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>
47 #include "transaction.h"
48 #include "btrfs_inode.h"
50 #include "print-tree.h"
53 #include "inode-map.h"
55 #include "rcu-string.h"
57 /* Mask out flags that are inappropriate for the given type of inode. */
58 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
62 else if (S_ISREG(mode))
63 return flags & ~FS_DIRSYNC_FL;
65 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
69 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
71 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
73 unsigned int iflags = 0;
75 if (flags & BTRFS_INODE_SYNC)
77 if (flags & BTRFS_INODE_IMMUTABLE)
78 iflags |= FS_IMMUTABLE_FL;
79 if (flags & BTRFS_INODE_APPEND)
80 iflags |= FS_APPEND_FL;
81 if (flags & BTRFS_INODE_NODUMP)
82 iflags |= FS_NODUMP_FL;
83 if (flags & BTRFS_INODE_NOATIME)
84 iflags |= FS_NOATIME_FL;
85 if (flags & BTRFS_INODE_DIRSYNC)
86 iflags |= FS_DIRSYNC_FL;
87 if (flags & BTRFS_INODE_NODATACOW)
88 iflags |= FS_NOCOW_FL;
90 if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
91 iflags |= FS_COMPR_FL;
92 else if (flags & BTRFS_INODE_NOCOMPRESS)
93 iflags |= FS_NOCOMP_FL;
99 * Update inode->i_flags based on the btrfs internal flags.
101 void btrfs_update_iflags(struct inode *inode)
103 struct btrfs_inode *ip = BTRFS_I(inode);
105 inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
107 if (ip->flags & BTRFS_INODE_SYNC)
108 inode->i_flags |= S_SYNC;
109 if (ip->flags & BTRFS_INODE_IMMUTABLE)
110 inode->i_flags |= S_IMMUTABLE;
111 if (ip->flags & BTRFS_INODE_APPEND)
112 inode->i_flags |= S_APPEND;
113 if (ip->flags & BTRFS_INODE_NOATIME)
114 inode->i_flags |= S_NOATIME;
115 if (ip->flags & BTRFS_INODE_DIRSYNC)
116 inode->i_flags |= S_DIRSYNC;
120 * Inherit flags from the parent inode.
122 * Currently only the compression flags and the cow flags are inherited.
124 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
131 flags = BTRFS_I(dir)->flags;
133 if (flags & BTRFS_INODE_NOCOMPRESS) {
134 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
135 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
136 } else if (flags & BTRFS_INODE_COMPRESS) {
137 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
138 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
141 if (flags & BTRFS_INODE_NODATACOW)
142 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
144 btrfs_update_iflags(inode);
147 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
149 struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode);
150 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
152 if (copy_to_user(arg, &flags, sizeof(flags)))
157 static int check_flags(unsigned int flags)
159 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
160 FS_NOATIME_FL | FS_NODUMP_FL | \
161 FS_SYNC_FL | FS_DIRSYNC_FL | \
162 FS_NOCOMP_FL | FS_COMPR_FL |
166 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
172 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
174 struct inode *inode = file->f_path.dentry->d_inode;
175 struct btrfs_inode *ip = BTRFS_I(inode);
176 struct btrfs_root *root = ip->root;
177 struct btrfs_trans_handle *trans;
178 unsigned int flags, oldflags;
181 unsigned int i_oldflags;
183 if (btrfs_root_readonly(root))
186 if (copy_from_user(&flags, arg, sizeof(flags)))
189 ret = check_flags(flags);
193 if (!inode_owner_or_capable(inode))
196 mutex_lock(&inode->i_mutex);
198 ip_oldflags = ip->flags;
199 i_oldflags = inode->i_flags;
201 flags = btrfs_mask_flags(inode->i_mode, flags);
202 oldflags = btrfs_flags_to_ioctl(ip->flags);
203 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
204 if (!capable(CAP_LINUX_IMMUTABLE)) {
210 ret = mnt_want_write_file(file);
214 if (flags & FS_SYNC_FL)
215 ip->flags |= BTRFS_INODE_SYNC;
217 ip->flags &= ~BTRFS_INODE_SYNC;
218 if (flags & FS_IMMUTABLE_FL)
219 ip->flags |= BTRFS_INODE_IMMUTABLE;
221 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
222 if (flags & FS_APPEND_FL)
223 ip->flags |= BTRFS_INODE_APPEND;
225 ip->flags &= ~BTRFS_INODE_APPEND;
226 if (flags & FS_NODUMP_FL)
227 ip->flags |= BTRFS_INODE_NODUMP;
229 ip->flags &= ~BTRFS_INODE_NODUMP;
230 if (flags & FS_NOATIME_FL)
231 ip->flags |= BTRFS_INODE_NOATIME;
233 ip->flags &= ~BTRFS_INODE_NOATIME;
234 if (flags & FS_DIRSYNC_FL)
235 ip->flags |= BTRFS_INODE_DIRSYNC;
237 ip->flags &= ~BTRFS_INODE_DIRSYNC;
238 if (flags & FS_NOCOW_FL)
239 ip->flags |= BTRFS_INODE_NODATACOW;
241 ip->flags &= ~BTRFS_INODE_NODATACOW;
244 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
245 * flag may be changed automatically if compression code won't make
248 if (flags & FS_NOCOMP_FL) {
249 ip->flags &= ~BTRFS_INODE_COMPRESS;
250 ip->flags |= BTRFS_INODE_NOCOMPRESS;
251 } else if (flags & FS_COMPR_FL) {
252 ip->flags |= BTRFS_INODE_COMPRESS;
253 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
255 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
258 trans = btrfs_start_transaction(root, 1);
260 ret = PTR_ERR(trans);
264 btrfs_update_iflags(inode);
265 inode_inc_iversion(inode);
266 inode->i_ctime = CURRENT_TIME;
267 ret = btrfs_update_inode(trans, root, inode);
269 btrfs_end_transaction(trans, root);
272 ip->flags = ip_oldflags;
273 inode->i_flags = i_oldflags;
276 mnt_drop_write_file(file);
278 mutex_unlock(&inode->i_mutex);
282 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
284 struct inode *inode = file->f_path.dentry->d_inode;
286 return put_user(inode->i_generation, arg);
289 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
291 struct btrfs_fs_info *fs_info = btrfs_sb(fdentry(file)->d_sb);
292 struct btrfs_device *device;
293 struct request_queue *q;
294 struct fstrim_range range;
295 u64 minlen = ULLONG_MAX;
297 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
300 if (!capable(CAP_SYS_ADMIN))
304 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
308 q = bdev_get_queue(device->bdev);
309 if (blk_queue_discard(q)) {
311 minlen = min((u64)q->limits.discard_granularity,
319 if (copy_from_user(&range, arg, sizeof(range)))
321 if (range.start > total_bytes)
324 range.len = min(range.len, total_bytes - range.start);
325 range.minlen = max(range.minlen, minlen);
326 ret = btrfs_trim_fs(fs_info->tree_root, &range);
330 if (copy_to_user(arg, &range, sizeof(range)))
336 static noinline int create_subvol(struct btrfs_root *root,
337 struct dentry *dentry,
338 char *name, int namelen,
341 struct btrfs_trans_handle *trans;
342 struct btrfs_key key;
343 struct btrfs_root_item root_item;
344 struct btrfs_inode_item *inode_item;
345 struct extent_buffer *leaf;
346 struct btrfs_root *new_root;
347 struct dentry *parent = dentry->d_parent;
352 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
355 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
359 dir = parent->d_inode;
367 trans = btrfs_start_transaction(root, 6);
369 return PTR_ERR(trans);
371 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
372 0, objectid, NULL, 0, 0, 0);
378 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
379 btrfs_set_header_bytenr(leaf, leaf->start);
380 btrfs_set_header_generation(leaf, trans->transid);
381 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
382 btrfs_set_header_owner(leaf, objectid);
384 write_extent_buffer(leaf, root->fs_info->fsid,
385 (unsigned long)btrfs_header_fsid(leaf),
387 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
388 (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
390 btrfs_mark_buffer_dirty(leaf);
392 inode_item = &root_item.inode;
393 memset(inode_item, 0, sizeof(*inode_item));
394 inode_item->generation = cpu_to_le64(1);
395 inode_item->size = cpu_to_le64(3);
396 inode_item->nlink = cpu_to_le32(1);
397 inode_item->nbytes = cpu_to_le64(root->leafsize);
398 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
401 root_item.byte_limit = 0;
402 inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
404 btrfs_set_root_bytenr(&root_item, leaf->start);
405 btrfs_set_root_generation(&root_item, trans->transid);
406 btrfs_set_root_level(&root_item, 0);
407 btrfs_set_root_refs(&root_item, 1);
408 btrfs_set_root_used(&root_item, leaf->len);
409 btrfs_set_root_last_snapshot(&root_item, 0);
411 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
412 root_item.drop_level = 0;
414 btrfs_tree_unlock(leaf);
415 free_extent_buffer(leaf);
418 btrfs_set_root_dirid(&root_item, new_dirid);
420 key.objectid = objectid;
422 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
423 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
428 key.offset = (u64)-1;
429 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
430 if (IS_ERR(new_root)) {
431 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
432 ret = PTR_ERR(new_root);
436 btrfs_record_root_in_trans(trans, new_root);
438 ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
440 /* We potentially lose an unused inode item here */
441 btrfs_abort_transaction(trans, root, ret);
446 * insert the directory item
448 ret = btrfs_set_inode_index(dir, &index);
450 btrfs_abort_transaction(trans, root, ret);
454 ret = btrfs_insert_dir_item(trans, root,
455 name, namelen, dir, &key,
456 BTRFS_FT_DIR, index);
458 btrfs_abort_transaction(trans, root, ret);
462 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
463 ret = btrfs_update_inode(trans, root, dir);
466 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
467 objectid, root->root_key.objectid,
468 btrfs_ino(dir), index, name, namelen);
472 d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
475 *async_transid = trans->transid;
476 err = btrfs_commit_transaction_async(trans, root, 1);
478 err = btrfs_commit_transaction(trans, root);
485 static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
486 char *name, int namelen, u64 *async_transid,
490 struct btrfs_pending_snapshot *pending_snapshot;
491 struct btrfs_trans_handle *trans;
497 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
498 if (!pending_snapshot)
501 btrfs_init_block_rsv(&pending_snapshot->block_rsv);
502 pending_snapshot->dentry = dentry;
503 pending_snapshot->root = root;
504 pending_snapshot->readonly = readonly;
506 trans = btrfs_start_transaction(root->fs_info->extent_root, 5);
508 ret = PTR_ERR(trans);
512 ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
515 spin_lock(&root->fs_info->trans_lock);
516 list_add(&pending_snapshot->list,
517 &trans->transaction->pending_snapshots);
518 spin_unlock(&root->fs_info->trans_lock);
520 *async_transid = trans->transid;
521 ret = btrfs_commit_transaction_async(trans,
522 root->fs_info->extent_root, 1);
524 ret = btrfs_commit_transaction(trans,
525 root->fs_info->extent_root);
529 ret = pending_snapshot->error;
533 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
537 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
539 ret = PTR_ERR(inode);
543 d_instantiate(dentry, inode);
546 kfree(pending_snapshot);
550 /* copy of check_sticky in fs/namei.c()
551 * It's inline, so penalty for filesystems that don't use sticky bit is
554 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
556 uid_t fsuid = current_fsuid();
558 if (!(dir->i_mode & S_ISVTX))
560 if (inode->i_uid == fsuid)
562 if (dir->i_uid == fsuid)
564 return !capable(CAP_FOWNER);
567 /* copy of may_delete in fs/namei.c()
568 * Check whether we can remove a link victim from directory dir, check
569 * whether the type of victim is right.
