projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Btrfs: disallow mutually exclusive admin operations from user mode
[deliverable/linux.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 umode_t mode;
185
186 if (btrfs_root_readonly(root))
187 return -EROFS;
188
189 if (copy_from_user(&flags, arg, sizeof(flags)))
190 return -EFAULT;
191
192 ret = check_flags(flags);
193 if (ret)
194 return ret;
195
196 if (!inode_owner_or_capable(inode))
197 return -EACCES;
198
199 ret = mnt_want_write_file(file);
200 if (ret)
201 return ret;
202
203 mutex_lock(&inode->i_mutex);
204
205 ip_oldflags = ip->flags;
206 i_oldflags = inode->i_flags;
207 mode = inode->i_mode;
208
209 flags = btrfs_mask_flags(inode->i_mode, flags);
210 oldflags = btrfs_flags_to_ioctl(ip->flags);
211 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
212 if (!capable(CAP_LINUX_IMMUTABLE)) {
213 ret = -EPERM;
214 goto out_unlock;
215 }
216 }
217
218 if (flags & FS_SYNC_FL)
219 ip->flags |= BTRFS_INODE_SYNC;
220 else
221 ip->flags &= ~BTRFS_INODE_SYNC;
222 if (flags & FS_IMMUTABLE_FL)
223 ip->flags |= BTRFS_INODE_IMMUTABLE;
224 else
225 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
226 if (flags & FS_APPEND_FL)
227 ip->flags |= BTRFS_INODE_APPEND;
228 else
229 ip->flags &= ~BTRFS_INODE_APPEND;
230 if (flags & FS_NODUMP_FL)
231 ip->flags |= BTRFS_INODE_NODUMP;
232 else
233 ip->flags &= ~BTRFS_INODE_NODUMP;
234 if (flags & FS_NOATIME_FL)
235 ip->flags |= BTRFS_INODE_NOATIME;
236 else
237 ip->flags &= ~BTRFS_INODE_NOATIME;
238 if (flags & FS_DIRSYNC_FL)
239 ip->flags |= BTRFS_INODE_DIRSYNC;
240 else
241 ip->flags &= ~BTRFS_INODE_DIRSYNC;
242 if (flags & FS_NOCOW_FL) {
243 if (S_ISREG(mode)) {
244 /*
245 * It's safe to turn csums off here, no extents exist.
246 * Otherwise we want the flag to reflect the real COW
247 * status of the file and will not set it.
248 */
249 if (inode->i_size == 0)
250 ip->flags |= BTRFS_INODE_NODATACOW
251 | BTRFS_INODE_NODATASUM;
252 } else {
253 ip->flags |= BTRFS_INODE_NODATACOW;
254 }
255 } else {
256 /*
257 * Revert back under same assuptions as above
258 */
259 if (S_ISREG(mode)) {
260 if (inode->i_size == 0)
261 ip->flags &= ~(BTRFS_INODE_NODATACOW
262 | BTRFS_INODE_NODATASUM);
263 } else {
264 ip->flags &= ~BTRFS_INODE_NODATACOW;
265 }
266 }
267
268 /*
269 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
270 * flag may be changed automatically if compression code won't make
271 * things smaller.
272 */
273 if (flags & FS_NOCOMP_FL) {
274 ip->flags &= ~BTRFS_INODE_COMPRESS;
275 ip->flags |= BTRFS_INODE_NOCOMPRESS;
276 } else if (flags & FS_COMPR_FL) {
277 ip->flags |= BTRFS_INODE_COMPRESS;
278 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
279 } else {
280 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
281 }
282
283 trans = btrfs_start_transaction(root, 1);
284 if (IS_ERR(trans)) {
285 ret = PTR_ERR(trans);
286 goto out_drop;
287 }
288
289 btrfs_update_iflags(inode);
290 inode_inc_iversion(inode);
291 inode->i_ctime = CURRENT_TIME;
292 ret = btrfs_update_inode(trans, root, inode);
293
294 btrfs_end_transaction(trans, root);
295 out_drop:
296 if (ret) {
297 ip->flags = ip_oldflags;
298 inode->i_flags = i_oldflags;
299 }
300
301 out_unlock:
302 mutex_unlock(&inode->i_mutex);
303 mnt_drop_write_file(file);
304 return ret;
305 }
306
307 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
308 {
309 struct inode *inode = file->f_path.dentry->d_inode;
310
311 return put_user(inode->i_generation, arg);
312 }
313
314 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
315 {
316 struct btrfs_fs_info *fs_info = btrfs_sb(fdentry(file)->d_sb);
317 struct btrfs_device *device;
318 struct request_queue *q;
319 struct fstrim_range range;
320 u64 minlen = ULLONG_MAX;
321 u64 num_devices = 0;
322 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
323 int ret;
324
325 if (!capable(CAP_SYS_ADMIN))
326 return -EPERM;
327
328 rcu_read_lock();
329 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
330 dev_list) {
331 if (!device->bdev)
332 continue;
333 q = bdev_get_queue(device->bdev);
334 if (blk_queue_discard(q)) {
335 num_devices++;
336 minlen = min((u64)q->limits.discard_granularity,
337 minlen);
338 }
339 }
340 rcu_read_unlock();
341
342 if (!num_devices)
343 return -EOPNOTSUPP;
344 if (copy_from_user(&range, arg, sizeof(range)))
345 return -EFAULT;
346 if (range.start > total_bytes ||
347 range.len < fs_info->sb->s_blocksize)
348 return -EINVAL;
349
350 range.len = min(range.len, total_bytes - range.start);
351 range.minlen = max(range.minlen, minlen);
352 ret = btrfs_trim_fs(fs_info->tree_root, &range);
353 if (ret < 0)
354 return ret;
355
356 if (copy_to_user(arg, &range, sizeof(range)))
357 return -EFAULT;
358
359 return 0;
360 }
361
362 static noinline int create_subvol(struct btrfs_root *root,
363 struct dentry *dentry,
364 char *name, int namelen,
365 u64 *async_transid,
366 struct btrfs_qgroup_inherit **inherit)
367 {
368 struct btrfs_trans_handle *trans;
369 struct btrfs_key key;
370 struct btrfs_root_item root_item;
371 struct btrfs_inode_item *inode_item;
372 struct extent_buffer *leaf;
373 struct btrfs_root *new_root;
374 struct dentry *parent = dentry->d_parent;
375 struct inode *dir;
376 struct timespec cur_time = CURRENT_TIME;
377 int ret;
378 int err;
379 u64 objectid;
380 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
381 u64 index = 0;
382 uuid_le new_uuid;
383
384 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
385 if (ret)
386 return ret;
387
388 dir = parent->d_inode;
389
390 /*
391 * 1 - inode item
392 * 2 - refs
393 * 1 - root item
394 * 2 - dir items
395 */
396 trans = btrfs_start_transaction(root, 6);
397 if (IS_ERR(trans))
398 return PTR_ERR(trans);
399
400 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid,
401 inherit ? *inherit : NULL);
402 if (ret)
403 goto fail;
404
405 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
406 0, objectid, NULL, 0, 0, 0);
407 if (IS_ERR(leaf)) {
408 ret = PTR_ERR(leaf);
409 goto fail;
410 }
411
412 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
413 btrfs_set_header_bytenr(leaf, leaf->start);
414 btrfs_set_header_generation(leaf, trans->transid);
415 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
416 btrfs_set_header_owner(leaf, objectid);
417
418 write_extent_buffer(leaf, root->fs_info->fsid,
419 (unsigned long)btrfs_header_fsid(leaf),
420 BTRFS_FSID_SIZE);
421 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
422 (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
423 BTRFS_UUID_SIZE);
424 btrfs_mark_buffer_dirty(leaf);
425
426 memset(&root_item, 0, sizeof(root_item));
427
428 inode_item = &root_item.inode;
429 inode_item->generation = cpu_to_le64(1);
430 inode_item->size = cpu_to_le64(3);
431 inode_item->nlink = cpu_to_le32(1);
432 inode_item->nbytes = cpu_to_le64(root->leafsize);
433 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
434
435 root_item.flags = 0;
436 root_item.byte_limit = 0;
437 inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
438
439 btrfs_set_root_bytenr(&root_item, leaf->start);
440 btrfs_set_root_generation(&root_item, trans->transid);
441 btrfs_set_root_level(&root_item, 0);
442 btrfs_set_root_refs(&root_item, 1);
443 btrfs_set_root_used(&root_item, leaf->len);
444 btrfs_set_root_last_snapshot(&root_item, 0);
445
446 btrfs_set_root_generation_v2(&root_item,
447 btrfs_root_generation(&root_item));
448 uuid_le_gen(&new_uuid);
449 memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
450 root_item.otime.sec = cpu_to_le64(cur_time.tv_sec);
451 root_item.otime.nsec = cpu_to_le32(cur_time.tv_nsec);
452 root_item.ctime = root_item.otime;
453 btrfs_set_root_ctransid(&root_item, trans->transid);
454 btrfs_set_root_otransid(&root_item, trans->transid);
455
456 btrfs_tree_unlock(leaf);
457 free_extent_buffer(leaf);
458 leaf = NULL;
459
460 btrfs_set_root_dirid(&root_item, new_dirid);
461
462 key.objectid = objectid;
463 key.offset = 0;
464 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
465 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
466 &root_item);
467 if (ret)
468 goto fail;
469
470 key.offset = (u64)-1;
471 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
472 if (IS_ERR(new_root)) {
473 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
474 ret = PTR_ERR(new_root);
475 goto fail;
476 }
477
478 btrfs_record_root_in_trans(trans, new_root);
479
480 ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
481 if (ret) {
482 /* We potentially lose an unused inode item here */
483 btrfs_abort_transaction(trans, root, ret);
484 goto fail;
485 }
486
487 /*
488 * insert the directory item
489 */
490 ret = btrfs_set_inode_index(dir, &index);
491 if (ret) {
492 btrfs_abort_transaction(trans, root, ret);
493 goto fail;
494 }
495
496 ret = btrfs_insert_dir_item(trans, root,
497 name, namelen, dir, &key,
498 BTRFS_FT_DIR, index);
499 if (ret) {
500 btrfs_abort_transaction(trans, root, ret);
501 goto fail;
502 }
503
504 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
505 ret = btrfs_update_inode(trans, root, dir);
506 BUG_ON(ret);
507
508 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
509 objectid, root->root_key.objectid,
510 btrfs_ino(dir), index, name, namelen);
511
512 BUG_ON(ret);
513
514 d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
515 fail:
516 if (async_transid) {
517 *async_transid = trans->transid;
518 err = btrfs_commit_transaction_async(trans, root, 1);
519 } else {
520 err = btrfs_commit_transaction(trans, root);
521 }
522 if (err && !ret)
523 ret = err;
524 return ret;
525 }
526
527 static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
528 char *name, int namelen, u64 *async_transid,
529 bool readonly, struct btrfs_qgroup_inherit **inherit)
530 {
531 struct inode *inode;
532 struct btrfs_pending_snapshot *pending_snapshot;
533 struct btrfs_trans_handle *trans;
534 int ret;
535
536 if (!root->ref_cows)
537 return -EINVAL;
538
539 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
540 if (!pending_snapshot)
541 return -ENOMEM;
542
543 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
544 BTRFS_BLOCK_RSV_TEMP);
545 pending_snapshot->dentry = dentry;
546 pending_snapshot->root = root;
547 pending_snapshot->readonly = readonly;
548 if (inherit) {
549 pending_snapshot->inherit = *inherit;
550 *inherit = NULL; /* take responsibility to free it */
551 }
552
553 trans = btrfs_start_transaction(root->fs_info->extent_root, 6);
554 if (IS_ERR(trans)) {
555 ret = PTR_ERR(trans);
556 goto fail;
557 }
558
559 ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
560 BUG_ON(ret);
561
562 spin_lock(&root->fs_info->trans_lock);
563 list_add(&pending_snapshot->list,
564 &trans->transaction->pending_snapshots);
565 spin_unlock(&root->fs_info->trans_lock);
566 if (async_transid) {
567 *async_transid = trans->transid;
568 ret = btrfs_commit_transaction_async(trans,
569 root->fs_info->extent_root, 1);
570 } else {
571 ret = btrfs_commit_transaction(trans,
572 root->fs_info->extent_root);
573 }
574 if (ret) {
575 /* cleanup_transaction has freed this for us */
576 if (trans->aborted)
577 pending_snapshot = NULL;
578 goto fail;
579 }
580
581 ret = pending_snapshot->error;
582 if (ret)
583 goto fail;
584
585 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
586 if (ret)
587 goto fail;
588
589 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
590 if (IS_ERR(inode)) {
591 ret = PTR_ERR(inode);
592 goto fail;
593 }
594 BUG_ON(!inode);
595 d_instantiate(dentry, inode);
596 ret = 0;
597 fail:
598 kfree(pending_snapshot);
599 return ret;
600 }
601
602 /* copy of check_sticky in fs/namei.c()
603 * It's inline, so penalty for filesystems that don't use sticky bit is
604 * minimal.
605 */
606 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
607 {
608 kuid_t fsuid = current_fsuid();
609
610 if (!(dir->i_mode & S_ISVTX))
611 return 0;
612 if (uid_eq(inode->i_uid, fsuid))
613 return 0;
614 if (uid_eq(dir->i_uid, fsuid))
615 return 0;
616 return !capable(CAP_FOWNER);
617 }
618
619 /* copy of may_delete in fs/namei.c()
620 * Check whether we can remove a link victim from directory dir, check
621 * whether the type of victim is right.
622 * 1. We can't do it if dir is read-only (done in permission())
623 * 2. We should have write and exec permissions on dir
624 * 3. We can't remove anything from append-only dir
625 * 4. We can't do anything with immutable dir (done in permission())
626 * 5. If the sticky bit on dir is set we should either
627 * a. be owner of dir, or
628 * b. be owner of victim, or
629 * c. have CAP_FOWNER capability
630 * 6. If the victim is append-only or immutable we can't do antyhing with
631 * links pointing to it.
