projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / fs / btrfs / ioctl.c
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/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>
44 #include <linux/uuid.h>
45 #include "compat.h"
46 #include "ctree.h"
47 #include "disk-io.h"
48 #include "transaction.h"
49 #include "btrfs_inode.h"
50 #include "ioctl.h"
51 #include "print-tree.h"
52 #include "volumes.h"
53 #include "locking.h"
54 #include "inode-map.h"
55 #include "backref.h"
56 #include "rcu-string.h"
57 #include "send.h"
58
59 /* Mask out flags that are inappropriate for the given type of inode. */
60 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
61 {
62 if (S_ISDIR(mode))
63 return flags;
64 else if (S_ISREG(mode))
65 return flags & ~FS_DIRSYNC_FL;
66 else
67 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
68 }
69
70 /*
71 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
72 */
73 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
74 {
75 unsigned int iflags = 0;
76
77 if (flags & BTRFS_INODE_SYNC)
78 iflags |= FS_SYNC_FL;
79 if (flags & BTRFS_INODE_IMMUTABLE)
80 iflags |= FS_IMMUTABLE_FL;
81 if (flags & BTRFS_INODE_APPEND)
82 iflags |= FS_APPEND_FL;
83 if (flags & BTRFS_INODE_NODUMP)
84 iflags |= FS_NODUMP_FL;
85 if (flags & BTRFS_INODE_NOATIME)
86 iflags |= FS_NOATIME_FL;
87 if (flags & BTRFS_INODE_DIRSYNC)
88 iflags |= FS_DIRSYNC_FL;
89 if (flags & BTRFS_INODE_NODATACOW)
90 iflags |= FS_NOCOW_FL;
91
92 if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
93 iflags |= FS_COMPR_FL;
94 else if (flags & BTRFS_INODE_NOCOMPRESS)
95 iflags |= FS_NOCOMP_FL;
96
97 return iflags;
98 }
99
100 /*
101 * Update inode->i_flags based on the btrfs internal flags.
102 */
103 void btrfs_update_iflags(struct inode *inode)
104 {
105 struct btrfs_inode *ip = BTRFS_I(inode);
106
107 inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
108
109 if (ip->flags & BTRFS_INODE_SYNC)
110 inode->i_flags |= S_SYNC;
111 if (ip->flags & BTRFS_INODE_IMMUTABLE)
112 inode->i_flags |= S_IMMUTABLE;
113 if (ip->flags & BTRFS_INODE_APPEND)
114 inode->i_flags |= S_APPEND;
115 if (ip->flags & BTRFS_INODE_NOATIME)
116 inode->i_flags |= S_NOATIME;
117 if (ip->flags & BTRFS_INODE_DIRSYNC)
118 inode->i_flags |= S_DIRSYNC;
119 }
120
121 /*
122 * Inherit flags from the parent inode.
123 *
124 * Currently only the compression flags and the cow flags are inherited.
125 */
126 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
127 {
128 unsigned int flags;
129
130 if (!dir)
131 return;
132
133 flags = BTRFS_I(dir)->flags;
134
135 if (flags & BTRFS_INODE_NOCOMPRESS) {
136 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
137 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
138 } else if (flags & BTRFS_INODE_COMPRESS) {
139 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
140 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
141 }
142
143 if (flags & BTRFS_INODE_NODATACOW)
144 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
145
146 btrfs_update_iflags(inode);
147 }
148
149 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
150 {
151 struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode);
152 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
153
154 if (copy_to_user(arg, &flags, sizeof(flags)))
155 return -EFAULT;
156 return 0;
157 }
158
159 static int check_flags(unsigned int flags)
160 {
161 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
162 FS_NOATIME_FL | FS_NODUMP_FL | \
163 FS_SYNC_FL | FS_DIRSYNC_FL | \
164 FS_NOCOMP_FL | FS_COMPR_FL |
165 FS_NOCOW_FL))
166 return -EOPNOTSUPP;
167
168 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
169 return -EINVAL;
170
171 return 0;
172 }
173
174 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
175 {
176 struct inode *inode = file->f_path.dentry->d_inode;
177 struct btrfs_inode *ip = BTRFS_I(inode);
178 struct btrfs_root *root = ip->root;
179 struct btrfs_trans_handle *trans;
180 unsigned int flags, oldflags;
181 int ret;
182 u64 ip_oldflags;
183 unsigned int i_oldflags;
184
185 if (btrfs_root_readonly(root))
186 return -EROFS;
187
188 if (copy_from_user(&flags, arg, sizeof(flags)))
189 return -EFAULT;
190
191 ret = check_flags(flags);
192 if (ret)
193 return ret;
194
195 if (!inode_owner_or_capable(inode))
196 return -EACCES;
197
198 ret = mnt_want_write_file(file);
199 if (ret)
200 return ret;
201
202 mutex_lock(&inode->i_mutex);
203
204 ip_oldflags = ip->flags;
205 i_oldflags = inode->i_flags;
206
207 flags = btrfs_mask_flags(inode->i_mode, flags);
208 oldflags = btrfs_flags_to_ioctl(ip->flags);
209 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
210 if (!capable(CAP_LINUX_IMMUTABLE)) {
211 ret = -EPERM;
212 goto out_unlock;
213 }
214 }
215
216 if (flags & FS_SYNC_FL)
217 ip->flags |= BTRFS_INODE_SYNC;
218 else
219 ip->flags &= ~BTRFS_INODE_SYNC;
220 if (flags & FS_IMMUTABLE_FL)
221 ip->flags |= BTRFS_INODE_IMMUTABLE;
222 else
223 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
224 if (flags & FS_APPEND_FL)
225 ip->flags |= BTRFS_INODE_APPEND;
226 else
227 ip->flags &= ~BTRFS_INODE_APPEND;
228 if (flags & FS_NODUMP_FL)
229 ip->flags |= BTRFS_INODE_NODUMP;
230 else
231 ip->flags &= ~BTRFS_INODE_NODUMP;
232 if (flags & FS_NOATIME_FL)
233 ip->flags |= BTRFS_INODE_NOATIME;
234 else
235 ip->flags &= ~BTRFS_INODE_NOATIME;
236 if (flags & FS_DIRSYNC_FL)
237 ip->flags |= BTRFS_INODE_DIRSYNC;
238 else
239 ip->flags &= ~BTRFS_INODE_DIRSYNC;
240 if (flags & FS_NOCOW_FL)
241 ip->flags |= BTRFS_INODE_NODATACOW;
242 else
243 ip->flags &= ~BTRFS_INODE_NODATACOW;
244
245 /*
246 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
247 * flag may be changed automatically if compression code won't make
248 * things smaller.
249 */
250 if (flags & FS_NOCOMP_FL) {
251 ip->flags &= ~BTRFS_INODE_COMPRESS;
252 ip->flags |= BTRFS_INODE_NOCOMPRESS;
253 } else if (flags & FS_COMPR_FL) {
254 ip->flags |= BTRFS_INODE_COMPRESS;
255 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
256 } else {
257 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
258 }
259
260 trans = btrfs_start_transaction(root, 1);
261 if (IS_ERR(trans)) {
262 ret = PTR_ERR(trans);
263 goto out_drop;
264 }
265
266 btrfs_update_iflags(inode);
267 inode_inc_iversion(inode);
268 inode->i_ctime = CURRENT_TIME;
269 ret = btrfs_update_inode(trans, root, inode);
270
271 btrfs_end_transaction(trans, root);
272 out_drop:
273 if (ret) {
274 ip->flags = ip_oldflags;
275 inode->i_flags = i_oldflags;
276 }
277
278 out_unlock:
279 mutex_unlock(&inode->i_mutex);
280 mnt_drop_write_file(file);
281 return ret;
282 }
283
284 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
285 {
286 struct inode *inode = file->f_path.dentry->d_inode;
287
288 return put_user(inode->i_generation, arg);
289 }
290
291 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
292 {
293 struct btrfs_fs_info *fs_info = btrfs_sb(fdentry(file)->d_sb);
294 struct btrfs_device *device;
295 struct request_queue *q;
296 struct fstrim_range range;
297 u64 minlen = ULLONG_MAX;
298 u64 num_devices = 0;
299 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
300 int ret;
301
302 if (!capable(CAP_SYS_ADMIN))
303 return -EPERM;
304
305 rcu_read_lock();
306 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
307 dev_list) {
308 if (!device->bdev)
309 continue;
310 q = bdev_get_queue(device->bdev);
311 if (blk_queue_discard(q)) {
312 num_devices++;
313 minlen = min((u64)q->limits.discard_granularity,
314 minlen);
315 }
316 }
317 rcu_read_unlock();
318
319 if (!num_devices)
320 return -EOPNOTSUPP;
321 if (copy_from_user(&range, arg, sizeof(range)))
322 return -EFAULT;
323 if (range.start > total_bytes)
324 return -EINVAL;
325
326 range.len = min(range.len, total_bytes - range.start);
327 range.minlen = max(range.minlen, minlen);
328 ret = btrfs_trim_fs(fs_info->tree_root, &range);
329 if (ret < 0)
330 return ret;
331
332 if (copy_to_user(arg, &range, sizeof(range)))
333 return -EFAULT;
334
335 return 0;
336 }
337
338 static noinline int create_subvol(struct btrfs_root *root,
339 struct dentry *dentry,
340 char *name, int namelen,
341 u64 *async_transid,
342 struct btrfs_qgroup_inherit **inherit)
343 {
344 struct btrfs_trans_handle *trans;
345 struct btrfs_key key;
346 struct btrfs_root_item root_item;
347 struct btrfs_inode_item *inode_item;
348 struct extent_buffer *leaf;
349 struct btrfs_root *new_root;
350 struct dentry *parent = dentry->d_parent;
351 struct inode *dir;
352 struct timespec cur_time = CURRENT_TIME;
353 int ret;
354 int err;
355 u64 objectid;
356 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
357 u64 index = 0;
358 uuid_le new_uuid;
359
360 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
361 if (ret)
362 return ret;
363
364 dir = parent->d_inode;
365
366 /*
367 * 1 - inode item
368 * 2 - refs
369 * 1 - root item
370 * 2 - dir items
371 */
372 trans = btrfs_start_transaction(root, 6);
373 if (IS_ERR(trans))
374 return PTR_ERR(trans);
375
376 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid,
377 inherit ? *inherit : NULL);
378 if (ret)
379 goto fail;
380
381 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
382 0, objectid, NULL, 0, 0, 0);
383 if (IS_ERR(leaf)) {
384 ret = PTR_ERR(leaf);
385 goto fail;
386 }
387
388 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
389 btrfs_set_header_bytenr(leaf, leaf->start);
390 btrfs_set_header_generation(leaf, trans->transid);
391 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
392 btrfs_set_header_owner(leaf, objectid);
393
394 write_extent_buffer(leaf, root->fs_info->fsid,
395 (unsigned long)btrfs_header_fsid(leaf),
396 BTRFS_FSID_SIZE);
397 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
398 (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
399 BTRFS_UUID_SIZE);
400 btrfs_mark_buffer_dirty(leaf);
401
402 memset(&root_item, 0, sizeof(root_item));
403
404 inode_item = &root_item.inode;
405 inode_item->generation = cpu_to_le64(1);
406 inode_item->size = cpu_to_le64(3);
407 inode_item->nlink = cpu_to_le32(1);
408 inode_item->nbytes = cpu_to_le64(root->leafsize);
409 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
410
411 root_item.flags = 0;
412 root_item.byte_limit = 0;
413 inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
414
415 btrfs_set_root_bytenr(&root_item, leaf->start);
416 btrfs_set_root_generation(&root_item, trans->transid);
417 btrfs_set_root_level(&root_item, 0);
418 btrfs_set_root_refs(&root_item, 1);
419 btrfs_set_root_used(&root_item, leaf->len);
420 btrfs_set_root_last_snapshot(&root_item, 0);
421
422 btrfs_set_root_generation_v2(&root_item,
423 btrfs_root_generation(&root_item));
424 uuid_le_gen(&new_uuid);
425 memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
426 root_item.otime.sec = cpu_to_le64(cur_time.tv_sec);
427 root_item.otime.nsec = cpu_to_le32(cur_time.tv_nsec);
428 root_item.ctime = root_item.otime;
429 btrfs_set_root_ctransid(&root_item, trans->transid);
430 btrfs_set_root_otransid(&root_item, trans->transid);
431
432 btrfs_tree_unlock(leaf);
433 free_extent_buffer(leaf);
434 leaf = NULL;
435
436 btrfs_set_root_dirid(&root_item, new_dirid);
437
438 key.objectid = objectid;
439 key.offset = 0;
440 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
441 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
442 &root_item);
443 if (ret)
444 goto fail;
445
446 key.offset = (u64)-1;
447 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
448 if (IS_ERR(new_root)) {
449 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
450 ret = PTR_ERR(new_root);
451 goto fail;
452 }
453
454 btrfs_record_root_in_trans(trans, new_root);
455
456 ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
457 if (ret) {
458 /* We potentially lose an unused inode item here */
459 btrfs_abort_transaction(trans, root, ret);
460 goto fail;
461 }
462
463 /*
464 * insert the directory item
465 */
466 ret = btrfs_set_inode_index(dir, &index);
467 if (ret) {
468 btrfs_abort_transaction(trans, root, ret);
469 goto fail;
470 }
471
472 ret = btrfs_insert_dir_item(trans, root,
473 name, namelen, dir, &key,
474 BTRFS_FT_DIR, index);
475 if (ret) {
476 btrfs_abort_transaction(trans, root, ret);
477 goto fail;
478 }
479
480 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
481 ret = btrfs_update_inode(trans, root, dir);
482 BUG_ON(ret);
483
484 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
485 objectid, root->root_key.objectid,
486 btrfs_ino(dir), index, name, namelen);
487
488 BUG_ON(ret);
489
490 d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
491 fail:
492 if (async_transid) {
493 *async_transid = trans->transid;
494 err = btrfs_commit_transaction_async(trans, root, 1);
495 } else {
496 err = btrfs_commit_transaction(trans, root);
497 }
498 if (err && !ret)
499 ret = err;
500 return ret;
501 }
502
503 static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
504 char *name, int namelen, u64 *async_transid,
505 bool readonly, struct btrfs_qgroup_inherit **inherit)
506 {
507 struct inode *inode;
508 struct btrfs_pending_snapshot *pending_snapshot;
509 struct btrfs_trans_handle *trans;
510 int ret;
511
512 if (!root->ref_cows)
513 return -EINVAL;
514
515 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
516 if (!pending_snapshot)
517 return -ENOMEM;
518
519 btrfs_init_block_rsv(&pending_snapshot->block_rsv);
520 pending_snapshot->dentry = dentry;
521 pending_snapshot->root = root;
522 pending_snapshot->readonly = readonly;
523 if (inherit) {
524 pending_snapshot->inherit = *inherit;
525 *inherit = NULL; /* take responsibility to free it */
526 }
527
528 trans = btrfs_start_transaction(root->fs_info->extent_root, 5);
529 if (IS_ERR(trans)) {
530 ret = PTR_ERR(trans);
531 goto fail;
532 }
533
534 ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
535 BUG_ON(ret);
536
537 spin_lock(&root->fs_info->trans_lock);
538 list_add(&pending_snapshot->list,
539 &trans->transaction->pending_snapshots);
540 spin_unlock(&root->fs_info->trans_lock);
541 if (async_transid) {
542 *async_transid = trans->transid;
543 ret = btrfs_commit_transaction_async(trans,
544 root->fs_info->extent_root, 1);
545 } else {
546 ret = btrfs_commit_transaction(trans,
547 root->fs_info->extent_root);
548 }
549 BUG_ON(ret);
550
551 ret = pending_snapshot->error;
552 if (ret)
553 goto fail;
554
555 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
556 if (ret)
557 goto fail;
558
559 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
560 if (IS_ERR(inode)) {
561 ret = PTR_ERR(inode);
562 goto fail;
563 }
564 BUG_ON(!inode);
565 d_instantiate(dentry, inode);
566 ret = 0;
567 fail:
568 kfree(pending_snapshot);
569 return ret;
570 }
571
572 /* copy of check_sticky in fs/namei.c()
573 * It's inline, so penalty for filesystems that don't use sticky bit is
574 * minimal.
