?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
MIPS: Fix sign-extension bug in 32-bit kernel on 32-bit hardware.
[linux-2.6/mini2440.git] / fs / btrfs / ioctl.c
blob5e94ea6e1cbe05ea164e5f6128b87c84b7bc8570
1 /*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
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.
7 *
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.
12 *
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.
17 */
18
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/smp_lock.h>
31 #include <linux/backing-dev.h>
32 #include <linux/mount.h>
33 #include <linux/mpage.h>
34 #include <linux/namei.h>
35 #include <linux/swap.h>
36 #include <linux/writeback.h>
37 #include <linux/statfs.h>
38 #include <linux/compat.h>
39 #include <linux/bit_spinlock.h>
40 #include <linux/security.h>
41 #include <linux/xattr.h>
42 #include <linux/vmalloc.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"
52
53
54
55 static noinline int create_subvol(struct btrfs_root *root,
56 struct dentry *dentry,
57 char *name, int namelen)
58 {
59 struct btrfs_trans_handle *trans;
60 struct btrfs_key key;
61 struct btrfs_root_item root_item;
62 struct btrfs_inode_item *inode_item;
63 struct extent_buffer *leaf;
64 struct btrfs_root *new_root = root;
65 struct inode *dir;
66 int ret;
67 int err;
68 u64 objectid;
69 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
70 u64 index = 0;
71 unsigned long nr = 1;
72
73 ret = btrfs_check_metadata_free_space(root);
74 if (ret)
75 goto fail_commit;
76
77 trans = btrfs_start_transaction(root, 1);
78 BUG_ON(!trans);
79
80 ret = btrfs_find_free_objectid(trans, root->fs_info->tree_root,
81 0, &objectid);
82 if (ret)
83 goto fail;
84
85 leaf = btrfs_alloc_free_block(trans, root, root->leafsize, 0,
86 objectid, trans->transid, 0, 0, 0);
87 if (IS_ERR(leaf)) {
88 ret = PTR_ERR(leaf);
89 goto fail;
90 }
91
92 btrfs_set_header_nritems(leaf, 0);
93 btrfs_set_header_level(leaf, 0);
94 btrfs_set_header_bytenr(leaf, leaf->start);
95 btrfs_set_header_generation(leaf, trans->transid);
96 btrfs_set_header_owner(leaf, objectid);
97
98 write_extent_buffer(leaf, root->fs_info->fsid,
99 (unsigned long)btrfs_header_fsid(leaf),
100 BTRFS_FSID_SIZE);
101 btrfs_mark_buffer_dirty(leaf);
102
103 inode_item = &root_item.inode;
104 memset(inode_item, 0, sizeof(*inode_item));
105 inode_item->generation = cpu_to_le64(1);
106 inode_item->size = cpu_to_le64(3);
107 inode_item->nlink = cpu_to_le32(1);
108 inode_item->nbytes = cpu_to_le64(root->leafsize);
109 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
110
111 btrfs_set_root_bytenr(&root_item, leaf->start);
112 btrfs_set_root_generation(&root_item, trans->transid);
113 btrfs_set_root_level(&root_item, 0);
114 btrfs_set_root_refs(&root_item, 1);
115 btrfs_set_root_used(&root_item, 0);
116 btrfs_set_root_last_snapshot(&root_item, 0);
117
118 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
119 root_item.drop_level = 0;
120
121 btrfs_tree_unlock(leaf);
122 free_extent_buffer(leaf);
123 leaf = NULL;
124
125 btrfs_set_root_dirid(&root_item, new_dirid);
126
127 key.objectid = objectid;
128 key.offset = 1;
129 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
130 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
131 &root_item);
132 if (ret)
133 goto fail;
134
135 /*
136 * insert the directory item
137 */
138 key.offset = (u64)-1;
139 dir = dentry->d_parent->d_inode;
140 ret = btrfs_set_inode_index(dir, &index);
141 BUG_ON(ret);
142
143 ret = btrfs_insert_dir_item(trans, root,
144 name, namelen, dir->i_ino, &key,
145 BTRFS_FT_DIR, index);
146 if (ret)
147 goto fail;
148
149 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
150 ret = btrfs_update_inode(trans, root, dir);
151 BUG_ON(ret);
152
