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"
56 /* Mask out flags that are inappropriate for the given type of inode. */
57 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
61 else if (S_ISREG(mode))
62 return flags & ~FS_DIRSYNC_FL;
64 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
68 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
70 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
72 unsigned int iflags = 0;
74 if (flags & BTRFS_INODE_SYNC)
76 if (flags & BTRFS_INODE_IMMUTABLE)
77 iflags |= FS_IMMUTABLE_FL;
78 if (flags & BTRFS_INODE_APPEND)
79 iflags |= FS_APPEND_FL;
80 if (flags & BTRFS_INODE_NODUMP)
81 iflags |= FS_NODUMP_FL;
82 if (flags & BTRFS_INODE_NOATIME)
83 iflags |= FS_NOATIME_FL;
84 if (flags & BTRFS_INODE_DIRSYNC)
85 iflags |= FS_DIRSYNC_FL;
86 if (flags & BTRFS_INODE_NODATACOW)
87 iflags |= FS_NOCOW_FL;
89 if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
90 iflags |= FS_COMPR_FL;
91 else if (flags & BTRFS_INODE_NOCOMPRESS)
92 iflags |= FS_NOCOMP_FL;
98 * Update inode->i_flags based on the btrfs internal flags.
100 void btrfs_update_iflags(struct inode *inode)
102 struct btrfs_inode *ip = BTRFS_I(inode);
104 inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
106 if (ip->flags & BTRFS_INODE_SYNC)
107 inode->i_flags |= S_SYNC;
108 if (ip->flags & BTRFS_INODE_IMMUTABLE)
109 inode->i_flags |= S_IMMUTABLE;
110 if (ip->flags & BTRFS_INODE_APPEND)
111 inode->i_flags |= S_APPEND;
112 if (ip->flags & BTRFS_INODE_NOATIME)
113 inode->i_flags |= S_NOATIME;
114 if (ip->flags & BTRFS_INODE_DIRSYNC)
115 inode->i_flags |= S_DIRSYNC;
119 * Inherit flags from the parent inode.
121 * Currently only the compression flags and the cow flags are inherited.
123 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
130 flags = BTRFS_I(dir)->flags;
132 if (flags & BTRFS_INODE_NOCOMPRESS) {
133 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
134 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
135 } else if (flags & BTRFS_INODE_COMPRESS) {
136 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
137 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
140 if (flags & BTRFS_INODE_NODATACOW)
141 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
143 btrfs_update_iflags(inode);
146 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
148 struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode);
149 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
151 if (copy_to_user(arg, &flags, sizeof(flags)))
156 static int check_flags(unsigned int flags)
158 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
159 FS_NOATIME_FL | FS_NODUMP_FL | \
160 FS_SYNC_FL | FS_DIRSYNC_FL | \
161 FS_NOCOMP_FL | FS_COMPR_FL |
165 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
171 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
173 struct inode *inode = file->f_path.dentry->d_inode;
174 struct btrfs_inode *ip = BTRFS_I(inode);
175 struct btrfs_root *root = ip->root;
176 struct btrfs_trans_handle *trans;
177 unsigned int flags, oldflags;
180 unsigned int i_oldflags;
182 if (btrfs_root_readonly(root))
185 if (copy_from_user(&flags, arg, sizeof(flags)))
188 ret = check_flags(flags);
192 if (!inode_owner_or_capable(inode))
195 mutex_lock(&inode->i_mutex);
197 ip_oldflags = ip->flags;
198 i_oldflags = inode->i_flags;
200 flags = btrfs_mask_flags(inode->i_mode, flags);
201 oldflags = btrfs_flags_to_ioctl(ip->flags);
202 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
203 if (!capable(CAP_LINUX_IMMUTABLE)) {
209 ret = mnt_want_write_file(file);
213 if (flags & FS_SYNC_FL)
214 ip->flags |= BTRFS_INODE_SYNC;
216 ip->flags &= ~BTRFS_INODE_SYNC;
217 if (flags & FS_IMMUTABLE_FL)
218 ip->flags |= BTRFS_INODE_IMMUTABLE;
220 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
221 if (flags & FS_APPEND_FL)
222 ip->flags |= BTRFS_INODE_APPEND;
224 ip->flags &= ~BTRFS_INODE_APPEND;
225 if (flags & FS_NODUMP_FL)
226 ip->flags |= BTRFS_INODE_NODUMP;
228 ip->flags &= ~BTRFS_INODE_NODUMP;
229 if (flags & FS_NOATIME_FL)
230 ip->flags |= BTRFS_INODE_NOATIME;
232 ip->flags &= ~BTRFS_INODE_NOATIME;
233 if (flags & FS_DIRSYNC_FL)
234 ip->flags |= BTRFS_INODE_DIRSYNC;
236 ip->flags &= ~BTRFS_INODE_DIRSYNC;
237 if (flags & FS_NOCOW_FL)
238 ip->flags |= BTRFS_INODE_NODATACOW;
240 ip->flags &= ~BTRFS_INODE_NODATACOW;
243 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
244 * flag may be changed automatically if compression code won't make
247 if (flags & FS_NOCOMP_FL) {
248 ip->flags &= ~BTRFS_INODE_COMPRESS;
249 ip->flags |= BTRFS_INODE_NOCOMPRESS;
250 } else if (flags & FS_COMPR_FL) {
251 ip->flags |= BTRFS_INODE_COMPRESS;
252 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
254 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
257 trans = btrfs_start_transaction(root, 1);
259 ret = PTR_ERR(trans);
263 btrfs_update_iflags(inode);
264 inode_inc_iversion(inode);
265 inode->i_ctime = CURRENT_TIME;
266 ret = btrfs_update_inode(trans, root, inode);
268 btrfs_end_transaction(trans, root);
271 ip->flags = ip_oldflags;
272 inode->i_flags = i_oldflags;
275 mnt_drop_write_file(file);
277 mutex_unlock(&inode->i_mutex);
281 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
283 struct inode *inode = file->f_path.dentry->d_inode;
285 return put_user(inode->i_generation, arg);
288 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
290 struct btrfs_fs_info *fs_info = btrfs_sb(fdentry(file)->d_sb);
291 struct btrfs_device *device;
292 struct request_queue *q;
293 struct fstrim_range range;
294 u64 minlen = ULLONG_MAX;
296 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
299 if (!capable(CAP_SYS_ADMIN))
303 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
307 q = bdev_get_queue(device->bdev);
308 if (blk_queue_discard(q)) {
310 minlen = min((u64)q->limits.discard_granularity,
318 if (copy_from_user(&range, arg, sizeof(range)))
320 if (range.start > total_bytes)
323 range.len = min(range.len, total_bytes - range.start);
324 range.minlen = max(range.minlen, minlen);
325 ret = btrfs_trim_fs(fs_info->tree_root, &range);
329 if (copy_to_user(arg, &range, sizeof(range)))
335 static noinline int create_subvol(struct btrfs_root *root,
336 struct dentry *dentry,
337 char *name, int namelen,
340 struct btrfs_trans_handle *trans;
341 struct btrfs_key key;
342 struct btrfs_root_item root_item;
343 struct btrfs_inode_item *inode_item;
344 struct extent_buffer *leaf;
345 struct btrfs_root *new_root;
346 struct dentry *parent = dentry->d_parent;
351 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
354 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
358 dir = parent->d_inode;
366 trans = btrfs_start_transaction(root, 6);
368 return PTR_ERR(trans);
370 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
371 0, objectid, NULL, 0, 0, 0, 0);
377 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
378 btrfs_set_header_bytenr(leaf, leaf->start);
379 btrfs_set_header_generation(leaf, trans->transid);
380 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
381 btrfs_set_header_owner(leaf, objectid);
383 write_extent_buffer(leaf, root->fs_info->fsid,
384 (unsigned long)btrfs_header_fsid(leaf),
386 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
387 (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
389 btrfs_mark_buffer_dirty(leaf);
391 inode_item = &root_item.inode;
392 memset(inode_item, 0, sizeof(*inode_item));
393 inode_item->generation = cpu_to_le64(1);
394 inode_item->size = cpu_to_le64(3);
395 inode_item->nlink = cpu_to_le32(1);
396 inode_item->nbytes = cpu_to_le64(root->leafsize);
397 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
400 root_item.byte_limit = 0;
401 inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
403 btrfs_set_root_bytenr(&root_item, leaf->start);
404 btrfs_set_root_generation(&root_item, trans->transid);
405 btrfs_set_root_level(&root_item, 0);
406 btrfs_set_root_refs(&root_item, 1);
407 btrfs_set_root_used(&root_item, leaf->len);
408 btrfs_set_root_last_snapshot(&root_item, 0);
410 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
411 root_item.drop_level = 0;
413 btrfs_tree_unlock(leaf);
414 free_extent_buffer(leaf);
417 btrfs_set_root_dirid(&root_item, new_dirid);
419 key.objectid = objectid;
421 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
422 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
427 key.offset = (u64)-1;
428 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
429 if (IS_ERR(new_root)) {
430 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
431 ret = PTR_ERR(new_root);
435 btrfs_record_root_in_trans(trans, new_root);
437 ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
439 /* We potentially lose an unused inode item here */
440 btrfs_abort_transaction(trans, root, ret);
445 * insert the directory item
447 ret = btrfs_set_inode_index(dir, &index);
449 btrfs_abort_transaction(trans, root, ret);
453 ret = btrfs_insert_dir_item(trans, root,
454 name, namelen, dir, &key,
455 BTRFS_FT_DIR, index);
457 btrfs_abort_transaction(trans, root, ret);
461 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
462 ret = btrfs_update_inode(trans, root, dir);
465 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
466 objectid, root->root_key.objectid,
467 btrfs_ino(dir), index, name, namelen);
471 d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
474 *async_transid = trans->transid;
475 err = btrfs_commit_transaction_async(trans, root, 1);
477 err = btrfs_commit_transaction(trans, root);
484 static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
485 char *name, int namelen, u64 *async_transid,
489 struct btrfs_pending_snapshot *pending_snapshot;
490 struct btrfs_trans_handle *trans;
496 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
497 if (!pending_snapshot)
500 btrfs_init_block_rsv(&pending_snapshot->block_rsv);
501 pending_snapshot->dentry = dentry;
502 pending_snapshot->root = root;
503 pending_snapshot->readonly = readonly;
505 trans = btrfs_start_transaction(root->fs_info->extent_root, 5);
507 ret = PTR_ERR(trans);
511 ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
514 spin_lock(&root->fs_info->trans_lock);
515 list_add(&pending_snapshot->list,
516 &trans->transaction->pending_snapshots);
517 spin_unlock(&root->fs_info->trans_lock);
519 *async_transid = trans->transid;
520 ret = btrfs_commit_transaction_async(trans,
521 root->fs_info->extent_root, 1);
523 ret = btrfs_commit_transaction(trans,
524 root->fs_info->extent_root);
528 ret = pending_snapshot->error;
532 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
536 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
538 ret = PTR_ERR(inode);
542 d_instantiate(dentry, inode);
545 kfree(pending_snapshot);
549 /* copy of check_sticky in fs/namei.c()
550 * It's inline, so penalty for filesystems that don't use sticky bit is
553 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
555 uid_t fsuid = current_fsuid();
557 if (!(dir->i_mode & S_ISVTX))
559 if (inode->i_uid == fsuid)
561 if (dir->i_uid == fsuid)
563 return !capable(CAP_FOWNER);
566 /* copy of may_delete in fs/namei.c()
567 * Check whether we can remove a link victim from directory dir, check
568 * whether the type of victim is right.