575 */
576 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
577 {
578 kuid_t fsuid = current_fsuid();
579
580 if (!(dir->i_mode & S_ISVTX))
581 return 0;
582 if (uid_eq(inode->i_uid, fsuid))
583 return 0;
584 if (uid_eq(dir->i_uid, fsuid))
585 return 0;
586 return !capable(CAP_FOWNER);
587 }
588
589 /* copy of may_delete in fs/namei.c()
590 * Check whether we can remove a link victim from directory dir, check
591 * whether the type of victim is right.
592 * 1. We can't do it if dir is read-only (done in permission())
593 * 2. We should have write and exec permissions on dir
594 * 3. We can't remove anything from append-only dir
595 * 4. We can't do anything with immutable dir (done in permission())
596 * 5. If the sticky bit on dir is set we should either
597 * a. be owner of dir, or
598 * b. be owner of victim, or
599 * c. have CAP_FOWNER capability
600 * 6. If the victim is append-only or immutable we can't do antyhing with
601 * links pointing to it.
602 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
603 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
604 * 9. We can't remove a root or mountpoint.
605 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
606 * nfs_async_unlink().
607 */
608
609 static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
610 {
611 int error;
612
613 if (!victim->d_inode)
614 return -ENOENT;
615
616 BUG_ON(victim->d_parent->d_inode != dir);
617 audit_inode_child(victim, dir);
618
619 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
620 if (error)
621 return error;
622 if (IS_APPEND(dir))
623 return -EPERM;
624 if (btrfs_check_sticky(dir, victim->d_inode)||
625 IS_APPEND(victim->d_inode)||
626 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
627 return -EPERM;
628 if (isdir) {
629 if (!S_ISDIR(victim->d_inode->i_mode))
630 return -ENOTDIR;
631 if (IS_ROOT(victim))
632 return -EBUSY;
633 } else if (S_ISDIR(victim->d_inode->i_mode))
634 return -EISDIR;
635 if (IS_DEADDIR(dir))
636 return -ENOENT;
637 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
638 return -EBUSY;
639 return 0;
640 }
641
642 /* copy of may_create in fs/namei.c() */
643 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
644 {
645 if (child->d_inode)
646 return -EEXIST;
647 if (IS_DEADDIR(dir))
648 return -ENOENT;
649 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
650 }
651
652 /*
653 * Create a new subvolume below @parent. This is largely modeled after
654 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
655 * inside this filesystem so it's quite a bit simpler.
656 */
657 static noinline int btrfs_mksubvol(struct path *parent,
658 char *name, int namelen,
659 struct btrfs_root *snap_src,
660 u64 *async_transid, bool readonly,
661 struct btrfs_qgroup_inherit **inherit)
662 {
663 struct inode *dir = parent->dentry->d_inode;
664 struct dentry *dentry;
665 int error;
666
667 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
668
669 dentry = lookup_one_len(name, parent->dentry, namelen);
670 error = PTR_ERR(dentry);
671 if (IS_ERR(dentry))
672 goto out_unlock;
673
674 error = -EEXIST;
675 if (dentry->d_inode)
676 goto out_dput;
677
678 error = btrfs_may_create(dir, dentry);
679 if (error)
680 goto out_dput;
681
682 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
683
684 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
685 goto out_up_read;
686
687 if (snap_src) {
688 error = create_snapshot(snap_src, dentry, name, namelen,
689 async_transid, readonly, inherit);
690 } else {
691 error = create_subvol(BTRFS_I(dir)->root, dentry,
692 name, namelen, async_transid, inherit);
693 }
694 if (!error)
695 fsnotify_mkdir(dir, dentry);
696 out_up_read:
697 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
698 out_dput:
699 dput(dentry);
700 out_unlock:
701 mutex_unlock(&dir->i_mutex);
702 return error;
703 }
704
705 /*
706 * When we're defragging a range, we don't want to kick it off again
707 * if it is really just waiting for delalloc to send it down.
708 * If we find a nice big extent or delalloc range for the bytes in the
709 * file you want to defrag, we return 0 to let you know to skip this
710 * part of the file
711 */
712 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
713 {
714 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
715 struct extent_map *em = NULL;
716 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
717 u64 end;
718
719 read_lock(&em_tree->lock);
720 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
721 read_unlock(&em_tree->lock);
722
723 if (em) {
724 end = extent_map_end(em);
725 free_extent_map(em);
726 if (end - offset > thresh)
727 return 0;
728 }
729 /* if we already have a nice delalloc here, just stop */
730 thresh /= 2;
731 end = count_range_bits(io_tree, &offset, offset + thresh,
732 thresh, EXTENT_DELALLOC, 1);
733 if (end >= thresh)
734 return 0;
735 return 1;
736 }
737
738 /*
739 * helper function to walk through a file and find extents
740 * newer than a specific transid, and smaller than thresh.
741 *
742 * This is used by the defragging code to find new and small
743 * extents
744 */
745 static int find_new_extents(struct btrfs_root *root,
746 struct inode *inode, u64 newer_than,
747 u64 *off, int thresh)
748 {
749 struct btrfs_path *path;
750 struct btrfs_key min_key;
751 struct btrfs_key max_key;
752 struct extent_buffer *leaf;
753 struct btrfs_file_extent_item *extent;
754 int type;
755 int ret;
756 u64 ino = btrfs_ino(inode);
757
758 path = btrfs_alloc_path();
759 if (!path)
760 return -ENOMEM;
761
762 min_key.objectid = ino;
763 min_key.type = BTRFS_EXTENT_DATA_KEY;
764 min_key.offset = *off;
765
766 max_key.objectid = ino;
767 max_key.type = (u8)-1;
768 max_key.offset = (u64)-1;
769
770 path->keep_locks = 1;
771
772 while(1) {
773 ret = btrfs_search_forward(root, &min_key, &max_key,
774 path, 0, newer_than);
775 if (ret != 0)
776 goto none;
777 if (min_key.objectid != ino)
778 goto none;
779 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
780 goto none;
781
782 leaf = path->nodes[0];
783 extent = btrfs_item_ptr(leaf, path->slots[0],
784 struct btrfs_file_extent_item);
785
786 type = btrfs_file_extent_type(leaf, extent);
787 if (type == BTRFS_FILE_EXTENT_REG &&
788 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
789 check_defrag_in_cache(inode, min_key.offset, thresh)) {
790 *off = min_key.offset;
791 btrfs_free_path(path);
792 return 0;
793 }
794
795 if (min_key.offset == (u64)-1)
796 goto none;
797
798 min_key.offset++;
799 btrfs_release_path(path);
800 }
801 none:
802 btrfs_free_path(path);
803 return -ENOENT;
804 }
805
806 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
807 {
808 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
809 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
810 struct extent_map *em;
811 u64 len = PAGE_CACHE_SIZE;
812
813 /*
814 * hopefully we have this extent in the tree already, try without
815 * the full extent lock
816 */
817 read_lock(&em_tree->lock);
818 em = lookup_extent_mapping(em_tree, start, len);
819 read_unlock(&em_tree->lock);
820
821 if (!em) {
822 /* get the big lock and read metadata off disk */
823 lock_extent(io_tree, start, start + len - 1);
824 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
825 unlock_extent(io_tree, start, start + len - 1);
826
827 if (IS_ERR(em))
828 return NULL;
829 }
830
831 return em;
832 }
833
834 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
835 {
836 struct extent_map *next;
837 bool ret = true;
838
839 /* this is the last extent */
840 if (em->start + em->len >= i_size_read(inode))
841 return false;
842
843 next = defrag_lookup_extent(inode, em->start + em->len);
844 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
845 ret = false;
846
847 free_extent_map(next);
848 return ret;
849 }
850
851 static int should_defrag_range(struct inode *inode, u64 start, int thresh,
852 u64 *last_len, u64 *skip, u64 *defrag_end,
853 int compress)
854 {
855 struct extent_map *em;
856 int ret = 1;
857 bool next_mergeable = true;
858
859 /*
860 * make sure that once we start defragging an extent, we keep on
861 * defragging it
862 */
863 if (start < *defrag_end)
864 return 1;
865
866 *skip = 0;
867
868 em = defrag_lookup_extent(inode, start);
869 if (!em)
870 return 0;
871
872 /* this will cover holes, and inline extents */
873 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
874 ret = 0;
875 goto out;
876 }
877
878 next_mergeable = defrag_check_next_extent(inode, em);
879
880 /*
881 * we hit a real extent, if it is big or the next extent is not a
882 * real extent, don't bother defragging it
883 */
884 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
885 (em->len >= thresh || !next_mergeable))
886 ret = 0;
887 out:
888 /*
889 * last_len ends up being a counter of how many bytes we've defragged.
890 * every time we choose not to defrag an extent, we reset *last_len
891 * so that the next tiny extent will force a defrag.
892 *
893 * The end result of this is that tiny extents before a single big
894 * extent will force at least part of that big extent to be defragged.
895 */
896 if (ret) {
897 *defrag_end = extent_map_end(em);
898 } else {
899 *last_len = 0;
900 *skip = extent_map_end(em);
901 *defrag_end = 0;
902 }
903
904 free_extent_map(em);
905 return ret;
906 }
907
908 /*
909 * it doesn't do much good to defrag one or two pages
910 * at a time. This pulls in a nice chunk of pages
911 * to COW and defrag.
912 *
913 * It also makes sure the delalloc code has enough
914 * dirty data to avoid making new small extents as part
915 * of the defrag
916 *
917 * It's a good idea to start RA on this range
918 * before calling this.
919 */
920 static int cluster_pages_for_defrag(struct inode *inode,
921 struct page **pages,
922 unsigned long start_index,
923 int num_pages)
924 {
925 unsigned long file_end;
926 u64 isize = i_size_read(inode);
927 u64 page_start;
928 u64 page_end;
929 u64 page_cnt;
930 int ret;
931 int i;
932 int i_done;
933 struct btrfs_ordered_extent *ordered;
934 struct extent_state *cached_state = NULL;
935 struct extent_io_tree *tree;
936 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
937
938 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
939 if (!isize || start_index > file_end)
940 return 0;
941
942 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
943
944 ret = btrfs_delalloc_reserve_space(inode,
945 page_cnt << PAGE_CACHE_SHIFT);
946 if (ret)
947 return ret;
948 i_done = 0;
949 tree = &BTRFS_I(inode)->io_tree;
950
951 /* step one, lock all the pages */
952 for (i = 0; i < page_cnt; i++) {
953 struct page *page;
954 again:
955 page = find_or_create_page(inode->i_mapping,
956 start_index + i, mask);
957 if (!page)
958 break;
959
960 page_start = page_offset(page);
961 page_end = page_start + PAGE_CACHE_SIZE - 1;
962 while (1) {
963 lock_extent(tree, page_start, page_end);
964 ordered = btrfs_lookup_ordered_extent(inode,
965 page_start);
966 unlock_extent(tree, page_start, page_end);
967 if (!ordered)
968 break;
969
970 unlock_page(page);
971 btrfs_start_ordered_extent(inode, ordered, 1);
972 btrfs_put_ordered_extent(ordered);
973 lock_page(page);
974 /*
975 * we unlocked the page above, so we need check if
976 * it was released or not.