632 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
633 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
634 * 9. We can't remove a root or mountpoint.
635 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
636 * nfs_async_unlink().
637 */
638
639 static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
640 {
641 int error;
642
643 if (!victim->d_inode)
644 return -ENOENT;
645
646 BUG_ON(victim->d_parent->d_inode != dir);
647 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
648
649 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
650 if (error)
651 return error;
652 if (IS_APPEND(dir))
653 return -EPERM;
654 if (btrfs_check_sticky(dir, victim->d_inode)||
655 IS_APPEND(victim->d_inode)||
656 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
657 return -EPERM;
658 if (isdir) {
659 if (!S_ISDIR(victim->d_inode->i_mode))
660 return -ENOTDIR;
661 if (IS_ROOT(victim))
662 return -EBUSY;
663 } else if (S_ISDIR(victim->d_inode->i_mode))
664 return -EISDIR;
665 if (IS_DEADDIR(dir))
666 return -ENOENT;
667 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
668 return -EBUSY;
669 return 0;
670 }
671
672 /* copy of may_create in fs/namei.c() */
673 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
674 {
675 if (child->d_inode)
676 return -EEXIST;
677 if (IS_DEADDIR(dir))
678 return -ENOENT;
679 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
680 }
681
682 /*
683 * Create a new subvolume below @parent. This is largely modeled after
684 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
685 * inside this filesystem so it's quite a bit simpler.
686 */
687 static noinline int btrfs_mksubvol(struct path *parent,
688 char *name, int namelen,
689 struct btrfs_root *snap_src,
690 u64 *async_transid, bool readonly,
691 struct btrfs_qgroup_inherit **inherit)
692 {
693 struct inode *dir = parent->dentry->d_inode;
694 struct dentry *dentry;
695 int error;
696
697 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
698
699 dentry = lookup_one_len(name, parent->dentry, namelen);
700 error = PTR_ERR(dentry);
701 if (IS_ERR(dentry))
702 goto out_unlock;
703
704 error = -EEXIST;
705 if (dentry->d_inode)
706 goto out_dput;
707
708 error = btrfs_may_create(dir, dentry);
709 if (error)
710 goto out_dput;
711
712 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
713
714 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
715 goto out_up_read;
716
717 if (snap_src) {
718 error = create_snapshot(snap_src, dentry, name, namelen,
719 async_transid, readonly, inherit);
720 } else {
721 error = create_subvol(BTRFS_I(dir)->root, dentry,
722 name, namelen, async_transid, inherit);
723 }
724 if (!error)
725 fsnotify_mkdir(dir, dentry);
726 out_up_read:
727 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
728 out_dput:
729 dput(dentry);
730 out_unlock:
731 mutex_unlock(&dir->i_mutex);
732 return error;
733 }
734
735 /*
736 * When we're defragging a range, we don't want to kick it off again
737 * if it is really just waiting for delalloc to send it down.
738 * If we find a nice big extent or delalloc range for the bytes in the
739 * file you want to defrag, we return 0 to let you know to skip this
740 * part of the file
741 */
742 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
743 {
744 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
745 struct extent_map *em = NULL;
746 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
747 u64 end;
748
749 read_lock(&em_tree->lock);
750 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
751 read_unlock(&em_tree->lock);
752
753 if (em) {
754 end = extent_map_end(em);
755 free_extent_map(em);
756 if (end - offset > thresh)
757 return 0;
758 }
759 /* if we already have a nice delalloc here, just stop */
760 thresh /= 2;
761 end = count_range_bits(io_tree, &offset, offset + thresh,
762 thresh, EXTENT_DELALLOC, 1);
763 if (end >= thresh)
764 return 0;
765 return 1;
766 }
767
768 /*
769 * helper function to walk through a file and find extents
770 * newer than a specific transid, and smaller than thresh.
771 *
772 * This is used by the defragging code to find new and small
773 * extents
774 */
775 static int find_new_extents(struct btrfs_root *root,
776 struct inode *inode, u64 newer_than,
777 u64 *off, int thresh)
778 {
779 struct btrfs_path *path;
780 struct btrfs_key min_key;
781 struct btrfs_key max_key;
782 struct extent_buffer *leaf;
783 struct btrfs_file_extent_item *extent;
784 int type;
785 int ret;
786 u64 ino = btrfs_ino(inode);
787
788 path = btrfs_alloc_path();
789 if (!path)
790 return -ENOMEM;
791
792 min_key.objectid = ino;
793 min_key.type = BTRFS_EXTENT_DATA_KEY;
794 min_key.offset = *off;
795
796 max_key.objectid = ino;
797 max_key.type = (u8)-1;
798 max_key.offset = (u64)-1;
799
800 path->keep_locks = 1;
801
802 while(1) {
803 ret = btrfs_search_forward(root, &min_key, &max_key,
804 path, 0, newer_than);
805 if (ret != 0)
806 goto none;
807 if (min_key.objectid != ino)
808 goto none;
809 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
810 goto none;
811
812 leaf = path->nodes[0];
813 extent = btrfs_item_ptr(leaf, path->slots[0],
814 struct btrfs_file_extent_item);
815
816 type = btrfs_file_extent_type(leaf, extent);
817 if (type == BTRFS_FILE_EXTENT_REG &&
818 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
819 check_defrag_in_cache(inode, min_key.offset, thresh)) {
820 *off = min_key.offset;
821 btrfs_free_path(path);
822 return 0;
823 }
824
825 if (min_key.offset == (u64)-1)
826 goto none;
827
828 min_key.offset++;
829 btrfs_release_path(path);
830 }
831 none:
832 btrfs_free_path(path);
833 return -ENOENT;
834 }
835
836 static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
837 {
838 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
839 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
840 struct extent_map *em;
841 u64 len = PAGE_CACHE_SIZE;
842
843 /*
844 * hopefully we have this extent in the tree already, try without
845 * the full extent lock
846 */
847 read_lock(&em_tree->lock);
848 em = lookup_extent_mapping(em_tree, start, len);
849 read_unlock(&em_tree->lock);
850
851 if (!em) {
852 /* get the big lock and read metadata off disk */
853 lock_extent(io_tree, start, start + len - 1);
854 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
855 unlock_extent(io_tree, start, start + len - 1);
856
857 if (IS_ERR(em))
858 return NULL;
859 }
860
861 return em;
862 }
863
864 static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
865 {
866 struct extent_map *next;
867 bool ret = true;
868
869 /* this is the last extent */
870 if (em->start + em->len >= i_size_read(inode))
871 return false;
872
873 next = defrag_lookup_extent(inode, em->start + em->len);
874 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
875 ret = false;
876
877 free_extent_map(next);
878 return ret;
879 }
880
881 static int should_defrag_range(struct inode *inode, u64 start, int thresh,
882 u64 *last_len, u64 *skip, u64 *defrag_end,
883 int compress)
884 {
885 struct extent_map *em;
886 int ret = 1;
887 bool next_mergeable = true;
888
889 /*
890 * make sure that once we start defragging an extent, we keep on
891 * defragging it
892 */
893 if (start < *defrag_end)
894 return 1;
895
896 *skip = 0;
897
898 em = defrag_lookup_extent(inode, start);
899 if (!em)
900 return 0;
901
902 /* this will cover holes, and inline extents */
903 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
904 ret = 0;
905 goto out;
906 }
907
908 next_mergeable = defrag_check_next_extent(inode, em);
909
910 /*
911 * we hit a real extent, if it is big or the next extent is not a
912 * real extent, don't bother defragging it
913 */
914 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
915 (em->len >= thresh || !next_mergeable))
916 ret = 0;
917 out:
918 /*
919 * last_len ends up being a counter of how many bytes we've defragged.
920 * every time we choose not to defrag an extent, we reset *last_len
921 * so that the next tiny extent will force a defrag.
922 *
923 * The end result of this is that tiny extents before a single big
924 * extent will force at least part of that big extent to be defragged.
925 */
926 if (ret) {
927 *defrag_end = extent_map_end(em);
928 } else {
929 *last_len = 0;
930 *skip = extent_map_end(em);
931 *defrag_end = 0;
932 }
933
934 free_extent_map(em);
935 return ret;
936 }
937
938 /*
939 * it doesn't do much good to defrag one or two pages
940 * at a time. This pulls in a nice chunk of pages
941 * to COW and defrag.
942 *
943 * It also makes sure the delalloc code has enough
944 * dirty data to avoid making new small extents as part
945 * of the defrag
946 *
947 * It's a good idea to start RA on this range
948 * before calling this.
949 */
950 static int cluster_pages_for_defrag(struct inode *inode,
951 struct page **pages,
952 unsigned long start_index,
953 int num_pages)
954 {
955 unsigned long file_end;
956 u64 isize = i_size_read(inode);
957 u64 page_start;
958 u64 page_end;
959 u64 page_cnt;
960 int ret;
961 int i;
962 int i_done;
963 struct btrfs_ordered_extent *ordered;
964 struct extent_state *cached_state = NULL;
965 struct extent_io_tree *tree;
966 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
967
968 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
969 if (!isize || start_index > file_end)
970 return 0;
971
972 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
973
974 ret = btrfs_delalloc_reserve_space(inode,
975 page_cnt << PAGE_CACHE_SHIFT);
976 if (ret)
977 return ret;
978 i_done = 0;
979 tree = &BTRFS_I(inode)->io_tree;
980
981 /* step one, lock all the pages */
982 for (i = 0; i < page_cnt; i++) {
983 struct page *page;
984 again:
985 page = find_or_create_page(inode->i_mapping,
986 start_index + i, mask);
987 if (!page)
988 break;
989
990 page_start = page_offset(page);
991 page_end = page_start + PAGE_CACHE_SIZE - 1;
992 while (1) {
993 lock_extent(tree, page_start, page_end);
994 ordered = btrfs_lookup_ordered_extent(inode,
995 page_start);
996 unlock_extent(tree, page_start, page_end);
997 if (!ordered)
998 break;
999
1000 unlock_page(page);
1001 btrfs_start_ordered_extent(inode, ordered, 1);
1002 btrfs_put_ordered_extent(ordered);
1003 lock_page(page);
1004 /*
1005 * we unlocked the page above, so we need check if
1006 * it was released or not.
1007 */
1008 if (page->mapping != inode->i_mapping) {
1009 unlock_page(page);
1010 page_cache_release(page);
1011 goto again;
1012 }
1013 }
1014
1015 if (!PageUptodate(page)) {
1016 btrfs_readpage(NULL, page);
1017 lock_page(page);
1018 if (!PageUptodate(page)) {
1019 unlock_page(page);
1020 page_cache_release(page);
1021 ret = -EIO;
1022 break;
1023 }
1024 }
1025
1026 if (page->mapping != inode->i_mapping) {
1027 unlock_page(page);
1028 page_cache_release(page);
1029 goto again;
1030 }
1031
1032 pages[i] = page;
1033 i_done++;
1034 }
1035 if (!i_done || ret)
1036 goto out;
1037
1038 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1039 goto out;
1040
1041 /*
1042 * so now we have a nice long stream of locked
1043 * and up to date pages, lets wait on them
1044 */
1045 for (i = 0; i < i_done; i++)
1046 wait_on_page_writeback(pages[i]);
1047
1048 page_start = page_offset(pages[0]);
1049 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1050
1051 lock_extent_bits(&BTRFS_I(inode)->io_tree,
1052 page_start, page_end - 1, 0, &cached_state);
1053 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1054 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1055 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1056 &cached_state, GFP_NOFS);
1057
1058 if (i_done != page_cnt) {
1059 spin_lock(&BTRFS_I(inode)->lock);
1060 BTRFS_I(inode)->outstanding_extents++;
1061 spin_unlock(&BTRFS_I(inode)->lock);
1062 btrfs_delalloc_release_space(inode,
1063 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
1064 }
1065
1066
1067 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1068 &cached_state, GFP_NOFS);
1069
1070 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1071 page_start, page_end - 1, &cached_state,
1072 GFP_NOFS);
1073
1074 for (i = 0; i < i_done; i++) {
1075 clear_page_dirty_for_io(pages[i]);
1076 ClearPageChecked(pages[i]);
1077 set_page_extent_mapped(pages[i]);
1078 set_page_dirty(pages[i]);
1079 unlock_page(pages[i]);
1080 page_cache_release(pages[i]);
1081 }
1082 return i_done;
1083 out:
1084 for (i = 0; i < i_done; i++) {
1085 unlock_page(pages[i]);
1086 page_cache_release(pages[i]);
1087 }
1088 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
1089 return ret;
1090
1091 }
1092
1093 int btrfs_defrag_file(struct inode *inode, struct file *file,
1094 struct btrfs_ioctl_defrag_range_args *range,
1095 u64 newer_than, unsigned long max_to_defrag)
1096 {
1097 struct btrfs_root *root = BTRFS_I(inode)->root;
1098 struct file_ra_state *ra = NULL;
1099 unsigned long last_index;
1100 u64 isize = i_size_read(inode);
1101 u64 last_len = 0;
1102 u64 skip = 0;
1103 u64 defrag_end = 0;
1104 u64 newer_off = range->start;
1105 unsigned long i;
1106 unsigned long ra_index = 0;
1107 int ret;
1108 int defrag_count = 0;
1109 int compress_type = BTRFS_COMPRESS_ZLIB;
1110 int extent_thresh = range->extent_thresh;
1111 int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1112 int cluster = max_cluster;
1113 u64 new_align = ~((u64)128 * 1024 - 1);
1114 struct page **pages = NULL;
1115
1116 if (extent_thresh == 0)
1117 extent_thresh = 256 * 1024;
1118
1119 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1120 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1121 return -EINVAL;
1122 if (range->compress_type)
1123 compress_type = range->compress_type;
1124 }
1125
1126 if (isize == 0)
1127 return 0;
1128
1129 /*
1130 * if we were not given a file, allocate a readahead
1131 * context
1132 */
1133 if (!file) {
1134 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1135 if (!ra)
1136 return -ENOMEM;
1137 file_ra_state_init(ra, inode->i_mapping);
1138 } else {
1139 ra = &file->f_ra;
1140 }
1141
1142 pages = kmalloc(sizeof(struct page *) * max_cluster,
1143 GFP_NOFS);
1144 if (!pages) {
1145 ret = -ENOMEM;
1146 goto out_ra;
1147 }
1148
1149 /* find the last page to defrag */
1150 if (range->start + range->len > range->start) {
1151 last_index = min_t(u64, isize - 1,
1152 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1153 } else {
1154 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1155 }
1156
1157 if (newer_than) {
1158 ret = find_new_extents(root, inode, newer_than,
1159 &newer_off, 64 * 1024);
1160 if (!ret) {
1161 range->start = newer_off;
1162 /*
1163 * we always align our defrag to help keep
1164 * the extents in the file evenly spaced
1165 */
1166 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1167 } else
1168 goto out_ra;
1169 } else {
1170 i = range->start >> PAGE_CACHE_SHIFT;
1171 }
1172 if (!max_to_defrag)
1173 max_to_defrag = last_index + 1;
1174
1175 /*
1176 * make writeback starts from i, so the defrag range can be
1177 * written sequentially.
