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->f_path.mnt);
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->i_ctime = CURRENT_TIME;
265 ret = btrfs_update_inode(trans, root, inode);
267 btrfs_end_transaction(trans, root);
270 ip->flags = ip_oldflags;
271 inode->i_flags = i_oldflags;
274 mnt_drop_write(file->f_path.mnt);
276 mutex_unlock(&inode->i_mutex);
280 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
282 struct inode *inode = file->f_path.dentry->d_inode;
284 return put_user(inode->i_generation, arg);
287 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
289 struct btrfs_root *root = fdentry(file)->d_sb->s_fs_info;
290 struct btrfs_fs_info *fs_info = root->fs_info;
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(root->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(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 BUG_ON(IS_ERR(new_root));
431 btrfs_record_root_in_trans(trans, new_root);
433 ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
435 * insert the directory item
437 ret = btrfs_set_inode_index(dir, &index);
440 ret = btrfs_insert_dir_item(trans, root,
441 name, namelen, dir, &key,
442 BTRFS_FT_DIR, index);
446 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
447 ret = btrfs_update_inode(trans, root, dir);
450 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
451 objectid, root->root_key.objectid,
452 btrfs_ino(dir), index, name, namelen);
456 d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
459 *async_transid = trans->transid;
460 err = btrfs_commit_transaction_async(trans, root, 1);
462 err = btrfs_commit_transaction(trans, root);
469 static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
470 char *name, int namelen, u64 *async_transid,
474 struct btrfs_pending_snapshot *pending_snapshot;
475 struct btrfs_trans_handle *trans;
481 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
482 if (!pending_snapshot)
485 btrfs_init_block_rsv(&pending_snapshot->block_rsv);
486 pending_snapshot->dentry = dentry;
487 pending_snapshot->root = root;
488 pending_snapshot->readonly = readonly;
490 trans = btrfs_start_transaction(root->fs_info->extent_root, 5);
492 ret = PTR_ERR(trans);
496 ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
499 spin_lock(&root->fs_info->trans_lock);
500 list_add(&pending_snapshot->list,
501 &trans->transaction->pending_snapshots);
502 spin_unlock(&root->fs_info->trans_lock);
504 *async_transid = trans->transid;
505 ret = btrfs_commit_transaction_async(trans,
506 root->fs_info->extent_root, 1);
508 ret = btrfs_commit_transaction(trans,
509 root->fs_info->extent_root);
513 ret = pending_snapshot->error;
517 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
521 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
523 ret = PTR_ERR(inode);
527 d_instantiate(dentry, inode);
530 kfree(pending_snapshot);
534 /* copy of check_sticky in fs/namei.c()
535 * It's inline, so penalty for filesystems that don't use sticky bit is
538 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
540 uid_t fsuid = current_fsuid();
542 if (!(dir->i_mode & S_ISVTX))
544 if (inode->i_uid == fsuid)
546 if (dir->i_uid == fsuid)
548 return !capable(CAP_FOWNER);
551 /* copy of may_delete in fs/namei.c()
552 * Check whether we can remove a link victim from directory dir, check
553 * whether the type of victim is right.
554 * 1. We can't do it if dir is read-only (done in permission())
555 * 2. We should have write and exec permissions on dir
556 * 3. We can't remove anything from append-only dir
557 * 4. We can't do anything with immutable dir (done in permission())
558 * 5. If the sticky bit on dir is set we should either
559 * a. be owner of dir, or
560 * b. be owner of victim, or
561 * c. have CAP_FOWNER capability
562 * 6. If the victim is append-only or immutable we can't do antyhing with
563 * links pointing to it.
564 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
565 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
566 * 9. We can't remove a root or mountpoint.
567 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
568 * nfs_async_unlink().
571 static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
575 if (!victim->d_inode)
578 BUG_ON(victim->d_parent->d_inode != dir);
579 audit_inode_child(victim, dir);
581 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
586 if (btrfs_check_sticky(dir, victim->d_inode)||
587 IS_APPEND(victim->d_inode)||
588 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
591 if (!S_ISDIR(victim->d_inode->i_mode))
595 } else if (S_ISDIR(victim->d_inode->i_mode))
599 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
604 /* copy of may_create in fs/namei.c() */
605 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
611 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
615 * Create a new subvolume below @parent. This is largely modeled after
616 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
617 * inside this filesystem so it's quite a bit simpler.
619 static noinline int btrfs_mksubvol(struct path *parent,
620 char *name, int namelen,
621 struct btrfs_root *snap_src,
622 u64 *async_transid, bool readonly)
624 struct inode *dir = parent->dentry->d_inode;
625 struct dentry *dentry;
628 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
630 dentry = lookup_one_len(name, parent->dentry, namelen);
631 error = PTR_ERR(dentry);
639 error = mnt_want_write(parent->mnt);
643 error = btrfs_may_create(dir, dentry);
647 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
649 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
653 error = create_snapshot(snap_src, dentry,
654 name, namelen, async_transid, readonly);
656 error = create_subvol(BTRFS_I(dir)->root, dentry,
657 name, namelen, async_transid);
660 fsnotify_mkdir(dir, dentry);
662 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
664 mnt_drop_write(parent->mnt);
668 mutex_unlock(&dir->i_mutex);
673 * When we're defragging a range, we don't want to kick it off again
674 * if it is really just waiting for delalloc to send it down.
675 * If we find a nice big extent or delalloc range for the bytes in the
676 * file you want to defrag, we return 0 to let you know to skip this
679 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
681 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
682 struct extent_map *em = NULL;
683 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
686 read_lock(&em_tree->lock);
687 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
688 read_unlock(&em_tree->lock);
691 end = extent_map_end(em);
693 if (end - offset > thresh)
696 /* if we already have a nice delalloc here, just stop */
698 end = count_range_bits(io_tree, &offset, offset + thresh,
699 thresh, EXTENT_DELALLOC, 1);
706 * helper function to walk through a file and find extents
707 * newer than a specific transid, and smaller than thresh.
709 * This is used by the defragging code to find new and small
712 static int find_new_extents(struct btrfs_root *root,
713 struct inode *inode, u64 newer_than,
714 u64 *off, int thresh)
716 struct btrfs_path *path;
717 struct btrfs_key min_key;
718 struct btrfs_key max_key;
719 struct extent_buffer *leaf;
720 struct btrfs_file_extent_item *extent;
723 u64 ino = btrfs_ino(inode);
725 path = btrfs_alloc_path();
729 min_key.objectid = ino;
730 min_key.type = BTRFS_EXTENT_DATA_KEY;
731 min_key.offset = *off;
733 max_key.objectid = ino;
734 max_key.type = (u8)-1;
735 max_key.offset = (u64)-1;
737 path->keep_locks = 1;
740 ret = btrfs_search_forward(root, &min_key, &max_key,
741 path, 0, newer_than);
744 if (min_key.objectid != ino)
746 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
749 leaf = path->nodes[0];
750 extent = btrfs_item_ptr(leaf, path->slots[0],
751 struct btrfs_file_extent_item);
753 type = btrfs_file_extent_type(leaf, extent);
754 if (type == BTRFS_FILE_EXTENT_REG &&
755 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
756 check_defrag_in_cache(inode, min_key.offset, thresh)) {
757 *off = min_key.offset;
758 btrfs_free_path(path);
762 if (min_key.offset == (u64)-1)
766 btrfs_release_path(path);
769 btrfs_free_path(path);
773 static int should_defrag_range(struct inode *inode, u64 start, u64 len,
774 int thresh, u64 *last_len, u64 *skip,
777 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
778 struct extent_map *em = NULL;
779 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
783 * make sure that once we start defragging an extent, we keep on
786 if (start < *defrag_end)
792 * hopefully we have this extent in the tree already, try without
793 * the full extent lock
795 read_lock(&em_tree->lock);
796 em = lookup_extent_mapping(em_tree, start, len);
797 read_unlock(&em_tree->lock);
800 /* get the big lock and read metadata off disk */
801 lock_extent(io_tree, start, start + len - 1, GFP_NOFS);
802 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
803 unlock_extent(io_tree, start, start + len - 1, GFP_NOFS);
809 /* this will cover holes, and inline extents */
810 if (em->block_start >= EXTENT_MAP_LAST_BYTE)
814 * we hit a real extent, if it is big don't bother defragging it again
816 if ((*last_len == 0 || *last_len >= thresh) && em->len >= thresh)
820 * last_len ends up being a counter of how many bytes we've defragged.
