228 max_to_defrag = last_index + 1;
1231 * make writeback starts from i, so the defrag range can be
1232 * written sequentially.
1234 if (i < inode->i_mapping->writeback_index)
1235 inode->i_mapping->writeback_index = i;
1237 while (i <= last_index && defrag_count < max_to_defrag &&
1238 (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1239 PAGE_CACHE_SHIFT)) {
1241 * make sure we stop running if someone unmounts
1244 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1247 if (btrfs_defrag_cancelled(root->fs_info)) {
1248 printk(KERN_DEBUG "btrfs: defrag_file cancelled\n");
1253 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1254 extent_thresh, &last_len, &skip,
1255 &defrag_end, range->flags &
1256 BTRFS_DEFRAG_RANGE_COMPRESS)) {
1259 * the should_defrag function tells us how much to skip
1260 * bump our counter by the suggested amount
1262 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1263 i = max(i + 1, next);
1268 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1269 PAGE_CACHE_SHIFT) - i;
1270 cluster = min(cluster, max_cluster);
1272 cluster = max_cluster;
1275 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1276 BTRFS_I(inode)->force_compress = compress_type;
1278 if (i + cluster > ra_index) {
1279 ra_index = max(i, ra_index);
1280 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1282 ra_index += max_cluster;
1285 mutex_lock(&inode->i_mutex);
1286 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1288 mutex_unlock(&inode->i_mutex);
1292 defrag_count += ret;
1293 balance_dirty_pages_ratelimited(inode->i_mapping);
1294 mutex_unlock(&inode->i_mutex);
1297 if (newer_off == (u64)-1)
1303 newer_off = max(newer_off + 1,
1304 (u64)i << PAGE_CACHE_SHIFT);
1306 ret = find_new_extents(root, inode,
1307 newer_than, &newer_off,
1310 range->start = newer_off;
1311 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1318 last_len += ret << PAGE_CACHE_SHIFT;
1326 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1327 filemap_flush(inode->i_mapping);
1329 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1330 /* the filemap_flush will queue IO into the worker threads, but
1331 * we have to make sure the IO is actually started and that
1332 * ordered extents get created before we return
1334 atomic_inc(&root->fs_info->async_submit_draining);
1335 while (atomic_read(&root->fs_info->nr_async_submits) ||
1336 atomic_read(&root->fs_info->async_delalloc_pages)) {
1337 wait_event(root->fs_info->async_submit_wait,
1338 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1339 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1341 atomic_dec(&root->fs_info->async_submit_draining);
1343 mutex_lock(&inode->i_mutex);
1344 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1345 mutex_unlock(&inode->i_mutex);
1348 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1349 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1361 static noinline int btrfs_ioctl_resize(struct file *file,
1367 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
1368 struct btrfs_ioctl_vol_args *vol_args;
1369 struct btrfs_trans_handle *trans;
1370 struct btrfs_device *device = NULL;
1372 char *devstr = NULL;
1376 if (!capable(CAP_SYS_ADMIN))
1379 ret = mnt_want_write_file(file);
1383 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1385 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
1386 mnt_drop_write_file(file);
1390 mutex_lock(&root->fs_info->volume_mutex);
1391 vol_args = memdup_user(arg, sizeof(*vol_args));
1392 if (IS_ERR(vol_args)) {
1393 ret = PTR_ERR(vol_args);
1397 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1399 sizestr = vol_args->name;
1400 devstr = strchr(sizestr, ':');
1403 sizestr = devstr + 1;
1405 devstr = vol_args->name;
1406 devid = simple_strtoull(devstr, &end, 10);
1411 printk(KERN_INFO "btrfs: resizing devid %llu\n",
1412 (unsigned long long)devid);
1415 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
1417 printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
1418 (unsigned long long)devid);
1423 if (!device->writeable) {
1424 printk(KERN_INFO "btrfs: resizer unable to apply on "
1425 "readonly device %llu\n",
1426 (unsigned long long)devid);
1431 if (!strcmp(sizestr, "max"))
1432 new_size = device->bdev->bd_inode->i_size;
1434 if (sizestr[0] == '-') {
1437 } else if (sizestr[0] == '+') {
1441 new_size = memparse(sizestr, NULL);
1442 if (new_size == 0) {
1448 if (device->is_tgtdev_for_dev_replace) {
1453 old_size = device->total_bytes;
1456 if (new_size > old_size) {
1460 new_size = old_size - new_size;
1461 } else if (mod > 0) {
1462 new_size = old_size + new_size;
1465 if (new_size < 256 * 1024 * 1024) {
1469 if (new_size > device->bdev->bd_inode->i_size) {
1474 do_div(new_size, root->sectorsize);
1475 new_size *= root->sectorsize;
1477 printk_in_rcu(KERN_INFO "btrfs: new size for %s is %llu\n",
1478 rcu_str_deref(device->name),
1479 (unsigned long long)new_size);
1481 if (new_size > old_size) {
1482 trans = btrfs_start_transaction(root, 0);
1483 if (IS_ERR(trans)) {
1484 ret = PTR_ERR(trans);
1487 ret = btrfs_grow_device(trans, device, new_size);
1488 btrfs_commit_transaction(trans, root);
1489 } else if (new_size < old_size) {
1490 ret = btrfs_shrink_device(device, new_size);
1491 } /* equal, nothing need to do */
1496 mutex_unlock(&root->fs_info->volume_mutex);
1497 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
1498 mnt_drop_write_file(file);
1502 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1503 char *name, unsigned long fd, int subvol,
1504 u64 *transid, bool readonly,
1505 struct btrfs_qgroup_inherit *inherit)
1510 ret = mnt_want_write_file(file);
1514 namelen = strlen(name);
1515 if (strchr(name, '/')) {
1517 goto out_drop_write;
1520 if (name[0] == '.' &&
1521 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1523 goto out_drop_write;
1527 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1528 NULL, transid, readonly, inherit);
1530 struct fd src = fdget(fd);
1531 struct inode *src_inode;
1534 goto out_drop_write;
1537 src_inode = file_inode(src.file);
1538 if (src_inode->i_sb != file_inode(file)->i_sb) {
1539 printk(KERN_INFO "btrfs: Snapshot src from "
1543 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1544 BTRFS_I(src_inode)->root,
1545 transid, readonly, inherit);
1550 mnt_drop_write_file(file);
1555 static noinline int btrfs_ioctl_snap_create(struct file *file,
1556 void __user *arg, int subvol)
1558 struct btrfs_ioctl_vol_args *vol_args;
1561 vol_args = memdup_user(arg, sizeof(*vol_args));
1562 if (IS_ERR(vol_args))
1563 return PTR_ERR(vol_args);
1564 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1566 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1567 vol_args->fd, subvol,
1574 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1575 void __user *arg, int subvol)
1577 struct btrfs_ioctl_vol_args_v2 *vol_args;
1581 bool readonly = false;
1582 struct btrfs_qgroup_inherit *inherit = NULL;
1584 vol_args = memdup_user(arg, sizeof(*vol_args));
1585 if (IS_ERR(vol_args))
1586 return PTR_ERR(vol_args);
1587 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1589 if (vol_args->flags &
1590 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1591 BTRFS_SUBVOL_QGROUP_INHERIT)) {
1596 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1598 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1600 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1601 if (vol_args->size > PAGE_CACHE_SIZE) {
1605 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1606 if (IS_ERR(inherit)) {
1607 ret = PTR_ERR(inherit);
1612 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1613 vol_args->fd, subvol, ptr,
1616 if (ret == 0 && ptr &&
1618 offsetof(struct btrfs_ioctl_vol_args_v2,
1619 transid), ptr, sizeof(*ptr)))
1627 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1630 struct inode *inode = file_inode(file);
1631 struct btrfs_root *root = BTRFS_I(inode)->root;
1635 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1638 down_read(&root->fs_info->subvol_sem);
1639 if (btrfs_root_readonly(root))
1640 flags |= BTRFS_SUBVOL_RDONLY;
1641 up_read(&root->fs_info->subvol_sem);
1643 if (copy_to_user(arg, &flags, sizeof(flags)))
1649 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1652 struct inode *inode = file_inode(file);
1653 struct btrfs_root *root = BTRFS_I(inode)->root;
1654 struct btrfs_trans_handle *trans;
1659 ret = mnt_want_write_file(file);
1663 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1665 goto out_drop_write;
1668 if (copy_from_user(&flags, arg, sizeof(flags))) {
1670 goto out_drop_write;
1673 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1675 goto out_drop_write;
1678 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1680 goto out_drop_write;
1683 if (!inode_owner_or_capable(inode)) {