977 */
978 if (page->mapping != inode->i_mapping) {
979 unlock_page(page);
980 page_cache_release(page);
981 goto again;
982 }
983 }
984
985 if (!PageUptodate(page)) {
986 btrfs_readpage(NULL, page);
987 lock_page(page);
988 if (!PageUptodate(page)) {
989 unlock_page(page);
990 page_cache_release(page);
991 ret = -EIO;
992 break;
993 }
994 }
995
996 if (page->mapping != inode->i_mapping) {
997 unlock_page(page);
998 page_cache_release(page);
999 goto again;
1000 }
1001
1002 pages[i] = page;
1003 i_done++;
1004 }
1005 if (!i_done || ret)
1006 goto out;
1007
1008 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1009 goto out;
1010
1011 /*
1012 * so now we have a nice long stream of locked
1013 * and up to date pages, lets wait on them
1014 */
1015 for (i = 0; i < i_done; i++)
1016 wait_on_page_writeback(pages[i]);
1017
1018 page_start = page_offset(pages[0]);
1019 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1020
1021 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1022 page_start, page_end - 1, 0, &cached_state);
1023 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1024 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1025 EXTENT_DO_ACCOUNTING, 0, 0, &cached_state,
1026 GFP_NOFS);
1027
1028 if (i_done != page_cnt) {
1029 spin_lock(&BTRFS_I(inode)->lock);
1030 BTRFS_I(inode)->outstanding_extents++;
1031 spin_unlock(&BTRFS_I(inode)->lock);
1032 btrfs_delalloc_release_space(inode,
1033 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1034 }
1035
1036
1037 btrfs_set_extent_delalloc(inode, page_start, page_end - 1,
1038 &cached_state);
1039
1040 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1041 page_start, page_end - 1, &cached_state,
1042 GFP_NOFS);
1043
1044 for (i = 0; i < i_done; i++) {
1045 clear_page_dirty_for_io(pages[i]);
1046 ClearPageChecked(pages[i]);
1047 set_page_extent_mapped(pages[i]);
1048 set_page_dirty(pages[i]);
1049 unlock_page(pages[i]);
1050 page_cache_release(pages[i]);
1051 }
1052 return i_done;
1053 out:
1054 for (i = 0; i < i_done; i++) {
1055 unlock_page(pages[i]);
1056 page_cache_release(pages[i]);
1057 }
1058 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1059 return ret;
1060
1061 }
1062
1063 int btrfs_defrag_file(struct inode *inode, struct file *file,
1064 struct btrfs_ioctl_defrag_range_args *range,
1065 u64 newer_than, unsigned long max_to_defrag)
1066 {
1067 struct btrfs_root *root = BTRFS_I(inode)->root;
1068 struct file_ra_state *ra = NULL;
1069 unsigned long last_index;
1070 u64 isize = i_size_read(inode);
1071 u64 last_len = 0;
1072 u64 skip = 0;
1073 u64 defrag_end = 0;
1074 u64 newer_off = range->start;
1075 unsigned long i;
1076 unsigned long ra_index = 0;
1077 int ret;
1078 int defrag_count = 0;
1079 int compress_type = BTRFS_COMPRESS_ZLIB;
1080 int extent_thresh = range->extent_thresh;
1081 int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1082 int cluster = max_cluster;
1083 u64 new_align = ~((u64)128 * 1024 - 1);
1084 struct page **pages = NULL;
1085
1086 if (extent_thresh == 0)
1087 extent_thresh = 256 * 1024;
1088
1089 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1090 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1091 return -EINVAL;
1092 if (range->compress_type)
1093 compress_type = range->compress_type;
1094 }
1095
1096 if (isize == 0)
1097 return 0;
1098
1099 /*
1100 * if we were not given a file, allocate a readahead
1101 * context
1102 */
1103 if (!file) {
1104 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1105 if (!ra)
1106 return -ENOMEM;
1107 file_ra_state_init(ra, inode->i_mapping);
1108 } else {
1109 ra = &file->f_ra;
1110 }
1111
1112 pages = kmalloc(sizeof(struct page *) * max_cluster,
1113 GFP_NOFS);
1114 if (!pages) {
1115 ret = -ENOMEM;
1116 goto out_ra;
1117 }
1118
1119 /* find the last page to defrag */
1120 if (range->start + range->len > range->start) {
1121 last_index = min_t(u64, isize - 1,
1122 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1123 } else {
1124 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1125 }
1126
1127 if (newer_than) {
1128 ret = find_new_extents(root, inode, newer_than,
1129 &newer_off, 64 * 1024);
1130 if (!ret) {
1131 range->start = newer_off;
1132 /*
1133 * we always align our defrag to help keep
1134 * the extents in the file evenly spaced
1135 */
1136 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1137 } else
1138 goto out_ra;
1139 } else {
1140 i = range->start >> PAGE_CACHE_SHIFT;
1141 }
1142 if (!max_to_defrag)
1143 max_to_defrag = last_index + 1;
1144
1145 /*
1146 * make writeback starts from i, so the defrag range can be
1147 * written sequentially.
1148 */
1149 if (i < inode->i_mapping->writeback_index)
1150 inode->i_mapping->writeback_index = i;
1151
1152 while (i <= last_index && defrag_count < max_to_defrag &&
1153 (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1154 PAGE_CACHE_SHIFT)) {
1155 /*
1156 * make sure we stop running if someone unmounts
1157 * the FS
1158 */
1159 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1160 break;
1161
1162 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1163 extent_thresh, &last_len, &skip,
1164 &defrag_end, range->flags &
1165 BTRFS_DEFRAG_RANGE_COMPRESS)) {
1166 unsigned long next;
1167 /*
1168 * the should_defrag function tells us how much to skip
1169 * bump our counter by the suggested amount
1170 */
1171 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1172 i = max(i + 1, next);
1173 continue;
1174 }
1175
1176 if (!newer_than) {
1177 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1178 PAGE_CACHE_SHIFT) - i;
1179 cluster = min(cluster, max_cluster);
1180 } else {
1181 cluster = max_cluster;
1182 }
1183
1184 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1185 BTRFS_I(inode)->force_compress = compress_type;
1186
1187 if (i + cluster > ra_index) {
1188 ra_index = max(i, ra_index);
1189 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1190 cluster);
1191 ra_index += max_cluster;
1192 }
1193
1194 mutex_lock(&inode->i_mutex);
1195 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1196 if (ret < 0) {
1197 mutex_unlock(&inode->i_mutex);
1198 goto out_ra;
1199 }
1200
1201 defrag_count += ret;
1202 balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret);
1203 mutex_unlock(&inode->i_mutex);
1204
1205 if (newer_than) {
1206 if (newer_off == (u64)-1)
1207 break;
1208
1209 if (ret > 0)
1210 i += ret;
1211
1212 newer_off = max(newer_off + 1,
1213 (u64)i << PAGE_CACHE_SHIFT);
1214
1215 ret = find_new_extents(root, inode,
1216 newer_than, &newer_off,
1217 64 * 1024);
1218 if (!ret) {
1219 range->start = newer_off;
1220 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1221 } else {
1222 break;
1223 }
1224 } else {
1225 if (ret > 0) {
1226 i += ret;
1227 last_len += ret << PAGE_CACHE_SHIFT;
1228 } else {
1229 i++;
1230 last_len = 0;
1231 }
1232 }
1233 }
1234
1235 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1236 filemap_flush(inode->i_mapping);
1237
1238 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1239 /* the filemap_flush will queue IO into the worker threads, but
1240 * we have to make sure the IO is actually started and that
1241 * ordered extents get created before we return
1242 */
1243 atomic_inc(&root->fs_info->async_submit_draining);
1244 while (atomic_read(&root->fs_info->nr_async_submits) ||
1245 atomic_read(&root->fs_info->async_delalloc_pages)) {
1246 wait_event(root->fs_info->async_submit_wait,
1247 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1248 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1249 }
1250 atomic_dec(&root->fs_info->async_submit_draining);
1251
1252 mutex_lock(&inode->i_mutex);
1253 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1254 mutex_unlock(&inode->i_mutex);
1255 }
1256
1257 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1258 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1259 }
1260
1261 ret = defrag_count;
1262
1263 out_ra:
1264 if (!file)
1265 kfree(ra);
1266 kfree(pages);
1267 return ret;
1268 }
1269
1270 static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
1271 void __user *arg)
1272 {
1273 u64 new_size;
1274 u64 old_size;
1275 u64 devid = 1;
1276 struct btrfs_ioctl_vol_args *vol_args;
1277 struct btrfs_trans_handle *trans;
1278 struct btrfs_device *device = NULL;
1279 char *sizestr;
1280 char *devstr = NULL;
1281 int ret = 0;
1282 int mod = 0;
1283
1284 if (root->fs_info->sb->s_flags & MS_RDONLY)
1285 return -EROFS;
1286
1287 if (!capable(CAP_SYS_ADMIN))
1288 return -EPERM;
1289
1290 mutex_lock(&root->fs_info->volume_mutex);
1291 if (root->fs_info->balance_ctl) {
1292 printk(KERN_INFO "btrfs: balance in progress\n");
1293 ret = -EINVAL;
1294 goto out;
1295 }
1296
1297 vol_args = memdup_user(arg, sizeof(*vol_args));
1298 if (IS_ERR(vol_args)) {
1299 ret = PTR_ERR(vol_args);
1300 goto out;
1301 }
1302
1303 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1304
1305 sizestr = vol_args->name;
1306 devstr = strchr(sizestr, ':');
1307 if (devstr) {
1308 char *end;
1309 sizestr = devstr + 1;
1310 *devstr = '\0';
1311 devstr = vol_args->name;
1312 devid = simple_strtoull(devstr, &end, 10);
1313 printk(KERN_INFO "btrfs: resizing devid %llu\n",
1314 (unsigned long long)devid);
1315 }
1316 device = btrfs_find_device(root, devid, NULL, NULL);
1317 if (!device) {
1318 printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
1319 (unsigned long long)devid);
1320 ret = -EINVAL;
1321 goto out_free;
1322 }
1323 if (device->fs_devices && device->fs_devices->seeding) {
1324 printk(KERN_INFO "btrfs: resizer unable to apply on "
1325 "seeding device %llu\n",
1326 (unsigned long long)devid);
1327 ret = -EINVAL;
1328 goto out_free;
1329 }
1330
1331 if (!strcmp(sizestr, "max"))
1332 new_size = device->bdev->bd_inode->i_size;
1333 else {
1334 if (sizestr[0] == '-') {
1335 mod = -1;
1336 sizestr++;
1337 } else if (sizestr[0] == '+') {
1338 mod = 1;
1339 sizestr++;
1340 }
1341 new_size = memparse(sizestr, NULL);
1342 if (new_size == 0) {
1343 ret = -EINVAL;
1344 goto out_free;
1345 }
1346 }
1347
1348 old_size = device->total_bytes;
1349
1350 if (mod < 0) {
1351 if (new_size > old_size) {
1352 ret = -EINVAL;
1353 goto out_free;
1354 }
1355 new_size = old_size - new_size;
1356 } else if (mod > 0) {
1357 new_size = old_size + new_size;
1358 }
1359
1360 if (new_size < 256 * 1024 * 1024) {
1361 ret = -EINVAL;
1362 goto out_free;
1363 }
1364 if (new_size > device->bdev->bd_inode->i_size) {
1365 ret = -EFBIG;
1366 goto out_free;
1367 }
1368
1369 do_div(new_size, root->sectorsize);
1370 new_size *= root->sectorsize;
1371
1372 printk_in_rcu(KERN_INFO "btrfs: new size for %s is %llu\n",
1373 rcu_str_deref(device->name),
1374 (unsigned long long)new_size);
1375
1376 if (new_size > old_size) {
1377 trans = btrfs_start_transaction(root, 0);
1378 if (IS_ERR(trans)) {
1379 ret = PTR_ERR(trans);
1380 goto out_free;
1381 }
1382 ret = btrfs_grow_device(trans, device, new_size);
1383 btrfs_commit_transaction(trans, root);
1384 } else if (new_size < old_size) {
1385 ret = btrfs_shrink_device(device, new_size);
1386 }
1387
1388 out_free:
1389 kfree(vol_args);
1390 out:
1391 mutex_unlock(&root->fs_info->volume_mutex);
1392 return ret;
1393 }
1394
1395 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1396 char *name, unsigned long fd, int subvol,
1397 u64 *transid, bool readonly,
1398 struct btrfs_qgroup_inherit **inherit)
1399 {
1400 int namelen;
1401 int ret = 0;
1402
1403 ret = mnt_want_write_file(file);
1404 if (ret)
1405 goto out;
1406
1407 namelen = strlen(name);
1408 if (strchr(name, '/')) {
1409 ret = -EINVAL;
1410 goto out_drop_write;
1411 }
1412
1413 if (name[0] == '.' &&
1414 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1415 ret = -EEXIST;
1416 goto out_drop_write;
1417 }
1418
1419 if (subvol) {
1420 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1421 NULL, transid, readonly, inherit);
1422 } else {
1423 struct fd src = fdget(fd);
1424 struct inode *src_inode;
1425 if (!src.file) {
1426 ret = -EINVAL;
1427 goto out_drop_write;
1428 }
1429
1430 src_inode = src.file->f_path.dentry->d_inode;
1431 if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
1432 printk(KERN_INFO "btrfs: Snapshot src from "
1433 "another FS\n");
1434 ret = -EINVAL;
1435 } else {
1436 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1437 BTRFS_I(src_inode)->root,
1438 transid, readonly, inherit);
1439 }
1440 fdput(src);
1441 }
1442 out_drop_write:
1443 mnt_drop_write_file(file);
1444 out:
1445 return ret;
1446 }
1447
1448 static noinline int btrfs_ioctl_snap_create(struct file *file,
1449 void __user *arg, int subvol)
1450 {
1451 struct btrfs_ioctl_vol_args *vol_args;
1452 int ret;
1453
1454 vol_args = memdup_user(arg, sizeof(*vol_args));
1455 if (IS_ERR(vol_args))
1456 return PTR_ERR(vol_args);
1457 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1458
1459 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1460 vol_args->fd, subvol,
1461 NULL, false, NULL);
1462
1463 kfree(vol_args);
1464 return ret;
1465 }
1466
1467 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1468 void __user *arg, int subvol)
1469 {
1470 struct btrfs_ioctl_vol_args_v2 *vol_args;
1471 int ret;
1472 u64 transid = 0;
1473 u64 *ptr = NULL;
1474 bool readonly = false;
1475 struct btrfs_qgroup_inherit *inherit = NULL;
1476
1477 vol_args = memdup_user(arg, sizeof(*vol_args));
1478 if (IS_ERR(vol_args))
1479 return PTR_ERR(vol_args);
1480 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1481
1482 if (vol_args->flags &
1483 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1484 BTRFS_SUBVOL_QGROUP_INHERIT)) {