1178 */
1179 if (i < inode->i_mapping->writeback_index)
1180 inode->i_mapping->writeback_index = i;
1181
1182 while (i <= last_index && defrag_count < max_to_defrag &&
1183 (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1184 PAGE_CACHE_SHIFT)) {
1185 /*
1186 * make sure we stop running if someone unmounts
1187 * the FS
1188 */
1189 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1190 break;
1191
1192 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1193 extent_thresh, &last_len, &skip,
1194 &defrag_end, range->flags &
1195 BTRFS_DEFRAG_RANGE_COMPRESS)) {
1196 unsigned long next;
1197 /*
1198 * the should_defrag function tells us how much to skip
1199 * bump our counter by the suggested amount
1200 */
1201 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1202 i = max(i + 1, next);
1203 continue;
1204 }
1205
1206 if (!newer_than) {
1207 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1208 PAGE_CACHE_SHIFT) - i;
1209 cluster = min(cluster, max_cluster);
1210 } else {
1211 cluster = max_cluster;
1212 }
1213
1214 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1215 BTRFS_I(inode)->force_compress = compress_type;
1216
1217 if (i + cluster > ra_index) {
1218 ra_index = max(i, ra_index);
1219 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1220 cluster);
1221 ra_index += max_cluster;
1222 }
1223
1224 mutex_lock(&inode->i_mutex);
1225 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
1226 if (ret < 0) {
1227 mutex_unlock(&inode->i_mutex);
1228 goto out_ra;
1229 }
1230
1231 defrag_count += ret;
1232 balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret);
1233 mutex_unlock(&inode->i_mutex);
1234
1235 if (newer_than) {
1236 if (newer_off == (u64)-1)
1237 break;
1238
1239 if (ret > 0)
1240 i += ret;
1241
1242 newer_off = max(newer_off + 1,
1243 (u64)i << PAGE_CACHE_SHIFT);
1244
1245 ret = find_new_extents(root, inode,
1246 newer_than, &newer_off,
1247 64 * 1024);
1248 if (!ret) {
1249 range->start = newer_off;
1250 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1251 } else {
1252 break;
1253 }
1254 } else {
1255 if (ret > 0) {
1256 i += ret;
1257 last_len += ret << PAGE_CACHE_SHIFT;
1258 } else {
1259 i++;
1260 last_len = 0;
1261 }
1262 }
1263 }
1264
1265 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1266 filemap_flush(inode->i_mapping);
1267
1268 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1269 /* the filemap_flush will queue IO into the worker threads, but
1270 * we have to make sure the IO is actually started and that
1271 * ordered extents get created before we return
1272 */
1273 atomic_inc(&root->fs_info->async_submit_draining);
1274 while (atomic_read(&root->fs_info->nr_async_submits) ||
1275 atomic_read(&root->fs_info->async_delalloc_pages)) {
1276 wait_event(root->fs_info->async_submit_wait,
1277 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1278 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1279 }
1280 atomic_dec(&root->fs_info->async_submit_draining);
1281
1282 mutex_lock(&inode->i_mutex);
1283 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1284 mutex_unlock(&inode->i_mutex);
1285 }
1286
1287 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1288 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1289 }
1290
1291 ret = defrag_count;
1292
1293 out_ra:
1294 if (!file)
1295 kfree(ra);
1296 kfree(pages);
1297 return ret;
1298 }
1299
1300 static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
1301 void __user *arg)
1302 {
1303 u64 new_size;
1304 u64 old_size;
1305 u64 devid = 1;
1306 struct btrfs_ioctl_vol_args *vol_args;
1307 struct btrfs_trans_handle *trans;
1308 struct btrfs_device *device = NULL;
1309 char *sizestr;
1310 char *devstr = NULL;
1311 int ret = 0;
1312 int mod = 0;
1313
1314 if (root->fs_info->sb->s_flags & MS_RDONLY)
1315 return -EROFS;
1316
1317 if (!capable(CAP_SYS_ADMIN))
1318 return -EPERM;
1319
1320 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1321 1)) {
1322 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
1323 return -EINPROGRESS;
1324 }
1325
1326 mutex_lock(&root->fs_info->volume_mutex);
1327 vol_args = memdup_user(arg, sizeof(*vol_args));
1328 if (IS_ERR(vol_args)) {
1329 ret = PTR_ERR(vol_args);
1330 goto out;
1331 }
1332
1333 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1334
1335 sizestr = vol_args->name;
1336 devstr = strchr(sizestr, ':');
1337 if (devstr) {
1338 char *end;
1339 sizestr = devstr + 1;
1340 *devstr = '\0';
1341 devstr = vol_args->name;
1342 devid = simple_strtoull(devstr, &end, 10);
1343 printk(KERN_INFO "btrfs: resizing devid %llu\n",
1344 (unsigned long long)devid);
1345 }
1346 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
1347 if (!device) {
1348 printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
1349 (unsigned long long)devid);
1350 ret = -EINVAL;
1351 goto out_free;
1352 }
1353 if (device->fs_devices && device->fs_devices->seeding) {
1354 printk(KERN_INFO "btrfs: resizer unable to apply on "
1355 "seeding device %llu\n",
1356 (unsigned long long)devid);
1357 ret = -EINVAL;
1358 goto out_free;
1359 }
1360
1361 if (!strcmp(sizestr, "max"))
1362 new_size = device->bdev->bd_inode->i_size;
1363 else {
1364 if (sizestr[0] == '-') {
1365 mod = -1;
1366 sizestr++;
1367 } else if (sizestr[0] == '+') {
1368 mod = 1;
1369 sizestr++;
1370 }
1371 new_size = memparse(sizestr, NULL);
1372 if (new_size == 0) {
1373 ret = -EINVAL;
1374 goto out_free;
1375 }
1376 }
1377
1378 old_size = device->total_bytes;
1379
1380 if (mod < 0) {
1381 if (new_size > old_size) {
1382 ret = -EINVAL;
1383 goto out_free;
1384 }
1385 new_size = old_size - new_size;
1386 } else if (mod > 0) {
1387 new_size = old_size + new_size;
1388 }
1389
1390 if (new_size < 256 * 1024 * 1024) {
1391 ret = -EINVAL;
1392 goto out_free;
1393 }
1394 if (new_size > device->bdev->bd_inode->i_size) {
1395 ret = -EFBIG;
1396 goto out_free;
1397 }
1398
1399 do_div(new_size, root->sectorsize);
1400 new_size *= root->sectorsize;
1401
1402 printk_in_rcu(KERN_INFO "btrfs: new size for %s is %llu\n",
1403 rcu_str_deref(device->name),
1404 (unsigned long long)new_size);
1405
1406 if (new_size > old_size) {
1407 trans = btrfs_start_transaction(root, 0);
1408 if (IS_ERR(trans)) {
1409 ret = PTR_ERR(trans);
1410 goto out_free;
1411 }
1412 ret = btrfs_grow_device(trans, device, new_size);
1413 btrfs_commit_transaction(trans, root);
1414 } else if (new_size < old_size) {
1415 ret = btrfs_shrink_device(device, new_size);
1416 } /* equal, nothing need to do */
1417
1418 out_free:
1419 kfree(vol_args);
1420 out:
1421 mutex_unlock(&root->fs_info->volume_mutex);
1422 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
1423 return ret;
1424 }
1425
1426 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1427 char *name, unsigned long fd, int subvol,
1428 u64 *transid, bool readonly,
1429 struct btrfs_qgroup_inherit **inherit)
1430 {
1431 int namelen;
1432 int ret = 0;
1433
1434 ret = mnt_want_write_file(file);
1435 if (ret)
1436 goto out;
1437
1438 namelen = strlen(name);
1439 if (strchr(name, '/')) {
1440 ret = -EINVAL;
1441 goto out_drop_write;
1442 }
1443
1444 if (name[0] == '.' &&
1445 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1446 ret = -EEXIST;
1447 goto out_drop_write;
1448 }
1449
1450 if (subvol) {
1451 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1452 NULL, transid, readonly, inherit);
1453 } else {
1454 struct fd src = fdget(fd);
1455 struct inode *src_inode;
1456 if (!src.file) {
1457 ret = -EINVAL;
1458 goto out_drop_write;
1459 }
1460
1461 src_inode = src.file->f_path.dentry->d_inode;
1462 if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
1463 printk(KERN_INFO "btrfs: Snapshot src from "
1464 "another FS\n");
1465 ret = -EINVAL;
1466 } else {
1467 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1468 BTRFS_I(src_inode)->root,
1469 transid, readonly, inherit);
1470 }
1471 fdput(src);
1472 }
1473 out_drop_write:
1474 mnt_drop_write_file(file);
1475 out:
1476 return ret;
1477 }
1478
1479 static noinline int btrfs_ioctl_snap_create(struct file *file,
1480 void __user *arg, int subvol)
1481 {
1482 struct btrfs_ioctl_vol_args *vol_args;
1483 int ret;
1484
1485 vol_args = memdup_user(arg, sizeof(*vol_args));
1486 if (IS_ERR(vol_args))
1487 return PTR_ERR(vol_args);
1488 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1489
1490 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1491 vol_args->fd, subvol,
1492 NULL, false, NULL);
1493
1494 kfree(vol_args);
1495 return ret;
1496 }
1497
1498 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1499 void __user *arg, int subvol)
1500 {
1501 struct btrfs_ioctl_vol_args_v2 *vol_args;
1502 int ret;
1503 u64 transid = 0;
1504 u64 *ptr = NULL;
1505 bool readonly = false;
1506 struct btrfs_qgroup_inherit *inherit = NULL;
1507
1508 vol_args = memdup_user(arg, sizeof(*vol_args));
1509 if (IS_ERR(vol_args))
1510 return PTR_ERR(vol_args);
1511 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1512
1513 if (vol_args->flags &
1514 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1515 BTRFS_SUBVOL_QGROUP_INHERIT)) {
1516 ret = -EOPNOTSUPP;
1517 goto out;
1518 }
1519
1520 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1521 ptr = &transid;
1522 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1523 readonly = true;
1524 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1525 if (vol_args->size > PAGE_CACHE_SIZE) {
1526 ret = -EINVAL;
1527 goto out;
1528 }
1529 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1530 if (IS_ERR(inherit)) {
1531 ret = PTR_ERR(inherit);
1532 goto out;
1533 }
1534 }
1535
1536 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1537 vol_args->fd, subvol, ptr,
1538 readonly, &inherit);
1539
1540 if (ret == 0 && ptr &&
1541 copy_to_user(arg +
1542 offsetof(struct btrfs_ioctl_vol_args_v2,
1543 transid), ptr, sizeof(*ptr)))
1544 ret = -EFAULT;
1545 out:
1546 kfree(vol_args);
1547 kfree(inherit);
1548 return ret;
1549 }
1550
1551 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1552 void __user *arg)
1553 {
1554 struct inode *inode = fdentry(file)->d_inode;
1555 struct btrfs_root *root = BTRFS_I(inode)->root;
1556 int ret = 0;
1557 u64 flags = 0;
1558
1559 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1560 return -EINVAL;
1561
1562 down_read(&root->fs_info->subvol_sem);
1563 if (btrfs_root_readonly(root))
1564 flags |= BTRFS_SUBVOL_RDONLY;
1565 up_read(&root->fs_info->subvol_sem);
1566
1567 if (copy_to_user(arg, &flags, sizeof(flags)))
1568 ret = -EFAULT;
1569
1570 return ret;
1571 }
1572
1573 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1574 void __user *arg)
1575 {
1576 struct inode *inode = fdentry(file)->d_inode;
1577 struct btrfs_root *root = BTRFS_I(inode)->root;
1578 struct btrfs_trans_handle *trans;
1579 u64 root_flags;
1580 u64 flags;
1581 int ret = 0;
1582
1583 ret = mnt_want_write_file(file);
1584 if (ret)
1585 goto out;
1586
1587 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1588 ret = -EINVAL;
1589 goto out_drop_write;
1590 }
1591
1592 if (copy_from_user(&flags, arg, sizeof(flags))) {
1593 ret = -EFAULT;
1594 goto out_drop_write;
1595 }
1596
1597 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1598 ret = -EINVAL;
1599 goto out_drop_write;
1600 }
1601
1602 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1603 ret = -EOPNOTSUPP;
1604 goto out_drop_write;
1605 }
1606
1607 if (!inode_owner_or_capable(inode)) {
1608 ret = -EACCES;
1609 goto out_drop_write;
1610 }
1611
1612 down_write(&root->fs_info->subvol_sem);
1613
1614 /* nothing to do */
1615 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1616 goto out_drop_sem;
