?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge remote-tracking branch 'remotes/josef/for-chris' into btrfs-next-stable
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / btrfs / ioctl.c
blob877727b28d88d8d58b903dd2896b43d495833344
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 "compat.h"
45 #include "ctree.h"
46 #include "disk-io.h"
47 #include "transaction.h"
48 #include "btrfs_inode.h"
49 #include "ioctl.h"
50 #include "print-tree.h"
51 #include "volumes.h"
52 #include "locking.h"
53 #include "inode-map.h"
54
55 /* Mask out flags that are inappropriate for the given type of inode. */
56 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
57 {
58 if (S_ISDIR(mode))
59 return flags;
60 else if (S_ISREG(mode))
61 return flags & ~FS_DIRSYNC_FL;
62 else
63 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
64 }
65
66 /*
67 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
68 */
69 static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
70 {
71 unsigned int iflags = 0;
72
73 if (flags & BTRFS_INODE_SYNC)
74 iflags |= FS_SYNC_FL;
75 if (flags & BTRFS_INODE_IMMUTABLE)
76 iflags |= FS_IMMUTABLE_FL;
77 if (flags & BTRFS_INODE_APPEND)
78 iflags |= FS_APPEND_FL;
79 if (flags & BTRFS_INODE_NODUMP)
80 iflags |= FS_NODUMP_FL;
81 if (flags & BTRFS_INODE_NOATIME)
82 iflags |= FS_NOATIME_FL;
83 if (flags & BTRFS_INODE_DIRSYNC)
84 iflags |= FS_DIRSYNC_FL;
85 if (flags & BTRFS_INODE_NODATACOW)
86 iflags |= FS_NOCOW_FL;
87
88 if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
89 iflags |= FS_COMPR_FL;
90 else if (flags & BTRFS_INODE_NOCOMPRESS)
91 iflags |= FS_NOCOMP_FL;
92
93 return iflags;
94 }
95
96 /*
97 * Update inode->i_flags based on the btrfs internal flags.
98 */
99 void btrfs_update_iflags(struct inode *inode)
100 {
101 struct btrfs_inode *ip = BTRFS_I(inode);
102
103 inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
104
105 if (ip->flags & BTRFS_INODE_SYNC)
106 inode->i_flags |= S_SYNC;
107 if (ip->flags & BTRFS_INODE_IMMUTABLE)
108 inode->i_flags |= S_IMMUTABLE;
109 if (ip->flags & BTRFS_INODE_APPEND)
110 inode->i_flags |= S_APPEND;
111 if (ip->flags & BTRFS_INODE_NOATIME)
112 inode->i_flags |= S_NOATIME;
113 if (ip->flags & BTRFS_INODE_DIRSYNC)
114 inode->i_flags |= S_DIRSYNC;
115 }
116
117 /*
118 * Inherit flags from the parent inode.
119 *
120 * Currently only the compression flags and the cow flags are inherited.
121 */
122 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
123 {
124 unsigned int flags;
125
126 if (!dir)
127 return;
128
129 flags = BTRFS_I(dir)->flags;
130
131 if (flags & BTRFS_INODE_NOCOMPRESS) {
132 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
133 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
134 } else if (flags & BTRFS_INODE_COMPRESS) {
135 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
136 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
137 }
138
139 if (flags & BTRFS_INODE_NODATACOW)
140 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
141
142 btrfs_update_iflags(inode);
143 }
144
145 static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
146 {
147 struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode);
148 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
149
150 if (copy_to_user(arg, &flags, sizeof(flags)))
151 return -EFAULT;
152 return 0;
153 }
154
155 static int check_flags(unsigned int flags)
156 {
157 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
158 FS_NOATIME_FL | FS_NODUMP_FL | \
159 FS_SYNC_FL | FS_DIRSYNC_FL | \
160 FS_NOCOMP_FL | FS_COMPR_FL |
161 FS_NOCOW_FL))
162 return -EOPNOTSUPP;
163
164 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
165 return -EINVAL;
166
167 return 0;
168 }
169
170 static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
171 {
172 struct inode *inode = file->f_path.dentry->d_inode;
173 struct btrfs_inode *ip = BTRFS_I(inode);
174 struct btrfs_root *root = ip->root;
175 struct btrfs_trans_handle *trans;
176 unsigned int flags, oldflags;
177 int ret;
178
179 if (btrfs_root_readonly(root))
180 return -EROFS;
181
182 if (copy_from_user(&flags, arg, sizeof(flags)))
183 return -EFAULT;
184
185 ret = check_flags(flags);
186 if (ret)
187 return ret;
188
189 if (!inode_owner_or_capable(inode))
190 return -EACCES;
191
192 mutex_lock(&inode->i_mutex);
193
194 flags = btrfs_mask_flags(inode->i_mode, flags);
195 oldflags = btrfs_flags_to_ioctl(ip->flags);
196 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
197 if (!capable(CAP_LINUX_IMMUTABLE)) {
198 ret = -EPERM;
199 goto out_unlock;
200 }
201 }
202
203 ret = mnt_want_write(file->f_path.mnt);
204 if (ret)
205 goto out_unlock;
206
207 if (flags & FS_SYNC_FL)
208 ip->flags |= BTRFS_INODE_SYNC;
209 else
210 ip->flags &= ~BTRFS_INODE_SYNC;
211 if (flags & FS_IMMUTABLE_FL)
212 ip->flags |= BTRFS_INODE_IMMUTABLE;
213 else
214 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
215 if (flags & FS_APPEND_FL)
216 ip->flags |= BTRFS_INODE_APPEND;
217 else
218 ip->flags &= ~BTRFS_INODE_APPEND;
219 if (flags & FS_NODUMP_FL)
220 ip->flags |= BTRFS_INODE_NODUMP;
221 else
222 ip->flags &= ~BTRFS_INODE_NODUMP;
223 if (flags & FS_NOATIME_FL)
224 ip->flags |= BTRFS_INODE_NOATIME;
225 else
226 ip->flags &= ~BTRFS_INODE_NOATIME;
227 if (flags & FS_DIRSYNC_FL)
228 ip->flags |= BTRFS_INODE_DIRSYNC;
229 else
230 ip->flags &= ~BTRFS_INODE_DIRSYNC;
231 if (flags & FS_NOCOW_FL)
232 ip->flags |= BTRFS_INODE_NODATACOW;
233 else
234 ip->flags &= ~BTRFS_INODE_NODATACOW;
235
236 /*
237 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
238 * flag may be changed automatically if compression code won't make
239 * things smaller.
240 */
241 if (flags & FS_NOCOMP_FL) {
242 ip->flags &= ~BTRFS_INODE_COMPRESS;
243 ip->flags |= BTRFS_INODE_NOCOMPRESS;
244 } else if (flags & FS_COMPR_FL) {
245 ip->flags |= BTRFS_INODE_COMPRESS;
246 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
247 } else {
248 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
249 }
250
251 trans = btrfs_join_transaction(root);
252 BUG_ON(IS_ERR(trans));
253
254 ret = btrfs_update_inode(trans, root, inode);
255 BUG_ON(ret);
256
257 btrfs_update_iflags(inode);
258 inode->i_ctime = CURRENT_TIME;
259 btrfs_end_transaction(trans, root);
260
261 mnt_drop_write(file->f_path.mnt);
262
263 ret = 0;
264 out_unlock:
265 mutex_unlock(&inode->i_mutex);
266 return ret;
267 }
268
269 static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
270 {
271 struct inode *inode = file->f_path.dentry->d_inode;
272
273 return put_user(inode->i_generation, arg);
274 }
275
276 static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
277 {
278 struct btrfs_root *root = fdentry(file)->d_sb->s_fs_info;
279 struct btrfs_fs_info *fs_info = root->fs_info;
280 struct btrfs_device *device;
281 struct request_queue *q;
282 struct fstrim_range range;
283 u64 minlen = ULLONG_MAX;
284 u64 num_devices = 0;
285 int ret;
286
287 if (!capable(CAP_SYS_ADMIN))
288 return -EPERM;
289
290 rcu_read_lock();
291 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
292 dev_list) {
293 if (!device->bdev)
294 continue;
295 q = bdev_get_queue(device->bdev);
296 if (blk_queue_discard(q)) {
297 num_devices++;
298 minlen = min((u64)q->limits.discard_granularity,
299 minlen);
300 }
301 }
302 rcu_read_unlock();
303 if (!num_devices)
304 return -EOPNOTSUPP;
305
306 if (copy_from_user(&range, arg, sizeof(range)))
307 return -EFAULT;
308
309 range.minlen = max(range.minlen, minlen);
310 ret = btrfs_trim_fs(root, &range);
311 if (ret < 0)
312 return ret;
313
314 if (copy_to_user(arg, &range, sizeof(range)))
315 return -EFAULT;
316
317 return 0;
318 }
319
320 static noinline int create_subvol(struct btrfs_root *root,
321 struct dentry *dentry,
322 char *name, int namelen,
323 u64 *async_transid)
324 {
325 struct btrfs_trans_handle *trans;
326 struct btrfs_key key;
327 struct btrfs_root_item root_item;
328 struct btrfs_inode_item *inode_item;
329 struct extent_buffer *leaf;
330 struct btrfs_root *new_root;
331 struct dentry *parent = dentry->d_parent;
332 struct inode *dir;
333 int ret;
334 int err;
335 u64 objectid;
336 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
337 u64 index = 0;
338
339 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
340 if (ret)
341 return ret;
342
343 dir = parent->d_inode;
344
345 /*
346 * 1 - inode item
347 * 2 - refs
348 * 1 - root item
349 * 2 - dir items
350 */
351 trans = btrfs_start_transaction(root, 6);
352 if (IS_ERR(trans))
353 return PTR_ERR(trans);
354
355 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
356 0, objectid, NULL, 0, 0, 0);
357 if (IS_ERR(leaf)) {
358 ret = PTR_ERR(leaf);
359 goto fail;
360 }
361
362 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
363 btrfs_set_header_bytenr(leaf, leaf->start);
364 btrfs_set_header_generation(leaf, trans->transid);
365 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
366 btrfs_set_header_owner(leaf, objectid);
367
368 write_extent_buffer(leaf, root->fs_info->fsid,
369 (unsigned long)btrfs_header_fsid(leaf),
370 BTRFS_FSID_SIZE);
371 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
372 (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
373 BTRFS_UUID_SIZE);
374 btrfs_mark_buffer_dirty(leaf);
375
376 inode_item = &root_item.inode;
377 memset(inode_item, 0, sizeof(*inode_item));
378 inode_item->generation = cpu_to_le64(1);
379 inode_item->size = cpu_to_le64(3);
380 inode_item->nlink = cpu_to_le32(1);
381 inode_item->nbytes = cpu_to_le64(root->leafsize);
382 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
383
384 root_item.flags = 0;
385 root_item.byte_limit = 0;
386 inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
387
388 btrfs_set_root_bytenr(&root_item, leaf->start);
389 btrfs_set_root_generation(&root_item, trans->transid);
390 btrfs_set_root_level(&root_item, 0);
391 btrfs_set_root_refs(&root_item, 1);
392 btrfs_set_root_used(&root_item, leaf->len);
393 btrfs_set_root_last_snapshot(&root_item, 0);
394
395 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
396 root_item.drop_level = 0;
397
398 btrfs_tree_unlock(leaf);
399 free_extent_buffer(leaf);
400 leaf = NULL;
401
402 btrfs_set_root_dirid(&root_item, new_dirid);
403
404 key.objectid = objectid;
405 key.offset = 0;
406 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
407 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
408 &root_item);
409 if (ret)
410 goto fail;
411
412 key.offset = (u64)-1;
413 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
414 BUG_ON(IS_ERR(new_root));
415
416 btrfs_record_root_in_trans(trans, new_root);
417
418 ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
419 /*
420 * insert the directory item
421 */
