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 if (btrfs_root_readonly(root))
183 if (copy_from_user(&flags, arg, sizeof(flags)))
186 ret = check_flags(flags);
190 if (!inode_owner_or_capable(inode))
193 mutex_lock(&inode->i_mutex);
195 flags = btrfs_mask_flags(inode->i_mode, flags);
196 oldflags = btrfs_flags_to_ioctl(ip->flags);
197 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
198 if (!capable(CAP_LINUX_IMMUTABLE)) {
204 ret = mnt_want_write(file->f_path.mnt);
208 if (flags & FS_SYNC_FL)
209 ip->flags |= BTRFS_INODE_SYNC;
211 ip->flags &= ~BTRFS_INODE_SYNC;
212 if (flags & FS_IMMUTABLE_FL)
213 ip->flags |= BTRFS_INODE_IMMUTABLE;
215 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
216 if (flags & FS_APPEND_FL)
217 ip->flags |= BTRFS_INODE_APPEND;
219 ip->flags &= ~BTRFS_INODE_APPEND;
220 if (flags & FS_NODUMP_FL)
221 ip->flags |= BTRFS_INODE_NODUMP;
223 ip->flags &= ~BTRFS_INODE_NODUMP;
224 if (flags & FS_NOATIME_FL)
225 ip->flags |= BTRFS_INODE_NOATIME;
227 ip->flags &= ~BTRFS_INODE_NOATIME;
228 if (flags & FS_DIRSYNC_FL)
229 ip->flags |= BTRFS_INODE_DIRSYNC;
231 ip->flags &= ~BTRFS_INODE_DIRSYNC;
232 if (flags & FS_NOCOW_FL)
233 ip->flags |= BTRFS_INODE_NODATACOW;
235 ip->flags &= ~BTRFS_INODE_NODATACOW;
238 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
239 * flag may be changed automatically if compression code won't make
242 if (flags & FS_NOCOMP_FL) {
243 ip->flags &= ~BTRFS_INODE_COMPRESS;
244 ip->flags |= BTRFS_INODE_NOCOMPRESS;
245 } else if (flags & FS_COMPR_FL) {
246 ip->flags |= BTRFS_INODE_COMPRESS;
247 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
249 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
252 trans = btrfs_join_transaction(root);
253 BUG_ON(IS_ERR(trans));
255 btrfs_update_iflags(inode);
256 inode->i_ctime = CURRENT_TIME;
257 ret = btrfs_update_inode(trans, root, inode);
260 btrfs_end_transaction(trans, root);
262 mnt_drop_write(file->f_path.mnt);
266 mutex_unlock(&inode->i_mutex);
270 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
272 struct inode *inode = file->f_path.dentry->d_inode;
274 return put_user(inode->i_generation, arg);
277 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
279 struct btrfs_root *root = fdentry(file)->d_sb->s_fs_info;
280 struct btrfs_fs_info *fs_info = root->fs_info;
281 struct btrfs_device *device;
282 struct request_queue *q;
283 struct fstrim_range range;
284 u64 minlen = ULLONG_MAX;
286 u64 total_bytes = btrfs_super_total_bytes(root->fs_info->super_copy);
289 if (!capable(CAP_SYS_ADMIN))
293 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
297 q = bdev_get_queue(device->bdev);
298 if (blk_queue_discard(q)) {
300 minlen = min((u64)q->limits.discard_granularity,
308 if (copy_from_user(&range, arg, sizeof(range)))
310 if (range.start > total_bytes)
313 range.len = min(range.len, total_bytes - range.start);
314 range.minlen = max(range.minlen, minlen);
315 ret = btrfs_trim_fs(root, &range);
319 if (copy_to_user(arg, &range, sizeof(range)))
325 static noinline int create_subvol(struct btrfs_root *root,
326 struct dentry *dentry,
327 char *name, int namelen,
330 struct btrfs_trans_handle *trans;
331 struct btrfs_key key;
332 struct btrfs_root_item root_item;
333 struct btrfs_inode_item *inode_item;
334 struct extent_buffer *leaf;
335 struct btrfs_root *new_root;
336 struct dentry *parent = dentry->d_parent;
341 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
344 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
348 dir = parent->d_inode;
356 trans = btrfs_start_transaction(root, 6);
358 return PTR_ERR(trans);
360 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
361 0, objectid, NULL, 0, 0, 0);
367 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
368 btrfs_set_header_bytenr(leaf, leaf->start);
369 btrfs_set_header_generation(leaf, trans->transid);
370 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
371 btrfs_set_header_owner(leaf, objectid);
373 write_extent_buffer(leaf, root->fs_info->fsid,
374 (unsigned long)btrfs_header_fsid(leaf),
376 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
377 (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
379 btrfs_mark_buffer_dirty(leaf);
381 inode_item = &root_item.inode;
382 memset(inode_item, 0, sizeof(*inode_item));
383 inode_item->generation = cpu_to_le64(1);
384 inode_item->size = cpu_to_le64(3);
385 inode_item->nlink = cpu_to_le32(1);
386 inode_item->nbytes = cpu_to_le64(root->leafsize);
387 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
390 root_item.byte_limit = 0;
391 inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
393 btrfs_set_root_bytenr(&root_item, leaf->start);
394 btrfs_set_root_generation(&root_item, trans->transid);
395 btrfs_set_root_level(&root_item, 0);
396 btrfs_set_root_refs(&root_item, 1);
397 btrfs_set_root_used(&root_item, leaf->len);
398 btrfs_set_root_last_snapshot(&root_item, 0);
400 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
401 root_item.drop_level = 0;
403 btrfs_tree_unlock(leaf);
404 free_extent_buffer(leaf);
407 btrfs_set_root_dirid(&root_item, new_dirid);
409 key.objectid = objectid;
411 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
412 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
417 key.offset = (u64)-1;
418 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
419 BUG_ON(IS_ERR(new_root));
421 btrfs_record_root_in_trans(trans, new_root);
423 ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
425 * insert the directory item
427 ret = btrfs_set_inode_index(dir, &index);
430 ret = btrfs_insert_dir_item(trans, root,
431 name, namelen, dir, &key,
432 BTRFS_FT_DIR, index);
436 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
437 ret = btrfs_update_inode(trans, root, dir);
440 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
441 objectid, root->root_key.objectid,
442 btrfs_ino(dir), index, name, namelen);
446 d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
449 *async_transid = trans->transid;
450 err = btrfs_commit_transaction_async(trans, root, 1);
452 err = btrfs_commit_transaction(trans, root);
459 static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
460 char *name, int namelen, u64 *async_transid,
464 struct btrfs_pending_snapshot *pending_snapshot;
465 struct btrfs_trans_handle *trans;
471 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
472 if (!pending_snapshot)
475 btrfs_init_block_rsv(&pending_snapshot->block_rsv);
476 pending_snapshot->dentry = dentry;
477 pending_snapshot->root = root;
478 pending_snapshot->readonly = readonly;
480 trans = btrfs_start_transaction(root->fs_info->extent_root, 5);
482 ret = PTR_ERR(trans);
486 ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
489 spin_lock(&root->fs_info->trans_lock);
490 list_add(&pending_snapshot->list,
491 &trans->transaction->pending_snapshots);
492 spin_unlock(&root->fs_info->trans_lock);
494 *async_transid = trans->transid;
495 ret = btrfs_commit_transaction_async(trans,
496 root->fs_info->extent_root, 1);
498 ret = btrfs_commit_transaction(trans,
499 root->fs_info->extent_root);
503 ret = pending_snapshot->error;
507 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
511 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
513 ret = PTR_ERR(inode);
517 d_instantiate(dentry, inode);
520 kfree(pending_snapshot);
524 /* copy of check_sticky in fs/namei.c()
525 * It's inline, so penalty for filesystems that don't use sticky bit is
528 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
530 uid_t fsuid = current_fsuid();
532 if (!(dir->i_mode & S_ISVTX))
534 if (inode->i_uid == fsuid)
536 if (dir->i_uid == fsuid)
538 return !capable(CAP_FOWNER);
541 /* copy of may_delete in fs/namei.c()
542 * Check whether we can remove a link victim from directory dir, check
543 * whether the type of victim is right.
