Btrfs: use memdup_user helpers
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / btrfs / ioctl.c
blob8079ebfeaf503e50add674c895f1d110089ae0c5
1 /*
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>
23 #include <linux/fs.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 "compat.h"
44 #include "ctree.h"
45 #include "disk-io.h"
46 #include "transaction.h"
47 #include "btrfs_inode.h"
48 #include "ioctl.h"
49 #include "print-tree.h"
50 #include "volumes.h"
51 #include "locking.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)
56 if (S_ISDIR(mode))
57 return flags;
58 else if (S_ISREG(mode))
59 return flags & ~FS_DIRSYNC_FL;
60 else
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)
72 iflags |= FS_SYNC_FL;
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;
84 return iflags;
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)
115 unsigned int flags;
117 if (!dir)
118 return;
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)))
137 return -EFAULT;
138 return 0;
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;
148 int ret;
150 if (copy_from_user(&flags, arg, sizeof(flags)))
151 return -EFAULT;
153 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
154 FS_NOATIME_FL | FS_NODUMP_FL | \
155 FS_SYNC_FL | FS_DIRSYNC_FL))
156 return -EOPNOTSUPP;
158 if (!is_owner_or_cap(inode))
159 return -EACCES;
161 mutex_lock(&inode->i_mutex);
163 flags = btrfs_mask_flags(inode->i_mode, flags);
164 oldflags = btrfs_flags_to_ioctl(ip->flags);
165 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
166 if (!capable(CAP_LINUX_IMMUTABLE)) {
167 ret = -EPERM;
168 goto out_unlock;
172 ret = mnt_want_write(file->f_path.mnt);
173 if (ret)
174 goto out_unlock;
176 if (flags & FS_SYNC_FL)
177 ip->flags |= BTRFS_INODE_SYNC;
178 else
179 ip->flags &= ~BTRFS_INODE_SYNC;
180 if (flags & FS_IMMUTABLE_FL)
181 ip->flags |= BTRFS_INODE_IMMUTABLE;
182 else
183 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
184 if (flags & FS_APPEND_FL)
185 ip->flags |= BTRFS_INODE_APPEND;
186 else
187 ip->flags &= ~BTRFS_INODE_APPEND;
188 if (flags & FS_NODUMP_FL)
189 ip->flags |= BTRFS_INODE_NODUMP;
190 else
191 ip->flags &= ~BTRFS_INODE_NODUMP;
192 if (flags & FS_NOATIME_FL)
193 ip->flags |= BTRFS_INODE_NOATIME;
194 else
195 ip->flags &= ~BTRFS_INODE_NOATIME;
196 if (flags & FS_DIRSYNC_FL)
197 ip->flags |= BTRFS_INODE_DIRSYNC;
198 else
199 ip->flags &= ~BTRFS_INODE_DIRSYNC;
202 trans = btrfs_join_transaction(root, 1);
203 BUG_ON(!trans);
205 ret = btrfs_update_inode(trans, root, inode);
206 BUG_ON(ret);
208 btrfs_update_iflags(inode);
209 inode->i_ctime = CURRENT_TIME;
210 btrfs_end_transaction(trans, root);
212 mnt_drop_write(file->f_path.mnt);
213 out_unlock:
214 mutex_unlock(&inode->i_mutex);
215 return 0;
218 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
220 struct inode *inode = file->f_path.dentry->d_inode;
222 return put_user(inode->i_generation, arg);
225 static noinline int create_subvol(struct btrfs_root *root,
226 struct dentry *dentry,
227 char *name, int namelen)
229 struct btrfs_trans_handle *trans;
230 struct btrfs_key key;
231 struct btrfs_root_item root_item;
232 struct btrfs_inode_item *inode_item;
233 struct extent_buffer *leaf;
234 struct btrfs_root *new_root;
235 struct inode *dir = dentry->d_parent->d_inode;
236 int ret;
237 int err;
238 u64 objectid;
239 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
240 u64 index = 0;
242 ret = btrfs_find_free_objectid(NULL, root->fs_info->tree_root,
243 0, &objectid);
244 if (ret)
245 return ret;
247 * 1 - inode item
248 * 2 - refs
249 * 1 - root item
250 * 2 - dir items
252 trans = btrfs_start_transaction(root, 6);
253 if (IS_ERR(trans))
254 return PTR_ERR(trans);
256 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
257 0, objectid, NULL, 0, 0, 0);
258 if (IS_ERR(leaf)) {
259 ret = PTR_ERR(leaf);
260 goto fail;
263 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
264 btrfs_set_header_bytenr(leaf, leaf->start);
265 btrfs_set_header_generation(leaf, trans->transid);
266 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
267 btrfs_set_header_owner(leaf, objectid);
269 write_extent_buffer(leaf, root->fs_info->fsid,
270 (unsigned long)btrfs_header_fsid(leaf),
271 BTRFS_FSID_SIZE);
272 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
273 (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
274 BTRFS_UUID_SIZE);
275 btrfs_mark_buffer_dirty(leaf);
277 inode_item = &root_item.inode;
278 memset(inode_item, 0, sizeof(*inode_item));
279 inode_item->generation = cpu_to_le64(1);
280 inode_item->size = cpu_to_le64(3);
281 inode_item->nlink = cpu_to_le32(1);
282 inode_item->nbytes = cpu_to_le64(root->leafsize);
283 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
285 btrfs_set_root_bytenr(&root_item, leaf->start);
286 btrfs_set_root_generation(&root_item, trans->transid);
287 btrfs_set_root_level(&root_item, 0);
288 btrfs_set_root_refs(&root_item, 1);
289 btrfs_set_root_used(&root_item, leaf->len);
290 btrfs_set_root_last_snapshot(&root_item, 0);
292 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
293 root_item.drop_level = 0;
295 btrfs_tree_unlock(leaf);
296 free_extent_buffer(leaf);
297 leaf = NULL;
299 btrfs_set_root_dirid(&root_item, new_dirid);
301 key.objectid = objectid;
302 key.offset = 0;
303 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
304 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
305 &root_item);
306 if (ret)
307 goto fail;
309 key.offset = (u64)-1;
310 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
311 BUG_ON(IS_ERR(new_root));
313 btrfs_record_root_in_trans(trans, new_root);
315 ret = btrfs_create_subvol_root(trans, new_root, new_dirid,
316 BTRFS_I(dir)->block_group);
318 * insert the directory item
320 ret = btrfs_set_inode_index(dir, &index);
321 BUG_ON(ret);
323 ret = btrfs_insert_dir_item(trans, root,
324 name, namelen, dir->i_ino, &key,
325 BTRFS_FT_DIR, index);
326 if (ret)
327 goto fail;
329 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
330 ret = btrfs_update_inode(trans, root, dir);
331 BUG_ON(ret);
333 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
334 objectid, root->root_key.objectid,
335 dir->i_ino, index, name, namelen);
337 BUG_ON(ret);
339 d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
340 fail:
341 err = btrfs_commit_transaction(trans, root);
342 if (err && !ret)
343 ret = err;
344 return ret;
347 static int create_snapshot(struct btrfs_root *root, struct dentry *dentry)
349 struct inode *inode;
350 struct btrfs_pending_snapshot *pending_snapshot;
351 struct btrfs_trans_handle *trans;
352 int ret;
354 if (!root->ref_cows)
355 return -EINVAL;
357 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
358 if (!pending_snapshot)
359 return -ENOMEM;
361 btrfs_init_block_rsv(&pending_snapshot->block_rsv);
362 pending_snapshot->dentry = dentry;
363 pending_snapshot->root = root;
365 trans = btrfs_start_transaction(root->fs_info->extent_root, 5);
366 if (IS_ERR(trans)) {
367 ret = PTR_ERR(trans);
368 goto fail;
371 ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
372 BUG_ON(ret);
374 list_add(&pending_snapshot->list,
375 &trans->transaction->pending_snapshots);
376 ret = btrfs_commit_transaction(trans, root->fs_info->extent_root);
377 BUG_ON(ret);
379 ret = pending_snapshot->error;
380 if (ret)
381 goto fail;
383 btrfs_orphan_cleanup(pending_snapshot->snap);
385 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
386 if (IS_ERR(inode)) {
387 ret = PTR_ERR(inode);
388 goto fail;
390 BUG_ON(!inode);
391 d_instantiate(dentry, inode);
392 ret = 0;
393 fail:
394 kfree(pending_snapshot);
395 return ret;
398 /* copy of may_create in fs/namei.c() */
399 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
401 if (child->d_inode)
402 return -EEXIST;
403 if (IS_DEADDIR(dir))
404 return -ENOENT;
405 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
409 * Create a new subvolume below @parent. This is largely modeled after
410 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
411 * inside this filesystem so it's quite a bit simpler.