422 ret = btrfs_set_inode_index(dir, &index);
423 BUG_ON(ret);
424
425 ret = btrfs_insert_dir_item(trans, root,
426 name, namelen, dir, &key,
427 BTRFS_FT_DIR, index);
428 if (ret)
429 goto fail;
430
431 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
432 ret = btrfs_update_inode(trans, root, dir);
433 BUG_ON(ret);
434
435 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
436 objectid, root->root_key.objectid,
437 btrfs_ino(dir), index, name, namelen);
438
439 BUG_ON(ret);
440
441 d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
442 fail:
443 if (async_transid) {
444 *async_transid = trans->transid;
445 err = btrfs_commit_transaction_async(trans, root, 1);
446 } else {
447 err = btrfs_commit_transaction(trans, root);
448 }
449 if (err && !ret)
450 ret = err;
451 return ret;
452 }
453
454 static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
455 char *name, int namelen, u64 *async_transid,
456 bool readonly)
457 {
458 struct inode *inode;
459 struct btrfs_pending_snapshot *pending_snapshot;
460 struct btrfs_trans_handle *trans;
461 int ret;
462
463 if (!root->ref_cows)
464 return -EINVAL;
465
466 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
467 if (!pending_snapshot)
468 return -ENOMEM;
469
470 btrfs_init_block_rsv(&pending_snapshot->block_rsv);
471 pending_snapshot->dentry = dentry;
472 pending_snapshot->root = root;
473 pending_snapshot->readonly = readonly;
474
475 trans = btrfs_start_transaction(root->fs_info->extent_root, 5);
476 if (IS_ERR(trans)) {
477 ret = PTR_ERR(trans);
478 goto fail;
479 }
480
481 ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
482 BUG_ON(ret);
483
484 spin_lock(&root->fs_info->trans_lock);
485 list_add(&pending_snapshot->list,
486 &trans->transaction->pending_snapshots);
487 spin_unlock(&root->fs_info->trans_lock);
488 if (async_transid) {
489 *async_transid = trans->transid;
490 ret = btrfs_commit_transaction_async(trans,
491 root->fs_info->extent_root, 1);
492 } else {
493 ret = btrfs_commit_transaction(trans,
494 root->fs_info->extent_root);
495 }
496 BUG_ON(ret);
497
498 ret = pending_snapshot->error;
499 if (ret)
500 goto fail;
501
502 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
503 if (ret)
504 goto fail;
505
506 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
507 if (IS_ERR(inode)) {
508 ret = PTR_ERR(inode);
509 goto fail;
510 }
511 BUG_ON(!inode);
512 d_instantiate(dentry, inode);
513 ret = 0;
514 fail:
515 kfree(pending_snapshot);
516 return ret;
517 }
518
519 /* copy of check_sticky in fs/namei.c()
520 * It's inline, so penalty for filesystems that don't use sticky bit is
521 * minimal.
522 */
523 static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
524 {
525 uid_t fsuid = current_fsuid();
526
527 if (!(dir->i_mode & S_ISVTX))
528 return 0;
529 if (inode->i_uid == fsuid)
530 return 0;
531 if (dir->i_uid == fsuid)
532 return 0;
533 return !capable(CAP_FOWNER);
534 }
535
536 /* copy of may_delete in fs/namei.c()
537 * Check whether we can remove a link victim from directory dir, check
538 * whether the type of victim is right.
539 * 1. We can't do it if dir is read-only (done in permission())
540 * 2. We should have write and exec permissions on dir
541 * 3. We can't remove anything from append-only dir
542 * 4. We can't do anything with immutable dir (done in permission())
543 * 5. If the sticky bit on dir is set we should either
544 * a. be owner of dir, or
545 * b. be owner of victim, or
546 * c. have CAP_FOWNER capability
547 * 6. If the victim is append-only or immutable we can't do antyhing with
548 * links pointing to it.
549 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
550 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
551 * 9. We can't remove a root or mountpoint.
552 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
553 * nfs_async_unlink().
554 */
555
556 static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
557 {
558 int error;
559
560 if (!victim->d_inode)
561 return -ENOENT;
562
563 BUG_ON(victim->d_parent->d_inode != dir);
564 audit_inode_child(victim, dir);
565
566 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
567 if (error)
568 return error;
569 if (IS_APPEND(dir))
570 return -EPERM;
571 if (btrfs_check_sticky(dir, victim->d_inode)||
572 IS_APPEND(victim->d_inode)||
573 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
574 return -EPERM;
575 if (isdir) {
576 if (!S_ISDIR(victim->d_inode->i_mode))
577 return -ENOTDIR;
578 if (IS_ROOT(victim))
579 return -EBUSY;
580 } else if (S_ISDIR(victim->d_inode->i_mode))
581 return -EISDIR;
582 if (IS_DEADDIR(dir))
583 return -ENOENT;
584 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
585 return -EBUSY;
586 return 0;
587 }
588
589 /* copy of may_create in fs/namei.c() */
590 static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
591 {
592 if (child->d_inode)
593 return -EEXIST;
594 if (IS_DEADDIR(dir))
595 return -ENOENT;
596 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
597 }
598
599 /*
600 * Create a new subvolume below @parent. This is largely modeled after
601 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
602 * inside this filesystem so it's quite a bit simpler.
603 */
604 static noinline int btrfs_mksubvol(struct path *parent,
605 char *name, int namelen,
606 struct btrfs_root *snap_src,
607 u64 *async_transid, bool readonly)
608 {
609 struct inode *dir = parent->dentry->d_inode;
610 struct dentry *dentry;
611 int error;
612
613 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
614
615 dentry = lookup_one_len(name, parent->dentry, namelen);
616 error = PTR_ERR(dentry);
617 if (IS_ERR(dentry))
618 goto out_unlock;
619
620 error = -EEXIST;
621 if (dentry->d_inode)
622 goto out_dput;
623
624 error = mnt_want_write(parent->mnt);
625 if (error)
626 goto out_dput;
627
628 error = btrfs_may_create(dir, dentry);
629 if (error)
630 goto out_drop_write;
631
632 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
633
634 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
635 goto out_up_read;
636
637 if (snap_src) {
638 error = create_snapshot(snap_src, dentry,
639 name, namelen, async_transid, readonly);
640 } else {
641 error = create_subvol(BTRFS_I(dir)->root, dentry,
642 name, namelen, async_transid);
643 }
644 if (!error)
645 fsnotify_mkdir(dir, dentry);
646 out_up_read:
647 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
648 out_drop_write:
649 mnt_drop_write(parent->mnt);
650 out_dput:
651 dput(dentry);
652 out_unlock:
653 mutex_unlock(&dir->i_mutex);
654 return error;
655 }
656
657 /*
658 * When we're defragging a range, we don't want to kick it off again
659 * if it is really just waiting for delalloc to send it down.
660 * If we find a nice big extent or delalloc range for the bytes in the
661 * file you want to defrag, we return 0 to let you know to skip this
662 * part of the file
663 */
664 static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
665 {
666 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
667 struct extent_map *em = NULL;
668 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
669 u64 end;
670
671 read_lock(&em_tree->lock);
672 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
673 read_unlock(&em_tree->lock);
674
675 if (em) {
676 end = extent_map_end(em);
677 free_extent_map(em);
678 if (end - offset > thresh)
679 return 0;
680 }
681 /* if we already have a nice delalloc here, just stop */
682 thresh /= 2;
683 end = count_range_bits(io_tree, &offset, offset + thresh,
684 thresh, EXTENT_DELALLOC, 1);
685 if (end >= thresh)
686 return 0;
687 return 1;
688 }
689
690 /*
691 * helper function to walk through a file and find extents
692 * newer than a specific transid, and smaller than thresh.
693 *
694 * This is used by the defragging code to find new and small
695 * extents
696 */
697 static int find_new_extents(struct btrfs_root *root,
698 struct inode *inode, u64 newer_than,
699 u64 *off, int thresh)
700 {
701 struct btrfs_path *path;
702 struct btrfs_key min_key;
703 struct btrfs_key max_key;
704 struct extent_buffer *leaf;
705 struct btrfs_file_extent_item *extent;
706 int type;
707 int ret;
708 u64 ino = btrfs_ino(inode);
709
710 path = btrfs_alloc_path();
711 if (!path)
712 return -ENOMEM;
713
714 min_key.objectid = ino;
715 min_key.type = BTRFS_EXTENT_DATA_KEY;
716 min_key.offset = *off;
717
718 max_key.objectid = ino;
719 max_key.type = (u8)-1;
720 max_key.offset = (u64)-1;
721
722 path->keep_locks = 1;
723
724 while(1) {
725 ret = btrfs_search_forward(root, &min_key, &max_key,
726 path, 0, newer_than);
727 if (ret != 0)
728 goto none;
729 if (min_key.objectid != ino)
730 goto none;
731 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
732 goto none;
733
734 leaf = path->nodes[0];
735 extent = btrfs_item_ptr(leaf, path->slots[0],
736 struct btrfs_file_extent_item);
737
738 type = btrfs_file_extent_type(leaf, extent);
739 if (type == BTRFS_FILE_EXTENT_REG &&
740 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
741 check_defrag_in_cache(inode, min_key.offset, thresh)) {
742 *off = min_key.offset;
743 btrfs_free_path(path);
744 return 0;
745 }
746
747 if (min_key.offset == (u64)-1)
748 goto none;
749
750 min_key.offset++;
751 btrfs_release_path(path);
752 }
753 none:
754 btrfs_free_path(path);
755 return -ENOENT;
756 }
757
758 static int should_defrag_range(struct inode *inode, u64 start, u64 len,
759 int thresh, u64 *last_len, u64 *skip,
760 u64 *defrag_end)
761 {
762 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
763 struct extent_map *em = NULL;
764 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
765 int ret = 1;
766
767 /*
768 * make sure that once we start defragging and extent, we keep on
769 * defragging it
770 */
771 if (start < *defrag_end)
772 return 1;
773
774 *skip = 0;
775
776 /*
777 * hopefully we have this extent in the tree already, try without
778 * the full extent lock
779 */
780 read_lock(&em_tree->lock);
781 em = lookup_extent_mapping(em_tree, start, len);
782 read_unlock(&em_tree->lock);
783
784 if (!em) {
785 /* get the big lock and read metadata off disk */
786 lock_extent(io_tree, start, start + len - 1, GFP_NOFS);
787 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
788 unlock_extent(io_tree, start, start + len - 1, GFP_NOFS);
789
790 if (IS_ERR(em))
791 return 0;
792 }
793
794 /* this will cover holes, and inline extents */
795 if (em->block_start >= EXTENT_MAP_LAST_BYTE)
796 ret = 0;
797
798 /*
799 * we hit a real extent, if it is big don't bother defragging it again
800 */
801 if ((*last_len == 0 || *last_len >= thresh) && em->len >= thresh)
802 ret = 0;
803
804 /*
805 * last_len ends up being a counter of how many bytes we've defragged.