544 * 1. We can't do it if dir is read-only (done in permission())
545 * 2. We should have write and exec permissions on dir
546 * 3. We can't remove anything from append-only dir
547 * 4. We can't do anything with immutable dir (done in permission())
548 * 5. If the sticky bit on dir is set we should either
549 * a. be owner of dir, or
550 * b. be owner of victim, or
551 * c. have CAP_FOWNER capability
552 * 6. If the victim is append-only or immutable we can't do antyhing with
553 * links pointing to it.
554 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
555 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
556 * 9. We can't remove a root or mountpoint.
557 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
558 * nfs_async_unlink().
561 static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
565 if (!victim->d_inode)
568 BUG_ON(victim->d_parent->d_inode != dir);
569 audit_inode_child(victim, dir);
571 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
576 if (btrfs_check_sticky(dir, victim->d_inode)||
577 IS_APPEND(victim->d_inode)||
578 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
581 if (!S_ISDIR(victim->d_inode->i_mode))
585 } else if (S_ISDIR(victim->d_inode->i_mode))
589 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
594 /* copy of may_create in fs/namei.c() */
595 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
601 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
605 * Create a new subvolume below @parent. This is largely modeled after
606 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
607 * inside this filesystem so it's quite a bit simpler.
609 static noinline int btrfs_mksubvol(struct path *parent,
610 char *name, int namelen,
611 struct btrfs_root *snap_src,
612 u64 *async_transid, bool readonly)
614 struct inode *dir = parent->dentry->d_inode;
615 struct dentry *dentry;
618 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
620 dentry = lookup_one_len(name, parent->dentry, namelen);
621 error = PTR_ERR(dentry);
629 error = mnt_want_write(parent->mnt);
633 error = btrfs_may_create(dir, dentry);
637 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
639 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
643 error = create_snapshot(snap_src, dentry,
644 name, namelen, async_transid, readonly);
646 error = create_subvol(BTRFS_I(dir)->root, dentry,
647 name, namelen, async_transid);
650 fsnotify_mkdir(dir, dentry);
652 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
654 mnt_drop_write(parent->mnt);
658 mutex_unlock(&dir->i_mutex);
663 * When we're defragging a range, we don't want to kick it off again
664 * if it is really just waiting for delalloc to send it down.
665 * If we find a nice big extent or delalloc range for the bytes in the
666 * file you want to defrag, we return 0 to let you know to skip this
669 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
671 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
672 struct extent_map *em = NULL;
673 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
676 read_lock(&em_tree->lock);
677 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
678 read_unlock(&em_tree->lock);
681 end = extent_map_end(em);
683 if (end - offset > thresh)
686 /* if we already have a nice delalloc here, just stop */
688 end = count_range_bits(io_tree, &offset, offset + thresh,
689 thresh, EXTENT_DELALLOC, 1);
696 * helper function to walk through a file and find extents
697 * newer than a specific transid, and smaller than thresh.
699 * This is used by the defragging code to find new and small
702 static int find_new_extents(struct btrfs_root *root,
703 struct inode *inode, u64 newer_than,
704 u64 *off, int thresh)
706 struct btrfs_path *path;
707 struct btrfs_key min_key;
708 struct btrfs_key max_key;
709 struct extent_buffer *leaf;
710 struct btrfs_file_extent_item *extent;
713 u64 ino = btrfs_ino(inode);
715 path = btrfs_alloc_path();
719 min_key.objectid = ino;
720 min_key.type = BTRFS_EXTENT_DATA_KEY;
721 min_key.offset = *off;
723 max_key.objectid = ino;
724 max_key.type = (u8)-1;
725 max_key.offset = (u64)-1;
727 path->keep_locks = 1;
730 ret = btrfs_search_forward(root, &min_key, &max_key,
731 path, 0, newer_than);
734 if (min_key.objectid != ino)
736 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
739 leaf = path->nodes[0];
740 extent = btrfs_item_ptr(leaf, path->slots[0],
741 struct btrfs_file_extent_item);
743 type = btrfs_file_extent_type(leaf, extent);
744 if (type == BTRFS_FILE_EXTENT_REG &&
745 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
746 check_defrag_in_cache(inode, min_key.offset, thresh)) {
747 *off = min_key.offset;
748 btrfs_free_path(path);
752 if (min_key.offset == (u64)-1)
756 btrfs_release_path(path);
759 btrfs_free_path(path);
763 static int should_defrag_range(struct inode *inode, u64 start, u64 len,
764 int thresh, u64 *last_len, u64 *skip,
767 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
768 struct extent_map *em = NULL;
769 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
773 * make sure that once we start defragging an extent, we keep on
776 if (start < *defrag_end)
782 * hopefully we have this extent in the tree already, try without
783 * the full extent lock
785 read_lock(&em_tree->lock);
786 em = lookup_extent_mapping(em_tree, start, len);
787 read_unlock(&em_tree->lock);
790 /* get the big lock and read metadata off disk */
791 lock_extent(io_tree, start, start + len - 1, GFP_NOFS);
792 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
793 unlock_extent(io_tree, start, start + len - 1, GFP_NOFS);
799 /* this will cover holes, and inline extents */
800 if (em->block_start >= EXTENT_MAP_LAST_BYTE)
804 * we hit a real extent, if it is big don't bother defragging it again
806 if ((*last_len == 0 || *last_len >= thresh) && em->len >= thresh)
810 * last_len ends up being a counter of how many bytes we've defragged.
811 * every time we choose not to defrag an extent, we reset *last_len
812 * so that the next tiny extent will force a defrag.
814 * The end result of this is that tiny extents before a single big
815 * extent will force at least part of that big extent to be defragged.
818 *defrag_end = extent_map_end(em);
821 *skip = extent_map_end(em);
830 * it doesn't do much good to defrag one or two pages
831 * at a time. This pulls in a nice chunk of pages
834 * It also makes sure the delalloc code has enough
835 * dirty data to avoid making new small extents as part
838 * It's a good idea to start RA on this range
839 * before calling this.