1485 ret = -EOPNOTSUPP;
1486 goto out;
1487 }
1488
1489 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1490 ptr = &transid;
1491 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1492 readonly = true;
1493 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1494 if (vol_args->size > PAGE_CACHE_SIZE) {
1495 ret = -EINVAL;
1496 goto out;
1497 }
1498 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1499 if (IS_ERR(inherit)) {
1500 ret = PTR_ERR(inherit);
1501 goto out;
1502 }
1503 }
1504
1505 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1506 vol_args->fd, subvol, ptr,
1507 readonly, &inherit);
1508
1509 if (ret == 0 && ptr &&
1510 copy_to_user(arg +
1511 offsetof(struct btrfs_ioctl_vol_args_v2,
1512 transid), ptr, sizeof(*ptr)))
1513 ret = -EFAULT;
1514 out:
1515 kfree(vol_args);
1516 kfree(inherit);
1517 return ret;
1518 }
1519
1520 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1521 void __user *arg)
1522 {
1523 struct inode *inode = fdentry(file)->d_inode;
1524 struct btrfs_root *root = BTRFS_I(inode)->root;
1525 int ret = 0;
1526 u64 flags = 0;
1527
1528 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1529 return -EINVAL;
1530
1531 down_read(&root->fs_info->subvol_sem);
1532 if (btrfs_root_readonly(root))
1533 flags |= BTRFS_SUBVOL_RDONLY;
1534 up_read(&root->fs_info->subvol_sem);
1535
1536 if (copy_to_user(arg, &flags, sizeof(flags)))
1537 ret = -EFAULT;
1538
1539 return ret;
1540 }
1541
1542 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1543 void __user *arg)
1544 {
1545 struct inode *inode = fdentry(file)->d_inode;
1546 struct btrfs_root *root = BTRFS_I(inode)->root;
1547 struct btrfs_trans_handle *trans;
1548 u64 root_flags;
1549 u64 flags;
1550 int ret = 0;
1551
1552 ret = mnt_want_write_file(file);
1553 if (ret)
1554 goto out;
1555
1556 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1557 ret = -EINVAL;
1558 goto out_drop_write;
1559 }
1560
1561 if (copy_from_user(&flags, arg, sizeof(flags))) {
1562 ret = -EFAULT;
1563 goto out_drop_write;
1564 }
1565
1566 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1567 ret = -EINVAL;
1568 goto out_drop_write;
1569 }
1570
1571 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1572 ret = -EOPNOTSUPP;
1573 goto out_drop_write;
1574 }
1575
1576 if (!inode_owner_or_capable(inode)) {
1577 ret = -EACCES;
1578 goto out_drop_write;
1579 }
1580
1581 down_write(&root->fs_info->subvol_sem);
1582
1583 /* nothing to do */
1584 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1585 goto out_drop_sem;
1586
1587 root_flags = btrfs_root_flags(&root->root_item);
1588 if (flags & BTRFS_SUBVOL_RDONLY)
1589 btrfs_set_root_flags(&root->root_item,
1590 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1591 else
1592 btrfs_set_root_flags(&root->root_item,
1593 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1594
1595 trans = btrfs_start_transaction(root, 1);
1596 if (IS_ERR(trans)) {
1597 ret = PTR_ERR(trans);
1598 goto out_reset;
1599 }
1600
1601 ret = btrfs_update_root(trans, root->fs_info->tree_root,
1602 &root->root_key, &root->root_item);
1603
1604 btrfs_commit_transaction(trans, root);
1605 out_reset:
1606 if (ret)
1607 btrfs_set_root_flags(&root->root_item, root_flags);
1608 out_drop_sem:
1609 up_write(&root->fs_info->subvol_sem);
1610 out_drop_write:
1611 mnt_drop_write_file(file);
1612 out:
1613 return ret;
1614 }
1615
1616 /*
1617 * helper to check if the subvolume references other subvolumes
1618 */
1619 static noinline int may_destroy_subvol(struct btrfs_root *root)
1620 {
1621 struct btrfs_path *path;
1622 struct btrfs_key key;
1623 int ret;
1624
1625 path = btrfs_alloc_path();
1626 if (!path)
1627 return -ENOMEM;
1628
1629 key.objectid = root->root_key.objectid;
1630 key.type = BTRFS_ROOT_REF_KEY;
1631 key.offset = (u64)-1;
1632
1633 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1634 &key, path, 0, 0);
1635 if (ret < 0)
1636 goto out;
1637 BUG_ON(ret == 0);
1638
1639 ret = 0;
1640 if (path->slots[0] > 0) {
1641 path->slots[0]--;
1642 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1643 if (key.objectid == root->root_key.objectid &&
1644 key.type == BTRFS_ROOT_REF_KEY)
1645 ret = -ENOTEMPTY;
1646 }
1647 out:
1648 btrfs_free_path(path);
1649 return ret;
1650 }
1651
1652 static noinline int key_in_sk(struct btrfs_key *key,
1653 struct btrfs_ioctl_search_key *sk)
1654 {
1655 struct btrfs_key test;
1656 int ret;
1657
1658 test.objectid = sk->min_objectid;
1659 test.type = sk->min_type;
1660 test.offset = sk->min_offset;
1661
1662 ret = btrfs_comp_cpu_keys(key, &test);
1663 if (ret < 0)
1664 return 0;
1665
1666 test.objectid = sk->max_objectid;
1667 test.type = sk->max_type;
1668 test.offset = sk->max_offset;
1669
1670 ret = btrfs_comp_cpu_keys(key, &test);
1671 if (ret > 0)
1672 return 0;
1673 return 1;
1674 }
1675
1676 static noinline int copy_to_sk(struct btrfs_root *root,
1677 struct btrfs_path *path,
1678 struct btrfs_key *key,
1679 struct btrfs_ioctl_search_key *sk,
1680 char *buf,
1681 unsigned long *sk_offset,
1682 int *num_found)
1683 {
1684 u64 found_transid;
1685 struct extent_buffer *leaf;
1686 struct btrfs_ioctl_search_header sh;
1687 unsigned long item_off;
1688 unsigned long item_len;
1689 int nritems;
1690 int i;
1691 int slot;
1692 int ret = 0;
1693
1694 leaf = path->nodes[0];
1695 slot = path->slots[0];
1696 nritems = btrfs_header_nritems(leaf);
1697
1698 if (btrfs_header_generation(leaf) > sk->max_transid) {
1699 i = nritems;
1700 goto advance_key;
1701 }
1702 found_transid = btrfs_header_generation(leaf);
1703
1704 for (i = slot; i < nritems; i++) {
1705 item_off = btrfs_item_ptr_offset(leaf, i);
1706 item_len = btrfs_item_size_nr(leaf, i);
1707
1708 if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1709 item_len = 0;
1710
1711 if (sizeof(sh) + item_len + *sk_offset >
1712 BTRFS_SEARCH_ARGS_BUFSIZE) {
1713 ret = 1;
1714 goto overflow;
1715 }
1716
1717 btrfs_item_key_to_cpu(leaf, key, i);
1718 if (!key_in_sk(key, sk))
1719 continue;
1720
1721 sh.objectid = key->objectid;
1722 sh.offset = key->offset;
1723 sh.type = key->type;
1724 sh.len = item_len;
1725 sh.transid = found_transid;
1726
1727 /* copy search result header */
1728 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1729 *sk_offset += sizeof(sh);
1730
1731 if (item_len) {
1732 char *p = buf + *sk_offset;
1733 /* copy the item */
1734 read_extent_buffer(leaf, p,
1735 item_off, item_len);
1736 *sk_offset += item_len;
1737 }
1738 (*num_found)++;
1739
1740 if (*num_found >= sk->nr_items)
1741 break;
1742 }
1743 advance_key:
1744 ret = 0;
1745 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1746 key->offset++;
1747 else if (key->type < (u8)-1 && key->type < sk->max_type) {
1748 key->offset = 0;
1749 key->type++;
1750 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1751 key->offset = 0;
1752 key->type = 0;
1753 key->objectid++;
1754 } else
1755 ret = 1;
1756 overflow:
1757 return ret;
1758 }
1759
1760 static noinline int search_ioctl(struct inode *inode,
1761 struct btrfs_ioctl_search_args *args)
1762 {
1763 struct btrfs_root *root;
1764 struct btrfs_key key;
1765 struct btrfs_key max_key;
1766 struct btrfs_path *path;
1767 struct btrfs_ioctl_search_key *sk = &args->key;
1768 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1769 int ret;
1770 int num_found = 0;
1771 unsigned long sk_offset = 0;
1772
1773 path = btrfs_alloc_path();
1774 if (!path)
1775 return -ENOMEM;
1776
1777 if (sk->tree_id == 0) {
1778 /* search the root of the inode that was passed */
1779 root = BTRFS_I(inode)->root;
1780 } else {
1781 key.objectid = sk->tree_id;
1782 key.type = BTRFS_ROOT_ITEM_KEY;
1783 key.offset = (u64)-1;
1784 root = btrfs_read_fs_root_no_name(info, &key);
1785 if (IS_ERR(root)) {
1786 printk(KERN_ERR "could not find root %llu\n",
1787 sk->tree_id);
1788 btrfs_free_path(path);
1789 return -ENOENT;
1790 }
1791 }
1792
1793 key.objectid = sk->min_objectid;
1794 key.type = sk->min_type;
1795 key.offset = sk->min_offset;
1796
1797 max_key.objectid = sk->max_objectid;
1798 max_key.type = sk->max_type;
1799 max_key.offset = sk->max_offset;
1800
1801 path->keep_locks = 1;
1802
1803 while(1) {
1804 ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1805 sk->min_transid);
1806 if (ret != 0) {
1807 if (ret > 0)
1808 ret = 0;
1809 goto err;
1810 }
1811 ret = copy_to_sk(root, path, &key, sk, args->buf,
1812 &sk_offset, &num_found);
1813 btrfs_release_path(path);
1814 if (ret || num_found >= sk->nr_items)
1815 break;
1816
1817 }
1818 ret = 0;
1819 err:
1820 sk->nr_items = num_found;
1821 btrfs_free_path(path);
1822 return ret;
1823 }
1824
1825 static noinline int btrfs_ioctl_tree_search(struct file *file,
1826 void __user *argp)
1827 {
1828 struct btrfs_ioctl_search_args *args;
1829 struct inode *inode;
1830 int ret;
1831
1832 if (!capable(CAP_SYS_ADMIN))
1833 return -EPERM;
1834
1835 args = memdup_user(argp, sizeof(*args));
1836 if (IS_ERR(args))
1837 return PTR_ERR(args);
1838
1839 inode = fdentry(file)->d_inode;
1840 ret = search_ioctl(inode, args);
1841 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1842 ret = -EFAULT;
1843 kfree(args);
1844 return ret;
1845 }
1846
1847 /*
1848 * Search INODE_REFs to identify path name of 'dirid' directory
1849 * in a 'tree_id' tree. and sets path name to 'name'.
1850 */
1851 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1852 u64 tree_id, u64 dirid, char *name)
1853 {
1854 struct btrfs_root *root;
1855 struct btrfs_key key;
1856 char *ptr;
1857 int ret = -1;
1858 int slot;
1859 int len;
1860 int total_len = 0;
1861 struct btrfs_inode_ref *iref;
1862 struct extent_buffer *l;
1863 struct btrfs_path *path;
1864
1865 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1866 name[0]='\0';
1867 return 0;
1868 }
1869
1870 path = btrfs_alloc_path();
1871 if (!path)
1872 return -ENOMEM;
1873
1874 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1875
1876 key.objectid = tree_id;
1877 key.type = BTRFS_ROOT_ITEM_KEY;
1878 key.offset = (u64)-1;
1879 root = btrfs_read_fs_root_no_name(info, &key);
1880 if (IS_ERR(root)) {
1881 printk(KERN_ERR "could not find root %llu\n", tree_id);
1882 ret = -ENOENT;
1883 goto out;
1884 }
1885
1886 key.objectid = dirid;
1887 key.type = BTRFS_INODE_REF_KEY;
1888 key.offset = (u64)-1;
1889
1890 while(1) {
1891 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1892 if (ret < 0)
1893 goto out;
1894
1895 l = path->nodes[0];
1896 slot = path->slots[0];
1897 if (ret > 0 && slot > 0)
1898 slot--;
1899 btrfs_item_key_to_cpu(l, &key, slot);
1900
1901 if (ret > 0 && (key.objectid != dirid ||
1902 key.type != BTRFS_INODE_REF_KEY)) {
1903 ret = -ENOENT;
1904 goto out;
1905 }
1906
1907 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1908 len = btrfs_inode_ref_name_len(l, iref);
1909 ptr -= len + 1;
1910 total_len += len + 1;
1911 if (ptr < name)
1912 goto out;
1913
1914 *(ptr + len) = '/';
1915 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1916
1917 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1918 break;
1919
1920 btrfs_release_path(path);
1921 key.objectid = key.offset;
1922 key.offset = (u64)-1;
1923 dirid = key.objectid;
1924 }
1925 if (ptr < name)
1926 goto out;
1927 memmove(name, ptr, total_len);
1928 name[total_len]='\0';
1929 ret = 0;
1930 out:
1931 btrfs_free_path(path);
1932 return ret;
1933 }
1934
1935 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1936 void __user *argp)
1937 {
1938 struct btrfs_ioctl_ino_lookup_args *args;
1939 struct inode *inode;
1940 int ret;
1941
1942 if (!capable(CAP_SYS_ADMIN))
1943 return -EPERM;
1944
1945 args = memdup_user(argp, sizeof(*args));
1946 if (IS_ERR(args))
1947 return PTR_ERR(args);
1948
1949 inode = fdentry(file)->d_inode;
1950
1951 if (args->treeid == 0)
1952 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1953
1954 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1955 args->treeid, args->objectid,
1956 args->name);
1957
1958 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1959 ret = -EFAULT;
1960
1961 kfree(args);
1962 return ret;
1963 }
1964
1965 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1966 void __user *arg)
1967 {
1968 struct dentry *parent = fdentry(file);
1969 struct dentry *dentry;
1970 struct inode *dir = parent->d_inode;
1971 struct inode *inode;
1972 struct btrfs_root *root = BTRFS_I(dir)->root;
1973 struct btrfs_root *dest = NULL;
1974 struct btrfs_ioctl_vol_args *vol_args;
1975 struct btrfs_trans_handle *trans;
1976 int namelen;
1977 int ret;
1978 int err = 0;
1979
1980 vol_args = memdup_user(arg, sizeof(*vol_args));
1981 if (IS_ERR(vol_args))
1982 return PTR_ERR(vol_args);
1983
1984 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1985 namelen = strlen(vol_args->name);
1986 if (strchr(vol_args->name, '/') ||
1987 strncmp(vol_args->name, "..", namelen) == 0) {
1988 err = -EINVAL;
1989 goto out;
1990 }
1991
1992 err = mnt_want_write_file(file);
1993 if (err)
1994 goto out;
1995
1996 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
1997 dentry = lookup_one_len(vol_args->name, parent, namelen);
1998 if (IS_ERR(dentry)) {
1999 err = PTR_ERR(dentry);
2000 goto out_unlock_dir;
2001 }
2002
2003 if (!dentry->d_inode) {
2004 err = -ENOENT;
2005 goto out_dput;
2006 }
2007
2008 inode = dentry->d_inode;
2009 dest = BTRFS_I(inode)->root;
2010 if (!capable(CAP_SYS_ADMIN)){
2011 /*
2012 * Regular user. Only allow this with a special mount
2013 * option, when the user has write+exec access to the
2014 * subvol root, and when rmdir(2) would have been
2015 * allowed.