821 * every time we choose not to defrag an extent, we reset *last_len
822 * so that the next tiny extent will force a defrag.
824 * The end result of this is that tiny extents before a single big
825 * extent will force at least part of that big extent to be defragged.
828 *defrag_end = extent_map_end(em);
831 *skip = extent_map_end(em);
840 * it doesn't do much good to defrag one or two pages
841 * at a time. This pulls in a nice chunk of pages
844 * It also makes sure the delalloc code has enough
845 * dirty data to avoid making new small extents as part
848 * It's a good idea to start RA on this range
849 * before calling this.
851 static int cluster_pages_for_defrag(struct inode *inode,
853 unsigned long start_index,
856 unsigned long file_end;
857 u64 isize = i_size_read(inode);
863 struct btrfs_ordered_extent *ordered;
864 struct extent_state *cached_state = NULL;
865 struct extent_io_tree *tree;
866 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
870 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
872 ret = btrfs_delalloc_reserve_space(inode,
873 num_pages << PAGE_CACHE_SHIFT);
877 tree = &BTRFS_I(inode)->io_tree;
879 /* step one, lock all the pages */
880 for (i = 0; i < num_pages; i++) {
883 page = find_or_create_page(inode->i_mapping,
884 start_index + i, mask);
888 page_start = page_offset(page);
889 page_end = page_start + PAGE_CACHE_SIZE - 1;
891 lock_extent(tree, page_start, page_end, GFP_NOFS);
892 ordered = btrfs_lookup_ordered_extent(inode,
894 unlock_extent(tree, page_start, page_end, GFP_NOFS);
899 btrfs_start_ordered_extent(inode, ordered, 1);
900 btrfs_put_ordered_extent(ordered);
904 if (!PageUptodate(page)) {
905 btrfs_readpage(NULL, page);
907 if (!PageUptodate(page)) {
909 page_cache_release(page);
915 isize = i_size_read(inode);
916 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
917 if (!isize || page->index > file_end) {
918 /* whoops, we blew past eof, skip this page */
920 page_cache_release(page);
924 if (page->mapping != inode->i_mapping) {
926 page_cache_release(page);
936 if (!(inode->i_sb->s_flags & MS_ACTIVE))
940 * so now we have a nice long stream of locked
941 * and up to date pages, lets wait on them
943 for (i = 0; i < i_done; i++)
944 wait_on_page_writeback(pages[i]);
946 page_start = page_offset(pages[0]);
947 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
949 lock_extent_bits(&BTRFS_I(inode)->io_tree,
950 page_start, page_end - 1, 0, &cached_state,
952 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
953 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
954 EXTENT_DO_ACCOUNTING, 0, 0, &cached_state,
957 if (i_done != num_pages) {
958 spin_lock(&BTRFS_I(inode)->lock);
959 BTRFS_I(inode)->outstanding_extents++;
960 spin_unlock(&BTRFS_I(inode)->lock);
961 btrfs_delalloc_release_space(inode,
962 (num_pages - i_done) << PAGE_CACHE_SHIFT);
966 btrfs_set_extent_delalloc(inode, page_start, page_end - 1,
969 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
970 page_start, page_end - 1, &cached_state,
973 for (i = 0; i < i_done; i++) {
974 clear_page_dirty_for_io(pages[i]);
975 ClearPageChecked(pages[i]);
976 set_page_extent_mapped(pages[i]);
977 set_page_dirty(pages[i]);
978 unlock_page(pages[i]);
979 page_cache_release(pages[i]);
983 for (i = 0; i < i_done; i++) {
984 unlock_page(pages[i]);
985 page_cache_release(pages[i]);
987 btrfs_delalloc_release_space(inode, num_pages << PAGE_CACHE_SHIFT);
992 int btrfs_defrag_file(struct inode *inode, struct file *file,
993 struct btrfs_ioctl_defrag_range_args *range,
994 u64 newer_than, unsigned long max_to_defrag)
996 struct btrfs_root *root = BTRFS_I(inode)->root;
997 struct btrfs_super_block *disk_super;
998 struct file_ra_state *ra = NULL;
999 unsigned long last_index;
1000 u64 isize = i_size_read(inode);
1005 u64 newer_off = range->start;
1007 unsigned long ra_index = 0;
1009 int defrag_count = 0;
1010 int compress_type = BTRFS_COMPRESS_ZLIB;
1011 int extent_thresh = range->extent_thresh;
1012 int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1013 int cluster = max_cluster;
1014 u64 new_align = ~((u64)128 * 1024 - 1);
1015 struct page **pages = NULL;
1017 if (extent_thresh == 0)
1018 extent_thresh = 256 * 1024;
1020 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1021 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1023 if (range->compress_type)
1024 compress_type = range->compress_type;
1031 * if we were not given a file, allocate a readahead
1035 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1038 file_ra_state_init(ra, inode->i_mapping);
1043 pages = kmalloc(sizeof(struct page *) * max_cluster,
1050 /* find the last page to defrag */
1051 if (range->start + range->len > range->start) {
1052 last_index = min_t(u64, isize - 1,
1053 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1055 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1059 ret = find_new_extents(root, inode, newer_than,
1060 &newer_off, 64 * 1024);
1062 range->start = newer_off;
1064 * we always align our defrag to help keep
1065 * the extents in the file evenly spaced
1067 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1071 i = range->start >> PAGE_CACHE_SHIFT;
1074 max_to_defrag = last_index + 1;
1077 * make writeback starts from i, so the defrag range can be
1078 * written sequentially.