841 static int cluster_pages_for_defrag(struct inode *inode,
843 unsigned long start_index,
846 unsigned long file_end;
847 u64 isize = i_size_read(inode);
853 struct btrfs_ordered_extent *ordered;
854 struct extent_state *cached_state = NULL;
855 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
859 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
861 mutex_lock(&inode->i_mutex);
862 ret = btrfs_delalloc_reserve_space(inode,
863 num_pages << PAGE_CACHE_SHIFT);
864 mutex_unlock(&inode->i_mutex);
871 /* step one, lock all the pages */
872 for (i = 0; i < num_pages; i++) {
874 page = find_or_create_page(inode->i_mapping,
875 start_index + i, mask);
879 if (!PageUptodate(page)) {
880 btrfs_readpage(NULL, page);
882 if (!PageUptodate(page)) {
884 page_cache_release(page);
889 isize = i_size_read(inode);
890 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
891 if (!isize || page->index > file_end ||
892 page->mapping != inode->i_mapping) {
893 /* whoops, we blew past eof, skip this page */
895 page_cache_release(page);
904 if (!(inode->i_sb->s_flags & MS_ACTIVE))
908 * so now we have a nice long stream of locked
909 * and up to date pages, lets wait on them
911 for (i = 0; i < i_done; i++)
912 wait_on_page_writeback(pages[i]);
914 page_start = page_offset(pages[0]);
915 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
917 lock_extent_bits(&BTRFS_I(inode)->io_tree,
918 page_start, page_end - 1, 0, &cached_state,
920 ordered = btrfs_lookup_first_ordered_extent(inode, page_end - 1);
922 ordered->file_offset + ordered->len > page_start &&
923 ordered->file_offset < page_end) {
924 btrfs_put_ordered_extent(ordered);
925 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
926 page_start, page_end - 1,
927 &cached_state, GFP_NOFS);
928 for (i = 0; i < i_done; i++) {
929 unlock_page(pages[i]);
930 page_cache_release(pages[i]);
932 btrfs_wait_ordered_range(inode, page_start,
933 page_end - page_start);
937 btrfs_put_ordered_extent(ordered);
939 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
940 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
941 EXTENT_DO_ACCOUNTING, 0, 0, &cached_state,
944 if (i_done != num_pages) {
945 spin_lock(&BTRFS_I(inode)->lock);
946 BTRFS_I(inode)->outstanding_extents++;
947 spin_unlock(&BTRFS_I(inode)->lock);
948 btrfs_delalloc_release_space(inode,
949 (num_pages - i_done) << PAGE_CACHE_SHIFT);
953 btrfs_set_extent_delalloc(inode, page_start, page_end - 1,
956 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
957 page_start, page_end - 1, &cached_state,
960 for (i = 0; i < i_done; i++) {
961 clear_page_dirty_for_io(pages[i]);
962 ClearPageChecked(pages[i]);
963 set_page_extent_mapped(pages[i]);
964 set_page_dirty(pages[i]);
965 unlock_page(pages[i]);
966 page_cache_release(pages[i]);
970 for (i = 0; i < i_done; i++) {
971 unlock_page(pages[i]);
972 page_cache_release(pages[i]);
974 btrfs_delalloc_release_space(inode, num_pages << PAGE_CACHE_SHIFT);
979 int btrfs_defrag_file(struct inode *inode, struct file *file,
980 struct btrfs_ioctl_defrag_range_args *range,
981 u64 newer_than, unsigned long max_to_defrag)
983 struct btrfs_root *root = BTRFS_I(inode)->root;
984 struct btrfs_super_block *disk_super;
985 struct file_ra_state *ra = NULL;
986 unsigned long last_index;
987 u64 isize = i_size_read(inode);
992 u64 newer_off = range->start;
994 unsigned long ra_index = 0;
996 int defrag_count = 0;
997 int compress_type = BTRFS_COMPRESS_ZLIB;
998 int extent_thresh = range->extent_thresh;
999 int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1000 int cluster = max_cluster;
1001 u64 new_align = ~((u64)128 * 1024 - 1);
1002 struct page **pages = NULL;
1004 if (extent_thresh == 0)
1005 extent_thresh = 256 * 1024;
1007 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1008 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1010 if (range->compress_type)
1011 compress_type = range->compress_type;
1018 * if we were not given a file, allocate a readahead
1022 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1025 file_ra_state_init(ra, inode->i_mapping);
1030 pages = kmalloc(sizeof(struct page *) * max_cluster,
1037 /* find the last page to defrag */
1038 if (range->start + range->len > range->start) {
1039 last_index = min_t(u64, isize - 1,
1040 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1042 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1046 ret = find_new_extents(root, inode, newer_than,
1047 &newer_off, 64 * 1024);
1049 range->start = newer_off;
1051 * we always align our defrag to help keep
1052 * the extents in the file evenly spaced
1054 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1058 i = range->start >> PAGE_CACHE_SHIFT;
1061 max_to_defrag = last_index;
1064 * make writeback starts from i, so the defrag range can be
1065 * written sequentially.
1067 if (i < inode->i_mapping->writeback_index)
1068 inode->i_mapping->writeback_index = i;
1070 while (i <= last_index && defrag_count < max_to_defrag &&
1071 (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1072 PAGE_CACHE_SHIFT)) {
1074 * make sure we stop running if someone unmounts
1077 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1081 !should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1088 * the should_defrag function tells us how much to skip
1089 * bump our counter by the suggested amount
1091 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1092 i = max(i + 1, next);
1097 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1098 PAGE_CACHE_SHIFT) - i;
1099 cluster = min(cluster, max_cluster);
1101 cluster = max_cluster;
1104 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1105 BTRFS_I(inode)->force_compress = compress_type;
1107 if (i + cluster > ra_index) {
1108 ra_index = max(i, ra_index);
1109 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1111 ra_index += max_cluster;
1114 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1118 defrag_count += ret;
1119 balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret);
1122 if (newer_off == (u64)-1)
1125 newer_off = max(newer_off + 1,
1126 (u64)i << PAGE_CACHE_SHIFT);
1128 ret = find_new_extents(root, inode,
1129 newer_than, &newer_off,
1132 range->start = newer_off;
1133 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1140 last_len += ret << PAGE_CACHE_SHIFT;
1148 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1149 filemap_flush(inode->i_mapping);
1151 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1152 /* the filemap_flush will queue IO into the worker threads, but
1153 * we have to make sure the IO is actually started and that
1154 * ordered extents get created before we return
1156 atomic_inc(&root->fs_info->async_submit_draining);
1157 while (atomic_read(&root->fs_info->nr_async_submits) ||
1158 atomic_read(&root->fs_info->async_delalloc_pages)) {
1159 wait_event(root->fs_info->async_submit_wait,
1160 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1161 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1163 atomic_dec(&root->fs_info->async_submit_draining);
1165 mutex_lock(&inode->i_mutex);
1166 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1167 mutex_unlock(&inode->i_mutex);
1170 disk_super = root->fs_info->super_copy;
1171 features = btrfs_super_incompat_flags(disk_super);
1172 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1173 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
1174 btrfs_set_super_incompat_flags(disk_super, features);
1186 static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
1192 struct btrfs_ioctl_vol_args *vol_args;
1193 struct btrfs_trans_handle *trans;
1194 struct btrfs_device *device = NULL;
1196 char *devstr = NULL;
1200 if (root->fs_info->sb->s_flags & MS_RDONLY)
1203 if (!capable(CAP_SYS_ADMIN))
1206 mutex_lock(&root->fs_info->volume_mutex);
1207 if (root->fs_info->balance_ctl) {
1208 printk(KERN_INFO "btrfs: balance in progress\n");
1213 vol_args = memdup_user(arg, sizeof(*vol_args));
1214 if (IS_ERR(vol_args)) {
1215 ret = PTR_ERR(vol_args);
1219 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1221 sizestr = vol_args->name;
1222 devstr = strchr(sizestr, ':');
1225 sizestr = devstr + 1;
1227 devstr = vol_args->name;
1228 devid = simple_strtoull(devstr, &end, 10);
1229 printk(KERN_INFO "btrfs: resizing devid %llu\n",
1230 (unsigned long long)devid);
1232 device = btrfs_find_device(root, devid, NULL, NULL);
1234 printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
1235 (unsigned long long)devid);
1239 if (!strcmp(sizestr, "max"))
1240 new_size = device->bdev->bd_inode->i_size;
1242 if (sizestr[0] == '-') {
1245 } else if (sizestr[0] == '+') {