569 * 1. We can't do it if dir is read-only (done in permission())
570 * 2. We should have write and exec permissions on dir
571 * 3. We can't remove anything from append-only dir
572 * 4. We can't do anything with immutable dir (done in permission())
573 * 5. If the sticky bit on dir is set we should either
574 * a. be owner of dir, or
575 * b. be owner of victim, or
576 * c. have CAP_FOWNER capability
577 * 6. If the victim is append-only or immutable we can't do antyhing with
578 * links pointing to it.
579 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
580 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
581 * 9. We can't remove a root or mountpoint.
582 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
583 * nfs_async_unlink().
586 static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
590 if (!victim->d_inode)
593 BUG_ON(victim->d_parent->d_inode != dir);
594 audit_inode_child(victim, dir);
596 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
601 if (btrfs_check_sticky(dir, victim->d_inode)||
602 IS_APPEND(victim->d_inode)||
603 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
606 if (!S_ISDIR(victim->d_inode->i_mode))
610 } else if (S_ISDIR(victim->d_inode->i_mode))
614 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
619 /* copy of may_create in fs/namei.c() */
620 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
626 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
630 * Create a new subvolume below @parent. This is largely modeled after
631 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
632 * inside this filesystem so it's quite a bit simpler.
634 static noinline int btrfs_mksubvol(struct path *parent,
635 char *name, int namelen,
636 struct btrfs_root *snap_src,
637 u64 *async_transid, bool readonly)
639 struct inode *dir = parent->dentry->d_inode;
640 struct dentry *dentry;
643 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
645 dentry = lookup_one_len(name, parent->dentry, namelen);
646 error = PTR_ERR(dentry);
654 error = mnt_want_write(parent->mnt);
658 error = btrfs_may_create(dir, dentry);
662 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
664 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
668 error = create_snapshot(snap_src, dentry,
669 name, namelen, async_transid, readonly);
671 error = create_subvol(BTRFS_I(dir)->root, dentry,
672 name, namelen, async_transid);
675 fsnotify_mkdir(dir, dentry);
677 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
679 mnt_drop_write(parent->mnt);
683 mutex_unlock(&dir->i_mutex);
688 * When we're defragging a range, we don't want to kick it off again
689 * if it is really just waiting for delalloc to send it down.
690 * If we find a nice big extent or delalloc range for the bytes in the
691 * file you want to defrag, we return 0 to let you know to skip this
694 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
696 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
697 struct extent_map *em = NULL;
698 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
701 read_lock(&em_tree->lock);
702 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
703 read_unlock(&em_tree->lock);
706 end = extent_map_end(em);
708 if (end - offset > thresh)
711 /* if we already have a nice delalloc here, just stop */
713 end = count_range_bits(io_tree, &offset, offset + thresh,
714 thresh, EXTENT_DELALLOC, 1);
721 * helper function to walk through a file and find extents
722 * newer than a specific transid, and smaller than thresh.
724 * This is used by the defragging code to find new and small
727 static int find_new_extents(struct btrfs_root *root,
728 struct inode *inode, u64 newer_than,
729 u64 *off, int thresh)
731 struct btrfs_path *path;
732 struct btrfs_key min_key;
733 struct btrfs_key max_key;
734 struct extent_buffer *leaf;
735 struct btrfs_file_extent_item *extent;
738 u64 ino = btrfs_ino(inode);
740 path = btrfs_alloc_path();
744 min_key.objectid = ino;
745 min_key.type = BTRFS_EXTENT_DATA_KEY;
746 min_key.offset = *off;
748 max_key.objectid = ino;
749 max_key.type = (u8)-1;
750 max_key.offset = (u64)-1;
752 path->keep_locks = 1;
755 ret = btrfs_search_forward(root, &min_key, &max_key,
756 path, 0, newer_than);
759 if (min_key.objectid != ino)
761 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
764 leaf = path->nodes[0];
765 extent = btrfs_item_ptr(leaf, path->slots[0],
766 struct btrfs_file_extent_item);
768 type = btrfs_file_extent_type(leaf, extent);
769 if (type == BTRFS_FILE_EXTENT_REG &&
770 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
771 check_defrag_in_cache(inode, min_key.offset, thresh)) {
772 *off = min_key.offset;
773 btrfs_free_path(path);
777 if (min_key.offset == (u64)-1)
781 btrfs_release_path(path);
784 btrfs_free_path(path);
789 * Validaty check of prev em and next em:
791 * 2) prev/next em is an hole/inline extent
793 static int check_adjacent_extents(struct inode *inode, struct extent_map *em)
795 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
796 struct extent_map *prev = NULL, *next = NULL;
799 read_lock(&em_tree->lock);
800 prev = lookup_extent_mapping(em_tree, em->start - 1, (u64)-1);
801 next = lookup_extent_mapping(em_tree, em->start + em->len, (u64)-1);
802 read_unlock(&em_tree->lock);
804 if ((!prev || prev->block_start >= EXTENT_MAP_LAST_BYTE) &&
805 (!next || next->block_start >= EXTENT_MAP_LAST_BYTE))
807 free_extent_map(prev);
808 free_extent_map(next);
813 static int should_defrag_range(struct inode *inode, u64 start, u64 len,
814 int thresh, u64 *last_len, u64 *skip,
817 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
818 struct extent_map *em = NULL;
819 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
823 * make sure that once we start defragging an extent, we keep on
826 if (start < *defrag_end)
832 * hopefully we have this extent in the tree already, try without
833 * the full extent lock
835 read_lock(&em_tree->lock);
836 em = lookup_extent_mapping(em_tree, start, len);
837 read_unlock(&em_tree->lock);
840 /* get the big lock and read metadata off disk */
841 lock_extent(io_tree, start, start + len - 1);
842 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
843 unlock_extent(io_tree, start, start + len - 1);
849 /* this will cover holes, and inline extents */
850 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
855 /* If we have nothing to merge with us, just skip. */
856 if (check_adjacent_extents(inode, em)) {
862 * we hit a real extent, if it is big don't bother defragging it again
864 if ((*last_len == 0 || *last_len >= thresh) && em->len >= thresh)
869 * last_len ends up being a counter of how many bytes we've defragged.
870 * every time we choose not to defrag an extent, we reset *last_len
871 * so that the next tiny extent will force a defrag.
873 * The end result of this is that tiny extents before a single big
874 * extent will force at least part of that big extent to be defragged.
