]> git.ipfire.org Git - people/ms/linux.git/blame - fs/btrfs/ioctl.c
projects / people / ms / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
btrfs: trivial fix, a potential memory leak in btrfs_parse_early_options()
[people/ms/linux.git] / fs / btrfs / ioctl.c
CommitLineData
f46b5a66
CH		 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>
cb8e7090 24#include <linux/fsnotify.h>
f46b5a66
CH		 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>
f46b5a66 30#include <linux/backing-dev.h>
cb8e7090 31#include <linux/mount.h>
f46b5a66 32#include <linux/mpage.h>
cb8e7090 33#include <linux/namei.h>
f46b5a66
CH		 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>
cb8e7090 39#include <linux/security.h>
f46b5a66 40#include <linux/xattr.h>
7ea394f1 41#include <linux/vmalloc.h>
5a0e3ad6 42#include <linux/slab.h>
f7039b1d 43#include <linux/blkdev.h>
4b4e25f2 44#include "compat.h"
f46b5a66
CH		 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"
925baedd 52#include "locking.h"
581bb050 53#include "inode-map.h"
f46b5a66 54
6cbff00f
CH		 55/* Mask out flags that are inappropriate for the given type of inode. */
56static 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 */
69static 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;
d0092bdd
LZ		 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;
6cbff00f
CH		 92
93 return iflags;
94}
95
96/*
97 * Update inode->i_flags based on the btrfs internal flags.
98 */
99void 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 *
e27425d6 120 * Currently only the compression flags and the cow flags are inherited.
6cbff00f
CH		 121 */
122void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
123{
0b4dcea5
CM		 124 unsigned int flags;
125
126 if (!dir)
127 return;
128
129 flags = BTRFS_I(dir)->flags;
6cbff00f 130
e27425d6
JB		 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;
6cbff00f 141
6cbff00f
CH		 142 btrfs_update_iflags(inode);
143}
144
145static 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
75e7cb7f
LB		 155static 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 | \
e1e8fb6a
LZ		 160 FS_NOCOMP_FL | FS_COMPR_FL |
161 FS_NOCOW_FL))
75e7cb7f
LB		 162 return -EOPNOTSUPP;
163
164 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
165 return -EINVAL;
166
75e7cb7f
LB		 167 return 0;
168}
169
6cbff00f
CH		 170static 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
b83cc969
LZ		 179 if (btrfs_root_readonly(root))
180 return -EROFS;
181
6cbff00f
CH		 182 if (copy_from_user(&flags, arg, sizeof(flags)))
183 return -EFAULT;
184
75e7cb7f
LB		 185 ret = check_flags(flags);
186 if (ret)
187 return ret;
f46b5a66 188
2e149670 189 if (!inode_owner_or_capable(inode))
6cbff00f
CH		 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;
e1e8fb6a
LZ		 231 if (flags & FS_NOCOW_FL)
232 ip->flags |= BTRFS_INODE_NODATACOW;
233 else
234 ip->flags &= ~BTRFS_INODE_NODATACOW;
6cbff00f 235
75e7cb7f
LB		 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;
ebcb904d
LZ		 247 } else {
248 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
75e7cb7f 249 }
6cbff00f 250
7a7eaa40 251 trans = btrfs_join_transaction(root);
3612b495 252 BUG_ON(IS_ERR(trans));
6cbff00f
CH		 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);
2d4e6f6a 262
263 ret = 0;
6cbff00f
CH		 264 out_unlock:
265 mutex_unlock(&inode->i_mutex);
2d4e6f6a 266 return ret;
6cbff00f
CH		 267}
268
269static 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}
f46b5a66 275
f7039b1d
LD		 276static 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
1f78160c
XG		 290 rcu_read_lock();
291 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
292 dev_list) {
f7039b1d
LD		 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 }
1f78160c 302 rcu_read_unlock();
f7039b1d