153 /* add the backref first */
154 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
155 objectid, BTRFS_ROOT_BACKREF_KEY,
156 root->root_key.objectid,
157 dir->i_ino, index, name, namelen);
158
159 BUG_ON(ret);
160
161 /* now add the forward ref */
162 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
163 root->root_key.objectid, BTRFS_ROOT_REF_KEY,
164 objectid,
165 dir->i_ino, index, name, namelen);
166
167 BUG_ON(ret);
168
169 ret = btrfs_commit_transaction(trans, root);
170 if (ret)
171 goto fail_commit;
172
173 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
174 BUG_ON(!new_root);
175
176 trans = btrfs_start_transaction(new_root, 1);
177 BUG_ON(!trans);
178
179 ret = btrfs_create_subvol_root(trans, new_root, dentry, new_dirid,
180 BTRFS_I(dir)->block_group);
181 if (ret)
182 goto fail;
183
184 fail:
185 nr = trans->blocks_used;
186 err = btrfs_commit_transaction(trans, new_root);
187 if (err && !ret)
188 ret = err;
189 fail_commit:
190 btrfs_btree_balance_dirty(root, nr);
191 return ret;
192 }
193
194 static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
195 char *name, int namelen)
196 {
197 struct btrfs_pending_snapshot *pending_snapshot;
198 struct btrfs_trans_handle *trans;
199 int ret = 0;
200 int err;
201 unsigned long nr = 0;
202
203 if (!root->ref_cows)
204 return -EINVAL;
205
206 ret = btrfs_check_metadata_free_space(root);
207 if (ret)
208 goto fail_unlock;
209
210 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
211 if (!pending_snapshot) {
212 ret = -ENOMEM;
213 goto fail_unlock;
214 }
215 pending_snapshot->name = kmalloc(namelen + 1, GFP_NOFS);
216 if (!pending_snapshot->name) {
217 ret = -ENOMEM;
218 kfree(pending_snapshot);
219 goto fail_unlock;
220 }
221 memcpy(pending_snapshot->name, name, namelen);
222 pending_snapshot->name[namelen] = '\0';
223 pending_snapshot->dentry = dentry;
224 trans = btrfs_start_transaction(root, 1);
225 BUG_ON(!trans);
226 pending_snapshot->root = root;
227 list_add(&pending_snapshot->list,
228 &trans->transaction->pending_snapshots);
229 err = btrfs_commit_transaction(trans, root);
230
231 fail_unlock:
232 btrfs_btree_balance_dirty(root, nr);
233 return ret;
234 }
235
236 /* copy of may_create in fs/namei.c() */
237 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
238 {
239 if (child->d_inode)
240 return -EEXIST;
241 if (IS_DEADDIR(dir))
242 return -ENOENT;
243 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
244 }
245
246 /*
247 * Create a new subvolume below @parent. This is largely modeled after
248 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
249 * inside this filesystem so it's quite a bit simpler.
250 */
251 static noinline int btrfs_mksubvol(struct path *parent, char *name,
252 int mode, int namelen,
253 struct btrfs_root *snap_src)
254 {
255 struct dentry *dentry;
256 int error;
257
258 mutex_lock_nested(&parent->dentry->d_inode->i_mutex, I_MUTEX_PARENT);
259
260 dentry = lookup_one_len(name, parent->dentry, namelen);
261 error = PTR_ERR(dentry);
262 if (IS_ERR(dentry))
263 goto out_unlock;
264
265 error = -EEXIST;
266 if (dentry->d_inode)
267 goto out_dput;
268
269 if (!IS_POSIXACL(parent->dentry->d_inode))
270 mode &= ~current_umask();
271
272 error = mnt_want_write(parent->mnt);
273 if (error)
274 goto out_dput;
275
276 error = btrfs_may_create(parent->dentry->d_inode, dentry);
277 if (error)
278 goto out_drop_write;
279
280 /*
281 * Actually perform the low-level subvolume creation after all
282 * this VFS fuzz.
283 *
284 * Eventually we want to pass in an inode under which we create this
285 * subvolume, but for now all are under the filesystem root.
286 *
287 * Also we should pass on the mode eventually to allow creating new
288 * subvolume with specific mode bits.