877 *defrag_end = extent_map_end(em);
880 *skip = extent_map_end(em);
889 * it doesn't do much good to defrag one or two pages
890 * at a time. This pulls in a nice chunk of pages
893 * It also makes sure the delalloc code has enough
894 * dirty data to avoid making new small extents as part
897 * It's a good idea to start RA on this range
898 * before calling this.
900 static int cluster_pages_for_defrag(struct inode *inode,
902 unsigned long start_index,
905 unsigned long file_end;
906 u64 isize = i_size_read(inode);
913 struct btrfs_ordered_extent *ordered;
914 struct extent_state *cached_state = NULL;
915 struct extent_io_tree *tree;
916 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
918 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
919 if (!isize || start_index > file_end)
922 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
924 ret = btrfs_delalloc_reserve_space(inode,
925 page_cnt << PAGE_CACHE_SHIFT);
929 tree = &BTRFS_I(inode)->io_tree;
931 /* step one, lock all the pages */
932 for (i = 0; i < page_cnt; i++) {
935 page = find_or_create_page(inode->i_mapping,
936 start_index + i, mask);
940 page_start = page_offset(page);
941 page_end = page_start + PAGE_CACHE_SIZE - 1;
943 lock_extent(tree, page_start, page_end);
944 ordered = btrfs_lookup_ordered_extent(inode,
946 unlock_extent(tree, page_start, page_end);
951 btrfs_start_ordered_extent(inode, ordered, 1);
952 btrfs_put_ordered_extent(ordered);
955 * we unlocked the page above, so we need check if
956 * it was released or not.
958 if (page->mapping != inode->i_mapping) {
960 page_cache_release(page);
965 if (!PageUptodate(page)) {
966 btrfs_readpage(NULL, page);
968 if (!PageUptodate(page)) {
970 page_cache_release(page);
976 if (page->mapping != inode->i_mapping) {
978 page_cache_release(page);
988 if (!(inode->i_sb->s_flags & MS_ACTIVE))
992 * so now we have a nice long stream of locked
993 * and up to date pages, lets wait on them
995 for (i = 0; i < i_done; i++)
996 wait_on_page_writeback(pages[i]);
998 page_start = page_offset(pages[0]);
999 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1001 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1002 page_start, page_end - 1, 0, &cached_state);
1003 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1004 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1005 EXTENT_DO_ACCOUNTING, 0, 0, &cached_state,
1008 if (i_done != page_cnt) {
1009 spin_lock(&BTRFS_I(inode)->lock);
1010 BTRFS_I(inode)->outstanding_extents++;
1011 spin_unlock(&BTRFS_I(inode)->lock);
1012 btrfs_delalloc_release_space(inode,
1013 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1017 btrfs_set_extent_delalloc(inode, page_start, page_end - 1,
1020 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1021 page_start, page_end - 1, &cached_state,
1024 for (i = 0; i < i_done; i++) {
1025 clear_page_dirty_for_io(pages[i]);
1026 ClearPageChecked(pages[i]);
1027 set_page_extent_mapped(pages[i]);
1028 set_page_dirty(pages[i]);
1029 unlock_page(pages[i]);
1030 page_cache_release(pages[i]);
1034 for (i = 0; i < i_done; i++) {
1035 unlock_page(pages[i]);
1036 page_cache_release(pages[i]);
1038 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1043 int btrfs_defrag_file(struct inode *inode, struct file *file,
1044 struct btrfs_ioctl_defrag_range_args *range,
1045 u64 newer_than, unsigned long max_to_defrag)
1047 struct btrfs_root *root = BTRFS_I(inode)->root;
1048 struct btrfs_super_block *disk_super;
1049 struct file_ra_state *ra = NULL;
1050 unsigned long last_index;
1051 u64 isize = i_size_read(inode);
1056 u64 newer_off = range->start;
1058 unsigned long ra_index = 0;
1060 int defrag_count = 0;
1061 int compress_type = BTRFS_COMPRESS_ZLIB;
1062 int extent_thresh = range->extent_thresh;
1063 int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1064 int cluster = max_cluster;
1065 u64 new_align = ~((u64)128 * 1024 - 1);
1066 struct page **pages = NULL;
1068 if (extent_thresh == 0)
1069 extent_thresh = 256 * 1024;
1071 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1072 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1074 if (range->compress_type)
1075 compress_type = range->compress_type;
1082 * if we were not given a file, allocate a readahead
1086 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1089 file_ra_state_init(ra, inode->i_mapping);
1094 pages = kmalloc(sizeof(struct page *) * max_cluster,
1101 /* find the last page to defrag */
1102 if (range->start + range->len > range->start) {
1103 last_index = min_t(u64, isize - 1,
1104 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1106 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1110 ret = find_new_extents(root, inode, newer_than,
1111 &newer_off, 64 * 1024);
1113 range->start = newer_off;
1115 * we always align our defrag to help keep
1116 * the extents in the file evenly spaced
1118 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1122 i = range->start >> PAGE_CACHE_SHIFT;
1125 max_to_defrag = last_index + 1;
1128 * make writeback starts from i, so the defrag range can be
1129 * written sequentially.
1131 if (i < inode->i_mapping->writeback_index)
1132 inode->i_mapping->writeback_index = i;
1134 while (i <= last_index && defrag_count < max_to_defrag &&
1135 (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1136 PAGE_CACHE_SHIFT)) {
1138 * make sure we stop running if someone unmounts
1141 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1144 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1145 PAGE_CACHE_SIZE, extent_thresh,
1146 &last_len, &skip, &defrag_end)) {
1149 * the should_defrag function tells us how much to skip
1150 * bump our counter by the suggested amount
1152 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1153 i = max(i + 1, next);
1158 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1159 PAGE_CACHE_SHIFT) - i;
1160 cluster = min(cluster, max_cluster);
1162 cluster = max_cluster;
1165 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1166 BTRFS_I(inode)->force_compress = compress_type;
1168 if (i + cluster > ra_index) {
1169 ra_index = max(i, ra_index);
1170 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1172 ra_index += max_cluster;
1175 mutex_lock(&inode->i_mutex);
1176 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1178 mutex_unlock(&inode->i_mutex);
1182 defrag_count += ret;
1183 balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret);
1184 mutex_unlock(&inode->i_mutex);
1187 if (newer_off == (u64)-1)
1193 newer_off = max(newer_off + 1,
1194 (u64)i << PAGE_CACHE_SHIFT);
1196 ret = find_new_extents(root, inode,
1197 newer_than, &newer_off,
1200 range->start = newer_off;
1201 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1208 last_len += ret << PAGE_CACHE_SHIFT;
1216 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1217 filemap_flush(inode->i_mapping);
1219 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1220 /* the filemap_flush will queue IO into the worker threads, but
1221 * we have to make sure the IO is actually started and that
1222 * ordered extents get created before we return
1224 atomic_inc(&root->fs_info->async_submit_draining);
1225 while (atomic_read(&root->fs_info->nr_async_submits) ||
1226 atomic_read(&root->fs_info->async_delalloc_pages)) {
1227 wait_event(root->fs_info->async_submit_wait,
1228 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1229 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1231 atomic_dec(&root->fs_info->async_submit_draining);
1233 mutex_lock(&inode->i_mutex);
1234 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1235 mutex_unlock(&inode->i_mutex);
1238 disk_super = root->fs_info->super_copy;
1239 features = btrfs_super_incompat_flags(disk_super);
1240 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1241 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
1242 btrfs_set_super_incompat_flags(disk_super, features);
1254 static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
1260 struct btrfs_ioctl_vol_args *vol_args;
1261 struct btrfs_trans_handle *trans;
1262 struct btrfs_device *device = NULL;
1264 char *devstr = NULL;
1268 if (root->fs_info->sb->s_flags & MS_RDONLY)
1271 if (!capable(CAP_SYS_ADMIN))
1274 mutex_lock(&root->fs_info->volume_mutex);
1275 if (root->fs_info->balance_ctl) {
1276 printk(KERN_INFO "btrfs: balance in progress\n");
1281 vol_args = memdup_user(arg, sizeof(*vol_args));
1282 if (IS_ERR(vol_args)) {
1283 ret = PTR_ERR(vol_args);
1287 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1289 sizestr = vol_args->name;
1290 devstr = strchr(sizestr, ':');
1293 sizestr = devstr + 1;
1295 devstr = vol_args->name;
1296 devid = simple_strtoull(devstr, &end, 10);
1297 printk(KERN_INFO "btrfs: resizing devid %llu\n",
1298 (unsigned long long)devid);
1300 device = btrfs_find_device(root, devid, NULL, NULL);
1302 printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