1617
1618 root_flags = btrfs_root_flags(&root->root_item);
1619 if (flags & BTRFS_SUBVOL_RDONLY)
1620 btrfs_set_root_flags(&root->root_item,
1621 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1622 else
1623 btrfs_set_root_flags(&root->root_item,
1624 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1625
1626 trans = btrfs_start_transaction(root, 1);
1627 if (IS_ERR(trans)) {
1628 ret = PTR_ERR(trans);
1629 goto out_reset;
1630 }
1631
1632 ret = btrfs_update_root(trans, root->fs_info->tree_root,
1633 &root->root_key, &root->root_item);
1634
1635 btrfs_commit_transaction(trans, root);
1636 out_reset:
1637 if (ret)
1638 btrfs_set_root_flags(&root->root_item, root_flags);
1639 out_drop_sem:
1640 up_write(&root->fs_info->subvol_sem);
1641 out_drop_write:
1642 mnt_drop_write_file(file);
1643 out:
1644 return ret;
1645 }
1646
1647 /*
1648 * helper to check if the subvolume references other subvolumes
1649 */
1650 static noinline int may_destroy_subvol(struct btrfs_root *root)
1651 {
1652 struct btrfs_path *path;
1653 struct btrfs_key key;
1654 int ret;
1655
1656 path = btrfs_alloc_path();
1657 if (!path)
1658 return -ENOMEM;
1659
1660 key.objectid = root->root_key.objectid;
1661 key.type = BTRFS_ROOT_REF_KEY;
1662 key.offset = (u64)-1;
1663
1664 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1665 &key, path, 0, 0);
1666 if (ret < 0)
1667 goto out;
1668 BUG_ON(ret == 0);
1669
1670 ret = 0;
1671 if (path->slots[0] > 0) {
1672 path->slots[0]--;
1673 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1674 if (key.objectid == root->root_key.objectid &&
1675 key.type == BTRFS_ROOT_REF_KEY)
1676 ret = -ENOTEMPTY;
1677 }
1678 out:
1679 btrfs_free_path(path);
1680 return ret;
1681 }
1682
1683 static noinline int key_in_sk(struct btrfs_key *key,
1684 struct btrfs_ioctl_search_key *sk)
1685 {
1686 struct btrfs_key test;
1687 int ret;
1688
1689 test.objectid = sk->min_objectid;
1690 test.type = sk->min_type;
1691 test.offset = sk->min_offset;
1692
1693 ret = btrfs_comp_cpu_keys(key, &test);
1694 if (ret < 0)
1695 return 0;
1696
1697 test.objectid = sk->max_objectid;
1698 test.type = sk->max_type;
1699 test.offset = sk->max_offset;
1700
1701 ret = btrfs_comp_cpu_keys(key, &test);
1702 if (ret > 0)
1703 return 0;
1704 return 1;
1705 }
1706
1707 static noinline int copy_to_sk(struct btrfs_root *root,
1708 struct btrfs_path *path,
1709 struct btrfs_key *key,
1710 struct btrfs_ioctl_search_key *sk,
1711 char *buf,
1712 unsigned long *sk_offset,
1713 int *num_found)
1714 {
1715 u64 found_transid;
1716 struct extent_buffer *leaf;
1717 struct btrfs_ioctl_search_header sh;
1718 unsigned long item_off;
1719 unsigned long item_len;
1720 int nritems;
1721 int i;
1722 int slot;
1723 int ret = 0;
1724
1725 leaf = path->nodes[0];
1726 slot = path->slots[0];
1727 nritems = btrfs_header_nritems(leaf);
1728
1729 if (btrfs_header_generation(leaf) > sk->max_transid) {
1730 i = nritems;
1731 goto advance_key;
1732 }
1733 found_transid = btrfs_header_generation(leaf);
1734
1735 for (i = slot; i < nritems; i++) {
1736 item_off = btrfs_item_ptr_offset(leaf, i);
1737 item_len = btrfs_item_size_nr(leaf, i);
1738
1739 if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1740 item_len = 0;
1741
1742 if (sizeof(sh) + item_len + *sk_offset >
1743 BTRFS_SEARCH_ARGS_BUFSIZE) {
1744 ret = 1;
1745 goto overflow;
1746 }
1747
1748 btrfs_item_key_to_cpu(leaf, key, i);
1749 if (!key_in_sk(key, sk))
1750 continue;
1751
1752 sh.objectid = key->objectid;
1753 sh.offset = key->offset;
1754 sh.type = key->type;
1755 sh.len = item_len;
1756 sh.transid = found_transid;
1757
1758 /* copy search result header */
1759 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1760 *sk_offset += sizeof(sh);
1761
1762 if (item_len) {
1763 char *p = buf + *sk_offset;
1764 /* copy the item */
1765 read_extent_buffer(leaf, p,
1766 item_off, item_len);
1767 *sk_offset += item_len;
1768 }
1769 (*num_found)++;
1770
1771 if (*num_found >= sk->nr_items)
1772 break;
1773 }
1774 advance_key:
1775 ret = 0;
1776 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1777 key->offset++;
1778 else if (key->type < (u8)-1 && key->type < sk->max_type) {
1779 key->offset = 0;
1780 key->type++;
1781 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1782 key->offset = 0;
1783 key->type = 0;
1784 key->objectid++;
1785 } else
1786 ret = 1;
1787 overflow:
1788 return ret;
1789 }
1790
1791 static noinline int search_ioctl(struct inode *inode,
1792 struct btrfs_ioctl_search_args *args)
1793 {
1794 struct btrfs_root *root;
1795 struct btrfs_key key;
1796 struct btrfs_key max_key;
1797 struct btrfs_path *path;
1798 struct btrfs_ioctl_search_key *sk = &args->key;
1799 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1800 int ret;
1801 int num_found = 0;
1802 unsigned long sk_offset = 0;
1803
1804 path = btrfs_alloc_path();
1805 if (!path)
1806 return -ENOMEM;
1807
1808 if (sk->tree_id == 0) {
1809 /* search the root of the inode that was passed */
1810 root = BTRFS_I(inode)->root;
1811 } else {
1812 key.objectid = sk->tree_id;
1813 key.type = BTRFS_ROOT_ITEM_KEY;
1814 key.offset = (u64)-1;
1815 root = btrfs_read_fs_root_no_name(info, &key);
1816 if (IS_ERR(root)) {
1817 printk(KERN_ERR "could not find root %llu\n",
1818 sk->tree_id);
1819 btrfs_free_path(path);
1820 return -ENOENT;
1821 }
1822 }
1823
1824 key.objectid = sk->min_objectid;
1825 key.type = sk->min_type;
1826 key.offset = sk->min_offset;
1827
1828 max_key.objectid = sk->max_objectid;
1829 max_key.type = sk->max_type;
1830 max_key.offset = sk->max_offset;
1831
1832 path->keep_locks = 1;
1833
1834 while(1) {
1835 ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1836 sk->min_transid);
1837 if (ret != 0) {
1838 if (ret > 0)
1839 ret = 0;
1840 goto err;
1841 }
1842 ret = copy_to_sk(root, path, &key, sk, args->buf,
1843 &sk_offset, &num_found);
1844 btrfs_release_path(path);
1845 if (ret || num_found >= sk->nr_items)
1846 break;
1847
1848 }
1849 ret = 0;
1850 err:
1851 sk->nr_items = num_found;
1852 btrfs_free_path(path);
1853 return ret;
1854 }
1855
1856 static noinline int btrfs_ioctl_tree_search(struct file *file,
1857 void __user *argp)
1858 {
1859 struct btrfs_ioctl_search_args *args;
1860 struct inode *inode;
1861 int ret;
1862
1863 if (!capable(CAP_SYS_ADMIN))
1864 return -EPERM;
1865
1866 args = memdup_user(argp, sizeof(*args));
1867 if (IS_ERR(args))
1868 return PTR_ERR(args);
1869
1870 inode = fdentry(file)->d_inode;
1871 ret = search_ioctl(inode, args);
1872 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1873 ret = -EFAULT;
1874 kfree(args);
1875 return ret;
1876 }
1877
1878 /*
1879 * Search INODE_REFs to identify path name of 'dirid' directory
1880 * in a 'tree_id' tree. and sets path name to 'name'.
1881 */
1882 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1883 u64 tree_id, u64 dirid, char *name)
1884 {
1885 struct btrfs_root *root;
1886 struct btrfs_key key;
1887 char *ptr;
1888 int ret = -1;
1889 int slot;
1890 int len;
1891 int total_len = 0;
1892 struct btrfs_inode_ref *iref;
1893 struct extent_buffer *l;
1894 struct btrfs_path *path;
1895
1896 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1897 name[0]='\0';
1898 return 0;
1899 }
1900
1901 path = btrfs_alloc_path();
1902 if (!path)
1903 return -ENOMEM;
1904
1905 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1906
1907 key.objectid = tree_id;
1908 key.type = BTRFS_ROOT_ITEM_KEY;
1909 key.offset = (u64)-1;
1910 root = btrfs_read_fs_root_no_name(info, &key);
1911 if (IS_ERR(root)) {
1912 printk(KERN_ERR "could not find root %llu\n", tree_id);
1913 ret = -ENOENT;
1914 goto out;
1915 }
1916
1917 key.objectid = dirid;
1918 key.type = BTRFS_INODE_REF_KEY;
1919 key.offset = (u64)-1;
1920
1921 while(1) {
1922 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1923 if (ret < 0)
1924 goto out;
1925
1926 l = path->nodes[0];
1927 slot = path->slots[0];
1928 if (ret > 0 && slot > 0)
1929 slot--;
1930 btrfs_item_key_to_cpu(l, &key, slot);
1931
1932 if (ret > 0 && (key.objectid != dirid ||
1933 key.type != BTRFS_INODE_REF_KEY)) {
1934 ret = -ENOENT;
1935 goto out;
1936 }
1937
1938 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1939 len = btrfs_inode_ref_name_len(l, iref);
1940 ptr -= len + 1;
1941 total_len += len + 1;
1942 if (ptr < name)
1943 goto out;
1944
1945 *(ptr + len) = '/';
1946 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1947
1948 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1949 break;
1950
1951 btrfs_release_path(path);
1952 key.objectid = key.offset;
1953 key.offset = (u64)-1;
1954 dirid = key.objectid;
1955 }
1956 if (ptr < name)
1957 goto out;
1958 memmove(name, ptr, total_len);
1959 name[total_len]='\0';
1960 ret = 0;
1961 out:
1962 btrfs_free_path(path);
1963 return ret;
1964 }
1965
1966 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1967 void __user *argp)
1968 {
1969 struct btrfs_ioctl_ino_lookup_args *args;
1970 struct inode *inode;
1971 int ret;
1972
1973 if (!capable(CAP_SYS_ADMIN))
1974 return -EPERM;
1975
1976 args = memdup_user(argp, sizeof(*args));
1977 if (IS_ERR(args))
1978 return PTR_ERR(args);
1979
1980 inode = fdentry(file)->d_inode;
1981
1982 if (args->treeid == 0)
1983 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1984
1985 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1986 args->treeid, args->objectid,
1987 args->name);
1988
1989 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1990 ret = -EFAULT;
1991
1992 kfree(args);
1993 return ret;
1994 }
1995
1996 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1997 void __user *arg)
1998 {
1999 struct dentry *parent = fdentry(file);
2000 struct dentry *dentry;
2001 struct inode *dir = parent->d_inode;
2002 struct inode *inode;
2003 struct btrfs_root *root = BTRFS_I(dir)->root;
2004 struct btrfs_root *dest = NULL;
2005 struct btrfs_ioctl_vol_args *vol_args;
2006 struct btrfs_trans_handle *trans;
2007 int namelen;
2008 int ret;
2009 int err = 0;
2010
2011 vol_args = memdup_user(arg, sizeof(*vol_args));
2012 if (IS_ERR(vol_args))
2013 return PTR_ERR(vol_args);
2014
2015 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2016 namelen = strlen(vol_args->name);
2017 if (strchr(vol_args->name, '/') ||
2018 strncmp(vol_args->name, "..", namelen) == 0) {
2019 err = -EINVAL;
2020 goto out;
2021 }
2022
2023 err = mnt_want_write_file(file);
2024 if (err)
2025 goto out;
2026
2027 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
2028 dentry = lookup_one_len(vol_args->name, parent, namelen);
2029 if (IS_ERR(dentry)) {
2030 err = PTR_ERR(dentry);
2031 goto out_unlock_dir;
2032 }
2033
2034 if (!dentry->d_inode) {
2035 err = -ENOENT;
2036 goto out_dput;
2037 }
2038
2039 inode = dentry->d_inode;
2040 dest = BTRFS_I(inode)->root;
2041 if (!capable(CAP_SYS_ADMIN)){
2042 /*
2043 * Regular user. Only allow this with a special mount
2044 * option, when the user has write+exec access to the
2045 * subvol root, and when rmdir(2) would have been
2046 * allowed.
2047 *
2048 * Note that this is _not_ check that the subvol is
2049 * empty or doesn't contain data that we wouldn't
2050 * otherwise be able to delete.
2051 *
2052 * Users who want to delete empty subvols should try
2053 * rmdir(2).
2054 */
2055 err = -EPERM;
2056 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2057 goto out_dput;
2058
2059 /*
2060 * Do not allow deletion if the parent dir is the same
2061 * as the dir to be deleted. That means the ioctl
2062 * must be called on the dentry referencing the root
2063 * of the subvol, not a random directory contained
2064 * within it.
