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"
54 /* Mask out flags that are inappropriate for the given type of inode. */
55 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
59 else if (S_ISREG(mode))
60 return flags & ~FS_DIRSYNC_FL;
62 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
66 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
68 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
70 unsigned int iflags = 0;
72 if (flags & BTRFS_INODE_SYNC)
74 if (flags & BTRFS_INODE_IMMUTABLE)
75 iflags |= FS_IMMUTABLE_FL;
76 if (flags & BTRFS_INODE_APPEND)
77 iflags |= FS_APPEND_FL;
78 if (flags & BTRFS_INODE_NODUMP)
79 iflags |= FS_NODUMP_FL;
80 if (flags & BTRFS_INODE_NOATIME)
81 iflags |= FS_NOATIME_FL;
82 if (flags & BTRFS_INODE_DIRSYNC)
83 iflags |= FS_DIRSYNC_FL;
89 * Update inode->i_flags based on the btrfs internal flags.
91 void btrfs_update_iflags(struct inode *inode)
93 struct btrfs_inode *ip = BTRFS_I(inode);
95 inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
97 if (ip->flags & BTRFS_INODE_SYNC)
98 inode->i_flags |= S_SYNC;
99 if (ip->flags & BTRFS_INODE_IMMUTABLE)
100 inode->i_flags |= S_IMMUTABLE;
101 if (ip->flags & BTRFS_INODE_APPEND)
102 inode->i_flags |= S_APPEND;
103 if (ip->flags & BTRFS_INODE_NOATIME)
104 inode->i_flags |= S_NOATIME;
105 if (ip->flags & BTRFS_INODE_DIRSYNC)
106 inode->i_flags |= S_DIRSYNC;
110 * Inherit flags from the parent inode.
112 * Unlike extN we don't have any flags we don't want to inherit currently.
114 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
121 flags = BTRFS_I(dir)->flags;
123 if (S_ISREG(inode->i_mode))
124 flags &= ~BTRFS_INODE_DIRSYNC;
125 else if (!S_ISDIR(inode->i_mode))
126 flags &= (BTRFS_INODE_NODUMP | BTRFS_INODE_NOATIME);
128 BTRFS_I(inode)->flags = flags;
129 btrfs_update_iflags(inode);
132 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
134 struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode);
135 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
137 if (copy_to_user(arg, &flags, sizeof(flags)))
142 static int check_flags(unsigned int flags)
144 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
145 FS_NOATIME_FL | FS_NODUMP_FL | \
146 FS_SYNC_FL | FS_DIRSYNC_FL | \
147 FS_NOCOMP_FL | FS_COMPR_FL | \
148 FS_NOCOW_FL | FS_COW_FL))
151 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
154 if ((flags & FS_NOCOW_FL) && (flags & FS_COW_FL))
160 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
162 struct inode *inode = file->f_path.dentry->d_inode;
163 struct btrfs_inode *ip = BTRFS_I(inode);
164 struct btrfs_root *root = ip->root;
165 struct btrfs_trans_handle *trans;
166 unsigned int flags, oldflags;
169 if (btrfs_root_readonly(root))
172 if (copy_from_user(&flags, arg, sizeof(flags)))
175 ret = check_flags(flags);
179 if (!is_owner_or_cap(inode))
182 mutex_lock(&inode->i_mutex);
184 flags = btrfs_mask_flags(inode->i_mode, flags);
185 oldflags = btrfs_flags_to_ioctl(ip->flags);
186 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
187 if (!capable(CAP_LINUX_IMMUTABLE)) {
193 ret = mnt_want_write(file->f_path.mnt);
197 if (flags & FS_SYNC_FL)
198 ip->flags |= BTRFS_INODE_SYNC;
200 ip->flags &= ~BTRFS_INODE_SYNC;
201 if (flags & FS_IMMUTABLE_FL)
202 ip->flags |= BTRFS_INODE_IMMUTABLE;
204 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
205 if (flags & FS_APPEND_FL)
206 ip->flags |= BTRFS_INODE_APPEND;
208 ip->flags &= ~BTRFS_INODE_APPEND;
209 if (flags & FS_NODUMP_FL)
210 ip->flags |= BTRFS_INODE_NODUMP;
212 ip->flags &= ~BTRFS_INODE_NODUMP;
213 if (flags & FS_NOATIME_FL)
214 ip->flags |= BTRFS_INODE_NOATIME;
216 ip->flags &= ~BTRFS_INODE_NOATIME;
217 if (flags & FS_DIRSYNC_FL)
218 ip->flags |= BTRFS_INODE_DIRSYNC;
220 ip->flags &= ~BTRFS_INODE_DIRSYNC;
223 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
224 * flag may be changed automatically if compression code won't make
227 if (flags & FS_NOCOMP_FL) {
228 ip->flags &= ~BTRFS_INODE_COMPRESS;
229 ip->flags |= BTRFS_INODE_NOCOMPRESS;
230 } else if (flags & FS_COMPR_FL) {
231 ip->flags |= BTRFS_INODE_COMPRESS;
232 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
234 if (flags & FS_NOCOW_FL)
235 ip->flags |= BTRFS_INODE_NODATACOW;
236 else if (flags & FS_COW_FL)
237 ip->flags &= ~BTRFS_INODE_NODATACOW;
239 trans = btrfs_join_transaction(root, 1);
240 BUG_ON(IS_ERR(trans));
242 ret = btrfs_update_inode(trans, root, inode);
245 btrfs_update_iflags(inode);
246 inode->i_ctime = CURRENT_TIME;
247 btrfs_end_transaction(trans, root);
249 mnt_drop_write(file->f_path.mnt);
253 mutex_unlock(&inode->i_mutex);
257 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
259 struct inode *inode = file->f_path.dentry->d_inode;
261 return put_user(inode->i_generation, arg);
264 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
266 struct btrfs_root *root = fdentry(file)->d_sb->s_fs_info;
267 struct btrfs_fs_info *fs_info = root->fs_info;
268 struct btrfs_device *device;
269 struct request_queue *q;
270 struct fstrim_range range;
271 u64 minlen = ULLONG_MAX;
275 if (!capable(CAP_SYS_ADMIN))
278 mutex_lock(&fs_info->fs_devices->device_list_mutex);
279 list_for_each_entry(device, &fs_info->fs_devices->devices, dev_list) {
282 q = bdev_get_queue(device->bdev);
283 if (blk_queue_discard(q)) {
285 minlen = min((u64)q->limits.discard_granularity,
289 mutex_unlock(&fs_info->fs_devices->device_list_mutex);
293 if (copy_from_user(&range, arg, sizeof(range)))
296 range.minlen = max(range.minlen, minlen);
297 ret = btrfs_trim_fs(root, &range);
301 if (copy_to_user(arg, &range, sizeof(range)))
307 static noinline int create_subvol(struct btrfs_root *root,
308 struct dentry *dentry,
309 char *name, int namelen,
312 struct btrfs_trans_handle *trans;
313 struct btrfs_key key;
314 struct btrfs_root_item root_item;
315 struct btrfs_inode_item *inode_item;
316 struct extent_buffer *leaf;
317 struct btrfs_root *new_root;
318 struct dentry *parent = dget_parent(dentry);
323 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
326 ret = btrfs_find_free_objectid(NULL, root->fs_info->tree_root,
333 dir = parent->d_inode;
341 trans = btrfs_start_transaction(root, 6);
344 return PTR_ERR(trans);
347 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
348 0, objectid, NULL, 0, 0, 0);
354 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
355 btrfs_set_header_bytenr(leaf, leaf->start);
356 btrfs_set_header_generation(leaf, trans->transid);
357 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
358 btrfs_set_header_owner(leaf, objectid);
360 write_extent_buffer(leaf, root->fs_info->fsid,
361 (unsigned long)btrfs_header_fsid(leaf),
363 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
364 (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
366 btrfs_mark_buffer_dirty(leaf);
368 inode_item = &root_item.inode;
369 memset(inode_item, 0, sizeof(*inode_item));
370 inode_item->generation = cpu_to_le64(1);
371 inode_item->size = cpu_to_le64(3);
372 inode_item->nlink = cpu_to_le32(1);
373 inode_item->nbytes = cpu_to_le64(root->leafsize);