1303 (unsigned long long)devid);
1307 if (!strcmp(sizestr, "max"))
1308 new_size = device->bdev->bd_inode->i_size;
1310 if (sizestr[0] == '-') {
1313 } else if (sizestr[0] == '+') {
1317 new_size = memparse(sizestr, NULL);
1318 if (new_size == 0) {
1324 old_size = device->total_bytes;
1327 if (new_size > old_size) {
1331 new_size = old_size - new_size;
1332 } else if (mod > 0) {
1333 new_size = old_size + new_size;
1336 if (new_size < 256 * 1024 * 1024) {
1340 if (new_size > device->bdev->bd_inode->i_size) {
1345 do_div(new_size, root->sectorsize);
1346 new_size *= root->sectorsize;
1348 printk(KERN_INFO "btrfs: new size for %s is %llu\n",
1349 device->name, (unsigned long long)new_size);
1351 if (new_size > old_size) {
1352 trans = btrfs_start_transaction(root, 0);
1353 if (IS_ERR(trans)) {
1354 ret = PTR_ERR(trans);
1357 ret = btrfs_grow_device(trans, device, new_size);
1358 btrfs_commit_transaction(trans, root);
1359 } else if (new_size < old_size) {
1360 ret = btrfs_shrink_device(device, new_size);
1366 mutex_unlock(&root->fs_info->volume_mutex);
1370 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1377 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
1378 struct file *src_file;
1382 if (root->fs_info->sb->s_flags & MS_RDONLY)
1385 namelen = strlen(name);
1386 if (strchr(name, '/')) {
1391 if (name[0] == '.' &&
1392 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1398 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1399 NULL, transid, readonly);
1401 struct inode *src_inode;
1402 src_file = fget(fd);
1408 src_inode = src_file->f_path.dentry->d_inode;
1409 if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
1410 printk(KERN_INFO "btrfs: Snapshot src from "
1416 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1417 BTRFS_I(src_inode)->root,
1425 static noinline int btrfs_ioctl_snap_create(struct file *file,
1426 void __user *arg, int subvol)
1428 struct btrfs_ioctl_vol_args *vol_args;
1431 vol_args = memdup_user(arg, sizeof(*vol_args));
1432 if (IS_ERR(vol_args))
1433 return PTR_ERR(vol_args);
1434 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1436 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1437 vol_args->fd, subvol,
1444 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1445 void __user *arg, int subvol)
1447 struct btrfs_ioctl_vol_args_v2 *vol_args;
1451 bool readonly = false;
1453 vol_args = memdup_user(arg, sizeof(*vol_args));
1454 if (IS_ERR(vol_args))
1455 return PTR_ERR(vol_args);
1456 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1458 if (vol_args->flags &
1459 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY)) {
1464 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1466 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1469 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1470 vol_args->fd, subvol,
1473 if (ret == 0 && ptr &&
1475 offsetof(struct btrfs_ioctl_vol_args_v2,
1476 transid), ptr, sizeof(*ptr)))
1483 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1486 struct inode *inode = fdentry(file)->d_inode;
1487 struct btrfs_root *root = BTRFS_I(inode)->root;
1491 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1494 down_read(&root->fs_info->subvol_sem);
1495 if (btrfs_root_readonly(root))
1496 flags |= BTRFS_SUBVOL_RDONLY;
1497 up_read(&root->fs_info->subvol_sem);
1499 if (copy_to_user(arg, &flags, sizeof(flags)))
1505 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1508 struct inode *inode = fdentry(file)->d_inode;
1509 struct btrfs_root *root = BTRFS_I(inode)->root;
1510 struct btrfs_trans_handle *trans;
1515 if (root->fs_info->sb->s_flags & MS_RDONLY)
1518 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1521 if (copy_from_user(&flags, arg, sizeof(flags)))
1524 if (flags & BTRFS_SUBVOL_CREATE_ASYNC)
1527 if (flags & ~BTRFS_SUBVOL_RDONLY)
1530 if (!inode_owner_or_capable(inode))
1533 down_write(&root->fs_info->subvol_sem);
1536 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1539 root_flags = btrfs_root_flags(&root->root_item);
1540 if (flags & BTRFS_SUBVOL_RDONLY)
1541 btrfs_set_root_flags(&root->root_item,
1542 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1544 btrfs_set_root_flags(&root->root_item,
1545 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1547 trans = btrfs_start_transaction(root, 1);
1548 if (IS_ERR(trans)) {
1549 ret = PTR_ERR(trans);
1553 ret = btrfs_update_root(trans, root->fs_info->tree_root,
1554 &root->root_key, &root->root_item);
1556 btrfs_commit_transaction(trans, root);
1559 btrfs_set_root_flags(&root->root_item, root_flags);
1561 up_write(&root->fs_info->subvol_sem);
1566 * helper to check if the subvolume references other subvolumes
1568 static noinline int may_destroy_subvol(struct btrfs_root *root)
1570 struct btrfs_path *path;
1571 struct btrfs_key key;
1574 path = btrfs_alloc_path();
1578 key.objectid = root->root_key.objectid;
1579 key.type = BTRFS_ROOT_REF_KEY;
1580 key.offset = (u64)-1;
1582 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1589 if (path->slots[0] > 0) {
1591 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1592 if (key.objectid == root->root_key.objectid &&
1593 key.type == BTRFS_ROOT_REF_KEY)
1597 btrfs_free_path(path);
1601 static noinline int key_in_sk(struct btrfs_key *key,
1602 struct btrfs_ioctl_search_key *sk)
1604 struct btrfs_key test;
1607 test.objectid = sk->min_objectid;
1608 test.type = sk->min_type;
1609 test.offset = sk->min_offset;
1611 ret = btrfs_comp_cpu_keys(key, &test);
1615 test.objectid = sk->max_objectid;
1616 test.type = sk->max_type;
1617 test.offset = sk->max_offset;
1619 ret = btrfs_comp_cpu_keys(key, &test);
1625 static noinline int copy_to_sk(struct btrfs_root *root,
1626 struct btrfs_path *path,
1627 struct btrfs_key *key,
1628 struct btrfs_ioctl_search_key *sk,
1630 unsigned long *sk_offset,
1634 struct extent_buffer *leaf;
1635 struct btrfs_ioctl_search_header sh;
1636 unsigned long item_off;
1637 unsigned long item_len;
1643 leaf = path->nodes[0];
1644 slot = path->slots[0];
1645 nritems = btrfs_header_nritems(leaf);
1647 if (btrfs_header_generation(leaf) > sk->max_transid) {
1651 found_transid = btrfs_header_generation(leaf);
1653 for (i = slot; i < nritems; i++) {
1654 item_off = btrfs_item_ptr_offset(leaf, i);
1655 item_len = btrfs_item_size_nr(leaf, i);
1657 if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1660 if (sizeof(sh) + item_len + *sk_offset >
1661 BTRFS_SEARCH_ARGS_BUFSIZE) {
1666 btrfs_item_key_to_cpu(leaf, key, i);
1667 if (!key_in_sk(key, sk))
1670 sh.objectid = key->objectid;
1671 sh.offset = key->offset;
1672 sh.type = key->type;
1674 sh.transid = found_transid;
1676 /* copy search result header */
1677 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1678 *sk_offset += sizeof(sh);
1681 char *p = buf + *sk_offset;
1683 read_extent_buffer(leaf, p,
1684 item_off, item_len);
1685 *sk_offset += item_len;
1689 if (*num_found >= sk->nr_items)
1694 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1696 else if (key->type < (u8)-1 && key->type < sk->max_type) {
1699 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1709 static noinline int search_ioctl(struct inode *inode,
1710 struct btrfs_ioctl_search_args *args)
1712 struct btrfs_root *root;
1713 struct btrfs_key key;
1714 struct btrfs_key max_key;
1715 struct btrfs_path *path;
1716 struct btrfs_ioctl_search_key *sk = &args->key;
1717 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1720 unsigned long sk_offset = 0;
1722 path = btrfs_alloc_path();
1726 if (sk->tree_id == 0) {
1727 /* search the root of the inode that was passed */
1728 root = BTRFS_I(inode)->root;
1730 key.objectid = sk->tree_id;
1731 key.type = BTRFS_ROOT_ITEM_KEY;
1732 key.offset = (u64)-1;
1733 root = btrfs_read_fs_root_no_name(info, &key);
1735 printk(KERN_ERR "could not find root %llu\n",
1737 btrfs_free_path(path);
1742 key.objectid = sk->min_objectid;
1743 key.type = sk->min_type;
1744 key.offset = sk->min_offset;
1746 max_key.objectid = sk->max_objectid;
1747 max_key.type = sk->max_type;
1748 max_key.offset = sk->max_offset;
1750 path->keep_locks = 1;
1753 ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1760 ret = copy_to_sk(root, path, &key, sk, args->buf,
1761 &sk_offset, &num_found);
1762 btrfs_release_path(path);
1763 if (ret || num_found >= sk->nr_items)
1769 sk->nr_items = num_found;
1770 btrfs_free_path(path);
1774 static noinline int btrfs_ioctl_tree_search(struct file *file,
1777 struct btrfs_ioctl_search_args *args;
1778 struct inode *inode;
1781 if (!capable(CAP_SYS_ADMIN))
1784 args = memdup_user(argp, sizeof(*args));
1786 return PTR_ERR(args);
1788 inode = fdentry(file)->d_inode;
1789 ret = search_ioctl(inode, args);
1790 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1797 * Search INODE_REFs to identify path name of 'dirid' directory
1798 * in a 'tree_id' tree. and sets path name to 'name'.