289 */
290 if (snap_src) {
291 struct dentry *dir = dentry->d_parent;
292 struct dentry *test = dir->d_parent;
293 struct btrfs_path *path = btrfs_alloc_path();
294 int ret;
295 u64 test_oid;
296 u64 parent_oid = BTRFS_I(dir->d_inode)->root->root_key.objectid;
297
298 test_oid = snap_src->root_key.objectid;
299
300 ret = btrfs_find_root_ref(snap_src->fs_info->tree_root,
301 path, parent_oid, test_oid);
302 if (ret == 0)
303 goto create;
304 btrfs_release_path(snap_src->fs_info->tree_root, path);
305
306 /* we need to make sure we aren't creating a directory loop
307 * by taking a snapshot of something that has our current
308 * subvol in its directory tree. So, this loops through
309 * the dentries and checks the forward refs for each subvolume
310 * to see if is references the subvolume where we are
311 * placing this new snapshot.
312 */
313 while (1) {
314 if (!test ||
315 dir == snap_src->fs_info->sb->s_root ||
316 test == snap_src->fs_info->sb->s_root ||
317 test->d_inode->i_sb != snap_src->fs_info->sb) {
318 break;
319 }
320 if (S_ISLNK(test->d_inode->i_mode)) {
321 printk(KERN_INFO "Btrfs symlink in snapshot "
322 "path, failed\n");
323 error = -EMLINK;
324 btrfs_free_path(path);
325 goto out_drop_write;
326 }
327 test_oid =
328 BTRFS_I(test->d_inode)->root->root_key.objectid;
329 ret = btrfs_find_root_ref(snap_src->fs_info->tree_root,
330 path, test_oid, parent_oid);
331 if (ret == 0) {
332 printk(KERN_INFO "Btrfs snapshot creation "
333 "failed, looping\n");
334 error = -EMLINK;
335 btrfs_free_path(path);
336 goto out_drop_write;
337 }
338 btrfs_release_path(snap_src->fs_info->tree_root, path);
339 test = test->d_parent;
340 }
341 create:
342 btrfs_free_path(path);
343 error = create_snapshot(snap_src, dentry, name, namelen);
344 } else {
345 error = create_subvol(BTRFS_I(parent->dentry->d_inode)->root,
346 dentry, name, namelen);
347 }
348 if (error)
349 goto out_drop_write;
350
351 fsnotify_mkdir(parent->dentry->d_inode, dentry);
352 out_drop_write:
353 mnt_drop_write(parent->mnt);
354 out_dput:
355 dput(dentry);
356 out_unlock:
357 mutex_unlock(&parent->dentry->d_inode->i_mutex);
358 return error;
359 }
360
361
362 static int btrfs_defrag_file(struct file *file)
363 {
364 struct inode *inode = fdentry(file)->d_inode;
365 struct btrfs_root *root = BTRFS_I(inode)->root;
366 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
367 struct btrfs_ordered_extent *ordered;
368 struct page *page;
369 unsigned long last_index;
370 unsigned long ra_pages = root->fs_info->bdi.ra_pages;
371 unsigned long total_read = 0;
372 u64 page_start;
373 u64 page_end;
374 unsigned long i;
375 int ret;
376
377 ret = btrfs_check_data_free_space(root, inode, inode->i_size);
378 if (ret)
379 return -ENOSPC;
380
381 mutex_lock(&inode->i_mutex);
382 last_index = inode->i_size >> PAGE_CACHE_SHIFT;
383 for (i = 0; i <= last_index; i++) {
384 if (total_read % ra_pages == 0) {
385 btrfs_force_ra(inode->i_mapping, &file->f_ra, file, i,
386 min(last_index, i + ra_pages - 1));
387 }
388 total_read++;
389 again:
390 page = grab_cache_page(inode->i_mapping, i);
391 if (!page)
392 goto out_unlock;
393 if (!PageUptodate(page)) {
394 btrfs_readpage(NULL, page);
395 lock_page(page);
396 if (!PageUptodate(page)) {
397 unlock_page(page);
398 page_cache_release(page);
399 goto out_unlock;
400 }
401 }
402
403 wait_on_page_writeback(page);
404
405 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
406 page_end = page_start + PAGE_CACHE_SIZE - 1;
407 lock_extent(io_tree, page_start, page_end, GFP_NOFS);
408
409 ordered = btrfs_lookup_ordered_extent(inode, page_start);
410 if (ordered) {
411 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
412 unlock_page(page);
413 page_cache_release(page);
414 btrfs_start_ordered_extent(inode, ordered, 1);