1249 new_size = memparse(sizestr, NULL);
1250 if (new_size == 0) {
1256 old_size = device->total_bytes;
1259 if (new_size > old_size) {
1263 new_size = old_size - new_size;
1264 } else if (mod > 0) {
1265 new_size = old_size + new_size;
1268 if (new_size < 256 * 1024 * 1024) {
1272 if (new_size > device->bdev->bd_inode->i_size) {
1277 do_div(new_size, root->sectorsize);
1278 new_size *= root->sectorsize;
1280 printk(KERN_INFO "btrfs: new size for %s is %llu\n",
1281 device->name, (unsigned long long)new_size);
1283 if (new_size > old_size) {
1284 trans = btrfs_start_transaction(root, 0);
1285 if (IS_ERR(trans)) {
1286 ret = PTR_ERR(trans);
1289 ret = btrfs_grow_device(trans, device, new_size);
1290 btrfs_commit_transaction(trans, root);
1291 } else if (new_size < old_size) {
1292 ret = btrfs_shrink_device(device, new_size);
1298 mutex_unlock(&root->fs_info->volume_mutex);
1302 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1309 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
1310 struct file *src_file;
1314 if (root->fs_info->sb->s_flags & MS_RDONLY)
1317 namelen = strlen(name);
1318 if (strchr(name, '/')) {
1324 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1325 NULL, transid, readonly);
1327 struct inode *src_inode;
1328 src_file = fget(fd);
1334 src_inode = src_file->f_path.dentry->d_inode;
1335 if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
1336 printk(KERN_INFO "btrfs: Snapshot src from "
1342 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1343 BTRFS_I(src_inode)->root,
1351 static noinline int btrfs_ioctl_snap_create(struct file *file,
1352 void __user *arg, int subvol)
1354 struct btrfs_ioctl_vol_args *vol_args;
1357 vol_args = memdup_user(arg, sizeof(*vol_args));
1358 if (IS_ERR(vol_args))
1359 return PTR_ERR(vol_args);
1360 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1362 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1363 vol_args->fd, subvol,
1370 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1371 void __user *arg, int subvol)
1373 struct btrfs_ioctl_vol_args_v2 *vol_args;
1377 bool readonly = false;
1379 vol_args = memdup_user(arg, sizeof(*vol_args));
1380 if (IS_ERR(vol_args))
1381 return PTR_ERR(vol_args);
1382 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1384 if (vol_args->flags &
1385 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY)) {
1390 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1392 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1395 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1396 vol_args->fd, subvol,
1399 if (ret == 0 && ptr &&
1401 offsetof(struct btrfs_ioctl_vol_args_v2,
1402 transid), ptr, sizeof(*ptr)))
1409 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1412 struct inode *inode = fdentry(file)->d_inode;
1413 struct btrfs_root *root = BTRFS_I(inode)->root;
1417 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1420 down_read(&root->fs_info->subvol_sem);
1421 if (btrfs_root_readonly(root))
1422 flags |= BTRFS_SUBVOL_RDONLY;
1423 up_read(&root->fs_info->subvol_sem);
1425 if (copy_to_user(arg, &flags, sizeof(flags)))
1431 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1434 struct inode *inode = fdentry(file)->d_inode;
1435 struct btrfs_root *root = BTRFS_I(inode)->root;
1436 struct btrfs_trans_handle *trans;
1441 if (root->fs_info->sb->s_flags & MS_RDONLY)
1444 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1447 if (copy_from_user(&flags, arg, sizeof(flags)))
1450 if (flags & BTRFS_SUBVOL_CREATE_ASYNC)
1453 if (flags & ~BTRFS_SUBVOL_RDONLY)
1456 if (!inode_owner_or_capable(inode))
1459 down_write(&root->fs_info->subvol_sem);
1462 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1465 root_flags = btrfs_root_flags(&root->root_item);
1466 if (flags & BTRFS_SUBVOL_RDONLY)
1467 btrfs_set_root_flags(&root->root_item,
1468 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1470 btrfs_set_root_flags(&root->root_item,
1471 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1473 trans = btrfs_start_transaction(root, 1);
1474 if (IS_ERR(trans)) {
1475 ret = PTR_ERR(trans);
1479 ret = btrfs_update_root(trans, root->fs_info->tree_root,
1480 &root->root_key, &root->root_item);
1482 btrfs_commit_transaction(trans, root);
1485 btrfs_set_root_flags(&root->root_item, root_flags);
1487 up_write(&root->fs_info->subvol_sem);
1492 * helper to check if the subvolume references other subvolumes
1494 static noinline int may_destroy_subvol(struct btrfs_root *root)
1496 struct btrfs_path *path;
1497 struct btrfs_key key;
1500 path = btrfs_alloc_path();
1504 key.objectid = root->root_key.objectid;
1505 key.type = BTRFS_ROOT_REF_KEY;
1506 key.offset = (u64)-1;
1508 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1515 if (path->slots[0] > 0) {
1517 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1518 if (key.objectid == root->root_key.objectid &&
1519 key.type == BTRFS_ROOT_REF_KEY)
1523 btrfs_free_path(path);
1527 static noinline int key_in_sk(struct btrfs_key *key,
1528 struct btrfs_ioctl_search_key *sk)
1530 struct btrfs_key test;
1533 test.objectid = sk->min_objectid;
1534 test.type = sk->min_type;
1535 test.offset = sk->min_offset;
1537 ret = btrfs_comp_cpu_keys(key, &test);
1541 test.objectid = sk->max_objectid;
1542 test.type = sk->max_type;
1543 test.offset = sk->max_offset;
1545 ret = btrfs_comp_cpu_keys(key, &test);
1551 static noinline int copy_to_sk(struct btrfs_root *root,
1552 struct btrfs_path *path,
1553 struct btrfs_key *key,
1554 struct btrfs_ioctl_search_key *sk,
1556 unsigned long *sk_offset,
1560 struct extent_buffer *leaf;
1561 struct btrfs_ioctl_search_header sh;
1562 unsigned long item_off;
1563 unsigned long item_len;
1569 leaf = path->nodes[0];
1570 slot = path->slots[0];
1571 nritems = btrfs_header_nritems(leaf);
1573 if (btrfs_header_generation(leaf) > sk->max_transid) {
1577 found_transid = btrfs_header_generation(leaf);
1579 for (i = slot; i < nritems; i++) {
1580 item_off = btrfs_item_ptr_offset(leaf, i);
1581 item_len = btrfs_item_size_nr(leaf, i);
1583 if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1586 if (sizeof(sh) + item_len + *sk_offset >
1587 BTRFS_SEARCH_ARGS_BUFSIZE) {
1592 btrfs_item_key_to_cpu(leaf, key, i);
1593 if (!key_in_sk(key, sk))
1596 sh.objectid = key->objectid;
1597 sh.offset = key->offset;
1598 sh.type = key->type;
1600 sh.transid = found_transid;
1602 /* copy search result header */
1603 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1604 *sk_offset += sizeof(sh);
1607 char *p = buf + *sk_offset;
1609 read_extent_buffer(leaf, p,
1610 item_off, item_len);
1611 *sk_offset += item_len;
1615 if (*num_found >= sk->nr_items)
1620 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1622 else if (key->type < (u8)-1 && key->type < sk->max_type) {
1625 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1635 static noinline int search_ioctl(struct inode *inode,
1636 struct btrfs_ioctl_search_args *args)
1638 struct btrfs_root *root;
1639 struct btrfs_key key;
1640 struct btrfs_key max_key;
1641 struct btrfs_path *path;
1642 struct btrfs_ioctl_search_key *sk = &args->key;
1643 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1646 unsigned long sk_offset = 0;
1648 path = btrfs_alloc_path();
1652 if (sk->tree_id == 0) {
1653 /* search the root of the inode that was passed */
1654 root = BTRFS_I(inode)->root;
1656 key.objectid = sk->tree_id;
1657 key.type = BTRFS_ROOT_ITEM_KEY;
1658 key.offset = (u64)-1;
1659 root = btrfs_read_fs_root_no_name(info, &key);
1661 printk(KERN_ERR "could not find root %llu\n",
1663 btrfs_free_path(path);
1668 key.objectid = sk->min_objectid;
1669 key.type = sk->min_type;
1670 key.offset = sk->min_offset;
1672 max_key.objectid = sk->max_objectid;
1673 max_key.type = sk->max_type;
1674 max_key.offset = sk->max_offset;
1676 path->keep_locks = 1;
1679 ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1686 ret = copy_to_sk(root, path, &key, sk, args->buf,
1687 &sk_offset, &num_found);
1688 btrfs_release_path(path);
1689 if (ret || num_found >= sk->nr_items)
1695 sk->nr_items = num_found;
1696 btrfs_free_path(path);
1700 static noinline int btrfs_ioctl_tree_search(struct file *file,
1703 struct btrfs_ioctl_search_args *args;
1704 struct inode *inode;
1707 if (!capable(CAP_SYS_ADMIN))
1710 args = memdup_user(argp, sizeof(*args));
1712 return PTR_ERR(args);
1714 inode = fdentry(file)->d_inode;
1715 ret = search_ioctl(inode, args);
1716 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1723 * Search INODE_REFs to identify path name of 'dirid' directory
1724 * in a 'tree_id' tree. and sets path name to 'name'.