374 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
377 root_item.byte_limit = 0;
378 inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
380 btrfs_set_root_bytenr(&root_item, leaf->start);
381 btrfs_set_root_generation(&root_item, trans->transid);
382 btrfs_set_root_level(&root_item, 0);
383 btrfs_set_root_refs(&root_item, 1);
384 btrfs_set_root_used(&root_item, leaf->len);
385 btrfs_set_root_last_snapshot(&root_item, 0);
387 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
388 root_item.drop_level = 0;
390 btrfs_tree_unlock(leaf);
391 free_extent_buffer(leaf);
394 btrfs_set_root_dirid(&root_item, new_dirid);
396 key.objectid = objectid;
398 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
399 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
404 key.offset = (u64)-1;
405 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
406 BUG_ON(IS_ERR(new_root));
408 btrfs_record_root_in_trans(trans, new_root);
410 ret = btrfs_create_subvol_root(trans, new_root, new_dirid,
411 BTRFS_I(dir)->block_group);
413 * insert the directory item
415 ret = btrfs_set_inode_index(dir, &index);
418 ret = btrfs_insert_dir_item(trans, root,
419 name, namelen, dir->i_ino, &key,
420 BTRFS_FT_DIR, index);
424 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
425 ret = btrfs_update_inode(trans, root, dir);
428 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
429 objectid, root->root_key.objectid,
430 dir->i_ino, index, name, namelen);
434 d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
438 *async_transid = trans->transid;
439 err = btrfs_commit_transaction_async(trans, root, 1);
441 err = btrfs_commit_transaction(trans, root);
448 static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
449 char *name, int namelen, u64 *async_transid,
453 struct dentry *parent;
454 struct btrfs_pending_snapshot *pending_snapshot;
455 struct btrfs_trans_handle *trans;
461 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
462 if (!pending_snapshot)
465 btrfs_init_block_rsv(&pending_snapshot->block_rsv);
466 pending_snapshot->dentry = dentry;
467 pending_snapshot->root = root;
468 pending_snapshot->readonly = readonly;
470 trans = btrfs_start_transaction(root->fs_info->extent_root, 5);
472 ret = PTR_ERR(trans);
476 ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
479 list_add(&pending_snapshot->list,
480 &trans->transaction->pending_snapshots);
482 *async_transid = trans->transid;
483 ret = btrfs_commit_transaction_async(trans,
484 root->fs_info->extent_root, 1);
486 ret = btrfs_commit_transaction(trans,
487 root->fs_info->extent_root);
491 ret = pending_snapshot->error;
495 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
499 parent = dget_parent(dentry);
500 inode = btrfs_lookup_dentry(parent->d_inode, dentry);
503 ret = PTR_ERR(inode);
507 d_instantiate(dentry, inode);
510 kfree(pending_snapshot);
514 /* copy of check_sticky in fs/namei.c()
515 * It's inline, so penalty for filesystems that don't use sticky bit is
518 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
520 uid_t fsuid = current_fsuid();
522 if (!(dir->i_mode & S_ISVTX))
524 if (inode->i_uid == fsuid)
526 if (dir->i_uid == fsuid)
528 return !capable(CAP_FOWNER);
531 /* copy of may_delete in fs/namei.c()
532 * Check whether we can remove a link victim from directory dir, check
533 * whether the type of victim is right.
534 * 1. We can't do it if dir is read-only (done in permission())
535 * 2. We should have write and exec permissions on dir
536 * 3. We can't remove anything from append-only dir
537 * 4. We can't do anything with immutable dir (done in permission())
538 * 5. If the sticky bit on dir is set we should either
539 * a. be owner of dir, or
540 * b. be owner of victim, or
541 * c. have CAP_FOWNER capability
542 * 6. If the victim is append-only or immutable we can't do antyhing with
543 * links pointing to it.
544 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
545 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
546 * 9. We can't remove a root or mountpoint.
547 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
548 * nfs_async_unlink().
551 static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
555 if (!victim->d_inode)
558 BUG_ON(victim->d_parent->d_inode != dir);
559 audit_inode_child(victim, dir);
561 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
566 if (btrfs_check_sticky(dir, victim->d_inode)||
567 IS_APPEND(victim->d_inode)||
568 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
571 if (!S_ISDIR(victim->d_inode->i_mode))
575 } else if (S_ISDIR(victim->d_inode->i_mode))
579 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
584 /* copy of may_create in fs/namei.c() */
585 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
591 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
595 * Create a new subvolume below @parent. This is largely modeled after
596 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
597 * inside this filesystem so it's quite a bit simpler.
599 static noinline int btrfs_mksubvol(struct path *parent,
600 char *name, int namelen,
601 struct btrfs_root *snap_src,
602 u64 *async_transid, bool readonly)
604 struct inode *dir = parent->dentry->d_inode;
605 struct dentry *dentry;
608 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
610 dentry = lookup_one_len(name, parent->dentry, namelen);
611 error = PTR_ERR(dentry);
619 error = mnt_want_write(parent->mnt);
623 error = btrfs_may_create(dir, dentry);
627 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
629 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
633 error = create_snapshot(snap_src, dentry,
634 name, namelen, async_transid, readonly);
636 error = create_subvol(BTRFS_I(dir)->root, dentry,
637 name, namelen, async_transid);
640 fsnotify_mkdir(dir, dentry);
642 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
644 mnt_drop_write(parent->mnt);
648 mutex_unlock(&dir->i_mutex);
652 static int should_defrag_range(struct inode *inode, u64 start, u64 len,
653 int thresh, u64 *last_len, u64 *skip,
656 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
657 struct extent_map *em = NULL;
658 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
666 * make sure that once we start defragging and extent, we keep on
669 if (start < *defrag_end)
675 * hopefully we have this extent in the tree already, try without
676 * the full extent lock
678 read_lock(&em_tree->lock);
679 em = lookup_extent_mapping(em_tree, start, len);
680 read_unlock(&em_tree->lock);
683 /* get the big lock and read metadata off disk */
684 lock_extent(io_tree, start, start + len - 1, GFP_NOFS);
685 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
686 unlock_extent(io_tree, start, start + len - 1, GFP_NOFS);
692 /* this will cover holes, and inline extents */
693 if (em->block_start >= EXTENT_MAP_LAST_BYTE)
697 * we hit a real extent, if it is big don't bother defragging it again
699 if ((*last_len == 0 || *last_len >= thresh) && em->len >= thresh)
703 * last_len ends up being a counter of how many bytes we've defragged.
