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>
46 #include "transaction.h"
47 #include "btrfs_inode.h"
49 #include "print-tree.h"
53 /* Mask out flags that are inappropriate for the given type of inode. */
54 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
58 else if (S_ISREG(mode))
59 return flags & ~FS_DIRSYNC_FL;
61 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
65 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
67 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
69 unsigned int iflags = 0;
71 if (flags & BTRFS_INODE_SYNC)
73 if (flags & BTRFS_INODE_IMMUTABLE)
74 iflags |= FS_IMMUTABLE_FL;
75 if (flags & BTRFS_INODE_APPEND)
76 iflags |= FS_APPEND_FL;
77 if (flags & BTRFS_INODE_NODUMP)
78 iflags |= FS_NODUMP_FL;
79 if (flags & BTRFS_INODE_NOATIME)
80 iflags |= FS_NOATIME_FL;
81 if (flags & BTRFS_INODE_DIRSYNC)
82 iflags |= FS_DIRSYNC_FL;
88 * Update inode->i_flags based on the btrfs internal flags.
90 void btrfs_update_iflags(struct inode *inode)
92 struct btrfs_inode *ip = BTRFS_I(inode);
94 inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
96 if (ip->flags & BTRFS_INODE_SYNC)
97 inode->i_flags |= S_SYNC;
98 if (ip->flags & BTRFS_INODE_IMMUTABLE)
99 inode->i_flags |= S_IMMUTABLE;
100 if (ip->flags & BTRFS_INODE_APPEND)
101 inode->i_flags |= S_APPEND;
102 if (ip->flags & BTRFS_INODE_NOATIME)
103 inode->i_flags |= S_NOATIME;
104 if (ip->flags & BTRFS_INODE_DIRSYNC)
105 inode->i_flags |= S_DIRSYNC;
109 * Inherit flags from the parent inode.
111 * Unlike extN we don't have any flags we don't want to inherit currently.
113 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
120 flags = BTRFS_I(dir)->flags;
122 if (S_ISREG(inode->i_mode))
123 flags &= ~BTRFS_INODE_DIRSYNC;
124 else if (!S_ISDIR(inode->i_mode))
125 flags &= (BTRFS_INODE_NODUMP | BTRFS_INODE_NOATIME);
127 BTRFS_I(inode)->flags = flags;
128 btrfs_update_iflags(inode);
131 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
133 struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode);
134 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
136 if (copy_to_user(arg, &flags, sizeof(flags)))
141 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
143 struct inode *inode = file->f_path.dentry->d_inode;
144 struct btrfs_inode *ip = BTRFS_I(inode);
145 struct btrfs_root *root = ip->root;
146 struct btrfs_trans_handle *trans;
147 unsigned int flags, oldflags;
150 if (btrfs_root_readonly(root))
153 if (copy_from_user(&flags, arg, sizeof(flags)))
156 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
157 FS_NOATIME_FL | FS_NODUMP_FL | \
158 FS_SYNC_FL | FS_DIRSYNC_FL))
161 if (!is_owner_or_cap(inode))
164 mutex_lock(&inode->i_mutex);
166 flags = btrfs_mask_flags(inode->i_mode, flags);
167 oldflags = btrfs_flags_to_ioctl(ip->flags);
168 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
169 if (!capable(CAP_LINUX_IMMUTABLE)) {
175 ret = mnt_want_write(file->f_path.mnt);
179 if (flags & FS_SYNC_FL)
180 ip->flags |= BTRFS_INODE_SYNC;
182 ip->flags &= ~BTRFS_INODE_SYNC;
183 if (flags & FS_IMMUTABLE_FL)
184 ip->flags |= BTRFS_INODE_IMMUTABLE;
186 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
187 if (flags & FS_APPEND_FL)
188 ip->flags |= BTRFS_INODE_APPEND;
190 ip->flags &= ~BTRFS_INODE_APPEND;
191 if (flags & FS_NODUMP_FL)
192 ip->flags |= BTRFS_INODE_NODUMP;
194 ip->flags &= ~BTRFS_INODE_NODUMP;
195 if (flags & FS_NOATIME_FL)
196 ip->flags |= BTRFS_INODE_NOATIME;
198 ip->flags &= ~BTRFS_INODE_NOATIME;
199 if (flags & FS_DIRSYNC_FL)
200 ip->flags |= BTRFS_INODE_DIRSYNC;
202 ip->flags &= ~BTRFS_INODE_DIRSYNC;
205 trans = btrfs_join_transaction(root, 1);
208 ret = btrfs_update_inode(trans, root, inode);
211 btrfs_update_iflags(inode);
212 inode->i_ctime = CURRENT_TIME;
213 btrfs_end_transaction(trans, root);
215 mnt_drop_write(file->f_path.mnt);
217 mutex_unlock(&inode->i_mutex);
221 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
223 struct inode *inode = file->f_path.dentry->d_inode;
225 return put_user(inode->i_generation, arg);
228 static noinline int create_subvol(struct btrfs_root *root,
229 struct dentry *dentry,
230 char *name, int namelen,
233 struct btrfs_trans_handle *trans;
234 struct btrfs_key key;
235 struct btrfs_root_item root_item;
236 struct btrfs_inode_item *inode_item;
237 struct extent_buffer *leaf;
238 struct btrfs_root *new_root;
239 struct dentry *parent = dget_parent(dentry);
244 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
247 ret = btrfs_find_free_objectid(NULL, root->fs_info->tree_root,
254 dir = parent->d_inode;
262 trans = btrfs_start_transaction(root, 6);
265 return PTR_ERR(trans);
268 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
269 0, objectid, NULL, 0, 0, 0);
275 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
276 btrfs_set_header_bytenr(leaf, leaf->start);
277 btrfs_set_header_generation(leaf, trans->transid);
278 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
279 btrfs_set_header_owner(leaf, objectid);
281 write_extent_buffer(leaf, root->fs_info->fsid,
282 (unsigned long)btrfs_header_fsid(leaf),
284 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
285 (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
287 btrfs_mark_buffer_dirty(leaf);
289 inode_item = &root_item.inode;
290 memset(inode_item, 0, sizeof(*inode_item));
291 inode_item->generation = cpu_to_le64(1);
292 inode_item->size = cpu_to_le64(3);
293 inode_item->nlink = cpu_to_le32(1);
294 inode_item->nbytes = cpu_to_le64(root->leafsize);
295 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
297 btrfs_set_root_bytenr(&root_item, leaf->start);
298 btrfs_set_root_generation(&root_item, trans->transid);
299 btrfs_set_root_level(&root_item, 0);
300 btrfs_set_root_refs(&root_item, 1);
301 btrfs_set_root_used(&root_item, leaf->len);
302 btrfs_set_root_last_snapshot(&root_item, 0);
304 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
305 root_item.drop_level = 0;
307 btrfs_tree_unlock(leaf);
308 free_extent_buffer(leaf);
311 btrfs_set_root_dirid(&root_item, new_dirid);
313 key.objectid = objectid;
315 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
316 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
321 key.offset = (u64)-1;
322 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
323 BUG_ON(IS_ERR(new_root));
325 btrfs_record_root_in_trans(trans, new_root);