LD		 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
cb8e7090
CH		 320static noinline int create_subvol(struct btrfs_root *root,
321 struct dentry *dentry,
72fd032e
SW		 322 char *name, int namelen,
323 u64 *async_transid)
f46b5a66
CH		 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;
76dda93c 330 struct btrfs_root *new_root;
2fbe8c8a 331 struct dentry *parent = dentry->d_parent;
6a912213 332 struct inode *dir;
f46b5a66
CH		 333 int ret;
334 int err;
335 u64 objectid;
336 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
3de4586c 337 u64 index = 0;
f46b5a66 338
581bb050 339 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
2fbe8c8a 340 if (ret)
a22285a6 341 return ret;
6a912213
JB		 342
343 dir = parent->d_inode;
344
9ed74f2d
JB		 345 /*
346 * 1 - inode item
347 * 2 - refs
348 * 1 - root item
349 * 2 - dir items
350 */
a22285a6 351 trans = btrfs_start_transaction(root, 6);
2fbe8c8a 352 if (IS_ERR(trans))
a22285a6 353 return PTR_ERR(trans);
f46b5a66 354
5d4f98a2
YZ		 355 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
356 0, objectid, NULL, 0, 0, 0);
8e8a1e31
JB		 357 if (IS_ERR(leaf)) {
358 ret = PTR_ERR(leaf);
359 goto fail;
360 }
f46b5a66 361
5d4f98a2 362 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
f46b5a66
CH		 363 btrfs_set_header_bytenr(leaf, leaf->start);
364 btrfs_set_header_generation(leaf, trans->transid);
5d4f98a2 365 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
f46b5a66
CH		 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);
5d4f98a2
YZ		 371 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
372 (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
373 BTRFS_UUID_SIZE);
f46b5a66
CH		 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);
a76a3cd4 381 inode_item->nbytes = cpu_to_le64(root->leafsize);
f46b5a66
CH		 382 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
383
08fe4db1
LZ		 384 root_item.flags = 0;
385 root_item.byte_limit = 0;
386 inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
387
f46b5a66 388 btrfs_set_root_bytenr(&root_item, leaf->start);
84234f3a 389 btrfs_set_root_generation(&root_item, trans->transid);
f46b5a66
CH		 390 btrfs_set_root_level(&root_item, 0);
391 btrfs_set_root_refs(&root_item, 1);
86b9f2ec 392 btrfs_set_root_used(&root_item, leaf->len);
80ff3856 393 btrfs_set_root_last_snapshot(&root_item, 0);
f46b5a66
CH		 394
395 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
396 root_item.drop_level = 0;
397
925baedd 398 btrfs_tree_unlock(leaf);
f46b5a66
CH		 399 free_extent_buffer(leaf);
400 leaf = NULL;
401
402 btrfs_set_root_dirid(&root_item, new_dirid);
403
404 key.objectid = objectid;
5d4f98a2 405 key.offset = 0;
f46b5a66
CH		 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
76dda93c
YZ		 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
d82a6f1d 418 ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
f46b5a66
CH		 419 /*
420 * insert the directory item
421 */
3de4586c
CM		 422 ret = btrfs_set_inode_index(dir, &index);
423 BUG_ON(ret);
424
425 ret = btrfs_insert_dir_item(trans, root,
16cdcec7 426 name, namelen, dir, &key,
3de4586c 427 BTRFS_FT_DIR, index);
f46b5a66
CH		 428 if (ret)
429 goto fail;
0660b5af 430
52c26179
YZ		 431 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
432 ret = btrfs_update_inode(trans, root, dir);
433 BUG_ON(ret);
434
0660b5af 435 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
4df27c4d 436 objectid, root->root_key.objectid,
33345d01 437 btrfs_ino(dir), index, name, namelen);
0660b5af 438
76dda93c 439 BUG_ON(ret);
f46b5a66 440
76dda93c 441 d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
f46b5a66 442fail:
72fd032e
SW		 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 }
f46b5a66
CH		 449 if (err && !ret)
450 ret = err;
f46b5a66