1685 goto out_drop_write;
1688 down_write(&root->fs_info->subvol_sem);
1691 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1694 root_flags = btrfs_root_flags(&root->root_item);
1695 if (flags & BTRFS_SUBVOL_RDONLY)
1696 btrfs_set_root_flags(&root->root_item,
1697 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1699 btrfs_set_root_flags(&root->root_item,
1700 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1702 trans = btrfs_start_transaction(root, 1);
1703 if (IS_ERR(trans)) {
1704 ret = PTR_ERR(trans);
1708 ret = btrfs_update_root(trans, root->fs_info->tree_root,
1709 &root->root_key, &root->root_item);
1711 btrfs_commit_transaction(trans, root);
1714 btrfs_set_root_flags(&root->root_item, root_flags);
1716 up_write(&root->fs_info->subvol_sem);
1718 mnt_drop_write_file(file);
1724 * helper to check if the subvolume references other subvolumes
1726 static noinline int may_destroy_subvol(struct btrfs_root *root)
1728 struct btrfs_path *path;
1729 struct btrfs_dir_item *di;
1730 struct btrfs_key key;
1734 path = btrfs_alloc_path();
1738 /* Make sure this root isn't set as the default subvol */
1739 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1740 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1741 dir_id, "default", 7, 0);
1742 if (di && !IS_ERR(di)) {
1743 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1744 if (key.objectid == root->root_key.objectid) {
1748 btrfs_release_path(path);
1751 key.objectid = root->root_key.objectid;
1752 key.type = BTRFS_ROOT_REF_KEY;
1753 key.offset = (u64)-1;
1755 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1762 if (path->slots[0] > 0) {
1764 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1765 if (key.objectid == root->root_key.objectid &&
1766 key.type == BTRFS_ROOT_REF_KEY)
1770 btrfs_free_path(path);
1774 static noinline int key_in_sk(struct btrfs_key *key,
1775 struct btrfs_ioctl_search_key *sk)
1777 struct btrfs_key test;
1780 test.objectid = sk->min_objectid;
1781 test.type = sk->min_type;
1782 test.offset = sk->min_offset;
1784 ret = btrfs_comp_cpu_keys(key, &test);
1788 test.objectid = sk->max_objectid;
1789 test.type = sk->max_type;
1790 test.offset = sk->max_offset;
1792 ret = btrfs_comp_cpu_keys(key, &test);
1798 static noinline int copy_to_sk(struct btrfs_root *root,
1799 struct btrfs_path *path,
1800 struct btrfs_key *key,
1801 struct btrfs_ioctl_search_key *sk,
1803 unsigned long *sk_offset,
1807 struct extent_buffer *leaf;
1808 struct btrfs_ioctl_search_header sh;
1809 unsigned long item_off;
1810 unsigned long item_len;
1816 leaf = path->nodes[0];
1817 slot = path->slots[0];
1818 nritems = btrfs_header_nritems(leaf);
1820 if (btrfs_header_generation(leaf) > sk->max_transid) {
1824 found_transid = btrfs_header_generation(leaf);
1826 for (i = slot; i < nritems; i++) {
1827 item_off = btrfs_item_ptr_offset(leaf, i);
1828 item_len = btrfs_item_size_nr(leaf, i);
1830 btrfs_item_key_to_cpu(leaf, key, i);
1831 if (!key_in_sk(key, sk))
1834 if (sizeof(sh) + item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1837 if (sizeof(sh) + item_len + *sk_offset >
1838 BTRFS_SEARCH_ARGS_BUFSIZE) {
1843 sh.objectid = key->objectid;
1844 sh.offset = key->offset;
1845 sh.type = key->type;
1847 sh.transid = found_transid;
1849 /* copy search result header */
1850 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1851 *sk_offset += sizeof(sh);
1854 char *p = buf + *sk_offset;
1856 read_extent_buffer(leaf, p,
1857 item_off, item_len);
1858 *sk_offset += item_len;
1862 if (*num_found >= sk->nr_items)
1867 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1869 else if (key->type < (u8)-1 && key->type < sk->max_type) {
1872 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1882 static noinline int search_ioctl(struct inode *inode,
1883 struct btrfs_ioctl_search_args *args)
1885 struct btrfs_root *root;
1886 struct btrfs_key key;
1887 struct btrfs_key max_key;
1888 struct btrfs_path *path;
1889 struct btrfs_ioctl_search_key *sk = &args->key;
1890 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1893 unsigned long sk_offset = 0;
1895 path = btrfs_alloc_path();
1899 if (sk->tree_id == 0) {
1900 /* search the root of the inode that was passed */
1901 root = BTRFS_I(inode)->root;
1903 key.objectid = sk->tree_id;
1904 key.type = BTRFS_ROOT_ITEM_KEY;
1905 key.offset = (u64)-1;
1906 root = btrfs_read_fs_root_no_name(info, &key);
1908 printk(KERN_ERR "could not find root %llu\n",
1910 btrfs_free_path(path);
1915 key.objectid = sk->min_objectid;
1916 key.type = sk->min_type;
1917 key.offset = sk->min_offset;
1919 max_key.objectid = sk->max_objectid;
1920 max_key.type = sk->max_type;
1921 max_key.offset = sk->max_offset;
1923 path->keep_locks = 1;
1926 ret = btrfs_search_forward(root, &key, &max_key, path,
1933 ret = copy_to_sk(root, path, &key, sk, args->buf,
1934 &sk_offset, &num_found);
1935 btrfs_release_path(path);
1936 if (ret || num_found >= sk->nr_items)
1942 sk->nr_items = num_found;
1943 btrfs_free_path(path);
1947 static noinline int btrfs_ioctl_tree_search(struct file *file,
1950 struct btrfs_ioctl_search_args *args;
1951 struct inode *inode;
1954 if (!capable(CAP_SYS_ADMIN))
1957 args = memdup_user(argp, sizeof(*args));
1959 return PTR_ERR(args);
1961 inode = file_inode(file);
1962 ret = search_ioctl(inode, args);
1963 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1970 * Search INODE_REFs to identify path name of 'dirid' directory
1971 * in a 'tree_id' tree. and sets path name to 'name'.
1973 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1974 u64 tree_id, u64 dirid, char *name)
1976 struct btrfs_root *root;
1977 struct btrfs_key key;
1983 struct btrfs_inode_ref *iref;
1984 struct extent_buffer *l;
1985 struct btrfs_path *path;
1987 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1992 path = btrfs_alloc_path();
1996 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1998 key.objectid = tree_id;
1999 key.type = BTRFS_ROOT_ITEM_KEY;
2000 key.offset = (u64)-1;
2001 root = btrfs_read_fs_root_no_name(info, &key);
2003 printk(KERN_ERR "could not find root %llu\n", tree_id);
2008 key.objectid = dirid;
2009 key.type = BTRFS_INODE_REF_KEY;
2010 key.offset = (u64)-1;
2013 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2018 slot = path->slots[0];
2019 if (ret > 0 && slot > 0)
2021 btrfs_item_key_to_cpu(l, &key, slot);
2023 if (ret > 0 && (key.objectid != dirid ||
2024 key.type != BTRFS_INODE_REF_KEY)) {
2029 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2030 len = btrfs_inode_ref_name_len(l, iref);
2032 total_len += len + 1;
2037 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
2039 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2042 btrfs_release_path(path);
2043 key.objectid = key.offset;
2044 key.offset = (u64)-1;
2045 dirid = key.objectid;
2049 memmove(name, ptr, total_len);
2050 name[total_len]='\0';
2053 btrfs_free_path(path);
2057 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2060 struct btrfs_ioctl_ino_lookup_args *args;
2061 struct inode *inode;
2064 if (!capable(CAP_SYS_ADMIN))
2067 args = memdup_user(argp, sizeof(*args));
2069 return PTR_ERR(args);
2071 inode = file_inode(file);
2073 if (args->treeid == 0)
2074 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2076 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2077 args->treeid, args->objectid,
2080 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2087 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2090 struct dentry *parent = fdentry(file);
2091 struct dentry *dentry;
2092 struct inode *dir = parent->d_inode;
2093 struct inode *inode;
2094 struct btrfs_root *root = BTRFS_I(dir)->root;
2095 struct btrfs_root *dest = NULL;
2096 struct btrfs_ioctl_vol_args *vol_args;
2097 struct btrfs_trans_handle *trans;
2098 struct btrfs_block_rsv block_rsv;
2099 u64 qgroup_reserved;
2104 vol_args = memdup_user(arg, sizeof(*vol_args));
2105 if (IS_ERR(vol_args))
2106 return PTR_ERR(vol_args);
2108 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2109 namelen = strlen(vol_args->name);
2110 if (strchr(vol_args->name, '/') ||
2111 strncmp(vol_args->name, "..", namelen) == 0) {
2116 err = mnt_want_write_file(file);
2120 err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2123 dentry = lookup_one_len(vol_args->name, parent, namelen);
2124 if (IS_ERR(dentry)) {
2125 err = PTR_ERR(dentry);
2126 goto out_unlock_dir;
2129 if (!dentry->d_inode) {
2134 inode = dentry->d_inode;
2135 dest = BTRFS_I(inode)->root;
2136 if (!capable(CAP_SYS_ADMIN)){
2138 * Regular user. Only allow this with a special mount
2139 * option, when the user has write+exec access to the
2140 * subvol root, and when rmdir(2) would have been
2143 * Note that this is _not_ check that the subvol is
2144 * empty or doesn't contain data that we wouldn't
2145 * otherwise be able to delete.