413 static noinline int btrfs_mksubvol(struct path *parent,
414 char *name, int namelen,
415 struct btrfs_root *snap_src)
417 struct inode *dir = parent->dentry->d_inode;
418 struct dentry *dentry;
419 int error;
421 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
423 dentry = lookup_one_len(name, parent->dentry, namelen);
424 error = PTR_ERR(dentry);
425 if (IS_ERR(dentry))
426 goto out_unlock;
428 error = -EEXIST;
429 if (dentry->d_inode)
430 goto out_dput;
432 error = mnt_want_write(parent->mnt);
433 if (error)
434 goto out_dput;
436 error = btrfs_may_create(dir, dentry);
437 if (error)
438 goto out_drop_write;
440 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
442 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
443 goto out_up_read;
445 if (snap_src) {
446 error = create_snapshot(snap_src, dentry);
447 } else {
448 error = create_subvol(BTRFS_I(dir)->root, dentry,
449 name, namelen);
451 if (!error)
452 fsnotify_mkdir(dir, dentry);
453 out_up_read:
454 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
455 out_drop_write:
456 mnt_drop_write(parent->mnt);
457 out_dput:
458 dput(dentry);
459 out_unlock:
460 mutex_unlock(&dir->i_mutex);
461 return error;
464 static int should_defrag_range(struct inode *inode, u64 start, u64 len,
465 int thresh, u64 *last_len, u64 *skip,
466 u64 *defrag_end)
468 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
469 struct extent_map *em = NULL;
470 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
471 int ret = 1;
474 if (thresh == 0)
475 thresh = 256 * 1024;
478 * make sure that once we start defragging and extent, we keep on
479 * defragging it
481 if (start < *defrag_end)
482 return 1;
484 *skip = 0;
487 * hopefully we have this extent in the tree already, try without
488 * the full extent lock
490 read_lock(&em_tree->lock);
491 em = lookup_extent_mapping(em_tree, start, len);
492 read_unlock(&em_tree->lock);
494 if (!em) {
495 /* get the big lock and read metadata off disk */
496 lock_extent(io_tree, start, start + len - 1, GFP_NOFS);
497 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
498 unlock_extent(io_tree, start, start + len - 1, GFP_NOFS);
500 if (IS_ERR(em))
501 return 0;
504 /* this will cover holes, and inline extents */
505 if (em->block_start >= EXTENT_MAP_LAST_BYTE)
506 ret = 0;
509 * we hit a real extent, if it is big don't bother defragging it again
511 if ((*last_len == 0 || *last_len >= thresh) && em->len >= thresh)
512 ret = 0;
515 * last_len ends up being a counter of how many bytes we've defragged.
516 * every time we choose not to defrag an extent, we reset *last_len
517 * so that the next tiny extent will force a defrag.
519 * The end result of this is that tiny extents before a single big
520 * extent will force at least part of that big extent to be defragged.
522 if (ret) {
523 *last_len += len;
524 *defrag_end = extent_map_end(em);
525 } else {
526 *last_len = 0;
527 *skip = extent_map_end(em);
528 *defrag_end = 0;
531 free_extent_map(em);
532 return ret;
535 static int btrfs_defrag_file(struct file *file,
536 struct btrfs_ioctl_defrag_range_args *range)
538 struct inode *inode = fdentry(file)->d_inode;
539 struct btrfs_root *root = BTRFS_I(inode)->root;
540 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
541 struct btrfs_ordered_extent *ordered;
542 struct page *page;
543 unsigned long last_index;
544 unsigned long ra_pages = root->fs_info->bdi.ra_pages;
545 unsigned long total_read = 0;
546 u64 page_start;
547 u64 page_end;
548 u64 last_len = 0;
549 u64 skip = 0;
550 u64 defrag_end = 0;
551 unsigned long i;
552 int ret;
554 if (inode->i_size == 0)
555 return 0;
557 if (range->start + range->len > range->start) {
558 last_index = min_t(u64, inode->i_size - 1,
559 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
560 } else {
561 last_index = (inode->i_size - 1) >> PAGE_CACHE_SHIFT;
564 i = range->start >> PAGE_CACHE_SHIFT;
565 while (i <= last_index) {
566 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
567 PAGE_CACHE_SIZE,
568 range->extent_thresh,
569 &last_len, &skip,
570 &defrag_end)) {
571 unsigned long next;
573 * the should_defrag function tells us how much to skip
574 * bump our counter by the suggested amount
576 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
577 i = max(i + 1, next);
578 continue;
581 if (total_read % ra_pages == 0) {
582 btrfs_force_ra(inode->i_mapping, &file->f_ra, file, i,
583 min(last_index, i + ra_pages - 1));
585 total_read++;
586 mutex_lock(&inode->i_mutex);
587 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
588 BTRFS_I(inode)->force_compress = 1;
590 ret = btrfs_delalloc_reserve_space(inode, PAGE_CACHE_SIZE);
591 if (ret)
592 goto err_unlock;
593 again:
594 if (inode->i_size == 0 ||
595 i > ((inode->i_size - 1) >> PAGE_CACHE_SHIFT)) {
596 ret = 0;
597 goto err_reservations;
600 page = grab_cache_page(inode->i_mapping, i);
601 if (!page) {
602 ret = -ENOMEM;
603 goto err_reservations;
606 if (!PageUptodate(page)) {
607 btrfs_readpage(NULL, page);
608 lock_page(page);
609 if (!PageUptodate(page)) {
610 unlock_page(page);
611 page_cache_release(page);
612 ret = -EIO;
613 goto err_reservations;
617 if (page->mapping != inode->i_mapping) {
618 unlock_page(page);
619 page_cache_release(page);
620 goto again;
623 wait_on_page_writeback(page);
625 if (PageDirty(page)) {
626 btrfs_delalloc_release_space(inode, PAGE_CACHE_SIZE);
627 goto loop_unlock;
630 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
631 page_end = page_start + PAGE_CACHE_SIZE - 1;
632 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