2016 *
2017 * Note that this is _not_ check that the subvol is
2018 * empty or doesn't contain data that we wouldn't
2019 * otherwise be able to delete.
2020 *
2021 * Users who want to delete empty subvols should try
2022 * rmdir(2).
2023 */
2024 err = -EPERM;
2025 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2026 goto out_dput;
2027
2028 /*
2029 * Do not allow deletion if the parent dir is the same
2030 * as the dir to be deleted. That means the ioctl
2031 * must be called on the dentry referencing the root
2032 * of the subvol, not a random directory contained
2033 * within it.
2034 */
2035 err = -EINVAL;
2036 if (root == dest)
2037 goto out_dput;
2038
2039 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2040 if (err)
2041 goto out_dput;
2042
2043 /* check if subvolume may be deleted by a non-root user */
2044 err = btrfs_may_delete(dir, dentry, 1);
2045 if (err)
2046 goto out_dput;
2047 }
2048
2049 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2050 err = -EINVAL;
2051 goto out_dput;
2052 }
2053
2054 mutex_lock(&inode->i_mutex);
2055 err = d_invalidate(dentry);
2056 if (err)
2057 goto out_unlock;
2058
2059 down_write(&root->fs_info->subvol_sem);
2060
2061 err = may_destroy_subvol(dest);
2062 if (err)
2063 goto out_up_write;
2064
2065 trans = btrfs_start_transaction(root, 0);
2066 if (IS_ERR(trans)) {
2067 err = PTR_ERR(trans);
2068 goto out_up_write;
2069 }
2070 trans->block_rsv = &root->fs_info->global_block_rsv;
2071
2072 ret = btrfs_unlink_subvol(trans, root, dir,
2073 dest->root_key.objectid,
2074 dentry->d_name.name,
2075 dentry->d_name.len);
2076 if (ret) {
2077 err = ret;
2078 btrfs_abort_transaction(trans, root, ret);
2079 goto out_end_trans;
2080 }
2081
2082 btrfs_record_root_in_trans(trans, dest);
2083
2084 memset(&dest->root_item.drop_progress, 0,
2085 sizeof(dest->root_item.drop_progress));
2086 dest->root_item.drop_level = 0;
2087 btrfs_set_root_refs(&dest->root_item, 0);
2088
2089 if (!xchg(&dest->orphan_item_inserted, 1)) {
2090 ret = btrfs_insert_orphan_item(trans,
2091 root->fs_info->tree_root,
2092 dest->root_key.objectid);
2093 if (ret) {
2094 btrfs_abort_transaction(trans, root, ret);
2095 err = ret;
2096 goto out_end_trans;
2097 }
2098 }
2099 out_end_trans:
2100 ret = btrfs_end_transaction(trans, root);
2101 if (ret && !err)
2102 err = ret;
2103 inode->i_flags |= S_DEAD;
2104 out_up_write:
2105 up_write(&root->fs_info->subvol_sem);
2106 out_unlock:
2107 mutex_unlock(&inode->i_mutex);
2108 if (!err) {
2109 shrink_dcache_sb(root->fs_info->sb);
2110 btrfs_invalidate_inodes(dest);
2111 d_delete(dentry);
2112 }
2113 out_dput:
2114 dput(dentry);
2115 out_unlock_dir:
2116 mutex_unlock(&dir->i_mutex);
2117 mnt_drop_write_file(file);
2118 out:
2119 kfree(vol_args);
2120 return err;
2121 }
2122
2123 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2124 {
2125 struct inode *inode = fdentry(file)->d_inode;
2126 struct btrfs_root *root = BTRFS_I(inode)->root;
2127 struct btrfs_ioctl_defrag_range_args *range;
2128 int ret;
2129
2130 if (btrfs_root_readonly(root))
2131 return -EROFS;
2132
2133 ret = mnt_want_write_file(file);
2134 if (ret)
2135 return ret;
2136
2137 switch (inode->i_mode & S_IFMT) {
2138 case S_IFDIR:
2139 if (!capable(CAP_SYS_ADMIN)) {
2140 ret = -EPERM;
2141 goto out;
2142 }
2143 ret = btrfs_defrag_root(root, 0);
2144 if (ret)
2145 goto out;
2146 ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
2147 break;
2148 case S_IFREG:
2149 if (!(file->f_mode & FMODE_WRITE)) {
2150 ret = -EINVAL;
2151 goto out;
2152 }
2153
2154 range = kzalloc(sizeof(*range), GFP_KERNEL);
2155 if (!range) {
2156 ret = -ENOMEM;
2157 goto out;
2158 }
2159
2160 if (argp) {
2161 if (copy_from_user(range, argp,
2162 sizeof(*range))) {
2163 ret = -EFAULT;
2164 kfree(range);
2165 goto out;
2166 }
2167 /* compression requires us to start the IO */
2168 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2169 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2170 range->extent_thresh = (u32)-1;
2171 }
2172 } else {
2173 /* the rest are all set to zero by kzalloc */
2174 range->len = (u64)-1;
2175 }
2176 ret = btrfs_defrag_file(fdentry(file)->d_inode, file,
2177 range, 0, 0);
2178 if (ret > 0)
2179 ret = 0;
2180 kfree(range);
2181 break;
2182 default:
2183 ret = -EINVAL;
2184 }
2185 out:
2186 mnt_drop_write_file(file);
2187 return ret;
2188 }
2189
2190 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2191 {
2192 struct btrfs_ioctl_vol_args *vol_args;
2193 int ret;
2194
2195 if (!capable(CAP_SYS_ADMIN))
2196 return -EPERM;
2197
2198 mutex_lock(&root->fs_info->volume_mutex);
2199 if (root->fs_info->balance_ctl) {
2200 printk(KERN_INFO "btrfs: balance in progress\n");
2201 ret = -EINVAL;
2202 goto out;
2203 }
2204
2205 vol_args = memdup_user(arg, sizeof(*vol_args));
2206 if (IS_ERR(vol_args)) {
2207 ret = PTR_ERR(vol_args);
2208 goto out;
2209 }
2210
2211 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2212 ret = btrfs_init_new_device(root, vol_args->name);
2213
2214 kfree(vol_args);
2215 out:
2216 mutex_unlock(&root->fs_info->volume_mutex);
2217 return ret;
2218 }
2219
2220 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
2221 {
2222 struct btrfs_ioctl_vol_args *vol_args;
2223 int ret;
2224
2225 if (!capable(CAP_SYS_ADMIN))
2226 return -EPERM;
2227
2228 if (root->fs_info->sb->s_flags & MS_RDONLY)
2229 return -EROFS;
2230
2231 mutex_lock(&root->fs_info->volume_mutex);
2232 if (root->fs_info->balance_ctl) {
2233 printk(KERN_INFO "btrfs: balance in progress\n");
2234 ret = -EINVAL;
2235 goto out;
2236 }
2237
2238 vol_args = memdup_user(arg, sizeof(*vol_args));
2239 if (IS_ERR(vol_args)) {
2240 ret = PTR_ERR(vol_args);
2241 goto out;
2242 }
2243
2244 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2245 ret = btrfs_rm_device(root, vol_args->name);
2246
2247 kfree(vol_args);
2248 out:
2249 mutex_unlock(&root->fs_info->volume_mutex);
2250 return ret;
2251 }
2252
2253 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2254 {
2255 struct btrfs_ioctl_fs_info_args *fi_args;
2256 struct btrfs_device *device;
2257 struct btrfs_device *next;
2258 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2259 int ret = 0;
2260
2261 if (!capable(CAP_SYS_ADMIN))
2262 return -EPERM;
2263
2264 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2265 if (!fi_args)
2266 return -ENOMEM;
2267
2268 fi_args->num_devices = fs_devices->num_devices;
2269 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2270
2271 mutex_lock(&fs_devices->device_list_mutex);
2272 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2273 if (device->devid > fi_args->max_id)
2274 fi_args->max_id = device->devid;
2275 }
2276 mutex_unlock(&fs_devices->device_list_mutex);
2277
2278 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2279 ret = -EFAULT;
2280
2281 kfree(fi_args);
2282 return ret;
2283 }
2284
2285 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2286 {
2287 struct btrfs_ioctl_dev_info_args *di_args;
2288 struct btrfs_device *dev;
2289 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2290 int ret = 0;
2291 char *s_uuid = NULL;
2292 char empty_uuid[BTRFS_UUID_SIZE] = {0};
2293
2294 if (!capable(CAP_SYS_ADMIN))
2295 return -EPERM;
2296
2297 di_args = memdup_user(arg, sizeof(*di_args));
2298 if (IS_ERR(di_args))
2299 return PTR_ERR(di_args);
2300
2301 if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2302 s_uuid = di_args->uuid;
2303
2304 mutex_lock(&fs_devices->device_list_mutex);
2305 dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL);
2306 mutex_unlock(&fs_devices->device_list_mutex);
2307
2308 if (!dev) {
2309 ret = -ENODEV;
2310 goto out;
2311 }
2312
2313 di_args->devid = dev->devid;
2314 di_args->bytes_used = dev->bytes_used;
2315 di_args->total_bytes = dev->total_bytes;
2316 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2317 if (dev->name) {
2318 struct rcu_string *name;
2319
2320 rcu_read_lock();
2321 name = rcu_dereference(dev->name);
2322 strncpy(di_args->path, name->str, sizeof(di_args->path));
2323 rcu_read_unlock();
2324 di_args->path[sizeof(di_args->path) - 1] = 0;
2325 } else {
2326 di_args->path[0] = '\0';
2327 }
2328
2329 out:
2330 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2331 ret = -EFAULT;
2332
2333 kfree(di_args);
2334 return ret;
2335 }
2336
2337 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2338 u64 off, u64 olen, u64 destoff)
2339 {
2340 struct inode *inode = fdentry(file)->d_inode;
2341 struct btrfs_root *root = BTRFS_I(inode)->root;
2342 struct fd src_file;
2343 struct inode *src;
2344 struct btrfs_trans_handle *trans;
2345 struct btrfs_path *path;
2346 struct extent_buffer *leaf;
2347 char *buf;
2348 struct btrfs_key key;
2349 u32 nritems;
2350 int slot;
2351 int ret;
2352 u64 len = olen;
2353 u64 bs = root->fs_info->sb->s_blocksize;
2354 u64 hint_byte;
2355
2356 /*
2357 * TODO:
2358 * - split compressed inline extents. annoying: we need to
2359 * decompress into destination's address_space (the file offset
2360 * may change, so source mapping won't do), then recompress (or
2361 * otherwise reinsert) a subrange.
2362 * - allow ranges within the same file to be cloned (provided
2363 * they don't overlap)?
2364 */
2365
2366 /* the destination must be opened for writing */
2367 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
2368 return -EINVAL;
2369
2370 if (btrfs_root_readonly(root))
2371 return -EROFS;
2372
2373 ret = mnt_want_write_file(file);
2374 if (ret)
2375 return ret;
2376
2377 src_file = fdget(srcfd);
2378 if (!src_file.file) {
2379 ret = -EBADF;
2380 goto out_drop_write;
2381 }
2382
2383 ret = -EXDEV;
2384 if (src_file.file->f_path.mnt != file->f_path.mnt)
2385 goto out_fput;
2386
2387 src = src_file.file->f_dentry->d_inode;
2388
2389 ret = -EINVAL;
2390 if (src == inode)
2391 goto out_fput;
2392
2393 /* the src must be open for reading */
2394 if (!(src_file.file->f_mode & FMODE_READ))
2395 goto out_fput;
2396
2397 /* don't make the dst file partly checksummed */
2398 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2399 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2400 goto out_fput;
2401
2402 ret = -EISDIR;
2403 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2404 goto out_fput;
2405
2406 ret = -EXDEV;
2407 if (src->i_sb != inode->i_sb)
2408 goto out_fput;
2409
2410 ret = -ENOMEM;
2411 buf = vmalloc(btrfs_level_size(root, 0));
2412 if (!buf)
2413 goto out_fput;
2414
2415 path = btrfs_alloc_path();
2416 if (!path) {
2417 vfree(buf);
2418 goto out_fput;
2419 }
2420 path->reada = 2;
2421
2422 if (inode < src) {
2423 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2424 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
2425 } else {
2426 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2427 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
2428 }
2429
2430 /* determine range to clone */
2431 ret = -EINVAL;
2432 if (off + len > src->i_size || off + len < off)
2433 goto out_unlock;
2434 if (len == 0)
2435 olen = len = src->i_size - off;
2436 /* if we extend to eof, continue to block boundary */
2437 if (off + len == src->i_size)
2438 len = ALIGN(src->i_size, bs) - off;
2439
2440 /* verify the end result is block aligned */
2441 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2442 !IS_ALIGNED(destoff, bs))
2443 goto out_unlock;
2444
2445 if (destoff > inode->i_size) {
2446 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2447 if (ret)
2448 goto out_unlock;
2449 }
2450
2451 /* truncate page cache pages from target inode range */
2452 truncate_inode_pages_range(&inode->i_data, destoff,
2453 PAGE_CACHE_ALIGN(destoff + len) - 1);
2454
2455 /* do any pending delalloc/csum calc on src, one way or
2456 another, and lock file content */
2457 while (1) {
2458 struct btrfs_ordered_extent *ordered;
2459 lock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2460 ordered = btrfs_lookup_first_ordered_extent(src, off+len);
2461 if (!ordered &&
2462 !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len,
2463 EXTENT_DELALLOC, 0, NULL))
2464 break;
2465 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2466 if (ordered)
2467 btrfs_put_ordered_extent(ordered);
2468 btrfs_wait_ordered_range(src, off, len);
2469 }
2470
2471 /* clone data */
2472 key.objectid = btrfs_ino(src);
2473 key.type = BTRFS_EXTENT_DATA_KEY;
2474 key.offset = 0;
2475
2476 while (1) {
2477 /*
2478 * note the key will change type as we walk through the
2479 * tree.