806 * every time we choose not to defrag an extent, we reset *last_len
807 * so that the next tiny extent will force a defrag.
808 *
809 * The end result of this is that tiny extents before a single big
810 * extent will force at least part of that big extent to be defragged.
811 */
812 if (ret) {
813 *last_len += len;
814 *defrag_end = extent_map_end(em);
815 } else {
816 *last_len = 0;
817 *skip = extent_map_end(em);
818 *defrag_end = 0;
819 }
820
821 free_extent_map(em);
822 return ret;
823 }
824
825 /*
826 * it doesn't do much good to defrag one or two pages
827 * at a time. This pulls in a nice chunk of pages
828 * to COW and defrag.
829 *
830 * It also makes sure the delalloc code has enough
831 * dirty data to avoid making new small extents as part
832 * of the defrag
833 *
834 * It's a good idea to start RA on this range
835 * before calling this.
836 */
837 static int cluster_pages_for_defrag(struct inode *inode,
838 struct page **pages,
839 unsigned long start_index,
840 int num_pages)
841 {
842 unsigned long file_end;
843 u64 isize = i_size_read(inode);
844 u64 page_start;
845 u64 page_end;
846 int ret;
847 int i;
848 int i_done;
849 struct btrfs_ordered_extent *ordered;
850 struct extent_state *cached_state = NULL;
851 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
852
853 if (isize == 0)
854 return 0;
855 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
856
857 ret = btrfs_delalloc_reserve_space(inode,
858 num_pages << PAGE_CACHE_SHIFT);
859 if (ret)
860 return ret;
861 again:
862 ret = 0;
863 i_done = 0;
864
865 /* step one, lock all the pages */
866 for (i = 0; i < num_pages; i++) {
867 struct page *page;
868 page = find_or_create_page(inode->i_mapping,
869 start_index + i, mask);
870 if (!page)
871 break;
872
873 if (!PageUptodate(page)) {
874 btrfs_readpage(NULL, page);
875 lock_page(page);
876 if (!PageUptodate(page)) {
877 unlock_page(page);
878 page_cache_release(page);
879 ret = -EIO;
880 break;
881 }
882 }
883 isize = i_size_read(inode);
884 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
885 if (!isize || page->index > file_end ||
886 page->mapping != inode->i_mapping) {
887 /* whoops, we blew past eof, skip this page */
888 unlock_page(page);
889 page_cache_release(page);
890 break;
891 }
892 pages[i] = page;
893 i_done++;
894 }
895 if (!i_done || ret)
896 goto out;
897
898 if (!(inode->i_sb->s_flags & MS_ACTIVE))
899 goto out;
900
901 /*
902 * so now we have a nice long stream of locked
903 * and up to date pages, lets wait on them
904 */
905 for (i = 0; i < i_done; i++)
906 wait_on_page_writeback(pages[i]);
907
908 page_start = page_offset(pages[0]);
909 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
910
911 lock_extent_bits(&BTRFS_I(inode)->io_tree,
912 page_start, page_end - 1, 0, &cached_state,
913 GFP_NOFS);
914 ordered = btrfs_lookup_first_ordered_extent(inode, page_end - 1);
915 if (ordered &&
916 ordered->file_offset + ordered->len > page_start &&
917 ordered->file_offset < page_end) {
918 btrfs_put_ordered_extent(ordered);
919 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
920 page_start, page_end - 1,
921 &cached_state, GFP_NOFS);
922 for (i = 0; i < i_done; i++) {
923 unlock_page(pages[i]);
924 page_cache_release(pages[i]);
925 }
926 btrfs_wait_ordered_range(inode, page_start,
927 page_end - page_start);
928 goto again;
929 }
930 if (ordered)
931 btrfs_put_ordered_extent(ordered);
932
933 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
934 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
935 EXTENT_DO_ACCOUNTING, 0, 0, &cached_state,
936 GFP_NOFS);
937
938 if (i_done != num_pages) {
939 spin_lock(&BTRFS_I(inode)->lock);
940 BTRFS_I(inode)->outstanding_extents++;
941 spin_unlock(&BTRFS_I(inode)->lock);
942 btrfs_delalloc_release_space(inode,
943 (num_pages - i_done) << PAGE_CACHE_SHIFT);
944 }
945
946
947 btrfs_set_extent_delalloc(inode, page_start, page_end - 1,
948 &cached_state);
949
950 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
951 page_start, page_end - 1, &cached_state,
952 GFP_NOFS);
953
954 for (i = 0; i < i_done; i++) {
955 clear_page_dirty_for_io(pages[i]);
956 ClearPageChecked(pages[i]);
957 set_page_extent_mapped(pages[i]);
958 set_page_dirty(pages[i]);
959 unlock_page(pages[i]);
960 page_cache_release(pages[i]);
961 }
962 return i_done;
963 out:
964 for (i = 0; i < i_done; i++) {
965 unlock_page(pages[i]);
966 page_cache_release(pages[i]);
967 }
968 btrfs_delalloc_release_space(inode, num_pages << PAGE_CACHE_SHIFT);
969 return ret;
970
971 }
972
973 int btrfs_defrag_file(struct inode *inode, struct file *file,
974 struct btrfs_ioctl_defrag_range_args *range,
975 u64 newer_than, unsigned long max_to_defrag)
976 {
977 struct btrfs_root *root = BTRFS_I(inode)->root;
978 struct btrfs_super_block *disk_super;
979 struct file_ra_state *ra = NULL;
980 unsigned long last_index;
981 u64 features;
982 u64 last_len = 0;
983 u64 skip = 0;
984 u64 defrag_end = 0;
985 u64 newer_off = range->start;
986 int newer_left = 0;
987 unsigned long i;
988 int ret;
989 int defrag_count = 0;
990 int compress_type = BTRFS_COMPRESS_ZLIB;
991 int extent_thresh = range->extent_thresh;
992 int newer_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
993 u64 new_align = ~((u64)128 * 1024 - 1);
994 struct page **pages = NULL;
995
996 if (extent_thresh == 0)
997 extent_thresh = 256 * 1024;
998
999 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1000 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1001 return -EINVAL;
1002 if (range->compress_type)
1003 compress_type = range->compress_type;
1004 }
1005
1006 if (inode->i_size == 0)
1007 return 0;
1008
1009 /*
1010 * if we were not given a file, allocate a readahead
1011 * context
1012 */
1013 if (!file) {
1014 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1015 if (!ra)
1016 return -ENOMEM;
1017 file_ra_state_init(ra, inode->i_mapping);
1018 } else {
1019 ra = &file->f_ra;
1020 }
1021
1022 pages = kmalloc(sizeof(struct page *) * newer_cluster,
1023 GFP_NOFS);
1024 if (!pages) {
1025 ret = -ENOMEM;
1026 goto out_ra;
1027 }
1028
1029 /* find the last page to defrag */
1030 if (range->start + range->len > range->start) {
1031 last_index = min_t(u64, inode->i_size - 1,
1032 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1033 } else {
1034 last_index = (inode->i_size - 1) >> PAGE_CACHE_SHIFT;
1035 }
1036
1037 if (newer_than) {
1038 ret = find_new_extents(root, inode, newer_than,
1039 &newer_off, 64 * 1024);
1040 if (!ret) {
1041 range->start = newer_off;
1042 /*
1043 * we always align our defrag to help keep
1044 * the extents in the file evenly spaced
1045 */
1046 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1047 newer_left = newer_cluster;
1048 } else
1049 goto out_ra;
1050 } else {
1051 i = range->start >> PAGE_CACHE_SHIFT;
1052 }
1053 if (!max_to_defrag)
1054 max_to_defrag = last_index - 1;
1055
1056 /*
1057 * make writeback starts from i, so the defrag range can be
1058 * written sequentially.