634 ordered = btrfs_lookup_ordered_extent(inode, page_start);
635 if (ordered) {
636 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
637 unlock_page(page);
638 page_cache_release(page);
639 btrfs_start_ordered_extent(inode, ordered, 1);
640 btrfs_put_ordered_extent(ordered);
641 goto again;
643 set_page_extent_mapped(page);
646 * this makes sure page_mkwrite is called on the
647 * page if it is dirtied again later
649 clear_page_dirty_for_io(page);
650 clear_extent_bits(&BTRFS_I(inode)->io_tree, page_start,
651 page_end, EXTENT_DIRTY | EXTENT_DELALLOC |
652 EXTENT_DO_ACCOUNTING, GFP_NOFS);
654 btrfs_set_extent_delalloc(inode, page_start, page_end, NULL);
655 ClearPageChecked(page);
656 set_page_dirty(page);
657 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
659 loop_unlock:
660 unlock_page(page);
661 page_cache_release(page);
662 mutex_unlock(&inode->i_mutex);
664 balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
665 i++;
668 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
669 filemap_flush(inode->i_mapping);
671 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
672 /* the filemap_flush will queue IO into the worker threads, but
673 * we have to make sure the IO is actually started and that
674 * ordered extents get created before we return
676 atomic_inc(&root->fs_info->async_submit_draining);
677 while (atomic_read(&root->fs_info->nr_async_submits) ||
678 atomic_read(&root->fs_info->async_delalloc_pages)) {
679 wait_event(root->fs_info->async_submit_wait,
680 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
681 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
683 atomic_dec(&root->fs_info->async_submit_draining);
685 mutex_lock(&inode->i_mutex);
686 BTRFS_I(inode)->force_compress = 0;
687 mutex_unlock(&inode->i_mutex);
690 return 0;
692 err_reservations:
693 btrfs_delalloc_release_space(inode, PAGE_CACHE_SIZE);
694 err_unlock:
695 mutex_unlock(&inode->i_mutex);
696 return ret;
699 static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
700 void __user *arg)
702 u64 new_size;
703 u64 old_size;
704 u64 devid = 1;
705 struct btrfs_ioctl_vol_args *vol_args;
706 struct btrfs_trans_handle *trans;
707 struct btrfs_device *device = NULL;
708 char *sizestr;
709 char *devstr = NULL;
710 int ret = 0;
711 int namelen;
712 int mod = 0;
714 if (root->fs_info->sb->s_flags & MS_RDONLY)
715 return -EROFS;
717 if (!capable(CAP_SYS_ADMIN))
718 return -EPERM;
720 vol_args = memdup_user(arg, sizeof(*vol_args));
721 if (IS_ERR(vol_args))
722 return PTR_ERR(vol_args);
724 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
725 namelen = strlen(vol_args->name);
727 mutex_lock(&root->fs_info->volume_mutex);
728 sizestr = vol_args->name;
729 devstr = strchr(sizestr, ':');
730 if (devstr) {
731 char *end;
732 sizestr = devstr + 1;
733 *devstr = '\0';
734 devstr = vol_args->name;
735 devid = simple_strtoull(devstr, &end, 10);
736 printk(KERN_INFO "resizing devid %llu\n",
737 (unsigned long long)devid);
739 device = btrfs_find_device(root, devid, NULL, NULL);
740 if (!device) {
741 printk(KERN_INFO "resizer unable to find device %llu\n",
742 (unsigned long long)devid);
743 ret = -EINVAL;
744 goto out_unlock;
746 if (!strcmp(sizestr, "max"))
747 new_size = device->bdev->bd_inode->i_size;
748 else {
749 if (sizestr[0] == '-') {
750 mod = -1;
751 sizestr++;
752 } else if (sizestr[0] == '+') {
753 mod = 1;
754 sizestr++;
756 new_size = memparse(sizestr, NULL);
757 if (new_size == 0) {
758 ret = -EINVAL;
759 goto out_unlock;
763 old_size = device->total_bytes;
765 if (mod < 0) {
766 if (new_size > old_size) {
767 ret = -EINVAL;
768 goto out_unlock;
770 new_size = old_size - new_size;
771 } else if (mod > 0) {
772 new_size = old_size + new_size;
775 if (new_size < 256 * 1024 * 1024) {
776 ret = -EINVAL;
777 goto out_unlock;
779 if (new_size > device->bdev->bd_inode->i_size) {
780 ret = -EFBIG;
781 goto out_unlock;
784 do_div(new_size, root->sectorsize);
785 new_size *= root->sectorsize;
787 printk(KERN_INFO "new size for %s is %llu\n",
788 device->name, (unsigned long long)new_size);
790 if (new_size > old_size) {
791 trans = btrfs_start_transaction(root, 0);
792 ret = btrfs_grow_device(trans, device, new_size);
793 btrfs_commit_transaction(trans, root);
794 } else {
795 ret = btrfs_shrink_device(device, new_size);
798 out_unlock:
799 mutex_unlock(&root->fs_info->volume_mutex);
800 kfree(vol_args);
801 return ret;
804 static noinline int btrfs_ioctl_snap_create(struct file *file,
805 void __user *arg, int subvol)
807 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
808 struct btrfs_ioctl_vol_args *vol_args;
809 struct file *src_file;
810 int namelen;
811 int ret = 0;
813 if (root->fs_info->sb->s_flags & MS_RDONLY)
814 return -EROFS;
816 vol_args = memdup_user(arg, sizeof(*vol_args));
817 if (IS_ERR(vol_args))
818 return PTR_ERR(vol_args);
820 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
821 namelen = strlen(vol_args->name);
822 if (strchr(vol_args->name, '/')) {
823 ret = -EINVAL;
824 goto out;
827 if (subvol) {
828 ret = btrfs_mksubvol(&file->f_path, vol_args->name, namelen,
829 NULL);
830 } else {
831 struct inode *src_inode;
832 src_file = fget(vol_args->fd);
833 if (!src_file) {
834 ret = -EINVAL;
835 goto out;
838 src_inode = src_file->f_path.dentry->d_inode;
839 if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
840 printk(KERN_INFO "btrfs: Snapshot src from "
841 "another FS\n");
842 ret = -EINVAL;
843 fput(src_file);
844 goto out;
846 ret = btrfs_mksubvol(&file->f_path, vol_args->name, namelen,
847 BTRFS_I(src_inode)->root);