327 ret = btrfs_create_subvol_root(trans, new_root, new_dirid,
328 BTRFS_I(dir)->block_group);
330 * insert the directory item
332 ret = btrfs_set_inode_index(dir, &index);
335 ret = btrfs_insert_dir_item(trans, root,
336 name, namelen, dir->i_ino, &key,
337 BTRFS_FT_DIR, index);
341 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
342 ret = btrfs_update_inode(trans, root, dir);
345 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
346 objectid, root->root_key.objectid,
347 dir->i_ino, index, name, namelen);
351 d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
355 *async_transid = trans->transid;
356 err = btrfs_commit_transaction_async(trans, root, 1);
358 err = btrfs_commit_transaction(trans, root);
365 static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
366 char *name, int namelen, u64 *async_transid,
370 struct dentry *parent;
371 struct btrfs_pending_snapshot *pending_snapshot;
372 struct btrfs_trans_handle *trans;
378 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
379 if (!pending_snapshot)
382 btrfs_init_block_rsv(&pending_snapshot->block_rsv);
383 pending_snapshot->dentry = dentry;
384 pending_snapshot->root = root;
385 pending_snapshot->readonly = readonly;
387 trans = btrfs_start_transaction(root->fs_info->extent_root, 5);
389 ret = PTR_ERR(trans);
393 ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
396 list_add(&pending_snapshot->list,
397 &trans->transaction->pending_snapshots);
399 *async_transid = trans->transid;
400 ret = btrfs_commit_transaction_async(trans,
401 root->fs_info->extent_root, 1);
403 ret = btrfs_commit_transaction(trans,
404 root->fs_info->extent_root);
408 ret = pending_snapshot->error;
412 btrfs_orphan_cleanup(pending_snapshot->snap);
414 parent = dget_parent(dentry);
415 inode = btrfs_lookup_dentry(parent->d_inode, dentry);
418 ret = PTR_ERR(inode);
422 d_instantiate(dentry, inode);
425 kfree(pending_snapshot);
429 /* copy of check_sticky in fs/namei.c()
430 * It's inline, so penalty for filesystems that don't use sticky bit is
433 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
435 uid_t fsuid = current_fsuid();
437 if (!(dir->i_mode & S_ISVTX))
439 if (inode->i_uid == fsuid)
441 if (dir->i_uid == fsuid)
443 return !capable(CAP_FOWNER);
446 /* copy of may_delete in fs/namei.c()
447 * Check whether we can remove a link victim from directory dir, check
448 * whether the type of victim is right.
449 * 1. We can't do it if dir is read-only (done in permission())
450 * 2. We should have write and exec permissions on dir
451 * 3. We can't remove anything from append-only dir
452 * 4. We can't do anything with immutable dir (done in permission())
453 * 5. If the sticky bit on dir is set we should either
454 * a. be owner of dir, or
455 * b. be owner of victim, or
456 * c. have CAP_FOWNER capability
457 * 6. If the victim is append-only or immutable we can't do antyhing with
458 * links pointing to it.
459 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
460 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
461 * 9. We can't remove a root or mountpoint.
462 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
463 * nfs_async_unlink().
466 static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
470 if (!victim->d_inode)
473 BUG_ON(victim->d_parent->d_inode != dir);
474 audit_inode_child(victim, dir);
476 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
481 if (btrfs_check_sticky(dir, victim->d_inode)||
482 IS_APPEND(victim->d_inode)||
483 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
486 if (!S_ISDIR(victim->d_inode->i_mode))
490 } else if (S_ISDIR(victim->d_inode->i_mode))
494 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
499 /* copy of may_create in fs/namei.c() */
500 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
506 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
510 * Create a new subvolume below @parent. This is largely modeled after
511 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
512 * inside this filesystem so it's quite a bit simpler.
514 static noinline int btrfs_mksubvol(struct path *parent,
515 char *name, int namelen,
516 struct btrfs_root *snap_src,
517 u64 *async_transid, bool readonly)
519 struct inode *dir = parent->dentry->d_inode;
520 struct dentry *dentry;
523 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
525 dentry = lookup_one_len(name, parent->dentry, namelen);
526 error = PTR_ERR(dentry);
534 error = mnt_want_write(parent->mnt);
538 error = btrfs_may_create(dir, dentry);
542 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
544 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
548 error = create_snapshot(snap_src, dentry,
549 name, namelen, async_transid, readonly);
551 error = create_subvol(BTRFS_I(dir)->root, dentry,
552 name, namelen, async_transid);
555 fsnotify_mkdir(dir, dentry);
557 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
559 mnt_drop_write(parent->mnt);
563 mutex_unlock(&dir->i_mutex);
567 static int should_defrag_range(struct inode *inode, u64 start, u64 len,
568 int thresh, u64 *last_len, u64 *skip,
571 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
572 struct extent_map *em = NULL;
573 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
581 * make sure that once we start defragging and extent, we keep on
584 if (start < *defrag_end)
590 * hopefully we have this extent in the tree already, try without
591 * the full extent lock
593 read_lock(&em_tree->lock);
594 em = lookup_extent_mapping(em_tree, start, len);
595 read_unlock(&em_tree->lock);
598 /* get the big lock and read metadata off disk */
599 lock_extent(io_tree, start, start + len - 1, GFP_NOFS);
600 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
601 unlock_extent(io_tree, start, start + len - 1, GFP_NOFS);
607 /* this will cover holes, and inline extents */
608 if (em->block_start >= EXTENT_MAP_LAST_BYTE)
612 * we hit a real extent, if it is big don't bother defragging it again
614 if ((*last_len == 0 || *last_len >= thresh) && em->len >= thresh)
618 * last_len ends up being a counter of how many bytes we've defragged.
619 * every time we choose not to defrag an extent, we reset *last_len
620 * so that the next tiny extent will force a defrag.
622 * The end result of this is that tiny extents before a single big
623 * extent will force at least part of that big extent to be defragged.