1059 */
1060 if (i < inode->i_mapping->writeback_index)
1061 inode->i_mapping->writeback_index = i;
1062
1063 while (i <= last_index && defrag_count < max_to_defrag &&
1064 (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1065 PAGE_CACHE_SHIFT)) {
1066 /*
1067 * make sure we stop running if someone unmounts
1068 * the FS
1069 */
1070 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1071 break;
1072
1073 if (!newer_than &&
1074 !should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1075 PAGE_CACHE_SIZE,
1076 extent_thresh,
1077 &last_len, &skip,
1078 &defrag_end)) {
1079 unsigned long next;
1080 /*
1081 * the should_defrag function tells us how much to skip
1082 * bump our counter by the suggested amount
1083 */
1084 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1085 i = max(i + 1, next);
1086 continue;
1087 }
1088 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1089 BTRFS_I(inode)->force_compress = compress_type;
1090
1091 btrfs_force_ra(inode->i_mapping, ra, file, i, newer_cluster);
1092
1093 ret = cluster_pages_for_defrag(inode, pages, i, newer_cluster);
1094 if (ret < 0)
1095 goto out_ra;
1096
1097 defrag_count += ret;
1098 balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret);
1099 i += ret;
1100
1101 if (newer_than) {
1102 if (newer_off == (u64)-1)
1103 break;
1104
1105 newer_off = max(newer_off + 1,
1106 (u64)i << PAGE_CACHE_SHIFT);
1107
1108 ret = find_new_extents(root, inode,
1109 newer_than, &newer_off,
1110 64 * 1024);
1111 if (!ret) {
1112 range->start = newer_off;
1113 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1114 newer_left = newer_cluster;
1115 } else {
1116 break;
1117 }
1118 } else {
1119 i++;
1120 }
1121 }
1122
1123 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1124 filemap_flush(inode->i_mapping);
1125
1126 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1127 /* the filemap_flush will queue IO into the worker threads, but
1128 * we have to make sure the IO is actually started and that
1129 * ordered extents get created before we return
1130 */
1131 atomic_inc(&root->fs_info->async_submit_draining);
1132 while (atomic_read(&root->fs_info->nr_async_submits) ||
1133 atomic_read(&root->fs_info->async_delalloc_pages)) {
1134 wait_event(root->fs_info->async_submit_wait,
1135 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1136 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1137 }
1138 atomic_dec(&root->fs_info->async_submit_draining);
1139
1140 mutex_lock(&inode->i_mutex);
1141 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1142 mutex_unlock(&inode->i_mutex);
1143 }
1144
1145 disk_super = &root->fs_info->super_copy;
1146 features = btrfs_super_incompat_flags(disk_super);
1147 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1148 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
1149 btrfs_set_super_incompat_flags(disk_super, features);
1150 }
1151
1152 if (!file)
1153 kfree(ra);
1154 return defrag_count;
1155
1156 out_ra:
1157 if (!file)
1158 kfree(ra);
1159 kfree(pages);
1160 return ret;
1161 }
1162
1163 static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
1164 void __user *arg)
1165 {
1166 u64 new_size;
1167 u64 old_size;
1168 u64 devid = 1;
1169 struct btrfs_ioctl_vol_args *vol_args;
1170 struct btrfs_trans_handle *trans;
1171 struct btrfs_device *device = NULL;
1172 char *sizestr;
1173 char *devstr = NULL;
1174 int ret = 0;
1175 int mod = 0;
1176
1177 if (root->fs_info->sb->s_flags & MS_RDONLY)
1178 return -EROFS;
1179
1180 if (!capable(CAP_SYS_ADMIN))
1181 return -EPERM;
1182
1183 vol_args = memdup_user(arg, sizeof(*vol_args));
1184 if (IS_ERR(vol_args))
1185 return PTR_ERR(vol_args);
1186
1187 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1188
1189 mutex_lock(&root->fs_info->volume_mutex);
1190 sizestr = vol_args->name;
1191 devstr = strchr(sizestr, ':');
1192 if (devstr) {
1193 char *end;
1194 sizestr = devstr + 1;
1195 *devstr = '\0';
1196 devstr = vol_args->name;
1197 devid = simple_strtoull(devstr, &end, 10);
1198 printk(KERN_INFO "resizing devid %llu\n",
1199 (unsigned long long)devid);
1200 }
1201 device = btrfs_find_device(root, devid, NULL, NULL);
1202 if (!device) {
1203 printk(KERN_INFO "resizer unable to find device %llu\n",
1204 (unsigned long long)devid);
1205 ret = -EINVAL;
1206 goto out_unlock;
1207 }
1208 if (!strcmp(sizestr, "max"))
1209 new_size = device->bdev->bd_inode->i_size;
1210 else {
1211 if (sizestr[0] == '-') {
1212 mod = -1;
1213 sizestr++;
1214 } else if (sizestr[0] == '+') {
1215 mod = 1;
1216 sizestr++;
1217 }
1218 new_size = memparse(sizestr, NULL);
1219 if (new_size == 0) {
1220 ret = -EINVAL;
1221 goto out_unlock;
1222 }
1223 }
1224
1225 old_size = device->total_bytes;
1226
1227 if (mod < 0) {
1228 if (new_size > old_size) {
1229 ret = -EINVAL;
1230 goto out_unlock;
1231 }
1232 new_size = old_size - new_size;
1233 } else if (mod > 0) {
1234 new_size = old_size + new_size;
1235 }
1236
1237 if (new_size < 256 * 1024 * 1024) {
1238 ret = -EINVAL;
1239 goto out_unlock;
1240 }
1241 if (new_size > device->bdev->bd_inode->i_size) {
1242 ret = -EFBIG;
1243 goto out_unlock;
1244 }
1245
1246 do_div(new_size, root->sectorsize);
1247 new_size *= root->sectorsize;
1248
1249 printk(KERN_INFO "new size for %s is %llu\n",
1250 device->name, (unsigned long long)new_size);
1251
1252 if (new_size > old_size) {
1253 trans = btrfs_start_transaction(root, 0);
1254 if (IS_ERR(trans)) {
1255 ret = PTR_ERR(trans);
1256 goto out_unlock;
1257 }
1258 ret = btrfs_grow_device(trans, device, new_size);
1259 btrfs_commit_transaction(trans, root);
1260 } else {
1261 ret = btrfs_shrink_device(device, new_size);
1262 }
1263
1264 out_unlock:
1265 mutex_unlock(&root->fs_info->volume_mutex);
1266 kfree(vol_args);
1267 return ret;
1268 }
1269
1270 static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1271 char *name,
1272 unsigned long fd,
1273 int subvol,
1274 u64 *transid,
1275 bool readonly)
1276 {
1277 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
1278 struct file *src_file;
1279 int namelen;
1280 int ret = 0;
1281
1282 if (root->fs_info->sb->s_flags & MS_RDONLY)
1283 return -EROFS;
1284
1285 namelen = strlen(name);
1286 if (strchr(name, '/')) {
1287 ret = -EINVAL;
1288 goto out;
1289 }
1290
1291 if (subvol) {
1292 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1293 NULL, transid, readonly);
1294 } else {
1295 struct inode *src_inode;
1296 src_file = fget(fd);
1297 if (!src_file) {
1298 ret = -EINVAL;
1299 goto out;
1300 }
1301
1302 src_inode = src_file->f_path.dentry->d_inode;
1303 if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
1304 printk(KERN_INFO "btrfs: Snapshot src from "
1305 "another FS\n");
1306 ret = -EINVAL;
1307 fput(src_file);
1308 goto out;
1309 }
1310 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1311 BTRFS_I(src_inode)->root,
1312 transid, readonly);
1313 fput(src_file);
1314 }
1315 out:
1316 return ret;
1317 }
1318
1319 static noinline int btrfs_ioctl_snap_create(struct file *file,
1320 void __user *arg, int subvol)
1321 {
1322 struct btrfs_ioctl_vol_args *vol_args;
1323 int ret;
1324
1325 vol_args = memdup_user(arg, sizeof(*vol_args));
1326 if (IS_ERR(vol_args))
1327 return PTR_ERR(vol_args);
1328 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1329
1330 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1331 vol_args->fd, subvol,
1332 NULL, false);
1333
1334 kfree(vol_args);
1335 return ret;
1336 }
1337
1338 static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1339 void __user *arg, int subvol)
1340 {
1341 struct btrfs_ioctl_vol_args_v2 *vol_args;
1342 int ret;
1343 u64 transid = 0;
1344 u64 *ptr = NULL;
1345 bool readonly = false;
1346
1347 vol_args = memdup_user(arg, sizeof(*vol_args));
1348 if (IS_ERR(vol_args))
1349 return PTR_ERR(vol_args);
1350 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
1351
1352 if (vol_args->flags &
1353 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY)) {
1354 ret = -EOPNOTSUPP;
1355 goto out;
1356 }
1357
1358 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1359 ptr = &transid;
1360 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1361 readonly = true;
1362
1363 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
1364 vol_args->fd, subvol,
1365 ptr, readonly);
1366
1367 if (ret == 0 && ptr &&
1368 copy_to_user(arg +
1369 offsetof(struct btrfs_ioctl_vol_args_v2,
1370 transid), ptr, sizeof(*ptr)))
1371 ret = -EFAULT;
1372 out:
1373 kfree(vol_args);
1374 return ret;
1375 }
1376
1377 static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1378 void __user *arg)
1379 {
1380 struct inode *inode = fdentry(file)->d_inode;
1381 struct btrfs_root *root = BTRFS_I(inode)->root;
1382 int ret = 0;
1383 u64 flags = 0;
1384
1385 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1386 return -EINVAL;
1387
1388 down_read(&root->fs_info->subvol_sem);
1389 if (btrfs_root_readonly(root))
1390 flags |= BTRFS_SUBVOL_RDONLY;
1391 up_read(&root->fs_info->subvol_sem);
1392
1393 if (copy_to_user(arg, &flags, sizeof(flags)))
1394 ret = -EFAULT;
1395
1396 return ret;
1397 }
1398
1399 static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1400 void __user *arg)
1401 {
1402 struct inode *inode = fdentry(file)->d_inode;
1403 struct btrfs_root *root = BTRFS_I(inode)->root;
1404 struct btrfs_trans_handle *trans;
1405 u64 root_flags;
1406 u64 flags;
1407 int ret = 0;
1408
1409 if (root->fs_info->sb->s_flags & MS_RDONLY)
1410 return -EROFS;
1411
1412 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
1413 return -EINVAL;
1414
1415 if (copy_from_user(&flags, arg, sizeof(flags)))
1416 return -EFAULT;
1417
1418 if (flags & BTRFS_SUBVOL_CREATE_ASYNC)
1419 return -EINVAL;
1420
1421 if (flags & ~BTRFS_SUBVOL_RDONLY)
1422 return -EOPNOTSUPP;
1423
1424 if (!inode_owner_or_capable(inode))
1425 return -EACCES;
1426
1427 down_write(&root->fs_info->subvol_sem);
1428
1429 /* nothing to do */
1430 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1431 goto out;
1432
1433 root_flags = btrfs_root_flags(&root->root_item);
1434 if (flags & BTRFS_SUBVOL_RDONLY)
1435 btrfs_set_root_flags(&root->root_item,
1436 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1437 else
1438 btrfs_set_root_flags(&root->root_item,
1439 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1440
1441 trans = btrfs_start_transaction(root, 1);
1442 if (IS_ERR(trans)) {
1443 ret = PTR_ERR(trans);
1444 goto out_reset;
1445 }
1446
1447 ret = btrfs_update_root(trans, root->fs_info->tree_root,
1448 &root->root_key, &root->root_item);
1449
1450 btrfs_commit_transaction(trans, root);
1451 out_reset:
1452 if (ret)
1453 btrfs_set_root_flags(&root->root_item, root_flags);
1454 out:
1455 up_write(&root->fs_info->subvol_sem);
1456 return ret;
1457 }
1458