1800 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1801 u64 tree_id, u64 dirid, char *name)
1803 struct btrfs_root *root;
1804 struct btrfs_key key;
1810 struct btrfs_inode_ref *iref;
1811 struct extent_buffer *l;
1812 struct btrfs_path *path;
1814 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1819 path = btrfs_alloc_path();
1823 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1825 key.objectid = tree_id;
1826 key.type = BTRFS_ROOT_ITEM_KEY;
1827 key.offset = (u64)-1;
1828 root = btrfs_read_fs_root_no_name(info, &key);
1830 printk(KERN_ERR "could not find root %llu\n", tree_id);
1835 key.objectid = dirid;
1836 key.type = BTRFS_INODE_REF_KEY;
1837 key.offset = (u64)-1;
1840 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1845 slot = path->slots[0];
1846 if (ret > 0 && slot > 0)
1848 btrfs_item_key_to_cpu(l, &key, slot);
1850 if (ret > 0 && (key.objectid != dirid ||
1851 key.type != BTRFS_INODE_REF_KEY)) {
1856 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1857 len = btrfs_inode_ref_name_len(l, iref);
1859 total_len += len + 1;
1864 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1866 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1869 btrfs_release_path(path);
1870 key.objectid = key.offset;
1871 key.offset = (u64)-1;
1872 dirid = key.objectid;
1876 memmove(name, ptr, total_len);
1877 name[total_len]='\0';
1880 btrfs_free_path(path);
1884 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1887 struct btrfs_ioctl_ino_lookup_args *args;
1888 struct inode *inode;
1891 if (!capable(CAP_SYS_ADMIN))
1894 args = memdup_user(argp, sizeof(*args));
1896 return PTR_ERR(args);
1898 inode = fdentry(file)->d_inode;
1900 if (args->treeid == 0)
1901 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1903 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1904 args->treeid, args->objectid,
1907 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1914 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1917 struct dentry *parent = fdentry(file);
1918 struct dentry *dentry;
1919 struct inode *dir = parent->d_inode;
1920 struct inode *inode;
1921 struct btrfs_root *root = BTRFS_I(dir)->root;
1922 struct btrfs_root *dest = NULL;
1923 struct btrfs_ioctl_vol_args *vol_args;
1924 struct btrfs_trans_handle *trans;
1929 vol_args = memdup_user(arg, sizeof(*vol_args));
1930 if (IS_ERR(vol_args))
1931 return PTR_ERR(vol_args);
1933 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1934 namelen = strlen(vol_args->name);
1935 if (strchr(vol_args->name, '/') ||
1936 strncmp(vol_args->name, "..", namelen) == 0) {
1941 err = mnt_want_write_file(file);
1945 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
1946 dentry = lookup_one_len(vol_args->name, parent, namelen);
1947 if (IS_ERR(dentry)) {
1948 err = PTR_ERR(dentry);
1949 goto out_unlock_dir;
1952 if (!dentry->d_inode) {
1957 inode = dentry->d_inode;
1958 dest = BTRFS_I(inode)->root;
1959 if (!capable(CAP_SYS_ADMIN)){
1961 * Regular user. Only allow this with a special mount
1962 * option, when the user has write+exec access to the
1963 * subvol root, and when rmdir(2) would have been
1966 * Note that this is _not_ check that the subvol is
1967 * empty or doesn't contain data that we wouldn't
1968 * otherwise be able to delete.
1970 * Users who want to delete empty subvols should try
1974 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1978 * Do not allow deletion if the parent dir is the same
1979 * as the dir to be deleted. That means the ioctl
1980 * must be called on the dentry referencing the root
1981 * of the subvol, not a random directory contained
1988 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
1992 /* check if subvolume may be deleted by a non-root user */
1993 err = btrfs_may_delete(dir, dentry, 1);
1998 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2003 mutex_lock(&inode->i_mutex);
2004 err = d_invalidate(dentry);
2008 down_write(&root->fs_info->subvol_sem);
2010 err = may_destroy_subvol(dest);
2014 trans = btrfs_start_transaction(root, 0);
2015 if (IS_ERR(trans)) {
2016 err = PTR_ERR(trans);
2019 trans->block_rsv = &root->fs_info->global_block_rsv;
2021 ret = btrfs_unlink_subvol(trans, root, dir,
2022 dest->root_key.objectid,
2023 dentry->d_name.name,
2024 dentry->d_name.len);
2027 btrfs_abort_transaction(trans, root, ret);
2031 btrfs_record_root_in_trans(trans, dest);
2033 memset(&dest->root_item.drop_progress, 0,
2034 sizeof(dest->root_item.drop_progress));
2035 dest->root_item.drop_level = 0;
2036 btrfs_set_root_refs(&dest->root_item, 0);
2038 if (!xchg(&dest->orphan_item_inserted, 1)) {
2039 ret = btrfs_insert_orphan_item(trans,
2040 root->fs_info->tree_root,
2041 dest->root_key.objectid);
2043 btrfs_abort_transaction(trans, root, ret);
2049 ret = btrfs_end_transaction(trans, root);
2052 inode->i_flags |= S_DEAD;
2054 up_write(&root->fs_info->subvol_sem);
2056 mutex_unlock(&inode->i_mutex);
2058 shrink_dcache_sb(root->fs_info->sb);
2059 btrfs_invalidate_inodes(dest);
2065 mutex_unlock(&dir->i_mutex);
2066 mnt_drop_write_file(file);
2072 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2074 struct inode *inode = fdentry(file)->d_inode;
2075 struct btrfs_root *root = BTRFS_I(inode)->root;
2076 struct btrfs_ioctl_defrag_range_args *range;
2079 if (btrfs_root_readonly(root))
2082 ret = mnt_want_write_file(file);
2086 switch (inode->i_mode & S_IFMT) {
2088 if (!capable(CAP_SYS_ADMIN)) {
2092 ret = btrfs_defrag_root(root, 0);
2095 ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
2098 if (!(file->f_mode & FMODE_WRITE)) {
2103 range = kzalloc(sizeof(*range), GFP_KERNEL);
2110 if (copy_from_user(range, argp,
2116 /* compression requires us to start the IO */
2117 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2118 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2119 range->extent_thresh = (u32)-1;
2122 /* the rest are all set to zero by kzalloc */
2123 range->len = (u64)-1;
2125 ret = btrfs_defrag_file(fdentry(file)->d_inode, file,
2135 mnt_drop_write_file(file);
2139 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2141 struct btrfs_ioctl_vol_args *vol_args;
2144 if (!capable(CAP_SYS_ADMIN))
2147 mutex_lock(&root->fs_info->volume_mutex);
2148 if (root->fs_info->balance_ctl) {
2149 printk(KERN_INFO "btrfs: balance in progress\n");
2154 vol_args = memdup_user(arg, sizeof(*vol_args));
2155 if (IS_ERR(vol_args)) {
2156 ret = PTR_ERR(vol_args);
2160 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2161 ret = btrfs_init_new_device(root, vol_args->name);
2165 mutex_unlock(&root->fs_info->volume_mutex);
2169 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
2171 struct btrfs_ioctl_vol_args *vol_args;
2174 if (!capable(CAP_SYS_ADMIN))
2177 if (root->fs_info->sb->s_flags & MS_RDONLY)
2180 mutex_lock(&root->fs_info->volume_mutex);
2181 if (root->fs_info->balance_ctl) {
2182 printk(KERN_INFO "btrfs: balance in progress\n");
2187 vol_args = memdup_user(arg, sizeof(*vol_args));
2188 if (IS_ERR(vol_args)) {
2189 ret = PTR_ERR(vol_args);
2193 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2194 ret = btrfs_rm_device(root, vol_args->name);
2198 mutex_unlock(&root->fs_info->volume_mutex);
2202 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2204 struct btrfs_ioctl_fs_info_args *fi_args;
2205 struct btrfs_device *device;
2206 struct btrfs_device *next;
2207 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2210 if (!capable(CAP_SYS_ADMIN))
2213 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2217 fi_args->num_devices = fs_devices->num_devices;
2218 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2220 mutex_lock(&fs_devices->device_list_mutex);
2221 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2222 if (device->devid > fi_args->max_id)
2223 fi_args->max_id = device->devid;
2225 mutex_unlock(&fs_devices->device_list_mutex);
2227 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2234 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2236 struct btrfs_ioctl_dev_info_args *di_args;
2237 struct btrfs_device *dev;
2238 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2240 char *s_uuid = NULL;
2241 char empty_uuid[BTRFS_UUID_SIZE] = {0};
2243 if (!capable(CAP_SYS_ADMIN))
2246 di_args = memdup_user(arg, sizeof(*di_args));
2247 if (IS_ERR(di_args))
2248 return PTR_ERR(di_args);
2250 if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2251 s_uuid = di_args->uuid;
2253 mutex_lock(&fs_devices->device_list_mutex);
2254 dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL);
2255 mutex_unlock(&fs_devices->device_list_mutex);
2262 di_args->devid = dev->devid;
2263 di_args->bytes_used = dev->bytes_used;
2264 di_args->total_bytes = dev->total_bytes;
2265 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2267 strncpy(di_args->path, dev->name, sizeof(di_args->path));
2269 di_args->path[0] = '\0';
2272 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2279 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2280 u64 off, u64 olen, u64 destoff)
2282 struct inode *inode = fdentry(file)->d_inode;
2283 struct btrfs_root *root = BTRFS_I(inode)->root;
2284 struct file *src_file;
2286 struct btrfs_trans_handle *trans;
2287 struct btrfs_path *path;
2288 struct extent_buffer *leaf;
2290 struct btrfs_key key;
2295 u64 bs = root->fs_info->sb->s_blocksize;
2300 * - split compressed inline extents. annoying: we need to
2301 * decompress into destination's address_space (the file offset
2302 * may change, so source mapping won't do), then recompress (or
2303 * otherwise reinsert) a subrange.