2147 * Users who want to delete empty subvols should try
2151 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2155 * Do not allow deletion if the parent dir is the same
2156 * as the dir to be deleted. That means the ioctl
2157 * must be called on the dentry referencing the root
2158 * of the subvol, not a random directory contained
2165 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2170 /* check if subvolume may be deleted by a user */
2171 err = btrfs_may_delete(dir, dentry, 1);
2175 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2180 mutex_lock(&inode->i_mutex);
2181 err = d_invalidate(dentry);
2185 down_write(&root->fs_info->subvol_sem);
2187 err = may_destroy_subvol(dest);
2191 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2193 * One for dir inode, two for dir entries, two for root
2196 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2197 5, &qgroup_reserved, true);
2201 trans = btrfs_start_transaction(root, 0);
2202 if (IS_ERR(trans)) {
2203 err = PTR_ERR(trans);
2206 trans->block_rsv = &block_rsv;
2207 trans->bytes_reserved = block_rsv.size;
2209 ret = btrfs_unlink_subvol(trans, root, dir,
2210 dest->root_key.objectid,
2211 dentry->d_name.name,
2212 dentry->d_name.len);
2215 btrfs_abort_transaction(trans, root, ret);
2219 btrfs_record_root_in_trans(trans, dest);
2221 memset(&dest->root_item.drop_progress, 0,
2222 sizeof(dest->root_item.drop_progress));
2223 dest->root_item.drop_level = 0;
2224 btrfs_set_root_refs(&dest->root_item, 0);
2226 if (!xchg(&dest->orphan_item_inserted, 1)) {
2227 ret = btrfs_insert_orphan_item(trans,
2228 root->fs_info->tree_root,
2229 dest->root_key.objectid);
2231 btrfs_abort_transaction(trans, root, ret);
2237 trans->block_rsv = NULL;
2238 trans->bytes_reserved = 0;
2239 ret = btrfs_end_transaction(trans, root);
2242 inode->i_flags |= S_DEAD;
2244 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
2246 up_write(&root->fs_info->subvol_sem);
2248 mutex_unlock(&inode->i_mutex);
2250 shrink_dcache_sb(root->fs_info->sb);
2251 btrfs_invalidate_inodes(dest);
2255 if (dest->cache_inode) {
2256 iput(dest->cache_inode);
2257 dest->cache_inode = NULL;
2263 mutex_unlock(&dir->i_mutex);
2264 mnt_drop_write_file(file);
2270 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2272 struct inode *inode = file_inode(file);
2273 struct btrfs_root *root = BTRFS_I(inode)->root;
2274 struct btrfs_ioctl_defrag_range_args *range;
2277 ret = mnt_want_write_file(file);
2281 if (btrfs_root_readonly(root)) {
2286 switch (inode->i_mode & S_IFMT) {
2288 if (!capable(CAP_SYS_ADMIN)) {
2292 ret = btrfs_defrag_root(root);
2295 ret = btrfs_defrag_root(root->fs_info->extent_root);
2298 if (!(file->f_mode & FMODE_WRITE)) {
2303 range = kzalloc(sizeof(*range), GFP_KERNEL);
2310 if (copy_from_user(range, argp,
2316 /* compression requires us to start the IO */
2317 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2318 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2319 range->extent_thresh = (u32)-1;
2322 /* the rest are all set to zero by kzalloc */
2323 range->len = (u64)-1;
2325 ret = btrfs_defrag_file(file_inode(file), file,
2335 mnt_drop_write_file(file);
2339 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2341 struct btrfs_ioctl_vol_args *vol_args;
2344 if (!capable(CAP_SYS_ADMIN))
2347 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2349 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2353 mutex_lock(&root->fs_info->volume_mutex);
2354 vol_args = memdup_user(arg, sizeof(*vol_args));
2355 if (IS_ERR(vol_args)) {
2356 ret = PTR_ERR(vol_args);
2360 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2361 ret = btrfs_init_new_device(root, vol_args->name);
2365 mutex_unlock(&root->fs_info->volume_mutex);
2366 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2370 static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
2372 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2373 struct btrfs_ioctl_vol_args *vol_args;
2376 if (!capable(CAP_SYS_ADMIN))
2379 ret = mnt_want_write_file(file);
2383 vol_args = memdup_user(arg, sizeof(*vol_args));
2384 if (IS_ERR(vol_args)) {
2385 ret = PTR_ERR(vol_args);
2389 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2391 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2393 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2397 mutex_lock(&root->fs_info->volume_mutex);
2398 ret = btrfs_rm_device(root, vol_args->name);
2399 mutex_unlock(&root->fs_info->volume_mutex);
2400 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2404 mnt_drop_write_file(file);
2408 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2410 struct btrfs_ioctl_fs_info_args *fi_args;
2411 struct btrfs_device *device;
2412 struct btrfs_device *next;
2413 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2416 if (!capable(CAP_SYS_ADMIN))
2419 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2423 fi_args->num_devices = fs_devices->num_devices;
2424 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2426 mutex_lock(&fs_devices->device_list_mutex);
2427 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2428 if (device->devid > fi_args->max_id)
2429 fi_args->max_id = device->devid;
2431 mutex_unlock(&fs_devices->device_list_mutex);
2433 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2440 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2442 struct btrfs_ioctl_dev_info_args *di_args;
2443 struct btrfs_device *dev;
2444 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2446 char *s_uuid = NULL;
2447 char empty_uuid[BTRFS_UUID_SIZE] = {0};
2449 if (!capable(CAP_SYS_ADMIN))
2452 di_args = memdup_user(arg, sizeof(*di_args));
2453 if (IS_ERR(di_args))
2454 return PTR_ERR(di_args);
2456 if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2457 s_uuid = di_args->uuid;
2459 mutex_lock(&fs_devices->device_list_mutex);
2460 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
2467 di_args->devid = dev->devid;
2468 di_args->bytes_used = dev->bytes_used;
2469 di_args->total_bytes = dev->total_bytes;
2470 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2472 struct rcu_string *name;
2475 name = rcu_dereference(dev->name);
2476 strncpy(di_args->path, name->str, sizeof(di_args->path));
2478 di_args->path[sizeof(di_args->path) - 1] = 0;
2480 di_args->path[0] = '\0';
2484 mutex_unlock(&fs_devices->device_list_mutex);
2485 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2492 static struct page *extent_same_get_page(struct inode *inode, u64 off)
2496 struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
2498 index = off >> PAGE_CACHE_SHIFT;
2500 page = grab_cache_page(inode->i_mapping, index);
2504 if (!PageUptodate(page)) {
2505 if (extent_read_full_page_nolock(tree, page, btrfs_get_extent,
2509 if (!PageUptodate(page)) {
2511 page_cache_release(page);
2520 static inline void lock_extent_range(struct inode *inode, u64 off, u64 len)
2522 /* do any pending delalloc/csum calc on src, one way or
2523 another, and lock file content */
2525 struct btrfs_ordered_extent *ordered;
2526 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2527 ordered = btrfs_lookup_first_ordered_extent(inode,
2530 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2531 off + len - 1, EXTENT_DELALLOC, 0, NULL))
2533 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2535 btrfs_put_ordered_extent(ordered);
2536 btrfs_wait_ordered_range(inode, off, len);
2540 static void btrfs_double_unlock(struct inode *inode1, u64 loff1,
2541 struct inode *inode2, u64 loff2, u64 len)
2543 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2544 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2546 mutex_unlock(&inode1->i_mutex);
2547 mutex_unlock(&inode2->i_mutex);
2550 static void btrfs_double_lock(struct inode *inode1, u64 loff1,
2551 struct inode *inode2, u64 loff2, u64 len)
2553 if (inode1 < inode2) {
2554 swap(inode1, inode2);
2558 mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT);
2559 lock_extent_range(inode1, loff1, len);
2560 if (inode1 != inode2) {
2561 mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD);
2562 lock_extent_range(inode2, loff2, len);
2566 static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
2567 u64 dst_loff, u64 len)
2570 struct page *src_page, *dst_page;
2571 unsigned int cmp_len = PAGE_CACHE_SIZE;
2572 void *addr, *dst_addr;
2575 if (len < PAGE_CACHE_SIZE)
2578 src_page = extent_same_get_page(src, loff);
2581 dst_page = extent_same_get_page(dst, dst_loff);
2583 page_cache_release(src_page);
2586 addr = kmap_atomic(src_page);
2587 dst_addr = kmap_atomic(dst_page);
2589 flush_dcache_page(src_page);
2590 flush_dcache_page(dst_page);
2592 if (memcmp(addr, dst_addr, cmp_len))
2593 ret = BTRFS_SAME_DATA_DIFFERS;
2595 kunmap_atomic(addr);
2596 kunmap_atomic(dst_addr);
2597 page_cache_release(src_page);
2598 page_cache_release(dst_page);
2604 dst_loff += cmp_len;
2611 static int extent_same_check_offsets(struct inode *inode, u64 off, u64 len)
2613 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
2615 if (off + len > inode->i_size || off + len < off)
2617 /* Check that we are block aligned - btrfs_clone() requires this */
2618 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
2624 static int btrfs_extent_same(struct inode *src, u64 loff, u64 len,
2625 struct inode *dst, u64 dst_loff)
2630 * btrfs_clone() can't handle extents in the same file
2631 * yet. Once that works, we can drop this check and replace it
2632 * with a check for the same inode, but overlapping extents.