848 fput(src_file);
850 out:
851 kfree(vol_args);
852 return ret;
856 * helper to check if the subvolume references other subvolumes
858 static noinline int may_destroy_subvol(struct btrfs_root *root)
860 struct btrfs_path *path;
861 struct btrfs_key key;
862 int ret;
864 path = btrfs_alloc_path();
865 if (!path)
866 return -ENOMEM;
868 key.objectid = root->root_key.objectid;
869 key.type = BTRFS_ROOT_REF_KEY;
870 key.offset = (u64)-1;
872 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
873 &key, path, 0, 0);
874 if (ret < 0)
875 goto out;
876 BUG_ON(ret == 0);
878 ret = 0;
879 if (path->slots[0] > 0) {
880 path->slots[0]--;
881 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
882 if (key.objectid == root->root_key.objectid &&
883 key.type == BTRFS_ROOT_REF_KEY)
884 ret = -ENOTEMPTY;
886 out:
887 btrfs_free_path(path);
888 return ret;
891 static noinline int key_in_sk(struct btrfs_key *key,
892 struct btrfs_ioctl_search_key *sk)
894 struct btrfs_key test;
895 int ret;
897 test.objectid = sk->min_objectid;
898 test.type = sk->min_type;
899 test.offset = sk->min_offset;
901 ret = btrfs_comp_cpu_keys(key, &test);
902 if (ret < 0)
903 return 0;
905 test.objectid = sk->max_objectid;
906 test.type = sk->max_type;
907 test.offset = sk->max_offset;
909 ret = btrfs_comp_cpu_keys(key, &test);
910 if (ret > 0)
911 return 0;
912 return 1;
915 static noinline int copy_to_sk(struct btrfs_root *root,
916 struct btrfs_path *path,
917 struct btrfs_key *key,
918 struct btrfs_ioctl_search_key *sk,
919 char *buf,
920 unsigned long *sk_offset,
921 int *num_found)
923 u64 found_transid;
924 struct extent_buffer *leaf;
925 struct btrfs_ioctl_search_header sh;
926 unsigned long item_off;
927 unsigned long item_len;
928 int nritems;
929 int i;
930 int slot;
931 int found = 0;
932 int ret = 0;
934 leaf = path->nodes[0];
935 slot = path->slots[0];
936 nritems = btrfs_header_nritems(leaf);
938 if (btrfs_header_generation(leaf) > sk->max_transid) {
939 i = nritems;
940 goto advance_key;
942 found_transid = btrfs_header_generation(leaf);
944 for (i = slot; i < nritems; i++) {
945 item_off = btrfs_item_ptr_offset(leaf, i);
946 item_len = btrfs_item_size_nr(leaf, i);
948 if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
949 item_len = 0;
951 if (sizeof(sh) + item_len + *sk_offset >
952 BTRFS_SEARCH_ARGS_BUFSIZE) {
953 ret = 1;
954 goto overflow;
957 btrfs_item_key_to_cpu(leaf, key, i);
958 if (!key_in_sk(key, sk))
959 continue;
961 sh.objectid = key->objectid;
962 sh.offset = key->offset;
963 sh.type = key->type;
964 sh.len = item_len;
965 sh.transid = found_transid;
967 /* copy search result header */
968 memcpy(buf + *sk_offset, &sh, sizeof(sh));
969 *sk_offset += sizeof(sh);
971 if (item_len) {
972 char *p = buf + *sk_offset;
973 /* copy the item */
974 read_extent_buffer(leaf, p,
975 item_off, item_len);
976 *sk_offset += item_len;
978 found++;
980 if (*num_found >= sk->nr_items)
981 break;
983 advance_key:
984 ret = 0;
985 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
986 key->offset++;
987 else if (key->type < (u8)-1 && key->type < sk->max_type) {
988 key->offset = 0;
989 key->type++;
990 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
991 key->offset = 0;
992 key->type = 0;
993 key->objectid++;
994 } else
995 ret = 1;
996 overflow:
997 *num_found += found;
998 return ret;
1001 static noinline int search_ioctl(struct inode *inode,
1002 struct btrfs_ioctl_search_args *args)
1004 struct btrfs_root *root;
1005 struct btrfs_key key;
1006 struct btrfs_key max_key;
1007 struct btrfs_path *path;
1008 struct btrfs_ioctl_search_key *sk = &args->key;
1009 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1010 int ret;
1011 int num_found = 0;
1012 unsigned long sk_offset = 0;
1014 path = btrfs_alloc_path();
1015 if (!path)
1016 return -ENOMEM;
1018 if (sk->tree_id == 0) {
1019 /* search the root of the inode that was passed */
1020 root = BTRFS_I(inode)->root;
1021 } else {
1022 key.objectid = sk->tree_id;
1023 key.type = BTRFS_ROOT_ITEM_KEY;
1024 key.offset = (u64)-1;
1025 root = btrfs_read_fs_root_no_name(info, &key);
1026 if (IS_ERR(root)) {
1027 printk(KERN_ERR "could not find root %llu\n",
1028 sk->tree_id);
1029 btrfs_free_path(path);
1030 return -ENOENT;
1034 key.objectid = sk->min_objectid;
1035 key.type = sk->min_type;
1036 key.offset = sk->min_offset;
1038 max_key.objectid = sk->max_objectid;
1039 max_key.type = sk->max_type;
1040 max_key.offset = sk->max_offset;
1042 path->keep_locks = 1;
1044 while(1) {
1045 ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1046 sk->min_transid);
1047 if (ret != 0) {
1048 if (ret > 0)
1049 ret = 0;
1050 goto err;
1052 ret = copy_to_sk(root, path, &key, sk, args->buf,
1053 &sk_offset, &num_found);
1054 btrfs_release_path(root, path);
1055 if (ret || num_found >= sk->nr_items)
1056 break;
1059 ret = 0;
1060 err:
1061 sk->nr_items = num_found;
1062 btrfs_free_path(path);
1063 return ret;
1066 static noinline int btrfs_ioctl_tree_search(struct file *file,
1067 void __user *argp)
1069 struct btrfs_ioctl_search_args *args;
1070 struct inode *inode;
1071 int ret;
1073 if (!capable(CAP_SYS_ADMIN))
1074 return -EPERM;
1076 args = memdup_user(argp, sizeof(*args));
1077 if (IS_ERR(args))
1078 return PTR_ERR(args);
1080 inode = fdentry(file)->d_inode;
1081 ret = search_ioctl(inode, args);
1082 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1083 ret = -EFAULT;
1084 kfree(args);
1085 return ret;
1089 * Search INODE_REFs to identify path name of 'dirid' directory
1090 * in a 'tree_id' tree. and sets path name to 'name'.