2480 */
2481 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
2482 0, 0);
2483 if (ret < 0)
2484 goto out;
2485
2486 nritems = btrfs_header_nritems(path->nodes[0]);
2487 if (path->slots[0] >= nritems) {
2488 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
2489 if (ret < 0)
2490 goto out;
2491 if (ret > 0)
2492 break;
2493 nritems = btrfs_header_nritems(path->nodes[0]);
2494 }
2495 leaf = path->nodes[0];
2496 slot = path->slots[0];
2497
2498 btrfs_item_key_to_cpu(leaf, &key, slot);
2499 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
2500 key.objectid != btrfs_ino(src))
2501 break;
2502
2503 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2504 struct btrfs_file_extent_item *extent;
2505 int type;
2506 u32 size;
2507 struct btrfs_key new_key;
2508 u64 disko = 0, diskl = 0;
2509 u64 datao = 0, datal = 0;
2510 u8 comp;
2511 u64 endoff;
2512
2513 size = btrfs_item_size_nr(leaf, slot);
2514 read_extent_buffer(leaf, buf,
2515 btrfs_item_ptr_offset(leaf, slot),
2516 size);
2517
2518 extent = btrfs_item_ptr(leaf, slot,
2519 struct btrfs_file_extent_item);
2520 comp = btrfs_file_extent_compression(leaf, extent);
2521 type = btrfs_file_extent_type(leaf, extent);
2522 if (type == BTRFS_FILE_EXTENT_REG ||
2523 type == BTRFS_FILE_EXTENT_PREALLOC) {
2524 disko = btrfs_file_extent_disk_bytenr(leaf,
2525 extent);
2526 diskl = btrfs_file_extent_disk_num_bytes(leaf,
2527 extent);
2528 datao = btrfs_file_extent_offset(leaf, extent);
2529 datal = btrfs_file_extent_num_bytes(leaf,
2530 extent);
2531 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2532 /* take upper bound, may be compressed */
2533 datal = btrfs_file_extent_ram_bytes(leaf,
2534 extent);
2535 }
2536 btrfs_release_path(path);
2537
2538 if (key.offset + datal <= off ||
2539 key.offset >= off+len)
2540 goto next;
2541
2542 memcpy(&new_key, &key, sizeof(new_key));
2543 new_key.objectid = btrfs_ino(inode);
2544 if (off <= key.offset)
2545 new_key.offset = key.offset + destoff - off;
2546 else
2547 new_key.offset = destoff;
2548
2549 /*
2550 * 1 - adjusting old extent (we may have to split it)
2551 * 1 - add new extent
2552 * 1 - inode update
2553 */
2554 trans = btrfs_start_transaction(root, 3);
2555 if (IS_ERR(trans)) {
2556 ret = PTR_ERR(trans);
2557 goto out;
2558 }
2559
2560 if (type == BTRFS_FILE_EXTENT_REG ||
2561 type == BTRFS_FILE_EXTENT_PREALLOC) {
2562 /*
2563 * a | --- range to clone ---| b
2564 * | ------------- extent ------------- |
2565 */
2566
2567 /* substract range b */
2568 if (key.offset + datal > off + len)
2569 datal = off + len - key.offset;
2570
2571 /* substract range a */
2572 if (off > key.offset) {
2573 datao += off - key.offset;
2574 datal -= off - key.offset;
2575 }
2576
2577 ret = btrfs_drop_extents(trans, inode,
2578 new_key.offset,
2579 new_key.offset + datal,
2580 &hint_byte, 1);
2581 if (ret) {
2582 btrfs_abort_transaction(trans, root,
2583 ret);
2584 btrfs_end_transaction(trans, root);
2585 goto out;
2586 }
2587
2588 ret = btrfs_insert_empty_item(trans, root, path,
2589 &new_key, size);
2590 if (ret) {
2591 btrfs_abort_transaction(trans, root,
2592 ret);
2593 btrfs_end_transaction(trans, root);
2594 goto out;
2595 }
2596
2597 leaf = path->nodes[0];
2598 slot = path->slots[0];
2599 write_extent_buffer(leaf, buf,
2600 btrfs_item_ptr_offset(leaf, slot),
2601 size);
2602
2603 extent = btrfs_item_ptr(leaf, slot,
2604 struct btrfs_file_extent_item);
2605
2606 /* disko == 0 means it's a hole */
2607 if (!disko)
2608 datao = 0;
2609
2610 btrfs_set_file_extent_offset(leaf, extent,
2611 datao);
2612 btrfs_set_file_extent_num_bytes(leaf, extent,
2613 datal);
2614 if (disko) {
2615 inode_add_bytes(inode, datal);
2616 ret = btrfs_inc_extent_ref(trans, root,
2617 disko, diskl, 0,
2618 root->root_key.objectid,
2619 btrfs_ino(inode),
2620 new_key.offset - datao,
2621 0);
2622 if (ret) {
2623 btrfs_abort_transaction(trans,
2624 root,
2625 ret);
2626 btrfs_end_transaction(trans,
2627 root);
2628 goto out;
2629
2630 }
2631 }
2632 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2633 u64 skip = 0;
2634 u64 trim = 0;
2635 if (off > key.offset) {
2636 skip = off - key.offset;
2637 new_key.offset += skip;
2638 }
2639
2640 if (key.offset + datal > off+len)
2641 trim = key.offset + datal - (off+len);
2642
2643 if (comp && (skip || trim)) {
2644 ret = -EINVAL;
2645 btrfs_end_transaction(trans, root);
2646 goto out;
2647 }
2648 size -= skip + trim;
2649 datal -= skip + trim;
2650
2651 ret = btrfs_drop_extents(trans, inode,
2652 new_key.offset,
2653 new_key.offset + datal,
2654 &hint_byte, 1);
2655 if (ret) {
2656 btrfs_abort_transaction(trans, root,
2657 ret);
2658 btrfs_end_transaction(trans, root);
2659 goto out;
2660 }
2661
2662 ret = btrfs_insert_empty_item(trans, root, path,
2663 &new_key, size);
2664 if (ret) {
2665 btrfs_abort_transaction(trans, root,
2666 ret);
2667 btrfs_end_transaction(trans, root);
2668 goto out;
2669 }
2670
2671 if (skip) {
2672 u32 start =
2673 btrfs_file_extent_calc_inline_size(0);
2674 memmove(buf+start, buf+start+skip,
2675 datal);
2676 }
2677
2678 leaf = path->nodes[0];
2679 slot = path->slots[0];
2680 write_extent_buffer(leaf, buf,
2681 btrfs_item_ptr_offset(leaf, slot),
2682 size);
2683 inode_add_bytes(inode, datal);
2684 }
2685
2686 btrfs_mark_buffer_dirty(leaf);
2687 btrfs_release_path(path);
2688
2689 inode_inc_iversion(inode);
2690 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2691
2692 /*
2693 * we round up to the block size at eof when
2694 * determining which extents to clone above,
2695 * but shouldn't round up the file size
2696 */
2697 endoff = new_key.offset + datal;
2698 if (endoff > destoff+olen)
2699 endoff = destoff+olen;
2700 if (endoff > inode->i_size)
2701 btrfs_i_size_write(inode, endoff);
2702
2703 ret = btrfs_update_inode(trans, root, inode);
2704 if (ret) {
2705 btrfs_abort_transaction(trans, root, ret);
2706 btrfs_end_transaction(trans, root);
2707 goto out;
2708 }
2709 ret = btrfs_end_transaction(trans, root);
2710 }
2711 next:
2712 btrfs_release_path(path);
2713 key.offset++;
2714 }
2715 ret = 0;
2716 out:
2717 btrfs_release_path(path);
2718 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len);
2719 out_unlock:
2720 mutex_unlock(&src->i_mutex);
2721 mutex_unlock(&inode->i_mutex);
2722 vfree(buf);
2723 btrfs_free_path(path);
2724 out_fput:
2725 fdput(src_file);
2726 out_drop_write:
2727 mnt_drop_write_file(file);
2728 return ret;
2729 }
2730
2731 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
2732 {
2733 struct btrfs_ioctl_clone_range_args args;
2734
2735 if (copy_from_user(&args, argp, sizeof(args)))
2736 return -EFAULT;
2737 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2738 args.src_length, args.dest_offset);
2739 }
2740
2741 /*
2742 * there are many ways the trans_start and trans_end ioctls can lead
2743 * to deadlocks. They should only be used by applications that
2744 * basically own the machine, and have a very in depth understanding
2745 * of all the possible deadlocks and enospc problems.
2746 */
2747 static long btrfs_ioctl_trans_start(struct file *file)
2748 {
2749 struct inode *inode = fdentry(file)->d_inode;
2750 struct btrfs_root *root = BTRFS_I(inode)->root;
2751 struct btrfs_trans_handle *trans;
2752 int ret;
2753
2754 ret = -EPERM;
2755 if (!capable(CAP_SYS_ADMIN))
2756 goto out;
2757
2758 ret = -EINPROGRESS;
2759 if (file->private_data)
2760 goto out;
2761
2762 ret = -EROFS;
2763 if (btrfs_root_readonly(root))
2764 goto out;
2765
2766 ret = mnt_want_write_file(file);
2767 if (ret)
2768 goto out;
2769
2770 atomic_inc(&root->fs_info->open_ioctl_trans);
2771
2772 ret = -ENOMEM;
2773 trans = btrfs_start_ioctl_transaction(root);
2774 if (IS_ERR(trans))
2775 goto out_drop;
2776
2777 file->private_data = trans;
2778 return 0;
2779
2780 out_drop:
2781 atomic_dec(&root->fs_info->open_ioctl_trans);
2782 mnt_drop_write_file(file);
2783 out:
2784 return ret;
2785 }
2786
2787 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2788 {
2789 struct inode *inode = fdentry(file)->d_inode;
2790 struct btrfs_root *root = BTRFS_I(inode)->root;
2791 struct btrfs_root *new_root;
2792 struct btrfs_dir_item *di;
2793 struct btrfs_trans_handle *trans;
2794 struct btrfs_path *path;
2795 struct btrfs_key location;
2796 struct btrfs_disk_key disk_key;
2797 u64 objectid = 0;
2798 u64 dir_id;
2799
2800 if (!capable(CAP_SYS_ADMIN))
2801 return -EPERM;
2802
2803 if (copy_from_user(&objectid, argp, sizeof(objectid)))
2804 return -EFAULT;
2805
2806 if (!objectid)
2807 objectid = root->root_key.objectid;
2808
2809 location.objectid = objectid;
2810 location.type = BTRFS_ROOT_ITEM_KEY;
2811 location.offset = (u64)-1;
2812
2813 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2814 if (IS_ERR(new_root))
2815 return PTR_ERR(new_root);
2816
2817 if (btrfs_root_refs(&new_root->root_item) == 0)
2818 return -ENOENT;
2819
2820 path = btrfs_alloc_path();
2821 if (!path)
2822 return -ENOMEM;
2823 path->leave_spinning = 1;
2824
2825 trans = btrfs_start_transaction(root, 1);
2826 if (IS_ERR(trans)) {
2827 btrfs_free_path(path);
2828 return PTR_ERR(trans);
2829 }
2830
2831 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
2832 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2833 dir_id, "default", 7, 1);
2834 if (IS_ERR_OR_NULL(di)) {
2835 btrfs_free_path(path);
2836 btrfs_end_transaction(trans, root);
2837 printk(KERN_ERR "Umm, you don't have the default dir item, "
2838 "this isn't going to work\n");
2839 return -ENOENT;
2840 }
2841
2842 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2843 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2844 btrfs_mark_buffer_dirty(path->nodes[0]);
2845 btrfs_free_path(path);
2846
2847 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
2848 btrfs_end_transaction(trans, root);
2849
2850 return 0;
2851 }
2852
2853 static void get_block_group_info(struct list_head *groups_list,
2854 struct btrfs_ioctl_space_info *space)
2855 {
2856 struct btrfs_block_group_cache *block_group;
2857
2858 space->total_bytes = 0;
2859 space->used_bytes = 0;
2860 space->flags = 0;
2861 list_for_each_entry(block_group, groups_list, list) {
2862 space->flags = block_group->flags;
2863 space->total_bytes += block_group->key.offset;
2864 space->used_bytes +=
2865 btrfs_block_group_used(&block_group->item);
2866 }
2867 }
2868
2869 long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2870 {
2871 struct btrfs_ioctl_space_args space_args;
2872 struct btrfs_ioctl_space_info space;
2873 struct btrfs_ioctl_space_info *dest;
2874 struct btrfs_ioctl_space_info *dest_orig;
2875 struct btrfs_ioctl_space_info __user *user_dest;
2876 struct btrfs_space_info *info;
2877 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2878 BTRFS_BLOCK_GROUP_SYSTEM,
2879 BTRFS_BLOCK_GROUP_METADATA,
2880 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2881 int num_types = 4;
2882 int alloc_size;
2883 int ret = 0;
2884 u64 slot_count = 0;
2885 int i, c;
2886
2887 if (copy_from_user(&space_args,
2888 (struct btrfs_ioctl_space_args __user *)arg,
2889 sizeof(space_args)))
2890 return -EFAULT;
2891
2892 for (i = 0; i < num_types; i++) {
2893 struct btrfs_space_info *tmp;
2894
2895 info = NULL;
2896 rcu_read_lock();
2897 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2898 list) {
2899 if (tmp->flags == types[i]) {
2900 info = tmp;
2901 break;
2902 }
2903 }
2904 rcu_read_unlock();
2905
2906 if (!info)
2907 continue;
2908
2909 down_read(&info->groups_sem);
2910 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2911 if (!list_empty(&info->block_groups[c]))
2912 slot_count++;
2913 }
2914 up_read(&info->groups_sem);
2915 }
2916
2917 /* space_slots == 0 means they are asking for a count */
2918 if (space_args.space_slots == 0) {
2919 space_args.total_spaces = slot_count;
2920 goto out;
2921 }
2922
2923 slot_count = min_t(u64, space_args.space_slots, slot_count);
2924
2925 alloc_size = sizeof(*dest) * slot_count;
2926
2927 /* we generally have at most 6 or so space infos, one for each raid
2928 * level. So, a whole page should be more than enough for everyone
2929 */
2930 if (alloc_size > PAGE_CACHE_SIZE)
2931 return -ENOMEM;
2932
2933 space_args.total_spaces = 0;
2934 dest = kmalloc(alloc_size, GFP_NOFS);
2935 if (!dest)
2936 return -ENOMEM;
2937 dest_orig = dest;
2938
2939 /* now we have a buffer to copy into */
2940 for (i = 0; i < num_types; i++) {
2941 struct btrfs_space_info *tmp;
2942
2943 if (!slot_count)
2944 break;
2945
2946 info = NULL;
2947 rcu_read_lock();
2948 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2949 list) {
2950 if (tmp->flags == types[i]) {
2951 info = tmp;
2952 break;
2953 }
2954 }
2955 rcu_read_unlock();
2956
2957 if (!info)
2958 continue;
2959 down_read(&info->groups_sem);
2960 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2961 if (!list_empty(&info->block_groups[c])) {
2962 get_block_group_info(&info->block_groups[c],
2963 &space);
2964 memcpy(dest, &space, sizeof(space));
2965 dest++;
2966 space_args.total_spaces++;
2967 slot_count--;
2968 }
2969 if (!slot_count)
2970 break;
2971 }
2972 up_read(&info->groups_sem);
2973 }
2974
2975 user_dest = (struct btrfs_ioctl_space_info __user *)
2976 (arg + sizeof(struct btrfs_ioctl_space_args));
2977
2978 if (copy_to_user(user_dest, dest_orig, alloc_size))
2979 ret = -EFAULT;
2980
2981 kfree(dest_orig);
2982 out:
2983 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
2984 ret = -EFAULT;
2985
2986 return ret;
2987 }
2988
2989 /*
2990 * there are many ways the trans_start and trans_end ioctls can lead
2991 * to deadlocks. They should only be used by applications that
2992 * basically own the machine, and have a very in depth understanding
2993 * of all the possible deadlocks and enospc problems.