570 * 1. We can't do it if dir is read-only (done in permission())
571 * 2. We should have write and exec permissions on dir
572 * 3. We can't remove anything from append-only dir
573 * 4. We can't do anything with immutable dir (done in permission())
574 * 5. If the sticky bit on dir is set we should either
575 * a. be owner of dir, or
576 * b. be owner of victim, or
577 * c. have CAP_FOWNER capability
578 * 6. If the victim is append-only or immutable we can't do antyhing with
579 * links pointing to it.
580 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
581 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
582 * 9. We can't remove a root or mountpoint.
583 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
584 * nfs_async_unlink().
587 static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
591 if (!victim->d_inode)
594 BUG_ON(victim->d_parent->d_inode != dir);
595 audit_inode_child(victim, dir);
597 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
602 if (btrfs_check_sticky(dir, victim->d_inode)||
603 IS_APPEND(victim->d_inode)||
604 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
607 if (!S_ISDIR(victim->d_inode->i_mode))
611 } else if (S_ISDIR(victim->d_inode->i_mode))
615 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
620 /* copy of may_create in fs/namei.c() */
621 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
627 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
631 * Create a new subvolume below @parent. This is largely modeled after
632 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
633 * inside this filesystem so it's quite a bit simpler.
635 static noinline int btrfs_mksubvol(struct path *parent,
636 char *name, int namelen,
637 struct btrfs_root *snap_src,
638 u64 *async_transid, bool readonly)
640 struct inode *dir = parent->dentry->d_inode;
641 struct dentry *dentry;
644 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
646 dentry = lookup_one_len(name, parent->dentry, namelen);
647 error = PTR_ERR(dentry);
655 error = mnt_want_write(parent->mnt);
659 error = btrfs_may_create(dir, dentry);
663 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
665 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
669 error = create_snapshot(snap_src, dentry,
670 name, namelen, async_transid, readonly);
672 error = create_subvol(BTRFS_I(dir)->root, dentry,
673 name, namelen, async_transid);
676 fsnotify_mkdir(dir, dentry);
678 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
680 mnt_drop_write(parent->mnt);
684 mutex_unlock(&dir->i_mutex);
689 * When we're defragging a range, we don't want to kick it off again
690 * if it is really just waiting for delalloc to send it down.
691 * If we find a nice big extent or delalloc range for the bytes in the
692 * file you want to defrag, we return 0 to let you know to skip this
695 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
697 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
698 struct extent_map *em = NULL;
699 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
702 read_lock(&em_tree->lock);
703 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
704 read_unlock(&em_tree->lock);
707 end = extent_map_end(em);
709 if (end - offset > thresh)
712 /* if we already have a nice delalloc here, just stop */
714 end = count_range_bits(io_tree, &offset, offset + thresh,
715 thresh, EXTENT_DELALLOC, 1);
722 * helper function to walk through a file and find extents
723 * newer than a specific transid, and smaller than thresh.
725 * This is used by the defragging code to find new and small
728 static int find_new_extents(struct btrfs_root *root,
729 struct inode *inode, u64 newer_than,
730 u64 *off, int thresh)
732 struct btrfs_path *path;
733 struct btrfs_key min_key;
734 struct btrfs_key max_key;
735 struct extent_buffer *leaf;
736 struct btrfs_file_extent_item *extent;
739 u64 ino = btrfs_ino(inode);
741 path = btrfs_alloc_path();
745 min_key.objectid = ino;
746 min_key.type = BTRFS_EXTENT_DATA_KEY;
747 min_key.offset = *off;
749 max_key.objectid = ino;
750 max_key.type = (u8)-1;
751 max_key.offset = (u64)-1;
753 path->keep_locks = 1;
756 ret = btrfs_search_forward(root, &min_key, &max_key,
757 path, 0, newer_than);
760 if (min_key.objectid != ino)
762 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
765 leaf = path->nodes[0];
766 extent = btrfs_item_ptr(leaf, path->slots[0],
767 struct btrfs_file_extent_item);
769 type = btrfs_file_extent_type(leaf, extent);
770 if (type == BTRFS_FILE_EXTENT_REG &&
771 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
772 check_defrag_in_cache(inode, min_key.offset, thresh)) {
773 *off = min_key.offset;
774 btrfs_free_path(path);
778 if (min_key.offset == (u64)-1)
782 btrfs_release_path(path);
785 btrfs_free_path(path);
789 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
791 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
792 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
793 struct extent_map *em;
794 u64 len = PAGE_CACHE_SIZE;
797 * hopefully we have this extent in the tree already, try without
798 * the full extent lock
800 read_lock(&em_tree->lock);
801 em = lookup_extent_mapping(em_tree, start, len);
802 read_unlock(&em_tree->lock);
805 /* get the big lock and read metadata off disk */
806 lock_extent(io_tree, start, start + len - 1);
807 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
808 unlock_extent(io_tree, start, start + len - 1);
817 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
819 struct extent_map *next;
822 /* this is the last extent */
823 if (em->start + em->len >= i_size_read(inode))
826 next = defrag_lookup_extent(inode, em->start + em->len);
827 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
830 free_extent_map(next);
834 static int should_defrag_range(struct inode *inode, u64 start, int thresh,
835 u64 *last_len, u64 *skip, u64 *defrag_end)
837 struct extent_map *em;
839 bool next_mergeable = true;
842 * make sure that once we start defragging an extent, we keep on
845 if (start < *defrag_end)
850 em = defrag_lookup_extent(inode, start);
854 /* this will cover holes, and inline extents */
855 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
860 next_mergeable = defrag_check_next_extent(inode, em);
863 * we hit a real extent, if it is big or the next extent is not a
864 * real extent, don't bother defragging it
866 if ((*last_len == 0 || *last_len >= thresh) &&
867 (em->len >= thresh || !next_mergeable))
871 * last_len ends up being a counter of how many bytes we've defragged.
872 * every time we choose not to defrag an extent, we reset *last_len
873 * so that the next tiny extent will force a defrag.
875 * The end result of this is that tiny extents before a single big
876 * extent will force at least part of that big extent to be defragged.
879 *defrag_end = extent_map_end(em);
882 *skip = extent_map_end(em);
891 * it doesn't do much good to defrag one or two pages
892 * at a time. This pulls in a nice chunk of pages
895 * It also makes sure the delalloc code has enough
896 * dirty data to avoid making new small extents as part
899 * It's a good idea to start RA on this range
900 * before calling this.