1726 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1727 u64 tree_id, u64 dirid, char *name)
1729 struct btrfs_root *root;
1730 struct btrfs_key key;
1736 struct btrfs_inode_ref *iref;
1737 struct extent_buffer *l;
1738 struct btrfs_path *path;
1740 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1745 path = btrfs_alloc_path();
1749 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1751 key.objectid = tree_id;
1752 key.type = BTRFS_ROOT_ITEM_KEY;
1753 key.offset = (u64)-1;
1754 root = btrfs_read_fs_root_no_name(info, &key);
1756 printk(KERN_ERR "could not find root %llu\n", tree_id);
1761 key.objectid = dirid;
1762 key.type = BTRFS_INODE_REF_KEY;
1763 key.offset = (u64)-1;
1766 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1771 slot = path->slots[0];
1772 if (ret > 0 && slot > 0)
1774 btrfs_item_key_to_cpu(l, &key, slot);
1776 if (ret > 0 && (key.objectid != dirid ||
1777 key.type != BTRFS_INODE_REF_KEY)) {
1782 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1783 len = btrfs_inode_ref_name_len(l, iref);
1785 total_len += len + 1;
1790 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1792 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1795 btrfs_release_path(path);
1796 key.objectid = key.offset;
1797 key.offset = (u64)-1;
1798 dirid = key.objectid;
1802 memmove(name, ptr, total_len);
1803 name[total_len]='\0';
1806 btrfs_free_path(path);
1810 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1813 struct btrfs_ioctl_ino_lookup_args *args;
1814 struct inode *inode;
1817 if (!capable(CAP_SYS_ADMIN))
1820 args = memdup_user(argp, sizeof(*args));
1822 return PTR_ERR(args);
1824 inode = fdentry(file)->d_inode;
1826 if (args->treeid == 0)
1827 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1829 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1830 args->treeid, args->objectid,
1833 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1840 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1843 struct dentry *parent = fdentry(file);
1844 struct dentry *dentry;
1845 struct inode *dir = parent->d_inode;
1846 struct inode *inode;
1847 struct btrfs_root *root = BTRFS_I(dir)->root;
1848 struct btrfs_root *dest = NULL;
1849 struct btrfs_ioctl_vol_args *vol_args;
1850 struct btrfs_trans_handle *trans;
1855 vol_args = memdup_user(arg, sizeof(*vol_args));
1856 if (IS_ERR(vol_args))
1857 return PTR_ERR(vol_args);
1859 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1860 namelen = strlen(vol_args->name);
1861 if (strchr(vol_args->name, '/') ||
1862 strncmp(vol_args->name, "..", namelen) == 0) {
1867 err = mnt_want_write(file->f_path.mnt);
1871 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
1872 dentry = lookup_one_len(vol_args->name, parent, namelen);
1873 if (IS_ERR(dentry)) {
1874 err = PTR_ERR(dentry);
1875 goto out_unlock_dir;
1878 if (!dentry->d_inode) {
1883 inode = dentry->d_inode;
1884 dest = BTRFS_I(inode)->root;
1885 if (!capable(CAP_SYS_ADMIN)){
1887 * Regular user. Only allow this with a special mount
1888 * option, when the user has write+exec access to the
1889 * subvol root, and when rmdir(2) would have been
1892 * Note that this is _not_ check that the subvol is
1893 * empty or doesn't contain data that we wouldn't
1894 * otherwise be able to delete.
1896 * Users who want to delete empty subvols should try
1900 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1904 * Do not allow deletion if the parent dir is the same
1905 * as the dir to be deleted. That means the ioctl
1906 * must be called on the dentry referencing the root
1907 * of the subvol, not a random directory contained
1914 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
1918 /* check if subvolume may be deleted by a non-root user */
1919 err = btrfs_may_delete(dir, dentry, 1);
1924 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1929 mutex_lock(&inode->i_mutex);
1930 err = d_invalidate(dentry);
1934 down_write(&root->fs_info->subvol_sem);
1936 err = may_destroy_subvol(dest);
1940 trans = btrfs_start_transaction(root, 0);
1941 if (IS_ERR(trans)) {
1942 err = PTR_ERR(trans);
1945 trans->block_rsv = &root->fs_info->global_block_rsv;
1947 ret = btrfs_unlink_subvol(trans, root, dir,
1948 dest->root_key.objectid,
1949 dentry->d_name.name,
1950 dentry->d_name.len);
1953 btrfs_record_root_in_trans(trans, dest);
1955 memset(&dest->root_item.drop_progress, 0,
1956 sizeof(dest->root_item.drop_progress));
1957 dest->root_item.drop_level = 0;
1958 btrfs_set_root_refs(&dest->root_item, 0);
1960 if (!xchg(&dest->orphan_item_inserted, 1)) {
1961 ret = btrfs_insert_orphan_item(trans,
1962 root->fs_info->tree_root,
1963 dest->root_key.objectid);
1967 ret = btrfs_end_transaction(trans, root);
1969 inode->i_flags |= S_DEAD;
1971 up_write(&root->fs_info->subvol_sem);
1973 mutex_unlock(&inode->i_mutex);
1975 shrink_dcache_sb(root->fs_info->sb);
1976 btrfs_invalidate_inodes(dest);
1982 mutex_unlock(&dir->i_mutex);
1983 mnt_drop_write(file->f_path.mnt);
1989 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
1991 struct inode *inode = fdentry(file)->d_inode;
1992 struct btrfs_root *root = BTRFS_I(inode)->root;
1993 struct btrfs_ioctl_defrag_range_args *range;
1996 if (btrfs_root_readonly(root))
1999 ret = mnt_want_write(file->f_path.mnt);
2003 switch (inode->i_mode & S_IFMT) {
2005 if (!capable(CAP_SYS_ADMIN)) {
2009 ret = btrfs_defrag_root(root, 0);
2012 ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
2015 if (!(file->f_mode & FMODE_WRITE)) {
2020 range = kzalloc(sizeof(*range), GFP_KERNEL);
2027 if (copy_from_user(range, argp,
2033 /* compression requires us to start the IO */
2034 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2035 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2036 range->extent_thresh = (u32)-1;
2039 /* the rest are all set to zero by kzalloc */
2040 range->len = (u64)-1;
2042 ret = btrfs_defrag_file(fdentry(file)->d_inode, file,
2052 mnt_drop_write(file->f_path.mnt);
2056 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2058 struct btrfs_ioctl_vol_args *vol_args;
2061 if (!capable(CAP_SYS_ADMIN))
2064 mutex_lock(&root->fs_info->volume_mutex);
2065 if (root->fs_info->balance_ctl) {
2066 printk(KERN_INFO "btrfs: balance in progress\n");
2071 vol_args = memdup_user(arg, sizeof(*vol_args));
2072 if (IS_ERR(vol_args)) {
2073 ret = PTR_ERR(vol_args);
2077 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2078 ret = btrfs_init_new_device(root, vol_args->name);
2082 mutex_unlock(&root->fs_info->volume_mutex);
2086 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
2088 struct btrfs_ioctl_vol_args *vol_args;
2091 if (!capable(CAP_SYS_ADMIN))
2094 if (root->fs_info->sb->s_flags & MS_RDONLY)
2097 mutex_lock(&root->fs_info->volume_mutex);
2098 if (root->fs_info->balance_ctl) {
2099 printk(KERN_INFO "btrfs: balance in progress\n");
2104 vol_args = memdup_user(arg, sizeof(*vol_args));
2105 if (IS_ERR(vol_args)) {
2106 ret = PTR_ERR(vol_args);
2110 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2111 ret = btrfs_rm_device(root, vol_args->name);
2115 mutex_unlock(&root->fs_info->volume_mutex);
2119 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2121 struct btrfs_ioctl_fs_info_args *fi_args;
2122 struct btrfs_device *device;
2123 struct btrfs_device *next;
2124 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2127 if (!capable(CAP_SYS_ADMIN))
2130 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2134 fi_args->num_devices = fs_devices->num_devices;
2135 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2137 mutex_lock(&fs_devices->device_list_mutex);
2138 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2139 if (device->devid > fi_args->max_id)
2140 fi_args->max_id = device->devid;
2142 mutex_unlock(&fs_devices->device_list_mutex);
2144 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2151 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2153 struct btrfs_ioctl_dev_info_args *di_args;
2154 struct btrfs_device *dev;
2155 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2157 char *s_uuid = NULL;
2158 char empty_uuid[BTRFS_UUID_SIZE] = {0};
2160 if (!capable(CAP_SYS_ADMIN))
2163 di_args = memdup_user(arg, sizeof(*di_args));
2164 if (IS_ERR(di_args))
2165 return PTR_ERR(di_args);
2167 if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2168 s_uuid = di_args->uuid;
2170 mutex_lock(&fs_devices->device_list_mutex);
2171 dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL);
2172 mutex_unlock(&fs_devices->device_list_mutex);
2179 di_args->devid = dev->devid;
2180 di_args->bytes_used = dev->bytes_used;
2181 di_args->total_bytes = dev->total_bytes;
2182 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2183 strncpy(di_args->path, dev->name, sizeof(di_args->path));
2186 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2193 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2194 u64 off, u64 olen, u64 destoff)
2196 struct inode *inode = fdentry(file)->d_inode;
2197 struct btrfs_root *root = BTRFS_I(inode)->root;
2198 struct file *src_file;
2200 struct btrfs_trans_handle *trans;
2201 struct btrfs_path *path;
2202 struct extent_buffer *leaf;
2204 struct btrfs_key key;
2209 u64 bs = root->fs_info->sb->s_blocksize;
2214 * - split compressed inline extents. annoying: we need to
2215 * decompress into destination's address_space (the file offset
2216 * may change, so source mapping won't do), then recompress (or
2217 * otherwise reinsert) a subrange.