1080 if (i < inode->i_mapping->writeback_index)
1081 inode->i_mapping->writeback_index = i;
1083 while (i <= last_index && defrag_count < max_to_defrag &&
1084 (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1085 PAGE_CACHE_SHIFT)) {
1087 * make sure we stop running if someone unmounts
1090 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1094 !should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1101 * the should_defrag function tells us how much to skip
1102 * bump our counter by the suggested amount
1104 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1105 i = max(i + 1, next);
1110 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1111 PAGE_CACHE_SHIFT) - i;
1112 cluster = min(cluster, max_cluster);
1114 cluster = max_cluster;
1117 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1118 BTRFS_I(inode)->force_compress = compress_type;
1120 if (i + cluster > ra_index) {
1121 ra_index = max(i, ra_index);
1122 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1124 ra_index += max_cluster;
1127 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1131 defrag_count += ret;
1132 balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret);
1135 if (newer_off == (u64)-1)
1138 newer_off = max(newer_off + 1,
1139 (u64)i << PAGE_CACHE_SHIFT);
1141 ret = find_new_extents(root, inode,
1142 newer_than, &newer_off,
1145 range->start = newer_off;
1146 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1153 last_len += ret << PAGE_CACHE_SHIFT;
1161 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1162 filemap_flush(inode->i_mapping);
1164 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1165 /* the filemap_flush will queue IO into the worker threads, but
1166 * we have to make sure the IO is actually started and that
1167 * ordered extents get created before we return
1169 atomic_inc(&root->fs_info->async_submit_draining);
1170 while (atomic_read(&root->fs_info->nr_async_submits) ||
1171 atomic_read(&root->fs_info->async_delalloc_pages)) {
1172 wait_event(root->fs_info->async_submit_wait,
1173 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1174 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1176 atomic_dec(&root->fs_info->async_submit_draining);
1178 mutex_lock(&inode->i_mutex);
1179 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1180 mutex_unlock(&inode->i_mutex);
1183 disk_super = root->fs_info->super_copy;
1184 features = btrfs_super_incompat_flags(disk_super);
1185 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1186 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
1187 btrfs_set_super_incompat_flags(disk_super, features);
1199 static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
1205 struct btrfs_ioctl_vol_args *vol_args;
1206 struct btrfs_trans_handle *trans;
1207 struct btrfs_device *device = NULL;
1209 char *devstr = NULL;
1213 if (root->fs_info->sb->s_flags & MS_RDONLY)
1216 if (!capable(CAP_SYS_ADMIN))
1219 mutex_lock(&root->fs_info->volume_mutex);
1220 if (root->fs_info->balance_ctl) {
1221 printk(KERN_INFO "btrfs: balance in progress\n");
1226 vol_args = memdup_user(arg, sizeof(*vol_args));
1227 if (IS_ERR(vol_args)) {
1228 ret = PTR_ERR(vol_args);
1232 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1234 sizestr = vol_args->name;
1235 devstr = strchr(sizestr, ':');
1238 sizestr = devstr + 1;
1240 devstr = vol_args->name;
1241 devid = simple_strtoull(devstr, &end, 10);
1242 printk(KERN_INFO "btrfs: resizing devid %llu\n",
1243 (unsigned long long)devid);
1245 device = btrfs_find_device(root, devid, NULL, NULL);
1247 printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
1248 (unsigned long long)devid);
1252 if (!strcmp(sizestr, "max"))
1253 new_size = device->bdev->bd_inode->i_size;
1255 if (sizestr[0] == '-') {
1258 } else if (sizestr[0] == '+') {
1262 new_size = memparse(sizestr, NULL);
1263 if (new_size == 0) {
1269 old_size = device->total_bytes;
1272 if (new_size > old_size) {
1276 new_size = old_size - new_size;
1277 } else if (mod > 0) {
1278 new_size = old_size + new_size;
1281 if (new_size < 256 * 1024 * 1024) {
1285 if (new_size > device->bdev->bd_inode->i_size) {
1290 do_div(new_size, root->sectorsize);
1291 new_size *= root->sectorsize;
1293 printk(KERN_INFO "btrfs: new size for %s is %llu\n",
1294 device->name, (unsigned long long)new_size);
1296 if (new_size > old_size) {
1297 trans = btrfs_start_transaction(root, 0);
1298 if (IS_ERR(trans)) {
1299 ret = PTR_ERR(trans);
1302 ret = btrfs_grow_device(trans, device, new_size);
1303 btrfs_commit_transaction(trans, root);
1304 } else if (new_size < old_size) {
1305 ret = btrfs_shrink_device(device, new_size);
1311 mutex_unlock(&root->fs_info->volume_mutex);
1315 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1322 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
1323 struct file *src_file;
1327 if (root->fs_info->sb->s_flags & MS_RDONLY)
1330 namelen = strlen(name);
1331 if (strchr(name, '/')) {
1336 if (name[0] == '.' &&
1337 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1343 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1344 NULL, transid, readonly);
1346 struct inode *src_inode;
1347 src_file = fget(fd);
1353 src_inode = src_file->f_path.dentry->d_inode;
1354 if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
1355 printk(KERN_INFO "btrfs: Snapshot src from "
1361 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1362 BTRFS_I(src_inode)->root,
1370 static noinline int btrfs_ioctl_snap_create(struct file *file,
1371 void __user *arg, int subvol)
1373 struct btrfs_ioctl_vol_args *vol_args;
1376 vol_args = memdup_user(arg, sizeof(*vol_args));
1377 if (IS_ERR(vol_args))
1378 return PTR_ERR(vol_args);
1379 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1381 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1382 vol_args->fd, subvol,
1389 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1390 void __user *arg, int subvol)
1392 struct btrfs_ioctl_vol_args_v2 *vol_args;
1396 bool readonly = false;
1398 vol_args = memdup_user(arg, sizeof(*vol_args));
1399 if (IS_ERR(vol_args))
1400 return PTR_ERR(vol_args);
1401 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1403 if (vol_args->flags &
1404 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY)) {
1409 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1411 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1414 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1415 vol_args->fd, subvol,
1418 if (ret == 0 && ptr &&
1420 offsetof(struct btrfs_ioctl_vol_args_v2,
1421 transid), ptr, sizeof(*ptr)))
1428 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1431 struct inode *inode = fdentry(file)->d_inode;
1432 struct btrfs_root *root = BTRFS_I(inode)->root;
1436 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1439 down_read(&root->fs_info->subvol_sem);
1440 if (btrfs_root_readonly(root))
1441 flags |= BTRFS_SUBVOL_RDONLY;
1442 up_read(&root->fs_info->subvol_sem);
1444 if (copy_to_user(arg, &flags, sizeof(flags)))
1450 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1453 struct inode *inode = fdentry(file)->d_inode;
1454 struct btrfs_root *root = BTRFS_I(inode)->root;
1455 struct btrfs_trans_handle *trans;
1460 if (root->fs_info->sb->s_flags & MS_RDONLY)
1463 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1466 if (copy_from_user(&flags, arg, sizeof(flags)))
1469 if (flags & BTRFS_SUBVOL_CREATE_ASYNC)
1472 if (flags & ~BTRFS_SUBVOL_RDONLY)
1475 if (!inode_owner_or_capable(inode))
1478 down_write(&root->fs_info->subvol_sem);
1481 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1484 root_flags = btrfs_root_flags(&root->root_item);
1485 if (flags & BTRFS_SUBVOL_RDONLY)
1486 btrfs_set_root_flags(&root->root_item,
1487 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1489 btrfs_set_root_flags(&root->root_item,
1490 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1492 trans = btrfs_start_transaction(root, 1);