902 static int cluster_pages_for_defrag(struct inode *inode,
904 unsigned long start_index,
907 unsigned long file_end;
908 u64 isize = i_size_read(inode);
915 struct btrfs_ordered_extent *ordered;
916 struct extent_state *cached_state = NULL;
917 struct extent_io_tree *tree;
918 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
920 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
921 if (!isize || start_index > file_end)
924 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
926 ret = btrfs_delalloc_reserve_space(inode,
927 page_cnt << PAGE_CACHE_SHIFT);
931 tree = &BTRFS_I(inode)->io_tree;
933 /* step one, lock all the pages */
934 for (i = 0; i < page_cnt; i++) {
937 page = find_or_create_page(inode->i_mapping,
938 start_index + i, mask);
942 page_start = page_offset(page);
943 page_end = page_start + PAGE_CACHE_SIZE - 1;
945 lock_extent(tree, page_start, page_end);
946 ordered = btrfs_lookup_ordered_extent(inode,
948 unlock_extent(tree, page_start, page_end);
953 btrfs_start_ordered_extent(inode, ordered, 1);
954 btrfs_put_ordered_extent(ordered);
957 * we unlocked the page above, so we need check if
958 * it was released or not.
960 if (page->mapping != inode->i_mapping) {
962 page_cache_release(page);
967 if (!PageUptodate(page)) {
968 btrfs_readpage(NULL, page);
970 if (!PageUptodate(page)) {
972 page_cache_release(page);
978 if (page->mapping != inode->i_mapping) {
980 page_cache_release(page);
990 if (!(inode->i_sb->s_flags & MS_ACTIVE))
994 * so now we have a nice long stream of locked
995 * and up to date pages, lets wait on them
997 for (i = 0; i < i_done; i++)
998 wait_on_page_writeback(pages[i]);
1000 page_start = page_offset(pages[0]);
1001 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1003 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1004 page_start, page_end - 1, 0, &cached_state);
1005 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1006 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1007 EXTENT_DO_ACCOUNTING, 0, 0, &cached_state,
1010 if (i_done != page_cnt) {
1011 spin_lock(&BTRFS_I(inode)->lock);
1012 BTRFS_I(inode)->outstanding_extents++;
1013 spin_unlock(&BTRFS_I(inode)->lock);
1014 btrfs_delalloc_release_space(inode,
1015 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1019 btrfs_set_extent_delalloc(inode, page_start, page_end - 1,
1022 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1023 page_start, page_end - 1, &cached_state,
1026 for (i = 0; i < i_done; i++) {
1027 clear_page_dirty_for_io(pages[i]);
1028 ClearPageChecked(pages[i]);
1029 set_page_extent_mapped(pages[i]);
1030 set_page_dirty(pages[i]);
1031 unlock_page(pages[i]);
1032 page_cache_release(pages[i]);
1036 for (i = 0; i < i_done; i++) {
1037 unlock_page(pages[i]);
1038 page_cache_release(pages[i]);
1040 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1045 int btrfs_defrag_file(struct inode *inode, struct file *file,
1046 struct btrfs_ioctl_defrag_range_args *range,
1047 u64 newer_than, unsigned long max_to_defrag)
1049 struct btrfs_root *root = BTRFS_I(inode)->root;
1050 struct btrfs_super_block *disk_super;
1051 struct file_ra_state *ra = NULL;
1052 unsigned long last_index;
1053 u64 isize = i_size_read(inode);
1058 u64 newer_off = range->start;
1060 unsigned long ra_index = 0;
1062 int defrag_count = 0;
1063 int compress_type = BTRFS_COMPRESS_ZLIB;
1064 int extent_thresh = range->extent_thresh;
1065 int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1066 int cluster = max_cluster;
1067 u64 new_align = ~((u64)128 * 1024 - 1);
1068 struct page **pages = NULL;
1070 if (extent_thresh == 0)
1071 extent_thresh = 256 * 1024;
1073 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1074 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1076 if (range->compress_type)
1077 compress_type = range->compress_type;
1084 * if we were not given a file, allocate a readahead
1088 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1091 file_ra_state_init(ra, inode->i_mapping);
1096 pages = kmalloc(sizeof(struct page *) * max_cluster,
1103 /* find the last page to defrag */
1104 if (range->start + range->len > range->start) {
1105 last_index = min_t(u64, isize - 1,
1106 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1108 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1112 ret = find_new_extents(root, inode, newer_than,
1113 &newer_off, 64 * 1024);
1115 range->start = newer_off;
1117 * we always align our defrag to help keep
1118 * the extents in the file evenly spaced
1120 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1124 i = range->start >> PAGE_CACHE_SHIFT;
1127 max_to_defrag = last_index + 1;
1130 * make writeback starts from i, so the defrag range can be
1131 * written sequentially.
1133 if (i < inode->i_mapping->writeback_index)
1134 inode->i_mapping->writeback_index = i;
1136 while (i <= last_index && defrag_count < max_to_defrag &&
1137 (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1138 PAGE_CACHE_SHIFT)) {
1140 * make sure we stop running if someone unmounts
1143 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1146 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1147 extent_thresh, &last_len, &skip,
1151 * the should_defrag function tells us how much to skip
1152 * bump our counter by the suggested amount
1154 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1155 i = max(i + 1, next);
1160 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1161 PAGE_CACHE_SHIFT) - i;
1162 cluster = min(cluster, max_cluster);
1164 cluster = max_cluster;
1167 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1168 BTRFS_I(inode)->force_compress = compress_type;
1170 if (i + cluster > ra_index) {
1171 ra_index = max(i, ra_index);
1172 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1174 ra_index += max_cluster;
1177 mutex_lock(&inode->i_mutex);
1178 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1180 mutex_unlock(&inode->i_mutex);
1184 defrag_count += ret;
1185 balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret);
1186 mutex_unlock(&inode->i_mutex);
1189 if (newer_off == (u64)-1)
1195 newer_off = max(newer_off + 1,
1196 (u64)i << PAGE_CACHE_SHIFT);
1198 ret = find_new_extents(root, inode,
1199 newer_than, &newer_off,
1202 range->start = newer_off;
1203 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1210 last_len += ret << PAGE_CACHE_SHIFT;
1218 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1219 filemap_flush(inode->i_mapping);
1221 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1222 /* the filemap_flush will queue IO into the worker threads, but
1223 * we have to make sure the IO is actually started and that
1224 * ordered extents get created before we return
1226 atomic_inc(&root->fs_info->async_submit_draining);
1227 while (atomic_read(&root->fs_info->nr_async_submits) ||
1228 atomic_read(&root->fs_info->async_delalloc_pages)) {
1229 wait_event(root->fs_info->async_submit_wait,
1230 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1231 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1233 atomic_dec(&root->fs_info->async_submit_draining);
1235 mutex_lock(&inode->i_mutex);
1236 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1237 mutex_unlock(&inode->i_mutex);
1240 disk_super = root->fs_info->super_copy;
1241 features = btrfs_super_incompat_flags(disk_super);
1242 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1243 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
1244 btrfs_set_super_incompat_flags(disk_super, features);
1256 static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
1262 struct btrfs_ioctl_vol_args *vol_args;
1263 struct btrfs_trans_handle *trans;
1264 struct btrfs_device *device = NULL;
1266 char *devstr = NULL;
1270 if (root->fs_info->sb->s_flags & MS_RDONLY)
1273 if (!capable(CAP_SYS_ADMIN))
1276 mutex_lock(&root->fs_info->volume_mutex);
1277 if (root->fs_info->balance_ctl) {
1278 printk(KERN_INFO "btrfs: balance in progress\n");
1283 vol_args = memdup_user(arg, sizeof(*vol_args));
1284 if (IS_ERR(vol_args)) {
1285 ret = PTR_ERR(vol_args);
1289 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1291 sizestr = vol_args->name;
1292 devstr = strchr(sizestr, ':');
1295 sizestr = devstr + 1;
1297 devstr = vol_args->name;
1298 devid = simple_strtoull(devstr, &end, 10);
1299 printk(KERN_INFO "btrfs: resizing devid %llu\n",
1300 (unsigned long long)devid);
1302 device = btrfs_find_device(root, devid, NULL, NULL);
1304 printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
1305 (unsigned long long)devid);
1309 if (device->fs_devices && device->fs_devices->seeding) {
1310 printk(KERN_INFO "btrfs: resizer unable to apply on "
1311 "seeding device %llu\n",
1312 (unsigned long long)devid);
1317 if (!strcmp(sizestr, "max"))
1318 new_size = device->bdev->bd_inode->i_size;