1092 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1093 u64 tree_id, u64 dirid, char *name)
1095 struct btrfs_root *root;
1096 struct btrfs_key key;
1097 char *ptr;
1098 int ret = -1;
1099 int slot;
1100 int len;
1101 int total_len = 0;
1102 struct btrfs_inode_ref *iref;
1103 struct extent_buffer *l;
1104 struct btrfs_path *path;
1106 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1107 name[0]='\0';
1108 return 0;
1111 path = btrfs_alloc_path();
1112 if (!path)
1113 return -ENOMEM;
1115 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1117 key.objectid = tree_id;
1118 key.type = BTRFS_ROOT_ITEM_KEY;
1119 key.offset = (u64)-1;
1120 root = btrfs_read_fs_root_no_name(info, &key);
1121 if (IS_ERR(root)) {
1122 printk(KERN_ERR "could not find root %llu\n", tree_id);
1123 ret = -ENOENT;
1124 goto out;
1127 key.objectid = dirid;
1128 key.type = BTRFS_INODE_REF_KEY;
1129 key.offset = (u64)-1;
1131 while(1) {
1132 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1133 if (ret < 0)
1134 goto out;
1136 l = path->nodes[0];
1137 slot = path->slots[0];
1138 if (ret > 0 && slot > 0)
1139 slot--;
1140 btrfs_item_key_to_cpu(l, &key, slot);
1142 if (ret > 0 && (key.objectid != dirid ||
1143 key.type != BTRFS_INODE_REF_KEY)) {
1144 ret = -ENOENT;
1145 goto out;
1148 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1149 len = btrfs_inode_ref_name_len(l, iref);
1150 ptr -= len + 1;
1151 total_len += len + 1;
1152 if (ptr < name)
1153 goto out;
1155 *(ptr + len) = '/';
1156 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1158 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1159 break;
1161 btrfs_release_path(root, path);
1162 key.objectid = key.offset;
1163 key.offset = (u64)-1;
1164 dirid = key.objectid;
1167 if (ptr < name)
1168 goto out;
1169 memcpy(name, ptr, total_len);
1170 name[total_len]='\0';
1171 ret = 0;
1172 out:
1173 btrfs_free_path(path);
1174 return ret;
1177 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1178 void __user *argp)
1180 struct btrfs_ioctl_ino_lookup_args *args;
1181 struct inode *inode;
1182 int ret;
1184 if (!capable(CAP_SYS_ADMIN))
1185 return -EPERM;
1187 args = memdup_user(argp, sizeof(*args));
1188 if (IS_ERR(args))
1189 return PTR_ERR(args);
1191 inode = fdentry(file)->d_inode;
1193 if (args->treeid == 0)
1194 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1196 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1197 args->treeid, args->objectid,
1198 args->name);
1200 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1201 ret = -EFAULT;
1203 kfree(args);
1204 return ret;
1207 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1208 void __user *arg)
1210 struct dentry *parent = fdentry(file);
1211 struct dentry *dentry;
1212 struct inode *dir = parent->d_inode;
1213 struct inode *inode;
1214 struct btrfs_root *root = BTRFS_I(dir)->root;
1215 struct btrfs_root *dest = NULL;
1216 struct btrfs_ioctl_vol_args *vol_args;
1217 struct btrfs_trans_handle *trans;
1218 int namelen;
1219 int ret;
1220 int err = 0;
1222 if (!capable(CAP_SYS_ADMIN))
1223 return -EPERM;
1225 vol_args = memdup_user(arg, sizeof(*vol_args));
1226 if (IS_ERR(vol_args))
1227 return PTR_ERR(vol_args);
1229 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1230 namelen = strlen(vol_args->name);
1231 if (strchr(vol_args->name, '/') ||
1232 strncmp(vol_args->name, "..", namelen) == 0) {
1233 err = -EINVAL;
1234 goto out;
1237 err = mnt_want_write(file->f_path.mnt);
1238 if (err)
1239 goto out;
1241 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
1242 dentry = lookup_one_len(vol_args->name, parent, namelen);
1243 if (IS_ERR(dentry)) {
1244 err = PTR_ERR(dentry);
1245 goto out_unlock_dir;
1248 if (!dentry->d_inode) {
1249 err = -ENOENT;
1250 goto out_dput;
1253 inode = dentry->d_inode;
1254 if (inode->i_ino != BTRFS_FIRST_FREE_OBJECTID) {
1255 err = -EINVAL;
1256 goto out_dput;
1259 dest = BTRFS_I(inode)->root;
1261 mutex_lock(&inode->i_mutex);
1262 err = d_invalidate(dentry);
1263 if (err)
1264 goto out_unlock;
1266 down_write(&root->fs_info->subvol_sem);
1268 err = may_destroy_subvol(dest);
1269 if (err)
1270 goto out_up_write;
1272 trans = btrfs_start_transaction(root, 0);
1273 if (IS_ERR(trans)) {
1274 err = PTR_ERR(trans);
1275 goto out_up_write;
1277 trans->block_rsv = &root->fs_info->global_block_rsv;
1279 ret = btrfs_unlink_subvol(trans, root, dir,
1280 dest->root_key.objectid,
1281 dentry->d_name.name,
1282 dentry->d_name.len);
1283 BUG_ON(ret);
1285 btrfs_record_root_in_trans(trans, dest);
1287 memset(&dest->root_item.drop_progress, 0,
1288 sizeof(dest->root_item.drop_progress));
1289 dest->root_item.drop_level = 0;
1290 btrfs_set_root_refs(&dest->root_item, 0);
1292 if (!xchg(&dest->orphan_item_inserted, 1)) {
1293 ret = btrfs_insert_orphan_item(trans,
1294 root->fs_info->tree_root,
1295 dest->root_key.objectid);
1296 BUG_ON(ret);
1299 ret = btrfs_commit_transaction(trans, root);
1300 BUG_ON(ret);
1301 inode->i_flags |= S_DEAD;
1302 out_up_write:
1303 up_write(&root->fs_info->subvol_sem);
1304 out_unlock:
1305 mutex_unlock(&inode->i_mutex);
1306 if (!err) {
1307 shrink_dcache_sb(root->fs_info->sb);
1308 btrfs_invalidate_inodes(dest);
1309 d_delete(dentry);
1311 out_dput:
1312 dput(dentry);
1313 out_unlock_dir:
1314 mutex_unlock(&dir->i_mutex);
1315 mnt_drop_write(file->f_path.mnt);
1316 out:
1317 kfree(vol_args);
1318 return err;
1321 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
1323 struct inode *inode = fdentry(file)->d_inode;
1324 struct btrfs_root *root = BTRFS_I(inode)->root;
1325 struct btrfs_ioctl_defrag_range_args *range;
1326 int ret;
1328 ret = mnt_want_write(file->f_path.mnt);
1329 if (ret)
1330 return ret;
1332 switch (inode->i_mode & S_IFMT) {
1333 case S_IFDIR:
1334 if (!capable(CAP_SYS_ADMIN)) {
1335 ret = -EPERM;
1336 goto out;
1338 ret = btrfs_defrag_root(root, 0);
1339 if (ret)
1340 goto out;
1341 ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
1342 break;
1343 case S_IFREG:
1344 if (!(file->f_mode & FMODE_WRITE)) {
1345 ret = -EINVAL;
1346 goto out;
1349 range = kzalloc(sizeof(*range), GFP_KERNEL);
1350 if (!range) {
1351 ret = -ENOMEM;
1352 goto out;
1355 if (argp) {
1356 if (copy_from_user(range, argp,
1357 sizeof(*range))) {
1358 ret = -EFAULT;
1359 kfree(range);
1360 goto out;
1362 /* compression requires us to start the IO */
1363 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1364 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
1365 range->extent_thresh = (u32)-1;
1367 } else {
1368 /* the rest are all set to zero by kzalloc */
1369 range->len = (u64)-1;
1371 ret = btrfs_defrag_file(file, range);
1372 kfree(range);
1373 break;
1374 default:
1375 ret = -EINVAL;
1377 out:
1378 mnt_drop_write(file->f_path.mnt);
1379 return ret;
1382 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
1384 struct btrfs_ioctl_vol_args *vol_args;
1385 int ret;
1387 if (!capable(CAP_SYS_ADMIN))
1388 return -EPERM;
1390 vol_args = memdup_user(arg, sizeof(*vol_args));
1391 if (IS_ERR(vol_args))
1392 return PTR_ERR(vol_args);
1394 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1395 ret = btrfs_init_new_device(root, vol_args->name);
1397 kfree(vol_args);
1398 return ret;
1401 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
1403 struct btrfs_ioctl_vol_args *vol_args;
1404 int ret;
1406 if (!capable(CAP_SYS_ADMIN))
1407 return -EPERM;
1409 if (root->fs_info->sb->s_flags & MS_RDONLY)
1410 return -EROFS;
1412 vol_args = memdup_user(arg, sizeof(*vol_args));
1413 if (IS_ERR(vol_args))
1414 return PTR_ERR(vol_args);
1416 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1417 ret = btrfs_rm_device(root, vol_args->name);
1419 kfree(vol_args);
1420 return ret;
1423 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
1424 u64 off, u64 olen, u64 destoff)
1426 struct inode *inode = fdentry(file)->d_inode;
1427 struct btrfs_root *root = BTRFS_I(inode)->root;
1428 struct file *src_file;
1429 struct inode *src;
1430 struct btrfs_trans_handle *trans;
1431 struct btrfs_path *path;
1432 struct extent_buffer *leaf;
1433 char *buf;
1434 struct btrfs_key key;
1435 u32 nritems;
1436 int slot;
1437 int ret;
1438 u64 len = olen;
1439 u64 bs = root->fs_info->sb->s_blocksize;
1440 u64 hint_byte;
1443 * TODO:
1444 * - split compressed inline extents. annoying: we need to
1445 * decompress into destination's address_space (the file offset
1446 * may change, so source mapping won't do), then recompress (or
1447 * otherwise reinsert) a subrange.