415 btrfs_put_ordered_extent(ordered);
416 goto again;
417 }
418 set_page_extent_mapped(page);
419
420 /*
421 * this makes sure page_mkwrite is called on the
422 * page if it is dirtied again later
423 */
424 clear_page_dirty_for_io(page);
425
426 btrfs_set_extent_delalloc(inode, page_start, page_end);
427
428 unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
429 set_page_dirty(page);
430 unlock_page(page);
431 page_cache_release(page);
432 balance_dirty_pages_ratelimited_nr(inode->i_mapping, 1);
433 }
434
435 out_unlock:
436 mutex_unlock(&inode->i_mutex);
437 return 0;
438 }
439
440 /*
441 * Called inside transaction, so use GFP_NOFS
442 */
443
444 static int btrfs_ioctl_resize(struct btrfs_root *root, void __user *arg)
445 {
446 u64 new_size;
447 u64 old_size;
448 u64 devid = 1;
449 struct btrfs_ioctl_vol_args *vol_args;
450 struct btrfs_trans_handle *trans;
451 struct btrfs_device *device = NULL;
452 char *sizestr;
453 char *devstr = NULL;
454 int ret = 0;
455 int namelen;
456 int mod = 0;
457
458 if (root->fs_info->sb->s_flags & MS_RDONLY)
459 return -EROFS;
460
461 if (!capable(CAP_SYS_ADMIN))
462 return -EPERM;
463
464 vol_args = memdup_user(arg, sizeof(*vol_args));
465 if (IS_ERR(vol_args))
466 return PTR_ERR(vol_args);
467
468 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
469 namelen = strlen(vol_args->name);
470
471 mutex_lock(&root->fs_info->volume_mutex);
472 sizestr = vol_args->name;
473 devstr = strchr(sizestr, ':');
474 if (devstr) {
475 char *end;
476 sizestr = devstr + 1;
477 *devstr = '\0';
478 devstr = vol_args->name;
479 devid = simple_strtoull(devstr, &end, 10);
480 printk(KERN_INFO "resizing devid %llu\n",
481 (unsigned long long)devid);
482 }
483 device = btrfs_find_device(root, devid, NULL, NULL);
484 if (!device) {
485 printk(KERN_INFO "resizer unable to find device %llu\n",
486 (unsigned long long)devid);
487 ret = -EINVAL;
488 goto out_unlock;
489 }
490 if (!strcmp(sizestr, "max"))
491 new_size = device->bdev->bd_inode->i_size;
492 else {
493 if (sizestr[0] == '-') {
494 mod = -1;
495 sizestr++;
496 } else if (sizestr[0] == '+') {
497 mod = 1;
498 sizestr++;
499 }
500 new_size = btrfs_parse_size(sizestr);
501 if (new_size == 0) {
502 ret = -EINVAL;
503 goto out_unlock;
504 }
505 }
506
507 old_size = device->total_bytes;
508
509 if (mod < 0) {
510 if (new_size > old_size) {
511 ret = -EINVAL;
512 goto out_unlock;
513 }
514 new_size = old_size - new_size;
515 } else if (mod > 0) {
516 new_size = old_size + new_size;
517 }
518
519 if (new_size < 256 * 1024 * 1024) {
520 ret = -EINVAL;
521 goto out_unlock;
522 }
523 if (new_size > device->bdev->bd_inode->i_size) {
524 ret = -EFBIG;
525 goto out_unlock;
526 }
527
528 do_div(new_size, root->sectorsize);
529 new_size *= root->sectorsize;
530
531 printk(KERN_INFO "new size for %s is %llu\n",
532 device->name, (unsigned long long)new_size);
533
534 if (new_size > old_size) {
535 trans = btrfs_start_transaction(root, 1);
536 ret = btrfs_grow_device(trans, device, new_size);
537 btrfs_commit_transaction(trans, root);
538 } else {
539 ret = btrfs_shrink_device(device, new_size);
540 }
541
542 out_unlock:
543 mutex_unlock(&root->fs_info->volume_mutex);
544 kfree(vol_args);
545 return ret;
546 }
547
548 static noinline int btrfs_ioctl_snap_create(struct file *file,
549 void __user *arg, int subvol)
550 {
551 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
552 struct btrfs_ioctl_vol_args *vol_args;
553 struct btrfs_dir_item *di;
554 struct btrfs_path *path;
555 struct file *src_file;
556 u64 root_dirid;
557 int namelen;
558 int ret = 0;
559
560 if (root->fs_info->sb->s_flags & MS_RDONLY)
561 return -EROFS;
562
563 vol_args = memdup_user(arg, sizeof(*vol_args));
564 if (IS_ERR(vol_args))
565 return PTR_ERR(vol_args);
566
567 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