2065 */
2066 err = -EINVAL;
2067 if (root == dest)
2068 goto out_dput;
2069
2070 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2071 if (err)
2072 goto out_dput;
2073
2074 /* check if subvolume may be deleted by a non-root user */
2075 err = btrfs_may_delete(dir, dentry, 1);
2076 if (err)
2077 goto out_dput;
2078 }
2079
2080 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
2081 err = -EINVAL;
2082 goto out_dput;
2083 }
2084
2085 mutex_lock(&inode->i_mutex);
2086 err = d_invalidate(dentry);
2087 if (err)
2088 goto out_unlock;
2089
2090 down_write(&root->fs_info->subvol_sem);
2091
2092 err = may_destroy_subvol(dest);
2093 if (err)
2094 goto out_up_write;
2095
2096 trans = btrfs_start_transaction(root, 0);
2097 if (IS_ERR(trans)) {
2098 err = PTR_ERR(trans);
2099 goto out_up_write;
2100 }
2101 trans->block_rsv = &root->fs_info->global_block_rsv;
2102
2103 ret = btrfs_unlink_subvol(trans, root, dir,
2104 dest->root_key.objectid,
2105 dentry->d_name.name,
2106 dentry->d_name.len);
2107 if (ret) {
2108 err = ret;
2109 btrfs_abort_transaction(trans, root, ret);
2110 goto out_end_trans;
2111 }
2112
2113 btrfs_record_root_in_trans(trans, dest);
2114
2115 memset(&dest->root_item.drop_progress, 0,
2116 sizeof(dest->root_item.drop_progress));
2117 dest->root_item.drop_level = 0;
2118 btrfs_set_root_refs(&dest->root_item, 0);
2119
2120 if (!xchg(&dest->orphan_item_inserted, 1)) {
2121 ret = btrfs_insert_orphan_item(trans,
2122 root->fs_info->tree_root,
2123 dest->root_key.objectid);
2124 if (ret) {
2125 btrfs_abort_transaction(trans, root, ret);
2126 err = ret;
2127 goto out_end_trans;
2128 }
2129 }
2130 out_end_trans:
2131 ret = btrfs_end_transaction(trans, root);
2132 if (ret && !err)
2133 err = ret;
2134 inode->i_flags |= S_DEAD;
2135 out_up_write:
2136 up_write(&root->fs_info->subvol_sem);
2137 out_unlock:
2138 mutex_unlock(&inode->i_mutex);
2139 if (!err) {
2140 shrink_dcache_sb(root->fs_info->sb);
2141 btrfs_invalidate_inodes(dest);
2142 d_delete(dentry);
2143 }
2144 out_dput:
2145 dput(dentry);
2146 out_unlock_dir:
2147 mutex_unlock(&dir->i_mutex);
2148 mnt_drop_write_file(file);
2149 out:
2150 kfree(vol_args);
2151 return err;
2152 }
2153
2154 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
2155 {
2156 struct inode *inode = fdentry(file)->d_inode;
2157 struct btrfs_root *root = BTRFS_I(inode)->root;
2158 struct btrfs_ioctl_defrag_range_args *range;
2159 int ret;
2160
2161 if (btrfs_root_readonly(root))
2162 return -EROFS;
2163
2164 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2165 1)) {
2166 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2167 return -EINPROGRESS;
2168 }
2169 ret = mnt_want_write_file(file);
2170 if (ret) {
2171 atomic_set(&root->fs_info->mutually_exclusive_operation_running,
2172 0);
2173 return ret;
2174 }
2175
2176 switch (inode->i_mode & S_IFMT) {
2177 case S_IFDIR:
2178 if (!capable(CAP_SYS_ADMIN)) {
2179 ret = -EPERM;
2180 goto out;
2181 }
2182 ret = btrfs_defrag_root(root, 0);
2183 if (ret)
2184 goto out;
2185 ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
2186 break;
2187 case S_IFREG:
2188 if (!(file->f_mode & FMODE_WRITE)) {
2189 ret = -EINVAL;
2190 goto out;
2191 }
2192
2193 range = kzalloc(sizeof(*range), GFP_KERNEL);
2194 if (!range) {
2195 ret = -ENOMEM;
2196 goto out;
2197 }
2198
2199 if (argp) {
2200 if (copy_from_user(range, argp,
2201 sizeof(*range))) {
2202 ret = -EFAULT;
2203 kfree(range);
2204 goto out;
2205 }
2206 /* compression requires us to start the IO */
2207 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2208 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2209 range->extent_thresh = (u32)-1;
2210 }
2211 } else {
2212 /* the rest are all set to zero by kzalloc */
2213 range->len = (u64)-1;
2214 }
2215 ret = btrfs_defrag_file(fdentry(file)->d_inode, file,
2216 range, 0, 0);
2217 if (ret > 0)
2218 ret = 0;
2219 kfree(range);
2220 break;
2221 default:
2222 ret = -EINVAL;
2223 }
2224 out:
2225 mnt_drop_write_file(file);
2226 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2227 return ret;
2228 }
2229
2230 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2231 {
2232 struct btrfs_ioctl_vol_args *vol_args;
2233 int ret;
2234
2235 if (!capable(CAP_SYS_ADMIN))
2236 return -EPERM;
2237
2238 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2239 1)) {
2240 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2241 return -EINPROGRESS;
2242 }
2243
2244 mutex_lock(&root->fs_info->volume_mutex);
2245 vol_args = memdup_user(arg, sizeof(*vol_args));
2246 if (IS_ERR(vol_args)) {
2247 ret = PTR_ERR(vol_args);
2248 goto out;
2249 }
2250
2251 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2252 ret = btrfs_init_new_device(root, vol_args->name);
2253
2254 kfree(vol_args);
2255 out:
2256 mutex_unlock(&root->fs_info->volume_mutex);
2257 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2258 return ret;
2259 }
2260
2261 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
2262 {
2263 struct btrfs_ioctl_vol_args *vol_args;
2264 int ret;
2265
2266 if (!capable(CAP_SYS_ADMIN))
2267 return -EPERM;
2268
2269 if (root->fs_info->sb->s_flags & MS_RDONLY)
2270 return -EROFS;
2271
2272 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2273 1)) {
2274 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2275 return -EINPROGRESS;
2276 }
2277
2278 mutex_lock(&root->fs_info->volume_mutex);
2279 vol_args = memdup_user(arg, sizeof(*vol_args));
2280 if (IS_ERR(vol_args)) {
2281 ret = PTR_ERR(vol_args);
2282 goto out;
2283 }
2284
2285 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2286 ret = btrfs_rm_device(root, vol_args->name);
2287
2288 kfree(vol_args);
2289 out:
2290 mutex_unlock(&root->fs_info->volume_mutex);
2291 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
2292 return ret;
2293 }
2294
2295 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2296 {
2297 struct btrfs_ioctl_fs_info_args *fi_args;
2298 struct btrfs_device *device;
2299 struct btrfs_device *next;
2300 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2301 int ret = 0;
2302
2303 if (!capable(CAP_SYS_ADMIN))
2304 return -EPERM;
2305
2306 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2307 if (!fi_args)
2308 return -ENOMEM;
2309
2310 fi_args->num_devices = fs_devices->num_devices;
2311 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2312
2313 mutex_lock(&fs_devices->device_list_mutex);
2314 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2315 if (device->devid > fi_args->max_id)
2316 fi_args->max_id = device->devid;
2317 }
2318 mutex_unlock(&fs_devices->device_list_mutex);
2319
2320 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2321 ret = -EFAULT;
2322
2323 kfree(fi_args);
2324 return ret;
2325 }
2326
2327 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2328 {
2329 struct btrfs_ioctl_dev_info_args *di_args;
2330 struct btrfs_device *dev;
2331 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2332 int ret = 0;
2333 char *s_uuid = NULL;
2334 char empty_uuid[BTRFS_UUID_SIZE] = {0};
2335
2336 if (!capable(CAP_SYS_ADMIN))
2337 return -EPERM;
2338
2339 di_args = memdup_user(arg, sizeof(*di_args));
2340 if (IS_ERR(di_args))
2341 return PTR_ERR(di_args);
2342
2343 if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2344 s_uuid = di_args->uuid;
2345
2346 mutex_lock(&fs_devices->device_list_mutex);
2347 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
2348 mutex_unlock(&fs_devices->device_list_mutex);
2349
2350 if (!dev) {
2351 ret = -ENODEV;
2352 goto out;
2353 }
2354
2355 di_args->devid = dev->devid;
2356 di_args->bytes_used = dev->bytes_used;
2357 di_args->total_bytes = dev->total_bytes;
2358 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2359 if (dev->name) {
2360 struct rcu_string *name;
2361
2362 rcu_read_lock();
2363 name = rcu_dereference(dev->name);
2364 strncpy(di_args->path, name->str, sizeof(di_args->path));
2365 rcu_read_unlock();
2366 di_args->path[sizeof(di_args->path) - 1] = 0;
2367 } else {
2368 di_args->path[0] = '\0';
2369 }
2370
2371 out:
2372 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2373 ret = -EFAULT;
2374
2375 kfree(di_args);
2376 return ret;
2377 }
2378
2379 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2380 u64 off, u64 olen, u64 destoff)
2381 {
2382 struct inode *inode = fdentry(file)->d_inode;
2383 struct btrfs_root *root = BTRFS_I(inode)->root;
2384 struct fd src_file;
2385 struct inode *src;
2386 struct btrfs_trans_handle *trans;
2387 struct btrfs_path *path;
2388 struct extent_buffer *leaf;
2389 char *buf;
2390 struct btrfs_key key;
2391 u32 nritems;
2392 int slot;
2393 int ret;
2394 u64 len = olen;
2395 u64 bs = root->fs_info->sb->s_blocksize;
2396
2397 /*
2398 * TODO:
2399 * - split compressed inline extents. annoying: we need to
2400 * decompress into destination's address_space (the file offset
2401 * may change, so source mapping won't do), then recompress (or
2402 * otherwise reinsert) a subrange.
2403 * - allow ranges within the same file to be cloned (provided
2404 * they don't overlap)?
2405 */
2406
2407 /* the destination must be opened for writing */
2408 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
2409 return -EINVAL;
2410
2411 if (btrfs_root_readonly(root))
2412 return -EROFS;
2413
2414 ret = mnt_want_write_file(file);
2415 if (ret)
2416 return ret;
2417
2418 src_file = fdget(srcfd);
2419 if (!src_file.file) {
2420 ret = -EBADF;
2421 goto out_drop_write;
2422 }
2423
2424 ret = -EXDEV;
2425 if (src_file.file->f_path.mnt != file->f_path.mnt)
2426 goto out_fput;
2427
2428 src = src_file.file->f_dentry->d_inode;
2429
2430 ret = -EINVAL;
2431 if (src == inode)
2432 goto out_fput;
2433
2434 /* the src must be open for reading */
2435 if (!(src_file.file->f_mode & FMODE_READ))
2436 goto out_fput;
2437
2438 /* don't make the dst file partly checksummed */
2439 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2440 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2441 goto out_fput;
2442
2443 ret = -EISDIR;
2444 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2445 goto out_fput;
2446
2447 ret = -EXDEV;
2448 if (src->i_sb != inode->i_sb)
2449 goto out_fput;
2450
2451 ret = -ENOMEM;
2452 buf = vmalloc(btrfs_level_size(root, 0));
2453 if (!buf)
2454 goto out_fput;
2455
2456 path = btrfs_alloc_path();
2457 if (!path) {
2458 vfree(buf);
2459 goto out_fput;
2460 }
2461 path->reada = 2;
2462
2463 if (inode < src) {
2464 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2465 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
2466 } else {
2467 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2468 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
2469 }
2470
2471 /* determine range to clone */
2472 ret = -EINVAL;
2473 if (off + len > src->i_size || off + len < off)
2474 goto out_unlock;
2475 if (len == 0)
2476 olen = len = src->i_size - off;
2477 /* if we extend to eof, continue to block boundary */
2478 if (off + len == src->i_size)
2479 len = ALIGN(src->i_size, bs) - off;
2480
2481 /* verify the end result is block aligned */
2482 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2483 !IS_ALIGNED(destoff, bs))
2484 goto out_unlock;
2485
2486 if (destoff > inode->i_size) {
2487 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2488 if (ret)
2489 goto out_unlock;
2490 }
2491
2492 /* truncate page cache pages from target inode range */
2493 truncate_inode_pages_range(&inode->i_data, destoff,
2494 PAGE_CACHE_ALIGN(destoff + len) - 1);
2495
2496 /* do any pending delalloc/csum calc on src, one way or
2497 another, and lock file content */
2498 while (1) {
2499 struct btrfs_ordered_extent *ordered;
2500 lock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
2501 ordered = btrfs_lookup_first_ordered_extent(src, off + len - 1);
2502 if (!ordered &&
2503 !test_range_bit(&BTRFS_I(src)->io_tree, off, off + len - 1,
2504 EXTENT_DELALLOC, 0, NULL))
2505 break;
2506 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
2507 if (ordered)
2508 btrfs_put_ordered_extent(ordered);
2509 btrfs_wait_ordered_range(src, off, len);
2510 }
2511
2512 /* clone data */
2513 key.objectid = btrfs_ino(src);
2514 key.type = BTRFS_EXTENT_DATA_KEY;
2515 key.offset = 0;
2516
2517 while (1) {
2518 /*
2519 * note the key will change type as we walk through the
2520 * tree.