2637 btrfs_double_lock(src, loff, dst, dst_loff, len);
2639 ret = extent_same_check_offsets(src, loff, len);
2643 ret = extent_same_check_offsets(dst, dst_loff, len);
2647 /* don't make the dst file partly checksummed */
2648 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2649 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
2654 ret = btrfs_cmp_data(src, loff, dst, dst_loff, len);
2656 ret = btrfs_clone(src, dst, loff, len, len, dst_loff);
2659 btrfs_double_unlock(src, loff, dst, dst_loff, len);
2664 #define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024)
2666 static long btrfs_ioctl_file_extent_same(struct file *file,
2669 struct btrfs_ioctl_same_args *args = argp;
2670 struct btrfs_ioctl_same_args same;
2671 struct btrfs_ioctl_same_extent_info info;
2672 struct inode *src = file->f_dentry->d_inode;
2673 struct file *dst_file = NULL;
2679 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
2680 bool is_admin = capable(CAP_SYS_ADMIN);
2682 if (!(file->f_mode & FMODE_READ))
2685 ret = mnt_want_write_file(file);
2689 if (copy_from_user(&same,
2690 (struct btrfs_ioctl_same_args __user *)argp,
2696 off = same.logical_offset;
2700 * Limit the total length we will dedupe for each operation.
2701 * This is intended to bound the total time spent in this
2702 * ioctl to something sane.
2704 if (len > BTRFS_MAX_DEDUPE_LEN)
2705 len = BTRFS_MAX_DEDUPE_LEN;
2707 if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) {
2709 * Btrfs does not support blocksize < page_size. As a
2710 * result, btrfs_cmp_data() won't correctly handle
2711 * this situation without an update.
2718 if (S_ISDIR(src->i_mode))
2722 if (!S_ISREG(src->i_mode))
2726 for (i = 0; i < same.dest_count; i++) {
2727 if (copy_from_user(&info, &args->info[i], sizeof(info))) {
2732 info.bytes_deduped = 0;
2734 dst_file = fget(info.fd);
2736 info.status = -EBADF;
2740 if (!(is_admin || (dst_file->f_mode & FMODE_WRITE))) {
2741 info.status = -EINVAL;
2745 info.status = -EXDEV;
2746 if (file->f_path.mnt != dst_file->f_path.mnt)
2749 dst = dst_file->f_dentry->d_inode;
2750 if (src->i_sb != dst->i_sb)
2753 if (S_ISDIR(dst->i_mode)) {
2754 info.status = -EISDIR;
2758 if (!S_ISREG(dst->i_mode)) {
2759 info.status = -EACCES;
2763 info.status = btrfs_extent_same(src, off, len, dst,
2764 info.logical_offset);
2765 if (info.status == 0)
2766 info.bytes_deduped += len;
2772 if (__put_user_unaligned(info.status, &args->info[i].status) ||
2773 __put_user_unaligned(info.bytes_deduped,
2774 &args->info[i].bytes_deduped)) {
2781 mnt_drop_write_file(file);
2786 * btrfs_clone() - clone a range from inode file to another
2788 * @src: Inode to clone from
2789 * @inode: Inode to clone to
2790 * @off: Offset within source to start clone from
2791 * @olen: Original length, passed by user, of range to clone
2792 * @olen_aligned: Block-aligned value of olen, extent_same uses
2793 * identical values here
2794 * @destoff: Offset within @inode to start clone
2796 static int btrfs_clone(struct inode *src, struct inode *inode,
2797 u64 off, u64 olen, u64 olen_aligned, u64 destoff)
2799 struct btrfs_root *root = BTRFS_I(inode)->root;
2800 struct btrfs_path *path = NULL;
2801 struct extent_buffer *leaf;
2802 struct btrfs_trans_handle *trans;
2804 struct btrfs_key key;
2808 u64 len = olen_aligned;
2811 buf = vmalloc(btrfs_level_size(root, 0));
2815 path = btrfs_alloc_path();
2823 key.objectid = btrfs_ino(src);
2824 key.type = BTRFS_EXTENT_DATA_KEY;
2829 * note the key will change type as we walk through the
2832 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
2837 nritems = btrfs_header_nritems(path->nodes[0]);
2838 if (path->slots[0] >= nritems) {
2839 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
2844 nritems = btrfs_header_nritems(path->nodes[0]);
2846 leaf = path->nodes[0];
2847 slot = path->slots[0];
2849 btrfs_item_key_to_cpu(leaf, &key, slot);
2850 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
2851 key.objectid != btrfs_ino(src))
2854 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2855 struct btrfs_file_extent_item *extent;
2858 struct btrfs_key new_key;
2859 u64 disko = 0, diskl = 0;
2860 u64 datao = 0, datal = 0;
2864 size = btrfs_item_size_nr(leaf, slot);
2865 read_extent_buffer(leaf, buf,
2866 btrfs_item_ptr_offset(leaf, slot),
2869 extent = btrfs_item_ptr(leaf, slot,
2870 struct btrfs_file_extent_item);
2871 comp = btrfs_file_extent_compression(leaf, extent);
2872 type = btrfs_file_extent_type(leaf, extent);
2873 if (type == BTRFS_FILE_EXTENT_REG ||
2874 type == BTRFS_FILE_EXTENT_PREALLOC) {
2875 disko = btrfs_file_extent_disk_bytenr(leaf,
2877 diskl = btrfs_file_extent_disk_num_bytes(leaf,
2879 datao = btrfs_file_extent_offset(leaf, extent);
2880 datal = btrfs_file_extent_num_bytes(leaf,
2882 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2883 /* take upper bound, may be compressed */
2884 datal = btrfs_file_extent_ram_bytes(leaf,
2887 btrfs_release_path(path);
2889 if (key.offset + datal <= off ||
2890 key.offset >= off + len - 1)
2893 memcpy(&new_key, &key, sizeof(new_key));
2894 new_key.objectid = btrfs_ino(inode);
2895 if (off <= key.offset)
2896 new_key.offset = key.offset + destoff - off;
2898 new_key.offset = destoff;
2901 * 1 - adjusting old extent (we may have to split it)
2902 * 1 - add new extent
2905 trans = btrfs_start_transaction(root, 3);
2906 if (IS_ERR(trans)) {
2907 ret = PTR_ERR(trans);
2911 if (type == BTRFS_FILE_EXTENT_REG ||
2912 type == BTRFS_FILE_EXTENT_PREALLOC) {
2914 * a | --- range to clone ---| b
2915 * | ------------- extent ------------- |
2918 /* substract range b */
2919 if (key.offset + datal > off + len)
2920 datal = off + len - key.offset;
2922 /* substract range a */
2923 if (off > key.offset) {