704 * every time we choose not to defrag an extent, we reset *last_len
705 * so that the next tiny extent will force a defrag.
707 * The end result of this is that tiny extents before a single big
708 * extent will force at least part of that big extent to be defragged.
712 *defrag_end = extent_map_end(em);
715 *skip = extent_map_end(em);
723 static int btrfs_defrag_file(struct file *file,
724 struct btrfs_ioctl_defrag_range_args *range)
726 struct inode *inode = fdentry(file)->d_inode;
727 struct btrfs_root *root = BTRFS_I(inode)->root;
728 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
729 struct btrfs_ordered_extent *ordered;
731 struct btrfs_super_block *disk_super;
732 unsigned long last_index;
733 unsigned long ra_pages = root->fs_info->bdi.ra_pages;
734 unsigned long total_read = 0;
743 int compress_type = BTRFS_COMPRESS_ZLIB;
745 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
746 if (range->compress_type > BTRFS_COMPRESS_TYPES)
748 if (range->compress_type)
749 compress_type = range->compress_type;
752 if (inode->i_size == 0)
755 if (range->start + range->len > range->start) {
756 last_index = min_t(u64, inode->i_size - 1,
757 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
759 last_index = (inode->i_size - 1) >> PAGE_CACHE_SHIFT;
762 i = range->start >> PAGE_CACHE_SHIFT;
763 while (i <= last_index) {
764 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
766 range->extent_thresh,
771 * the should_defrag function tells us how much to skip
772 * bump our counter by the suggested amount
774 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
775 i = max(i + 1, next);
779 if (total_read % ra_pages == 0) {
780 btrfs_force_ra(inode->i_mapping, &file->f_ra, file, i,
781 min(last_index, i + ra_pages - 1));
784 mutex_lock(&inode->i_mutex);
785 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
786 BTRFS_I(inode)->force_compress = compress_type;
788 ret = btrfs_delalloc_reserve_space(inode, PAGE_CACHE_SIZE);
792 if (inode->i_size == 0 ||
793 i > ((inode->i_size - 1) >> PAGE_CACHE_SHIFT)) {
795 goto err_reservations;
798 page = grab_cache_page(inode->i_mapping, i);
801 goto err_reservations;
804 if (!PageUptodate(page)) {
805 btrfs_readpage(NULL, page);
807 if (!PageUptodate(page)) {
809 page_cache_release(page);
811 goto err_reservations;
815 if (page->mapping != inode->i_mapping) {
817 page_cache_release(page);
821 wait_on_page_writeback(page);
823 if (PageDirty(page)) {
824 btrfs_delalloc_release_space(inode, PAGE_CACHE_SIZE);
828 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
829 page_end = page_start + PAGE_CACHE_SIZE - 1;
830 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
832 ordered = btrfs_lookup_ordered_extent(inode, page_start);
834 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
836 page_cache_release(page);
837 btrfs_start_ordered_extent(inode, ordered, 1);
838 btrfs_put_ordered_extent(ordered);
841 set_page_extent_mapped(page);
844 * this makes sure page_mkwrite is called on the
845 * page if it is dirtied again later
847 clear_page_dirty_for_io(page);
848 clear_extent_bits(&BTRFS_I(inode)->io_tree, page_start,
849 page_end, EXTENT_DIRTY | EXTENT_DELALLOC |
850 EXTENT_DO_ACCOUNTING, GFP_NOFS);
852 btrfs_set_extent_delalloc(inode, page_start, page_end, NULL);
853 ClearPageChecked(page);
854 set_page_dirty(page);
855 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
859 page_cache_release(page);
860 mutex_unlock(&inode->i_mutex);
862 balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
866 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
867 filemap_flush(inode->i_mapping);
869 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
870 /* the filemap_flush will queue IO into the worker threads, but
871 * we have to make sure the IO is actually started and that
872 * ordered extents get created before we return
874 atomic_inc(&root->fs_info->async_submit_draining);
875 while (atomic_read(&root->fs_info->nr_async_submits) ||
876 atomic_read(&root->fs_info->async_delalloc_pages)) {
877 wait_event(root->fs_info->async_submit_wait,
878 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
879 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
881 atomic_dec(&root->fs_info->async_submit_draining);
883 mutex_lock(&inode->i_mutex);
884 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
885 mutex_unlock(&inode->i_mutex);
888 disk_super = &root->fs_info->super_copy;
889 features = btrfs_super_incompat_flags(disk_super);
890 if (range->compress_type == BTRFS_COMPRESS_LZO) {
891 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
892 btrfs_set_super_incompat_flags(disk_super, features);
898 btrfs_delalloc_release_space(inode, PAGE_CACHE_SIZE);
900 mutex_unlock(&inode->i_mutex);
904 static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
910 struct btrfs_ioctl_vol_args *vol_args;
911 struct btrfs_trans_handle *trans;
912 struct btrfs_device *device = NULL;
918 if (root->fs_info->sb->s_flags & MS_RDONLY)
921 if (!capable(CAP_SYS_ADMIN))
924 vol_args = memdup_user(arg, sizeof(*vol_args));
925 if (IS_ERR(vol_args))
926 return PTR_ERR(vol_args);
928 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
930 mutex_lock(&root->fs_info->volume_mutex);
931 sizestr = vol_args->name;
932 devstr = strchr(sizestr, ':');
935 sizestr = devstr + 1;
937 devstr = vol_args->name;
938 devid = simple_strtoull(devstr, &end, 10);
939 printk(KERN_INFO "resizing devid %llu\n",
940 (unsigned long long)devid);
942 device = btrfs_find_device(root, devid, NULL, NULL);
944 printk(KERN_INFO "resizer unable to find device %llu\n",
945 (unsigned long long)devid);
949 if (!strcmp(sizestr, "max"))
950 new_size = device->bdev->bd_inode->i_size;
952 if (sizestr[0] == '-') {
955 } else if (sizestr[0] == '+') {
959 new_size = memparse(sizestr, NULL);
966 old_size = device->total_bytes;
969 if (new_size > old_size) {
973 new_size = old_size - new_size;
974 } else if (mod > 0) {
975 new_size = old_size + new_size;
978 if (new_size < 256 * 1024 * 1024) {
982 if (new_size > device->bdev->bd_inode->i_size) {
987 do_div(new_size, root->sectorsize);
988 new_size *= root->sectorsize;
990 printk(KERN_INFO "new size for %s is %llu\n",
991 device->name, (unsigned long long)new_size);
993 if (new_size > old_size) {
994 trans = btrfs_start_transaction(root, 0);
996 ret = PTR_ERR(trans);
999 ret = btrfs_grow_device(trans, device, new_size);
1000 btrfs_commit_transaction(trans, root);
1002 ret = btrfs_shrink_device(device, new_size);
1006 mutex_unlock(&root->fs_info->volume_mutex);
1011 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1018 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
1019 struct file *src_file;
1023 if (root->fs_info->sb->s_flags & MS_RDONLY)
1026 namelen = strlen(name);
1027 if (strchr(name, '/')) {
1033 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1034 NULL, transid, readonly);
1036 struct inode *src_inode;
1037 src_file = fget(fd);
1043 src_inode = src_file->f_path.dentry->d_inode;
1044 if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
1045 printk(KERN_INFO "btrfs: Snapshot src from "