1320 if (sizestr[0] == '-') {
1323 } else if (sizestr[0] == '+') {
1327 new_size = memparse(sizestr, NULL);
1328 if (new_size == 0) {
1334 old_size = device->total_bytes;
1337 if (new_size > old_size) {
1341 new_size = old_size - new_size;
1342 } else if (mod > 0) {
1343 new_size = old_size + new_size;
1346 if (new_size < 256 * 1024 * 1024) {
1350 if (new_size > device->bdev->bd_inode->i_size) {
1355 do_div(new_size, root->sectorsize);
1356 new_size *= root->sectorsize;
1358 printk_in_rcu(KERN_INFO "btrfs: new size for %s is %llu\n",
1359 rcu_str_deref(device->name),
1360 (unsigned long long)new_size);
1362 if (new_size > old_size) {
1363 trans = btrfs_start_transaction(root, 0);
1364 if (IS_ERR(trans)) {
1365 ret = PTR_ERR(trans);
1368 ret = btrfs_grow_device(trans, device, new_size);
1369 btrfs_commit_transaction(trans, root);
1370 } else if (new_size < old_size) {
1371 ret = btrfs_shrink_device(device, new_size);
1377 mutex_unlock(&root->fs_info->volume_mutex);
1381 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1388 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
1389 struct file *src_file;
1393 if (root->fs_info->sb->s_flags & MS_RDONLY)
1396 namelen = strlen(name);
1397 if (strchr(name, '/')) {
1402 if (name[0] == '.' &&
1403 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1409 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1410 NULL, transid, readonly);
1412 struct inode *src_inode;
1413 src_file = fget(fd);
1419 src_inode = src_file->f_path.dentry->d_inode;
1420 if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
1421 printk(KERN_INFO "btrfs: Snapshot src from "
1427 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1428 BTRFS_I(src_inode)->root,
1436 static noinline int btrfs_ioctl_snap_create(struct file *file,
1437 void __user *arg, int subvol)
1439 struct btrfs_ioctl_vol_args *vol_args;
1442 vol_args = memdup_user(arg, sizeof(*vol_args));
1443 if (IS_ERR(vol_args))
1444 return PTR_ERR(vol_args);
1445 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1447 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1448 vol_args->fd, subvol,
1455 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1456 void __user *arg, int subvol)
1458 struct btrfs_ioctl_vol_args_v2 *vol_args;
1462 bool readonly = false;
1464 vol_args = memdup_user(arg, sizeof(*vol_args));
1465 if (IS_ERR(vol_args))
1466 return PTR_ERR(vol_args);
1467 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1469 if (vol_args->flags &
1470 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY)) {
1475 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1477 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1480 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1481 vol_args->fd, subvol,
1484 if (ret == 0 && ptr &&
1486 offsetof(struct btrfs_ioctl_vol_args_v2,
1487 transid), ptr, sizeof(*ptr)))
1494 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1497 struct inode *inode = fdentry(file)->d_inode;
1498 struct btrfs_root *root = BTRFS_I(inode)->root;
1502 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1505 down_read(&root->fs_info->subvol_sem);
1506 if (btrfs_root_readonly(root))
1507 flags |= BTRFS_SUBVOL_RDONLY;
1508 up_read(&root->fs_info->subvol_sem);
1510 if (copy_to_user(arg, &flags, sizeof(flags)))
1516 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1519 struct inode *inode = fdentry(file)->d_inode;
1520 struct btrfs_root *root = BTRFS_I(inode)->root;
1521 struct btrfs_trans_handle *trans;
1526 if (root->fs_info->sb->s_flags & MS_RDONLY)
1529 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1532 if (copy_from_user(&flags, arg, sizeof(flags)))
1535 if (flags & BTRFS_SUBVOL_CREATE_ASYNC)
1538 if (flags & ~BTRFS_SUBVOL_RDONLY)
1541 if (!inode_owner_or_capable(inode))
1544 down_write(&root->fs_info->subvol_sem);
1547 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1550 root_flags = btrfs_root_flags(&root->root_item);
1551 if (flags & BTRFS_SUBVOL_RDONLY)
1552 btrfs_set_root_flags(&root->root_item,
1553 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1555 btrfs_set_root_flags(&root->root_item,
1556 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1558 trans = btrfs_start_transaction(root, 1);
1559 if (IS_ERR(trans)) {
1560 ret = PTR_ERR(trans);
1564 ret = btrfs_update_root(trans, root->fs_info->tree_root,
1565 &root->root_key, &root->root_item);
1567 btrfs_commit_transaction(trans, root);
1570 btrfs_set_root_flags(&root->root_item, root_flags);
1572 up_write(&root->fs_info->subvol_sem);
1577 * helper to check if the subvolume references other subvolumes
1579 static noinline int may_destroy_subvol(struct btrfs_root *root)
1581 struct btrfs_path *path;
1582 struct btrfs_key key;
1585 path = btrfs_alloc_path();
1589 key.objectid = root->root_key.objectid;
1590 key.type = BTRFS_ROOT_REF_KEY;
1591 key.offset = (u64)-1;
1593 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1600 if (path->slots[0] > 0) {
1602 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1603 if (key.objectid == root->root_key.objectid &&
1604 key.type == BTRFS_ROOT_REF_KEY)
1608 btrfs_free_path(path);
1612 static noinline int key_in_sk(struct btrfs_key *key,
1613 struct btrfs_ioctl_search_key *sk)
1615 struct btrfs_key test;
1618 test.objectid = sk->min_objectid;
1619 test.type = sk->min_type;
1620 test.offset = sk->min_offset;
1622 ret = btrfs_comp_cpu_keys(key, &test);
1626 test.objectid = sk->max_objectid;
1627 test.type = sk->max_type;
1628 test.offset = sk->max_offset;
1630 ret = btrfs_comp_cpu_keys(key, &test);
1636 static noinline int copy_to_sk(struct btrfs_root *root,
1637 struct btrfs_path *path,
1638 struct btrfs_key *key,
1639 struct btrfs_ioctl_search_key *sk,
1641 unsigned long *sk_offset,
1645 struct extent_buffer *leaf;
1646 struct btrfs_ioctl_search_header sh;
1647 unsigned long item_off;
1648 unsigned long item_len;
1654 leaf = path->nodes[0];
1655 slot = path->slots[0];
1656 nritems = btrfs_header_nritems(leaf);
1658 if (btrfs_header_generation(leaf) > sk->max_transid) {
1662 found_transid = btrfs_header_generation(leaf);
1664 for (i = slot; i < nritems; i++) {
1665 item_off = btrfs_item_ptr_offset(leaf, i);
1666 item_len = btrfs_item_size_nr(leaf, i);
1668 if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1671 if (sizeof(sh) + item_len + *sk_offset >
1672 BTRFS_SEARCH_ARGS_BUFSIZE) {
1677 btrfs_item_key_to_cpu(leaf, key, i);
1678 if (!key_in_sk(key, sk))
1681 sh.objectid = key->objectid;
1682 sh.offset = key->offset;
1683 sh.type = key->type;
1685 sh.transid = found_transid;
1687 /* copy search result header */
1688 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1689 *sk_offset += sizeof(sh);
1692 char *p = buf + *sk_offset;
1694 read_extent_buffer(leaf, p,
1695 item_off, item_len);
1696 *sk_offset += item_len;
1700 if (*num_found >= sk->nr_items)
1705 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1707 else if (key->type < (u8)-1 && key->type < sk->max_type) {
1710 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1720 static noinline int search_ioctl(struct inode *inode,
1721 struct btrfs_ioctl_search_args *args)
1723 struct btrfs_root *root;
1724 struct btrfs_key key;
1725 struct btrfs_key max_key;
1726 struct btrfs_path *path;
1727 struct btrfs_ioctl_search_key *sk = &args->key;
1728 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1731 unsigned long sk_offset = 0;
1733 path = btrfs_alloc_path();
1737 if (sk->tree_id == 0) {
1738 /* search the root of the inode that was passed */
1739 root = BTRFS_I(inode)->root;
1741 key.objectid = sk->tree_id;
1742 key.type = BTRFS_ROOT_ITEM_KEY;
1743 key.offset = (u64)-1;
1744 root = btrfs_read_fs_root_no_name(info, &key);
1746 printk(KERN_ERR "could not find root %llu\n",
1748 btrfs_free_path(path);
1753 key.objectid = sk->min_objectid;
1754 key.type = sk->min_type;
1755 key.offset = sk->min_offset;
1757 max_key.objectid = sk->max_objectid;
1758 max_key.type = sk->max_type;
1759 max_key.offset = sk->max_offset;
1761 path->keep_locks = 1;
1764 ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1771 ret = copy_to_sk(root, path, &key, sk, args->buf,
1772 &sk_offset, &num_found);
1773 btrfs_release_path(path);
1774 if (ret || num_found >= sk->nr_items)
1780 sk->nr_items = num_found;
1781 btrfs_free_path(path);
1785 static noinline int btrfs_ioctl_tree_search(struct file *file,
1788 struct btrfs_ioctl_search_args *args;
1789 struct inode *inode;
1792 if (!capable(CAP_SYS_ADMIN))
1795 args = memdup_user(argp, sizeof(*args));
1797 return PTR_ERR(args);
1799 inode = fdentry(file)->d_inode;
1800 ret = search_ioctl(inode, args);
1801 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1808 * Search INODE_REFs to identify path name of 'dirid' directory
1809 * in a 'tree_id' tree. and sets path name to 'name'.