2924 datao += off - key.offset;
2925 datal -= off - key.offset;
2928 ret = btrfs_drop_extents(trans, root, inode,
2930 new_key.offset + datal,
2933 btrfs_abort_transaction(trans, root,
2935 btrfs_end_transaction(trans, root);
2939 ret = btrfs_insert_empty_item(trans, root, path,
2942 btrfs_abort_transaction(trans, root,
2944 btrfs_end_transaction(trans, root);
2948 leaf = path->nodes[0];
2949 slot = path->slots[0];
2950 write_extent_buffer(leaf, buf,
2951 btrfs_item_ptr_offset(leaf, slot),
2954 extent = btrfs_item_ptr(leaf, slot,
2955 struct btrfs_file_extent_item);
2957 /* disko == 0 means it's a hole */
2961 btrfs_set_file_extent_offset(leaf, extent,
2963 btrfs_set_file_extent_num_bytes(leaf, extent,
2966 inode_add_bytes(inode, datal);
2967 ret = btrfs_inc_extent_ref(trans, root,
2969 root->root_key.objectid,
2971 new_key.offset - datao,
2974 btrfs_abort_transaction(trans,
2977 btrfs_end_transaction(trans,
2983 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2986 if (off > key.offset) {
2987 skip = off - key.offset;
2988 new_key.offset += skip;
2991 if (key.offset + datal > off + len)
2992 trim = key.offset + datal - (off + len);
2994 if (comp && (skip || trim)) {
2996 btrfs_end_transaction(trans, root);
2999 size -= skip + trim;
3000 datal -= skip + trim;
3002 ret = btrfs_drop_extents(trans, root, inode,
3004 new_key.offset + datal,
3007 btrfs_abort_transaction(trans, root,
3009 btrfs_end_transaction(trans, root);
3013 ret = btrfs_insert_empty_item(trans, root, path,
3016 btrfs_abort_transaction(trans, root,
3018 btrfs_end_transaction(trans, root);
3024 btrfs_file_extent_calc_inline_size(0);
3025 memmove(buf+start, buf+start+skip,
3029 leaf = path->nodes[0];
3030 slot = path->slots[0];
3031 write_extent_buffer(leaf, buf,
3032 btrfs_item_ptr_offset(leaf, slot),
3034 inode_add_bytes(inode, datal);
3037 btrfs_mark_buffer_dirty(leaf);
3038 btrfs_release_path(path);
3040 inode_inc_iversion(inode);
3041 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
3044 * we round up to the block size at eof when
3045 * determining which extents to clone above,
3046 * but shouldn't round up the file size
3048 endoff = new_key.offset + datal;
3049 if (endoff > destoff+olen)
3050 endoff = destoff+olen;
3051 if (endoff > inode->i_size)
3052 btrfs_i_size_write(inode, endoff);
3054 ret = btrfs_update_inode(trans, root, inode);
3056 btrfs_abort_transaction(trans, root, ret);
3057 btrfs_end_transaction(trans, root);
3060 ret = btrfs_end_transaction(trans, root);
3063 btrfs_release_path(path);
3069 btrfs_release_path(path);
3070 btrfs_free_path(path);
3075 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
3076 u64 off, u64 olen, u64 destoff)
3078 struct inode *inode = fdentry(file)->d_inode;
3079 struct btrfs_root *root = BTRFS_I(inode)->root;
3084 u64 bs = root->fs_info->sb->s_blocksize;
3089 * - split compressed inline extents. annoying: we need to
3090 * decompress into destination's address_space (the file offset
3091 * may change, so source mapping won't do), then recompress (or
3092 * otherwise reinsert) a subrange.
3093 * - allow ranges within the same file to be cloned (provided
3094 * they don't overlap)?
3097 /* the destination must be opened for writing */
3098 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
3101 if (btrfs_root_readonly(root))
3104 ret = mnt_want_write_file(file);
3108 src_file = fdget(srcfd);
3109 if (!src_file.file) {
3111 goto out_drop_write;
3115 if (src_file.file->f_path.mnt != file->f_path.mnt)
3118 src = file_inode(src_file.file);
3124 /* the src must be open for reading */
3125 if (!(src_file.file->f_mode & FMODE_READ))
3128 /* don't make the dst file partly checksummed */
3129 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3130 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3134 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3138 if (src->i_sb != inode->i_sb)
3143 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
3144 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
3146 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
3147 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
3150 mutex_lock(&src->i_mutex);
3153 /* determine range to clone */
3155 if (off + len > src->i_size || off + len < off)
3158 olen = len = src->i_size - off;
3159 /* if we extend to eof, continue to block boundary */
3160 if (off + len == src->i_size)
3161 len = ALIGN(src->i_size, bs) - off;
3163 /* verify the end result is block aligned */
3164 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3165 !IS_ALIGNED(destoff, bs))
3168 /* verify if ranges are overlapped within the same file */
3170 if (destoff + len > off && destoff < off + len)
3174 if (destoff > inode->i_size) {
3175 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3180 /* truncate page cache pages from target inode range */
3181 truncate_inode_pages_range(&inode->i_data, destoff,
3182 PAGE_CACHE_ALIGN(destoff + len) - 1);
3184 lock_extent_range(src, off, len);
3186 ret = btrfs_clone(src, inode, off, olen, len, destoff);
3188 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
3190 mutex_unlock(&src->i_mutex);
3192 mutex_unlock(&inode->i_mutex);
3196 mnt_drop_write_file(file);
3200 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
3202 struct btrfs_ioctl_clone_range_args args;
3204 if (copy_from_user(&args, argp, sizeof(args)))
3206 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
3207 args.src_length, args.dest_offset);
3211 * there are many ways the trans_start and trans_end ioctls can lead
3212 * to deadlocks. They should only be used by applications that
3213 * basically own the machine, and have a very in depth understanding
3214 * of all the possible deadlocks and enospc problems.