1051 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1052 BTRFS_I(src_inode)->root,
1060 static noinline int btrfs_ioctl_snap_create(struct file *file,
1061 void __user *arg, int subvol)
1063 struct btrfs_ioctl_vol_args *vol_args;
1066 vol_args = memdup_user(arg, sizeof(*vol_args));
1067 if (IS_ERR(vol_args))
1068 return PTR_ERR(vol_args);
1069 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1071 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1072 vol_args->fd, subvol,
1079 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1080 void __user *arg, int subvol)
1082 struct btrfs_ioctl_vol_args_v2 *vol_args;
1086 bool readonly = false;
1088 vol_args = memdup_user(arg, sizeof(*vol_args));
1089 if (IS_ERR(vol_args))
1090 return PTR_ERR(vol_args);
1091 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1093 if (vol_args->flags &
1094 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY)) {
1099 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1101 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1104 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1105 vol_args->fd, subvol,
1108 if (ret == 0 && ptr &&
1110 offsetof(struct btrfs_ioctl_vol_args_v2,
1111 transid), ptr, sizeof(*ptr)))
1118 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1121 struct inode *inode = fdentry(file)->d_inode;
1122 struct btrfs_root *root = BTRFS_I(inode)->root;
1126 if (inode->i_ino != BTRFS_FIRST_FREE_OBJECTID)
1129 down_read(&root->fs_info->subvol_sem);
1130 if (btrfs_root_readonly(root))
1131 flags |= BTRFS_SUBVOL_RDONLY;
1132 up_read(&root->fs_info->subvol_sem);
1134 if (copy_to_user(arg, &flags, sizeof(flags)))
1140 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1143 struct inode *inode = fdentry(file)->d_inode;
1144 struct btrfs_root *root = BTRFS_I(inode)->root;
1145 struct btrfs_trans_handle *trans;
1150 if (root->fs_info->sb->s_flags & MS_RDONLY)
1153 if (inode->i_ino != BTRFS_FIRST_FREE_OBJECTID)
1156 if (copy_from_user(&flags, arg, sizeof(flags)))
1159 if (flags & BTRFS_SUBVOL_CREATE_ASYNC)
1162 if (flags & ~BTRFS_SUBVOL_RDONLY)
1165 if (!is_owner_or_cap(inode))
1168 down_write(&root->fs_info->subvol_sem);
1171 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1174 root_flags = btrfs_root_flags(&root->root_item);
1175 if (flags & BTRFS_SUBVOL_RDONLY)
1176 btrfs_set_root_flags(&root->root_item,
1177 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1179 btrfs_set_root_flags(&root->root_item,
1180 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1182 trans = btrfs_start_transaction(root, 1);
1183 if (IS_ERR(trans)) {
1184 ret = PTR_ERR(trans);
1188 ret = btrfs_update_root(trans, root->fs_info->tree_root,
1189 &root->root_key, &root->root_item);
1191 btrfs_commit_transaction(trans, root);
1194 btrfs_set_root_flags(&root->root_item, root_flags);
1196 up_write(&root->fs_info->subvol_sem);
1201 * helper to check if the subvolume references other subvolumes
1203 static noinline int may_destroy_subvol(struct btrfs_root *root)
1205 struct btrfs_path *path;
1206 struct btrfs_key key;
1209 path = btrfs_alloc_path();
1213 key.objectid = root->root_key.objectid;
1214 key.type = BTRFS_ROOT_REF_KEY;
1215 key.offset = (u64)-1;
1217 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1224 if (path->slots[0] > 0) {
1226 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1227 if (key.objectid == root->root_key.objectid &&
1228 key.type == BTRFS_ROOT_REF_KEY)
1232 btrfs_free_path(path);
1236 static noinline int key_in_sk(struct btrfs_key *key,
1237 struct btrfs_ioctl_search_key *sk)
1239 struct btrfs_key test;
1242 test.objectid = sk->min_objectid;
1243 test.type = sk->min_type;
1244 test.offset = sk->min_offset;
1246 ret = btrfs_comp_cpu_keys(key, &test);
1250 test.objectid = sk->max_objectid;
1251 test.type = sk->max_type;
1252 test.offset = sk->max_offset;
1254 ret = btrfs_comp_cpu_keys(key, &test);
1260 static noinline int copy_to_sk(struct btrfs_root *root,
1261 struct btrfs_path *path,
1262 struct btrfs_key *key,
1263 struct btrfs_ioctl_search_key *sk,
1265 unsigned long *sk_offset,
1269 struct extent_buffer *leaf;
1270 struct btrfs_ioctl_search_header sh;
1271 unsigned long item_off;
1272 unsigned long item_len;
1279 leaf = path->nodes[0];
1280 slot = path->slots[0];
1281 nritems = btrfs_header_nritems(leaf);
1283 if (btrfs_header_generation(leaf) > sk->max_transid) {
1287 found_transid = btrfs_header_generation(leaf);
1289 for (i = slot; i < nritems; i++) {
1290 item_off = btrfs_item_ptr_offset(leaf, i);
1291 item_len = btrfs_item_size_nr(leaf, i);
1293 if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1296 if (sizeof(sh) + item_len + *sk_offset >
1297 BTRFS_SEARCH_ARGS_BUFSIZE) {
1302 btrfs_item_key_to_cpu(leaf, key, i);
1303 if (!key_in_sk(key, sk))
1306 sh.objectid = key->objectid;
1307 sh.offset = key->offset;
1308 sh.type = key->type;
1310 sh.transid = found_transid;
1312 /* copy search result header */
1313 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1314 *sk_offset += sizeof(sh);
1317 char *p = buf + *sk_offset;
1319 read_extent_buffer(leaf, p,
1320 item_off, item_len);
1321 *sk_offset += item_len;
1325 if (*num_found >= sk->nr_items)
1330 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1332 else if (key->type < (u8)-1 && key->type < sk->max_type) {
1335 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1342 *num_found += found;
1346 static noinline int search_ioctl(struct inode *inode,
1347 struct btrfs_ioctl_search_args *args)
1349 struct btrfs_root *root;
1350 struct btrfs_key key;
1351 struct btrfs_key max_key;
1352 struct btrfs_path *path;
1353 struct btrfs_ioctl_search_key *sk = &args->key;
1354 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1357 unsigned long sk_offset = 0;
1359 path = btrfs_alloc_path();
1363 if (sk->tree_id == 0) {
1364 /* search the root of the inode that was passed */
1365 root = BTRFS_I(inode)->root;
1367 key.objectid = sk->tree_id;
1368 key.type = BTRFS_ROOT_ITEM_KEY;
1369 key.offset = (u64)-1;
1370 root = btrfs_read_fs_root_no_name(info, &key);
1372 printk(KERN_ERR "could not find root %llu\n",
1374 btrfs_free_path(path);
1379 key.objectid = sk->min_objectid;
1380 key.type = sk->min_type;
1381 key.offset = sk->min_offset;
1383 max_key.objectid = sk->max_objectid;
1384 max_key.type = sk->max_type;
1385 max_key.offset = sk->max_offset;
1387 path->keep_locks = 1;
1390 ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1397 ret = copy_to_sk(root, path, &key, sk, args->buf,
1398 &sk_offset, &num_found);
1399 btrfs_release_path(root, path);
1400 if (ret || num_found >= sk->nr_items)
1406 sk->nr_items = num_found;
1407 btrfs_free_path(path);
1411 static noinline int btrfs_ioctl_tree_search(struct file *file,
1414 struct btrfs_ioctl_search_args *args;
1415 struct inode *inode;
1418 if (!capable(CAP_SYS_ADMIN))
1421 args = memdup_user(argp, sizeof(*args));
1423 return PTR_ERR(args);
1425 inode = fdentry(file)->d_inode;
1426 ret = search_ioctl(inode, args);
1427 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1434 * Search INODE_REFs to identify path name of 'dirid' directory
1435 * in a 'tree_id' tree. and sets path name to 'name'.