2994 */
2995 long btrfs_ioctl_trans_end(struct file *file)
2996 {
2997 struct inode *inode = fdentry(file)->d_inode;
2998 struct btrfs_root *root = BTRFS_I(inode)->root;
2999 struct btrfs_trans_handle *trans;
3000
3001 trans = file->private_data;
3002 if (!trans)
3003 return -EINVAL;
3004 file->private_data = NULL;
3005
3006 btrfs_end_transaction(trans, root);
3007
3008 atomic_dec(&root->fs_info->open_ioctl_trans);
3009
3010 mnt_drop_write_file(file);
3011 return 0;
3012 }
3013
3014 static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
3015 {
3016 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
3017 struct btrfs_trans_handle *trans;
3018 u64 transid;
3019 int ret;
3020
3021 trans = btrfs_start_transaction(root, 0);
3022 if (IS_ERR(trans))
3023 return PTR_ERR(trans);
3024 transid = trans->transid;
3025 ret = btrfs_commit_transaction_async(trans, root, 0);
3026 if (ret) {
3027 btrfs_end_transaction(trans, root);
3028 return ret;
3029 }
3030
3031 if (argp)
3032 if (copy_to_user(argp, &transid, sizeof(transid)))
3033 return -EFAULT;
3034 return 0;
3035 }
3036
3037 static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
3038 {
3039 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
3040 u64 transid;
3041
3042 if (argp) {
3043 if (copy_from_user(&transid, argp, sizeof(transid)))
3044 return -EFAULT;
3045 } else {
3046 transid = 0; /* current trans */
3047 }
3048 return btrfs_wait_for_commit(root, transid);
3049 }
3050
3051 static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg)
3052 {
3053 int ret;
3054 struct btrfs_ioctl_scrub_args *sa;
3055
3056 if (!capable(CAP_SYS_ADMIN))
3057 return -EPERM;
3058
3059 sa = memdup_user(arg, sizeof(*sa));
3060 if (IS_ERR(sa))
3061 return PTR_ERR(sa);
3062
3063 ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end,
3064 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY);
3065
3066 if (copy_to_user(arg, sa, sizeof(*sa)))
3067 ret = -EFAULT;
3068
3069 kfree(sa);
3070 return ret;
3071 }
3072
3073 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3074 {
3075 if (!capable(CAP_SYS_ADMIN))
3076 return -EPERM;
3077
3078 return btrfs_scrub_cancel(root);
3079 }
3080
3081 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3082 void __user *arg)
3083 {
3084 struct btrfs_ioctl_scrub_args *sa;
3085 int ret;
3086
3087 if (!capable(CAP_SYS_ADMIN))
3088 return -EPERM;
3089
3090 sa = memdup_user(arg, sizeof(*sa));
3091 if (IS_ERR(sa))
3092 return PTR_ERR(sa);
3093
3094 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3095
3096 if (copy_to_user(arg, sa, sizeof(*sa)))
3097 ret = -EFAULT;
3098
3099 kfree(sa);
3100 return ret;
3101 }
3102
3103 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
3104 void __user *arg)
3105 {
3106 struct btrfs_ioctl_get_dev_stats *sa;
3107 int ret;
3108
3109 sa = memdup_user(arg, sizeof(*sa));
3110 if (IS_ERR(sa))
3111 return PTR_ERR(sa);
3112
3113 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3114 kfree(sa);
3115 return -EPERM;
3116 }
3117
3118 ret = btrfs_get_dev_stats(root, sa);
3119
3120 if (copy_to_user(arg, sa, sizeof(*sa)))
3121 ret = -EFAULT;
3122
3123 kfree(sa);
3124 return ret;
3125 }
3126
3127 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3128 {
3129 int ret = 0;
3130 int i;
3131 u64 rel_ptr;
3132 int size;
3133 struct btrfs_ioctl_ino_path_args *ipa = NULL;
3134 struct inode_fs_paths *ipath = NULL;
3135 struct btrfs_path *path;
3136
3137 if (!capable(CAP_SYS_ADMIN))
3138 return -EPERM;
3139
3140 path = btrfs_alloc_path();
3141 if (!path) {
3142 ret = -ENOMEM;
3143 goto out;
3144 }
3145
3146 ipa = memdup_user(arg, sizeof(*ipa));
3147 if (IS_ERR(ipa)) {
3148 ret = PTR_ERR(ipa);
3149 ipa = NULL;
3150 goto out;
3151 }
3152
3153 size = min_t(u32, ipa->size, 4096);
3154 ipath = init_ipath(size, root, path);
3155 if (IS_ERR(ipath)) {
3156 ret = PTR_ERR(ipath);
3157 ipath = NULL;
3158 goto out;
3159 }
3160
3161 ret = paths_from_inode(ipa->inum, ipath);
3162 if (ret < 0)
3163 goto out;
3164
3165 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3166 rel_ptr = ipath->fspath->val[i] -
3167 (u64)(unsigned long)ipath->fspath->val;
3168 ipath->fspath->val[i] = rel_ptr;
3169 }
3170
3171 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3172 (void *)(unsigned long)ipath->fspath, size);
3173 if (ret) {
3174 ret = -EFAULT;
3175 goto out;
3176 }
3177
3178 out:
3179 btrfs_free_path(path);
3180 free_ipath(ipath);
3181 kfree(ipa);
3182
3183 return ret;
3184 }
3185
3186 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3187 {
3188 struct btrfs_data_container *inodes = ctx;
3189 const size_t c = 3 * sizeof(u64);
3190
3191 if (inodes->bytes_left >= c) {
3192 inodes->bytes_left -= c;
3193 inodes->val[inodes->elem_cnt] = inum;
3194 inodes->val[inodes->elem_cnt + 1] = offset;
3195 inodes->val[inodes->elem_cnt + 2] = root;
3196 inodes->elem_cnt += 3;
3197 } else {
3198 inodes->bytes_missing += c - inodes->bytes_left;
3199 inodes->bytes_left = 0;
3200 inodes->elem_missed += 3;
3201 }
3202
3203 return 0;
3204 }
3205
3206 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3207 void __user *arg)
3208 {
3209 int ret = 0;
3210 int size;
3211 u64 extent_item_pos;
3212 struct btrfs_ioctl_logical_ino_args *loi;
3213 struct btrfs_data_container *inodes = NULL;
3214 struct btrfs_path *path = NULL;
3215 struct btrfs_key key;
3216
3217 if (!capable(CAP_SYS_ADMIN))
3218 return -EPERM;
3219
3220 loi = memdup_user(arg, sizeof(*loi));
3221 if (IS_ERR(loi)) {
3222 ret = PTR_ERR(loi);
3223 loi = NULL;
3224 goto out;
3225 }
3226
3227 path = btrfs_alloc_path();
3228 if (!path) {
3229 ret = -ENOMEM;
3230 goto out;
3231 }
3232
3233 size = min_t(u32, loi->size, 4096);
3234 inodes = init_data_container(size);
3235 if (IS_ERR(inodes)) {
3236 ret = PTR_ERR(inodes);
3237 inodes = NULL;
3238 goto out;
3239 }
3240
3241 ret = extent_from_logical(root->fs_info, loi->logical, path, &key);
3242 btrfs_release_path(path);
3243
3244 if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK)
3245 ret = -ENOENT;
3246 if (ret < 0)
3247 goto out;
3248
3249 extent_item_pos = loi->logical - key.objectid;
3250 ret = iterate_extent_inodes(root->fs_info, key.objectid,
3251 extent_item_pos, 0, build_ino_list,
3252 inodes);
3253
3254 if (ret < 0)
3255 goto out;
3256
3257 ret = copy_to_user((void *)(unsigned long)loi->inodes,
3258 (void *)(unsigned long)inodes, size);
3259 if (ret)
3260 ret = -EFAULT;
3261
3262 out:
3263 btrfs_free_path(path);
3264 kfree(inodes);
3265 kfree(loi);
3266
3267 return ret;
3268 }
3269
3270 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
3271 struct btrfs_ioctl_balance_args *bargs)
3272 {
3273 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3274
3275 bargs->flags = bctl->flags;
3276
3277 if (atomic_read(&fs_info->balance_running))
3278 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3279 if (atomic_read(&fs_info->balance_pause_req))
3280 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3281 if (atomic_read(&fs_info->balance_cancel_req))
3282 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3283
3284 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3285 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3286 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3287
3288 if (lock) {
3289 spin_lock(&fs_info->balance_lock);
3290 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3291 spin_unlock(&fs_info->balance_lock);
3292 } else {
3293 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3294 }
3295 }
3296
3297 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3298 {
3299 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3300 struct btrfs_fs_info *fs_info = root->fs_info;
3301 struct btrfs_ioctl_balance_args *bargs;
3302 struct btrfs_balance_control *bctl;
3303 int ret;
3304
3305 if (!capable(CAP_SYS_ADMIN))
3306 return -EPERM;
3307
3308 ret = mnt_want_write_file(file);
3309 if (ret)
3310 return ret;
3311
3312 mutex_lock(&fs_info->volume_mutex);
3313 mutex_lock(&fs_info->balance_mutex);
3314
3315 if (arg) {
3316 bargs = memdup_user(arg, sizeof(*bargs));
3317 if (IS_ERR(bargs)) {
3318 ret = PTR_ERR(bargs);
3319 goto out;
3320 }
3321
3322 if (bargs->flags & BTRFS_BALANCE_RESUME) {
3323 if (!fs_info->balance_ctl) {
3324 ret = -ENOTCONN;
3325 goto out_bargs;
3326 }
3327
3328 bctl = fs_info->balance_ctl;
3329 spin_lock(&fs_info->balance_lock);
3330 bctl->flags |= BTRFS_BALANCE_RESUME;
3331 spin_unlock(&fs_info->balance_lock);
3332
3333 goto do_balance;
3334 }
3335 } else {
3336 bargs = NULL;
3337 }
3338
3339 if (fs_info->balance_ctl) {
3340 ret = -EINPROGRESS;
3341 goto out_bargs;
3342 }
3343
3344 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3345 if (!bctl) {
3346 ret = -ENOMEM;
3347 goto out_bargs;
3348 }
3349
3350 bctl->fs_info = fs_info;
3351 if (arg) {
3352 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3353 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3354 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3355
3356 bctl->flags = bargs->flags;
3357 } else {
3358 /* balance everything - no filters */
3359 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
3360 }
3361
3362 do_balance:
3363 ret = btrfs_balance(bctl, bargs);
3364 /*
3365 * bctl is freed in __cancel_balance or in free_fs_info if
3366 * restriper was paused all the way until unmount
3367 */
3368 if (arg) {
3369 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3370 ret = -EFAULT;
3371 }
3372
3373 out_bargs:
3374 kfree(bargs);
3375 out:
3376 mutex_unlock(&fs_info->balance_mutex);
3377 mutex_unlock(&fs_info->volume_mutex);
3378 mnt_drop_write_file(file);
3379 return ret;
3380 }
3381
3382 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
3383 {
3384 if (!capable(CAP_SYS_ADMIN))
3385 return -EPERM;
3386
3387 switch (cmd) {
3388 case BTRFS_BALANCE_CTL_PAUSE:
3389 return btrfs_pause_balance(root->fs_info);
3390 case BTRFS_BALANCE_CTL_CANCEL:
3391 return btrfs_cancel_balance(root->fs_info);
3392 }
3393
3394 return -EINVAL;
3395 }
3396
3397 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
3398 void __user *arg)
3399 {
3400 struct btrfs_fs_info *fs_info = root->fs_info;
3401 struct btrfs_ioctl_balance_args *bargs;
3402 int ret = 0;
3403
3404 if (!capable(CAP_SYS_ADMIN))
3405 return -EPERM;
3406
3407 mutex_lock(&fs_info->balance_mutex);
3408 if (!fs_info->balance_ctl) {
3409 ret = -ENOTCONN;
3410 goto out;
3411 }
3412
3413 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
3414 if (!bargs) {
3415 ret = -ENOMEM;
3416 goto out;
3417 }
3418