3216 static long btrfs_ioctl_trans_start(struct file *file)
3218 struct inode *inode = file_inode(file);
3219 struct btrfs_root *root = BTRFS_I(inode)->root;
3220 struct btrfs_trans_handle *trans;
3224 if (!capable(CAP_SYS_ADMIN))
3228 if (file->private_data)
3232 if (btrfs_root_readonly(root))
3235 ret = mnt_want_write_file(file);
3239 atomic_inc(&root->fs_info->open_ioctl_trans);
3242 trans = btrfs_start_ioctl_transaction(root);
3246 file->private_data = trans;
3250 atomic_dec(&root->fs_info->open_ioctl_trans);
3251 mnt_drop_write_file(file);
3256 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3258 struct inode *inode = file_inode(file);
3259 struct btrfs_root *root = BTRFS_I(inode)->root;
3260 struct btrfs_root *new_root;
3261 struct btrfs_dir_item *di;
3262 struct btrfs_trans_handle *trans;
3263 struct btrfs_path *path;
3264 struct btrfs_key location;
3265 struct btrfs_disk_key disk_key;
3270 if (!capable(CAP_SYS_ADMIN))
3273 ret = mnt_want_write_file(file);
3277 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3283 objectid = root->root_key.objectid;
3285 location.objectid = objectid;
3286 location.type = BTRFS_ROOT_ITEM_KEY;
3287 location.offset = (u64)-1;
3289 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
3290 if (IS_ERR(new_root)) {
3291 ret = PTR_ERR(new_root);
3295 path = btrfs_alloc_path();
3300 path->leave_spinning = 1;
3302 trans = btrfs_start_transaction(root, 1);
3303 if (IS_ERR(trans)) {
3304 btrfs_free_path(path);
3305 ret = PTR_ERR(trans);
3309 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
3310 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
3311 dir_id, "default", 7, 1);
3312 if (IS_ERR_OR_NULL(di)) {
3313 btrfs_free_path(path);
3314 btrfs_end_transaction(trans, root);
3315 printk(KERN_ERR "Umm, you don't have the default dir item, "
3316 "this isn't going to work\n");
3321 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3322 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3323 btrfs_mark_buffer_dirty(path->nodes[0]);
3324 btrfs_free_path(path);
3326 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
3327 btrfs_end_transaction(trans, root);
3329 mnt_drop_write_file(file);
3333 void btrfs_get_block_group_info(struct list_head *groups_list,
3334 struct btrfs_ioctl_space_info *space)
3336 struct btrfs_block_group_cache *block_group;
3338 space->total_bytes = 0;
3339 space->used_bytes = 0;
3341 list_for_each_entry(block_group, groups_list, list) {
3342 space->flags = block_group->flags;
3343 space->total_bytes += block_group->key.offset;
3344 space->used_bytes +=
3345 btrfs_block_group_used(&block_group->item);
3349 static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
3351 struct btrfs_ioctl_space_args space_args;
3352 struct btrfs_ioctl_space_info space;
3353 struct btrfs_ioctl_space_info *dest;
3354 struct btrfs_ioctl_space_info *dest_orig;
3355 struct btrfs_ioctl_space_info __user *user_dest;
3356 struct btrfs_space_info *info;
3357 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3358 BTRFS_BLOCK_GROUP_SYSTEM,
3359 BTRFS_BLOCK_GROUP_METADATA,
3360 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3367 if (copy_from_user(&space_args,
3368 (struct btrfs_ioctl_space_args __user *)arg,
3369 sizeof(space_args)))
3372 for (i = 0; i < num_types; i++) {
3373 struct btrfs_space_info *tmp;
3377 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3379 if (tmp->flags == types[i]) {
3389 down_read(&info->groups_sem);
3390 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3391 if (!list_empty(&info->block_groups[c]))
3394 up_read(&info->groups_sem);
3397 /* space_slots == 0 means they are asking for a count */
3398 if (space_args.space_slots == 0) {
3399 space_args.total_spaces = slot_count;
3403 slot_count = min_t(u64, space_args.space_slots, slot_count);
3405 alloc_size = sizeof(*dest) * slot_count;
3407 /* we generally have at most 6 or so space infos, one for each raid
3408 * level. So, a whole page should be more than enough for everyone
3410 if (alloc_size > PAGE_CACHE_SIZE)
3413 space_args.total_spaces = 0;
3414 dest = kmalloc(alloc_size, GFP_NOFS);
3419 /* now we have a buffer to copy into */
3420 for (i = 0; i < num_types; i++) {
3421 struct btrfs_space_info *tmp;
3428 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3430 if (tmp->flags == types[i]) {
3439 down_read(&info->groups_sem);
3440 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3441 if (!list_empty(&info->block_groups[c])) {
3442 btrfs_get_block_group_info(
3443 &info->block_groups[c], &space);
3444 memcpy(dest, &space, sizeof(space));
3446 space_args.total_spaces++;
3452 up_read(&info->groups_sem);
3455 user_dest = (struct btrfs_ioctl_space_info __user *)
3456 (arg + sizeof(struct btrfs_ioctl_space_args));
3458 if (copy_to_user(user_dest, dest_orig, alloc_size))
3463 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3470 * there are many ways the trans_start and trans_end ioctls can lead
3471 * to deadlocks. They should only be used by applications that
3472 * basically own the machine, and have a very in depth understanding
3473 * of all the possible deadlocks and enospc problems.
3475 long btrfs_ioctl_trans_end(struct file *file)
3477 struct inode *inode = file_inode(file);
3478 struct btrfs_root *root = BTRFS_I(inode)->root;
3479 struct btrfs_trans_handle *trans;
3481 trans = file->private_data;
3484 file->private_data = NULL;
3486 btrfs_end_transaction(trans, root);
3488 atomic_dec(&root->fs_info->open_ioctl_trans);
3490 mnt_drop_write_file(file);
3494 static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3497 struct btrfs_trans_handle *trans;
3501 trans = btrfs_attach_transaction_barrier(root);
3502 if (IS_ERR(trans)) {
3503 if (PTR_ERR(trans) != -ENOENT)
3504 return PTR_ERR(trans);
3506 /* No running transaction, don't bother */
3507 transid = root->fs_info->last_trans_committed;
3510 transid = trans->transid;
3511 ret = btrfs_commit_transaction_async(trans, root, 0);
3513 btrfs_end_transaction(trans, root);
3518 if (copy_to_user(argp, &transid, sizeof(transid)))
3523 static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
3529 if (copy_from_user(&transid, argp, sizeof(transid)))
3532 transid = 0; /* current trans */
3534 return btrfs_wait_for_commit(root, transid);
3537 static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
3539 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3540 struct btrfs_ioctl_scrub_args *sa;
3543 if (!capable(CAP_SYS_ADMIN))
3546 sa = memdup_user(arg, sizeof(*sa));
3550 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3551 ret = mnt_want_write_file(file);
3556 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
3557 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3560 if (copy_to_user(arg, sa, sizeof(*sa)))
3563 if (!(sa->flags & BTRFS_SCRUB_READONLY))
3564 mnt_drop_write_file(file);
3570 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3572 if (!capable(CAP_SYS_ADMIN))
3575 return btrfs_scrub_cancel(root->fs_info);
3578 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3581 struct btrfs_ioctl_scrub_args *sa;
3584 if (!capable(CAP_SYS_ADMIN))
3587 sa = memdup_user(arg, sizeof(*sa));
3591 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3593 if (copy_to_user(arg, sa, sizeof(*sa)))
3600 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
3603 struct btrfs_ioctl_get_dev_stats *sa;
3606 sa = memdup_user(arg, sizeof(*sa));
3610 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3615 ret = btrfs_get_dev_stats(root, sa);
3617 if (copy_to_user(arg, sa, sizeof(*sa)))
3624 static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
3626 struct btrfs_ioctl_dev_replace_args *p;
3629 if (!capable(CAP_SYS_ADMIN))
3632 p = memdup_user(arg, sizeof(*p));
3637 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
3639 &root->fs_info->mutually_exclusive_operation_running,
3641 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3644 ret = btrfs_dev_replace_start(root, p);
3646 &root->fs_info->mutually_exclusive_operation_running,
3650 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3651 btrfs_dev_replace_status(root->fs_info, p);
3654 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3655 ret = btrfs_dev_replace_cancel(root->fs_info, p);
3662 if (copy_to_user(arg, p, sizeof(*p)))
3669 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3675 struct btrfs_ioctl_ino_path_args *ipa = NULL;
3676 struct inode_fs_paths *ipath = NULL;
3677 struct btrfs_path *path;
3679 if (!capable(CAP_DAC_READ_SEARCH))
3682 path = btrfs_alloc_path();
3688 ipa = memdup_user(arg, sizeof(*ipa));
3695 size = min_t(u32, ipa->size, 4096);
3696 ipath = init_ipath(size, root, path);