1437 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1438 u64 tree_id, u64 dirid, char *name)
1440 struct btrfs_root *root;
1441 struct btrfs_key key;
1447 struct btrfs_inode_ref *iref;
1448 struct extent_buffer *l;
1449 struct btrfs_path *path;
1451 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1456 path = btrfs_alloc_path();
1460 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1462 key.objectid = tree_id;
1463 key.type = BTRFS_ROOT_ITEM_KEY;
1464 key.offset = (u64)-1;
1465 root = btrfs_read_fs_root_no_name(info, &key);
1467 printk(KERN_ERR "could not find root %llu\n", tree_id);
1472 key.objectid = dirid;
1473 key.type = BTRFS_INODE_REF_KEY;
1474 key.offset = (u64)-1;
1477 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1482 slot = path->slots[0];
1483 if (ret > 0 && slot > 0)
1485 btrfs_item_key_to_cpu(l, &key, slot);
1487 if (ret > 0 && (key.objectid != dirid ||
1488 key.type != BTRFS_INODE_REF_KEY)) {
1493 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1494 len = btrfs_inode_ref_name_len(l, iref);
1496 total_len += len + 1;
1501 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1503 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1506 btrfs_release_path(root, path);
1507 key.objectid = key.offset;
1508 key.offset = (u64)-1;
1509 dirid = key.objectid;
1514 memcpy(name, ptr, total_len);
1515 name[total_len]='\0';
1518 btrfs_free_path(path);
1522 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1525 struct btrfs_ioctl_ino_lookup_args *args;
1526 struct inode *inode;
1529 if (!capable(CAP_SYS_ADMIN))
1532 args = memdup_user(argp, sizeof(*args));
1534 return PTR_ERR(args);
1536 inode = fdentry(file)->d_inode;
1538 if (args->treeid == 0)
1539 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1541 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1542 args->treeid, args->objectid,
1545 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1552 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1555 struct dentry *parent = fdentry(file);
1556 struct dentry *dentry;
1557 struct inode *dir = parent->d_inode;
1558 struct inode *inode;
1559 struct btrfs_root *root = BTRFS_I(dir)->root;
1560 struct btrfs_root *dest = NULL;
1561 struct btrfs_ioctl_vol_args *vol_args;
1562 struct btrfs_trans_handle *trans;
1567 vol_args = memdup_user(arg, sizeof(*vol_args));
1568 if (IS_ERR(vol_args))
1569 return PTR_ERR(vol_args);
1571 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1572 namelen = strlen(vol_args->name);
1573 if (strchr(vol_args->name, '/') ||
1574 strncmp(vol_args->name, "..", namelen) == 0) {
1579 err = mnt_want_write(file->f_path.mnt);
1583 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
1584 dentry = lookup_one_len(vol_args->name, parent, namelen);
1585 if (IS_ERR(dentry)) {
1586 err = PTR_ERR(dentry);
1587 goto out_unlock_dir;
1590 if (!dentry->d_inode) {
1595 inode = dentry->d_inode;
1596 dest = BTRFS_I(inode)->root;
1597 if (!capable(CAP_SYS_ADMIN)){
1599 * Regular user. Only allow this with a special mount
1600 * option, when the user has write+exec access to the
1601 * subvol root, and when rmdir(2) would have been
1604 * Note that this is _not_ check that the subvol is
1605 * empty or doesn't contain data that we wouldn't
1606 * otherwise be able to delete.
1608 * Users who want to delete empty subvols should try
1612 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1616 * Do not allow deletion if the parent dir is the same
1617 * as the dir to be deleted. That means the ioctl
1618 * must be called on the dentry referencing the root
1619 * of the subvol, not a random directory contained
1626 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
1630 /* check if subvolume may be deleted by a non-root user */
1631 err = btrfs_may_delete(dir, dentry, 1);
1636 if (inode->i_ino != BTRFS_FIRST_FREE_OBJECTID) {
1641 mutex_lock(&inode->i_mutex);
1642 err = d_invalidate(dentry);
1646 down_write(&root->fs_info->subvol_sem);
1648 err = may_destroy_subvol(dest);
1652 trans = btrfs_start_transaction(root, 0);
1653 if (IS_ERR(trans)) {
1654 err = PTR_ERR(trans);
1657 trans->block_rsv = &root->fs_info->global_block_rsv;
1659 ret = btrfs_unlink_subvol(trans, root, dir,
1660 dest->root_key.objectid,
1661 dentry->d_name.name,
1662 dentry->d_name.len);
1665 btrfs_record_root_in_trans(trans, dest);
1667 memset(&dest->root_item.drop_progress, 0,
1668 sizeof(dest->root_item.drop_progress));
1669 dest->root_item.drop_level = 0;
1670 btrfs_set_root_refs(&dest->root_item, 0);
1672 if (!xchg(&dest->orphan_item_inserted, 1)) {
1673 ret = btrfs_insert_orphan_item(trans,
1674 root->fs_info->tree_root,
1675 dest->root_key.objectid);
1679 ret = btrfs_end_transaction(trans, root);
1681 inode->i_flags |= S_DEAD;
1683 up_write(&root->fs_info->subvol_sem);
1685 mutex_unlock(&inode->i_mutex);
1687 shrink_dcache_sb(root->fs_info->sb);
1688 btrfs_invalidate_inodes(dest);
1694 mutex_unlock(&dir->i_mutex);
1695 mnt_drop_write(file->f_path.mnt);
1701 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
1703 struct inode *inode = fdentry(file)->d_inode;
1704 struct btrfs_root *root = BTRFS_I(inode)->root;
1705 struct btrfs_ioctl_defrag_range_args *range;
1708 if (btrfs_root_readonly(root))
1711 ret = mnt_want_write(file->f_path.mnt);
1715 switch (inode->i_mode & S_IFMT) {
1717 if (!capable(CAP_SYS_ADMIN)) {
1721 ret = btrfs_defrag_root(root, 0);
1724 ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
1727 if (!(file->f_mode & FMODE_WRITE)) {
1732 range = kzalloc(sizeof(*range), GFP_KERNEL);
1739 if (copy_from_user(range, argp,
1745 /* compression requires us to start the IO */
1746 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1747 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
1748 range->extent_thresh = (u32)-1;
1751 /* the rest are all set to zero by kzalloc */
1752 range->len = (u64)-1;
1754 ret = btrfs_defrag_file(file, range);
1761 mnt_drop_write(file->f_path.mnt);
1765 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
1767 struct btrfs_ioctl_vol_args *vol_args;
1770 if (!capable(CAP_SYS_ADMIN))
1773 vol_args = memdup_user(arg, sizeof(*vol_args));
1774 if (IS_ERR(vol_args))
1775 return PTR_ERR(vol_args);
1777 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1778 ret = btrfs_init_new_device(root, vol_args->name);
1784 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
1786 struct btrfs_ioctl_vol_args *vol_args;
1789 if (!capable(CAP_SYS_ADMIN))
1792 if (root->fs_info->sb->s_flags & MS_RDONLY)
1795 vol_args = memdup_user(arg, sizeof(*vol_args));
1796 if (IS_ERR(vol_args))
1797 return PTR_ERR(vol_args);
1799 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1800 ret = btrfs_rm_device(root, vol_args->name);
1806 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
1808 struct btrfs_ioctl_fs_info_args fi_args;
1809 struct btrfs_device *device;
1810 struct btrfs_device *next;
1811 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
1813 if (!capable(CAP_SYS_ADMIN))
1816 fi_args.num_devices = fs_devices->num_devices;
1818 memcpy(&fi_args.fsid, root->fs_info->fsid, sizeof(fi_args.fsid));