2521 */
2522 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
2523 0, 0);
2524 if (ret < 0)
2525 goto out;
2526
2527 nritems = btrfs_header_nritems(path->nodes[0]);
2528 if (path->slots[0] >= nritems) {
2529 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
2530 if (ret < 0)
2531 goto out;
2532 if (ret > 0)
2533 break;
2534 nritems = btrfs_header_nritems(path->nodes[0]);
2535 }
2536 leaf = path->nodes[0];
2537 slot = path->slots[0];
2538
2539 btrfs_item_key_to_cpu(leaf, &key, slot);
2540 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
2541 key.objectid != btrfs_ino(src))
2542 break;
2543
2544 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2545 struct btrfs_file_extent_item *extent;
2546 int type;
2547 u32 size;
2548 struct btrfs_key new_key;
2549 u64 disko = 0, diskl = 0;
2550 u64 datao = 0, datal = 0;
2551 u8 comp;
2552 u64 endoff;
2553
2554 size = btrfs_item_size_nr(leaf, slot);
2555 read_extent_buffer(leaf, buf,
2556 btrfs_item_ptr_offset(leaf, slot),
2557 size);
2558
2559 extent = btrfs_item_ptr(leaf, slot,
2560 struct btrfs_file_extent_item);
2561 comp = btrfs_file_extent_compression(leaf, extent);
2562 type = btrfs_file_extent_type(leaf, extent);
2563 if (type == BTRFS_FILE_EXTENT_REG ||
2564 type == BTRFS_FILE_EXTENT_PREALLOC) {
2565 disko = btrfs_file_extent_disk_bytenr(leaf,
2566 extent);
2567 diskl = btrfs_file_extent_disk_num_bytes(leaf,
2568 extent);
2569 datao = btrfs_file_extent_offset(leaf, extent);
2570 datal = btrfs_file_extent_num_bytes(leaf,
2571 extent);
2572 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2573 /* take upper bound, may be compressed */
2574 datal = btrfs_file_extent_ram_bytes(leaf,
2575 extent);
2576 }
2577 btrfs_release_path(path);
2578
2579 if (key.offset + datal <= off ||
2580 key.offset >= off + len - 1)
2581 goto next;
2582
2583 memcpy(&new_key, &key, sizeof(new_key));
2584 new_key.objectid = btrfs_ino(inode);
2585 if (off <= key.offset)
2586 new_key.offset = key.offset + destoff - off;
2587 else
2588 new_key.offset = destoff;
2589
2590 /*
2591 * 1 - adjusting old extent (we may have to split it)
2592 * 1 - add new extent
2593 * 1 - inode update
2594 */
2595 trans = btrfs_start_transaction(root, 3);
2596 if (IS_ERR(trans)) {
2597 ret = PTR_ERR(trans);
2598 goto out;
2599 }
2600
2601 if (type == BTRFS_FILE_EXTENT_REG ||
2602 type == BTRFS_FILE_EXTENT_PREALLOC) {
2603 /*
2604 * a | --- range to clone ---| b
2605 * | ------------- extent ------------- |
2606 */
2607
2608 /* substract range b */
2609 if (key.offset + datal > off + len)
2610 datal = off + len - key.offset;
2611
2612 /* substract range a */
2613 if (off > key.offset) {
2614 datao += off - key.offset;
2615 datal -= off - key.offset;
2616 }
2617
2618 ret = btrfs_drop_extents(trans, root, inode,
2619 new_key.offset,
2620 new_key.offset + datal,
2621 1);
2622 if (ret) {
2623 btrfs_abort_transaction(trans, root,
2624 ret);
2625 btrfs_end_transaction(trans, root);
2626 goto out;
2627 }
2628
2629 ret = btrfs_insert_empty_item(trans, root, path,
2630 &new_key, size);
2631 if (ret) {
2632 btrfs_abort_transaction(trans, root,
2633 ret);
2634 btrfs_end_transaction(trans, root);
2635 goto out;
2636 }
2637
2638 leaf = path->nodes[0];
2639 slot = path->slots[0];
2640 write_extent_buffer(leaf, buf,
2641 btrfs_item_ptr_offset(leaf, slot),
2642 size);
2643
2644 extent = btrfs_item_ptr(leaf, slot,
2645 struct btrfs_file_extent_item);
2646
2647 /* disko == 0 means it's a hole */
2648 if (!disko)
2649 datao = 0;
2650
2651 btrfs_set_file_extent_offset(leaf, extent,
2652 datao);
2653 btrfs_set_file_extent_num_bytes(leaf, extent,
2654 datal);
2655 if (disko) {
2656 inode_add_bytes(inode, datal);
2657 ret = btrfs_inc_extent_ref(trans, root,
2658 disko, diskl, 0,
2659 root->root_key.objectid,
2660 btrfs_ino(inode),
2661 new_key.offset - datao,
2662 0);
2663 if (ret) {
2664 btrfs_abort_transaction(trans,
2665 root,
2666 ret);
2667 btrfs_end_transaction(trans,
2668 root);
2669 goto out;
2670
2671 }
2672 }
2673 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2674 u64 skip = 0;
2675 u64 trim = 0;
2676 if (off > key.offset) {
2677 skip = off - key.offset;
2678 new_key.offset += skip;
2679 }
2680
2681 if (key.offset + datal > off + len)
2682 trim = key.offset + datal - (off + len);
2683
2684 if (comp && (skip || trim)) {
2685 ret = -EINVAL;
2686 btrfs_end_transaction(trans, root);
2687 goto out;
2688 }
2689 size -= skip + trim;
2690 datal -= skip + trim;
2691
2692 ret = btrfs_drop_extents(trans, root, inode,
2693 new_key.offset,
2694 new_key.offset + datal,
2695 1);
2696 if (ret) {
2697 btrfs_abort_transaction(trans, root,
2698 ret);
2699 btrfs_end_transaction(trans, root);
2700 goto out;
2701 }
2702
2703 ret = btrfs_insert_empty_item(trans, root, path,
2704 &new_key, size);
2705 if (ret) {
2706 btrfs_abort_transaction(trans, root,
2707 ret);
2708 btrfs_end_transaction(trans, root);
2709 goto out;
2710 }
2711
2712 if (skip) {
2713 u32 start =
2714 btrfs_file_extent_calc_inline_size(0);
2715 memmove(buf+start, buf+start+skip,
2716 datal);
2717 }
2718
2719 leaf = path->nodes[0];
2720 slot = path->slots[0];
2721 write_extent_buffer(leaf, buf,
2722 btrfs_item_ptr_offset(leaf, slot),
2723 size);
2724 inode_add_bytes(inode, datal);
2725 }
2726
2727 btrfs_mark_buffer_dirty(leaf);
2728 btrfs_release_path(path);
2729
2730 inode_inc_iversion(inode);
2731 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2732
2733 /*
2734 * we round up to the block size at eof when
2735 * determining which extents to clone above,
2736 * but shouldn't round up the file size
2737 */
2738 endoff = new_key.offset + datal;
2739 if (endoff > destoff+olen)
2740 endoff = destoff+olen;
2741 if (endoff > inode->i_size)
2742 btrfs_i_size_write(inode, endoff);
2743
2744 ret = btrfs_update_inode(trans, root, inode);
2745 if (ret) {
2746 btrfs_abort_transaction(trans, root, ret);
2747 btrfs_end_transaction(trans, root);
2748 goto out;
2749 }
2750 ret = btrfs_end_transaction(trans, root);
2751 }
2752 next:
2753 btrfs_release_path(path);
2754 key.offset++;
2755 }
2756 ret = 0;
2757 out:
2758 btrfs_release_path(path);
2759 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
2760 out_unlock:
2761 mutex_unlock(&src->i_mutex);
2762 mutex_unlock(&inode->i_mutex);
2763 vfree(buf);
2764 btrfs_free_path(path);
2765 out_fput:
2766 fdput(src_file);
2767 out_drop_write:
2768 mnt_drop_write_file(file);
2769 return ret;
2770 }
2771
2772 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
2773 {
2774 struct btrfs_ioctl_clone_range_args args;
2775
2776 if (copy_from_user(&args, argp, sizeof(args)))
2777 return -EFAULT;
2778 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2779 args.src_length, args.dest_offset);
2780 }
2781
2782 /*
2783 * there are many ways the trans_start and trans_end ioctls can lead
2784 * to deadlocks. They should only be used by applications that
2785 * basically own the machine, and have a very in depth understanding
2786 * of all the possible deadlocks and enospc problems.
2787 */
2788 static long btrfs_ioctl_trans_start(struct file *file)
2789 {
2790 struct inode *inode = fdentry(file)->d_inode;
2791 struct btrfs_root *root = BTRFS_I(inode)->root;
2792 struct btrfs_trans_handle *trans;
2793 int ret;
2794
2795 ret = -EPERM;
2796 if (!capable(CAP_SYS_ADMIN))
2797 goto out;
2798
2799 ret = -EINPROGRESS;
2800 if (file->private_data)
2801 goto out;
2802
2803 ret = -EROFS;
2804 if (btrfs_root_readonly(root))
2805 goto out;
2806
2807 ret = mnt_want_write_file(file);
2808 if (ret)
2809 goto out;
2810
2811 atomic_inc(&root->fs_info->open_ioctl_trans);
2812
2813 ret = -ENOMEM;
2814 trans = btrfs_start_ioctl_transaction(root);
2815 if (IS_ERR(trans))
2816 goto out_drop;
2817
2818 file->private_data = trans;
2819 return 0;
2820
2821 out_drop:
2822 atomic_dec(&root->fs_info->open_ioctl_trans);
2823 mnt_drop_write_file(file);
2824 out:
2825 return ret;
2826 }
2827
2828 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2829 {
2830 struct inode *inode = fdentry(file)->d_inode;
2831 struct btrfs_root *root = BTRFS_I(inode)->root;
2832 struct btrfs_root *new_root;
2833 struct btrfs_dir_item *di;
2834 struct btrfs_trans_handle *trans;
2835 struct btrfs_path *path;
2836 struct btrfs_key location;
2837 struct btrfs_disk_key disk_key;
2838 u64 objectid = 0;
2839 u64 dir_id;
2840
2841 if (!capable(CAP_SYS_ADMIN))
2842 return -EPERM;
2843
2844 if (copy_from_user(&objectid, argp, sizeof(objectid)))
2845 return -EFAULT;
2846
2847 if (!objectid)
2848 objectid = root->root_key.objectid;
2849
2850 location.objectid = objectid;
2851 location.type = BTRFS_ROOT_ITEM_KEY;
2852 location.offset = (u64)-1;
2853
2854 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2855 if (IS_ERR(new_root))
2856 return PTR_ERR(new_root);
2857
2858 if (btrfs_root_refs(&new_root->root_item) == 0)
2859 return -ENOENT;
2860
2861 path = btrfs_alloc_path();
2862 if (!path)
2863 return -ENOMEM;
2864 path->leave_spinning = 1;
2865
2866 trans = btrfs_start_transaction(root, 1);
2867 if (IS_ERR(trans)) {
2868 btrfs_free_path(path);
2869 return PTR_ERR(trans);
2870 }
2871
2872 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
2873 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2874 dir_id, "default", 7, 1);
2875 if (IS_ERR_OR_NULL(di)) {
2876 btrfs_free_path(path);
2877 btrfs_end_transaction(trans, root);
2878 printk(KERN_ERR "Umm, you don't have the default dir item, "
2879 "this isn't going to work\n");
2880 return -ENOENT;
2881 }
2882
2883 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2884 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2885 btrfs_mark_buffer_dirty(path->nodes[0]);
2886 btrfs_free_path(path);
2887
2888 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
2889 btrfs_end_transaction(trans, root);
2890
2891 return 0;
2892 }
2893
2894 void btrfs_get_block_group_info(struct list_head *groups_list,
2895 struct btrfs_ioctl_space_info *space)
2896 {
2897 struct btrfs_block_group_cache *block_group;
2898
2899 space->total_bytes = 0;
2900 space->used_bytes = 0;
2901 space->flags = 0;
2902 list_for_each_entry(block_group, groups_list, list) {
2903 space->flags = block_group->flags;
2904 space->total_bytes += block_group->key.offset;
2905 space->used_bytes +=
2906 btrfs_block_group_used(&block_group->item);
2907 }
2908 }
2909
2910 long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2911 {
2912 struct btrfs_ioctl_space_args space_args;
2913 struct btrfs_ioctl_space_info space;
2914 struct btrfs_ioctl_space_info *dest;
2915 struct btrfs_ioctl_space_info *dest_orig;
2916 struct btrfs_ioctl_space_info __user *user_dest;
2917 struct btrfs_space_info *info;
2918 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2919 BTRFS_BLOCK_GROUP_SYSTEM,
2920 BTRFS_BLOCK_GROUP_METADATA,
2921 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2922 int num_types = 4;
2923 int alloc_size;
2924 int ret = 0;
2925 u64 slot_count = 0;
2926 int i, c;
2927
2928 if (copy_from_user(&space_args,
2929 (struct btrfs_ioctl_space_args __user *)arg,
2930 sizeof(space_args)))
2931 return -EFAULT;
2932
2933 for (i = 0; i < num_types; i++) {
2934 struct btrfs_space_info *tmp;
2935
2936 info = NULL;
2937 rcu_read_lock();
2938 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2939 list) {
2940 if (tmp->flags == types[i]) {
2941 info = tmp;
2942 break;
2943 }
2944 }
2945 rcu_read_unlock();
2946
2947 if (!info)
2948 continue;
2949
2950 down_read(&info->groups_sem);
2951 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2952 if (!list_empty(&info->block_groups[c]))
2953 slot_count++;
2954 }
2955 up_read(&info->groups_sem);
2956 }
2957
2958 /* space_slots == 0 means they are asking for a count */
2959 if (space_args.space_slots == 0) {
2960 space_args.total_spaces = slot_count;
2961 goto out;
2962 }
2963
2964 slot_count = min_t(u64, space_args.space_slots, slot_count);
2965
2966 alloc_size = sizeof(*dest) * slot_count;
2967
2968 /* we generally have at most 6 or so space infos, one for each raid
2969 * level. So, a whole page should be more than enough for everyone
2970 */
2971 if (alloc_size > PAGE_CACHE_SIZE)
2972 return -ENOMEM;
2973
2974 space_args.total_spaces = 0;
2975 dest = kmalloc(alloc_size, GFP_NOFS);
2976 if (!dest)
2977 return -ENOMEM;
2978 dest_orig = dest;
2979
2980 /* now we have a buffer to copy into */
2981 for (i = 0; i < num_types; i++) {
2982 struct btrfs_space_info *tmp;
2983
2984 if (!slot_count)
2985 break;
2986
2987 info = NULL;
2988 rcu_read_lock();
2989 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2990 list) {
2991 if (tmp->flags == types[i]) {
2992 info = tmp;
2993 break;
2994 }
2995 }
2996 rcu_read_unlock();
2997
2998 if (!info)
2999 continue;
3000 down_read(&info->groups_sem);
3001 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3002 if (!list_empty(&info->block_groups[c])) {
3003 btrfs_get_block_group_info(
3004 &info->block_groups[c], &space);
3005 memcpy(dest, &space, sizeof(space));
3006 dest++;
3007 space_args.total_spaces++;
3008 slot_count--;
3009 }
3010 if (!slot_count)
3011 break;
3012 }
3013 up_read(&info->groups_sem);
3014 }
3015
3016 user_dest = (struct btrfs_ioctl_space_info __user *)
3017 (arg + sizeof(struct btrfs_ioctl_space_args));
3018
3019 if (copy_to_user(user_dest, dest_orig, alloc_size))
3020 ret = -EFAULT;
3021
3022 kfree(dest_orig);
3023 out:
3024 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
3025 ret = -EFAULT;
3026
3027 return ret;
3028 }
3029
3030 /*
3031 * there are many ways the trans_start and trans_end ioctls can lead
3032 * to deadlocks. They should only be used by applications that
3033 * basically own the machine, and have a very in depth understanding
3034 * of all the possible deadlocks and enospc problems.