3697 if (IS_ERR(ipath)) {
3698 ret = PTR_ERR(ipath);
3703 ret = paths_from_inode(ipa->inum, ipath);
3707 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3708 rel_ptr = ipath->fspath->val[i] -
3709 (u64)(unsigned long)ipath->fspath->val;
3710 ipath->fspath->val[i] = rel_ptr;
3713 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3714 (void *)(unsigned long)ipath->fspath, size);
3721 btrfs_free_path(path);
3728 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3730 struct btrfs_data_container *inodes = ctx;
3731 const size_t c = 3 * sizeof(u64);
3733 if (inodes->bytes_left >= c) {
3734 inodes->bytes_left -= c;
3735 inodes->val[inodes->elem_cnt] = inum;
3736 inodes->val[inodes->elem_cnt + 1] = offset;
3737 inodes->val[inodes->elem_cnt + 2] = root;
3738 inodes->elem_cnt += 3;
3740 inodes->bytes_missing += c - inodes->bytes_left;
3741 inodes->bytes_left = 0;
3742 inodes->elem_missed += 3;
3748 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3753 struct btrfs_ioctl_logical_ino_args *loi;
3754 struct btrfs_data_container *inodes = NULL;
3755 struct btrfs_path *path = NULL;
3757 if (!capable(CAP_SYS_ADMIN))
3760 loi = memdup_user(arg, sizeof(*loi));
3767 path = btrfs_alloc_path();
3773 size = min_t(u32, loi->size, 64 * 1024);
3774 inodes = init_data_container(size);
3775 if (IS_ERR(inodes)) {
3776 ret = PTR_ERR(inodes);
3781 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
3782 build_ino_list, inodes);
3788 ret = copy_to_user((void *)(unsigned long)loi->inodes,
3789 (void *)(unsigned long)inodes, size);
3794 btrfs_free_path(path);
3801 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
3802 struct btrfs_ioctl_balance_args *bargs)
3804 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3806 bargs->flags = bctl->flags;
3808 if (atomic_read(&fs_info->balance_running))
3809 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3810 if (atomic_read(&fs_info->balance_pause_req))
3811 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3812 if (atomic_read(&fs_info->balance_cancel_req))
3813 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3815 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3816 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3817 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3820 spin_lock(&fs_info->balance_lock);
3821 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3822 spin_unlock(&fs_info->balance_lock);
3824 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3828 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3830 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
3831 struct btrfs_fs_info *fs_info = root->fs_info;
3832 struct btrfs_ioctl_balance_args *bargs;
3833 struct btrfs_balance_control *bctl;
3834 bool need_unlock; /* for mut. excl. ops lock */
3837 if (!capable(CAP_SYS_ADMIN))
3840 ret = mnt_want_write_file(file);
3845 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
3846 mutex_lock(&fs_info->volume_mutex);
3847 mutex_lock(&fs_info->balance_mutex);
3853 * mut. excl. ops lock is locked. Three possibilites:
3854 * (1) some other op is running
3855 * (2) balance is running
3856 * (3) balance is paused -- special case (think resume)
3858 mutex_lock(&fs_info->balance_mutex);
3859 if (fs_info->balance_ctl) {
3860 /* this is either (2) or (3) */
3861 if (!atomic_read(&fs_info->balance_running)) {
3862 mutex_unlock(&fs_info->balance_mutex);
3863 if (!mutex_trylock(&fs_info->volume_mutex))
3865 mutex_lock(&fs_info->balance_mutex);
3867 if (fs_info->balance_ctl &&
3868 !atomic_read(&fs_info->balance_running)) {
3870 need_unlock = false;
3874 mutex_unlock(&fs_info->balance_mutex);
3875 mutex_unlock(&fs_info->volume_mutex);
3879 mutex_unlock(&fs_info->balance_mutex);
3885 mutex_unlock(&fs_info->balance_mutex);
3886 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3892 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
3895 bargs = memdup_user(arg, sizeof(*bargs));
3896 if (IS_ERR(bargs)) {
3897 ret = PTR_ERR(bargs);
3901 if (bargs->flags & BTRFS_BALANCE_RESUME) {
3902 if (!fs_info->balance_ctl) {
3907 bctl = fs_info->balance_ctl;
3908 spin_lock(&fs_info->balance_lock);
3909 bctl->flags |= BTRFS_BALANCE_RESUME;
3910 spin_unlock(&fs_info->balance_lock);
3918 if (fs_info->balance_ctl) {
3923 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3929 bctl->fs_info = fs_info;
3931 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3932 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3933 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3935 bctl->flags = bargs->flags;
3937 /* balance everything - no filters */
3938 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
3943 * Ownership of bctl and mutually_exclusive_operation_running
3944 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
3945 * or, if restriper was paused all the way until unmount, in
3946 * free_fs_info. mutually_exclusive_operation_running is
3947 * cleared in __cancel_balance.
3949 need_unlock = false;
3951 ret = btrfs_balance(bctl, bargs);
3954 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3961 mutex_unlock(&fs_info->balance_mutex);
3962 mutex_unlock(&fs_info->volume_mutex);
3964 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
3966 mnt_drop_write_file(file);
3970 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
3972 if (!capable(CAP_SYS_ADMIN))
3976 case BTRFS_BALANCE_CTL_PAUSE:
3977 return btrfs_pause_balance(root->fs_info);
3978 case BTRFS_BALANCE_CTL_CANCEL:
3979 return btrfs_cancel_balance(root->fs_info);
3985 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
3988 struct btrfs_fs_info *fs_info = root->fs_info;
3989 struct btrfs_ioctl_balance_args *bargs;
3992 if (!capable(CAP_SYS_ADMIN))
3995 mutex_lock(&fs_info->balance_mutex);
3996 if (!fs_info->balance_ctl) {
4001 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
4007 update_ioctl_balance_args(fs_info, 1, bargs);
4009 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4014 mutex_unlock(&fs_info->balance_mutex);
4018 static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
4020 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4021 struct btrfs_ioctl_quota_ctl_args *sa;
4022 struct btrfs_trans_handle *trans = NULL;
4026 if (!capable(CAP_SYS_ADMIN))
4029 ret = mnt_want_write_file(file);
4033 sa = memdup_user(arg, sizeof(*sa));
4039 down_write(&root->fs_info->subvol_sem);
4040 trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4041 if (IS_ERR(trans)) {
4042 ret = PTR_ERR(trans);
4047 case BTRFS_QUOTA_CTL_ENABLE:
4048 ret = btrfs_quota_enable(trans, root->fs_info);
4050 case BTRFS_QUOTA_CTL_DISABLE:
4051 ret = btrfs_quota_disable(trans, root->fs_info);
4058 err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4063 up_write(&root->fs_info->subvol_sem);
4065 mnt_drop_write_file(file);
4069 static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
4071 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4072 struct btrfs_ioctl_qgroup_assign_args *sa;
4073 struct btrfs_trans_handle *trans;
4077 if (!capable(CAP_SYS_ADMIN))
4080 ret = mnt_want_write_file(file);
4084 sa = memdup_user(arg, sizeof(*sa));
4090 trans = btrfs_join_transaction(root);
4091 if (IS_ERR(trans)) {
4092 ret = PTR_ERR(trans);
4096 /* FIXME: check if the IDs really exist */
4098 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4101 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4105 err = btrfs_end_transaction(trans, root);
4112 mnt_drop_write_file(file);
4116 static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
4118 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4119 struct btrfs_ioctl_qgroup_create_args *sa;
4120 struct btrfs_trans_handle *trans;
4124 if (!capable(CAP_SYS_ADMIN))
4127 ret = mnt_want_write_file(file);
4131 sa = memdup_user(arg, sizeof(*sa));
4137 if (!sa->qgroupid) {
4142 trans = btrfs_join_transaction(root);
4143 if (IS_ERR(trans)) {
4144 ret = PTR_ERR(trans);
4148 /* FIXME: check if the IDs really exist */
4150 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
4153 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4156 err = btrfs_end_transaction(trans, root);
4163 mnt_drop_write_file(file);
4167 static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
4169 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4170 struct btrfs_ioctl_qgroup_limit_args *sa;
4171 struct btrfs_trans_handle *trans;
4176 if (!capable(CAP_SYS_ADMIN))
4179 ret = mnt_want_write_file(file);
4183 sa = memdup_user(arg, sizeof(*sa));
4189 trans = btrfs_join_transaction(root);
4190 if (IS_ERR(trans)) {
4191 ret = PTR_ERR(trans);
4195 qgroupid = sa->qgroupid;
4197 /* take the current subvol as qgroup */