1448 * - allow ranges within the same file to be cloned (provided
1449 * they don't overlap)?
1452 /* the destination must be opened for writing */
1453 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
1454 return -EINVAL;
1456 ret = mnt_want_write(file->f_path.mnt);
1457 if (ret)
1458 return ret;
1460 src_file = fget(srcfd);
1461 if (!src_file) {
1462 ret = -EBADF;
1463 goto out_drop_write;
1466 src = src_file->f_dentry->d_inode;
1468 ret = -EINVAL;
1469 if (src == inode)
1470 goto out_fput;
1472 /* the src must be open for reading */
1473 if (!(src_file->f_mode & FMODE_READ))
1474 goto out_fput;
1476 ret = -EISDIR;
1477 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
1478 goto out_fput;
1480 ret = -EXDEV;
1481 if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
1482 goto out_fput;
1484 ret = -ENOMEM;
1485 buf = vmalloc(btrfs_level_size(root, 0));
1486 if (!buf)
1487 goto out_fput;
1489 path = btrfs_alloc_path();
1490 if (!path) {
1491 vfree(buf);
1492 goto out_fput;
1494 path->reada = 2;
1496 if (inode < src) {
1497 mutex_lock(&inode->i_mutex);
1498 mutex_lock(&src->i_mutex);
1499 } else {
1500 mutex_lock(&src->i_mutex);
1501 mutex_lock(&inode->i_mutex);
1504 /* determine range to clone */
1505 ret = -EINVAL;
1506 if (off + len > src->i_size || off + len < off)
1507 goto out_unlock;
1508 if (len == 0)
1509 olen = len = src->i_size - off;
1510 /* if we extend to eof, continue to block boundary */
1511 if (off + len == src->i_size)
1512 len = ((src->i_size + bs-1) & ~(bs-1))
1513 - off;
1515 /* verify the end result is block aligned */
1516 if ((off & (bs-1)) ||
1517 ((off + len) & (bs-1)))
1518 goto out_unlock;
1520 /* do any pending delalloc/csum calc on src, one way or
1521 another, and lock file content */
1522 while (1) {
1523 struct btrfs_ordered_extent *ordered;
1524 lock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
1525 ordered = btrfs_lookup_first_ordered_extent(inode, off+len);
1526 if (BTRFS_I(src)->delalloc_bytes == 0 && !ordered)
1527 break;
1528 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
1529 if (ordered)
1530 btrfs_put_ordered_extent(ordered);
1531 btrfs_wait_ordered_range(src, off, off+len);
1534 /* clone data */
1535 key.objectid = src->i_ino;
1536 key.type = BTRFS_EXTENT_DATA_KEY;
1537 key.offset = 0;
1539 while (1) {
1541 * note the key will change type as we walk through the
1542 * tree.
1544 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1545 if (ret < 0)
1546 goto out;
1548 nritems = btrfs_header_nritems(path->nodes[0]);
1549 if (path->slots[0] >= nritems) {
1550 ret = btrfs_next_leaf(root, path);
1551 if (ret < 0)
1552 goto out;
1553 if (ret > 0)
1554 break;
1555 nritems = btrfs_header_nritems(path->nodes[0]);
1557 leaf = path->nodes[0];
1558 slot = path->slots[0];
1560 btrfs_item_key_to_cpu(leaf, &key, slot);
1561 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
1562 key.objectid != src->i_ino)
1563 break;
1565 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
1566 struct btrfs_file_extent_item *extent;
1567 int type;
1568 u32 size;
1569 struct btrfs_key new_key;
1570 u64 disko = 0, diskl = 0;
1571 u64 datao = 0, datal = 0;
1572 u8 comp;
1573 u64 endoff;
1575 size = btrfs_item_size_nr(leaf, slot);
1576 read_extent_buffer(leaf, buf,
1577 btrfs_item_ptr_offset(leaf, slot),
1578 size);
1580 extent = btrfs_item_ptr(leaf, slot,
1581 struct btrfs_file_extent_item);
1582 comp = btrfs_file_extent_compression(leaf, extent);
1583 type = btrfs_file_extent_type(leaf, extent);
1584 if (type == BTRFS_FILE_EXTENT_REG ||
1585 type == BTRFS_FILE_EXTENT_PREALLOC) {
1586 disko = btrfs_file_extent_disk_bytenr(leaf,
1587 extent);
1588 diskl = btrfs_file_extent_disk_num_bytes(leaf,
1589 extent);
1590 datao = btrfs_file_extent_offset(leaf, extent);
1591 datal = btrfs_file_extent_num_bytes(leaf,
1592 extent);
1593 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
1594 /* take upper bound, may be compressed */
1595 datal = btrfs_file_extent_ram_bytes(leaf,
1596 extent);
1598 btrfs_release_path(root, path);
1600 if (key.offset + datal < off ||
1601 key.offset >= off+len)
1602 goto next;
1604 memcpy(&new_key, &key, sizeof(new_key));
1605 new_key.objectid = inode->i_ino;
1606 new_key.offset = key.offset + destoff - off;
1608 trans = btrfs_start_transaction(root, 1);
1609 if (IS_ERR(trans)) {
1610 ret = PTR_ERR(trans);
1611 goto out;
1614 if (type == BTRFS_FILE_EXTENT_REG ||
1615 type == BTRFS_FILE_EXTENT_PREALLOC) {
1616 if (off > key.offset) {
1617 datao += off - key.offset;
1618 datal -= off - key.offset;
1621 if (key.offset + datal > off + len)
1622 datal = off + len - key.offset;
1624 ret = btrfs_drop_extents(trans, inode,
1625 new_key.offset,
1626 new_key.offset + datal,
1627 &hint_byte, 1);
1628 BUG_ON(ret);
1630 ret = btrfs_insert_empty_item(trans, root, path,
1631 &new_key, size);
1632 BUG_ON(ret);
1634 leaf = path->nodes[0];
1635 slot = path->slots[0];
1636 write_extent_buffer(leaf, buf,
1637 btrfs_item_ptr_offset(leaf, slot),
1638 size);
1640 extent = btrfs_item_ptr(leaf, slot,
1641 struct btrfs_file_extent_item);
1643 /* disko == 0 means it's a hole */
1644 if (!disko)
1645 datao = 0;
1647 btrfs_set_file_extent_offset(leaf, extent,
1648 datao);
1649 btrfs_set_file_extent_num_bytes(leaf, extent,
1650 datal);
1651 if (disko) {
1652 inode_add_bytes(inode, datal);
1653 ret = btrfs_inc_extent_ref(trans, root,
1654 disko, diskl, 0,
1655 root->root_key.objectid,
1656 inode->i_ino,
1657 new_key.offset - datao);
1658 BUG_ON(ret);
1660 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
1661 u64 skip = 0;
1662 u64 trim = 0;
1663 if (off > key.offset) {
1664 skip = off - key.offset;
1665 new_key.offset += skip;