2218 * - allow ranges within the same file to be cloned (provided
2219 * they don't overlap)?
2222 /* the destination must be opened for writing */
2223 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
2226 if (btrfs_root_readonly(root))
2229 ret = mnt_want_write(file->f_path.mnt);
2233 src_file = fget(srcfd);
2236 goto out_drop_write;
2239 src = src_file->f_dentry->d_inode;
2245 /* the src must be open for reading */
2246 if (!(src_file->f_mode & FMODE_READ))
2249 /* don't make the dst file partly checksummed */
2250 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2251 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2255 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2259 if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
2263 buf = vmalloc(btrfs_level_size(root, 0));
2267 path = btrfs_alloc_path();
2275 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2276 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
2278 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2279 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
2282 /* determine range to clone */
2284 if (off + len > src->i_size || off + len < off)
2287 olen = len = src->i_size - off;
2288 /* if we extend to eof, continue to block boundary */
2289 if (off + len == src->i_size)
2290 len = ALIGN(src->i_size, bs) - off;
2292 /* verify the end result is block aligned */
2293 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2294 !IS_ALIGNED(destoff, bs))
2297 if (destoff > inode->i_size) {
2298 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2303 /* truncate page cache pages from target inode range */
2304 truncate_inode_pages_range(&inode->i_data, destoff,
2305 PAGE_CACHE_ALIGN(destoff + len) - 1);
2307 /* do any pending delalloc/csum calc on src, one way or
2308 another, and lock file content */
2310 struct btrfs_ordered_extent *ordered;
2311 lock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
2312 ordered = btrfs_lookup_first_ordered_extent(src, off+len);
2314 !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len,
2315 EXTENT_DELALLOC, 0, NULL))
2317 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
2319 btrfs_put_ordered_extent(ordered);
2320 btrfs_wait_ordered_range(src, off, len);
2324 key.objectid = btrfs_ino(src);
2325 key.type = BTRFS_EXTENT_DATA_KEY;
2330 * note the key will change type as we walk through the
2333 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2337 nritems = btrfs_header_nritems(path->nodes[0]);
2338 if (path->slots[0] >= nritems) {
2339 ret = btrfs_next_leaf(root, path);
2344 nritems = btrfs_header_nritems(path->nodes[0]);
2346 leaf = path->nodes[0];
2347 slot = path->slots[0];
2349 btrfs_item_key_to_cpu(leaf, &key, slot);
2350 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
2351 key.objectid != btrfs_ino(src))
2354 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2355 struct btrfs_file_extent_item *extent;
2358 struct btrfs_key new_key;
2359 u64 disko = 0, diskl = 0;
2360 u64 datao = 0, datal = 0;
2364 size = btrfs_item_size_nr(leaf, slot);
2365 read_extent_buffer(leaf, buf,
2366 btrfs_item_ptr_offset(leaf, slot),
2369 extent = btrfs_item_ptr(leaf, slot,
2370 struct btrfs_file_extent_item);
2371 comp = btrfs_file_extent_compression(leaf, extent);
2372 type = btrfs_file_extent_type(leaf, extent);
2373 if (type == BTRFS_FILE_EXTENT_REG ||
2374 type == BTRFS_FILE_EXTENT_PREALLOC) {
2375 disko = btrfs_file_extent_disk_bytenr(leaf,
2377 diskl = btrfs_file_extent_disk_num_bytes(leaf,
2379 datao = btrfs_file_extent_offset(leaf, extent);
2380 datal = btrfs_file_extent_num_bytes(leaf,
2382 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2383 /* take upper bound, may be compressed */
2384 datal = btrfs_file_extent_ram_bytes(leaf,
2387 btrfs_release_path(path);
2389 if (key.offset + datal <= off ||
2390 key.offset >= off+len)
2393 memcpy(&new_key, &key, sizeof(new_key));
2394 new_key.objectid = btrfs_ino(inode);
2395 if (off <= key.offset)
2396 new_key.offset = key.offset + destoff - off;
2398 new_key.offset = destoff;
2401 * 1 - adjusting old extent (we may have to split it)
2402 * 1 - add new extent
2405 trans = btrfs_start_transaction(root, 3);
2406 if (IS_ERR(trans)) {
2407 ret = PTR_ERR(trans);
2411 if (type == BTRFS_FILE_EXTENT_REG ||
2412 type == BTRFS_FILE_EXTENT_PREALLOC) {
2414 * a | --- range to clone ---| b
2415 * | ------------- extent ------------- |
2418 /* substract range b */
2419 if (key.offset + datal > off + len)
2420 datal = off + len - key.offset;
2422 /* substract range a */
2423 if (off > key.offset) {
2424 datao += off - key.offset;
2425 datal -= off - key.offset;
2428 ret = btrfs_drop_extents(trans, inode,
2430 new_key.offset + datal,
2434 ret = btrfs_insert_empty_item(trans, root, path,
2438 leaf = path->nodes[0];
2439 slot = path->slots[0];
2440 write_extent_buffer(leaf, buf,
2441 btrfs_item_ptr_offset(leaf, slot),
2444 extent = btrfs_item_ptr(leaf, slot,
2445 struct btrfs_file_extent_item);
2447 /* disko == 0 means it's a hole */
2451 btrfs_set_file_extent_offset(leaf, extent,
2453 btrfs_set_file_extent_num_bytes(leaf, extent,
2456 inode_add_bytes(inode, datal);
2457 ret = btrfs_inc_extent_ref(trans, root,
2459 root->root_key.objectid,
2461 new_key.offset - datao);
2464 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2467 if (off > key.offset) {
2468 skip = off - key.offset;
2469 new_key.offset += skip;
2472 if (key.offset + datal > off+len)
2473 trim = key.offset + datal - (off+len);
2475 if (comp && (skip || trim)) {
2477 btrfs_end_transaction(trans, root);
2480 size -= skip + trim;
2481 datal -= skip + trim;
2483 ret = btrfs_drop_extents(trans, inode,
2485 new_key.offset + datal,
2489 ret = btrfs_insert_empty_item(trans, root, path,
2495 btrfs_file_extent_calc_inline_size(0);
2496 memmove(buf+start, buf+start+skip,
2500 leaf = path->nodes[0];
2501 slot = path->slots[0];
2502 write_extent_buffer(leaf, buf,
2503 btrfs_item_ptr_offset(leaf, slot),
2505 inode_add_bytes(inode, datal);
2508 btrfs_mark_buffer_dirty(leaf);
2509 btrfs_release_path(path);
2511 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2514 * we round up to the block size at eof when
2515 * determining which extents to clone above,
2516 * but shouldn't round up the file size
2518 endoff = new_key.offset + datal;
2519 if (endoff > destoff+olen)
2520 endoff = destoff+olen;
2521 if (endoff > inode->i_size)
2522 btrfs_i_size_write(inode, endoff);
2524 ret = btrfs_update_inode(trans, root, inode);
2526 btrfs_end_transaction(trans, root);
2529 btrfs_release_path(path);
2534 btrfs_release_path(path);
2535 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
2537 mutex_unlock(&src->i_mutex);
2538 mutex_unlock(&inode->i_mutex);
2540 btrfs_free_path(path);
2544 mnt_drop_write(file->f_path.mnt);
2548 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
2550 struct btrfs_ioctl_clone_range_args args;
2552 if (copy_from_user(&args, argp, sizeof(args)))
2554 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2555 args.src_length, args.dest_offset);
2559 * there are many ways the trans_start and trans_end ioctls can lead
2560 * to deadlocks. They should only be used by applications that
2561 * basically own the machine, and have a very in depth understanding
2562 * of all the possible deadlocks and enospc problems.