2304 * - allow ranges within the same file to be cloned (provided
2305 * they don't overlap)?
2308 /* the destination must be opened for writing */
2309 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
2312 if (btrfs_root_readonly(root))
2315 ret = mnt_want_write_file(file);
2319 src_file = fget(srcfd);
2322 goto out_drop_write;
2325 src = src_file->f_dentry->d_inode;
2331 /* the src must be open for reading */
2332 if (!(src_file->f_mode & FMODE_READ))
2335 /* don't make the dst file partly checksummed */
2336 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2337 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2341 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2345 if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
2349 buf = vmalloc(btrfs_level_size(root, 0));
2353 path = btrfs_alloc_path();
2361 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2362 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
2364 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2365 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
2368 /* determine range to clone */
2370 if (off + len > src->i_size || off + len < off)
2373 olen = len = src->i_size - off;
2374 /* if we extend to eof, continue to block boundary */
2375 if (off + len == src->i_size)
2376 len = ALIGN(src->i_size, bs) - off;
2378 /* verify the end result is block aligned */
2379 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2380 !IS_ALIGNED(destoff, bs))
2383 if (destoff > inode->i_size) {
2384 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2389 /* truncate page cache pages from target inode range */
2390 truncate_inode_pages_range(&inode->i_data, destoff,
2391 PAGE_CACHE_ALIGN(destoff + len) - 1);
2393 /* do any pending delalloc/csum calc on src, one way or
2394 another, and lock file content */
2396 struct btrfs_ordered_extent *ordered;
2397 lock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2398 ordered = btrfs_lookup_first_ordered_extent(src, off+len);
2400 !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len,
2401 EXTENT_DELALLOC, 0, NULL))
2403 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2405 btrfs_put_ordered_extent(ordered);
2406 btrfs_wait_ordered_range(src, off, len);
2410 key.objectid = btrfs_ino(src);
2411 key.type = BTRFS_EXTENT_DATA_KEY;
2416 * note the key will change type as we walk through the
2419 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2423 nritems = btrfs_header_nritems(path->nodes[0]);
2424 if (path->slots[0] >= nritems) {
2425 ret = btrfs_next_leaf(root, path);
2430 nritems = btrfs_header_nritems(path->nodes[0]);
2432 leaf = path->nodes[0];
2433 slot = path->slots[0];
2435 btrfs_item_key_to_cpu(leaf, &key, slot);
2436 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
2437 key.objectid != btrfs_ino(src))
2440 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2441 struct btrfs_file_extent_item *extent;
2444 struct btrfs_key new_key;
2445 u64 disko = 0, diskl = 0;
2446 u64 datao = 0, datal = 0;
2450 size = btrfs_item_size_nr(leaf, slot);
2451 read_extent_buffer(leaf, buf,
2452 btrfs_item_ptr_offset(leaf, slot),
2455 extent = btrfs_item_ptr(leaf, slot,
2456 struct btrfs_file_extent_item);
2457 comp = btrfs_file_extent_compression(leaf, extent);
2458 type = btrfs_file_extent_type(leaf, extent);
2459 if (type == BTRFS_FILE_EXTENT_REG ||
2460 type == BTRFS_FILE_EXTENT_PREALLOC) {
2461 disko = btrfs_file_extent_disk_bytenr(leaf,
2463 diskl = btrfs_file_extent_disk_num_bytes(leaf,
2465 datao = btrfs_file_extent_offset(leaf, extent);
2466 datal = btrfs_file_extent_num_bytes(leaf,
2468 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2469 /* take upper bound, may be compressed */
2470 datal = btrfs_file_extent_ram_bytes(leaf,
2473 btrfs_release_path(path);
2475 if (key.offset + datal <= off ||
2476 key.offset >= off+len)
2479 memcpy(&new_key, &key, sizeof(new_key));
2480 new_key.objectid = btrfs_ino(inode);
2481 if (off <= key.offset)
2482 new_key.offset = key.offset + destoff - off;
2484 new_key.offset = destoff;
2487 * 1 - adjusting old extent (we may have to split it)
2488 * 1 - add new extent
2491 trans = btrfs_start_transaction(root, 3);
2492 if (IS_ERR(trans)) {
2493 ret = PTR_ERR(trans);
2497 if (type == BTRFS_FILE_EXTENT_REG ||
2498 type == BTRFS_FILE_EXTENT_PREALLOC) {
2500 * a | --- range to clone ---| b
2501 * | ------------- extent ------------- |
2504 /* substract range b */
2505 if (key.offset + datal > off + len)
2506 datal = off + len - key.offset;
2508 /* substract range a */
2509 if (off > key.offset) {
2510 datao += off - key.offset;
2511 datal -= off - key.offset;
2514 ret = btrfs_drop_extents(trans, inode,
2516 new_key.offset + datal,
2519 btrfs_abort_transaction(trans, root,
2521 btrfs_end_transaction(trans, root);
2525 ret = btrfs_insert_empty_item(trans, root, path,
2528 btrfs_abort_transaction(trans, root,
2530 btrfs_end_transaction(trans, root);
2534 leaf = path->nodes[0];
2535 slot = path->slots[0];
2536 write_extent_buffer(leaf, buf,
2537 btrfs_item_ptr_offset(leaf, slot),
2540 extent = btrfs_item_ptr(leaf, slot,
2541 struct btrfs_file_extent_item);
2543 /* disko == 0 means it's a hole */
2547 btrfs_set_file_extent_offset(leaf, extent,
2549 btrfs_set_file_extent_num_bytes(leaf, extent,
2552 inode_add_bytes(inode, datal);
2553 ret = btrfs_inc_extent_ref(trans, root,
2555 root->root_key.objectid,
2557 new_key.offset - datao,
2560 btrfs_abort_transaction(trans,
2563 btrfs_end_transaction(trans,
2569 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2572 if (off > key.offset) {
2573 skip = off - key.offset;
2574 new_key.offset += skip;
2577 if (key.offset + datal > off+len)
2578 trim = key.offset + datal - (off+len);
2580 if (comp && (skip || trim)) {
2582 btrfs_end_transaction(trans, root);
2585 size -= skip + trim;
2586 datal -= skip + trim;
2588 ret = btrfs_drop_extents(trans, inode,
2590 new_key.offset + datal,
2593 btrfs_abort_transaction(trans, root,
2595 btrfs_end_transaction(trans, root);
2599 ret = btrfs_insert_empty_item(trans, root, path,
2602 btrfs_abort_transaction(trans, root,
2604 btrfs_end_transaction(trans, root);
2610 btrfs_file_extent_calc_inline_size(0);
2611 memmove(buf+start, buf+start+skip,
2615 leaf = path->nodes[0];
2616 slot = path->slots[0];
2617 write_extent_buffer(leaf, buf,
2618 btrfs_item_ptr_offset(leaf, slot),
2620 inode_add_bytes(inode, datal);
2623 btrfs_mark_buffer_dirty(leaf);
2624 btrfs_release_path(path);
2626 inode_inc_iversion(inode);
2627 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2630 * we round up to the block size at eof when
2631 * determining which extents to clone above,
2632 * but shouldn't round up the file size
2634 endoff = new_key.offset + datal;
2635 if (endoff > destoff+olen)
2636 endoff = destoff+olen;
2637 if (endoff > inode->i_size)
2638 btrfs_i_size_write(inode, endoff);
2640 ret = btrfs_update_inode(trans, root, inode);
2642 btrfs_abort_transaction(trans, root, ret);
2643 btrfs_end_transaction(trans, root);
2646 ret = btrfs_end_transaction(trans, root);
2649 btrfs_release_path(path);
2654 btrfs_release_path(path);
2655 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2657 mutex_unlock(&src->i_mutex);
2658 mutex_unlock(&inode->i_mutex);
2660 btrfs_free_path(path);
2664 mnt_drop_write_file(file);
2668 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
2670 struct btrfs_ioctl_clone_range_args args;
2672 if (copy_from_user(&args, argp, sizeof(args)))
2674 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2675 args.src_length, args.dest_offset);
2679 * there are many ways the trans_start and trans_end ioctls can lead
2680 * to deadlocks. They should only be used by applications that
2681 * basically own the machine, and have a very in depth understanding
2682 * of all the possible deadlocks and enospc problems.