1493 if (IS_ERR(trans)) {
1494 ret = PTR_ERR(trans);
1498 ret = btrfs_update_root(trans, root->fs_info->tree_root,
1499 &root->root_key, &root->root_item);
1501 btrfs_commit_transaction(trans, root);
1504 btrfs_set_root_flags(&root->root_item, root_flags);
1506 up_write(&root->fs_info->subvol_sem);
1511 * helper to check if the subvolume references other subvolumes
1513 static noinline int may_destroy_subvol(struct btrfs_root *root)
1515 struct btrfs_path *path;
1516 struct btrfs_key key;
1519 path = btrfs_alloc_path();
1523 key.objectid = root->root_key.objectid;
1524 key.type = BTRFS_ROOT_REF_KEY;
1525 key.offset = (u64)-1;
1527 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1534 if (path->slots[0] > 0) {
1536 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1537 if (key.objectid == root->root_key.objectid &&
1538 key.type == BTRFS_ROOT_REF_KEY)
1542 btrfs_free_path(path);
1546 static noinline int key_in_sk(struct btrfs_key *key,
1547 struct btrfs_ioctl_search_key *sk)
1549 struct btrfs_key test;
1552 test.objectid = sk->min_objectid;
1553 test.type = sk->min_type;
1554 test.offset = sk->min_offset;
1556 ret = btrfs_comp_cpu_keys(key, &test);
1560 test.objectid = sk->max_objectid;
1561 test.type = sk->max_type;
1562 test.offset = sk->max_offset;
1564 ret = btrfs_comp_cpu_keys(key, &test);
1570 static noinline int copy_to_sk(struct btrfs_root *root,
1571 struct btrfs_path *path,
1572 struct btrfs_key *key,
1573 struct btrfs_ioctl_search_key *sk,
1575 unsigned long *sk_offset,
1579 struct extent_buffer *leaf;
1580 struct btrfs_ioctl_search_header sh;
1581 unsigned long item_off;
1582 unsigned long item_len;
1588 leaf = path->nodes[0];
1589 slot = path->slots[0];
1590 nritems = btrfs_header_nritems(leaf);
1592 if (btrfs_header_generation(leaf) > sk->max_transid) {
1596 found_transid = btrfs_header_generation(leaf);
1598 for (i = slot; i < nritems; i++) {
1599 item_off = btrfs_item_ptr_offset(leaf, i);
1600 item_len = btrfs_item_size_nr(leaf, i);
1602 if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1605 if (sizeof(sh) + item_len + *sk_offset >
1606 BTRFS_SEARCH_ARGS_BUFSIZE) {
1611 btrfs_item_key_to_cpu(leaf, key, i);
1612 if (!key_in_sk(key, sk))
1615 sh.objectid = key->objectid;
1616 sh.offset = key->offset;
1617 sh.type = key->type;
1619 sh.transid = found_transid;
1621 /* copy search result header */
1622 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1623 *sk_offset += sizeof(sh);
1626 char *p = buf + *sk_offset;
1628 read_extent_buffer(leaf, p,
1629 item_off, item_len);
1630 *sk_offset += item_len;
1634 if (*num_found >= sk->nr_items)
1639 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1641 else if (key->type < (u8)-1 && key->type < sk->max_type) {
1644 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1654 static noinline int search_ioctl(struct inode *inode,
1655 struct btrfs_ioctl_search_args *args)
1657 struct btrfs_root *root;
1658 struct btrfs_key key;
1659 struct btrfs_key max_key;
1660 struct btrfs_path *path;
1661 struct btrfs_ioctl_search_key *sk = &args->key;
1662 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1665 unsigned long sk_offset = 0;
1667 path = btrfs_alloc_path();
1671 if (sk->tree_id == 0) {
1672 /* search the root of the inode that was passed */
1673 root = BTRFS_I(inode)->root;
1675 key.objectid = sk->tree_id;
1676 key.type = BTRFS_ROOT_ITEM_KEY;
1677 key.offset = (u64)-1;
1678 root = btrfs_read_fs_root_no_name(info, &key);
1680 printk(KERN_ERR "could not find root %llu\n",
1682 btrfs_free_path(path);
1687 key.objectid = sk->min_objectid;
1688 key.type = sk->min_type;
1689 key.offset = sk->min_offset;
1691 max_key.objectid = sk->max_objectid;
1692 max_key.type = sk->max_type;
1693 max_key.offset = sk->max_offset;
1695 path->keep_locks = 1;
1698 ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1705 ret = copy_to_sk(root, path, &key, sk, args->buf,
1706 &sk_offset, &num_found);
1707 btrfs_release_path(path);
1708 if (ret || num_found >= sk->nr_items)
1714 sk->nr_items = num_found;
1715 btrfs_free_path(path);
1719 static noinline int btrfs_ioctl_tree_search(struct file *file,
1722 struct btrfs_ioctl_search_args *args;
1723 struct inode *inode;
1726 if (!capable(CAP_SYS_ADMIN))
1729 args = memdup_user(argp, sizeof(*args));
1731 return PTR_ERR(args);
1733 inode = fdentry(file)->d_inode;
1734 ret = search_ioctl(inode, args);
1735 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1742 * Search INODE_REFs to identify path name of 'dirid' directory
1743 * in a 'tree_id' tree. and sets path name to 'name'.
1745 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1746 u64 tree_id, u64 dirid, char *name)
1748 struct btrfs_root *root;
1749 struct btrfs_key key;
1755 struct btrfs_inode_ref *iref;
1756 struct extent_buffer *l;
1757 struct btrfs_path *path;
1759 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1764 path = btrfs_alloc_path();
1768 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1770 key.objectid = tree_id;
1771 key.type = BTRFS_ROOT_ITEM_KEY;
1772 key.offset = (u64)-1;
1773 root = btrfs_read_fs_root_no_name(info, &key);
1775 printk(KERN_ERR "could not find root %llu\n", tree_id);
1780 key.objectid = dirid;
1781 key.type = BTRFS_INODE_REF_KEY;
1782 key.offset = (u64)-1;
1785 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1790 slot = path->slots[0];
1791 if (ret > 0 && slot > 0)
1793 btrfs_item_key_to_cpu(l, &key, slot);
1795 if (ret > 0 && (key.objectid != dirid ||
1796 key.type != BTRFS_INODE_REF_KEY)) {
1801 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1802 len = btrfs_inode_ref_name_len(l, iref);
1804 total_len += len + 1;
1809 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1811 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1814 btrfs_release_path(path);
1815 key.objectid = key.offset;
1816 key.offset = (u64)-1;
1817 dirid = key.objectid;
1821 memmove(name, ptr, total_len);
1822 name[total_len]='\0';
1825 btrfs_free_path(path);
1829 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1832 struct btrfs_ioctl_ino_lookup_args *args;
1833 struct inode *inode;
1836 if (!capable(CAP_SYS_ADMIN))
1839 args = memdup_user(argp, sizeof(*args));
1841 return PTR_ERR(args);
1843 inode = fdentry(file)->d_inode;
1845 if (args->treeid == 0)
1846 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1848 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1849 args->treeid, args->objectid,
1852 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1859 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1862 struct dentry *parent = fdentry(file);
1863 struct dentry *dentry;
1864 struct inode *dir = parent->d_inode;
1865 struct inode *inode;
1866 struct btrfs_root *root = BTRFS_I(dir)->root;
1867 struct btrfs_root *dest = NULL;
1868 struct btrfs_ioctl_vol_args *vol_args;
1869 struct btrfs_trans_handle *trans;
1874 vol_args = memdup_user(arg, sizeof(*vol_args));
1875 if (IS_ERR(vol_args))
1876 return PTR_ERR(vol_args);
1878 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1879 namelen = strlen(vol_args->name);
1880 if (strchr(vol_args->name, '/') ||
1881 strncmp(vol_args->name, "..", namelen) == 0) {
1886 err = mnt_want_write(file->f_path.mnt);
1890 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
1891 dentry = lookup_one_len(vol_args->name, parent, namelen);
1892 if (IS_ERR(dentry)) {
1893 err = PTR_ERR(dentry);
1894 goto out_unlock_dir;
1897 if (!dentry->d_inode) {
1902 inode = dentry->d_inode;
1903 dest = BTRFS_I(inode)->root;
1904 if (!capable(CAP_SYS_ADMIN)){
1906 * Regular user. Only allow this with a special mount
1907 * option, when the user has write+exec access to the
1908 * subvol root, and when rmdir(2) would have been
1911 * Note that this is _not_ check that the subvol is
1912 * empty or doesn't contain data that we wouldn't
1913 * otherwise be able to delete.