CH		 451 return ret;
452}
453
72fd032e 454static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
b83cc969
LZ		 455 char *name, int namelen, u64 *async_transid,
456 bool readonly)
f46b5a66 457{
2e4bfab9 458 struct inode *inode;
f46b5a66
CH		 459 struct btrfs_pending_snapshot *pending_snapshot;
460 struct btrfs_trans_handle *trans;
2e4bfab9 461 int ret;
f46b5a66
CH		 462
463 if (!root->ref_cows)
464 return -EINVAL;
465
3de4586c 466 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
a22285a6
YZ		 467 if (!pending_snapshot)
468 return -ENOMEM;
469
470 btrfs_init_block_rsv(&pending_snapshot->block_rsv);
3de4586c 471 pending_snapshot->dentry = dentry;
f46b5a66 472 pending_snapshot->root = root;
b83cc969 473 pending_snapshot->readonly = readonly;
a22285a6
YZ		 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
8351583e 484 spin_lock(&root->fs_info->trans_lock);
f46b5a66
CH		 485 list_add(&pending_snapshot->list,
486 &trans->transaction->pending_snapshots);
8351583e 487 spin_unlock(&root->fs_info->trans_lock);
72fd032e
SW		 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 }
2e4bfab9 496 BUG_ON(ret);
a22285a6
YZ		 497
498 ret = pending_snapshot->error;
499 if (ret)
500 goto fail;
501
66b4ffd1
JB		 502 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
503 if (ret)
504 goto fail;
f46b5a66 505
2fbe8c8a 506 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
2e4bfab9
YZ		 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;
514fail:
a22285a6 515 kfree(pending_snapshot);
f46b5a66
CH		 516 return ret;
517}
518
4260f7c7
SW		 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*/
523static 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
556static 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
cb8e7090
CH		 589/* copy of may_create in fs/namei.c() */
590static 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 */
76dda93c
YZ		 604static noinline int btrfs_mksubvol(struct path *parent,
605 char *name, int namelen,
72fd032e 606 struct btrfs_root *snap_src,
b83cc969 607 u64 *async_transid, bool readonly)
cb8e7090 608{
76dda93c 609 struct inode *dir = parent->dentry->d_inode;
cb8e7090
CH		 610 struct dentry *dentry;
611 int error;
612
76dda93c 613 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
cb8e7090
CH		 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
cb8e7090
CH		 624 error = mnt_want_write(parent->mnt);
625 if (error)
626 goto out_dput;
627
76dda93c 628 error = btrfs_may_create(dir, dentry);
cb8e7090
CH		 629 if (error)
630 goto out_drop_write;
631
76dda93c
YZ		 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
3de4586c 637 if (snap_src) {
72fd032e 638 error = create_snapshot(snap_src, dentry,
b83cc969 639 name, namelen, async_transid, readonly);
3de4586c 640 } else {
76dda93c 641 error = create_subvol(BTRFS_I(dir)->root, dentry,
72fd032e 642 name, namelen, async_transid);
3de4586c 643 }
76dda93c
YZ		 644 if (!error)
645 fsnotify_mkdir(dir, dentry);
646out_up_read:
647 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
cb8e7090
CH		 648out_drop_write:
649 mnt_drop_write(parent->mnt);
650out_dput:
651 dput(dentry);
652out_unlock:
76dda93c 653 mutex_unlock(&dir->i_mutex);
cb8e7090
CH		 654 return error;
655}
656
4cb5300b
CM		 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 */
664static 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 */
697static 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;
a4689d2b 708 u64 ino = btrfs_ino(inode);
4cb5300b
CM		 709
710 path = btrfs_alloc_path();
711 if (!path)
712 return -ENOMEM;
713
a4689d2b 714 min_key.objectid = ino;
4cb5300b
CM		 715 min_key.type = BTRFS_EXTENT_DATA_KEY;
716 min_key.offset = *off;
717
a4689d2b 718 max_key.objectid = ino;
4cb5300b
CM		 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;