1459 /*
1460 * helper to check if the subvolume references other subvolumes
1461 */
1462 static noinline int may_destroy_subvol(struct btrfs_root *root)
1463 {
1464 struct btrfs_path *path;
1465 struct btrfs_key key;
1466 int ret;
1467
1468 path = btrfs_alloc_path();
1469 if (!path)
1470 return -ENOMEM;
1471
1472 key.objectid = root->root_key.objectid;
1473 key.type = BTRFS_ROOT_REF_KEY;
1474 key.offset = (u64)-1;
1475
1476 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1477 &key, path, 0, 0);
1478 if (ret < 0)
1479 goto out;
1480 BUG_ON(ret == 0);
1481
1482 ret = 0;
1483 if (path->slots[0] > 0) {
1484 path->slots[0]--;
1485 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1486 if (key.objectid == root->root_key.objectid &&
1487 key.type == BTRFS_ROOT_REF_KEY)
1488 ret = -ENOTEMPTY;
1489 }
1490 out:
1491 btrfs_free_path(path);
1492 return ret;
1493 }
1494
1495 static noinline int key_in_sk(struct btrfs_key *key,
1496 struct btrfs_ioctl_search_key *sk)
1497 {
1498 struct btrfs_key test;
1499 int ret;
1500
1501 test.objectid = sk->min_objectid;
1502 test.type = sk->min_type;
1503 test.offset = sk->min_offset;
1504
1505 ret = btrfs_comp_cpu_keys(key, &test);
1506 if (ret < 0)
1507 return 0;
1508
1509 test.objectid = sk->max_objectid;
1510 test.type = sk->max_type;
1511 test.offset = sk->max_offset;
1512
1513 ret = btrfs_comp_cpu_keys(key, &test);
1514 if (ret > 0)
1515 return 0;
1516 return 1;
1517 }
1518
1519 static noinline int copy_to_sk(struct btrfs_root *root,
1520 struct btrfs_path *path,
1521 struct btrfs_key *key,
1522 struct btrfs_ioctl_search_key *sk,
1523 char *buf,
1524 unsigned long *sk_offset,
1525 int *num_found)
1526 {
1527 u64 found_transid;
1528 struct extent_buffer *leaf;
1529 struct btrfs_ioctl_search_header sh;
1530 unsigned long item_off;
1531 unsigned long item_len;
1532 int nritems;
1533 int i;
1534 int slot;
1535 int ret = 0;
1536
1537 leaf = path->nodes[0];
1538 slot = path->slots[0];
1539 nritems = btrfs_header_nritems(leaf);
1540
1541 if (btrfs_header_generation(leaf) > sk->max_transid) {
1542 i = nritems;
1543 goto advance_key;
1544 }
1545 found_transid = btrfs_header_generation(leaf);
1546
1547 for (i = slot; i < nritems; i++) {
1548 item_off = btrfs_item_ptr_offset(leaf, i);
1549 item_len = btrfs_item_size_nr(leaf, i);
1550
1551 if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1552 item_len = 0;
1553
1554 if (sizeof(sh) + item_len + *sk_offset >
1555 BTRFS_SEARCH_ARGS_BUFSIZE) {
1556 ret = 1;
1557 goto overflow;
1558 }
1559
1560 btrfs_item_key_to_cpu(leaf, key, i);
1561 if (!key_in_sk(key, sk))
1562 continue;
1563
1564 sh.objectid = key->objectid;
1565 sh.offset = key->offset;
1566 sh.type = key->type;
1567 sh.len = item_len;
1568 sh.transid = found_transid;
1569
1570 /* copy search result header */
1571 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1572 *sk_offset += sizeof(sh);
1573
1574 if (item_len) {
1575 char *p = buf + *sk_offset;
1576 /* copy the item */
1577 read_extent_buffer(leaf, p,
1578 item_off, item_len);
1579 *sk_offset += item_len;
1580 }
1581 (*num_found)++;
1582
1583 if (*num_found >= sk->nr_items)
1584 break;
1585 }
1586 advance_key:
1587 ret = 0;
1588 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
1589 key->offset++;
1590 else if (key->type < (u8)-1 && key->type < sk->max_type) {
1591 key->offset = 0;
1592 key->type++;
1593 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1594 key->offset = 0;
1595 key->type = 0;
1596 key->objectid++;
1597 } else
1598 ret = 1;
1599 overflow:
1600 return ret;
1601 }
1602
1603 static noinline int search_ioctl(struct inode *inode,
1604 struct btrfs_ioctl_search_args *args)
1605 {
1606 struct btrfs_root *root;
1607 struct btrfs_key key;
1608 struct btrfs_key max_key;
1609 struct btrfs_path *path;
1610 struct btrfs_ioctl_search_key *sk = &args->key;
1611 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1612 int ret;
1613 int num_found = 0;
1614 unsigned long sk_offset = 0;
1615
1616 path = btrfs_alloc_path();
1617 if (!path)
1618 return -ENOMEM;
1619
1620 if (sk->tree_id == 0) {
1621 /* search the root of the inode that was passed */
1622 root = BTRFS_I(inode)->root;
1623 } else {
1624 key.objectid = sk->tree_id;
1625 key.type = BTRFS_ROOT_ITEM_KEY;
1626 key.offset = (u64)-1;
1627 root = btrfs_read_fs_root_no_name(info, &key);
1628 if (IS_ERR(root)) {
1629 printk(KERN_ERR "could not find root %llu\n",
1630 sk->tree_id);
1631 btrfs_free_path(path);
1632 return -ENOENT;
1633 }
1634 }
1635
1636 key.objectid = sk->min_objectid;
1637 key.type = sk->min_type;
1638 key.offset = sk->min_offset;
1639
1640 max_key.objectid = sk->max_objectid;
1641 max_key.type = sk->max_type;
1642 max_key.offset = sk->max_offset;
1643
1644 path->keep_locks = 1;
1645
1646 while(1) {
1647 ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1648 sk->min_transid);
1649 if (ret != 0) {
1650 if (ret > 0)
1651 ret = 0;
1652 goto err;
1653 }
1654 ret = copy_to_sk(root, path, &key, sk, args->buf,
1655 &sk_offset, &num_found);
1656 btrfs_release_path(path);
1657 if (ret || num_found >= sk->nr_items)
1658 break;
1659
1660 }
1661 ret = 0;
1662 err:
1663 sk->nr_items = num_found;
1664 btrfs_free_path(path);
1665 return ret;
1666 }
1667
1668 static noinline int btrfs_ioctl_tree_search(struct file *file,
1669 void __user *argp)
1670 {
1671 struct btrfs_ioctl_search_args *args;
1672 struct inode *inode;
1673 int ret;
1674
1675 if (!capable(CAP_SYS_ADMIN))
1676 return -EPERM;
1677
1678 args = memdup_user(argp, sizeof(*args));
1679 if (IS_ERR(args))
1680 return PTR_ERR(args);
1681
1682 inode = fdentry(file)->d_inode;
1683 ret = search_ioctl(inode, args);
1684 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1685 ret = -EFAULT;
1686 kfree(args);
1687 return ret;
1688 }
1689
1690 /*
1691 * Search INODE_REFs to identify path name of 'dirid' directory
1692 * in a 'tree_id' tree. and sets path name to 'name'.
1693 */
1694 static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1695 u64 tree_id, u64 dirid, char *name)
1696 {
1697 struct btrfs_root *root;
1698 struct btrfs_key key;
1699 char *ptr;
1700 int ret = -1;
1701 int slot;
1702 int len;
1703 int total_len = 0;
1704 struct btrfs_inode_ref *iref;
1705 struct extent_buffer *l;
1706 struct btrfs_path *path;
1707
1708 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1709 name[0]='\0';
1710 return 0;
1711 }
1712
1713 path = btrfs_alloc_path();
1714 if (!path)
1715 return -ENOMEM;
1716
1717 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
1718
1719 key.objectid = tree_id;
1720 key.type = BTRFS_ROOT_ITEM_KEY;
1721 key.offset = (u64)-1;
1722 root = btrfs_read_fs_root_no_name(info, &key);
1723 if (IS_ERR(root)) {
1724 printk(KERN_ERR "could not find root %llu\n", tree_id);
1725 ret = -ENOENT;
1726 goto out;
1727 }
1728
1729 key.objectid = dirid;
1730 key.type = BTRFS_INODE_REF_KEY;
1731 key.offset = (u64)-1;
1732
1733 while(1) {
1734 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1735 if (ret < 0)
1736 goto out;
1737
1738 l = path->nodes[0];
1739 slot = path->slots[0];
1740 if (ret > 0 && slot > 0)
1741 slot--;
1742 btrfs_item_key_to_cpu(l, &key, slot);
1743
1744 if (ret > 0 && (key.objectid != dirid ||
1745 key.type != BTRFS_INODE_REF_KEY)) {
1746 ret = -ENOENT;
1747 goto out;
1748 }
1749
1750 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1751 len = btrfs_inode_ref_name_len(l, iref);
1752 ptr -= len + 1;
1753 total_len += len + 1;
1754 if (ptr < name)
1755 goto out;
1756
1757 *(ptr + len) = '/';
1758 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1759
1760 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1761 break;
1762
1763 btrfs_release_path(path);
1764 key.objectid = key.offset;
1765 key.offset = (u64)-1;
1766 dirid = key.objectid;
1767 }
1768 if (ptr < name)
1769 goto out;
1770 memmove(name, ptr, total_len);
1771 name[total_len]='\0';
1772 ret = 0;
1773 out:
1774 btrfs_free_path(path);
1775 return ret;
1776 }
1777
1778 static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1779 void __user *argp)
1780 {
1781 struct btrfs_ioctl_ino_lookup_args *args;
1782 struct inode *inode;
1783 int ret;
1784
1785 if (!capable(CAP_SYS_ADMIN))
1786 return -EPERM;
1787
1788 args = memdup_user(argp, sizeof(*args));
1789 if (IS_ERR(args))
1790 return PTR_ERR(args);
1791
1792 inode = fdentry(file)->d_inode;
1793
1794 if (args->treeid == 0)
1795 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1796
1797 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1798 args->treeid, args->objectid,
1799 args->name);
1800
1801 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1802 ret = -EFAULT;
1803
1804 kfree(args);
1805 return ret;
1806 }
1807
1808 static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1809 void __user *arg)
1810 {
1811 struct dentry *parent = fdentry(file);
1812 struct dentry *dentry;
1813 struct inode *dir = parent->d_inode;
1814 struct inode *inode;
1815 struct btrfs_root *root = BTRFS_I(dir)->root;
1816 struct btrfs_root *dest = NULL;
1817 struct btrfs_ioctl_vol_args *vol_args;
1818 struct btrfs_trans_handle *trans;
1819 int namelen;
1820 int ret;
1821 int err = 0;
1822
1823 vol_args = memdup_user(arg, sizeof(*vol_args));
1824 if (IS_ERR(vol_args))
1825 return PTR_ERR(vol_args);
1826
1827 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1828 namelen = strlen(vol_args->name);
1829 if (strchr(vol_args->name, '/') ||
1830 strncmp(vol_args->name, "..", namelen) == 0) {
1831 err = -EINVAL;
1832 goto out;
1833 }
1834
1835 err = mnt_want_write(file->f_path.mnt);
1836 if (err)
1837 goto out;
1838
1839 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
1840 dentry = lookup_one_len(vol_args->name, parent, namelen);
1841 if (IS_ERR(dentry)) {
1842 err = PTR_ERR(dentry);
1843 goto out_unlock_dir;
1844 }
1845
1846 if (!dentry->d_inode) {
1847 err = -ENOENT;
1848 goto out_dput;
1849 }
1850
1851 inode = dentry->d_inode;
1852 dest = BTRFS_I(inode)->root;
1853 if (!capable(CAP_SYS_ADMIN)){
1854 /*
1855 * Regular user. Only allow this with a special mount
1856 * option, when the user has write+exec access to the
1857 * subvol root, and when rmdir(2) would have been
1858 * allowed.
1859 *
1860 * Note that this is _not_ check that the subvol is
1861 * empty or doesn't contain data that we wouldn't
1862 * otherwise be able to delete.
1863 *
1864 * Users who want to delete empty subvols should try
1865 * rmdir(2).
1866 */
1867 err = -EPERM;
1868 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1869 goto out_dput;
1870
1871 /*
1872 * Do not allow deletion if the parent dir is the same
1873 * as the dir to be deleted. That means the ioctl
1874 * must be called on the dentry referencing the root
1875 * of the subvol, not a random directory contained
1876 * within it.