1915 * Users who want to delete empty subvols should try
1919 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1923 * Do not allow deletion if the parent dir is the same
1924 * as the dir to be deleted. That means the ioctl
1925 * must be called on the dentry referencing the root
1926 * of the subvol, not a random directory contained
1933 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
1937 /* check if subvolume may be deleted by a non-root user */
1938 err = btrfs_may_delete(dir, dentry, 1);
1943 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1948 mutex_lock(&inode->i_mutex);
1949 err = d_invalidate(dentry);
1953 down_write(&root->fs_info->subvol_sem);
1955 err = may_destroy_subvol(dest);
1959 trans = btrfs_start_transaction(root, 0);
1960 if (IS_ERR(trans)) {
1961 err = PTR_ERR(trans);
1964 trans->block_rsv = &root->fs_info->global_block_rsv;
1966 ret = btrfs_unlink_subvol(trans, root, dir,
1967 dest->root_key.objectid,
1968 dentry->d_name.name,
1969 dentry->d_name.len);
1972 btrfs_record_root_in_trans(trans, dest);
1974 memset(&dest->root_item.drop_progress, 0,
1975 sizeof(dest->root_item.drop_progress));
1976 dest->root_item.drop_level = 0;
1977 btrfs_set_root_refs(&dest->root_item, 0);
1979 if (!xchg(&dest->orphan_item_inserted, 1)) {
1980 ret = btrfs_insert_orphan_item(trans,
1981 root->fs_info->tree_root,
1982 dest->root_key.objectid);
1986 ret = btrfs_end_transaction(trans, root);
1988 inode->i_flags |= S_DEAD;
1990 up_write(&root->fs_info->subvol_sem);
1992 mutex_unlock(&inode->i_mutex);
1994 shrink_dcache_sb(root->fs_info->sb);
1995 btrfs_invalidate_inodes(dest);
2001 mutex_unlock(&dir->i_mutex);
2002 mnt_drop_write(file->f_path.mnt);
2008 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2010 struct inode *inode = fdentry(file)->d_inode;
2011 struct btrfs_root *root = BTRFS_I(inode)->root;
2012 struct btrfs_ioctl_defrag_range_args *range;
2015 if (btrfs_root_readonly(root))
2018 ret = mnt_want_write(file->f_path.mnt);
2022 switch (inode->i_mode & S_IFMT) {
2024 if (!capable(CAP_SYS_ADMIN)) {
2028 ret = btrfs_defrag_root(root, 0);
2031 ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
2034 if (!(file->f_mode & FMODE_WRITE)) {
2039 range = kzalloc(sizeof(*range), GFP_KERNEL);
2046 if (copy_from_user(range, argp,
2052 /* compression requires us to start the IO */
2053 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2054 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2055 range->extent_thresh = (u32)-1;
2058 /* the rest are all set to zero by kzalloc */
2059 range->len = (u64)-1;
2061 ret = btrfs_defrag_file(fdentry(file)->d_inode, file,
2071 mnt_drop_write(file->f_path.mnt);
2075 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2077 struct btrfs_ioctl_vol_args *vol_args;
2080 if (!capable(CAP_SYS_ADMIN))
2083 mutex_lock(&root->fs_info->volume_mutex);
2084 if (root->fs_info->balance_ctl) {
2085 printk(KERN_INFO "btrfs: balance in progress\n");
2090 vol_args = memdup_user(arg, sizeof(*vol_args));
2091 if (IS_ERR(vol_args)) {
2092 ret = PTR_ERR(vol_args);
2096 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2097 ret = btrfs_init_new_device(root, vol_args->name);
2101 mutex_unlock(&root->fs_info->volume_mutex);
2105 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
2107 struct btrfs_ioctl_vol_args *vol_args;
2110 if (!capable(CAP_SYS_ADMIN))
2113 if (root->fs_info->sb->s_flags & MS_RDONLY)
2116 mutex_lock(&root->fs_info->volume_mutex);
2117 if (root->fs_info->balance_ctl) {
2118 printk(KERN_INFO "btrfs: balance in progress\n");
2123 vol_args = memdup_user(arg, sizeof(*vol_args));
2124 if (IS_ERR(vol_args)) {
2125 ret = PTR_ERR(vol_args);
2129 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2130 ret = btrfs_rm_device(root, vol_args->name);
2134 mutex_unlock(&root->fs_info->volume_mutex);
2138 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2140 struct btrfs_ioctl_fs_info_args *fi_args;
2141 struct btrfs_device *device;
2142 struct btrfs_device *next;
2143 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2146 if (!capable(CAP_SYS_ADMIN))
2149 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2153 fi_args->num_devices = fs_devices->num_devices;
2154 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2156 mutex_lock(&fs_devices->device_list_mutex);
2157 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2158 if (device->devid > fi_args->max_id)
2159 fi_args->max_id = device->devid;
2161 mutex_unlock(&fs_devices->device_list_mutex);
2163 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2170 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2172 struct btrfs_ioctl_dev_info_args *di_args;
2173 struct btrfs_device *dev;
2174 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2176 char *s_uuid = NULL;
2177 char empty_uuid[BTRFS_UUID_SIZE] = {0};
2179 if (!capable(CAP_SYS_ADMIN))
2182 di_args = memdup_user(arg, sizeof(*di_args));
2183 if (IS_ERR(di_args))
2184 return PTR_ERR(di_args);
2186 if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2187 s_uuid = di_args->uuid;
2189 mutex_lock(&fs_devices->device_list_mutex);
2190 dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL);
2191 mutex_unlock(&fs_devices->device_list_mutex);
2198 di_args->devid = dev->devid;
2199 di_args->bytes_used = dev->bytes_used;
2200 di_args->total_bytes = dev->total_bytes;
2201 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2202 strncpy(di_args->path, dev->name, sizeof(di_args->path));
2205 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2212 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2213 u64 off, u64 olen, u64 destoff)
2215 struct inode *inode = fdentry(file)->d_inode;
2216 struct btrfs_root *root = BTRFS_I(inode)->root;
2217 struct file *src_file;
2219 struct btrfs_trans_handle *trans;
2220 struct btrfs_path *path;
2221 struct extent_buffer *leaf;
2223 struct btrfs_key key;
2228 u64 bs = root->fs_info->sb->s_blocksize;
2233 * - split compressed inline extents. annoying: we need to
2234 * decompress into destination's address_space (the file offset
2235 * may change, so source mapping won't do), then recompress (or
2236 * otherwise reinsert) a subrange.
2237 * - allow ranges within the same file to be cloned (provided
2238 * they don't overlap)?