568 namelen = strlen(vol_args->name);
569 if (strchr(vol_args->name, '/')) {
570 ret = -EINVAL;
571 goto out;
572 }
573
574 path = btrfs_alloc_path();
575 if (!path) {
576 ret = -ENOMEM;
577 goto out;
578 }
579
580 root_dirid = root->fs_info->sb->s_root->d_inode->i_ino,
581 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root,
582 path, root_dirid,
583 vol_args->name, namelen, 0);
584 btrfs_free_path(path);
585
586 if (di && !IS_ERR(di)) {
587 ret = -EEXIST;
588 goto out;
589 }
590
591 if (IS_ERR(di)) {
592 ret = PTR_ERR(di);
593 goto out;
594 }
595
596 if (subvol) {
597 ret = btrfs_mksubvol(&file->f_path, vol_args->name,
598 file->f_path.dentry->d_inode->i_mode,
599 namelen, NULL);
600 } else {
601 struct inode *src_inode;
602 src_file = fget(vol_args->fd);
603 if (!src_file) {
604 ret = -EINVAL;
605 goto out;
606 }
607
608 src_inode = src_file->f_path.dentry->d_inode;
609 if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
610 printk(KERN_INFO "btrfs: Snapshot src from "
611 "another FS\n");
612 ret = -EINVAL;
613 fput(src_file);
614 goto out;
615 }
616 ret = btrfs_mksubvol(&file->f_path, vol_args->name,
617 file->f_path.dentry->d_inode->i_mode,
618 namelen, BTRFS_I(src_inode)->root);
619 fput(src_file);
620 }
621
622 out:
623 kfree(vol_args);
624 return ret;
625 }
626
627 static int btrfs_ioctl_defrag(struct file *file)
628 {
629 struct inode *inode = fdentry(file)->d_inode;
630 struct btrfs_root *root = BTRFS_I(inode)->root;
631 int ret;
632
633 ret = mnt_want_write(file->f_path.mnt);
634 if (ret)
635 return ret;
636
637 switch (inode->i_mode & S_IFMT) {
638 case S_IFDIR:
639 if (!capable(CAP_SYS_ADMIN)) {
640 ret = -EPERM;
641 goto out;
642 }
643 btrfs_defrag_root(root, 0);
644 btrfs_defrag_root(root->fs_info->extent_root, 0);
645 break;
646 case S_IFREG:
647 if (!(file->f_mode & FMODE_WRITE)) {
648 ret = -EINVAL;
649 goto out;
650 }
651 btrfs_defrag_file(file);
652 break;
653 }
654 out:
655 mnt_drop_write(file->f_path.mnt);
656 return ret;
657 }
658
659 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
660 {
661 struct btrfs_ioctl_vol_args *vol_args;
662 int ret;
663
664 if (!capable(CAP_SYS_ADMIN))
665 return -EPERM;
666
667 vol_args = memdup_user(arg, sizeof(*vol_args));
668 if (IS_ERR(vol_args))
669 return PTR_ERR(vol_args);
670
671 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
672 ret = btrfs_init_new_device(root, vol_args->name);
673
674 kfree(vol_args);
675 return ret;
676 }
677
678 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
679 {
680 struct btrfs_ioctl_vol_args *vol_args;
681 int ret;
682
683 if (!capable(CAP_SYS_ADMIN))
684 return -EPERM;
685
686 if (root->fs_info->sb->s_flags & MS_RDONLY)
687 return -EROFS;
688
689 vol_args = memdup_user(arg, sizeof(*vol_args));
690 if (IS_ERR(vol_args))
691 return PTR_ERR(vol_args);
692
693 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
694 ret = btrfs_rm_device(root, vol_args->name);
695
696 kfree(vol_args);
697 return ret;
698 }
699
700 static long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
701 u64 off, u64 olen, u64 destoff)
702 {
703 struct inode *inode = fdentry(file)->d_inode;
704 struct btrfs_root *root = BTRFS_I(inode)->root;
705 struct file *src_file;
706 struct inode *src;
707 struct btrfs_trans_handle *trans;
708 struct btrfs_path *path;
709 struct extent_buffer *leaf;
710 char *buf;
711 struct btrfs_key key;
712 u32 nritems;
713 int slot;
714 int ret;
715 u64 len = olen;
716 u64 bs = root->fs_info->sb->s_blocksize;
717 u64 hint_byte;
718
719 /*
720 * TODO:
721 * - split compressed inline extents. annoying: we need to
722 * decompress into destination's address_space (the file offset
723 * may change, so source mapping won't do), then recompress (or
724 * otherwise reinsert) a subrange.