3035 */
3036 long btrfs_ioctl_trans_end(struct file *file)
3037 {
3038 struct inode *inode = fdentry(file)->d_inode;
3039 struct btrfs_root *root = BTRFS_I(inode)->root;
3040 struct btrfs_trans_handle *trans;
3041
3042 trans = file->private_data;
3043 if (!trans)
3044 return -EINVAL;
3045 file->private_data = NULL;
3046
3047 btrfs_end_transaction(trans, root);
3048
3049 atomic_dec(&root->fs_info->open_ioctl_trans);
3050
3051 mnt_drop_write_file(file);
3052 return 0;
3053 }
3054
3055 static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
3056 {
3057 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
3058 struct btrfs_trans_handle *trans;
3059 u64 transid;
3060 int ret;
3061
3062 trans = btrfs_start_transaction(root, 0);
3063 if (IS_ERR(trans))
3064 return PTR_ERR(trans);
3065 transid = trans->transid;
3066 ret = btrfs_commit_transaction_async(trans, root, 0);
3067 if (ret) {
3068 btrfs_end_transaction(trans, root);
3069 return ret;
3070 }
3071
3072 if (argp)
3073 if (copy_to_user(argp, &transid, sizeof(transid)))
3074 return -EFAULT;
3075 return 0;
3076 }
3077
3078 static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
3079 {
3080 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
3081 u64 transid;
3082
3083 if (argp) {
3084 if (copy_from_user(&transid, argp, sizeof(transid)))
3085 return -EFAULT;
3086 } else {
3087 transid = 0; /* current trans */
3088 }
3089 return btrfs_wait_for_commit(root, transid);
3090 }
3091
3092 static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg)
3093 {
3094 int ret;
3095 struct btrfs_ioctl_scrub_args *sa;
3096
3097 if (!capable(CAP_SYS_ADMIN))
3098 return -EPERM;
3099
3100 sa = memdup_user(arg, sizeof(*sa));
3101 if (IS_ERR(sa))
3102 return PTR_ERR(sa);
3103
3104 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
3105 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY);
3106
3107 if (copy_to_user(arg, sa, sizeof(*sa)))
3108 ret = -EFAULT;
3109
3110 kfree(sa);
3111 return ret;
3112 }
3113
3114 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3115 {
3116 if (!capable(CAP_SYS_ADMIN))
3117 return -EPERM;
3118
3119 return btrfs_scrub_cancel(root->fs_info);
3120 }
3121
3122 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3123 void __user *arg)
3124 {
3125 struct btrfs_ioctl_scrub_args *sa;
3126 int ret;
3127
3128 if (!capable(CAP_SYS_ADMIN))
3129 return -EPERM;
3130
3131 sa = memdup_user(arg, sizeof(*sa));
3132 if (IS_ERR(sa))
3133 return PTR_ERR(sa);
3134
3135 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3136
3137 if (copy_to_user(arg, sa, sizeof(*sa)))
3138 ret = -EFAULT;
3139
3140 kfree(sa);
3141 return ret;
3142 }
3143
3144 static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
3145 void __user *arg)
3146 {
3147 struct btrfs_ioctl_get_dev_stats *sa;
3148 int ret;
3149
3150 sa = memdup_user(arg, sizeof(*sa));
3151 if (IS_ERR(sa))
3152 return PTR_ERR(sa);
3153
3154 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3155 kfree(sa);
3156 return -EPERM;
3157 }
3158
3159 ret = btrfs_get_dev_stats(root, sa);
3160
3161 if (copy_to_user(arg, sa, sizeof(*sa)))
3162 ret = -EFAULT;
3163
3164 kfree(sa);
3165 return ret;
3166 }
3167
3168 static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3169 {
3170 int ret = 0;
3171 int i;
3172 u64 rel_ptr;
3173 int size;
3174 struct btrfs_ioctl_ino_path_args *ipa = NULL;
3175 struct inode_fs_paths *ipath = NULL;
3176 struct btrfs_path *path;
3177
3178 if (!capable(CAP_SYS_ADMIN))
3179 return -EPERM;
3180
3181 path = btrfs_alloc_path();
3182 if (!path) {
3183 ret = -ENOMEM;
3184 goto out;
3185 }
3186
3187 ipa = memdup_user(arg, sizeof(*ipa));
3188 if (IS_ERR(ipa)) {
3189 ret = PTR_ERR(ipa);
3190 ipa = NULL;
3191 goto out;
3192 }
3193
3194 size = min_t(u32, ipa->size, 4096);
3195 ipath = init_ipath(size, root, path);
3196 if (IS_ERR(ipath)) {
3197 ret = PTR_ERR(ipath);
3198 ipath = NULL;
3199 goto out;
3200 }
3201
3202 ret = paths_from_inode(ipa->inum, ipath);
3203 if (ret < 0)
3204 goto out;
3205
3206 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
3207 rel_ptr = ipath->fspath->val[i] -
3208 (u64)(unsigned long)ipath->fspath->val;
3209 ipath->fspath->val[i] = rel_ptr;
3210 }
3211
3212 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3213 (void *)(unsigned long)ipath->fspath, size);
3214 if (ret) {
3215 ret = -EFAULT;
3216 goto out;
3217 }
3218
3219 out:
3220 btrfs_free_path(path);
3221 free_ipath(ipath);
3222 kfree(ipa);
3223
3224 return ret;
3225 }
3226
3227 static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3228 {
3229 struct btrfs_data_container *inodes = ctx;
3230 const size_t c = 3 * sizeof(u64);
3231
3232 if (inodes->bytes_left >= c) {
3233 inodes->bytes_left -= c;
3234 inodes->val[inodes->elem_cnt] = inum;
3235 inodes->val[inodes->elem_cnt + 1] = offset;
3236 inodes->val[inodes->elem_cnt + 2] = root;
3237 inodes->elem_cnt += 3;
3238 } else {
3239 inodes->bytes_missing += c - inodes->bytes_left;
3240 inodes->bytes_left = 0;
3241 inodes->elem_missed += 3;
3242 }
3243
3244 return 0;
3245 }
3246
3247 static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3248 void __user *arg)
3249 {
3250 int ret = 0;
3251 int size;
3252 struct btrfs_ioctl_logical_ino_args *loi;
3253 struct btrfs_data_container *inodes = NULL;
3254 struct btrfs_path *path = NULL;
3255
3256 if (!capable(CAP_SYS_ADMIN))
3257 return -EPERM;
3258
3259 loi = memdup_user(arg, sizeof(*loi));
3260 if (IS_ERR(loi)) {
3261 ret = PTR_ERR(loi);
3262 loi = NULL;
3263 goto out;
3264 }
3265
3266 path = btrfs_alloc_path();
3267 if (!path) {
3268 ret = -ENOMEM;
3269 goto out;
3270 }
3271
3272 size = min_t(u32, loi->size, 64 * 1024);
3273 inodes = init_data_container(size);
3274 if (IS_ERR(inodes)) {
3275 ret = PTR_ERR(inodes);
3276 inodes = NULL;
3277 goto out;
3278 }
3279
3280 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
3281 build_ino_list, inodes);
3282 if (ret == -EINVAL)
3283 ret = -ENOENT;
3284 if (ret < 0)
3285 goto out;
3286
3287 ret = copy_to_user((void *)(unsigned long)loi->inodes,
3288 (void *)(unsigned long)inodes, size);
3289 if (ret)
3290 ret = -EFAULT;
3291
3292 out:
3293 btrfs_free_path(path);
3294 vfree(inodes);
3295 kfree(loi);
3296
3297 return ret;
3298 }
3299
3300 void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
3301 struct btrfs_ioctl_balance_args *bargs)
3302 {
3303 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3304
3305 bargs->flags = bctl->flags;
3306
3307 if (atomic_read(&fs_info->balance_running))
3308 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3309 if (atomic_read(&fs_info->balance_pause_req))
3310 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
3311 if (atomic_read(&fs_info->balance_cancel_req))
3312 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
3313
3314 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3315 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3316 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
3317
3318 if (lock) {
3319 spin_lock(&fs_info->balance_lock);
3320 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3321 spin_unlock(&fs_info->balance_lock);
3322 } else {
3323 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3324 }
3325 }
3326
3327 static long btrfs_ioctl_balance(struct file *file, void __user *arg)
3328 {
3329 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3330 struct btrfs_fs_info *fs_info = root->fs_info;
3331 struct btrfs_ioctl_balance_args *bargs;
3332 struct btrfs_balance_control *bctl;
3333 int ret;
3334 int need_to_clear_lock = 0;
3335
3336 if (!capable(CAP_SYS_ADMIN))
3337 return -EPERM;
3338
3339 ret = mnt_want_write_file(file);
3340 if (ret)
3341 return ret;
3342
3343 mutex_lock(&fs_info->volume_mutex);
3344 mutex_lock(&fs_info->balance_mutex);
3345
3346 if (arg) {
3347 bargs = memdup_user(arg, sizeof(*bargs));
3348 if (IS_ERR(bargs)) {
3349 ret = PTR_ERR(bargs);
3350 goto out;
3351 }
3352
3353 if (bargs->flags & BTRFS_BALANCE_RESUME) {
3354 if (!fs_info->balance_ctl) {
3355 ret = -ENOTCONN;
3356 goto out_bargs;
3357 }
3358
3359 bctl = fs_info->balance_ctl;
3360 spin_lock(&fs_info->balance_lock);
3361 bctl->flags |= BTRFS_BALANCE_RESUME;
3362 spin_unlock(&fs_info->balance_lock);
3363
3364 goto do_balance;
3365 }
3366 } else {
3367 bargs = NULL;
3368 }
3369
3370 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
3371 1)) {
3372 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3373 ret = -EINPROGRESS;
3374 goto out_bargs;
3375 }
3376 need_to_clear_lock = 1;
3377
3378 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3379 if (!bctl) {
3380 ret = -ENOMEM;
3381 goto out_bargs;
3382 }
3383
3384 bctl->fs_info = fs_info;
3385 if (arg) {
3386 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3387 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3388 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3389
3390 bctl->flags = bargs->flags;
3391 } else {
3392 /* balance everything - no filters */
3393 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
3394 }
3395
3396 do_balance:
3397 ret = btrfs_balance(bctl, bargs);
3398 /*
3399 * bctl is freed in __cancel_balance or in free_fs_info if
3400 * restriper was paused all the way until unmount
3401 */
3402 if (arg) {
3403 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3404 ret = -EFAULT;
3405 }
3406
3407 out_bargs:
3408 kfree(bargs);
3409 out:
3410 if (need_to_clear_lock)
3411 atomic_set(&root->fs_info->mutually_exclusive_operation_running,
3412 0);
3413 mutex_unlock(&fs_info->balance_mutex);
3414 mutex_unlock(&fs_info->volume_mutex);
3415 mnt_drop_write_file(file);
3416 return ret;
3417 }
3418
3419 static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
3420 {
3421 if (!capable(CAP_SYS_ADMIN))
3422 return -EPERM;
3423
3424 switch (cmd) {
3425 case BTRFS_BALANCE_CTL_PAUSE:
3426 return btrfs_pause_balance(root->fs_info);
3427 case BTRFS_BALANCE_CTL_CANCEL:
3428 return btrfs_cancel_balance(root->fs_info);
3429 }
3430
3431 return -EINVAL;
3432 }
3433
3434 static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
3435 void __user *arg)
3436 {
3437 struct btrfs_fs_info *fs_info = root->fs_info;
3438 struct btrfs_ioctl_balance_args *bargs;
3439 int ret = 0;
3440
3441 if (!capable(CAP_SYS_ADMIN))
3442 return -EPERM;
3443
3444 mutex_lock(&fs_info->balance_mutex);
3445 if (!fs_info->balance_ctl) {
3446 ret = -ENOTCONN;
3447 goto out;
3448 }
3449
3450 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
3451 if (!bargs) {
3452 ret = -ENOMEM;
3453 goto out;
3454 }
3455
3456 update_ioctl_balance_args(fs_info, 1, bargs);