2684 static long btrfs_ioctl_trans_start(struct file *file)
2686 struct inode *inode = fdentry(file)->d_inode;
2687 struct btrfs_root *root = BTRFS_I(inode)->root;
2688 struct btrfs_trans_handle *trans;
2692 if (!capable(CAP_SYS_ADMIN))
2696 if (file->private_data)
2700 if (btrfs_root_readonly(root))
2703 ret = mnt_want_write_file(file);
2707 atomic_inc(&root->fs_info->open_ioctl_trans);
2710 trans = btrfs_start_ioctl_transaction(root);
2714 file->private_data = trans;
2718 atomic_dec(&root->fs_info->open_ioctl_trans);
2719 mnt_drop_write_file(file);
2724 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2726 struct inode *inode = fdentry(file)->d_inode;
2727 struct btrfs_root *root = BTRFS_I(inode)->root;
2728 struct btrfs_root *new_root;
2729 struct btrfs_dir_item *di;
2730 struct btrfs_trans_handle *trans;
2731 struct btrfs_path *path;
2732 struct btrfs_key location;
2733 struct btrfs_disk_key disk_key;
2734 struct btrfs_super_block *disk_super;
2739 if (!capable(CAP_SYS_ADMIN))
2742 if (copy_from_user(&objectid, argp, sizeof(objectid)))
2746 objectid = root->root_key.objectid;
2748 location.objectid = objectid;
2749 location.type = BTRFS_ROOT_ITEM_KEY;
2750 location.offset = (u64)-1;
2752 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2753 if (IS_ERR(new_root))
2754 return PTR_ERR(new_root);
2756 if (btrfs_root_refs(&new_root->root_item) == 0)
2759 path = btrfs_alloc_path();
2762 path->leave_spinning = 1;
2764 trans = btrfs_start_transaction(root, 1);
2765 if (IS_ERR(trans)) {
2766 btrfs_free_path(path);
2767 return PTR_ERR(trans);
2770 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
2771 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2772 dir_id, "default", 7, 1);
2773 if (IS_ERR_OR_NULL(di)) {
2774 btrfs_free_path(path);
2775 btrfs_end_transaction(trans, root);
2776 printk(KERN_ERR "Umm, you don't have the default dir item, "
2777 "this isn't going to work\n");
2781 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2782 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2783 btrfs_mark_buffer_dirty(path->nodes[0]);
2784 btrfs_free_path(path);
2786 disk_super = root->fs_info->super_copy;
2787 features = btrfs_super_incompat_flags(disk_super);
2788 if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) {
2789 features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL;
2790 btrfs_set_super_incompat_flags(disk_super, features);
2792 btrfs_end_transaction(trans, root);
2797 static void get_block_group_info(struct list_head *groups_list,
2798 struct btrfs_ioctl_space_info *space)
2800 struct btrfs_block_group_cache *block_group;
2802 space->total_bytes = 0;
2803 space->used_bytes = 0;
2805 list_for_each_entry(block_group, groups_list, list) {
2806 space->flags = block_group->flags;
2807 space->total_bytes += block_group->key.offset;
2808 space->used_bytes +=
2809 btrfs_block_group_used(&block_group->item);
2813 long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2815 struct btrfs_ioctl_space_args space_args;
2816 struct btrfs_ioctl_space_info space;
2817 struct btrfs_ioctl_space_info *dest;
2818 struct btrfs_ioctl_space_info *dest_orig;
2819 struct btrfs_ioctl_space_info __user *user_dest;
2820 struct btrfs_space_info *info;
2821 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2822 BTRFS_BLOCK_GROUP_SYSTEM,
2823 BTRFS_BLOCK_GROUP_METADATA,
2824 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2831 if (copy_from_user(&space_args,
2832 (struct btrfs_ioctl_space_args __user *)arg,
2833 sizeof(space_args)))
2836 for (i = 0; i < num_types; i++) {
2837 struct btrfs_space_info *tmp;
2841 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2843 if (tmp->flags == types[i]) {
2853 down_read(&info->groups_sem);
2854 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2855 if (!list_empty(&info->block_groups[c]))
2858 up_read(&info->groups_sem);
2861 /* space_slots == 0 means they are asking for a count */
2862 if (space_args.space_slots == 0) {
2863 space_args.total_spaces = slot_count;
2867 slot_count = min_t(u64, space_args.space_slots, slot_count);
2869 alloc_size = sizeof(*dest) * slot_count;
2871 /* we generally have at most 6 or so space infos, one for each raid
2872 * level. So, a whole page should be more than enough for everyone
2874 if (alloc_size > PAGE_CACHE_SIZE)
2877 space_args.total_spaces = 0;
2878 dest = kmalloc(alloc_size, GFP_NOFS);
2883 /* now we have a buffer to copy into */
2884 for (i = 0; i < num_types; i++) {
2885 struct btrfs_space_info *tmp;
2892 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2894 if (tmp->flags == types[i]) {
2903 down_read(&info->groups_sem);
2904 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2905 if (!list_empty(&info->block_groups[c])) {
2906 get_block_group_info(&info->block_groups[c],
2908 memcpy(dest, &space, sizeof(space));
2910 space_args.total_spaces++;
2916 up_read(&info->groups_sem);
2919 user_dest = (struct btrfs_ioctl_space_info *)
2920 (arg + sizeof(struct btrfs_ioctl_space_args));
2922 if (copy_to_user(user_dest, dest_orig, alloc_size))
2927 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
2934 * there are many ways the trans_start and trans_end ioctls can lead
2935 * to deadlocks. They should only be used by applications that
2936 * basically own the machine, and have a very in depth understanding
2937 * of all the possible deadlocks and enospc problems.
2939 long btrfs_ioctl_trans_end(struct file *file)
2941 struct inode *inode = fdentry(file)->d_inode;
2942 struct btrfs_root *root = BTRFS_I(inode)->root;
2943 struct btrfs_trans_handle *trans;
2945 trans = file->private_data;
2948 file->private_data = NULL;
2950 btrfs_end_transaction(trans, root);
2952 atomic_dec(&root->fs_info->open_ioctl_trans);
2954 mnt_drop_write_file(file);
2958 static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
2960 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2961 struct btrfs_trans_handle *trans;
2965 trans = btrfs_start_transaction(root, 0);
2967 return PTR_ERR(trans);
2968 transid = trans->transid;
2969 ret = btrfs_commit_transaction_async(trans, root, 0);
2971 btrfs_end_transaction(trans, root);
2976 if (copy_to_user(argp, &transid, sizeof(transid)))
2981 static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
2983 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2987 if (copy_from_user(&transid, argp, sizeof(transid)))
2990 transid = 0; /* current trans */
2992 return btrfs_wait_for_commit(root, transid);
2995 static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg)
2998 struct btrfs_ioctl_scrub_args *sa;
3000 if (!capable(CAP_SYS_ADMIN))
3003 sa = memdup_user(arg, sizeof(*sa));
3007 ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end,
3008 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY);
3010 if (copy_to_user(arg, sa, sizeof(*sa)))
3017 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3019 if (!capable(CAP_SYS_ADMIN))
3022 return btrfs_scrub_cancel(root);
3025 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3028 struct btrfs_ioctl_scrub_args *sa;
3031 if (!capable(CAP_SYS_ADMIN))
3034 sa = memdup_user(arg, sizeof(*sa));
3038 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3040 if (copy_to_user(arg, sa, sizeof(*sa)))
3047 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3053 struct btrfs_ioctl_ino_path_args *ipa = NULL;
3054 struct inode_fs_paths *ipath = NULL;
3055 struct btrfs_path *path;
3057 if (!capable(CAP_SYS_ADMIN))
3060 path = btrfs_alloc_path();
3066 ipa = memdup_user(arg, sizeof(*ipa));
3073 size = min_t(u32, ipa->size, 4096);
3074 ipath = init_ipath(size, root, path);
3075 if (IS_ERR(ipath)) {
3076 ret = PTR_ERR(ipath);
3081 ret = paths_from_inode(ipa->inum, ipath);
3085 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3086 rel_ptr = ipath->fspath->val[i] -
3087 (u64)(unsigned long)ipath->fspath->val;