725 * - allow ranges within the same file to be cloned (provided
726 * they don't overlap)?
727 */
728
729 /* the destination must be opened for writing */
730 if (!(file->f_mode & FMODE_WRITE))
731 return -EINVAL;
732
733 ret = mnt_want_write(file->f_path.mnt);
734 if (ret)
735 return ret;
736
737 src_file = fget(srcfd);
738 if (!src_file) {
739 ret = -EBADF;
740 goto out_drop_write;
741 }
742 src = src_file->f_dentry->d_inode;
743
744 ret = -EINVAL;
745 if (src == inode)
746 goto out_fput;
747
748 ret = -EISDIR;
749 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
750 goto out_fput;
751
752 ret = -EXDEV;
753 if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
754 goto out_fput;
755
756 ret = -ENOMEM;
757 buf = vmalloc(btrfs_level_size(root, 0));
758 if (!buf)
759 goto out_fput;
760
761 path = btrfs_alloc_path();
762 if (!path) {
763 vfree(buf);
764 goto out_fput;
765 }
766 path->reada = 2;
767
768 if (inode < src) {
769 mutex_lock(&inode->i_mutex);
770 mutex_lock(&src->i_mutex);
771 } else {
772 mutex_lock(&src->i_mutex);
773 mutex_lock(&inode->i_mutex);
774 }
775
776 /* determine range to clone */
777 ret = -EINVAL;
778 if (off >= src->i_size || off + len > src->i_size)
779 goto out_unlock;
780 if (len == 0)
781 olen = len = src->i_size - off;
782 /* if we extend to eof, continue to block boundary */
783 if (off + len == src->i_size)
784 len = ((src->i_size + bs-1) & ~(bs-1))
785 - off;
786
787 /* verify the end result is block aligned */
788 if ((off & (bs-1)) ||
789 ((off + len) & (bs-1)))
790 goto out_unlock;
791
792 /* do any pending delalloc/csum calc on src, one way or
793 another, and lock file content */
794 while (1) {
795 struct btrfs_ordered_extent *ordered;
796 lock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
797 ordered = btrfs_lookup_first_ordered_extent(inode, off+len);
798 if (BTRFS_I(src)->delalloc_bytes == 0 && !ordered)
799 break;
800 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
801 if (ordered)
802 btrfs_put_ordered_extent(ordered);
803 btrfs_wait_ordered_range(src, off, off+len);
804 }
805
806 trans = btrfs_start_transaction(root, 1);
807 BUG_ON(!trans);
808
809 /* punch hole in destination first */
810 btrfs_drop_extents(trans, root, inode, off, off + len,
811 off + len, 0, &hint_byte);
812
813 /* clone data */
814 key.objectid = src->i_ino;
815 key.type = BTRFS_EXTENT_DATA_KEY;
816 key.offset = 0;
817
818 while (1) {
819 /*
820 * note the key will change type as we walk through the
821 * tree.