1811 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1812 u64 tree_id, u64 dirid, char *name)
1814 struct btrfs_root *root;
1815 struct btrfs_key key;
1821 struct btrfs_inode_ref *iref;
1822 struct extent_buffer *l;
1823 struct btrfs_path *path;
1825 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1830 path = btrfs_alloc_path();
1834 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1836 key.objectid = tree_id;
1837 key.type = BTRFS_ROOT_ITEM_KEY;
1838 key.offset = (u64)-1;
1839 root = btrfs_read_fs_root_no_name(info, &key);
1841 printk(KERN_ERR "could not find root %llu\n", tree_id);
1846 key.objectid = dirid;
1847 key.type = BTRFS_INODE_REF_KEY;
1848 key.offset = (u64)-1;
1851 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1856 slot = path->slots[0];
1857 if (ret > 0 && slot > 0)
1859 btrfs_item_key_to_cpu(l, &key, slot);
1861 if (ret > 0 && (key.objectid != dirid ||
1862 key.type != BTRFS_INODE_REF_KEY)) {
1867 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1868 len = btrfs_inode_ref_name_len(l, iref);
1870 total_len += len + 1;
1875 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1877 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1880 btrfs_release_path(path);
1881 key.objectid = key.offset;
1882 key.offset = (u64)-1;
1883 dirid = key.objectid;
1887 memmove(name, ptr, total_len);
1888 name[total_len]='\0';
1891 btrfs_free_path(path);
1895 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1898 struct btrfs_ioctl_ino_lookup_args *args;
1899 struct inode *inode;
1902 if (!capable(CAP_SYS_ADMIN))
1905 args = memdup_user(argp, sizeof(*args));
1907 return PTR_ERR(args);
1909 inode = fdentry(file)->d_inode;
1911 if (args->treeid == 0)
1912 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1914 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1915 args->treeid, args->objectid,
1918 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1925 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1928 struct dentry *parent = fdentry(file);
1929 struct dentry *dentry;
1930 struct inode *dir = parent->d_inode;
1931 struct inode *inode;
1932 struct btrfs_root *root = BTRFS_I(dir)->root;
1933 struct btrfs_root *dest = NULL;
1934 struct btrfs_ioctl_vol_args *vol_args;
1935 struct btrfs_trans_handle *trans;
1940 vol_args = memdup_user(arg, sizeof(*vol_args));
1941 if (IS_ERR(vol_args))
1942 return PTR_ERR(vol_args);
1944 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1945 namelen = strlen(vol_args->name);
1946 if (strchr(vol_args->name, '/') ||
1947 strncmp(vol_args->name, "..", namelen) == 0) {
1952 err = mnt_want_write_file(file);
1956 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
1957 dentry = lookup_one_len(vol_args->name, parent, namelen);
1958 if (IS_ERR(dentry)) {
1959 err = PTR_ERR(dentry);
1960 goto out_unlock_dir;
1963 if (!dentry->d_inode) {
1968 inode = dentry->d_inode;
1969 dest = BTRFS_I(inode)->root;
1970 if (!capable(CAP_SYS_ADMIN)){
1972 * Regular user. Only allow this with a special mount
1973 * option, when the user has write+exec access to the
1974 * subvol root, and when rmdir(2) would have been
1977 * Note that this is _not_ check that the subvol is
1978 * empty or doesn't contain data that we wouldn't
1979 * otherwise be able to delete.
1981 * Users who want to delete empty subvols should try
1985 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1989 * Do not allow deletion if the parent dir is the same
1990 * as the dir to be deleted. That means the ioctl
1991 * must be called on the dentry referencing the root
1992 * of the subvol, not a random directory contained
1999 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2003 /* check if subvolume may be deleted by a non-root user */
2004 err = btrfs_may_delete(dir, dentry, 1);
2009 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2014 mutex_lock(&inode->i_mutex);
2015 err = d_invalidate(dentry);
2019 down_write(&root->fs_info->subvol_sem);
2021 err = may_destroy_subvol(dest);
2025 trans = btrfs_start_transaction(root, 0);
2026 if (IS_ERR(trans)) {
2027 err = PTR_ERR(trans);
2030 trans->block_rsv = &root->fs_info->global_block_rsv;
2032 ret = btrfs_unlink_subvol(trans, root, dir,
2033 dest->root_key.objectid,
2034 dentry->d_name.name,
2035 dentry->d_name.len);
2038 btrfs_abort_transaction(trans, root, ret);
2042 btrfs_record_root_in_trans(trans, dest);
2044 memset(&dest->root_item.drop_progress, 0,
2045 sizeof(dest->root_item.drop_progress));
2046 dest->root_item.drop_level = 0;
2047 btrfs_set_root_refs(&dest->root_item, 0);
2049 if (!xchg(&dest->orphan_item_inserted, 1)) {
2050 ret = btrfs_insert_orphan_item(trans,
2051 root->fs_info->tree_root,
2052 dest->root_key.objectid);
2054 btrfs_abort_transaction(trans, root, ret);
2060 ret = btrfs_end_transaction(trans, root);
2063 inode->i_flags |= S_DEAD;
2065 up_write(&root->fs_info->subvol_sem);
2067 mutex_unlock(&inode->i_mutex);
2069 shrink_dcache_sb(root->fs_info->sb);
2070 btrfs_invalidate_inodes(dest);
2076 mutex_unlock(&dir->i_mutex);
2077 mnt_drop_write_file(file);
2083 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2085 struct inode *inode = fdentry(file)->d_inode;
2086 struct btrfs_root *root = BTRFS_I(inode)->root;
2087 struct btrfs_ioctl_defrag_range_args *range;
2090 if (btrfs_root_readonly(root))
2093 ret = mnt_want_write_file(file);
2097 switch (inode->i_mode & S_IFMT) {
2099 if (!capable(CAP_SYS_ADMIN)) {
2103 ret = btrfs_defrag_root(root, 0);
2106 ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
2109 if (!(file->f_mode & FMODE_WRITE)) {
2114 range = kzalloc(sizeof(*range), GFP_KERNEL);
2121 if (copy_from_user(range, argp,
2127 /* compression requires us to start the IO */
2128 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2129 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2130 range->extent_thresh = (u32)-1;
2133 /* the rest are all set to zero by kzalloc */
2134 range->len = (u64)-1;
2136 ret = btrfs_defrag_file(fdentry(file)->d_inode, file,
2146 mnt_drop_write_file(file);
2150 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2152 struct btrfs_ioctl_vol_args *vol_args;
2155 if (!capable(CAP_SYS_ADMIN))
2158 mutex_lock(&root->fs_info->volume_mutex);
2159 if (root->fs_info->balance_ctl) {
2160 printk(KERN_INFO "btrfs: balance in progress\n");
2165 vol_args = memdup_user(arg, sizeof(*vol_args));
2166 if (IS_ERR(vol_args)) {
2167 ret = PTR_ERR(vol_args);
2171 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2172 ret = btrfs_init_new_device(root, vol_args->name);
2176 mutex_unlock(&root->fs_info->volume_mutex);
2180 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
2182 struct btrfs_ioctl_vol_args *vol_args;
2185 if (!capable(CAP_SYS_ADMIN))
2188 if (root->fs_info->sb->s_flags & MS_RDONLY)
2191 mutex_lock(&root->fs_info->volume_mutex);
2192 if (root->fs_info->balance_ctl) {
2193 printk(KERN_INFO "btrfs: balance in progress\n");
2198 vol_args = memdup_user(arg, sizeof(*vol_args));
2199 if (IS_ERR(vol_args)) {
2200 ret = PTR_ERR(vol_args);
2204 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2205 ret = btrfs_rm_device(root, vol_args->name);
2209 mutex_unlock(&root->fs_info->volume_mutex);
2213 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2215 struct btrfs_ioctl_fs_info_args *fi_args;
2216 struct btrfs_device *device;
2217 struct btrfs_device *next;
2218 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2221 if (!capable(CAP_SYS_ADMIN))
2224 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2228 fi_args->num_devices = fs_devices->num_devices;
2229 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2231 mutex_lock(&fs_devices->device_list_mutex);
2232 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2233 if (device->devid > fi_args->max_id)
2234 fi_args->max_id = device->devid;
2236 mutex_unlock(&fs_devices->device_list_mutex);
2238 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2245 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2247 struct btrfs_ioctl_dev_info_args *di_args;
2248 struct btrfs_device *dev;
2249 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2251 char *s_uuid = NULL;
2252 char empty_uuid[BTRFS_UUID_SIZE] = {0};
2254 if (!capable(CAP_SYS_ADMIN))
2257 di_args = memdup_user(arg, sizeof(*di_args));
2258 if (IS_ERR(di_args))
2259 return PTR_ERR(di_args);
2261 if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2262 s_uuid = di_args->uuid;
2264 mutex_lock(&fs_devices->device_list_mutex);
2265 dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL);
2266 mutex_unlock(&fs_devices->device_list_mutex);
2273 di_args->devid = dev->devid;
2274 di_args->bytes_used = dev->bytes_used;
2275 di_args->total_bytes = dev->total_bytes;
2276 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2278 struct rcu_string *name;
2281 name = rcu_dereference(dev->name);
2282 strncpy(di_args->path, name->str, sizeof(di_args->path));
2284 di_args->path[sizeof(di_args->path) - 1] = 0;
2286 di_args->path[0] = '\0';
2290 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2297 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2298 u64 off, u64 olen, u64 destoff)
2300 struct inode *inode = fdentry(file)->d_inode;
2301 struct btrfs_root *root = BTRFS_I(inode)->root;
2302 struct file *src_file;
2304 struct btrfs_trans_handle *trans;
2305 struct btrfs_path *path;
2306 struct extent_buffer *leaf;
2308 struct btrfs_key key;
2313 u64 bs = root->fs_info->sb->s_blocksize;
2318 * - split compressed inline extents. annoying: we need to
2319 * decompress into destination's address_space (the file offset
2320 * may change, so source mapping won't do), then recompress (or
2321 * otherwise reinsert) a subrange.