a4689d2b 729 if (min_key.objectid != ino)
4cb5300b
CM		 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 }
753none:
754 btrfs_free_path(path);
755 return -ENOENT;
756}
757
940100a4 758static int should_defrag_range(struct inode *inode, u64 start, u64 len,
1e701a32
CM		 759 int thresh, u64 *last_len, u64 *skip,
760 u64 *defrag_end)
940100a4
CM		 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
6cf8bfbf 790 if (IS_ERR(em))
940100a4
CM		 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 */
1e701a32 801 if ((*last_len == 0 || *last_len >= thresh) && em->len >= thresh)
940100a4
CM		 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
4cb5300b
CM		 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 */
837static int cluster_pages_for_defrag(struct inode *inode,
838 struct page **pages,
839 unsigned long start_index,
840 int num_pages)
f46b5a66 841{
4cb5300b
CM		 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;
3eaa2885 849 struct btrfs_ordered_extent *ordered;
4cb5300b 850 struct extent_state *cached_state = NULL;
3b16a4e3 851 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
4cb5300b
CM		 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;
861again:
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;
a94733d0 868 page = find_or_create_page(inode->i_mapping,
3b16a4e3 869 start_index + i, mask);
4cb5300b
CM		 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) {
9e0baf60
JB		 939 spin_lock(&BTRFS_I(inode)->lock);
940 BTRFS_I(inode)->outstanding_extents++;
941 spin_unlock(&BTRFS_I(inode)->lock);
4cb5300b
CM		 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;
963out:
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
973int 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;
1a419d85 978 struct btrfs_super_block *disk_super;
4cb5300b 979 struct file_ra_state *ra = NULL;
f46b5a66 980 unsigned long last_index;
151a31b2 981 u64 isize = i_size_read(inode);
1a419d85 982 u64 features;
940100a4
CM		 983 u64 last_len = 0;
984 u64 skip = 0;
985 u64 defrag_end = 0;
4cb5300b
CM		 986 u64 newer_off = range->start;
987 int newer_left = 0;
f46b5a66
CH		 988 unsigned long i;
989 int ret;
4cb5300b 990 int defrag_count = 0;
1a419d85 991 int compress_type = BTRFS_COMPRESS_ZLIB;
4cb5300b
CM		 992 int extent_thresh = range->extent_thresh;
993 int newer_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
994 u64 new_align = ~((u64)128 * 1024 - 1);
995 struct page **pages = NULL;
996
997 if (extent_thresh == 0)
998 extent_thresh = 256 * 1024;
1a419d85
LZ		 999
1000 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1001 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1002 return -EINVAL;
1003 if (range->compress_type)
1004 compress_type = range->compress_type;
1005 }
f46b5a66 1006
151a31b2 1007 if (isize == 0)
940100a4
CM		 1008 return 0;
1009
4cb5300b
CM		 1010 /*
1011 * if we were not given a file, allocate a readahead
1012 * context
1013 */
1014 if (!file) {
1015 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1016 if (!ra)
1017 return -ENOMEM;
1018 file_ra_state_init(ra, inode->i_mapping);
1019 } else {
1020 ra = &file->f_ra;
1021 }
1022
1023 pages = kmalloc(sizeof(struct page *) * newer_cluster,
1024 GFP_NOFS);
1025 if (!pages) {
1026 ret = -ENOMEM;
1027 goto out_ra;
1028 }
1029
1030 /* find the last page to defrag */
1e701a32 1031 if (range->start + range->len > range->start) {
151a31b2 1032 last_index = min_t(u64, isize - 1,
1e701a32
CM		 1033 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1034 } else {
151a31b2 1035 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1e701a32
CM		 1036 }
1037
4cb5300b
CM		 1038 if (newer_than) {
1039 ret = find_new_extents(root, inode, newer_than,
1040 &newer_off, 64 * 1024);
1041 if (!ret) {