1877 */
1878 err = -EINVAL;
1879 if (root == dest)
1880 goto out_dput;
1881
1882 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
1883 if (err)
1884 goto out_dput;
1885
1886 /* check if subvolume may be deleted by a non-root user */
1887 err = btrfs_may_delete(dir, dentry, 1);
1888 if (err)
1889 goto out_dput;
1890 }
1891
1892 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1893 err = -EINVAL;
1894 goto out_dput;
1895 }
1896
1897 mutex_lock(&inode->i_mutex);
1898 err = d_invalidate(dentry);
1899 if (err)
1900 goto out_unlock;
1901
1902 down_write(&root->fs_info->subvol_sem);
1903
1904 err = may_destroy_subvol(dest);
1905 if (err)
1906 goto out_up_write;
1907
1908 trans = btrfs_start_transaction(root, 0);
1909 if (IS_ERR(trans)) {
1910 err = PTR_ERR(trans);
1911 goto out_up_write;
1912 }
1913 trans->block_rsv = &root->fs_info->global_block_rsv;
1914
1915 ret = btrfs_unlink_subvol(trans, root, dir,
1916 dest->root_key.objectid,
1917 dentry->d_name.name,
1918 dentry->d_name.len);
1919 BUG_ON(ret);
1920
1921 btrfs_record_root_in_trans(trans, dest);
1922
1923 memset(&dest->root_item.drop_progress, 0,
1924 sizeof(dest->root_item.drop_progress));
1925 dest->root_item.drop_level = 0;
1926 btrfs_set_root_refs(&dest->root_item, 0);
1927
1928 if (!xchg(&dest->orphan_item_inserted, 1)) {
1929 ret = btrfs_insert_orphan_item(trans,
1930 root->fs_info->tree_root,
1931 dest->root_key.objectid);
1932 BUG_ON(ret);
1933 }
1934
1935 ret = btrfs_end_transaction(trans, root);
1936 BUG_ON(ret);
1937 inode->i_flags |= S_DEAD;
1938 out_up_write:
1939 up_write(&root->fs_info->subvol_sem);
1940 out_unlock:
1941 mutex_unlock(&inode->i_mutex);
1942 if (!err) {
1943 shrink_dcache_sb(root->fs_info->sb);
1944 btrfs_invalidate_inodes(dest);
1945 d_delete(dentry);
1946 }
1947 out_dput:
1948 dput(dentry);
1949 out_unlock_dir:
1950 mutex_unlock(&dir->i_mutex);
1951 mnt_drop_write(file->f_path.mnt);
1952 out:
1953 kfree(vol_args);
1954 return err;
1955 }
1956
1957 static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
1958 {
1959 struct inode *inode = fdentry(file)->d_inode;
1960 struct btrfs_root *root = BTRFS_I(inode)->root;
1961 struct btrfs_ioctl_defrag_range_args *range;
1962 int ret;
1963
1964 if (btrfs_root_readonly(root))
1965 return -EROFS;
1966
1967 ret = mnt_want_write(file->f_path.mnt);
1968 if (ret)
1969 return ret;
1970
1971 switch (inode->i_mode & S_IFMT) {
1972 case S_IFDIR:
1973 if (!capable(CAP_SYS_ADMIN)) {
1974 ret = -EPERM;
1975 goto out;
1976 }
1977 ret = btrfs_defrag_root(root, 0);
1978 if (ret)
1979 goto out;
1980 ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
1981 break;
1982 case S_IFREG:
1983 if (!(file->f_mode & FMODE_WRITE)) {
1984 ret = -EINVAL;
1985 goto out;
1986 }
1987
1988 range = kzalloc(sizeof(*range), GFP_KERNEL);
1989 if (!range) {
1990 ret = -ENOMEM;
1991 goto out;
1992 }
1993
1994 if (argp) {
1995 if (copy_from_user(range, argp,
1996 sizeof(*range))) {
1997 ret = -EFAULT;
1998 kfree(range);
1999 goto out;
2000 }
2001 /* compression requires us to start the IO */
2002 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2003 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2004 range->extent_thresh = (u32)-1;
2005 }
2006 } else {
2007 /* the rest are all set to zero by kzalloc */
2008 range->len = (u64)-1;
2009 }
2010 ret = btrfs_defrag_file(fdentry(file)->d_inode, file,
2011 range, 0, 0);
2012 if (ret > 0)
2013 ret = 0;
2014 kfree(range);
2015 break;
2016 default:
2017 ret = -EINVAL;
2018 }
2019 out:
2020 mnt_drop_write(file->f_path.mnt);
2021 return ret;
2022 }
2023
2024 static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
2025 {
2026 struct btrfs_ioctl_vol_args *vol_args;
2027 int ret;
2028
2029 if (!capable(CAP_SYS_ADMIN))
2030 return -EPERM;
2031
2032 vol_args = memdup_user(arg, sizeof(*vol_args));
2033 if (IS_ERR(vol_args))
2034 return PTR_ERR(vol_args);
2035
2036 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2037 ret = btrfs_init_new_device(root, vol_args->name);
2038
2039 kfree(vol_args);
2040 return ret;
2041 }
2042
2043 static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
2044 {
2045 struct btrfs_ioctl_vol_args *vol_args;
2046 int ret;
2047
2048 if (!capable(CAP_SYS_ADMIN))
2049 return -EPERM;
2050
2051 if (root->fs_info->sb->s_flags & MS_RDONLY)
2052 return -EROFS;
2053
2054 vol_args = memdup_user(arg, sizeof(*vol_args));
2055 if (IS_ERR(vol_args))
2056 return PTR_ERR(vol_args);
2057
2058 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2059 ret = btrfs_rm_device(root, vol_args->name);
2060
2061 kfree(vol_args);
2062 return ret;
2063 }
2064
2065 static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2066 {
2067 struct btrfs_ioctl_fs_info_args *fi_args;
2068 struct btrfs_device *device;
2069 struct btrfs_device *next;
2070 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2071 int ret = 0;
2072
2073 if (!capable(CAP_SYS_ADMIN))
2074 return -EPERM;
2075
2076 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2077 if (!fi_args)
2078 return -ENOMEM;
2079
2080 fi_args->num_devices = fs_devices->num_devices;
2081 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
2082
2083 mutex_lock(&fs_devices->device_list_mutex);
2084 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2085 if (device->devid > fi_args->max_id)
2086 fi_args->max_id = device->devid;
2087 }
2088 mutex_unlock(&fs_devices->device_list_mutex);
2089
2090 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2091 ret = -EFAULT;
2092
2093 kfree(fi_args);
2094 return ret;
2095 }
2096
2097 static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2098 {
2099 struct btrfs_ioctl_dev_info_args *di_args;
2100 struct btrfs_device *dev;
2101 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2102 int ret = 0;
2103 char *s_uuid = NULL;
2104 char empty_uuid[BTRFS_UUID_SIZE] = {0};
2105
2106 if (!capable(CAP_SYS_ADMIN))
2107 return -EPERM;
2108
2109 di_args = memdup_user(arg, sizeof(*di_args));
2110 if (IS_ERR(di_args))
2111 return PTR_ERR(di_args);
2112
2113 if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2114 s_uuid = di_args->uuid;
2115
2116 mutex_lock(&fs_devices->device_list_mutex);
2117 dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL);
2118 mutex_unlock(&fs_devices->device_list_mutex);
2119
2120 if (!dev) {
2121 ret = -ENODEV;
2122 goto out;
2123 }
2124
2125 di_args->devid = dev->devid;
2126 di_args->bytes_used = dev->bytes_used;
2127 di_args->total_bytes = dev->total_bytes;
2128 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2129 strncpy(di_args->path, dev->name, sizeof(di_args->path));
2130
2131 out:
2132 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2133 ret = -EFAULT;
2134
2135 kfree(di_args);
2136 return ret;
2137 }
2138
2139 static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2140 u64 off, u64 olen, u64 destoff)
2141 {
2142 struct inode *inode = fdentry(file)->d_inode;
2143 struct btrfs_root *root = BTRFS_I(inode)->root;
2144 struct file *src_file;
2145 struct inode *src;
2146 struct btrfs_trans_handle *trans;
2147 struct btrfs_path *path;
2148 struct extent_buffer *leaf;
2149 char *buf;
2150 struct btrfs_key key;
2151 u32 nritems;
2152 int slot;
2153 int ret;
2154 u64 len = olen;
2155 u64 bs = root->fs_info->sb->s_blocksize;
2156 u64 hint_byte;
2157
2158 /*
2159 * TODO:
2160 * - split compressed inline extents. annoying: we need to
2161 * decompress into destination's address_space (the file offset
2162 * may change, so source mapping won't do), then recompress (or
2163 * otherwise reinsert) a subrange.
2164 * - allow ranges within the same file to be cloned (provided
2165 * they don't overlap)?
2166 */
2167
2168 /* the destination must be opened for writing */
2169 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
2170 return -EINVAL;
2171
2172 if (btrfs_root_readonly(root))
2173 return -EROFS;
2174
2175 ret = mnt_want_write(file->f_path.mnt);
2176 if (ret)
2177 return ret;
2178
2179 src_file = fget(srcfd);
2180 if (!src_file) {
2181 ret = -EBADF;
2182 goto out_drop_write;
2183 }
2184
2185 src = src_file->f_dentry->d_inode;
2186
2187 ret = -EINVAL;
2188 if (src == inode)
2189 goto out_fput;
2190
2191 /* the src must be open for reading */
2192 if (!(src_file->f_mode & FMODE_READ))
2193 goto out_fput;
2194
2195 /* don't make the dst file partly checksummed */
2196 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2197 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2198 goto out_fput;
2199
2200 ret = -EISDIR;
2201 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2202 goto out_fput;
2203
2204 ret = -EXDEV;
2205 if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
2206 goto out_fput;
2207
2208 ret = -ENOMEM;
2209 buf = vmalloc(btrfs_level_size(root, 0));
2210 if (!buf)
2211 goto out_fput;
2212
2213 path = btrfs_alloc_path();
2214 if (!path) {
2215 vfree(buf);
2216 goto out_fput;
2217 }
2218 path->reada = 2;
2219
2220 if (inode < src) {
2221 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2222 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
2223 } else {
2224 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2225 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
2226 }
2227
2228 /* determine range to clone */
2229 ret = -EINVAL;
2230 if (off + len > src->i_size || off + len < off)
2231 goto out_unlock;
2232 if (len == 0)
2233 olen = len = src->i_size - off;
2234 /* if we extend to eof, continue to block boundary */
2235 if (off + len == src->i_size)
2236 len = ALIGN(src->i_size, bs) - off;
2237
2238 /* verify the end result is block aligned */
2239 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2240 !IS_ALIGNED(destoff, bs))
2241 goto out_unlock;
2242
2243 if (destoff > inode->i_size) {
2244 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2245 if (ret)
2246 goto out_unlock;
2247 }
2248
2249 /* truncate page cache pages from target inode range */
2250 truncate_inode_pages_range(&inode->i_data, destoff,
2251 PAGE_CACHE_ALIGN(destoff + len) - 1);
2252
2253 /* do any pending delalloc/csum calc on src, one way or
2254 another, and lock file content */
2255 while (1) {
2256 struct btrfs_ordered_extent *ordered;
2257 lock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
2258 ordered = btrfs_lookup_first_ordered_extent(src, off+len);
2259 if (!ordered &&
2260 !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len,
2261 EXTENT_DELALLOC, 0, NULL))
2262 break;
2263 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
2264 if (ordered)
2265 btrfs_put_ordered_extent(ordered);
2266 btrfs_wait_ordered_range(src, off, len);
2267 }
2268
2269 /* clone data */
2270 key.objectid = btrfs_ino(src);
2271 key.type = BTRFS_EXTENT_DATA_KEY;
2272 key.offset = 0;
2273
2274 while (1) {
2275 /*
2276 * note the key will change type as we walk through the
2277 * tree.