2322 * - allow ranges within the same file to be cloned (provided
2323 * they don't overlap)?
2326 /* the destination must be opened for writing */
2327 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
2330 if (btrfs_root_readonly(root))
2333 ret = mnt_want_write_file(file);
2337 src_file = fget(srcfd);
2340 goto out_drop_write;
2343 src = src_file->f_dentry->d_inode;
2349 /* the src must be open for reading */
2350 if (!(src_file->f_mode & FMODE_READ))
2353 /* don't make the dst file partly checksummed */
2354 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2355 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2359 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2363 if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
2367 buf = vmalloc(btrfs_level_size(root, 0));
2371 path = btrfs_alloc_path();
2379 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2380 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
2382 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2383 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
2386 /* determine range to clone */
2388 if (off + len > src->i_size || off + len < off)
2391 olen = len = src->i_size - off;
2392 /* if we extend to eof, continue to block boundary */
2393 if (off + len == src->i_size)
2394 len = ALIGN(src->i_size, bs) - off;
2396 /* verify the end result is block aligned */
2397 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2398 !IS_ALIGNED(destoff, bs))
2401 if (destoff > inode->i_size) {
2402 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2407 /* truncate page cache pages from target inode range */
2408 truncate_inode_pages_range(&inode->i_data, destoff,
2409 PAGE_CACHE_ALIGN(destoff + len) - 1);
2411 /* do any pending delalloc/csum calc on src, one way or
2412 another, and lock file content */
2414 struct btrfs_ordered_extent *ordered;
2415 lock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2416 ordered = btrfs_lookup_first_ordered_extent(src, off+len);
2418 !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len,
2419 EXTENT_DELALLOC, 0, NULL))
2421 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2423 btrfs_put_ordered_extent(ordered);
2424 btrfs_wait_ordered_range(src, off, len);
2428 key.objectid = btrfs_ino(src);
2429 key.type = BTRFS_EXTENT_DATA_KEY;
2434 * note the key will change type as we walk through the
2437 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2441 nritems = btrfs_header_nritems(path->nodes[0]);
2442 if (path->slots[0] >= nritems) {
2443 ret = btrfs_next_leaf(root, path);
2448 nritems = btrfs_header_nritems(path->nodes[0]);
2450 leaf = path->nodes[0];
2451 slot = path->slots[0];
2453 btrfs_item_key_to_cpu(leaf, &key, slot);
2454 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
2455 key.objectid != btrfs_ino(src))
2458 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2459 struct btrfs_file_extent_item *extent;
2462 struct btrfs_key new_key;
2463 u64 disko = 0, diskl = 0;
2464 u64 datao = 0, datal = 0;
2468 size = btrfs_item_size_nr(leaf, slot);
2469 read_extent_buffer(leaf, buf,
2470 btrfs_item_ptr_offset(leaf, slot),
2473 extent = btrfs_item_ptr(leaf, slot,
2474 struct btrfs_file_extent_item);
2475 comp = btrfs_file_extent_compression(leaf, extent);
2476 type = btrfs_file_extent_type(leaf, extent);
2477 if (type == BTRFS_FILE_EXTENT_REG ||
2478 type == BTRFS_FILE_EXTENT_PREALLOC) {
2479 disko = btrfs_file_extent_disk_bytenr(leaf,
2481 diskl = btrfs_file_extent_disk_num_bytes(leaf,
2483 datao = btrfs_file_extent_offset(leaf, extent);
2484 datal = btrfs_file_extent_num_bytes(leaf,
2486 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2487 /* take upper bound, may be compressed */
2488 datal = btrfs_file_extent_ram_bytes(leaf,
2491 btrfs_release_path(path);
2493 if (key.offset + datal <= off ||
2494 key.offset >= off+len)
2497 memcpy(&new_key, &key, sizeof(new_key));
2498 new_key.objectid = btrfs_ino(inode);
2499 if (off <= key.offset)
2500 new_key.offset = key.offset + destoff - off;
2502 new_key.offset = destoff;
2505 * 1 - adjusting old extent (we may have to split it)
2506 * 1 - add new extent
2509 trans = btrfs_start_transaction(root, 3);
2510 if (IS_ERR(trans)) {
2511 ret = PTR_ERR(trans);
2515 if (type == BTRFS_FILE_EXTENT_REG ||
2516 type == BTRFS_FILE_EXTENT_PREALLOC) {
2518 * a | --- range to clone ---| b
2519 * | ------------- extent ------------- |
2522 /* substract range b */
2523 if (key.offset + datal > off + len)
2524 datal = off + len - key.offset;
2526 /* substract range a */
2527 if (off > key.offset) {
2528 datao += off - key.offset;
2529 datal -= off - key.offset;
2532 ret = btrfs_drop_extents(trans, inode,
2534 new_key.offset + datal,
2537 btrfs_abort_transaction(trans, root,
2539 btrfs_end_transaction(trans, root);
2543 ret = btrfs_insert_empty_item(trans, root, path,
2546 btrfs_abort_transaction(trans, root,
2548 btrfs_end_transaction(trans, root);
2552 leaf = path->nodes[0];
2553 slot = path->slots[0];
2554 write_extent_buffer(leaf, buf,
2555 btrfs_item_ptr_offset(leaf, slot),
2558 extent = btrfs_item_ptr(leaf, slot,
2559 struct btrfs_file_extent_item);
2561 /* disko == 0 means it's a hole */
2565 btrfs_set_file_extent_offset(leaf, extent,
2567 btrfs_set_file_extent_num_bytes(leaf, extent,
2570 inode_add_bytes(inode, datal);
2571 ret = btrfs_inc_extent_ref(trans, root,
2573 root->root_key.objectid,
2575 new_key.offset - datao,
2578 btrfs_abort_transaction(trans,
2581 btrfs_end_transaction(trans,
2587 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2590 if (off > key.offset) {
2591 skip = off - key.offset;
2592 new_key.offset += skip;
2595 if (key.offset + datal > off+len)
2596 trim = key.offset + datal - (off+len);
2598 if (comp && (skip || trim)) {
2600 btrfs_end_transaction(trans, root);
2603 size -= skip + trim;
2604 datal -= skip + trim;
2606 ret = btrfs_drop_extents(trans, inode,
2608 new_key.offset + datal,
2611 btrfs_abort_transaction(trans, root,
2613 btrfs_end_transaction(trans, root);
2617 ret = btrfs_insert_empty_item(trans, root, path,
2620 btrfs_abort_transaction(trans, root,
2622 btrfs_end_transaction(trans, root);
2628 btrfs_file_extent_calc_inline_size(0);
2629 memmove(buf+start, buf+start+skip,
2633 leaf = path->nodes[0];
2634 slot = path->slots[0];
2635 write_extent_buffer(leaf, buf,
2636 btrfs_item_ptr_offset(leaf, slot),
2638 inode_add_bytes(inode, datal);
2641 btrfs_mark_buffer_dirty(leaf);
2642 btrfs_release_path(path);
2644 inode_inc_iversion(inode);
2645 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2648 * we round up to the block size at eof when
2649 * determining which extents to clone above,
2650 * but shouldn't round up the file size
2652 endoff = new_key.offset + datal;
2653 if (endoff > destoff+olen)
2654 endoff = destoff+olen;
2655 if (endoff > inode->i_size)
2656 btrfs_i_size_write(inode, endoff);
2658 ret = btrfs_update_inode(trans, root, inode);
2660 btrfs_abort_transaction(trans, root, ret);
2661 btrfs_end_transaction(trans, root);
2664 ret = btrfs_end_transaction(trans, root);
2667 btrfs_release_path(path);
2672 btrfs_release_path(path);
2673 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2675 mutex_unlock(&src->i_mutex);
2676 mutex_unlock(&inode->i_mutex);
2678 btrfs_free_path(path);
2682 mnt_drop_write_file(file);
2686 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
2688 struct btrfs_ioctl_clone_range_args args;
2690 if (copy_from_user(&args, argp, sizeof(args)))
2692 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2693 args.src_length, args.dest_offset);
2697 * there are many ways the trans_start and trans_end ioctls can lead
2698 * to deadlocks. They should only be used by applications that
2699 * basically own the machine, and have a very in depth understanding
2700 * of all the possible deadlocks and enospc problems.