1820 mutex_lock(&fs_devices->device_list_mutex);
1821 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
1822 if (device->devid > fi_args.max_id)
1823 fi_args.max_id = device->devid;
1825 mutex_unlock(&fs_devices->device_list_mutex);
1827 if (copy_to_user(arg, &fi_args, sizeof(fi_args)))
1833 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
1835 struct btrfs_ioctl_dev_info_args *di_args;
1836 struct btrfs_device *dev;
1837 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
1839 char *s_uuid = NULL;
1840 char empty_uuid[BTRFS_UUID_SIZE] = {0};
1842 if (!capable(CAP_SYS_ADMIN))
1845 di_args = memdup_user(arg, sizeof(*di_args));
1846 if (IS_ERR(di_args))
1847 return PTR_ERR(di_args);
1849 if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
1850 s_uuid = di_args->uuid;
1852 mutex_lock(&fs_devices->device_list_mutex);
1853 dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL);
1854 mutex_unlock(&fs_devices->device_list_mutex);
1861 di_args->devid = dev->devid;
1862 di_args->bytes_used = dev->bytes_used;
1863 di_args->total_bytes = dev->total_bytes;
1864 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
1865 strncpy(di_args->path, dev->name, sizeof(di_args->path));
1868 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
1875 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
1876 u64 off, u64 olen, u64 destoff)
1878 struct inode *inode = fdentry(file)->d_inode;
1879 struct btrfs_root *root = BTRFS_I(inode)->root;
1880 struct file *src_file;
1882 struct btrfs_trans_handle *trans;
1883 struct btrfs_path *path;
1884 struct extent_buffer *leaf;
1886 struct btrfs_key key;
1891 u64 bs = root->fs_info->sb->s_blocksize;
1896 * - split compressed inline extents. annoying: we need to
1897 * decompress into destination's address_space (the file offset
1898 * may change, so source mapping won't do), then recompress (or
1899 * otherwise reinsert) a subrange.
1900 * - allow ranges within the same file to be cloned (provided
1901 * they don't overlap)?
1904 /* the destination must be opened for writing */
1905 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
1908 if (btrfs_root_readonly(root))
1911 ret = mnt_want_write(file->f_path.mnt);
1915 src_file = fget(srcfd);
1918 goto out_drop_write;
1921 src = src_file->f_dentry->d_inode;
1927 /* the src must be open for reading */
1928 if (!(src_file->f_mode & FMODE_READ))
1932 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
1936 if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
1940 buf = vmalloc(btrfs_level_size(root, 0));
1944 path = btrfs_alloc_path();
1952 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
1953 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
1955 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
1956 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
1959 /* determine range to clone */
1961 if (off + len > src->i_size || off + len < off)
1964 olen = len = src->i_size - off;
1965 /* if we extend to eof, continue to block boundary */
1966 if (off + len == src->i_size)
1967 len = ALIGN(src->i_size, bs) - off;
1969 /* verify the end result is block aligned */
1970 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
1971 !IS_ALIGNED(destoff, bs))
1974 /* do any pending delalloc/csum calc on src, one way or
1975 another, and lock file content */
1977 struct btrfs_ordered_extent *ordered;
1978 lock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
1979 ordered = btrfs_lookup_first_ordered_extent(src, off+len);
1981 !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len,
1982 EXTENT_DELALLOC, 0, NULL))
1984 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
1986 btrfs_put_ordered_extent(ordered);
1987 btrfs_wait_ordered_range(src, off, len);
1991 key.objectid = src->i_ino;
1992 key.type = BTRFS_EXTENT_DATA_KEY;
1997 * note the key will change type as we walk through the
2000 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2004 nritems = btrfs_header_nritems(path->nodes[0]);
2005 if (path->slots[0] >= nritems) {
2006 ret = btrfs_next_leaf(root, path);
2011 nritems = btrfs_header_nritems(path->nodes[0]);
2013 leaf = path->nodes[0];
2014 slot = path->slots[0];
2016 btrfs_item_key_to_cpu(leaf, &key, slot);
2017 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
2018 key.objectid != src->i_ino)
2021 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2022 struct btrfs_file_extent_item *extent;
2025 struct btrfs_key new_key;
2026 u64 disko = 0, diskl = 0;
2027 u64 datao = 0, datal = 0;
2031 size = btrfs_item_size_nr(leaf, slot);
2032 read_extent_buffer(leaf, buf,
2033 btrfs_item_ptr_offset(leaf, slot),
2036 extent = btrfs_item_ptr(leaf, slot,
2037 struct btrfs_file_extent_item);
2038 comp = btrfs_file_extent_compression(leaf, extent);
2039 type = btrfs_file_extent_type(leaf, extent);
2040 if (type == BTRFS_FILE_EXTENT_REG ||
2041 type == BTRFS_FILE_EXTENT_PREALLOC) {
2042 disko = btrfs_file_extent_disk_bytenr(leaf,
2044 diskl = btrfs_file_extent_disk_num_bytes(leaf,
2046 datao = btrfs_file_extent_offset(leaf, extent);
2047 datal = btrfs_file_extent_num_bytes(leaf,
2049 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2050 /* take upper bound, may be compressed */
2051 datal = btrfs_file_extent_ram_bytes(leaf,
2054 btrfs_release_path(root, path);
2056 if (key.offset + datal <= off ||
2057 key.offset >= off+len)
2060 memcpy(&new_key, &key, sizeof(new_key));
2061 new_key.objectid = inode->i_ino;
2062 if (off <= key.offset)
2063 new_key.offset = key.offset + destoff - off;
2065 new_key.offset = destoff;
2067 trans = btrfs_start_transaction(root, 1);
2068 if (IS_ERR(trans)) {
2069 ret = PTR_ERR(trans);
2073 if (type == BTRFS_FILE_EXTENT_REG ||
2074 type == BTRFS_FILE_EXTENT_PREALLOC) {
2075 if (off > key.offset) {
2076 datao += off - key.offset;
2077 datal -= off - key.offset;
2080 if (key.offset + datal > off + len)
2081 datal = off + len - key.offset;
2083 ret = btrfs_drop_extents(trans, inode,
2085 new_key.offset + datal,
2089 ret = btrfs_insert_empty_item(trans, root, path,
2093 leaf = path->nodes[0];
2094 slot = path->slots[0];
2095 write_extent_buffer(leaf, buf,
2096 btrfs_item_ptr_offset(leaf, slot),
2099 extent = btrfs_item_ptr(leaf, slot,
2100 struct btrfs_file_extent_item);
2102 /* disko == 0 means it's a hole */
2106 btrfs_set_file_extent_offset(leaf, extent,
2108 btrfs_set_file_extent_num_bytes(leaf, extent,
2111 inode_add_bytes(inode, datal);
2112 ret = btrfs_inc_extent_ref(trans, root,
2114 root->root_key.objectid,
2116 new_key.offset - datao);
2119 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2122 if (off > key.offset) {
2123 skip = off - key.offset;
2124 new_key.offset += skip;
2127 if (key.offset + datal > off+len)
2128 trim = key.offset + datal - (off+len);
2130 if (comp && (skip || trim)) {
2132 btrfs_end_transaction(trans, root);
2135 size -= skip + trim;
2136 datal -= skip + trim;
2138 ret = btrfs_drop_extents(trans, inode,
2140 new_key.offset + datal,
2144 ret = btrfs_insert_empty_item(trans, root, path,
2150 btrfs_file_extent_calc_inline_size(0);
2151 memmove(buf+start, buf+start+skip,
2155 leaf = path->nodes[0];