2241 /* the destination must be opened for writing */
2242 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
2245 if (btrfs_root_readonly(root))
2248 ret = mnt_want_write(file->f_path.mnt);
2252 src_file = fget(srcfd);
2255 goto out_drop_write;
2258 src = src_file->f_dentry->d_inode;
2264 /* the src must be open for reading */
2265 if (!(src_file->f_mode & FMODE_READ))
2268 /* don't make the dst file partly checksummed */
2269 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2270 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2274 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2278 if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
2282 buf = vmalloc(btrfs_level_size(root, 0));
2286 path = btrfs_alloc_path();
2294 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2295 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
2297 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2298 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
2301 /* determine range to clone */
2303 if (off + len > src->i_size || off + len < off)
2306 olen = len = src->i_size - off;
2307 /* if we extend to eof, continue to block boundary */
2308 if (off + len == src->i_size)
2309 len = ALIGN(src->i_size, bs) - off;
2311 /* verify the end result is block aligned */
2312 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2313 !IS_ALIGNED(destoff, bs))
2316 if (destoff > inode->i_size) {
2317 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2322 /* truncate page cache pages from target inode range */
2323 truncate_inode_pages_range(&inode->i_data, destoff,
2324 PAGE_CACHE_ALIGN(destoff + len) - 1);
2326 /* do any pending delalloc/csum calc on src, one way or
2327 another, and lock file content */
2329 struct btrfs_ordered_extent *ordered;
2330 lock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
2331 ordered = btrfs_lookup_first_ordered_extent(src, off+len);
2333 !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len,
2334 EXTENT_DELALLOC, 0, NULL))
2336 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
2338 btrfs_put_ordered_extent(ordered);
2339 btrfs_wait_ordered_range(src, off, len);
2343 key.objectid = btrfs_ino(src);
2344 key.type = BTRFS_EXTENT_DATA_KEY;
2349 * note the key will change type as we walk through the
2352 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2356 nritems = btrfs_header_nritems(path->nodes[0]);
2357 if (path->slots[0] >= nritems) {
2358 ret = btrfs_next_leaf(root, path);
2363 nritems = btrfs_header_nritems(path->nodes[0]);
2365 leaf = path->nodes[0];
2366 slot = path->slots[0];
2368 btrfs_item_key_to_cpu(leaf, &key, slot);
2369 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
2370 key.objectid != btrfs_ino(src))
2373 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2374 struct btrfs_file_extent_item *extent;
2377 struct btrfs_key new_key;
2378 u64 disko = 0, diskl = 0;
2379 u64 datao = 0, datal = 0;
2383 size = btrfs_item_size_nr(leaf, slot);
2384 read_extent_buffer(leaf, buf,
2385 btrfs_item_ptr_offset(leaf, slot),
2388 extent = btrfs_item_ptr(leaf, slot,
2389 struct btrfs_file_extent_item);
2390 comp = btrfs_file_extent_compression(leaf, extent);
2391 type = btrfs_file_extent_type(leaf, extent);
2392 if (type == BTRFS_FILE_EXTENT_REG ||
2393 type == BTRFS_FILE_EXTENT_PREALLOC) {
2394 disko = btrfs_file_extent_disk_bytenr(leaf,
2396 diskl = btrfs_file_extent_disk_num_bytes(leaf,
2398 datao = btrfs_file_extent_offset(leaf, extent);
2399 datal = btrfs_file_extent_num_bytes(leaf,
2401 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2402 /* take upper bound, may be compressed */
2403 datal = btrfs_file_extent_ram_bytes(leaf,
2406 btrfs_release_path(path);
2408 if (key.offset + datal <= off ||
2409 key.offset >= off+len)
2412 memcpy(&new_key, &key, sizeof(new_key));
2413 new_key.objectid = btrfs_ino(inode);
2414 if (off <= key.offset)
2415 new_key.offset = key.offset + destoff - off;
2417 new_key.offset = destoff;
2420 * 1 - adjusting old extent (we may have to split it)
2421 * 1 - add new extent
2424 trans = btrfs_start_transaction(root, 3);
2425 if (IS_ERR(trans)) {
2426 ret = PTR_ERR(trans);
2430 if (type == BTRFS_FILE_EXTENT_REG ||
2431 type == BTRFS_FILE_EXTENT_PREALLOC) {
2433 * a | --- range to clone ---| b
2434 * | ------------- extent ------------- |
2437 /* substract range b */
2438 if (key.offset + datal > off + len)
2439 datal = off + len - key.offset;
2441 /* substract range a */
2442 if (off > key.offset) {
2443 datao += off - key.offset;
2444 datal -= off - key.offset;
2447 ret = btrfs_drop_extents(trans, inode,
2449 new_key.offset + datal,
2453 ret = btrfs_insert_empty_item(trans, root, path,
2457 leaf = path->nodes[0];
2458 slot = path->slots[0];
2459 write_extent_buffer(leaf, buf,
2460 btrfs_item_ptr_offset(leaf, slot),
2463 extent = btrfs_item_ptr(leaf, slot,
2464 struct btrfs_file_extent_item);
2466 /* disko == 0 means it's a hole */
2470 btrfs_set_file_extent_offset(leaf, extent,
2472 btrfs_set_file_extent_num_bytes(leaf, extent,
2475 inode_add_bytes(inode, datal);
2476 ret = btrfs_inc_extent_ref(trans, root,
2478 root->root_key.objectid,
2480 new_key.offset - datao,
2484 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2487 if (off > key.offset) {
2488 skip = off - key.offset;
2489 new_key.offset += skip;
2492 if (key.offset + datal > off+len)
2493 trim = key.offset + datal - (off+len);
2495 if (comp && (skip || trim)) {
2497 btrfs_end_transaction(trans, root);
2500 size -= skip + trim;
2501 datal -= skip + trim;
2503 ret = btrfs_drop_extents(trans, inode,
2505 new_key.offset + datal,
2509 ret = btrfs_insert_empty_item(trans, root, path,
2515 btrfs_file_extent_calc_inline_size(0);
2516 memmove(buf+start, buf+start+skip,
2520 leaf = path->nodes[0];
2521 slot = path->slots[0];
2522 write_extent_buffer(leaf, buf,
2523 btrfs_item_ptr_offset(leaf, slot),
2525 inode_add_bytes(inode, datal);
2528 btrfs_mark_buffer_dirty(leaf);
2529 btrfs_release_path(path);
2531 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2534 * we round up to the block size at eof when
2535 * determining which extents to clone above,
2536 * but shouldn't round up the file size
2538 endoff = new_key.offset + datal;
2539 if (endoff > destoff+olen)
2540 endoff = destoff+olen;
2541 if (endoff > inode->i_size)
2542 btrfs_i_size_write(inode, endoff);
2544 ret = btrfs_update_inode(trans, root, inode);
2546 btrfs_end_transaction(trans, root);
2549 btrfs_release_path(path);
2554 btrfs_release_path(path);
2555 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
2557 mutex_unlock(&src->i_mutex);
2558 mutex_unlock(&inode->i_mutex);
2560 btrfs_free_path(path);
2564 mnt_drop_write(file->f_path.mnt);
2568 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
2570 struct btrfs_ioctl_clone_range_args args;
2572 if (copy_from_user(&args, argp, sizeof(args)))
2574 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2575 args.src_length, args.dest_offset);
2579 * there are many ways the trans_start and trans_end ioctls can lead
2580 * to deadlocks. They should only be used by applications that
2581 * basically own the machine, and have a very in depth understanding
2582 * of all the possible deadlocks and enospc problems.