627 *defrag_end = extent_map_end(em);
630 *skip = extent_map_end(em);
638 static int btrfs_defrag_file(struct file *file,
639 struct btrfs_ioctl_defrag_range_args *range)
641 struct inode *inode = fdentry(file)->d_inode;
642 struct btrfs_root *root = BTRFS_I(inode)->root;
643 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
644 struct btrfs_ordered_extent *ordered;
646 unsigned long last_index;
647 unsigned long ra_pages = root->fs_info->bdi.ra_pages;
648 unsigned long total_read = 0;
657 if (inode->i_size == 0)
660 if (range->start + range->len > range->start) {
661 last_index = min_t(u64, inode->i_size - 1,
662 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
664 last_index = (inode->i_size - 1) >> PAGE_CACHE_SHIFT;
667 i = range->start >> PAGE_CACHE_SHIFT;
668 while (i <= last_index) {
669 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
671 range->extent_thresh,
676 * the should_defrag function tells us how much to skip
677 * bump our counter by the suggested amount
679 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
680 i = max(i + 1, next);
684 if (total_read % ra_pages == 0) {
685 btrfs_force_ra(inode->i_mapping, &file->f_ra, file, i,
686 min(last_index, i + ra_pages - 1));
689 mutex_lock(&inode->i_mutex);
690 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
691 BTRFS_I(inode)->force_compress = 1;
693 ret = btrfs_delalloc_reserve_space(inode, PAGE_CACHE_SIZE);
697 if (inode->i_size == 0 ||
698 i > ((inode->i_size - 1) >> PAGE_CACHE_SHIFT)) {
700 goto err_reservations;
703 page = grab_cache_page(inode->i_mapping, i);
706 goto err_reservations;
709 if (!PageUptodate(page)) {
710 btrfs_readpage(NULL, page);
712 if (!PageUptodate(page)) {
714 page_cache_release(page);
716 goto err_reservations;
720 if (page->mapping != inode->i_mapping) {
722 page_cache_release(page);
726 wait_on_page_writeback(page);
728 if (PageDirty(page)) {
729 btrfs_delalloc_release_space(inode, PAGE_CACHE_SIZE);
733 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
734 page_end = page_start + PAGE_CACHE_SIZE - 1;
735 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
737 ordered = btrfs_lookup_ordered_extent(inode, page_start);
739 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
741 page_cache_release(page);
742 btrfs_start_ordered_extent(inode, ordered, 1);
743 btrfs_put_ordered_extent(ordered);
746 set_page_extent_mapped(page);
749 * this makes sure page_mkwrite is called on the
750 * page if it is dirtied again later
752 clear_page_dirty_for_io(page);
753 clear_extent_bits(&BTRFS_I(inode)->io_tree, page_start,
754 page_end, EXTENT_DIRTY | EXTENT_DELALLOC |
755 EXTENT_DO_ACCOUNTING, GFP_NOFS);
757 btrfs_set_extent_delalloc(inode, page_start, page_end, NULL);
758 ClearPageChecked(page);
759 set_page_dirty(page);
760 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
764 page_cache_release(page);
765 mutex_unlock(&inode->i_mutex);
767 balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
771 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
772 filemap_flush(inode->i_mapping);
774 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
775 /* the filemap_flush will queue IO into the worker threads, but
776 * we have to make sure the IO is actually started and that
777 * ordered extents get created before we return
779 atomic_inc(&root->fs_info->async_submit_draining);
780 while (atomic_read(&root->fs_info->nr_async_submits) ||
781 atomic_read(&root->fs_info->async_delalloc_pages)) {
782 wait_event(root->fs_info->async_submit_wait,
783 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
784 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
786 atomic_dec(&root->fs_info->async_submit_draining);
788 mutex_lock(&inode->i_mutex);
789 BTRFS_I(inode)->force_compress = 0;
790 mutex_unlock(&inode->i_mutex);
796 btrfs_delalloc_release_space(inode, PAGE_CACHE_SIZE);
798 mutex_unlock(&inode->i_mutex);
802 static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
808 struct btrfs_ioctl_vol_args *vol_args;
809 struct btrfs_trans_handle *trans;
810 struct btrfs_device *device = NULL;
816 if (root->fs_info->sb->s_flags & MS_RDONLY)
819 if (!capable(CAP_SYS_ADMIN))
822 vol_args = memdup_user(arg, sizeof(*vol_args));
823 if (IS_ERR(vol_args))
824 return PTR_ERR(vol_args);
826 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
828 mutex_lock(&root->fs_info->volume_mutex);
829 sizestr = vol_args->name;
830 devstr = strchr(sizestr, ':');
833 sizestr = devstr + 1;
835 devstr = vol_args->name;
836 devid = simple_strtoull(devstr, &end, 10);
837 printk(KERN_INFO "resizing devid %llu\n",
838 (unsigned long long)devid);
840 device = btrfs_find_device(root, devid, NULL, NULL);
842 printk(KERN_INFO "resizer unable to find device %llu\n",
843 (unsigned long long)devid);
847 if (!strcmp(sizestr, "max"))
848 new_size = device->bdev->bd_inode->i_size;
850 if (sizestr[0] == '-') {
853 } else if (sizestr[0] == '+') {
857 new_size = memparse(sizestr, NULL);
864 old_size = device->total_bytes;
867 if (new_size > old_size) {
871 new_size = old_size - new_size;
872 } else if (mod > 0) {
873 new_size = old_size + new_size;
876 if (new_size < 256 * 1024 * 1024) {
880 if (new_size > device->bdev->bd_inode->i_size) {
885 do_div(new_size, root->sectorsize);
886 new_size *= root->sectorsize;
888 printk(KERN_INFO "new size for %s is %llu\n",
889 device->name, (unsigned long long)new_size);
891 if (new_size > old_size) {
892 trans = btrfs_start_transaction(root, 0);
893 ret = btrfs_grow_device(trans, device, new_size);
894 btrfs_commit_transaction(trans, root);
896 ret = btrfs_shrink_device(device, new_size);
900 mutex_unlock(&root->fs_info->volume_mutex);
905 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
912 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
913 struct file *src_file;
917 if (root->fs_info->sb->s_flags & MS_RDONLY)
920 namelen = strlen(name);
921 if (strchr(name, '/')) {
927 ret = btrfs_mksubvol(&file->f_path, name, namelen,
928 NULL, transid, readonly);
930 struct inode *src_inode;
937 src_inode = src_file->f_path.dentry->d_inode;
938 if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
939 printk(KERN_INFO "btrfs: Snapshot src from "
945 ret = btrfs_mksubvol(&file->f_path, name, namelen,
946 BTRFS_I(src_inode)->root,
954 static noinline int btrfs_ioctl_snap_create(struct file *file,
955 void __user *arg, int subvol)
957 struct btrfs_ioctl_vol_args *vol_args;
960 vol_args = memdup_user(arg, sizeof(*vol_args));
961 if (IS_ERR(vol_args))
962 return PTR_ERR(vol_args);
963 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
965 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