3419 update_ioctl_balance_args(fs_info, 1, bargs);
3420
3421 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3422 ret = -EFAULT;
3423
3424 kfree(bargs);
3425 out:
3426 mutex_unlock(&fs_info->balance_mutex);
3427 return ret;
3428 }
3429
3430 static long btrfs_ioctl_quota_ctl(struct btrfs_root *root, void __user *arg)
3431 {
3432 struct btrfs_ioctl_quota_ctl_args *sa;
3433 struct btrfs_trans_handle *trans = NULL;
3434 int ret;
3435 int err;
3436
3437 if (!capable(CAP_SYS_ADMIN))
3438 return -EPERM;
3439
3440 if (root->fs_info->sb->s_flags & MS_RDONLY)
3441 return -EROFS;
3442
3443 sa = memdup_user(arg, sizeof(*sa));
3444 if (IS_ERR(sa))
3445 return PTR_ERR(sa);
3446
3447 if (sa->cmd != BTRFS_QUOTA_CTL_RESCAN) {
3448 trans = btrfs_start_transaction(root, 2);
3449 if (IS_ERR(trans)) {
3450 ret = PTR_ERR(trans);
3451 goto out;
3452 }
3453 }
3454
3455 switch (sa->cmd) {
3456 case BTRFS_QUOTA_CTL_ENABLE:
3457 ret = btrfs_quota_enable(trans, root->fs_info);
3458 break;
3459 case BTRFS_QUOTA_CTL_DISABLE:
3460 ret = btrfs_quota_disable(trans, root->fs_info);
3461 break;
3462 case BTRFS_QUOTA_CTL_RESCAN:
3463 ret = btrfs_quota_rescan(root->fs_info);
3464 break;
3465 default:
3466 ret = -EINVAL;
3467 break;
3468 }
3469
3470 if (copy_to_user(arg, sa, sizeof(*sa)))
3471 ret = -EFAULT;
3472
3473 if (trans) {
3474 err = btrfs_commit_transaction(trans, root);
3475 if (err && !ret)
3476 ret = err;
3477 }
3478
3479 out:
3480 kfree(sa);
3481 return ret;
3482 }
3483
3484 static long btrfs_ioctl_qgroup_assign(struct btrfs_root *root, void __user *arg)
3485 {
3486 struct btrfs_ioctl_qgroup_assign_args *sa;
3487 struct btrfs_trans_handle *trans;
3488 int ret;
3489 int err;
3490
3491 if (!capable(CAP_SYS_ADMIN))
3492 return -EPERM;
3493
3494 if (root->fs_info->sb->s_flags & MS_RDONLY)
3495 return -EROFS;
3496
3497 sa = memdup_user(arg, sizeof(*sa));
3498 if (IS_ERR(sa))
3499 return PTR_ERR(sa);
3500
3501 trans = btrfs_join_transaction(root);
3502 if (IS_ERR(trans)) {
3503 ret = PTR_ERR(trans);
3504 goto out;
3505 }
3506
3507 /* FIXME: check if the IDs really exist */
3508 if (sa->assign) {
3509 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
3510 sa->src, sa->dst);
3511 } else {
3512 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
3513 sa->src, sa->dst);
3514 }
3515
3516 err = btrfs_end_transaction(trans, root);
3517 if (err && !ret)
3518 ret = err;
3519
3520 out:
3521 kfree(sa);
3522 return ret;
3523 }
3524
3525 static long btrfs_ioctl_qgroup_create(struct btrfs_root *root, void __user *arg)
3526 {
3527 struct btrfs_ioctl_qgroup_create_args *sa;
3528 struct btrfs_trans_handle *trans;
3529 int ret;
3530 int err;
3531
3532 if (!capable(CAP_SYS_ADMIN))
3533 return -EPERM;
3534
3535 if (root->fs_info->sb->s_flags & MS_RDONLY)
3536 return -EROFS;
3537
3538 sa = memdup_user(arg, sizeof(*sa));
3539 if (IS_ERR(sa))
3540 return PTR_ERR(sa);
3541
3542 trans = btrfs_join_transaction(root);
3543 if (IS_ERR(trans)) {
3544 ret = PTR_ERR(trans);
3545 goto out;
3546 }
3547
3548 /* FIXME: check if the IDs really exist */
3549 if (sa->create) {
3550 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
3551 NULL);
3552 } else {
3553 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
3554 }
3555
3556 err = btrfs_end_transaction(trans, root);
3557 if (err && !ret)
3558 ret = err;
3559
3560 out:
3561 kfree(sa);
3562 return ret;
3563 }
3564
3565 static long btrfs_ioctl_qgroup_limit(struct btrfs_root *root, void __user *arg)
3566 {
3567 struct btrfs_ioctl_qgroup_limit_args *sa;
3568 struct btrfs_trans_handle *trans;
3569 int ret;
3570 int err;
3571 u64 qgroupid;
3572
3573 if (!capable(CAP_SYS_ADMIN))
3574 return -EPERM;
3575
3576 if (root->fs_info->sb->s_flags & MS_RDONLY)
3577 return -EROFS;
3578
3579 sa = memdup_user(arg, sizeof(*sa));
3580 if (IS_ERR(sa))
3581 return PTR_ERR(sa);
3582
3583 trans = btrfs_join_transaction(root);
3584 if (IS_ERR(trans)) {
3585 ret = PTR_ERR(trans);
3586 goto out;
3587 }
3588
3589 qgroupid = sa->qgroupid;
3590 if (!qgroupid) {
3591 /* take the current subvol as qgroup */
3592 qgroupid = root->root_key.objectid;
3593 }
3594
3595 /* FIXME: check if the IDs really exist */
3596 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
3597
3598 err = btrfs_end_transaction(trans, root);
3599 if (err && !ret)
3600 ret = err;
3601
3602 out:
3603 kfree(sa);
3604 return ret;
3605 }
3606
3607 static long btrfs_ioctl_set_received_subvol(struct file *file,
3608 void __user *arg)
3609 {
3610 struct btrfs_ioctl_received_subvol_args *sa = NULL;
3611 struct inode *inode = fdentry(file)->d_inode;
3612 struct btrfs_root *root = BTRFS_I(inode)->root;
3613 struct btrfs_root_item *root_item = &root->root_item;
3614 struct btrfs_trans_handle *trans;
3615 struct timespec ct = CURRENT_TIME;
3616 int ret = 0;
3617
3618 ret = mnt_want_write_file(file);
3619 if (ret < 0)
3620 return ret;
3621
3622 down_write(&root->fs_info->subvol_sem);
3623
3624 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
3625 ret = -EINVAL;
3626 goto out;
3627 }
3628
3629 if (btrfs_root_readonly(root)) {
3630 ret = -EROFS;
3631 goto out;
3632 }
3633
3634 if (!inode_owner_or_capable(inode)) {
3635 ret = -EACCES;
3636 goto out;
3637 }
3638
3639 sa = memdup_user(arg, sizeof(*sa));
3640 if (IS_ERR(sa)) {
3641 ret = PTR_ERR(sa);
3642 sa = NULL;
3643 goto out;
3644 }
3645
3646 trans = btrfs_start_transaction(root, 1);
3647 if (IS_ERR(trans)) {
3648 ret = PTR_ERR(trans);
3649 trans = NULL;
3650 goto out;
3651 }
3652
3653 sa->rtransid = trans->transid;
3654 sa->rtime.sec = ct.tv_sec;
3655 sa->rtime.nsec = ct.tv_nsec;
3656
3657 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
3658 btrfs_set_root_stransid(root_item, sa->stransid);
3659 btrfs_set_root_rtransid(root_item, sa->rtransid);
3660 root_item->stime.sec = cpu_to_le64(sa->stime.sec);
3661 root_item->stime.nsec = cpu_to_le32(sa->stime.nsec);
3662 root_item->rtime.sec = cpu_to_le64(sa->rtime.sec);
3663 root_item->rtime.nsec = cpu_to_le32(sa->rtime.nsec);
3664
3665 ret = btrfs_update_root(trans, root->fs_info->tree_root,
3666 &root->root_key, &root->root_item);
3667 if (ret < 0) {
3668 btrfs_end_transaction(trans, root);
3669 trans = NULL;
3670 goto out;
3671 } else {
3672 ret = btrfs_commit_transaction(trans, root);
3673 if (ret < 0)
3674 goto out;
3675 }
3676
3677 ret = copy_to_user(arg, sa, sizeof(*sa));
3678 if (ret)
3679 ret = -EFAULT;
3680
3681 out:
3682 kfree(sa);
3683 up_write(&root->fs_info->subvol_sem);
3684 mnt_drop_write_file(file);
3685 return ret;
3686 }
3687
3688 long btrfs_ioctl(struct file *file, unsigned int
3689 cmd, unsigned long arg)
3690 {
3691 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3692 void __user *argp = (void __user *)arg;
3693
3694 switch (cmd) {
3695 case FS_IOC_GETFLAGS:
3696 return btrfs_ioctl_getflags(file, argp);
3697 case FS_IOC_SETFLAGS:
3698 return btrfs_ioctl_setflags(file, argp);
3699 case FS_IOC_GETVERSION:
3700 return btrfs_ioctl_getversion(file, argp);
3701 case FITRIM:
3702 return btrfs_ioctl_fitrim(file, argp);
3703 case BTRFS_IOC_SNAP_CREATE:
3704 return btrfs_ioctl_snap_create(file, argp, 0);
3705 case BTRFS_IOC_SNAP_CREATE_V2:
3706 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3707 case BTRFS_IOC_SUBVOL_CREATE:
3708 return btrfs_ioctl_snap_create(file, argp, 1);
3709 case BTRFS_IOC_SUBVOL_CREATE_V2:
3710 return btrfs_ioctl_snap_create_v2(file, argp, 1);
3711 case BTRFS_IOC_SNAP_DESTROY:
3712 return btrfs_ioctl_snap_destroy(file, argp);
3713 case BTRFS_IOC_SUBVOL_GETFLAGS:
3714 return btrfs_ioctl_subvol_getflags(file, argp);
3715 case BTRFS_IOC_SUBVOL_SETFLAGS:
3716 return btrfs_ioctl_subvol_setflags(file, argp);
3717 case BTRFS_IOC_DEFAULT_SUBVOL:
3718 return btrfs_ioctl_default_subvol(file, argp);
3719 case BTRFS_IOC_DEFRAG:
3720 return btrfs_ioctl_defrag(file, NULL);
3721 case BTRFS_IOC_DEFRAG_RANGE:
3722 return btrfs_ioctl_defrag(file, argp);
3723 case BTRFS_IOC_RESIZE:
3724 return btrfs_ioctl_resize(root, argp);
3725 case BTRFS_IOC_ADD_DEV:
3726 return btrfs_ioctl_add_dev(root, argp);
3727 case BTRFS_IOC_RM_DEV:
3728 return btrfs_ioctl_rm_dev(root, argp);
3729 case BTRFS_IOC_FS_INFO:
3730 return btrfs_ioctl_fs_info(root, argp);
3731 case BTRFS_IOC_DEV_INFO:
3732 return btrfs_ioctl_dev_info(root, argp);
3733 case BTRFS_IOC_BALANCE:
3734 return btrfs_ioctl_balance(file, NULL);
3735 case BTRFS_IOC_CLONE:
3736 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
3737 case BTRFS_IOC_CLONE_RANGE:
3738 return btrfs_ioctl_clone_range(file, argp);
3739 case BTRFS_IOC_TRANS_START:
3740 return btrfs_ioctl_trans_start(file);
3741 case BTRFS_IOC_TRANS_END:
3742 return btrfs_ioctl_trans_end(file);
3743 case BTRFS_IOC_TREE_SEARCH:
3744 return btrfs_ioctl_tree_search(file, argp);
3745 case BTRFS_IOC_INO_LOOKUP:
3746 return btrfs_ioctl_ino_lookup(file, argp);
3747 case BTRFS_IOC_INO_PATHS:
3748 return btrfs_ioctl_ino_to_path(root, argp);
3749 case BTRFS_IOC_LOGICAL_INO:
3750 return btrfs_ioctl_logical_to_ino(root, argp);
3751 case BTRFS_IOC_SPACE_INFO:
3752 return btrfs_ioctl_space_info(root, argp);
3753 case BTRFS_IOC_SYNC:
3754 btrfs_sync_fs(file->f_dentry->d_sb, 1);
3755 return 0;
3756 case BTRFS_IOC_START_SYNC:
3757 return btrfs_ioctl_start_sync(file, argp);
3758 case BTRFS_IOC_WAIT_SYNC:
3759 return btrfs_ioctl_wait_sync(file, argp);
3760 case BTRFS_IOC_SCRUB:
3761 return btrfs_ioctl_scrub(root, argp);
3762 case BTRFS_IOC_SCRUB_CANCEL:
3763 return btrfs_ioctl_scrub_cancel(root, argp);
3764 case BTRFS_IOC_SCRUB_PROGRESS:
3765 return btrfs_ioctl_scrub_progress(root, argp);
3766 case BTRFS_IOC_BALANCE_V2:
3767 return btrfs_ioctl_balance(file, argp);
3768 case BTRFS_IOC_BALANCE_CTL:
3769 return btrfs_ioctl_balance_ctl(root, arg);
3770 case BTRFS_IOC_BALANCE_PROGRESS:
3771 return btrfs_ioctl_balance_progress(root, argp);
3772 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
3773 return btrfs_ioctl_set_received_subvol(file, argp);
3774 case BTRFS_IOC_SEND:
3775 return btrfs_ioctl_send(file, argp);
3776 case BTRFS_IOC_GET_DEV_STATS:
3777 return btrfs_ioctl_get_dev_stats(root, argp);
3778 case BTRFS_IOC_QUOTA_CTL:
3779 return btrfs_ioctl_quota_ctl(root, argp);
3780 case BTRFS_IOC_QGROUP_ASSIGN:
3781 return btrfs_ioctl_qgroup_assign(root, argp);
3782 case BTRFS_IOC_QGROUP_CREATE:
3783 return btrfs_ioctl_qgroup_create(root, argp);
3784 case BTRFS_IOC_QGROUP_LIMIT:
3785 return btrfs_ioctl_qgroup_limit(root, argp);
3786 }
3787
3788 return -ENOTTY;
3789 }
kernel source for telekom puls (alcatel/mediatek)