2702 static long btrfs_ioctl_trans_start(struct file *file)
2704 struct inode *inode = fdentry(file)->d_inode;
2705 struct btrfs_root *root = BTRFS_I(inode)->root;
2706 struct btrfs_trans_handle *trans;
2710 if (!capable(CAP_SYS_ADMIN))
2714 if (file->private_data)
2718 if (btrfs_root_readonly(root))
2721 ret = mnt_want_write_file(file);
2725 atomic_inc(&root->fs_info->open_ioctl_trans);
2728 trans = btrfs_start_ioctl_transaction(root);
2732 file->private_data = trans;
2736 atomic_dec(&root->fs_info->open_ioctl_trans);
2737 mnt_drop_write_file(file);
2742 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2744 struct inode *inode = fdentry(file)->d_inode;
2745 struct btrfs_root *root = BTRFS_I(inode)->root;
2746 struct btrfs_root *new_root;
2747 struct btrfs_dir_item *di;
2748 struct btrfs_trans_handle *trans;
2749 struct btrfs_path *path;
2750 struct btrfs_key location;
2751 struct btrfs_disk_key disk_key;
2752 struct btrfs_super_block *disk_super;
2757 if (!capable(CAP_SYS_ADMIN))
2760 if (copy_from_user(&objectid, argp, sizeof(objectid)))
2764 objectid = root->root_key.objectid;
2766 location.objectid = objectid;
2767 location.type = BTRFS_ROOT_ITEM_KEY;
2768 location.offset = (u64)-1;
2770 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2771 if (IS_ERR(new_root))
2772 return PTR_ERR(new_root);
2774 if (btrfs_root_refs(&new_root->root_item) == 0)
2777 path = btrfs_alloc_path();
2780 path->leave_spinning = 1;
2782 trans = btrfs_start_transaction(root, 1);
2783 if (IS_ERR(trans)) {
2784 btrfs_free_path(path);
2785 return PTR_ERR(trans);
2788 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
2789 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2790 dir_id, "default", 7, 1);
2791 if (IS_ERR_OR_NULL(di)) {
2792 btrfs_free_path(path);
2793 btrfs_end_transaction(trans, root);
2794 printk(KERN_ERR "Umm, you don't have the default dir item, "
2795 "this isn't going to work\n");
2799 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2800 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2801 btrfs_mark_buffer_dirty(path->nodes[0]);
2802 btrfs_free_path(path);
2804 disk_super = root->fs_info->super_copy;
2805 features = btrfs_super_incompat_flags(disk_super);
2806 if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) {
2807 features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL;
2808 btrfs_set_super_incompat_flags(disk_super, features);
2810 btrfs_end_transaction(trans, root);
2815 static void get_block_group_info(struct list_head *groups_list,
2816 struct btrfs_ioctl_space_info *space)
2818 struct btrfs_block_group_cache *block_group;
2820 space->total_bytes = 0;
2821 space->used_bytes = 0;
2823 list_for_each_entry(block_group, groups_list, list) {
2824 space->flags = block_group->flags;
2825 space->total_bytes += block_group->key.offset;
2826 space->used_bytes +=
2827 btrfs_block_group_used(&block_group->item);
2831 long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2833 struct btrfs_ioctl_space_args space_args;
2834 struct btrfs_ioctl_space_info space;
2835 struct btrfs_ioctl_space_info *dest;
2836 struct btrfs_ioctl_space_info *dest_orig;
2837 struct btrfs_ioctl_space_info __user *user_dest;
2838 struct btrfs_space_info *info;
2839 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2840 BTRFS_BLOCK_GROUP_SYSTEM,
2841 BTRFS_BLOCK_GROUP_METADATA,
2842 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2849 if (copy_from_user(&space_args,
2850 (struct btrfs_ioctl_space_args __user *)arg,
2851 sizeof(space_args)))
2854 for (i = 0; i < num_types; i++) {
2855 struct btrfs_space_info *tmp;
2859 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2861 if (tmp->flags == types[i]) {
2871 down_read(&info->groups_sem);
2872 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2873 if (!list_empty(&info->block_groups[c]))
2876 up_read(&info->groups_sem);
2879 /* space_slots == 0 means they are asking for a count */
2880 if (space_args.space_slots == 0) {
2881 space_args.total_spaces = slot_count;
2885 slot_count = min_t(u64, space_args.space_slots, slot_count);
2887 alloc_size = sizeof(*dest) * slot_count;
2889 /* we generally have at most 6 or so space infos, one for each raid
2890 * level. So, a whole page should be more than enough for everyone
2892 if (alloc_size > PAGE_CACHE_SIZE)
2895 space_args.total_spaces = 0;
2896 dest = kmalloc(alloc_size, GFP_NOFS);
2901 /* now we have a buffer to copy into */
2902 for (i = 0; i < num_types; i++) {
2903 struct btrfs_space_info *tmp;
2910 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2912 if (tmp->flags == types[i]) {
2921 down_read(&info->groups_sem);
2922 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2923 if (!list_empty(&info->block_groups[c])) {
2924 get_block_group_info(&info->block_groups[c],
2926 memcpy(dest, &space, sizeof(space));
2928 space_args.total_spaces++;
2934 up_read(&info->groups_sem);
2937 user_dest = (struct btrfs_ioctl_space_info __user *)
2938 (arg + sizeof(struct btrfs_ioctl_space_args));
2940 if (copy_to_user(user_dest, dest_orig, alloc_size))
2945 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
2952 * there are many ways the trans_start and trans_end ioctls can lead
2953 * to deadlocks. They should only be used by applications that
2954 * basically own the machine, and have a very in depth understanding
2955 * of all the possible deadlocks and enospc problems.
2957 long btrfs_ioctl_trans_end(struct file *file)
2959 struct inode *inode = fdentry(file)->d_inode;
2960 struct btrfs_root *root = BTRFS_I(inode)->root;
2961 struct btrfs_trans_handle *trans;
2963 trans = file->private_data;
2966 file->private_data = NULL;
2968 btrfs_end_transaction(trans, root);
2970 atomic_dec(&root->fs_info->open_ioctl_trans);
2972 mnt_drop_write_file(file);
2976 static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
2978 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2979 struct btrfs_trans_handle *trans;
2983 trans = btrfs_start_transaction(root, 0);
2985 return PTR_ERR(trans);
2986 transid = trans->transid;
2987 ret = btrfs_commit_transaction_async(trans, root, 0);
2989 btrfs_end_transaction(trans, root);
2994 if (copy_to_user(argp, &transid, sizeof(transid)))
2999 static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
3001 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
3005 if (copy_from_user(&transid, argp, sizeof(transid)))
3008 transid = 0; /* current trans */
3010 return btrfs_wait_for_commit(root, transid);
3013 static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg)
3016 struct btrfs_ioctl_scrub_args *sa;
3018 if (!capable(CAP_SYS_ADMIN))
3021 sa = memdup_user(arg, sizeof(*sa));
3025 ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end,
3026 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY);
3028 if (copy_to_user(arg, sa, sizeof(*sa)))
3035 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3037 if (!capable(CAP_SYS_ADMIN))
3040 return btrfs_scrub_cancel(root);
3043 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3046 struct btrfs_ioctl_scrub_args *sa;
3049 if (!capable(CAP_SYS_ADMIN))
3052 sa = memdup_user(arg, sizeof(*sa));
3056 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3058 if (copy_to_user(arg, sa, sizeof(*sa)))
3065 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
3066 void __user *arg, int reset_after_read)
3068 struct btrfs_ioctl_get_dev_stats *sa;
3071 if (reset_after_read && !capable(CAP_SYS_ADMIN))
3074 sa = memdup_user(arg, sizeof(*sa));
3078 ret = btrfs_get_dev_stats(root, sa, reset_after_read);
3080 if (copy_to_user(arg, sa, sizeof(*sa)))
3087 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3093 struct btrfs_ioctl_ino_path_args *ipa = NULL;
3094 struct inode_fs_paths *ipath = NULL;
3095 struct btrfs_path *path;
3097 if (!capable(CAP_SYS_ADMIN))
3100 path = btrfs_alloc_path();
3106 ipa = memdup_user(arg, sizeof(*ipa));
3113 size = min_t(u32, ipa->size, 4096);
3114 ipath = init_ipath(size, root, path);
3115 if (IS_ERR(ipath)) {
3116 ret = PTR_ERR(ipath);
3121 ret = paths_from_inode(ipa->inum, ipath);
3125 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3126 rel_ptr = ipath->fspath->val[i] -