2278 */
2279 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2280 if (ret < 0)
2281 goto out;
2282
2283 nritems = btrfs_header_nritems(path->nodes[0]);
2284 if (path->slots[0] >= nritems) {
2285 ret = btrfs_next_leaf(root, path);
2286 if (ret < 0)
2287 goto out;
2288 if (ret > 0)
2289 break;
2290 nritems = btrfs_header_nritems(path->nodes[0]);
2291 }
2292 leaf = path->nodes[0];
2293 slot = path->slots[0];
2294
2295 btrfs_item_key_to_cpu(leaf, &key, slot);
2296 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
2297 key.objectid != btrfs_ino(src))
2298 break;
2299
2300 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2301 struct btrfs_file_extent_item *extent;
2302 int type;
2303 u32 size;
2304 struct btrfs_key new_key;
2305 u64 disko = 0, diskl = 0;
2306 u64 datao = 0, datal = 0;
2307 u8 comp;
2308 u64 endoff;
2309
2310 size = btrfs_item_size_nr(leaf, slot);
2311 read_extent_buffer(leaf, buf,
2312 btrfs_item_ptr_offset(leaf, slot),
2313 size);
2314
2315 extent = btrfs_item_ptr(leaf, slot,
2316 struct btrfs_file_extent_item);
2317 comp = btrfs_file_extent_compression(leaf, extent);
2318 type = btrfs_file_extent_type(leaf, extent);
2319 if (type == BTRFS_FILE_EXTENT_REG ||
2320 type == BTRFS_FILE_EXTENT_PREALLOC) {
2321 disko = btrfs_file_extent_disk_bytenr(leaf,
2322 extent);
2323 diskl = btrfs_file_extent_disk_num_bytes(leaf,
2324 extent);
2325 datao = btrfs_file_extent_offset(leaf, extent);
2326 datal = btrfs_file_extent_num_bytes(leaf,
2327 extent);
2328 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2329 /* take upper bound, may be compressed */
2330 datal = btrfs_file_extent_ram_bytes(leaf,
2331 extent);
2332 }
2333 btrfs_release_path(path);
2334
2335 if (key.offset + datal <= off ||
2336 key.offset >= off+len)
2337 goto next;
2338
2339 memcpy(&new_key, &key, sizeof(new_key));
2340 new_key.objectid = btrfs_ino(inode);
2341 if (off <= key.offset)
2342 new_key.offset = key.offset + destoff - off;
2343 else
2344 new_key.offset = destoff;
2345
2346 /*
2347 * 1 - adjusting old extent (we may have to split it)
2348 * 1 - add new extent
2349 * 1 - inode update
2350 */
2351 trans = btrfs_start_transaction(root, 3);
2352 if (IS_ERR(trans)) {
2353 ret = PTR_ERR(trans);
2354 goto out;
2355 }
2356
2357 if (type == BTRFS_FILE_EXTENT_REG ||
2358 type == BTRFS_FILE_EXTENT_PREALLOC) {
2359 /*
2360 * a | --- range to clone ---| b
2361 * | ------------- extent ------------- |
2362 */
2363
2364 /* substract range b */
2365 if (key.offset + datal > off + len)
2366 datal = off + len - key.offset;
2367
2368 /* substract range a */
2369 if (off > key.offset) {
2370 datao += off - key.offset;
2371 datal -= off - key.offset;
2372 }
2373
2374 ret = btrfs_drop_extents(trans, inode,
2375 new_key.offset,
2376 new_key.offset + datal,
2377 &hint_byte, 1);
2378 BUG_ON(ret);
2379
2380 ret = btrfs_insert_empty_item(trans, root, path,
2381 &new_key, size);
2382 BUG_ON(ret);
2383
2384 leaf = path->nodes[0];
2385 slot = path->slots[0];
2386 write_extent_buffer(leaf, buf,
2387 btrfs_item_ptr_offset(leaf, slot),
2388 size);
2389
2390 extent = btrfs_item_ptr(leaf, slot,
2391 struct btrfs_file_extent_item);
2392
2393 /* disko == 0 means it's a hole */
2394 if (!disko)
2395 datao = 0;
2396
2397 btrfs_set_file_extent_offset(leaf, extent,
2398 datao);
2399 btrfs_set_file_extent_num_bytes(leaf, extent,
2400 datal);
2401 if (disko) {
2402 inode_add_bytes(inode, datal);
2403 ret = btrfs_inc_extent_ref(trans, root,
2404 disko, diskl, 0,
2405 root->root_key.objectid,
2406 btrfs_ino(inode),
2407 new_key.offset - datao);
2408 BUG_ON(ret);
2409 }
2410 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2411 u64 skip = 0;
2412 u64 trim = 0;
2413 if (off > key.offset) {
2414 skip = off - key.offset;
2415 new_key.offset += skip;
2416 }
2417
2418 if (key.offset + datal > off+len)
2419 trim = key.offset + datal - (off+len);
2420
2421 if (comp && (skip || trim)) {
2422 ret = -EINVAL;
2423 btrfs_end_transaction(trans, root);
2424 goto out;
2425 }
2426 size -= skip + trim;
2427 datal -= skip + trim;
2428
2429 ret = btrfs_drop_extents(trans, inode,
2430 new_key.offset,
2431 new_key.offset + datal,
2432 &hint_byte, 1);
2433 BUG_ON(ret);
2434
2435 ret = btrfs_insert_empty_item(trans, root, path,
2436 &new_key, size);
2437 BUG_ON(ret);
2438
2439 if (skip) {
2440 u32 start =
2441 btrfs_file_extent_calc_inline_size(0);
2442 memmove(buf+start, buf+start+skip,
2443 datal);
2444 }
2445
2446 leaf = path->nodes[0];
2447 slot = path->slots[0];
2448 write_extent_buffer(leaf, buf,
2449 btrfs_item_ptr_offset(leaf, slot),
2450 size);
2451 inode_add_bytes(inode, datal);
2452 }
2453
2454 btrfs_mark_buffer_dirty(leaf);
2455 btrfs_release_path(path);
2456
2457 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
2458
2459 /*
2460 * we round up to the block size at eof when
2461 * determining which extents to clone above,
2462 * but shouldn't round up the file size
2463 */
2464 endoff = new_key.offset + datal;
2465 if (endoff > destoff+olen)
2466 endoff = destoff+olen;
2467 if (endoff > inode->i_size)
2468 btrfs_i_size_write(inode, endoff);
2469
2470 ret = btrfs_update_inode(trans, root, inode);
2471 BUG_ON(ret);
2472 btrfs_end_transaction(trans, root);
2473 }
2474 next:
2475 btrfs_release_path(path);
2476 key.offset++;
2477 }
2478 ret = 0;
2479 out:
2480 btrfs_release_path(path);
2481 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
2482 out_unlock:
2483 mutex_unlock(&src->i_mutex);
2484 mutex_unlock(&inode->i_mutex);
2485 vfree(buf);
2486 btrfs_free_path(path);
2487 out_fput:
2488 fput(src_file);
2489 out_drop_write:
2490 mnt_drop_write(file->f_path.mnt);
2491 return ret;
2492 }
2493
2494 static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
2495 {
2496 struct btrfs_ioctl_clone_range_args args;
2497
2498 if (copy_from_user(&args, argp, sizeof(args)))
2499 return -EFAULT;
2500 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2501 args.src_length, args.dest_offset);
2502 }
2503
2504 /*
2505 * there are many ways the trans_start and trans_end ioctls can lead
2506 * to deadlocks. They should only be used by applications that
2507 * basically own the machine, and have a very in depth understanding
2508 * of all the possible deadlocks and enospc problems.
2509 */
2510 static long btrfs_ioctl_trans_start(struct file *file)
2511 {
2512 struct inode *inode = fdentry(file)->d_inode;
2513 struct btrfs_root *root = BTRFS_I(inode)->root;
2514 struct btrfs_trans_handle *trans;
2515 int ret;
2516
2517 ret = -EPERM;
2518 if (!capable(CAP_SYS_ADMIN))
2519 goto out;
2520
2521 ret = -EINPROGRESS;
2522 if (file->private_data)
2523 goto out;
2524
2525 ret = -EROFS;
2526 if (btrfs_root_readonly(root))
2527 goto out;
2528
2529 ret = mnt_want_write(file->f_path.mnt);
2530 if (ret)
2531 goto out;
2532
2533 atomic_inc(&root->fs_info->open_ioctl_trans);
2534
2535 ret = -ENOMEM;
2536 trans = btrfs_start_ioctl_transaction(root);
2537 if (IS_ERR(trans))
2538 goto out_drop;
2539
2540 file->private_data = trans;
2541 return 0;
2542
2543 out_drop:
2544 atomic_dec(&root->fs_info->open_ioctl_trans);
2545 mnt_drop_write(file->f_path.mnt);
2546 out:
2547 return ret;
2548 }
2549
2550 static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2551 {
2552 struct inode *inode = fdentry(file)->d_inode;
2553 struct btrfs_root *root = BTRFS_I(inode)->root;
2554 struct btrfs_root *new_root;
2555 struct btrfs_dir_item *di;
2556 struct btrfs_trans_handle *trans;
2557 struct btrfs_path *path;
2558 struct btrfs_key location;
2559 struct btrfs_disk_key disk_key;
2560 struct btrfs_super_block *disk_super;
2561 u64 features;
2562 u64 objectid = 0;
2563 u64 dir_id;
2564
2565 if (!capable(CAP_SYS_ADMIN))
2566 return -EPERM;
2567
2568 if (copy_from_user(&objectid, argp, sizeof(objectid)))
2569 return -EFAULT;
2570
2571 if (!objectid)
2572 objectid = root->root_key.objectid;
2573
2574 location.objectid = objectid;
2575 location.type = BTRFS_ROOT_ITEM_KEY;
2576 location.offset = (u64)-1;
2577
2578 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2579 if (IS_ERR(new_root))
2580 return PTR_ERR(new_root);
2581
2582 if (btrfs_root_refs(&new_root->root_item) == 0)
2583 return -ENOENT;
2584
2585 path = btrfs_alloc_path();
2586 if (!path)
2587 return -ENOMEM;
2588 path->leave_spinning = 1;
2589
2590 trans = btrfs_start_transaction(root, 1);
2591 if (IS_ERR(trans)) {
2592 btrfs_free_path(path);
2593 return PTR_ERR(trans);
2594 }
2595
2596 dir_id = btrfs_super_root_dir(&root->fs_info->super_copy);
2597 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2598 dir_id, "default", 7, 1);
2599 if (IS_ERR_OR_NULL(di)) {
2600 btrfs_free_path(path);
2601 btrfs_end_transaction(trans, root);
2602 printk(KERN_ERR "Umm, you don't have the default dir item, "
2603 "this isn't going to work\n");
2604 return -ENOENT;
2605 }
2606
2607 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2608 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2609 btrfs_mark_buffer_dirty(path->nodes[0]);
2610 btrfs_free_path(path);
2611
2612 disk_super = &root->fs_info->super_copy;
2613 features = btrfs_super_incompat_flags(disk_super);
2614 if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) {
2615 features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL;
2616 btrfs_set_super_incompat_flags(disk_super, features);
2617 }
2618 btrfs_end_transaction(trans, root);
2619
2620 return 0;
2621 }
2622