3088 ipath->fspath->val[i] = rel_ptr;
3091 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3092 (void *)(unsigned long)ipath->fspath, size);
3099 btrfs_free_path(path);
3106 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3108 struct btrfs_data_container *inodes = ctx;
3109 const size_t c = 3 * sizeof(u64);
3111 if (inodes->bytes_left >= c) {
3112 inodes->bytes_left -= c;
3113 inodes->val[inodes->elem_cnt] = inum;
3114 inodes->val[inodes->elem_cnt + 1] = offset;
3115 inodes->val[inodes->elem_cnt + 2] = root;
3116 inodes->elem_cnt += 3;
3118 inodes->bytes_missing += c - inodes->bytes_left;
3119 inodes->bytes_left = 0;
3120 inodes->elem_missed += 3;
3126 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3131 u64 extent_item_pos;
3132 struct btrfs_ioctl_logical_ino_args *loi;
3133 struct btrfs_data_container *inodes = NULL;
3134 struct btrfs_path *path = NULL;
3135 struct btrfs_key key;
3137 if (!capable(CAP_SYS_ADMIN))
3140 loi = memdup_user(arg, sizeof(*loi));
3147 path = btrfs_alloc_path();
3153 size = min_t(u32, loi->size, 4096);
3154 inodes = init_data_container(size);
3155 if (IS_ERR(inodes)) {
3156 ret = PTR_ERR(inodes);
3161 ret = extent_from_logical(root->fs_info, loi->logical, path, &key);
3162 btrfs_release_path(path);
3164 if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK)
3169 extent_item_pos = loi->logical - key.objectid;
3170 ret = iterate_extent_inodes(root->fs_info, key.objectid,
3171 extent_item_pos, 0, build_ino_list,
3177 ret = copy_to_user((void *)(unsigned long)loi->inodes,
3178 (void *)(unsigned long)inodes, size);
3183 btrfs_free_path(path);
3190 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
3191 struct btrfs_ioctl_balance_args *bargs)
3193 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3195 bargs->flags = bctl->flags;
3197 if (atomic_read(&fs_info->balance_running))
3198 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3199 if (atomic_read(&fs_info->balance_pause_req))
3200 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3201 if (atomic_read(&fs_info->balance_cancel_req))
3202 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3204 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3205 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3206 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3209 spin_lock(&fs_info->balance_lock);
3210 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3211 spin_unlock(&fs_info->balance_lock);
3213 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3217 static long btrfs_ioctl_balance(struct btrfs_root *root, void __user *arg)
3219 struct btrfs_fs_info *fs_info = root->fs_info;
3220 struct btrfs_ioctl_balance_args *bargs;
3221 struct btrfs_balance_control *bctl;
3224 if (!capable(CAP_SYS_ADMIN))
3227 if (fs_info->sb->s_flags & MS_RDONLY)
3230 mutex_lock(&fs_info->volume_mutex);
3231 mutex_lock(&fs_info->balance_mutex);
3234 bargs = memdup_user(arg, sizeof(*bargs));
3235 if (IS_ERR(bargs)) {
3236 ret = PTR_ERR(bargs);
3240 if (bargs->flags & BTRFS_BALANCE_RESUME) {
3241 if (!fs_info->balance_ctl) {
3246 bctl = fs_info->balance_ctl;
3247 spin_lock(&fs_info->balance_lock);
3248 bctl->flags |= BTRFS_BALANCE_RESUME;
3249 spin_unlock(&fs_info->balance_lock);
3257 if (fs_info->balance_ctl) {
3262 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3268 bctl->fs_info = fs_info;
3270 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3271 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3272 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3274 bctl->flags = bargs->flags;
3276 /* balance everything - no filters */
3277 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
3281 ret = btrfs_balance(bctl, bargs);
3283 * bctl is freed in __cancel_balance or in free_fs_info if
3284 * restriper was paused all the way until unmount
3287 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3294 mutex_unlock(&fs_info->balance_mutex);
3295 mutex_unlock(&fs_info->volume_mutex);
3299 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
3301 if (!capable(CAP_SYS_ADMIN))
3305 case BTRFS_BALANCE_CTL_PAUSE:
3306 return btrfs_pause_balance(root->fs_info);
3307 case BTRFS_BALANCE_CTL_CANCEL:
3308 return btrfs_cancel_balance(root->fs_info);
3314 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
3317 struct btrfs_fs_info *fs_info = root->fs_info;
3318 struct btrfs_ioctl_balance_args *bargs;
3321 if (!capable(CAP_SYS_ADMIN))
3324 mutex_lock(&fs_info->balance_mutex);
3325 if (!fs_info->balance_ctl) {
3330 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
3336 update_ioctl_balance_args(fs_info, 1, bargs);
3338 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3343 mutex_unlock(&fs_info->balance_mutex);
3347 long btrfs_ioctl(struct file *file, unsigned int
3348 cmd, unsigned long arg)
3350 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3351 void __user *argp = (void __user *)arg;
3354 case FS_IOC_GETFLAGS:
3355 return btrfs_ioctl_getflags(file, argp);
3356 case FS_IOC_SETFLAGS:
3357 return btrfs_ioctl_setflags(file, argp);
3358 case FS_IOC_GETVERSION:
3359 return btrfs_ioctl_getversion(file, argp);
3361 return btrfs_ioctl_fitrim(file, argp);
3362 case BTRFS_IOC_SNAP_CREATE:
3363 return btrfs_ioctl_snap_create(file, argp, 0);
3364 case BTRFS_IOC_SNAP_CREATE_V2:
3365 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3366 case BTRFS_IOC_SUBVOL_CREATE:
3367 return btrfs_ioctl_snap_create(file, argp, 1);
3368 case BTRFS_IOC_SNAP_DESTROY:
3369 return btrfs_ioctl_snap_destroy(file, argp);
3370 case BTRFS_IOC_SUBVOL_GETFLAGS:
3371 return btrfs_ioctl_subvol_getflags(file, argp);
3372 case BTRFS_IOC_SUBVOL_SETFLAGS:
3373 return btrfs_ioctl_subvol_setflags(file, argp);
3374 case BTRFS_IOC_DEFAULT_SUBVOL:
3375 return btrfs_ioctl_default_subvol(file, argp);
3376 case BTRFS_IOC_DEFRAG:
3377 return btrfs_ioctl_defrag(file, NULL);
3378 case BTRFS_IOC_DEFRAG_RANGE:
3379 return btrfs_ioctl_defrag(file, argp);
3380 case BTRFS_IOC_RESIZE:
3381 return btrfs_ioctl_resize(root, argp);
3382 case BTRFS_IOC_ADD_DEV:
3383 return btrfs_ioctl_add_dev(root, argp);
3384 case BTRFS_IOC_RM_DEV:
3385 return btrfs_ioctl_rm_dev(root, argp);
3386 case BTRFS_IOC_FS_INFO:
3387 return btrfs_ioctl_fs_info(root, argp);
3388 case BTRFS_IOC_DEV_INFO:
3389 return btrfs_ioctl_dev_info(root, argp);
3390 case BTRFS_IOC_BALANCE:
3391 return btrfs_ioctl_balance(root, NULL);
3392 case BTRFS_IOC_CLONE:
3393 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
3394 case BTRFS_IOC_CLONE_RANGE:
3395 return btrfs_ioctl_clone_range(file, argp);
3396 case BTRFS_IOC_TRANS_START:
3397 return btrfs_ioctl_trans_start(file);
3398 case BTRFS_IOC_TRANS_END:
3399 return btrfs_ioctl_trans_end(file);
3400 case BTRFS_IOC_TREE_SEARCH:
3401 return btrfs_ioctl_tree_search(file, argp);
3402 case BTRFS_IOC_INO_LOOKUP:
3403 return btrfs_ioctl_ino_lookup(file, argp);
3404 case BTRFS_IOC_INO_PATHS:
3405 return btrfs_ioctl_ino_to_path(root, argp);
3406 case BTRFS_IOC_LOGICAL_INO:
3407 return btrfs_ioctl_logical_to_ino(root, argp);
3408 case BTRFS_IOC_SPACE_INFO:
3409 return btrfs_ioctl_space_info(root, argp);
3410 case BTRFS_IOC_SYNC:
3411 btrfs_sync_fs(file->f_dentry->d_sb, 1);
3413 case BTRFS_IOC_START_SYNC:
3414 return btrfs_ioctl_start_sync(file, argp);
3415 case BTRFS_IOC_WAIT_SYNC:
3416 return btrfs_ioctl_wait_sync(file, argp);
3417 case BTRFS_IOC_SCRUB:
3418 return btrfs_ioctl_scrub(root, argp);
3419 case BTRFS_IOC_SCRUB_CANCEL:
3420 return btrfs_ioctl_scrub_cancel(root, argp);
3421 case BTRFS_IOC_SCRUB_PROGRESS:
3422 return btrfs_ioctl_scrub_progress(root, argp);
3423 case BTRFS_IOC_BALANCE_V2:
3424 return btrfs_ioctl_balance(root, argp);
3425 case BTRFS_IOC_BALANCE_CTL:
3426 return btrfs_ioctl_balance_ctl(root, arg);
3427 case BTRFS_IOC_BALANCE_PROGRESS:
3428 return btrfs_ioctl_balance_progress(root, argp);