966 vol_args->fd, subvol,
973 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
974 void __user *arg, int subvol)
976 struct btrfs_ioctl_vol_args_v2 *vol_args;
980 bool readonly = false;
982 vol_args = memdup_user(arg, sizeof(*vol_args));
983 if (IS_ERR(vol_args))
984 return PTR_ERR(vol_args);
985 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
987 if (vol_args->flags &
988 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY)) {
993 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
995 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
998 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
999 vol_args->fd, subvol,
1002 if (ret == 0 && ptr &&
1004 offsetof(struct btrfs_ioctl_vol_args_v2,
1005 transid), ptr, sizeof(*ptr)))
1012 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1015 struct inode *inode = fdentry(file)->d_inode;
1016 struct btrfs_root *root = BTRFS_I(inode)->root;
1020 if (inode->i_ino != BTRFS_FIRST_FREE_OBJECTID)
1023 down_read(&root->fs_info->subvol_sem);
1024 if (btrfs_root_readonly(root))
1025 flags |= BTRFS_SUBVOL_RDONLY;
1026 up_read(&root->fs_info->subvol_sem);
1028 if (copy_to_user(arg, &flags, sizeof(flags)))
1034 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1037 struct inode *inode = fdentry(file)->d_inode;
1038 struct btrfs_root *root = BTRFS_I(inode)->root;
1039 struct btrfs_trans_handle *trans;
1044 if (root->fs_info->sb->s_flags & MS_RDONLY)
1047 if (inode->i_ino != BTRFS_FIRST_FREE_OBJECTID)
1050 if (copy_from_user(&flags, arg, sizeof(flags)))
1053 if (flags & ~BTRFS_SUBVOL_CREATE_ASYNC)
1056 if (flags & ~BTRFS_SUBVOL_RDONLY)
1059 down_write(&root->fs_info->subvol_sem);
1062 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1065 root_flags = btrfs_root_flags(&root->root_item);
1066 if (flags & BTRFS_SUBVOL_RDONLY)
1067 btrfs_set_root_flags(&root->root_item,
1068 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1070 btrfs_set_root_flags(&root->root_item,
1071 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1073 trans = btrfs_start_transaction(root, 1);
1074 if (IS_ERR(trans)) {
1075 ret = PTR_ERR(trans);
1079 ret = btrfs_update_root(trans, root,
1080 &root->root_key, &root->root_item);
1082 btrfs_commit_transaction(trans, root);
1085 btrfs_set_root_flags(&root->root_item, root_flags);
1087 up_write(&root->fs_info->subvol_sem);
1092 * helper to check if the subvolume references other subvolumes
1094 static noinline int may_destroy_subvol(struct btrfs_root *root)
1096 struct btrfs_path *path;
1097 struct btrfs_key key;
1100 path = btrfs_alloc_path();
1104 key.objectid = root->root_key.objectid;
1105 key.type = BTRFS_ROOT_REF_KEY;
1106 key.offset = (u64)-1;
1108 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1115 if (path->slots[0] > 0) {
1117 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1118 if (key.objectid == root->root_key.objectid &&
1119 key.type == BTRFS_ROOT_REF_KEY)
1123 btrfs_free_path(path);
1127 static noinline int key_in_sk(struct btrfs_key *key,
1128 struct btrfs_ioctl_search_key *sk)
1130 struct btrfs_key test;
1133 test.objectid = sk->min_objectid;
1134 test.type = sk->min_type;
1135 test.offset = sk->min_offset;
1137 ret = btrfs_comp_cpu_keys(key, &test);
1141 test.objectid = sk->max_objectid;
1142 test.type = sk->max_type;
1143 test.offset = sk->max_offset;
1145 ret = btrfs_comp_cpu_keys(key, &test);
1151 static noinline int copy_to_sk(struct btrfs_root *root,
1152 struct btrfs_path *path,
1153 struct btrfs_key *key,
1154 struct btrfs_ioctl_search_key *sk,
1156 unsigned long *sk_offset,
1160 struct extent_buffer *leaf;
1161 struct btrfs_ioctl_search_header sh;
1162 unsigned long item_off;
1163 unsigned long item_len;
1170 leaf = path->nodes[0];
1171 slot = path->slots[0];
1172 nritems = btrfs_header_nritems(leaf);
1174 if (btrfs_header_generation(leaf) > sk->max_transid) {
1178 found_transid = btrfs_header_generation(leaf);
1180 for (i = slot; i < nritems; i++) {
1181 item_off = btrfs_item_ptr_offset(leaf, i);
1182 item_len = btrfs_item_size_nr(leaf, i);
1184 if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1187 if (sizeof(sh) + item_len + *sk_offset >
1188 BTRFS_SEARCH_ARGS_BUFSIZE) {
1193 btrfs_item_key_to_cpu(leaf, key, i);
1194 if (!key_in_sk(key, sk))
1197 sh.objectid = key->objectid;
1198 sh.offset = key->offset;
1199 sh.type = key->type;
1201 sh.transid = found_transid;
1203 /* copy search result header */
1204 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1205 *sk_offset += sizeof(sh);
1208 char *p = buf + *sk_offset;
1210 read_extent_buffer(leaf, p,
1211 item_off, item_len);
1212 *sk_offset += item_len;
1216 if (*num_found >= sk->nr_items)
1221 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1223 else if (key->type < (u8)-1 && key->type < sk->max_type) {
1226 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1233 *num_found += found;
1237 static noinline int search_ioctl(struct inode *inode,
1238 struct btrfs_ioctl_search_args *args)
1240 struct btrfs_root *root;
1241 struct btrfs_key key;
1242 struct btrfs_key max_key;
1243 struct btrfs_path *path;
1244 struct btrfs_ioctl_search_key *sk = &args->key;
1245 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1248 unsigned long sk_offset = 0;
1250 path = btrfs_alloc_path();
1254 if (sk->tree_id == 0) {
1255 /* search the root of the inode that was passed */
1256 root = BTRFS_I(inode)->root;
1258 key.objectid = sk->tree_id;
1259 key.type = BTRFS_ROOT_ITEM_KEY;
1260 key.offset = (u64)-1;
1261 root = btrfs_read_fs_root_no_name(info, &key);
1263 printk(KERN_ERR "could not find root %llu\n",
1265 btrfs_free_path(path);
1270 key.objectid = sk->min_objectid;
1271 key.type = sk->min_type;
1272 key.offset = sk->min_offset;
1274 max_key.objectid = sk->max_objectid;
1275 max_key.type = sk->max_type;
1276 max_key.offset = sk->max_offset;
1278 path->keep_locks = 1;
1281 ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1288 ret = copy_to_sk(root, path, &key, sk, args->buf,
1289 &sk_offset, &num_found);
1290 btrfs_release_path(root, path);
1291 if (ret || num_found >= sk->nr_items)
1297 sk->nr_items = num_found;
1298 btrfs_free_path(path);
1302 static noinline int btrfs_ioctl_tree_search(struct file *file,
1305 struct btrfs_ioctl_search_args *args;
1306 struct inode *inode;
1309 if (!capable(CAP_SYS_ADMIN))
1312 args = memdup_user(argp, sizeof(*args));
1314 return PTR_ERR(args);
1316 inode = fdentry(file)->d_inode;
1317 ret = search_ioctl(inode, args);
1318 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1325 * Search INODE_REFs to identify path name of 'dirid' directory
1326 * in a 'tree_id' tree. and sets path name to 'name'.