822 */
823 ret = btrfs_search_slot(trans, root, &key, path, 0, 0);
824 if (ret < 0)
825 goto out;
826
827 nritems = btrfs_header_nritems(path->nodes[0]);
828 if (path->slots[0] >= nritems) {
829 ret = btrfs_next_leaf(root, path);
830 if (ret < 0)
831 goto out;
832 if (ret > 0)
833 break;
834 nritems = btrfs_header_nritems(path->nodes[0]);
835 }
836 leaf = path->nodes[0];
837 slot = path->slots[0];
838
839 btrfs_item_key_to_cpu(leaf, &key, slot);
840 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
841 key.objectid != src->i_ino)
842 break;
843
844 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
845 struct btrfs_file_extent_item *extent;
846 int type;
847 u32 size;
848 struct btrfs_key new_key;
849 u64 disko = 0, diskl = 0;
850 u64 datao = 0, datal = 0;
851 u8 comp;
852
853 size = btrfs_item_size_nr(leaf, slot);
854 read_extent_buffer(leaf, buf,
855 btrfs_item_ptr_offset(leaf, slot),
856 size);
857
858 extent = btrfs_item_ptr(leaf, slot,
859 struct btrfs_file_extent_item);
860 comp = btrfs_file_extent_compression(leaf, extent);
861 type = btrfs_file_extent_type(leaf, extent);
862 if (type == BTRFS_FILE_EXTENT_REG) {
863 disko = btrfs_file_extent_disk_bytenr(leaf,
864 extent);
865 diskl = btrfs_file_extent_disk_num_bytes(leaf,
866 extent);
867 datao = btrfs_file_extent_offset(leaf, extent);
868 datal = btrfs_file_extent_num_bytes(leaf,
869 extent);
870 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
871 /* take upper bound, may be compressed */
872 datal = btrfs_file_extent_ram_bytes(leaf,
873 extent);
874 }
875 btrfs_release_path(root, path);
876
877 if (key.offset + datal < off ||
878 key.offset >= off+len)
879 goto next;
880
881 memcpy(&new_key, &key, sizeof(new_key));
882 new_key.objectid = inode->i_ino;
883 new_key.offset = key.offset + destoff - off;
884
885 if (type == BTRFS_FILE_EXTENT_REG) {
886 ret = btrfs_insert_empty_item(trans, root, path,
887 &new_key, size);
888 if (ret)
889 goto out;
890
891 leaf = path->nodes[0];
892 slot = path->slots[0];
893 write_extent_buffer(leaf, buf,
894 btrfs_item_ptr_offset(leaf, slot),
895 size);
896
897 extent = btrfs_item_ptr(leaf, slot,
898 struct btrfs_file_extent_item);
899
900 if (off > key.offset) {
901 datao += off - key.offset;
902 datal -= off - key.offset;
903 }
904 if (key.offset + datao + datal + key.offset >
905 off + len)
906 datal = off + len - key.offset - datao;
907 /* disko == 0 means it's a hole */
908 if (!disko)
909 datao = 0;
910
911 btrfs_set_file_extent_offset(leaf, extent,
912 datao);
913 btrfs_set_file_extent_num_bytes(leaf, extent,
914 datal);
915 if (disko) {
916 inode_add_bytes(inode, datal);
917 ret = btrfs_inc_extent_ref(trans, root,
918 disko, diskl, leaf->start,
919 root->root_key.objectid,
920 trans->transid,
921 inode->i_ino);
922 BUG_ON(ret);
923 }
924 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
925 u64 skip = 0;
926 u64 trim = 0;
927 if (off > key.offset) {
928 skip = off - key.offset;
929 new_key.offset += skip;
930 }
931
932 if (key.offset + datal > off+len)
933 trim = key.offset + datal - (off+len);
934
935 if (comp && (skip || trim)) {
936 ret = -EINVAL;
937 goto out;
938 }
939 size -= skip + trim;
940 datal -= skip + trim;
941 ret = btrfs_insert_empty_item(trans, root, path,
942 &new_key, size);
943 if (ret)
944 goto out;
945
946 if (skip) {
947 u32 start =
948 btrfs_file_extent_calc_inline_size(0);
949 memmove(buf+start, buf+start+skip,
950 datal);
951 }
952
953 leaf = path->nodes[0];
954 slot = path->slots[0];
955 write_extent_buffer(leaf, buf,
956 btrfs_item_ptr_offset(leaf, slot),
957 size);
958 inode_add_bytes(inode, datal);
959 }
960
961 btrfs_mark_buffer_dirty(leaf);
962 }
963
964 next:
965 btrfs_release_path(root, path);
966 key.offset++;
967 }
968 ret = 0;
969 out:
970 btrfs_release_path(root, path);
971 if (ret == 0) {
972 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
973 if (destoff + olen > inode->i_size)
974 btrfs_i_size_write(inode, destoff + olen);
975 BTRFS_I(inode)->flags = BTRFS_I(src)->flags;
976 ret = btrfs_update_inode(trans, root, inode);
977 }
978 btrfs_end_transaction(trans, root);
979 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
980 if (ret)
981 vmtruncate(inode, 0);
982 out_unlock:
983 mutex_unlock(&src->i_mutex);
984 mutex_unlock(&inode->i_mutex);
985 vfree(buf);
986 btrfs_free_path(path);
987 out_fput:
988 fput(src_file);
989 out_drop_write:
990 mnt_drop_write(file->f_path.mnt);
991 return ret;
992 }
993
994 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
995 {
996 struct btrfs_ioctl_clone_range_args args;
997
998 if (copy_from_user(&args, argp, sizeof(args)))
999 return -EFAULT;
1000 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
1001 args.src_length, args.dest_offset);
1002 }
1003
1004 /*
1005 * there are many ways the trans_start and trans_end ioctls can lead
1006 * to deadlocks. They should only be used by applications that
1007 * basically own the machine, and have a very in depth understanding
1008 * of all the possible deadlocks and enospc problems.