4198 qgroupid = root->root_key.objectid;
4201 /* FIXME: check if the IDs really exist */
4202 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
4204 err = btrfs_end_transaction(trans, root);
4211 mnt_drop_write_file(file);
4215 static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4217 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4218 struct btrfs_ioctl_quota_rescan_args *qsa;
4221 if (!capable(CAP_SYS_ADMIN))
4224 ret = mnt_want_write_file(file);
4228 qsa = memdup_user(arg, sizeof(*qsa));
4239 ret = btrfs_qgroup_rescan(root->fs_info);
4244 mnt_drop_write_file(file);
4248 static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
4250 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4251 struct btrfs_ioctl_quota_rescan_args *qsa;
4254 if (!capable(CAP_SYS_ADMIN))
4257 qsa = kzalloc(sizeof(*qsa), GFP_NOFS);
4261 if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4263 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
4266 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4273 static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
4275 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
4277 if (!capable(CAP_SYS_ADMIN))
4280 return btrfs_qgroup_wait_for_completion(root->fs_info);
4283 static long btrfs_ioctl_set_received_subvol(struct file *file,
4286 struct btrfs_ioctl_received_subvol_args *sa = NULL;
4287 struct inode *inode = file_inode(file);
4288 struct btrfs_root *root = BTRFS_I(inode)->root;
4289 struct btrfs_root_item *root_item = &root->root_item;
4290 struct btrfs_trans_handle *trans;
4291 struct timespec ct = CURRENT_TIME;
4294 ret = mnt_want_write_file(file);
4298 down_write(&root->fs_info->subvol_sem);
4300 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
4305 if (btrfs_root_readonly(root)) {
4310 if (!inode_owner_or_capable(inode)) {
4315 sa = memdup_user(arg, sizeof(*sa));
4322 trans = btrfs_start_transaction(root, 1);
4323 if (IS_ERR(trans)) {
4324 ret = PTR_ERR(trans);
4329 sa->rtransid = trans->transid;
4330 sa->rtime.sec = ct.tv_sec;
4331 sa->rtime.nsec = ct.tv_nsec;
4333 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4334 btrfs_set_root_stransid(root_item, sa->stransid);
4335 btrfs_set_root_rtransid(root_item, sa->rtransid);
4336 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4337 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4338 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4339 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
4341 ret = btrfs_update_root(trans, root->fs_info->tree_root,
4342 &root->root_key, &root->root_item);
4344 btrfs_end_transaction(trans, root);
4348 ret = btrfs_commit_transaction(trans, root);
4353 ret = copy_to_user(arg, sa, sizeof(*sa));
4359 up_write(&root->fs_info->subvol_sem);
4360 mnt_drop_write_file(file);
4364 static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
4366 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4369 char label[BTRFS_LABEL_SIZE];
4371 spin_lock(&root->fs_info->super_lock);
4372 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4373 spin_unlock(&root->fs_info->super_lock);
4375 len = strnlen(label, BTRFS_LABEL_SIZE);
4377 if (len == BTRFS_LABEL_SIZE) {
4378 pr_warn("btrfs: label is too long, return the first %zu bytes\n",
4382 ret = copy_to_user(arg, label, len);
4384 return ret ? -EFAULT : 0;
4387 static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4389 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4390 struct btrfs_super_block *super_block = root->fs_info->super_copy;
4391 struct btrfs_trans_handle *trans;
4392 char label[BTRFS_LABEL_SIZE];
4395 if (!capable(CAP_SYS_ADMIN))
4398 if (copy_from_user(label, arg, sizeof(label)))
4401 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
4402 pr_err("btrfs: unable to set label with more than %d bytes\n",
4403 BTRFS_LABEL_SIZE - 1);
4407 ret = mnt_want_write_file(file);
4411 trans = btrfs_start_transaction(root, 0);
4412 if (IS_ERR(trans)) {
4413 ret = PTR_ERR(trans);
4417 spin_lock(&root->fs_info->super_lock);
4418 strcpy(super_block->label, label);
4419 spin_unlock(&root->fs_info->super_lock);
4420 ret = btrfs_end_transaction(trans, root);
4423 mnt_drop_write_file(file);
4427 long btrfs_ioctl(struct file *file, unsigned int
4428 cmd, unsigned long arg)
4430 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4431 void __user *argp = (void __user *)arg;
4434 case FS_IOC_GETFLAGS:
4435 return btrfs_ioctl_getflags(file, argp);
4436 case FS_IOC_SETFLAGS:
4437 return btrfs_ioctl_setflags(file, argp);
4438 case FS_IOC_GETVERSION:
4439 return btrfs_ioctl_getversion(file, argp);
4441 return btrfs_ioctl_fitrim(file, argp);
4442 case BTRFS_IOC_SNAP_CREATE:
4443 return btrfs_ioctl_snap_create(file, argp, 0);
4444 case BTRFS_IOC_SNAP_CREATE_V2:
4445 return btrfs_ioctl_snap_create_v2(file, argp, 0);
4446 case BTRFS_IOC_SUBVOL_CREATE:
4447 return btrfs_ioctl_snap_create(file, argp, 1);
4448 case BTRFS_IOC_SUBVOL_CREATE_V2:
4449 return btrfs_ioctl_snap_create_v2(file, argp, 1);
4450 case BTRFS_IOC_SNAP_DESTROY:
4451 return btrfs_ioctl_snap_destroy(file, argp);
4452 case BTRFS_IOC_SUBVOL_GETFLAGS:
4453 return btrfs_ioctl_subvol_getflags(file, argp);
4454 case BTRFS_IOC_SUBVOL_SETFLAGS:
4455 return btrfs_ioctl_subvol_setflags(file, argp);
4456 case BTRFS_IOC_DEFAULT_SUBVOL:
4457 return btrfs_ioctl_default_subvol(file, argp);
4458 case BTRFS_IOC_DEFRAG:
4459 return btrfs_ioctl_defrag(file, NULL);
4460 case BTRFS_IOC_DEFRAG_RANGE:
4461 return btrfs_ioctl_defrag(file, argp);
4462 case BTRFS_IOC_RESIZE:
4463 return btrfs_ioctl_resize(file, argp);
4464 case BTRFS_IOC_ADD_DEV:
4465 return btrfs_ioctl_add_dev(root, argp);
4466 case BTRFS_IOC_RM_DEV:
4467 return btrfs_ioctl_rm_dev(file, argp);
4468 case BTRFS_IOC_FS_INFO:
4469 return btrfs_ioctl_fs_info(root, argp);
4470 case BTRFS_IOC_DEV_INFO:
4471 return btrfs_ioctl_dev_info(root, argp);
4472 case BTRFS_IOC_BALANCE:
4473 return btrfs_ioctl_balance(file, NULL);
4474 case BTRFS_IOC_CLONE:
4475 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
4476 case BTRFS_IOC_CLONE_RANGE:
4477 return btrfs_ioctl_clone_range(file, argp);
4478 case BTRFS_IOC_TRANS_START:
4479 return btrfs_ioctl_trans_start(file);
4480 case BTRFS_IOC_TRANS_END:
4481 return btrfs_ioctl_trans_end(file);
4482 case BTRFS_IOC_TREE_SEARCH:
4483 return btrfs_ioctl_tree_search(file, argp);
4484 case BTRFS_IOC_INO_LOOKUP:
4485 return btrfs_ioctl_ino_lookup(file, argp);
4486 case BTRFS_IOC_INO_PATHS:
4487 return btrfs_ioctl_ino_to_path(root, argp);
4488 case BTRFS_IOC_LOGICAL_INO:
4489 return btrfs_ioctl_logical_to_ino(root, argp);
4490 case BTRFS_IOC_SPACE_INFO:
4491 return btrfs_ioctl_space_info(root, argp);
4492 case BTRFS_IOC_SYNC:
4493 btrfs_sync_fs(file->f_dentry->d_sb, 1);
4495 case BTRFS_IOC_START_SYNC:
4496 return btrfs_ioctl_start_sync(root, argp);
4497 case BTRFS_IOC_WAIT_SYNC:
4498 return btrfs_ioctl_wait_sync(root, argp);
4499 case BTRFS_IOC_SCRUB:
4500 return btrfs_ioctl_scrub(file, argp);
4501 case BTRFS_IOC_SCRUB_CANCEL:
4502 return btrfs_ioctl_scrub_cancel(root, argp);
4503 case BTRFS_IOC_SCRUB_PROGRESS:
4504 return btrfs_ioctl_scrub_progress(root, argp);
4505 case BTRFS_IOC_BALANCE_V2:
4506 return btrfs_ioctl_balance(file, argp);
4507 case BTRFS_IOC_BALANCE_CTL:
4508 return btrfs_ioctl_balance_ctl(root, arg);
4509 case BTRFS_IOC_BALANCE_PROGRESS:
4510 return btrfs_ioctl_balance_progress(root, argp);
4511 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
4512 return btrfs_ioctl_set_received_subvol(file, argp);
4513 case BTRFS_IOC_SEND:
4514 return btrfs_ioctl_send(file, argp);
4515 case BTRFS_IOC_GET_DEV_STATS:
4516 return btrfs_ioctl_get_dev_stats(root, argp);
4517 case BTRFS_IOC_QUOTA_CTL:
4518 return btrfs_ioctl_quota_ctl(file, argp);
4519 case BTRFS_IOC_QGROUP_ASSIGN:
4520 return btrfs_ioctl_qgroup_assign(file, argp);
4521 case BTRFS_IOC_QGROUP_CREATE:
4522 return btrfs_ioctl_qgroup_create(file, argp);
4523 case BTRFS_IOC_QGROUP_LIMIT:
4524 return btrfs_ioctl_qgroup_limit(file, argp);
4525 case BTRFS_IOC_QUOTA_RESCAN:
4526 return btrfs_ioctl_quota_rescan(file, argp);
4527 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
4528 return btrfs_ioctl_quota_rescan_status(file, argp);
4529 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
4530 return btrfs_ioctl_quota_rescan_wait(file, argp);
4531 case BTRFS_IOC_DEV_REPLACE:
4532 return btrfs_ioctl_dev_replace(root, argp);
4533 case BTRFS_IOC_GET_FSLABEL:
4534 return btrfs_ioctl_get_fslabel(file, argp);
4535 case BTRFS_IOC_SET_FSLABEL:
4536 return btrfs_ioctl_set_fslabel(file, argp);
4537 case BTRFS_IOC_FILE_EXTENT_SAME:
4538 return btrfs_ioctl_file_extent_same(file, argp);