1328 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1329 u64 tree_id, u64 dirid, char *name)
1331 struct btrfs_root *root;
1332 struct btrfs_key key;
1338 struct btrfs_inode_ref *iref;
1339 struct extent_buffer *l;
1340 struct btrfs_path *path;
1342 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1347 path = btrfs_alloc_path();
1351 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1353 key.objectid = tree_id;
1354 key.type = BTRFS_ROOT_ITEM_KEY;
1355 key.offset = (u64)-1;
1356 root = btrfs_read_fs_root_no_name(info, &key);
1358 printk(KERN_ERR "could not find root %llu\n", tree_id);
1363 key.objectid = dirid;
1364 key.type = BTRFS_INODE_REF_KEY;
1365 key.offset = (u64)-1;
1368 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1373 slot = path->slots[0];
1374 if (ret > 0 && slot > 0)
1376 btrfs_item_key_to_cpu(l, &key, slot);
1378 if (ret > 0 && (key.objectid != dirid ||
1379 key.type != BTRFS_INODE_REF_KEY)) {
1384 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1385 len = btrfs_inode_ref_name_len(l, iref);
1387 total_len += len + 1;
1392 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1394 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1397 btrfs_release_path(root, path);
1398 key.objectid = key.offset;
1399 key.offset = (u64)-1;
1400 dirid = key.objectid;
1405 memcpy(name, ptr, total_len);
1406 name[total_len]='\0';
1409 btrfs_free_path(path);
1413 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1416 struct btrfs_ioctl_ino_lookup_args *args;
1417 struct inode *inode;
1420 if (!capable(CAP_SYS_ADMIN))
1423 args = memdup_user(argp, sizeof(*args));
1425 return PTR_ERR(args);
1427 inode = fdentry(file)->d_inode;
1429 if (args->treeid == 0)
1430 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1432 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1433 args->treeid, args->objectid,
1436 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1443 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1446 struct dentry *parent = fdentry(file);
1447 struct dentry *dentry;
1448 struct inode *dir = parent->d_inode;
1449 struct inode *inode;
1450 struct btrfs_root *root = BTRFS_I(dir)->root;
1451 struct btrfs_root *dest = NULL;
1452 struct btrfs_ioctl_vol_args *vol_args;
1453 struct btrfs_trans_handle *trans;
1458 vol_args = memdup_user(arg, sizeof(*vol_args));
1459 if (IS_ERR(vol_args))
1460 return PTR_ERR(vol_args);
1462 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1463 namelen = strlen(vol_args->name);
1464 if (strchr(vol_args->name, '/') ||
1465 strncmp(vol_args->name, "..", namelen) == 0) {
1470 err = mnt_want_write(file->f_path.mnt);
1474 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
1475 dentry = lookup_one_len(vol_args->name, parent, namelen);
1476 if (IS_ERR(dentry)) {
1477 err = PTR_ERR(dentry);
1478 goto out_unlock_dir;
1481 if (!dentry->d_inode) {
1486 inode = dentry->d_inode;
1487 dest = BTRFS_I(inode)->root;
1488 if (!capable(CAP_SYS_ADMIN)){
1490 * Regular user. Only allow this with a special mount
1491 * option, when the user has write+exec access to the
1492 * subvol root, and when rmdir(2) would have been
1495 * Note that this is _not_ check that the subvol is
1496 * empty or doesn't contain data that we wouldn't
1497 * otherwise be able to delete.
1499 * Users who want to delete empty subvols should try
1503 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1507 * Do not allow deletion if the parent dir is the same
1508 * as the dir to be deleted. That means the ioctl
1509 * must be called on the dentry referencing the root
1510 * of the subvol, not a random directory contained
1517 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
1521 /* check if subvolume may be deleted by a non-root user */
1522 err = btrfs_may_delete(dir, dentry, 1);
1527 if (inode->i_ino != BTRFS_FIRST_FREE_OBJECTID) {
1532 mutex_lock(&inode->i_mutex);
1533 err = d_invalidate(dentry);
1537 down_write(&root->fs_info->subvol_sem);
1539 err = may_destroy_subvol(dest);
1543 trans = btrfs_start_transaction(root, 0);
1544 if (IS_ERR(trans)) {
1545 err = PTR_ERR(trans);
1548 trans->block_rsv = &root->fs_info->global_block_rsv;
1550 ret = btrfs_unlink_subvol(trans, root, dir,
1551 dest->root_key.objectid,
1552 dentry->d_name.name,
1553 dentry->d_name.len);
1556 btrfs_record_root_in_trans(trans, dest);
1558 memset(&dest->root_item.drop_progress, 0,
1559 sizeof(dest->root_item.drop_progress));
1560 dest->root_item.drop_level = 0;
1561 btrfs_set_root_refs(&dest->root_item, 0);
1563 if (!xchg(&dest->orphan_item_inserted, 1)) {
1564 ret = btrfs_insert_orphan_item(trans,
1565 root->fs_info->tree_root,
1566 dest->root_key.objectid);
1570 ret = btrfs_end_transaction(trans, root);
1572 inode->i_flags |= S_DEAD;
1574 up_write(&root->fs_info->subvol_sem);
1576 mutex_unlock(&inode->i_mutex);
1578 shrink_dcache_sb(root->fs_info->sb);
1579 btrfs_invalidate_inodes(dest);
1585 mutex_unlock(&dir->i_mutex);
1586 mnt_drop_write(file->f_path.mnt);
1592 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
1594 struct inode *inode = fdentry(file)->d_inode;
1595 struct btrfs_root *root = BTRFS_I(inode)->root;
1596 struct btrfs_ioctl_defrag_range_args *range;
1599 if (btrfs_root_readonly(root))
1602 ret = mnt_want_write(file->f_path.mnt);
1606 switch (inode->i_mode & S_IFMT) {
1608 if (!capable(CAP_SYS_ADMIN)) {
1612 ret = btrfs_defrag_root(root, 0);
1615 ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
1618 if (!(file->f_mode & FMODE_WRITE)) {
1623 range = kzalloc(sizeof(*range), GFP_KERNEL);
1630 if (copy_from_user(range, argp,
1636 /* compression requires us to start the IO */
1637 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1638 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