2584 static long btrfs_ioctl_trans_start(struct file *file)
2586 struct inode *inode = fdentry(file)->d_inode;
2587 struct btrfs_root *root = BTRFS_I(inode)->root;
2588 struct btrfs_trans_handle *trans;
2592 if (!capable(CAP_SYS_ADMIN))
2596 if (file->private_data)
2600 if (btrfs_root_readonly(root))
2603 ret = mnt_want_write(file->f_path.mnt);
2607 atomic_inc(&root->fs_info->open_ioctl_trans);
2610 trans = btrfs_start_ioctl_transaction(root);
2614 file->private_data = trans;
2618 atomic_dec(&root->fs_info->open_ioctl_trans);
2619 mnt_drop_write(file->f_path.mnt);
2624 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2626 struct inode *inode = fdentry(file)->d_inode;
2627 struct btrfs_root *root = BTRFS_I(inode)->root;
2628 struct btrfs_root *new_root;
2629 struct btrfs_dir_item *di;
2630 struct btrfs_trans_handle *trans;
2631 struct btrfs_path *path;
2632 struct btrfs_key location;
2633 struct btrfs_disk_key disk_key;
2634 struct btrfs_super_block *disk_super;
2639 if (!capable(CAP_SYS_ADMIN))
2642 if (copy_from_user(&objectid, argp, sizeof(objectid)))
2646 objectid = root->root_key.objectid;
2648 location.objectid = objectid;
2649 location.type = BTRFS_ROOT_ITEM_KEY;
2650 location.offset = (u64)-1;
2652 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2653 if (IS_ERR(new_root))
2654 return PTR_ERR(new_root);
2656 if (btrfs_root_refs(&new_root->root_item) == 0)
2659 path = btrfs_alloc_path();
2662 path->leave_spinning = 1;
2664 trans = btrfs_start_transaction(root, 1);
2665 if (IS_ERR(trans)) {
2666 btrfs_free_path(path);
2667 return PTR_ERR(trans);
2670 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
2671 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2672 dir_id, "default", 7, 1);
2673 if (IS_ERR_OR_NULL(di)) {
2674 btrfs_free_path(path);
2675 btrfs_end_transaction(trans, root);
2676 printk(KERN_ERR "Umm, you don't have the default dir item, "
2677 "this isn't going to work\n");
2681 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2682 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2683 btrfs_mark_buffer_dirty(path->nodes[0]);
2684 btrfs_free_path(path);
2686 disk_super = root->fs_info->super_copy;
2687 features = btrfs_super_incompat_flags(disk_super);
2688 if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) {
2689 features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL;
2690 btrfs_set_super_incompat_flags(disk_super, features);
2692 btrfs_end_transaction(trans, root);
2697 static void get_block_group_info(struct list_head *groups_list,
2698 struct btrfs_ioctl_space_info *space)
2700 struct btrfs_block_group_cache *block_group;
2702 space->total_bytes = 0;
2703 space->used_bytes = 0;
2705 list_for_each_entry(block_group, groups_list, list) {
2706 space->flags = block_group->flags;
2707 space->total_bytes += block_group->key.offset;
2708 space->used_bytes +=
2709 btrfs_block_group_used(&block_group->item);
2713 long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2715 struct btrfs_ioctl_space_args space_args;
2716 struct btrfs_ioctl_space_info space;
2717 struct btrfs_ioctl_space_info *dest;
2718 struct btrfs_ioctl_space_info *dest_orig;
2719 struct btrfs_ioctl_space_info __user *user_dest;
2720 struct btrfs_space_info *info;
2721 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2722 BTRFS_BLOCK_GROUP_SYSTEM,
2723 BTRFS_BLOCK_GROUP_METADATA,
2724 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2731 if (copy_from_user(&space_args,
2732 (struct btrfs_ioctl_space_args __user *)arg,
2733 sizeof(space_args)))
2736 for (i = 0; i < num_types; i++) {
2737 struct btrfs_space_info *tmp;
2741 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2743 if (tmp->flags == types[i]) {
2753 down_read(&info->groups_sem);
2754 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2755 if (!list_empty(&info->block_groups[c]))
2758 up_read(&info->groups_sem);
2761 /* space_slots == 0 means they are asking for a count */
2762 if (space_args.space_slots == 0) {
2763 space_args.total_spaces = slot_count;
2767 slot_count = min_t(u64, space_args.space_slots, slot_count);
2769 alloc_size = sizeof(*dest) * slot_count;
2771 /* we generally have at most 6 or so space infos, one for each raid
2772 * level. So, a whole page should be more than enough for everyone
2774 if (alloc_size > PAGE_CACHE_SIZE)
2777 space_args.total_spaces = 0;
2778 dest = kmalloc(alloc_size, GFP_NOFS);
2783 /* now we have a buffer to copy into */
2784 for (i = 0; i < num_types; i++) {
2785 struct btrfs_space_info *tmp;
2792 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2794 if (tmp->flags == types[i]) {
2803 down_read(&info->groups_sem);
2804 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2805 if (!list_empty(&info->block_groups[c])) {
2806 get_block_group_info(&info->block_groups[c],
2808 memcpy(dest, &space, sizeof(space));
2810 space_args.total_spaces++;
2816 up_read(&info->groups_sem);
2819 user_dest = (struct btrfs_ioctl_space_info *)
2820 (arg + sizeof(struct btrfs_ioctl_space_args));
2822 if (copy_to_user(user_dest, dest_orig, alloc_size))
2827 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
2834 * there are many ways the trans_start and trans_end ioctls can lead
2835 * to deadlocks. They should only be used by applications that
2836 * basically own the machine, and have a very in depth understanding
2837 * of all the possible deadlocks and enospc problems.
2839 long btrfs_ioctl_trans_end(struct file *file)
2841 struct inode *inode = fdentry(file)->d_inode;
2842 struct btrfs_root *root = BTRFS_I(inode)->root;
2843 struct btrfs_trans_handle *trans;
2845 trans = file->private_data;
2848 file->private_data = NULL;
2850 btrfs_end_transaction(trans, root);
2852 atomic_dec(&root->fs_info->open_ioctl_trans);
2854 mnt_drop_write(file->f_path.mnt);
2858 static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
2860 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2861 struct btrfs_trans_handle *trans;
2865 trans = btrfs_start_transaction(root, 0);
2867 return PTR_ERR(trans);
2868 transid = trans->transid;
2869 ret = btrfs_commit_transaction_async(trans, root, 0);
2871 btrfs_end_transaction(trans, root);
2876 if (copy_to_user(argp, &transid, sizeof(transid)))
2881 static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
2883 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2887 if (copy_from_user(&transid, argp, sizeof(transid)))
2890 transid = 0; /* current trans */
2892 return btrfs_wait_for_commit(root, transid);
2895 static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg)
2898 struct btrfs_ioctl_scrub_args *sa;
2900 if (!capable(CAP_SYS_ADMIN))
2903 sa = memdup_user(arg, sizeof(*sa));
2907 ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end,
2908 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY);
2910 if (copy_to_user(arg, sa, sizeof(*sa)))
2917 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
2919 if (!capable(CAP_SYS_ADMIN))
2922 return btrfs_scrub_cancel(root);
2925 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
2928 struct btrfs_ioctl_scrub_args *sa;
2931 if (!capable(CAP_SYS_ADMIN))
2934 sa = memdup_user(arg, sizeof(*sa));
2938 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
2940 if (copy_to_user(arg, sa, sizeof(*sa)))
2947 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
2953 struct btrfs_ioctl_ino_path_args *ipa = NULL;
2954 struct inode_fs_paths *ipath = NULL;
2955 struct btrfs_path *path;
2957 if (!capable(CAP_SYS_ADMIN))
2960 path = btrfs_alloc_path();
2966 ipa = memdup_user(arg, sizeof(*ipa));
2973 size = min_t(u32, ipa->size, 4096);
2974 ipath = init_ipath(size, root, path);
2975 if (IS_ERR(ipath)) {
2976 ret = PTR_ERR(ipath);
2981 ret = paths_from_inode(ipa->inum, ipath);
2985 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
2986 rel_ptr = ipath->fspath->val[i] -