2564 static long btrfs_ioctl_trans_start(struct file *file)
2566 struct inode *inode = fdentry(file)->d_inode;
2567 struct btrfs_root *root = BTRFS_I(inode)->root;
2568 struct btrfs_trans_handle *trans;
2572 if (!capable(CAP_SYS_ADMIN))
2576 if (file->private_data)
2580 if (btrfs_root_readonly(root))
2583 ret = mnt_want_write(file->f_path.mnt);
2587 atomic_inc(&root->fs_info->open_ioctl_trans);
2590 trans = btrfs_start_ioctl_transaction(root);
2594 file->private_data = trans;
2598 atomic_dec(&root->fs_info->open_ioctl_trans);
2599 mnt_drop_write(file->f_path.mnt);
2604 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2606 struct inode *inode = fdentry(file)->d_inode;
2607 struct btrfs_root *root = BTRFS_I(inode)->root;
2608 struct btrfs_root *new_root;
2609 struct btrfs_dir_item *di;
2610 struct btrfs_trans_handle *trans;
2611 struct btrfs_path *path;
2612 struct btrfs_key location;
2613 struct btrfs_disk_key disk_key;
2614 struct btrfs_super_block *disk_super;
2619 if (!capable(CAP_SYS_ADMIN))
2622 if (copy_from_user(&objectid, argp, sizeof(objectid)))
2626 objectid = root->root_key.objectid;
2628 location.objectid = objectid;
2629 location.type = BTRFS_ROOT_ITEM_KEY;
2630 location.offset = (u64)-1;
2632 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2633 if (IS_ERR(new_root))
2634 return PTR_ERR(new_root);
2636 if (btrfs_root_refs(&new_root->root_item) == 0)
2639 path = btrfs_alloc_path();
2642 path->leave_spinning = 1;
2644 trans = btrfs_start_transaction(root, 1);
2645 if (IS_ERR(trans)) {
2646 btrfs_free_path(path);
2647 return PTR_ERR(trans);
2650 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
2651 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2652 dir_id, "default", 7, 1);
2653 if (IS_ERR_OR_NULL(di)) {
2654 btrfs_free_path(path);
2655 btrfs_end_transaction(trans, root);
2656 printk(KERN_ERR "Umm, you don't have the default dir item, "
2657 "this isn't going to work\n");
2661 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2662 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2663 btrfs_mark_buffer_dirty(path->nodes[0]);
2664 btrfs_free_path(path);
2666 disk_super = root->fs_info->super_copy;
2667 features = btrfs_super_incompat_flags(disk_super);
2668 if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) {
2669 features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL;
2670 btrfs_set_super_incompat_flags(disk_super, features);
2672 btrfs_end_transaction(trans, root);
2677 static void get_block_group_info(struct list_head *groups_list,
2678 struct btrfs_ioctl_space_info *space)
2680 struct btrfs_block_group_cache *block_group;
2682 space->total_bytes = 0;
2683 space->used_bytes = 0;
2685 list_for_each_entry(block_group, groups_list, list) {
2686 space->flags = block_group->flags;
2687 space->total_bytes += block_group->key.offset;
2688 space->used_bytes +=
2689 btrfs_block_group_used(&block_group->item);
2693 long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2695 struct btrfs_ioctl_space_args space_args;
2696 struct btrfs_ioctl_space_info space;
2697 struct btrfs_ioctl_space_info *dest;
2698 struct btrfs_ioctl_space_info *dest_orig;
2699 struct btrfs_ioctl_space_info __user *user_dest;
2700 struct btrfs_space_info *info;
2701 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2702 BTRFS_BLOCK_GROUP_SYSTEM,
2703 BTRFS_BLOCK_GROUP_METADATA,
2704 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2711 if (copy_from_user(&space_args,
2712 (struct btrfs_ioctl_space_args __user *)arg,
2713 sizeof(space_args)))
2716 for (i = 0; i < num_types; i++) {
2717 struct btrfs_space_info *tmp;
2721 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2723 if (tmp->flags == types[i]) {
2733 down_read(&info->groups_sem);
2734 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2735 if (!list_empty(&info->block_groups[c]))
2738 up_read(&info->groups_sem);
2741 /* space_slots == 0 means they are asking for a count */
2742 if (space_args.space_slots == 0) {
2743 space_args.total_spaces = slot_count;
2747 slot_count = min_t(u64, space_args.space_slots, slot_count);
2749 alloc_size = sizeof(*dest) * slot_count;
2751 /* we generally have at most 6 or so space infos, one for each raid
2752 * level. So, a whole page should be more than enough for everyone
2754 if (alloc_size > PAGE_CACHE_SIZE)
2757 space_args.total_spaces = 0;
2758 dest = kmalloc(alloc_size, GFP_NOFS);
2763 /* now we have a buffer to copy into */
2764 for (i = 0; i < num_types; i++) {
2765 struct btrfs_space_info *tmp;
2772 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2774 if (tmp->flags == types[i]) {
2783 down_read(&info->groups_sem);
2784 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2785 if (!list_empty(&info->block_groups[c])) {
2786 get_block_group_info(&info->block_groups[c],
2788 memcpy(dest, &space, sizeof(space));
2790 space_args.total_spaces++;
2796 up_read(&info->groups_sem);
2799 user_dest = (struct btrfs_ioctl_space_info *)
2800 (arg + sizeof(struct btrfs_ioctl_space_args));
2802 if (copy_to_user(user_dest, dest_orig, alloc_size))
2807 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
2814 * there are many ways the trans_start and trans_end ioctls can lead
2815 * to deadlocks. They should only be used by applications that
2816 * basically own the machine, and have a very in depth understanding
2817 * of all the possible deadlocks and enospc problems.
2819 long btrfs_ioctl_trans_end(struct file *file)
2821 struct inode *inode = fdentry(file)->d_inode;
2822 struct btrfs_root *root = BTRFS_I(inode)->root;
2823 struct btrfs_trans_handle *trans;
2825 trans = file->private_data;
2828 file->private_data = NULL;
2830 btrfs_end_transaction(trans, root);
2832 atomic_dec(&root->fs_info->open_ioctl_trans);
2834 mnt_drop_write(file->f_path.mnt);
2838 static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
2840 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2841 struct btrfs_trans_handle *trans;
2845 trans = btrfs_start_transaction(root, 0);
2847 return PTR_ERR(trans);
2848 transid = trans->transid;
2849 ret = btrfs_commit_transaction_async(trans, root, 0);
2851 btrfs_end_transaction(trans, root);
2856 if (copy_to_user(argp, &transid, sizeof(transid)))
2861 static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
2863 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2867 if (copy_from_user(&transid, argp, sizeof(transid)))
2870 transid = 0; /* current trans */
2872 return btrfs_wait_for_commit(root, transid);
2875 static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg)
2878 struct btrfs_ioctl_scrub_args *sa;
2880 if (!capable(CAP_SYS_ADMIN))
2883 sa = memdup_user(arg, sizeof(*sa));
2887 ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end,
2888 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY);
2890 if (copy_to_user(arg, sa, sizeof(*sa)))
2897 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
2899 if (!capable(CAP_SYS_ADMIN))
2902 return btrfs_scrub_cancel(root);
2905 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
2908 struct btrfs_ioctl_scrub_args *sa;
2911 if (!capable(CAP_SYS_ADMIN))
2914 sa = memdup_user(arg, sizeof(*sa));
2918 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
2920 if (copy_to_user(arg, sa, sizeof(*sa)))
2927 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
2933 struct btrfs_ioctl_ino_path_args *ipa = NULL;
2934 struct inode_fs_paths *ipath = NULL;
2935 struct btrfs_path *path;