1668 if (key.offset + datal > off+len)
1669 trim = key.offset + datal - (off+len);
1671 if (comp && (skip || trim)) {
1672 ret = -EINVAL;
1673 btrfs_end_transaction(trans, root);
1674 goto out;
1676 size -= skip + trim;
1677 datal -= skip + trim;
1679 ret = btrfs_drop_extents(trans, inode,
1680 new_key.offset,
1681 new_key.offset + datal,
1682 &hint_byte, 1);
1683 BUG_ON(ret);
1685 ret = btrfs_insert_empty_item(trans, root, path,
1686 &new_key, size);
1687 BUG_ON(ret);
1689 if (skip) {
1690 u32 start =
1691 btrfs_file_extent_calc_inline_size(0);
1692 memmove(buf+start, buf+start+skip,
1693 datal);
1696 leaf = path->nodes[0];
1697 slot = path->slots[0];
1698 write_extent_buffer(leaf, buf,
1699 btrfs_item_ptr_offset(leaf, slot),
1700 size);
1701 inode_add_bytes(inode, datal);
1704 btrfs_mark_buffer_dirty(leaf);
1705 btrfs_release_path(root, path);
1707 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
1710 * we round up to the block size at eof when
1711 * determining which extents to clone above,
1712 * but shouldn't round up the file size
1714 endoff = new_key.offset + datal;
1715 if (endoff > off+olen)
1716 endoff = off+olen;
1717 if (endoff > inode->i_size)
1718 btrfs_i_size_write(inode, endoff);
1720 BTRFS_I(inode)->flags = BTRFS_I(src)->flags;
1721 ret = btrfs_update_inode(trans, root, inode);
1722 BUG_ON(ret);
1723 btrfs_end_transaction(trans, root);
1725 next:
1726 btrfs_release_path(root, path);
1727 key.offset++;
1729 ret = 0;
1730 out:
1731 btrfs_release_path(root, path);
1732 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
1733 out_unlock:
1734 mutex_unlock(&src->i_mutex);
1735 mutex_unlock(&inode->i_mutex);
1736 vfree(buf);
1737 btrfs_free_path(path);
1738 out_fput:
1739 fput(src_file);
1740 out_drop_write:
1741 mnt_drop_write(file->f_path.mnt);
1742 return ret;
1745 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
1747 struct btrfs_ioctl_clone_range_args args;
1749 if (copy_from_user(&args, argp, sizeof(args)))
1750 return -EFAULT;
1751 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
1752 args.src_length, args.dest_offset);
1756 * there are many ways the trans_start and trans_end ioctls can lead
1757 * to deadlocks. They should only be used by applications that
1758 * basically own the machine, and have a very in depth understanding
1759 * of all the possible deadlocks and enospc problems.
1761 static long btrfs_ioctl_trans_start(struct file *file)
1763 struct inode *inode = fdentry(file)->d_inode;
1764 struct btrfs_root *root = BTRFS_I(inode)->root;
1765 struct btrfs_trans_handle *trans;
1766 int ret;
1768 ret = -EPERM;
1769 if (!capable(CAP_SYS_ADMIN))
1770 goto out;
1772 ret = -EINPROGRESS;
1773 if (file->private_data)
1774 goto out;
1776 ret = mnt_want_write(file->f_path.mnt);
1777 if (ret)
1778 goto out;
1780 mutex_lock(&root->fs_info->trans_mutex);
1781 root->fs_info->open_ioctl_trans++;
1782 mutex_unlock(&root->fs_info->trans_mutex);
1784 ret = -ENOMEM;
1785 trans = btrfs_start_ioctl_transaction(root, 0);
1786 if (!trans)
1787 goto out_drop;
1789 file->private_data = trans;
1790 return 0;
1792 out_drop:
1793 mutex_lock(&root->fs_info->trans_mutex);
1794 root->fs_info->open_ioctl_trans--;
1795 mutex_unlock(&root->fs_info->trans_mutex);
1796 mnt_drop_write(file->f_path.mnt);
1797 out:
1798 return ret;
1801 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
1803 struct inode *inode = fdentry(file)->d_inode;
1804 struct btrfs_root *root = BTRFS_I(inode)->root;
1805 struct btrfs_root *new_root;
1806 struct btrfs_dir_item *di;
1807 struct btrfs_trans_handle *trans;
1808 struct btrfs_path *path;
1809 struct btrfs_key location;
1810 struct btrfs_disk_key disk_key;
1811 struct btrfs_super_block *disk_super;
1812 u64 features;
1813 u64 objectid = 0;
1814 u64 dir_id;
1816 if (!capable(CAP_SYS_ADMIN))
1817 return -EPERM;
1819 if (copy_from_user(&objectid, argp, sizeof(objectid)))
1820 return -EFAULT;
1822 if (!objectid)
1823 objectid = root->root_key.objectid;
1825 location.objectid = objectid;
1826 location.type = BTRFS_ROOT_ITEM_KEY;
1827 location.offset = (u64)-1;
1829 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
1830 if (IS_ERR(new_root))
1831 return PTR_ERR(new_root);
1833 if (btrfs_root_refs(&new_root->root_item) == 0)
1834 return -ENOENT;
1836 path = btrfs_alloc_path();
1837 if (!path)
1838 return -ENOMEM;
1839 path->leave_spinning = 1;
1841 trans = btrfs_start_transaction(root, 1);
1842 if (!trans) {
1843 btrfs_free_path(path);
1844 return -ENOMEM;
1847 dir_id = btrfs_super_root_dir(&root->fs_info->super_copy);
1848 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
1849 dir_id, "default", 7, 1);
1850 if (IS_ERR_OR_NULL(di)) {
1851 btrfs_free_path(path);
1852 btrfs_end_transaction(trans, root);
1853 printk(KERN_ERR "Umm, you don't have the default dir item, "
1854 "this isn't going to work\n");
1855 return -ENOENT;
1858 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
1859 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
1860 btrfs_mark_buffer_dirty(path->nodes[0]);
1861 btrfs_free_path(path);
1863 disk_super = &root->fs_info->super_copy;
1864 features = btrfs_super_incompat_flags(disk_super);
1865 if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) {
1866 features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL;
1867 btrfs_set_super_incompat_flags(disk_super, features);
1869 btrfs_end_transaction(trans, root);
1871 return 0;
1874 static void get_block_group_info(struct list_head *groups_list,
1875 struct btrfs_ioctl_space_info *space)
1877 struct btrfs_block_group_cache *block_group;
1879 space->total_bytes = 0;
1880 space->used_bytes = 0;
1881 space->flags = 0;
1882 list_for_each_entry(block_group, groups_list, list) {