2987 (u64)(unsigned long)ipath->fspath->val;
2988 ipath->fspath->val[i] = rel_ptr;
2991 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
2992 (void *)(unsigned long)ipath->fspath, size);
2999 btrfs_free_path(path);
3006 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3008 struct btrfs_data_container *inodes = ctx;
3009 const size_t c = 3 * sizeof(u64);
3011 if (inodes->bytes_left >= c) {
3012 inodes->bytes_left -= c;
3013 inodes->val[inodes->elem_cnt] = inum;
3014 inodes->val[inodes->elem_cnt + 1] = offset;
3015 inodes->val[inodes->elem_cnt + 2] = root;
3016 inodes->elem_cnt += 3;
3018 inodes->bytes_missing += c - inodes->bytes_left;
3019 inodes->bytes_left = 0;
3020 inodes->elem_missed += 3;
3026 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3031 u64 extent_item_pos;
3032 struct btrfs_ioctl_logical_ino_args *loi;
3033 struct btrfs_data_container *inodes = NULL;
3034 struct btrfs_path *path = NULL;
3035 struct btrfs_key key;
3037 if (!capable(CAP_SYS_ADMIN))
3040 loi = memdup_user(arg, sizeof(*loi));
3047 path = btrfs_alloc_path();
3053 size = min_t(u32, loi->size, 4096);
3054 inodes = init_data_container(size);
3055 if (IS_ERR(inodes)) {
3056 ret = PTR_ERR(inodes);
3061 ret = extent_from_logical(root->fs_info, loi->logical, path, &key);
3062 btrfs_release_path(path);
3064 if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK)
3069 extent_item_pos = loi->logical - key.objectid;
3070 ret = iterate_extent_inodes(root->fs_info, path, key.objectid,
3071 extent_item_pos, build_ino_list,
3077 ret = copy_to_user((void *)(unsigned long)loi->inodes,
3078 (void *)(unsigned long)inodes, size);
3083 btrfs_free_path(path);
3090 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
3091 struct btrfs_ioctl_balance_args *bargs)
3093 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3095 bargs->flags = bctl->flags;
3097 if (atomic_read(&fs_info->balance_running))
3098 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3099 if (atomic_read(&fs_info->balance_pause_req))
3100 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3101 if (atomic_read(&fs_info->balance_cancel_req))
3102 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3104 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3105 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3106 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3109 spin_lock(&fs_info->balance_lock);
3110 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3111 spin_unlock(&fs_info->balance_lock);
3113 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3117 static long btrfs_ioctl_balance(struct btrfs_root *root, void __user *arg)
3119 struct btrfs_fs_info *fs_info = root->fs_info;
3120 struct btrfs_ioctl_balance_args *bargs;
3121 struct btrfs_balance_control *bctl;
3124 if (!capable(CAP_SYS_ADMIN))
3127 if (fs_info->sb->s_flags & MS_RDONLY)
3130 mutex_lock(&fs_info->volume_mutex);
3131 mutex_lock(&fs_info->balance_mutex);
3134 bargs = memdup_user(arg, sizeof(*bargs));
3135 if (IS_ERR(bargs)) {
3136 ret = PTR_ERR(bargs);
3140 if (bargs->flags & BTRFS_BALANCE_RESUME) {
3141 if (!fs_info->balance_ctl) {
3146 bctl = fs_info->balance_ctl;
3147 spin_lock(&fs_info->balance_lock);
3148 bctl->flags |= BTRFS_BALANCE_RESUME;
3149 spin_unlock(&fs_info->balance_lock);
3157 if (fs_info->balance_ctl) {
3162 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3168 bctl->fs_info = fs_info;
3170 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3171 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3172 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3174 bctl->flags = bargs->flags;
3176 /* balance everything - no filters */
3177 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
3181 ret = btrfs_balance(bctl, bargs);
3183 * bctl is freed in __cancel_balance or in free_fs_info if
3184 * restriper was paused all the way until unmount
3187 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3194 mutex_unlock(&fs_info->balance_mutex);
3195 mutex_unlock(&fs_info->volume_mutex);
3199 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
3201 if (!capable(CAP_SYS_ADMIN))
3205 case BTRFS_BALANCE_CTL_PAUSE:
3206 return btrfs_pause_balance(root->fs_info);
3207 case BTRFS_BALANCE_CTL_CANCEL:
3208 return btrfs_cancel_balance(root->fs_info);
3214 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
3217 struct btrfs_fs_info *fs_info = root->fs_info;
3218 struct btrfs_ioctl_balance_args *bargs;
3221 if (!capable(CAP_SYS_ADMIN))
3224 mutex_lock(&fs_info->balance_mutex);
3225 if (!fs_info->balance_ctl) {
3230 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
3236 update_ioctl_balance_args(fs_info, 1, bargs);
3238 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3243 mutex_unlock(&fs_info->balance_mutex);
3247 long btrfs_ioctl(struct file *file, unsigned int
3248 cmd, unsigned long arg)
3250 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3251 void __user *argp = (void __user *)arg;
3254 case FS_IOC_GETFLAGS:
3255 return btrfs_ioctl_getflags(file, argp);
3256 case FS_IOC_SETFLAGS:
3257 return btrfs_ioctl_setflags(file, argp);
3258 case FS_IOC_GETVERSION:
3259 return btrfs_ioctl_getversion(file, argp);
3261 return btrfs_ioctl_fitrim(file, argp);
3262 case BTRFS_IOC_SNAP_CREATE:
3263 return btrfs_ioctl_snap_create(file, argp, 0);
3264 case BTRFS_IOC_SNAP_CREATE_V2:
3265 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3266 case BTRFS_IOC_SUBVOL_CREATE:
3267 return btrfs_ioctl_snap_create(file, argp, 1);
3268 case BTRFS_IOC_SNAP_DESTROY:
3269 return btrfs_ioctl_snap_destroy(file, argp);
3270 case BTRFS_IOC_SUBVOL_GETFLAGS:
3271 return btrfs_ioctl_subvol_getflags(file, argp);
3272 case BTRFS_IOC_SUBVOL_SETFLAGS:
3273 return btrfs_ioctl_subvol_setflags(file, argp);
3274 case BTRFS_IOC_DEFAULT_SUBVOL:
3275 return btrfs_ioctl_default_subvol(file, argp);
3276 case BTRFS_IOC_DEFRAG:
3277 return btrfs_ioctl_defrag(file, NULL);
3278 case BTRFS_IOC_DEFRAG_RANGE:
3279 return btrfs_ioctl_defrag(file, argp);
3280 case BTRFS_IOC_RESIZE:
3281 return btrfs_ioctl_resize(root, argp);
3282 case BTRFS_IOC_ADD_DEV:
3283 return btrfs_ioctl_add_dev(root, argp);
3284 case BTRFS_IOC_RM_DEV:
3285 return btrfs_ioctl_rm_dev(root, argp);
3286 case BTRFS_IOC_FS_INFO:
3287 return btrfs_ioctl_fs_info(root, argp);
3288 case BTRFS_IOC_DEV_INFO:
3289 return btrfs_ioctl_dev_info(root, argp);
3290 case BTRFS_IOC_BALANCE:
3291 return btrfs_ioctl_balance(root, NULL);
3292 case BTRFS_IOC_CLONE:
3293 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
3294 case BTRFS_IOC_CLONE_RANGE:
3295 return btrfs_ioctl_clone_range(file, argp);
3296 case BTRFS_IOC_TRANS_START:
3297 return btrfs_ioctl_trans_start(file);
3298 case BTRFS_IOC_TRANS_END:
3299 return btrfs_ioctl_trans_end(file);
3300 case BTRFS_IOC_TREE_SEARCH:
3301 return btrfs_ioctl_tree_search(file, argp);
3302 case BTRFS_IOC_INO_LOOKUP:
3303 return btrfs_ioctl_ino_lookup(file, argp);
3304 case BTRFS_IOC_INO_PATHS:
3305 return btrfs_ioctl_ino_to_path(root, argp);
3306 case BTRFS_IOC_LOGICAL_INO:
3307 return btrfs_ioctl_logical_to_ino(root, argp);
3308 case BTRFS_IOC_SPACE_INFO:
3309 return btrfs_ioctl_space_info(root, argp);
3310 case BTRFS_IOC_SYNC:
3311 btrfs_sync_fs(file->f_dentry->d_sb, 1);
3313 case BTRFS_IOC_START_SYNC:
3314 return btrfs_ioctl_start_sync(file, argp);
3315 case BTRFS_IOC_WAIT_SYNC:
3316 return btrfs_ioctl_wait_sync(file, argp);
3317 case BTRFS_IOC_SCRUB:
3318 return btrfs_ioctl_scrub(root, argp);
3319 case BTRFS_IOC_SCRUB_CANCEL:
3320 return btrfs_ioctl_scrub_cancel(root, argp);
3321 case BTRFS_IOC_SCRUB_PROGRESS:
3322 return btrfs_ioctl_scrub_progress(root, argp);
3323 case BTRFS_IOC_BALANCE_V2:
3324 return btrfs_ioctl_balance(root, argp);
3325 case BTRFS_IOC_BALANCE_CTL:
3326 return btrfs_ioctl_balance_ctl(root, arg);
3327 case BTRFS_IOC_BALANCE_PROGRESS:
3328 return btrfs_ioctl_balance_progress(root, argp);