3127 (u64)(unsigned long)ipath->fspath->val;
3128 ipath->fspath->val[i] = rel_ptr;
3131 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3132 (void *)(unsigned long)ipath->fspath, size);
3139 btrfs_free_path(path);
3146 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3148 struct btrfs_data_container *inodes = ctx;
3149 const size_t c = 3 * sizeof(u64);
3151 if (inodes->bytes_left >= c) {
3152 inodes->bytes_left -= c;
3153 inodes->val[inodes->elem_cnt] = inum;
3154 inodes->val[inodes->elem_cnt + 1] = offset;
3155 inodes->val[inodes->elem_cnt + 2] = root;
3156 inodes->elem_cnt += 3;
3158 inodes->bytes_missing += c - inodes->bytes_left;
3159 inodes->bytes_left = 0;
3160 inodes->elem_missed += 3;
3166 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3171 u64 extent_item_pos;
3172 struct btrfs_ioctl_logical_ino_args *loi;
3173 struct btrfs_data_container *inodes = NULL;
3174 struct btrfs_path *path = NULL;
3175 struct btrfs_key key;
3177 if (!capable(CAP_SYS_ADMIN))
3180 loi = memdup_user(arg, sizeof(*loi));
3187 path = btrfs_alloc_path();
3193 size = min_t(u32, loi->size, 4096);
3194 inodes = init_data_container(size);
3195 if (IS_ERR(inodes)) {
3196 ret = PTR_ERR(inodes);
3201 ret = extent_from_logical(root->fs_info, loi->logical, path, &key);
3202 btrfs_release_path(path);
3204 if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK)
3209 extent_item_pos = loi->logical - key.objectid;
3210 ret = iterate_extent_inodes(root->fs_info, key.objectid,
3211 extent_item_pos, 0, build_ino_list,
3217 ret = copy_to_user((void *)(unsigned long)loi->inodes,
3218 (void *)(unsigned long)inodes, size);
3223 btrfs_free_path(path);
3230 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
3231 struct btrfs_ioctl_balance_args *bargs)
3233 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3235 bargs->flags = bctl->flags;
3237 if (atomic_read(&fs_info->balance_running))
3238 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3239 if (atomic_read(&fs_info->balance_pause_req))
3240 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3241 if (atomic_read(&fs_info->balance_cancel_req))
3242 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3244 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3245 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3246 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3249 spin_lock(&fs_info->balance_lock);
3250 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3251 spin_unlock(&fs_info->balance_lock);
3253 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3257 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3259 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3260 struct btrfs_fs_info *fs_info = root->fs_info;
3261 struct btrfs_ioctl_balance_args *bargs;
3262 struct btrfs_balance_control *bctl;
3265 if (!capable(CAP_SYS_ADMIN))
3268 if (fs_info->sb->s_flags & MS_RDONLY)
3271 ret = mnt_want_write(file->f_path.mnt);
3275 mutex_lock(&fs_info->volume_mutex);
3276 mutex_lock(&fs_info->balance_mutex);
3279 bargs = memdup_user(arg, sizeof(*bargs));
3280 if (IS_ERR(bargs)) {
3281 ret = PTR_ERR(bargs);
3285 if (bargs->flags & BTRFS_BALANCE_RESUME) {
3286 if (!fs_info->balance_ctl) {
3291 bctl = fs_info->balance_ctl;
3292 spin_lock(&fs_info->balance_lock);
3293 bctl->flags |= BTRFS_BALANCE_RESUME;
3294 spin_unlock(&fs_info->balance_lock);
3302 if (fs_info->balance_ctl) {
3307 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3313 bctl->fs_info = fs_info;
3315 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3316 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3317 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3319 bctl->flags = bargs->flags;
3321 /* balance everything - no filters */
3322 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
3326 ret = btrfs_balance(bctl, bargs);
3328 * bctl is freed in __cancel_balance or in free_fs_info if
3329 * restriper was paused all the way until unmount
3332 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3339 mutex_unlock(&fs_info->balance_mutex);
3340 mutex_unlock(&fs_info->volume_mutex);
3341 mnt_drop_write(file->f_path.mnt);
3345 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
3347 if (!capable(CAP_SYS_ADMIN))
3351 case BTRFS_BALANCE_CTL_PAUSE:
3352 return btrfs_pause_balance(root->fs_info);
3353 case BTRFS_BALANCE_CTL_CANCEL:
3354 return btrfs_cancel_balance(root->fs_info);
3360 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
3363 struct btrfs_fs_info *fs_info = root->fs_info;
3364 struct btrfs_ioctl_balance_args *bargs;
3367 if (!capable(CAP_SYS_ADMIN))
3370 mutex_lock(&fs_info->balance_mutex);
3371 if (!fs_info->balance_ctl) {
3376 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
3382 update_ioctl_balance_args(fs_info, 1, bargs);
3384 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3389 mutex_unlock(&fs_info->balance_mutex);
3393 long btrfs_ioctl(struct file *file, unsigned int
3394 cmd, unsigned long arg)
3396 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3397 void __user *argp = (void __user *)arg;
3400 case FS_IOC_GETFLAGS:
3401 return btrfs_ioctl_getflags(file, argp);
3402 case FS_IOC_SETFLAGS:
3403 return btrfs_ioctl_setflags(file, argp);
3404 case FS_IOC_GETVERSION:
3405 return btrfs_ioctl_getversion(file, argp);
3407 return btrfs_ioctl_fitrim(file, argp);
3408 case BTRFS_IOC_SNAP_CREATE:
3409 return btrfs_ioctl_snap_create(file, argp, 0);
3410 case BTRFS_IOC_SNAP_CREATE_V2:
3411 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3412 case BTRFS_IOC_SUBVOL_CREATE:
3413 return btrfs_ioctl_snap_create(file, argp, 1);
3414 case BTRFS_IOC_SNAP_DESTROY:
3415 return btrfs_ioctl_snap_destroy(file, argp);
3416 case BTRFS_IOC_SUBVOL_GETFLAGS:
3417 return btrfs_ioctl_subvol_getflags(file, argp);
3418 case BTRFS_IOC_SUBVOL_SETFLAGS:
3419 return btrfs_ioctl_subvol_setflags(file, argp);
3420 case BTRFS_IOC_DEFAULT_SUBVOL:
3421 return btrfs_ioctl_default_subvol(file, argp);
3422 case BTRFS_IOC_DEFRAG:
3423 return btrfs_ioctl_defrag(file, NULL);
3424 case BTRFS_IOC_DEFRAG_RANGE:
3425 return btrfs_ioctl_defrag(file, argp);
3426 case BTRFS_IOC_RESIZE:
3427 return btrfs_ioctl_resize(root, argp);
3428 case BTRFS_IOC_ADD_DEV:
3429 return btrfs_ioctl_add_dev(root, argp);
3430 case BTRFS_IOC_RM_DEV:
3431 return btrfs_ioctl_rm_dev(root, argp);
3432 case BTRFS_IOC_FS_INFO:
3433 return btrfs_ioctl_fs_info(root, argp);
3434 case BTRFS_IOC_DEV_INFO:
3435 return btrfs_ioctl_dev_info(root, argp);
3436 case BTRFS_IOC_BALANCE:
3437 return btrfs_ioctl_balance(file, NULL);
3438 case BTRFS_IOC_CLONE:
3439 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
3440 case BTRFS_IOC_CLONE_RANGE:
3441 return btrfs_ioctl_clone_range(file, argp);
3442 case BTRFS_IOC_TRANS_START:
3443 return btrfs_ioctl_trans_start(file);
3444 case BTRFS_IOC_TRANS_END:
3445 return btrfs_ioctl_trans_end(file);
3446 case BTRFS_IOC_TREE_SEARCH:
3447 return btrfs_ioctl_tree_search(file, argp);
3448 case BTRFS_IOC_INO_LOOKUP:
3449 return btrfs_ioctl_ino_lookup(file, argp);
3450 case BTRFS_IOC_INO_PATHS:
3451 return btrfs_ioctl_ino_to_path(root, argp);
3452 case BTRFS_IOC_LOGICAL_INO:
3453 return btrfs_ioctl_logical_to_ino(root, argp);
3454 case BTRFS_IOC_SPACE_INFO:
3455 return btrfs_ioctl_space_info(root, argp);
3456 case BTRFS_IOC_SYNC:
3457 btrfs_sync_fs(file->f_dentry->d_sb, 1);
3459 case BTRFS_IOC_START_SYNC:
3460 return btrfs_ioctl_start_sync(file, argp);
3461 case BTRFS_IOC_WAIT_SYNC:
3462 return btrfs_ioctl_wait_sync(file, argp);
3463 case BTRFS_IOC_SCRUB:
3464 return btrfs_ioctl_scrub(root, argp);
3465 case BTRFS_IOC_SCRUB_CANCEL:
3466 return btrfs_ioctl_scrub_cancel(root, argp);
3467 case BTRFS_IOC_SCRUB_PROGRESS:
3468 return btrfs_ioctl_scrub_progress(root, argp);
3469 case BTRFS_IOC_BALANCE_V2:
3470 return btrfs_ioctl_balance(file, argp);
3471 case BTRFS_IOC_BALANCE_CTL:
3472 return btrfs_ioctl_balance_ctl(root, arg);
3473 case BTRFS_IOC_BALANCE_PROGRESS:
3474 return btrfs_ioctl_balance_progress(root, argp);
3475 case BTRFS_IOC_GET_DEV_STATS:
3476 return btrfs_ioctl_get_dev_stats(root, argp, 0);
3477 case BTRFS_IOC_GET_AND_RESET_DEV_STATS:
3478 return btrfs_ioctl_get_dev_stats(root, argp, 1);