2937 if (!capable(CAP_SYS_ADMIN))
2940 path = btrfs_alloc_path();
2946 ipa = memdup_user(arg, sizeof(*ipa));
2953 size = min_t(u32, ipa->size, 4096);
2954 ipath = init_ipath(size, root, path);
2955 if (IS_ERR(ipath)) {
2956 ret = PTR_ERR(ipath);
2961 ret = paths_from_inode(ipa->inum, ipath);
2965 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
2966 rel_ptr = ipath->fspath->val[i] -
2967 (u64)(unsigned long)ipath->fspath->val;
2968 ipath->fspath->val[i] = rel_ptr;
2971 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
2972 (void *)(unsigned long)ipath->fspath, size);
2979 btrfs_free_path(path);
2986 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
2988 struct btrfs_data_container *inodes = ctx;
2989 const size_t c = 3 * sizeof(u64);
2991 if (inodes->bytes_left >= c) {
2992 inodes->bytes_left -= c;
2993 inodes->val[inodes->elem_cnt] = inum;
2994 inodes->val[inodes->elem_cnt + 1] = offset;
2995 inodes->val[inodes->elem_cnt + 2] = root;
2996 inodes->elem_cnt += 3;
2998 inodes->bytes_missing += c - inodes->bytes_left;
2999 inodes->bytes_left = 0;
3000 inodes->elem_missed += 3;
3006 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3012 struct btrfs_ioctl_logical_ino_args *loi;
3013 struct btrfs_data_container *inodes = NULL;
3014 struct btrfs_path *path = NULL;
3015 struct btrfs_key key;
3017 if (!capable(CAP_SYS_ADMIN))
3020 loi = memdup_user(arg, sizeof(*loi));
3027 path = btrfs_alloc_path();
3033 size = min_t(u32, loi->size, 4096);
3034 inodes = init_data_container(size);
3035 if (IS_ERR(inodes)) {
3036 ret = PTR_ERR(inodes);
3041 ret = extent_from_logical(root->fs_info, loi->logical, path, &key);
3043 if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK)
3048 extent_offset = loi->logical - key.objectid;
3049 ret = iterate_extent_inodes(root->fs_info, path, key.objectid,
3050 extent_offset, build_ino_list, inodes);
3055 ret = copy_to_user((void *)(unsigned long)loi->inodes,
3056 (void *)(unsigned long)inodes, size);
3061 btrfs_free_path(path);
3068 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info,
3069 struct btrfs_ioctl_balance_args *bargs)
3071 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3073 bargs->flags = bctl->flags;
3075 if (atomic_read(&fs_info->balance_running))
3076 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3077 if (atomic_read(&fs_info->balance_pause_req))
3078 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3079 if (atomic_read(&fs_info->balance_cancel_req))
3080 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3082 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3083 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3084 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3087 static long btrfs_ioctl_balance(struct btrfs_root *root, void __user *arg)
3089 struct btrfs_fs_info *fs_info = root->fs_info;
3090 struct btrfs_ioctl_balance_args *bargs;
3091 struct btrfs_balance_control *bctl;
3094 if (!capable(CAP_SYS_ADMIN))
3097 if (fs_info->sb->s_flags & MS_RDONLY)
3100 mutex_lock(&fs_info->volume_mutex);
3101 mutex_lock(&fs_info->balance_mutex);
3104 bargs = memdup_user(arg, sizeof(*bargs));
3105 if (IS_ERR(bargs)) {
3106 ret = PTR_ERR(bargs);
3110 if (bargs->flags & BTRFS_BALANCE_RESUME) {
3111 if (!fs_info->balance_ctl) {
3116 bctl = fs_info->balance_ctl;
3117 spin_lock(&fs_info->balance_lock);
3118 bctl->flags |= BTRFS_BALANCE_RESUME;
3119 spin_unlock(&fs_info->balance_lock);
3127 if (fs_info->balance_ctl) {
3132 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3138 bctl->fs_info = fs_info;
3140 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3141 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3142 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3144 bctl->flags = bargs->flags;
3146 /* balance everything - no filters */
3147 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
3151 ret = btrfs_balance(bctl, bargs);
3153 * bctl is freed in __cancel_balance or in free_fs_info if
3154 * restriper was paused all the way until unmount
3157 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3164 mutex_unlock(&fs_info->balance_mutex);
3165 mutex_unlock(&fs_info->volume_mutex);
3169 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
3171 if (!capable(CAP_SYS_ADMIN))
3175 case BTRFS_BALANCE_CTL_PAUSE:
3176 return btrfs_pause_balance(root->fs_info);
3177 case BTRFS_BALANCE_CTL_CANCEL:
3178 return btrfs_cancel_balance(root->fs_info);
3184 long btrfs_ioctl(struct file *file, unsigned int
3185 cmd, unsigned long arg)
3187 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3188 void __user *argp = (void __user *)arg;
3191 case FS_IOC_GETFLAGS:
3192 return btrfs_ioctl_getflags(file, argp);
3193 case FS_IOC_SETFLAGS:
3194 return btrfs_ioctl_setflags(file, argp);
3195 case FS_IOC_GETVERSION:
3196 return btrfs_ioctl_getversion(file, argp);
3198 return btrfs_ioctl_fitrim(file, argp);
3199 case BTRFS_IOC_SNAP_CREATE:
3200 return btrfs_ioctl_snap_create(file, argp, 0);
3201 case BTRFS_IOC_SNAP_CREATE_V2:
3202 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3203 case BTRFS_IOC_SUBVOL_CREATE:
3204 return btrfs_ioctl_snap_create(file, argp, 1);
3205 case BTRFS_IOC_SNAP_DESTROY:
3206 return btrfs_ioctl_snap_destroy(file, argp);
3207 case BTRFS_IOC_SUBVOL_GETFLAGS:
3208 return btrfs_ioctl_subvol_getflags(file, argp);
3209 case BTRFS_IOC_SUBVOL_SETFLAGS:
3210 return btrfs_ioctl_subvol_setflags(file, argp);
3211 case BTRFS_IOC_DEFAULT_SUBVOL:
3212 return btrfs_ioctl_default_subvol(file, argp);
3213 case BTRFS_IOC_DEFRAG:
3214 return btrfs_ioctl_defrag(file, NULL);
3215 case BTRFS_IOC_DEFRAG_RANGE:
3216 return btrfs_ioctl_defrag(file, argp);
3217 case BTRFS_IOC_RESIZE:
3218 return btrfs_ioctl_resize(root, argp);
3219 case BTRFS_IOC_ADD_DEV:
3220 return btrfs_ioctl_add_dev(root, argp);
3221 case BTRFS_IOC_RM_DEV:
3222 return btrfs_ioctl_rm_dev(root, argp);
3223 case BTRFS_IOC_FS_INFO:
3224 return btrfs_ioctl_fs_info(root, argp);
3225 case BTRFS_IOC_DEV_INFO:
3226 return btrfs_ioctl_dev_info(root, argp);
3227 case BTRFS_IOC_BALANCE:
3228 return btrfs_ioctl_balance(root, NULL);
3229 case BTRFS_IOC_CLONE:
3230 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
3231 case BTRFS_IOC_CLONE_RANGE:
3232 return btrfs_ioctl_clone_range(file, argp);
3233 case BTRFS_IOC_TRANS_START:
3234 return btrfs_ioctl_trans_start(file);
3235 case BTRFS_IOC_TRANS_END:
3236 return btrfs_ioctl_trans_end(file);
3237 case BTRFS_IOC_TREE_SEARCH:
3238 return btrfs_ioctl_tree_search(file, argp);
3239 case BTRFS_IOC_INO_LOOKUP:
3240 return btrfs_ioctl_ino_lookup(file, argp);
3241 case BTRFS_IOC_INO_PATHS:
3242 return btrfs_ioctl_ino_to_path(root, argp);
3243 case BTRFS_IOC_LOGICAL_INO:
3244 return btrfs_ioctl_logical_to_ino(root, argp);
3245 case BTRFS_IOC_SPACE_INFO:
3246 return btrfs_ioctl_space_info(root, argp);
3247 case BTRFS_IOC_SYNC:
3248 btrfs_sync_fs(file->f_dentry->d_sb, 1);
3250 case BTRFS_IOC_START_SYNC:
3251 return btrfs_ioctl_start_sync(file, argp);
3252 case BTRFS_IOC_WAIT_SYNC:
3253 return btrfs_ioctl_wait_sync(file, argp);
3254 case BTRFS_IOC_SCRUB:
3255 return btrfs_ioctl_scrub(root, argp);
3256 case BTRFS_IOC_SCRUB_CANCEL:
3257 return btrfs_ioctl_scrub_cancel(root, argp);
3258 case BTRFS_IOC_SCRUB_PROGRESS:
3259 return btrfs_ioctl_scrub_progress(root, argp);
3260 case BTRFS_IOC_BALANCE_V2:
3261 return btrfs_ioctl_balance(root, argp);
3262 case BTRFS_IOC_BALANCE_CTL:
3263 return btrfs_ioctl_balance_ctl(root, arg);