3457
3458 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3459 ret = -EFAULT;
3460
3461 kfree(bargs);
3462 out:
3463 mutex_unlock(&fs_info->balance_mutex);
3464 return ret;
3465 }
3466
3467 static long btrfs_ioctl_quota_ctl(struct btrfs_root *root, void __user *arg)
3468 {
3469 struct btrfs_ioctl_quota_ctl_args *sa;
3470 struct btrfs_trans_handle *trans = NULL;
3471 int ret;
3472 int err;
3473
3474 if (!capable(CAP_SYS_ADMIN))
3475 return -EPERM;
3476
3477 if (root->fs_info->sb->s_flags & MS_RDONLY)
3478 return -EROFS;
3479
3480 sa = memdup_user(arg, sizeof(*sa));
3481 if (IS_ERR(sa))
3482 return PTR_ERR(sa);
3483
3484 if (sa->cmd != BTRFS_QUOTA_CTL_RESCAN) {
3485 trans = btrfs_start_transaction(root, 2);
3486 if (IS_ERR(trans)) {
3487 ret = PTR_ERR(trans);
3488 goto out;
3489 }
3490 }
3491
3492 switch (sa->cmd) {
3493 case BTRFS_QUOTA_CTL_ENABLE:
3494 ret = btrfs_quota_enable(trans, root->fs_info);
3495 break;
3496 case BTRFS_QUOTA_CTL_DISABLE:
3497 ret = btrfs_quota_disable(trans, root->fs_info);
3498 break;
3499 case BTRFS_QUOTA_CTL_RESCAN:
3500 ret = btrfs_quota_rescan(root->fs_info);
3501 break;
3502 default:
3503 ret = -EINVAL;
3504 break;
3505 }
3506
3507 if (copy_to_user(arg, sa, sizeof(*sa)))
3508 ret = -EFAULT;
3509
3510 if (trans) {
3511 err = btrfs_commit_transaction(trans, root);
3512 if (err && !ret)
3513 ret = err;
3514 }
3515
3516 out:
3517 kfree(sa);
3518 return ret;
3519 }
3520
3521 static long btrfs_ioctl_qgroup_assign(struct btrfs_root *root, void __user *arg)
3522 {
3523 struct btrfs_ioctl_qgroup_assign_args *sa;
3524 struct btrfs_trans_handle *trans;
3525 int ret;
3526 int err;
3527
3528 if (!capable(CAP_SYS_ADMIN))
3529 return -EPERM;
3530
3531 if (root->fs_info->sb->s_flags & MS_RDONLY)
3532 return -EROFS;
3533
3534 sa = memdup_user(arg, sizeof(*sa));
3535 if (IS_ERR(sa))
3536 return PTR_ERR(sa);
3537
3538 trans = btrfs_join_transaction(root);
3539 if (IS_ERR(trans)) {
3540 ret = PTR_ERR(trans);
3541 goto out;
3542 }
3543
3544 /* FIXME: check if the IDs really exist */
3545 if (sa->assign) {
3546 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
3547 sa->src, sa->dst);
3548 } else {
3549 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
3550 sa->src, sa->dst);
3551 }
3552
3553 err = btrfs_end_transaction(trans, root);
3554 if (err && !ret)
3555 ret = err;
3556
3557 out:
3558 kfree(sa);
3559 return ret;
3560 }
3561
3562 static long btrfs_ioctl_qgroup_create(struct btrfs_root *root, void __user *arg)
3563 {
3564 struct btrfs_ioctl_qgroup_create_args *sa;
3565 struct btrfs_trans_handle *trans;
3566 int ret;
3567 int err;
3568
3569 if (!capable(CAP_SYS_ADMIN))
3570 return -EPERM;
3571
3572 if (root->fs_info->sb->s_flags & MS_RDONLY)
3573 return -EROFS;
3574
3575 sa = memdup_user(arg, sizeof(*sa));
3576 if (IS_ERR(sa))
3577 return PTR_ERR(sa);
3578
3579 trans = btrfs_join_transaction(root);
3580 if (IS_ERR(trans)) {
3581 ret = PTR_ERR(trans);
3582 goto out;
3583 }
3584
3585 /* FIXME: check if the IDs really exist */
3586 if (sa->create) {
3587 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
3588 NULL);
3589 } else {
3590 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
3591 }
3592
3593 err = btrfs_end_transaction(trans, root);
3594 if (err && !ret)
3595 ret = err;
3596
3597 out:
3598 kfree(sa);
3599 return ret;
3600 }
3601
3602 static long btrfs_ioctl_qgroup_limit(struct btrfs_root *root, void __user *arg)
3603 {
3604 struct btrfs_ioctl_qgroup_limit_args *sa;
3605 struct btrfs_trans_handle *trans;
3606 int ret;
3607 int err;
3608 u64 qgroupid;
3609
3610 if (!capable(CAP_SYS_ADMIN))
3611 return -EPERM;
3612
3613 if (root->fs_info->sb->s_flags & MS_RDONLY)
3614 return -EROFS;
3615
3616 sa = memdup_user(arg, sizeof(*sa));
3617 if (IS_ERR(sa))
3618 return PTR_ERR(sa);
3619
3620 trans = btrfs_join_transaction(root);
3621 if (IS_ERR(trans)) {
3622 ret = PTR_ERR(trans);
3623 goto out;
3624 }
3625
3626 qgroupid = sa->qgroupid;
3627 if (!qgroupid) {
3628 /* take the current subvol as qgroup */
3629 qgroupid = root->root_key.objectid;
3630 }
3631
3632 /* FIXME: check if the IDs really exist */
3633 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
3634
3635 err = btrfs_end_transaction(trans, root);
3636 if (err && !ret)
3637 ret = err;
3638
3639 out:
3640 kfree(sa);
3641 return ret;
3642 }
3643
3644 static long btrfs_ioctl_set_received_subvol(struct file *file,
3645 void __user *arg)
3646 {
3647 struct btrfs_ioctl_received_subvol_args *sa = NULL;
3648 struct inode *inode = fdentry(file)->d_inode;
3649 struct btrfs_root *root = BTRFS_I(inode)->root;
3650 struct btrfs_root_item *root_item = &root->root_item;
3651 struct btrfs_trans_handle *trans;
3652 struct timespec ct = CURRENT_TIME;
3653 int ret = 0;
3654
3655 ret = mnt_want_write_file(file);
3656 if (ret < 0)
3657 return ret;
3658
3659 down_write(&root->fs_info->subvol_sem);
3660
3661 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
3662 ret = -EINVAL;
3663 goto out;
3664 }
3665
3666 if (btrfs_root_readonly(root)) {
3667 ret = -EROFS;
3668 goto out;
3669 }
3670
3671 if (!inode_owner_or_capable(inode)) {
3672 ret = -EACCES;
3673 goto out;
3674 }
3675
3676 sa = memdup_user(arg, sizeof(*sa));
3677 if (IS_ERR(sa)) {
3678 ret = PTR_ERR(sa);
3679 sa = NULL;
3680 goto out;
3681 }
3682
3683 trans = btrfs_start_transaction(root, 1);
3684 if (IS_ERR(trans)) {
3685 ret = PTR_ERR(trans);
3686 trans = NULL;
3687 goto out;
3688 }
3689
3690 sa->rtransid = trans->transid;
3691 sa->rtime.sec = ct.tv_sec;
3692 sa->rtime.nsec = ct.tv_nsec;
3693
3694 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
3695 btrfs_set_root_stransid(root_item, sa->stransid);
3696 btrfs_set_root_rtransid(root_item, sa->rtransid);
3697 root_item->stime.sec = cpu_to_le64(sa->stime.sec);
3698 root_item->stime.nsec = cpu_to_le32(sa->stime.nsec);
3699 root_item->rtime.sec = cpu_to_le64(sa->rtime.sec);
3700 root_item->rtime.nsec = cpu_to_le32(sa->rtime.nsec);
3701
3702 ret = btrfs_update_root(trans, root->fs_info->tree_root,
3703 &root->root_key, &root->root_item);
3704 if (ret < 0) {
3705 btrfs_end_transaction(trans, root);
3706 trans = NULL;
3707 goto out;
3708 } else {
3709 ret = btrfs_commit_transaction(trans, root);
3710 if (ret < 0)
3711 goto out;
3712 }
3713
3714 ret = copy_to_user(arg, sa, sizeof(*sa));
3715 if (ret)
3716 ret = -EFAULT;
3717
3718 out:
3719 kfree(sa);
3720 up_write(&root->fs_info->subvol_sem);
3721 mnt_drop_write_file(file);
3722 return ret;
3723 }
3724
3725 long btrfs_ioctl(struct file *file, unsigned int
3726 cmd, unsigned long arg)
3727 {
3728 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3729 void __user *argp = (void __user *)arg;
3730
3731 switch (cmd) {
3732 case FS_IOC_GETFLAGS:
3733 return btrfs_ioctl_getflags(file, argp);
3734 case FS_IOC_SETFLAGS:
3735 return btrfs_ioctl_setflags(file, argp);
3736 case FS_IOC_GETVERSION:
3737 return btrfs_ioctl_getversion(file, argp);
3738 case FITRIM:
3739 return btrfs_ioctl_fitrim(file, argp);
3740 case BTRFS_IOC_SNAP_CREATE:
3741 return btrfs_ioctl_snap_create(file, argp, 0);
3742 case BTRFS_IOC_SNAP_CREATE_V2:
3743 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3744 case BTRFS_IOC_SUBVOL_CREATE:
3745 return btrfs_ioctl_snap_create(file, argp, 1);
3746 case BTRFS_IOC_SUBVOL_CREATE_V2:
3747 return btrfs_ioctl_snap_create_v2(file, argp, 1);
3748 case BTRFS_IOC_SNAP_DESTROY:
3749 return btrfs_ioctl_snap_destroy(file, argp);
3750 case BTRFS_IOC_SUBVOL_GETFLAGS:
3751 return btrfs_ioctl_subvol_getflags(file, argp);
3752 case BTRFS_IOC_SUBVOL_SETFLAGS:
3753 return btrfs_ioctl_subvol_setflags(file, argp);
3754 case BTRFS_IOC_DEFAULT_SUBVOL:
3755 return btrfs_ioctl_default_subvol(file, argp);
3756 case BTRFS_IOC_DEFRAG:
3757 return btrfs_ioctl_defrag(file, NULL);
3758 case BTRFS_IOC_DEFRAG_RANGE:
3759 return btrfs_ioctl_defrag(file, argp);
3760 case BTRFS_IOC_RESIZE:
3761 return btrfs_ioctl_resize(root, argp);
3762 case BTRFS_IOC_ADD_DEV:
3763 return btrfs_ioctl_add_dev(root, argp);
3764 case BTRFS_IOC_RM_DEV:
3765 return btrfs_ioctl_rm_dev(root, argp);
3766 case BTRFS_IOC_FS_INFO:
3767 return btrfs_ioctl_fs_info(root, argp);
3768 case BTRFS_IOC_DEV_INFO:
3769 return btrfs_ioctl_dev_info(root, argp);
3770 case BTRFS_IOC_BALANCE:
3771 return btrfs_ioctl_balance(file, NULL);
3772 case BTRFS_IOC_CLONE:
3773 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
3774 case BTRFS_IOC_CLONE_RANGE:
3775 return btrfs_ioctl_clone_range(file, argp);
3776 case BTRFS_IOC_TRANS_START:
3777 return btrfs_ioctl_trans_start(file);
3778 case BTRFS_IOC_TRANS_END:
3779 return btrfs_ioctl_trans_end(file);
3780 case BTRFS_IOC_TREE_SEARCH:
3781 return btrfs_ioctl_tree_search(file, argp);
3782 case BTRFS_IOC_INO_LOOKUP:
3783 return btrfs_ioctl_ino_lookup(file, argp);
3784 case BTRFS_IOC_INO_PATHS:
3785 return btrfs_ioctl_ino_to_path(root, argp);
3786 case BTRFS_IOC_LOGICAL_INO:
3787 return btrfs_ioctl_logical_to_ino(root, argp);
3788 case BTRFS_IOC_SPACE_INFO:
3789 return btrfs_ioctl_space_info(root, argp);
3790 case BTRFS_IOC_SYNC:
3791 btrfs_sync_fs(file->f_dentry->d_sb, 1);
3792 return 0;
3793 case BTRFS_IOC_START_SYNC:
3794 return btrfs_ioctl_start_sync(file, argp);
3795 case BTRFS_IOC_WAIT_SYNC:
3796 return btrfs_ioctl_wait_sync(file, argp);
3797 case BTRFS_IOC_SCRUB:
3798 return btrfs_ioctl_scrub(root, argp);
3799 case BTRFS_IOC_SCRUB_CANCEL:
3800 return btrfs_ioctl_scrub_cancel(root, argp);
3801 case BTRFS_IOC_SCRUB_PROGRESS:
3802 return btrfs_ioctl_scrub_progress(root, argp);
3803 case BTRFS_IOC_BALANCE_V2:
3804 return btrfs_ioctl_balance(file, argp);
3805 case BTRFS_IOC_BALANCE_CTL:
3806 return btrfs_ioctl_balance_ctl(root, arg);
3807 case BTRFS_IOC_BALANCE_PROGRESS:
3808 return btrfs_ioctl_balance_progress(root, argp);
3809 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
3810 return btrfs_ioctl_set_received_subvol(file, argp);
3811 case BTRFS_IOC_SEND:
3812 return btrfs_ioctl_send(file, argp);
3813 case BTRFS_IOC_GET_DEV_STATS:
3814 return btrfs_ioctl_get_dev_stats(root, argp);
3815 case BTRFS_IOC_QUOTA_CTL:
3816 return btrfs_ioctl_quota_ctl(root, argp);
3817 case BTRFS_IOC_QGROUP_ASSIGN:
3818 return btrfs_ioctl_qgroup_assign(root, argp);
3819 case BTRFS_IOC_QGROUP_CREATE:
3820 return btrfs_ioctl_qgroup_create(root, argp);
3821 case BTRFS_IOC_QGROUP_LIMIT:
3822 return btrfs_ioctl_qgroup_limit(root, argp);
3823 }
3824
3825 return -ENOTTY;
3826 }
Linux kernel - Deliverables
This page took 0.111565 seconds and 5 git commands to generate.