1009 */
1010 static long btrfs_ioctl_trans_start(struct file *file)
1011 {
1012 struct inode *inode = fdentry(file)->d_inode;
1013 struct btrfs_root *root = BTRFS_I(inode)->root;
1014 struct btrfs_trans_handle *trans;
1015 int ret = 0;
1016
1017 if (!capable(CAP_SYS_ADMIN))
1018 return -EPERM;
1019
1020 if (file->private_data) {
1021 ret = -EINPROGRESS;
1022 goto out;
1023 }
1024
1025 ret = mnt_want_write(file->f_path.mnt);
1026 if (ret)
1027 goto out;
1028
1029 mutex_lock(&root->fs_info->trans_mutex);
1030 root->fs_info->open_ioctl_trans++;
1031 mutex_unlock(&root->fs_info->trans_mutex);
1032
1033 trans = btrfs_start_ioctl_transaction(root, 0);
1034 if (trans)
1035 file->private_data = trans;
1036 else
1037 ret = -ENOMEM;
1038 /*printk(KERN_INFO "btrfs_ioctl_trans_start on %p\n", file);*/
1039 out:
1040 return ret;
1041 }
1042
1043 /*
1044 * there are many ways the trans_start and trans_end ioctls can lead
1045 * to deadlocks. They should only be used by applications that
1046 * basically own the machine, and have a very in depth understanding
1047 * of all the possible deadlocks and enospc problems.
1048 */
1049 long btrfs_ioctl_trans_end(struct file *file)
1050 {
1051 struct inode *inode = fdentry(file)->d_inode;
1052 struct btrfs_root *root = BTRFS_I(inode)->root;
1053 struct btrfs_trans_handle *trans;
1054 int ret = 0;
1055
1056 trans = file->private_data;
1057 if (!trans) {
1058 ret = -EINVAL;
1059 goto out;
1060 }
1061 btrfs_end_transaction(trans, root);
1062 file->private_data = NULL;
1063
1064 mutex_lock(&root->fs_info->trans_mutex);
1065 root->fs_info->open_ioctl_trans--;
1066 mutex_unlock(&root->fs_info->trans_mutex);
1067
1068 mnt_drop_write(file->f_path.mnt);
1069
1070 out:
1071 return ret;
1072 }
1073
1074 long btrfs_ioctl(struct file *file, unsigned int
1075 cmd, unsigned long arg)
1076 {
1077 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
1078 void __user *argp = (void __user *)arg;
1079
1080 switch (cmd) {
1081 case BTRFS_IOC_SNAP_CREATE:
1082 return btrfs_ioctl_snap_create(file, argp, 0);
1083 case BTRFS_IOC_SUBVOL_CREATE:
1084 return btrfs_ioctl_snap_create(file, argp, 1);
1085 case BTRFS_IOC_DEFRAG:
1086 return btrfs_ioctl_defrag(file);
1087 case BTRFS_IOC_RESIZE:
1088 return btrfs_ioctl_resize(root, argp);
1089 case BTRFS_IOC_ADD_DEV:
1090 return btrfs_ioctl_add_dev(root, argp);
1091 case BTRFS_IOC_RM_DEV:
1092 return btrfs_ioctl_rm_dev(root, argp);
1093 case BTRFS_IOC_BALANCE:
1094 return btrfs_balance(root->fs_info->dev_root);
1095 case BTRFS_IOC_CLONE:
1096 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
1097 case BTRFS_IOC_CLONE_RANGE:
1098 return btrfs_ioctl_clone_range(file, argp);
1099 case BTRFS_IOC_TRANS_START:
1100 return btrfs_ioctl_trans_start(file);
1101 case BTRFS_IOC_TRANS_END:
1102 return btrfs_ioctl_trans_end(file);
1103 case BTRFS_IOC_SYNC:
1104 btrfs_sync_fs(file->f_dentry->d_sb, 1);
1105 return 0;
1106 }
1107
1108 return -ENOTTY;
1109 }
Support for the Chinese Samsung S3C2440 based development boards