2623 static void get_block_group_info(struct list_head *groups_list,
2624 struct btrfs_ioctl_space_info *space)
2625 {
2626 struct btrfs_block_group_cache *block_group;
2627
2628 space->total_bytes = 0;
2629 space->used_bytes = 0;
2630 space->flags = 0;
2631 list_for_each_entry(block_group, groups_list, list) {
2632 space->flags = block_group->flags;
2633 space->total_bytes += block_group->key.offset;
2634 space->used_bytes +=
2635 btrfs_block_group_used(&block_group->item);
2636 }
2637 }
2638
2639 long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2640 {
2641 struct btrfs_ioctl_space_args space_args;
2642 struct btrfs_ioctl_space_info space;
2643 struct btrfs_ioctl_space_info *dest;
2644 struct btrfs_ioctl_space_info *dest_orig;
2645 struct btrfs_ioctl_space_info __user *user_dest;
2646 struct btrfs_space_info *info;
2647 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2648 BTRFS_BLOCK_GROUP_SYSTEM,
2649 BTRFS_BLOCK_GROUP_METADATA,
2650 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2651 int num_types = 4;
2652 int alloc_size;
2653 int ret = 0;
2654 u64 slot_count = 0;
2655 int i, c;
2656
2657 if (copy_from_user(&space_args,
2658 (struct btrfs_ioctl_space_args __user *)arg,
2659 sizeof(space_args)))
2660 return -EFAULT;
2661
2662 for (i = 0; i < num_types; i++) {
2663 struct btrfs_space_info *tmp;
2664
2665 info = NULL;
2666 rcu_read_lock();
2667 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2668 list) {
2669 if (tmp->flags == types[i]) {
2670 info = tmp;
2671 break;
2672 }
2673 }
2674 rcu_read_unlock();
2675
2676 if (!info)
2677 continue;
2678
2679 down_read(&info->groups_sem);
2680 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2681 if (!list_empty(&info->block_groups[c]))
2682 slot_count++;
2683 }
2684 up_read(&info->groups_sem);
2685 }
2686
2687 /* space_slots == 0 means they are asking for a count */
2688 if (space_args.space_slots == 0) {
2689 space_args.total_spaces = slot_count;
2690 goto out;
2691 }
2692
2693 slot_count = min_t(u64, space_args.space_slots, slot_count);
2694
2695 alloc_size = sizeof(*dest) * slot_count;
2696
2697 /* we generally have at most 6 or so space infos, one for each raid
2698 * level. So, a whole page should be more than enough for everyone
2699 */
2700 if (alloc_size > PAGE_CACHE_SIZE)
2701 return -ENOMEM;
2702
2703 space_args.total_spaces = 0;
2704 dest = kmalloc(alloc_size, GFP_NOFS);
2705 if (!dest)
2706 return -ENOMEM;
2707 dest_orig = dest;
2708
2709 /* now we have a buffer to copy into */
2710 for (i = 0; i < num_types; i++) {
2711 struct btrfs_space_info *tmp;
2712
2713 if (!slot_count)
2714 break;
2715
2716 info = NULL;
2717 rcu_read_lock();
2718 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2719 list) {
2720 if (tmp->flags == types[i]) {
2721 info = tmp;
2722 break;
2723 }
2724 }
2725 rcu_read_unlock();
2726
2727 if (!info)
2728 continue;
2729 down_read(&info->groups_sem);
2730 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2731 if (!list_empty(&info->block_groups[c])) {
2732 get_block_group_info(&info->block_groups[c],
2733 &space);
2734 memcpy(dest, &space, sizeof(space));
2735 dest++;
2736 space_args.total_spaces++;
2737 slot_count--;
2738 }
2739 if (!slot_count)
2740 break;
2741 }
2742 up_read(&info->groups_sem);
2743 }
2744
2745 user_dest = (struct btrfs_ioctl_space_info *)
2746 (arg + sizeof(struct btrfs_ioctl_space_args));
2747
2748 if (copy_to_user(user_dest, dest_orig, alloc_size))
2749 ret = -EFAULT;
2750
2751 kfree(dest_orig);
2752 out:
2753 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
2754 ret = -EFAULT;
2755
2756 return ret;
2757 }
2758
2759 /*
2760 * there are many ways the trans_start and trans_end ioctls can lead
2761 * to deadlocks. They should only be used by applications that
2762 * basically own the machine, and have a very in depth understanding
2763 * of all the possible deadlocks and enospc problems.
2764 */
2765 long btrfs_ioctl_trans_end(struct file *file)
2766 {
2767 struct inode *inode = fdentry(file)->d_inode;
2768 struct btrfs_root *root = BTRFS_I(inode)->root;
2769 struct btrfs_trans_handle *trans;
2770
2771 trans = file->private_data;
2772 if (!trans)
2773 return -EINVAL;
2774 file->private_data = NULL;
2775
2776 btrfs_end_transaction(trans, root);
2777
2778 atomic_dec(&root->fs_info->open_ioctl_trans);
2779
2780 mnt_drop_write(file->f_path.mnt);
2781 return 0;
2782 }
2783
2784 static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
2785 {
2786 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2787 struct btrfs_trans_handle *trans;
2788 u64 transid;
2789 int ret;
2790
2791 trans = btrfs_start_transaction(root, 0);
2792 if (IS_ERR(trans))
2793 return PTR_ERR(trans);
2794 transid = trans->transid;
2795 ret = btrfs_commit_transaction_async(trans, root, 0);
2796 if (ret) {
2797 btrfs_end_transaction(trans, root);
2798 return ret;
2799 }
2800
2801 if (argp)
2802 if (copy_to_user(argp, &transid, sizeof(transid)))
2803 return -EFAULT;
2804 return 0;
2805 }
2806
2807 static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
2808 {
2809 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2810 u64 transid;
2811
2812 if (argp) {
2813 if (copy_from_user(&transid, argp, sizeof(transid)))
2814 return -EFAULT;
2815 } else {
2816 transid = 0; /* current trans */
2817 }
2818 return btrfs_wait_for_commit(root, transid);
2819 }
2820
2821 static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg)
2822 {
2823 int ret;
2824 struct btrfs_ioctl_scrub_args *sa;
2825
2826 if (!capable(CAP_SYS_ADMIN))
2827 return -EPERM;
2828
2829 sa = memdup_user(arg, sizeof(*sa));
2830 if (IS_ERR(sa))
2831 return PTR_ERR(sa);
2832
2833 ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end,
2834 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY);
2835
2836 if (copy_to_user(arg, sa, sizeof(*sa)))
2837 ret = -EFAULT;
2838
2839 kfree(sa);
2840 return ret;
2841 }
2842
2843 static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
2844 {
2845 if (!capable(CAP_SYS_ADMIN))
2846 return -EPERM;
2847
2848 return btrfs_scrub_cancel(root);
2849 }
2850
2851 static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
2852 void __user *arg)
2853 {
2854 struct btrfs_ioctl_scrub_args *sa;
2855 int ret;
2856
2857 if (!capable(CAP_SYS_ADMIN))
2858 return -EPERM;
2859
2860 sa = memdup_user(arg, sizeof(*sa));
2861 if (IS_ERR(sa))
2862 return PTR_ERR(sa);
2863
2864 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
2865
2866 if (copy_to_user(arg, sa, sizeof(*sa)))
2867 ret = -EFAULT;
2868
2869 kfree(sa);
2870 return ret;
2871 }
2872
2873 long btrfs_ioctl(struct file *file, unsigned int
2874 cmd, unsigned long arg)
2875 {
2876 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
2877 void __user *argp = (void __user *)arg;
2878
2879 switch (cmd) {
2880 case FS_IOC_GETFLAGS:
2881 return btrfs_ioctl_getflags(file, argp);
2882 case FS_IOC_SETFLAGS:
2883 return btrfs_ioctl_setflags(file, argp);
2884 case FS_IOC_GETVERSION:
2885 return btrfs_ioctl_getversion(file, argp);
2886 case FITRIM:
2887 return btrfs_ioctl_fitrim(file, argp);
2888 case BTRFS_IOC_SNAP_CREATE:
2889 return btrfs_ioctl_snap_create(file, argp, 0);
2890 case BTRFS_IOC_SNAP_CREATE_V2:
2891 return btrfs_ioctl_snap_create_v2(file, argp, 0);
2892 case BTRFS_IOC_SUBVOL_CREATE:
2893 return btrfs_ioctl_snap_create(file, argp, 1);
2894 case BTRFS_IOC_SNAP_DESTROY:
2895 return btrfs_ioctl_snap_destroy(file, argp);
2896 case BTRFS_IOC_SUBVOL_GETFLAGS:
2897 return btrfs_ioctl_subvol_getflags(file, argp);
2898 case BTRFS_IOC_SUBVOL_SETFLAGS:
2899 return btrfs_ioctl_subvol_setflags(file, argp);
2900 case BTRFS_IOC_DEFAULT_SUBVOL:
2901 return btrfs_ioctl_default_subvol(file, argp);
2902 case BTRFS_IOC_DEFRAG:
2903 return btrfs_ioctl_defrag(file, NULL);
2904 case BTRFS_IOC_DEFRAG_RANGE:
2905 return btrfs_ioctl_defrag(file, argp);
2906 case BTRFS_IOC_RESIZE:
2907 return btrfs_ioctl_resize(root, argp);
2908 case BTRFS_IOC_ADD_DEV:
2909 return btrfs_ioctl_add_dev(root, argp);
2910 case BTRFS_IOC_RM_DEV:
2911 return btrfs_ioctl_rm_dev(root, argp);
2912 case BTRFS_IOC_FS_INFO:
2913 return btrfs_ioctl_fs_info(root, argp);
2914 case BTRFS_IOC_DEV_INFO:
2915 return btrfs_ioctl_dev_info(root, argp);
2916 case BTRFS_IOC_BALANCE:
2917 return btrfs_balance(root->fs_info->dev_root);
2918 case BTRFS_IOC_CLONE:
2919 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
2920 case BTRFS_IOC_CLONE_RANGE:
2921 return btrfs_ioctl_clone_range(file, argp);
2922 case BTRFS_IOC_TRANS_START:
2923 return btrfs_ioctl_trans_start(file);
2924 case BTRFS_IOC_TRANS_END:
2925 return btrfs_ioctl_trans_end(file);
2926 case BTRFS_IOC_TREE_SEARCH:
2927 return btrfs_ioctl_tree_search(file, argp);
2928 case BTRFS_IOC_INO_LOOKUP:
2929 return btrfs_ioctl_ino_lookup(file, argp);
2930 case BTRFS_IOC_SPACE_INFO:
2931 return btrfs_ioctl_space_info(root, argp);
2932 case BTRFS_IOC_SYNC:
2933 btrfs_sync_fs(file->f_dentry->d_sb, 1);
2934 return 0;
2935 case BTRFS_IOC_START_SYNC:
2936 return btrfs_ioctl_start_sync(file, argp);
2937 case BTRFS_IOC_WAIT_SYNC:
2938 return btrfs_ioctl_wait_sync(file, argp);
2939 case BTRFS_IOC_SCRUB:
2940 return btrfs_ioctl_scrub(root, argp);
2941 case BTRFS_IOC_SCRUB_CANCEL:
2942 return btrfs_ioctl_scrub_cancel(root, argp);
2943 case BTRFS_IOC_SCRUB_PROGRESS:
2944 return btrfs_ioctl_scrub_progress(root, argp);
2945 }
2946
2947 return -ENOTTY;
2948 }
Linux 2.6 ibm-acpi/thinkpad-acpi driver git tree