1639 range->extent_thresh = (u32)-1;
1642 /* the rest are all set to zero by kzalloc */
1643 range->len = (u64)-1;
1645 ret = btrfs_defrag_file(file, range);
1652 mnt_drop_write(file->f_path.mnt);
1656 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
1658 struct btrfs_ioctl_vol_args *vol_args;
1661 if (!capable(CAP_SYS_ADMIN))
1664 vol_args = memdup_user(arg, sizeof(*vol_args));
1665 if (IS_ERR(vol_args))
1666 return PTR_ERR(vol_args);
1668 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1669 ret = btrfs_init_new_device(root, vol_args->name);
1675 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
1677 struct btrfs_ioctl_vol_args *vol_args;
1680 if (!capable(CAP_SYS_ADMIN))
1683 if (root->fs_info->sb->s_flags & MS_RDONLY)
1686 vol_args = memdup_user(arg, sizeof(*vol_args));
1687 if (IS_ERR(vol_args))
1688 return PTR_ERR(vol_args);
1690 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1691 ret = btrfs_rm_device(root, vol_args->name);
1697 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
1698 u64 off, u64 olen, u64 destoff)
1700 struct inode *inode = fdentry(file)->d_inode;
1701 struct btrfs_root *root = BTRFS_I(inode)->root;
1702 struct file *src_file;
1704 struct btrfs_trans_handle *trans;
1705 struct btrfs_path *path;
1706 struct extent_buffer *leaf;
1708 struct btrfs_key key;
1713 u64 bs = root->fs_info->sb->s_blocksize;
1718 * - split compressed inline extents. annoying: we need to
1719 * decompress into destination's address_space (the file offset
1720 * may change, so source mapping won't do), then recompress (or
1721 * otherwise reinsert) a subrange.
1722 * - allow ranges within the same file to be cloned (provided
1723 * they don't overlap)?
1726 /* the destination must be opened for writing */
1727 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
1730 if (btrfs_root_readonly(root))
1733 ret = mnt_want_write(file->f_path.mnt);
1737 src_file = fget(srcfd);
1740 goto out_drop_write;
1743 src = src_file->f_dentry->d_inode;
1749 /* the src must be open for reading */
1750 if (!(src_file->f_mode & FMODE_READ))
1754 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
1758 if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
1762 buf = vmalloc(btrfs_level_size(root, 0));
1766 path = btrfs_alloc_path();
1774 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
1775 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
1777 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
1778 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
1781 /* determine range to clone */
1783 if (off + len > src->i_size || off + len < off)
1786 olen = len = src->i_size - off;
1787 /* if we extend to eof, continue to block boundary */
1788 if (off + len == src->i_size)
1789 len = ALIGN(src->i_size, bs) - off;
1791 /* verify the end result is block aligned */
1792 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
1793 !IS_ALIGNED(destoff, bs))
1796 /* do any pending delalloc/csum calc on src, one way or
1797 another, and lock file content */
1799 struct btrfs_ordered_extent *ordered;
1800 lock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
1801 ordered = btrfs_lookup_first_ordered_extent(src, off+len);
1803 !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len,
1804 EXTENT_DELALLOC, 0, NULL))
1806 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
1808 btrfs_put_ordered_extent(ordered);
1809 btrfs_wait_ordered_range(src, off, len);
1813 key.objectid = src->i_ino;
1814 key.type = BTRFS_EXTENT_DATA_KEY;
1819 * note the key will change type as we walk through the
1822 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1826 nritems = btrfs_header_nritems(path->nodes[0]);
1827 if (path->slots[0] >= nritems) {
1828 ret = btrfs_next_leaf(root, path);
1833 nritems = btrfs_header_nritems(path->nodes[0]);
1835 leaf = path->nodes[0];
1836 slot = path->slots[0];
1838 btrfs_item_key_to_cpu(leaf, &key, slot);
1839 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
1840 key.objectid != src->i_ino)
1843 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
1844 struct btrfs_file_extent_item *extent;
1847 struct btrfs_key new_key;
1848 u64 disko = 0, diskl = 0;
1849 u64 datao = 0, datal = 0;
1853 size = btrfs_item_size_nr(leaf, slot);
1854 read_extent_buffer(leaf, buf,
1855 btrfs_item_ptr_offset(leaf, slot),
1858 extent = btrfs_item_ptr(leaf, slot,
1859 struct btrfs_file_extent_item);
1860 comp = btrfs_file_extent_compression(leaf, extent);
1861 type = btrfs_file_extent_type(leaf, extent);
1862 if (type == BTRFS_FILE_EXTENT_REG ||
1863 type == BTRFS_FILE_EXTENT_PREALLOC) {
1864 disko = btrfs_file_extent_disk_bytenr(leaf,
1866 diskl = btrfs_file_extent_disk_num_bytes(leaf,
1868 datao = btrfs_file_extent_offset(leaf, extent);
1869 datal = btrfs_file_extent_num_bytes(leaf,
1871 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
1872 /* take upper bound, may be compressed */
1873 datal = btrfs_file_extent_ram_bytes(leaf,
1876 btrfs_release_path(root, path);
1878 if (key.offset + datal <= off ||
1879 key.offset >= off+len)
1882 memcpy(&new_key, &key, sizeof(new_key));
1883 new_key.objectid = inode->i_ino;
1884 new_key.offset = key.offset + destoff - off;
1886 trans = btrfs_start_transaction(root, 1);
1887 if (IS_ERR(trans)) {
1888 ret = PTR_ERR(trans);
1892 if (type == BTRFS_FILE_EXTENT_REG ||
1893 type == BTRFS_FILE_EXTENT_PREALLOC) {
1894 if (off > key.offset) {
1895 datao += off - key.offset;
1896 datal -= off - key.offset;
1899 if (key.offset + datal > off + len)
1900 datal = off + len - key.offset;
1902 ret = btrfs_drop_extents(trans, inode,
1904 new_key.offset + datal,
1908 ret = btrfs_insert_empty_item(trans, root, path,
1912 leaf = path->nodes[0];
1913 slot = path->slots[0];