2156 slot = path->slots[0];
2157 write_extent_buffer(leaf, buf,
2158 btrfs_item_ptr_offset(leaf, slot),
2160 inode_add_bytes(inode, datal);
2163 btrfs_mark_buffer_dirty(leaf);
2164 btrfs_release_path(root, path);
2166 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2169 * we round up to the block size at eof when
2170 * determining which extents to clone above,
2171 * but shouldn't round up the file size
2173 endoff = new_key.offset + datal;
2174 if (endoff > destoff+olen)
2175 endoff = destoff+olen;
2176 if (endoff > inode->i_size)
2177 btrfs_i_size_write(inode, endoff);
2179 BTRFS_I(inode)->flags = BTRFS_I(src)->flags;
2180 ret = btrfs_update_inode(trans, root, inode);
2182 btrfs_end_transaction(trans, root);
2185 btrfs_release_path(root, path);
2190 btrfs_release_path(root, path);
2191 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
2193 mutex_unlock(&src->i_mutex);
2194 mutex_unlock(&inode->i_mutex);
2196 btrfs_free_path(path);
2200 mnt_drop_write(file->f_path.mnt);
2204 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
2206 struct btrfs_ioctl_clone_range_args args;
2208 if (copy_from_user(&args, argp, sizeof(args)))
2210 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2211 args.src_length, args.dest_offset);
2215 * there are many ways the trans_start and trans_end ioctls can lead
2216 * to deadlocks. They should only be used by applications that
2217 * basically own the machine, and have a very in depth understanding
2218 * of all the possible deadlocks and enospc problems.
2220 static long btrfs_ioctl_trans_start(struct file *file)
2222 struct inode *inode = fdentry(file)->d_inode;
2223 struct btrfs_root *root = BTRFS_I(inode)->root;
2224 struct btrfs_trans_handle *trans;
2228 if (!capable(CAP_SYS_ADMIN))
2232 if (file->private_data)
2236 if (btrfs_root_readonly(root))
2239 ret = mnt_want_write(file->f_path.mnt);
2243 mutex_lock(&root->fs_info->trans_mutex);
2244 root->fs_info->open_ioctl_trans++;
2245 mutex_unlock(&root->fs_info->trans_mutex);
2248 trans = btrfs_start_ioctl_transaction(root, 0);
2252 file->private_data = trans;
2256 mutex_lock(&root->fs_info->trans_mutex);
2257 root->fs_info->open_ioctl_trans--;
2258 mutex_unlock(&root->fs_info->trans_mutex);
2259 mnt_drop_write(file->f_path.mnt);
2264 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2266 struct inode *inode = fdentry(file)->d_inode;
2267 struct btrfs_root *root = BTRFS_I(inode)->root;
2268 struct btrfs_root *new_root;
2269 struct btrfs_dir_item *di;
2270 struct btrfs_trans_handle *trans;
2271 struct btrfs_path *path;
2272 struct btrfs_key location;
2273 struct btrfs_disk_key disk_key;
2274 struct btrfs_super_block *disk_super;
2279 if (!capable(CAP_SYS_ADMIN))
2282 if (copy_from_user(&objectid, argp, sizeof(objectid)))
2286 objectid = root->root_key.objectid;
2288 location.objectid = objectid;
2289 location.type = BTRFS_ROOT_ITEM_KEY;
2290 location.offset = (u64)-1;
2292 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2293 if (IS_ERR(new_root))
2294 return PTR_ERR(new_root);
2296 if (btrfs_root_refs(&new_root->root_item) == 0)
2299 path = btrfs_alloc_path();
2302 path->leave_spinning = 1;
2304 trans = btrfs_start_transaction(root, 1);
2305 if (IS_ERR(trans)) {
2306 btrfs_free_path(path);
2307 return PTR_ERR(trans);
2310 dir_id = btrfs_super_root_dir(&root->fs_info->super_copy);
2311 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2312 dir_id, "default", 7, 1);
2313 if (IS_ERR_OR_NULL(di)) {
2314 btrfs_free_path(path);
2315 btrfs_end_transaction(trans, root);
2316 printk(KERN_ERR "Umm, you don't have the default dir item, "
2317 "this isn't going to work\n");
2321 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2322 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2323 btrfs_mark_buffer_dirty(path->nodes[0]);
2324 btrfs_free_path(path);
2326 disk_super = &root->fs_info->super_copy;
2327 features = btrfs_super_incompat_flags(disk_super);
2328 if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) {
2329 features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL;
2330 btrfs_set_super_incompat_flags(disk_super, features);
2332 btrfs_end_transaction(trans, root);
2337 static void get_block_group_info(struct list_head *groups_list,
2338 struct btrfs_ioctl_space_info *space)
2340 struct btrfs_block_group_cache *block_group;
2342 space->total_bytes = 0;
2343 space->used_bytes = 0;
2345 list_for_each_entry(block_group, groups_list, list) {
2346 space->flags = block_group->flags;
2347 space->total_bytes += block_group->key.offset;
2348 space->used_bytes +=
2349 btrfs_block_group_used(&block_group->item);
2353 long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2355 struct btrfs_ioctl_space_args space_args;
2356 struct btrfs_ioctl_space_info space;
2357 struct btrfs_ioctl_space_info *dest;
2358 struct btrfs_ioctl_space_info *dest_orig;
2359 struct btrfs_ioctl_space_info __user *user_dest;
2360 struct btrfs_space_info *info;
2361 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2362 BTRFS_BLOCK_GROUP_SYSTEM,
2363 BTRFS_BLOCK_GROUP_METADATA,
2364 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2371 if (copy_from_user(&space_args,
2372 (struct btrfs_ioctl_space_args __user *)arg,
2373 sizeof(space_args)))
2376 for (i = 0; i < num_types; i++) {
2377 struct btrfs_space_info *tmp;
2381 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2383 if (tmp->flags == types[i]) {
2393 down_read(&info->groups_sem);
2394 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2395 if (!list_empty(&info->block_groups[c]))
2398 up_read(&info->groups_sem);
2401 /* space_slots == 0 means they are asking for a count */
2402 if (space_args.space_slots == 0) {
2403 space_args.total_spaces = slot_count;
2407 slot_count = min_t(u64, space_args.space_slots, slot_count);
2409 alloc_size = sizeof(*dest) * slot_count;
2411 /* we generally have at most 6 or so space infos, one for each raid
2412 * level. So, a whole page should be more than enough for everyone
2414 if (alloc_size > PAGE_CACHE_SIZE)
2417 space_args.total_spaces = 0;
2418 dest = kmalloc(alloc_size, GFP_NOFS);
2423 /* now we have a buffer to copy into */
2424 for (i = 0; i < num_types; i++) {
2425 struct btrfs_space_info *tmp;
2432 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2434 if (tmp->flags == types[i]) {
2443 down_read(&info->groups_sem);
2444 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2445 if (!list_empty(&info->block_groups[c])) {
2446 get_block_group_info(&info->block_groups[c],
2448 memcpy(dest, &space, sizeof(space));
2450 space_args.total_spaces++;
2456 up_read(&info->groups_sem);
2459 user_dest = (struct btrfs_ioctl_space_info *)
2460 (arg + sizeof(struct btrfs_ioctl_space_args));
2462 if (copy_to_user(user_dest, dest_orig, alloc_size))
2467 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
2474 * there are many ways the trans_start and trans_end ioctls can lead
2475 * to deadlocks. They should only be used by applications that
2476 * basically own the machine, and have a very in depth understanding
2477 * of all the possible deadlocks and enospc problems.