1883 space->flags = block_group->flags;
1884 space->total_bytes += block_group->key.offset;
1885 space->used_bytes +=
1886 btrfs_block_group_used(&block_group->item);
1890 long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
1892 struct btrfs_ioctl_space_args space_args;
1893 struct btrfs_ioctl_space_info space;
1894 struct btrfs_ioctl_space_info *dest;
1895 struct btrfs_ioctl_space_info *dest_orig;
1896 struct btrfs_ioctl_space_info *user_dest;
1897 struct btrfs_space_info *info;
1898 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
1899 BTRFS_BLOCK_GROUP_SYSTEM,
1900 BTRFS_BLOCK_GROUP_METADATA,
1901 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
1902 int num_types = 4;
1903 int alloc_size;
1904 int ret = 0;
1905 int slot_count = 0;
1906 int i, c;
1908 if (copy_from_user(&space_args,
1909 (struct btrfs_ioctl_space_args __user *)arg,
1910 sizeof(space_args)))
1911 return -EFAULT;
1913 for (i = 0; i < num_types; i++) {
1914 struct btrfs_space_info *tmp;
1916 info = NULL;
1917 rcu_read_lock();
1918 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
1919 list) {
1920 if (tmp->flags == types[i]) {
1921 info = tmp;
1922 break;
1925 rcu_read_unlock();
1927 if (!info)
1928 continue;
1930 down_read(&info->groups_sem);
1931 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
1932 if (!list_empty(&info->block_groups[c]))
1933 slot_count++;
1935 up_read(&info->groups_sem);
1938 /* space_slots == 0 means they are asking for a count */
1939 if (space_args.space_slots == 0) {
1940 space_args.total_spaces = slot_count;
1941 goto out;
1944 slot_count = min_t(int, space_args.space_slots, slot_count);
1946 alloc_size = sizeof(*dest) * slot_count;
1948 /* we generally have at most 6 or so space infos, one for each raid
1949 * level. So, a whole page should be more than enough for everyone
1951 if (alloc_size > PAGE_CACHE_SIZE)
1952 return -ENOMEM;
1954 space_args.total_spaces = 0;
1955 dest = kmalloc(alloc_size, GFP_NOFS);
1956 if (!dest)
1957 return -ENOMEM;
1958 dest_orig = dest;
1960 /* now we have a buffer to copy into */
1961 for (i = 0; i < num_types; i++) {
1962 struct btrfs_space_info *tmp;
1964 info = NULL;
1965 rcu_read_lock();
1966 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
1967 list) {
1968 if (tmp->flags == types[i]) {
1969 info = tmp;
1970 break;
1973 rcu_read_unlock();
1975 if (!info)
1976 continue;
1977 down_read(&info->groups_sem);
1978 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
1979 if (!list_empty(&info->block_groups[c])) {
1980 get_block_group_info(&info->block_groups[c],
1981 &space);
1982 memcpy(dest, &space, sizeof(space));
1983 dest++;
1984 space_args.total_spaces++;
1987 up_read(&info->groups_sem);
1990 user_dest = (struct btrfs_ioctl_space_info *)
1991 (arg + sizeof(struct btrfs_ioctl_space_args));
1993 if (copy_to_user(user_dest, dest_orig, alloc_size))
1994 ret = -EFAULT;
1996 kfree(dest_orig);
1997 out:
1998 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
1999 ret = -EFAULT;
2001 return ret;
2005 * there are many ways the trans_start and trans_end ioctls can lead
2006 * to deadlocks. They should only be used by applications that
2007 * basically own the machine, and have a very in depth understanding
2008 * of all the possible deadlocks and enospc problems.
2010 long btrfs_ioctl_trans_end(struct file *file)
2012 struct inode *inode = fdentry(file)->d_inode;
2013 struct btrfs_root *root = BTRFS_I(inode)->root;
2014 struct btrfs_trans_handle *trans;
2016 trans = file->private_data;
2017 if (!trans)
2018 return -EINVAL;
2019 file->private_data = NULL;
2021 btrfs_end_transaction(trans, root);
2023 mutex_lock(&root->fs_info->trans_mutex);
2024 root->fs_info->open_ioctl_trans--;
2025 mutex_unlock(&root->fs_info->trans_mutex);
2027 mnt_drop_write(file->f_path.mnt);
2028 return 0;
2031 long btrfs_ioctl(struct file *file, unsigned int
2032 cmd, unsigned long arg)
2034 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
2035 void __user *argp = (void __user *)arg;
2037 switch (cmd) {
2038 case FS_IOC_GETFLAGS:
2039 return btrfs_ioctl_getflags(file, argp);
2040 case FS_IOC_SETFLAGS:
2041 return btrfs_ioctl_setflags(file, argp);
2042 case FS_IOC_GETVERSION:
2043 return btrfs_ioctl_getversion(file, argp);
2044 case BTRFS_IOC_SNAP_CREATE:
2045 return btrfs_ioctl_snap_create(file, argp, 0);
2046 case BTRFS_IOC_SUBVOL_CREATE:
2047 return btrfs_ioctl_snap_create(file, argp, 1);
2048 case BTRFS_IOC_SNAP_DESTROY:
2049 return btrfs_ioctl_snap_destroy(file, argp);
2050 case BTRFS_IOC_DEFAULT_SUBVOL:
2051 return btrfs_ioctl_default_subvol(file, argp);
2052 case BTRFS_IOC_DEFRAG:
2053 return btrfs_ioctl_defrag(file, NULL);
2054 case BTRFS_IOC_DEFRAG_RANGE:
2055 return btrfs_ioctl_defrag(file, argp);
2056 case BTRFS_IOC_RESIZE:
2057 return btrfs_ioctl_resize(root, argp);
2058 case BTRFS_IOC_ADD_DEV:
2059 return btrfs_ioctl_add_dev(root, argp);
2060 case BTRFS_IOC_RM_DEV:
2061 return btrfs_ioctl_rm_dev(root, argp);
2062 case BTRFS_IOC_BALANCE:
2063 return btrfs_balance(root->fs_info->dev_root);
2064 case BTRFS_IOC_CLONE:
2065 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
2066 case BTRFS_IOC_CLONE_RANGE:
2067 return btrfs_ioctl_clone_range(file, argp);
2068 case BTRFS_IOC_TRANS_START:
2069 return btrfs_ioctl_trans_start(file);
2070 case BTRFS_IOC_TRANS_END:
2071 return btrfs_ioctl_trans_end(file);
2072 case BTRFS_IOC_TREE_SEARCH:
2073 return btrfs_ioctl_tree_search(file, argp);
2074 case BTRFS_IOC_INO_LOOKUP:
2075 return btrfs_ioctl_ino_lookup(file, argp);
2076 case BTRFS_IOC_SPACE_INFO:
2077 return btrfs_ioctl_space_info(root, argp);
2078 case BTRFS_IOC_SYNC:
2079 btrfs_sync_fs(file->f_dentry->d_sb, 1);
2080 return 0;
2083 return -ENOTTY;