1914 write_extent_buffer(leaf, buf,
1915 btrfs_item_ptr_offset(leaf, slot),
1918 extent = btrfs_item_ptr(leaf, slot,
1919 struct btrfs_file_extent_item);
1921 /* disko == 0 means it's a hole */
1925 btrfs_set_file_extent_offset(leaf, extent,
1927 btrfs_set_file_extent_num_bytes(leaf, extent,
1930 inode_add_bytes(inode, datal);
1931 ret = btrfs_inc_extent_ref(trans, root,
1933 root->root_key.objectid,
1935 new_key.offset - datao);
1938 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
1941 if (off > key.offset) {
1942 skip = off - key.offset;
1943 new_key.offset += skip;
1946 if (key.offset + datal > off+len)
1947 trim = key.offset + datal - (off+len);
1949 if (comp && (skip || trim)) {
1951 btrfs_end_transaction(trans, root);
1954 size -= skip + trim;
1955 datal -= skip + trim;
1957 ret = btrfs_drop_extents(trans, inode,
1959 new_key.offset + datal,
1963 ret = btrfs_insert_empty_item(trans, root, path,
1969 btrfs_file_extent_calc_inline_size(0);
1970 memmove(buf+start, buf+start+skip,
1974 leaf = path->nodes[0];
1975 slot = path->slots[0];
1976 write_extent_buffer(leaf, buf,
1977 btrfs_item_ptr_offset(leaf, slot),
1979 inode_add_bytes(inode, datal);
1982 btrfs_mark_buffer_dirty(leaf);
1983 btrfs_release_path(root, path);
1985 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1988 * we round up to the block size at eof when
1989 * determining which extents to clone above,
1990 * but shouldn't round up the file size
1992 endoff = new_key.offset + datal;
1993 if (endoff > destoff+olen)
1994 endoff = destoff+olen;
1995 if (endoff > inode->i_size)
1996 btrfs_i_size_write(inode, endoff);
1998 BTRFS_I(inode)->flags = BTRFS_I(src)->flags;
1999 ret = btrfs_update_inode(trans, root, inode);
2001 btrfs_end_transaction(trans, root);
2004 btrfs_release_path(root, path);
2009 btrfs_release_path(root, path);
2010 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
2012 mutex_unlock(&src->i_mutex);
2013 mutex_unlock(&inode->i_mutex);
2015 btrfs_free_path(path);
2019 mnt_drop_write(file->f_path.mnt);
2023 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
2025 struct btrfs_ioctl_clone_range_args args;
2027 if (copy_from_user(&args, argp, sizeof(args)))
2029 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2030 args.src_length, args.dest_offset);
2034 * there are many ways the trans_start and trans_end ioctls can lead
2035 * to deadlocks. They should only be used by applications that
2036 * basically own the machine, and have a very in depth understanding
2037 * of all the possible deadlocks and enospc problems.
2039 static long btrfs_ioctl_trans_start(struct file *file)
2041 struct inode *inode = fdentry(file)->d_inode;
2042 struct btrfs_root *root = BTRFS_I(inode)->root;
2043 struct btrfs_trans_handle *trans;
2047 if (!capable(CAP_SYS_ADMIN))
2051 if (file->private_data)
2055 if (btrfs_root_readonly(root))
2058 ret = mnt_want_write(file->f_path.mnt);
2062 mutex_lock(&root->fs_info->trans_mutex);
2063 root->fs_info->open_ioctl_trans++;
2064 mutex_unlock(&root->fs_info->trans_mutex);
2067 trans = btrfs_start_ioctl_transaction(root, 0);
2071 file->private_data = trans;
2075 mutex_lock(&root->fs_info->trans_mutex);
2076 root->fs_info->open_ioctl_trans--;
2077 mutex_unlock(&root->fs_info->trans_mutex);
2078 mnt_drop_write(file->f_path.mnt);
2083 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2085 struct inode *inode = fdentry(file)->d_inode;
2086 struct btrfs_root *root = BTRFS_I(inode)->root;
2087 struct btrfs_root *new_root;
2088 struct btrfs_dir_item *di;
2089 struct btrfs_trans_handle *trans;
2090 struct btrfs_path *path;
2091 struct btrfs_key location;
2092 struct btrfs_disk_key disk_key;
2093 struct btrfs_super_block *disk_super;
2098 if (!capable(CAP_SYS_ADMIN))
2101 if (copy_from_user(&objectid, argp, sizeof(objectid)))
2105 objectid = root->root_key.objectid;
2107 location.objectid = objectid;
2108 location.type = BTRFS_ROOT_ITEM_KEY;
2109 location.offset = (u64)-1;
2111 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2112 if (IS_ERR(new_root))
2113 return PTR_ERR(new_root);
2115 if (btrfs_root_refs(&new_root->root_item) == 0)
2118 path = btrfs_alloc_path();
2121 path->leave_spinning = 1;
2123 trans = btrfs_start_transaction(root, 1);
2125 btrfs_free_path(path);
2129 dir_id = btrfs_super_root_dir(&root->fs_info->super_copy);
2130 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2131 dir_id, "default", 7, 1);
2132 if (IS_ERR_OR_NULL(di)) {
2133 btrfs_free_path(path);
2134 btrfs_end_transaction(trans, root);
2135 printk(KERN_ERR "Umm, you don't have the default dir item, "
2136 "this isn't going to work\n");
2140 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2141 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2142 btrfs_mark_buffer_dirty(path->nodes[0]);
2143 btrfs_free_path(path);
2145 disk_super = &root->fs_info->super_copy;
2146 features = btrfs_super_incompat_flags(disk_super);
2147 if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) {
2148 features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL;
2149 btrfs_set_super_incompat_flags(disk_super, features);
2151 btrfs_end_transaction(trans, root);
2156 static void get_block_group_info(struct list_head *groups_list,
2157 struct btrfs_ioctl_space_info *space)
2159 struct btrfs_block_group_cache *block_group;
2161 space->total_bytes = 0;
2162 space->used_bytes = 0;
2164 list_for_each_entry(block_group, groups_list, list) {
2165 space->flags = block_group->flags;
2166 space->total_bytes += block_group->key.offset;
2167 space->used_bytes +=
2168 btrfs_block_group_used(&block_group->item);
2172 long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2174 struct btrfs_ioctl_space_args space_args;