2479 long btrfs_ioctl_trans_end(struct file *file)
2481 struct inode *inode = fdentry(file)->d_inode;
2482 struct btrfs_root *root = BTRFS_I(inode)->root;
2483 struct btrfs_trans_handle *trans;
2485 trans = file->private_data;
2488 file->private_data = NULL;
2490 btrfs_end_transaction(trans, root);
2492 mutex_lock(&root->fs_info->trans_mutex);
2493 root->fs_info->open_ioctl_trans--;
2494 mutex_unlock(&root->fs_info->trans_mutex);
2496 mnt_drop_write(file->f_path.mnt);
2500 static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
2502 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2503 struct btrfs_trans_handle *trans;
2507 trans = btrfs_start_transaction(root, 0);
2509 return PTR_ERR(trans);
2510 transid = trans->transid;
2511 ret = btrfs_commit_transaction_async(trans, root, 0);
2513 btrfs_end_transaction(trans, root);
2518 if (copy_to_user(argp, &transid, sizeof(transid)))
2523 static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
2525 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2529 if (copy_from_user(&transid, argp, sizeof(transid)))
2532 transid = 0; /* current trans */
2534 return btrfs_wait_for_commit(root, transid);
2537 static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg)
2540 struct btrfs_ioctl_scrub_args *sa;
2542 if (!capable(CAP_SYS_ADMIN))
2545 sa = memdup_user(arg, sizeof(*sa));
2549 ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end,
2552 if (copy_to_user(arg, sa, sizeof(*sa)))
2559 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
2561 if (!capable(CAP_SYS_ADMIN))
2564 return btrfs_scrub_cancel(root);
2567 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
2570 struct btrfs_ioctl_scrub_args *sa;
2573 if (!capable(CAP_SYS_ADMIN))
2576 sa = memdup_user(arg, sizeof(*sa));
2580 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
2582 if (copy_to_user(arg, sa, sizeof(*sa)))
2589 long btrfs_ioctl(struct file *file, unsigned int
2590 cmd, unsigned long arg)
2592 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
2593 void __user *argp = (void __user *)arg;
2596 case FS_IOC_GETFLAGS:
2597 return btrfs_ioctl_getflags(file, argp);
2598 case FS_IOC_SETFLAGS:
2599 return btrfs_ioctl_setflags(file, argp);
2600 case FS_IOC_GETVERSION:
2601 return btrfs_ioctl_getversion(file, argp);
2603 return btrfs_ioctl_fitrim(file, argp);
2604 case BTRFS_IOC_SNAP_CREATE:
2605 return btrfs_ioctl_snap_create(file, argp, 0);
2606 case BTRFS_IOC_SNAP_CREATE_V2:
2607 return btrfs_ioctl_snap_create_v2(file, argp, 0);
2608 case BTRFS_IOC_SUBVOL_CREATE:
2609 return btrfs_ioctl_snap_create(file, argp, 1);
2610 case BTRFS_IOC_SNAP_DESTROY:
2611 return btrfs_ioctl_snap_destroy(file, argp);
2612 case BTRFS_IOC_SUBVOL_GETFLAGS:
2613 return btrfs_ioctl_subvol_getflags(file, argp);
2614 case BTRFS_IOC_SUBVOL_SETFLAGS:
2615 return btrfs_ioctl_subvol_setflags(file, argp);
2616 case BTRFS_IOC_DEFAULT_SUBVOL:
2617 return btrfs_ioctl_default_subvol(file, argp);
2618 case BTRFS_IOC_DEFRAG:
2619 return btrfs_ioctl_defrag(file, NULL);
2620 case BTRFS_IOC_DEFRAG_RANGE:
2621 return btrfs_ioctl_defrag(file, argp);
2622 case BTRFS_IOC_RESIZE:
2623 return btrfs_ioctl_resize(root, argp);
2624 case BTRFS_IOC_ADD_DEV:
2625 return btrfs_ioctl_add_dev(root, argp);
2626 case BTRFS_IOC_RM_DEV:
2627 return btrfs_ioctl_rm_dev(root, argp);
2628 case BTRFS_IOC_FS_INFO:
2629 return btrfs_ioctl_fs_info(root, argp);
2630 case BTRFS_IOC_DEV_INFO:
2631 return btrfs_ioctl_dev_info(root, argp);
2632 case BTRFS_IOC_BALANCE:
2633 return btrfs_balance(root->fs_info->dev_root);
2634 case BTRFS_IOC_CLONE:
2635 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
2636 case BTRFS_IOC_CLONE_RANGE:
2637 return btrfs_ioctl_clone_range(file, argp);
2638 case BTRFS_IOC_TRANS_START:
2639 return btrfs_ioctl_trans_start(file);
2640 case BTRFS_IOC_TRANS_END:
2641 return btrfs_ioctl_trans_end(file);
2642 case BTRFS_IOC_TREE_SEARCH:
2643 return btrfs_ioctl_tree_search(file, argp);
2644 case BTRFS_IOC_INO_LOOKUP:
2645 return btrfs_ioctl_ino_lookup(file, argp);
2646 case BTRFS_IOC_SPACE_INFO:
2647 return btrfs_ioctl_space_info(root, argp);
2648 case BTRFS_IOC_SYNC:
2649 btrfs_sync_fs(file->f_dentry->d_sb, 1);
2651 case BTRFS_IOC_START_SYNC:
2652 return btrfs_ioctl_start_sync(file, argp);
2653 case BTRFS_IOC_WAIT_SYNC:
2654 return btrfs_ioctl_wait_sync(file, argp);
2655 case BTRFS_IOC_SCRUB:
2656 return btrfs_ioctl_scrub(root, argp);
2657 case BTRFS_IOC_SCRUB_CANCEL:
2658 return btrfs_ioctl_scrub_cancel(root, argp);
2659 case BTRFS_IOC_SCRUB_PROGRESS:
2660 return btrfs_ioctl_scrub_progress(root, argp);