2175 struct btrfs_ioctl_space_info space;
2176 struct btrfs_ioctl_space_info *dest;
2177 struct btrfs_ioctl_space_info *dest_orig;
2178 struct btrfs_ioctl_space_info *user_dest;
2179 struct btrfs_space_info *info;
2180 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2181 BTRFS_BLOCK_GROUP_SYSTEM,
2182 BTRFS_BLOCK_GROUP_METADATA,
2183 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2190 if (copy_from_user(&space_args,
2191 (struct btrfs_ioctl_space_args __user *)arg,
2192 sizeof(space_args)))
2195 for (i = 0; i < num_types; i++) {
2196 struct btrfs_space_info *tmp;
2200 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2202 if (tmp->flags == types[i]) {
2212 down_read(&info->groups_sem);
2213 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2214 if (!list_empty(&info->block_groups[c]))
2217 up_read(&info->groups_sem);
2220 /* space_slots == 0 means they are asking for a count */
2221 if (space_args.space_slots == 0) {
2222 space_args.total_spaces = slot_count;
2226 slot_count = min_t(int, space_args.space_slots, slot_count);
2228 alloc_size = sizeof(*dest) * slot_count;
2230 /* we generally have at most 6 or so space infos, one for each raid
2231 * level. So, a whole page should be more than enough for everyone
2233 if (alloc_size > PAGE_CACHE_SIZE)
2236 space_args.total_spaces = 0;
2237 dest = kmalloc(alloc_size, GFP_NOFS);
2242 /* now we have a buffer to copy into */
2243 for (i = 0; i < num_types; i++) {
2244 struct btrfs_space_info *tmp;
2248 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2250 if (tmp->flags == types[i]) {
2259 down_read(&info->groups_sem);
2260 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2261 if (!list_empty(&info->block_groups[c])) {
2262 get_block_group_info(&info->block_groups[c],
2264 memcpy(dest, &space, sizeof(space));
2266 space_args.total_spaces++;
2269 up_read(&info->groups_sem);
2272 user_dest = (struct btrfs_ioctl_space_info *)
2273 (arg + sizeof(struct btrfs_ioctl_space_args));
2275 if (copy_to_user(user_dest, dest_orig, alloc_size))
2280 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
2287 * there are many ways the trans_start and trans_end ioctls can lead
2288 * to deadlocks. They should only be used by applications that
2289 * basically own the machine, and have a very in depth understanding
2290 * of all the possible deadlocks and enospc problems.
2292 long btrfs_ioctl_trans_end(struct file *file)
2294 struct inode *inode = fdentry(file)->d_inode;
2295 struct btrfs_root *root = BTRFS_I(inode)->root;
2296 struct btrfs_trans_handle *trans;
2298 trans = file->private_data;
2301 file->private_data = NULL;
2303 btrfs_end_transaction(trans, root);
2305 mutex_lock(&root->fs_info->trans_mutex);
2306 root->fs_info->open_ioctl_trans--;
2307 mutex_unlock(&root->fs_info->trans_mutex);
2309 mnt_drop_write(file->f_path.mnt);
2313 static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
2315 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2316 struct btrfs_trans_handle *trans;
2319 trans = btrfs_start_transaction(root, 0);
2320 transid = trans->transid;
2321 btrfs_commit_transaction_async(trans, root, 0);
2324 if (copy_to_user(argp, &transid, sizeof(transid)))
2329 static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
2331 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2335 if (copy_from_user(&transid, argp, sizeof(transid)))
2338 transid = 0; /* current trans */
2340 return btrfs_wait_for_commit(root, transid);
2343 long btrfs_ioctl(struct file *file, unsigned int
2344 cmd, unsigned long arg)
2346 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
2347 void __user *argp = (void __user *)arg;
2350 case FS_IOC_GETFLAGS:
2351 return btrfs_ioctl_getflags(file, argp);
2352 case FS_IOC_SETFLAGS:
2353 return btrfs_ioctl_setflags(file, argp);
2354 case FS_IOC_GETVERSION:
2355 return btrfs_ioctl_getversion(file, argp);
2356 case BTRFS_IOC_SNAP_CREATE:
2357 return btrfs_ioctl_snap_create(file, argp, 0);
2358 case BTRFS_IOC_SNAP_CREATE_V2:
2359 return btrfs_ioctl_snap_create_v2(file, argp, 0);
2360 case BTRFS_IOC_SUBVOL_CREATE:
2361 return btrfs_ioctl_snap_create(file, argp, 1);
2362 case BTRFS_IOC_SNAP_DESTROY:
2363 return btrfs_ioctl_snap_destroy(file, argp);
2364 case BTRFS_IOC_SUBVOL_GETFLAGS:
2365 return btrfs_ioctl_subvol_getflags(file, argp);
2366 case BTRFS_IOC_SUBVOL_SETFLAGS:
2367 return btrfs_ioctl_subvol_setflags(file, argp);
2368 case BTRFS_IOC_DEFAULT_SUBVOL:
2369 return btrfs_ioctl_default_subvol(file, argp);
2370 case BTRFS_IOC_DEFRAG:
2371 return btrfs_ioctl_defrag(file, NULL);
2372 case BTRFS_IOC_DEFRAG_RANGE:
2373 return btrfs_ioctl_defrag(file, argp);
2374 case BTRFS_IOC_RESIZE:
2375 return btrfs_ioctl_resize(root, argp);
2376 case BTRFS_IOC_ADD_DEV:
2377 return btrfs_ioctl_add_dev(root, argp);
2378 case BTRFS_IOC_RM_DEV:
2379 return btrfs_ioctl_rm_dev(root, argp);
2380 case BTRFS_IOC_BALANCE:
2381 return btrfs_balance(root->fs_info->dev_root);
2382 case BTRFS_IOC_CLONE:
2383 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
2384 case BTRFS_IOC_CLONE_RANGE:
2385 return btrfs_ioctl_clone_range(file, argp);
2386 case BTRFS_IOC_TRANS_START:
2387 return btrfs_ioctl_trans_start(file);
2388 case BTRFS_IOC_TRANS_END:
2389 return btrfs_ioctl_trans_end(file);
2390 case BTRFS_IOC_TREE_SEARCH:
2391 return btrfs_ioctl_tree_search(file, argp);
2392 case BTRFS_IOC_INO_LOOKUP:
2393 return btrfs_ioctl_ino_lookup(file, argp);
2394 case BTRFS_IOC_SPACE_INFO:
2395 return btrfs_ioctl_space_info(root, argp);
2396 case BTRFS_IOC_SYNC:
2397 btrfs_sync_fs(file->f_dentry->d_sb, 1);
2399 case BTRFS_IOC_START_SYNC:
2400 return btrfs_ioctl_start_sync(file, argp);
2401 case BTRFS_IOC_WAIT_SYNC:
2402 return btrfs_ioctl_wait_sync(file, argp);