1042 range->start = newer_off;
1043 /*
1044 * we always align our defrag to help keep
1045 * the extents in the file evenly spaced
1046 */
1047 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1048 newer_left = newer_cluster;
1049 } else
1050 goto out_ra;
1051 } else {
1052 i = range->start >> PAGE_CACHE_SHIFT;
1053 }
1054 if (!max_to_defrag)
5ca49660 1055 max_to_defrag = last_index;
4cb5300b
CM		 1056
1057 while (i <= last_index && defrag_count < max_to_defrag) {
1058 /*
1059 * make sure we stop running if someone unmounts
1060 * the FS
1061 */
1062 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1063 break;
1064
1065 if (!newer_than &&
1066 !should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1e701a32 1067 PAGE_CACHE_SIZE,
4cb5300b 1068 extent_thresh,
1e701a32 1069 &last_len, &skip,
940100a4
CM		 1070 &defrag_end)) {
1071 unsigned long next;
1072 /*
1073 * the should_defrag function tells us how much to skip
1074 * bump our counter by the suggested amount
1075 */
1076 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1077 i = max(i + 1, next);
1078 continue;
1079 }
1e701a32 1080 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1a419d85 1081 BTRFS_I(inode)->force_compress = compress_type;
940100a4 1082
4cb5300b 1083 btrfs_force_ra(inode->i_mapping, ra, file, i, newer_cluster);
940100a4 1084
4cb5300b
CM		 1085 ret = cluster_pages_for_defrag(inode, pages, i, newer_cluster);
1086 if (ret < 0)
1087 goto out_ra;
1088
1089 defrag_count += ret;
1090 balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret);
4cb5300b
CM		 1091
1092 if (newer_than) {
1093 if (newer_off == (u64)-1)
1094 break;
1095
1096 newer_off = max(newer_off + 1,
1097 (u64)i << PAGE_CACHE_SHIFT);
1098
1099 ret = find_new_extents(root, inode,
1100 newer_than, &newer_off,
1101 64 * 1024);
1102 if (!ret) {
1103 range->start = newer_off;
1104 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1105 newer_left = newer_cluster;
1106 } else {
1107 break;
f46b5a66 1108 }
4cb5300b 1109 } else {
cbcc8326
LZ		 1110 if (ret > 0)
1111 i += ret;
1112 else
1113 i++;
f46b5a66 1114 }
f46b5a66
CH		 1115 }
1116
1e701a32
CM		 1117 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1118 filemap_flush(inode->i_mapping);
1119
1120 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1121 /* the filemap_flush will queue IO into the worker threads, but
1122 * we have to make sure the IO is actually started and that
1123 * ordered extents get created before we return
1124 */
1125 atomic_inc(&root->fs_info->async_submit_draining);
1126 while (atomic_read(&root->fs_info->nr_async_submits) ||
1127 atomic_read(&root->fs_info->async_delalloc_pages)) {
1128 wait_event(root->fs_info->async_submit_wait,
1129 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1130 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1131 }
1132 atomic_dec(&root->fs_info->async_submit_draining);
1133
1134 mutex_lock(&inode->i_mutex);
261507a0 1135 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1e701a32
CM		 1136 mutex_unlock(&inode->i_mutex);
1137 }
1138
1a419d85
LZ		 1139 disk_super = &root->fs_info->super_copy;
1140 features = btrfs_super_incompat_flags(disk_super);
1141 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1142 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
1143 btrfs_set_super_incompat_flags(disk_super, features);
1144 }
1145
60ccf82f 1146 ret = defrag_count;
940100a4 1147
4cb5300b
CM		 1148out_ra:
1149 if (!file)
1150 kfree(ra);
1151 kfree(pages);
940100a4 1152 return ret;
f46b5a66
CH		 1153}
1154
76dda93c
YZ		 1155static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
1156 void __user *arg)
f46b5a66
CH		 1157{
1158 u64 new_size;
1159 u64 old_size;
1160 u64 devid = 1;
1161 struct btrfs_ioctl_vol_args *vol_args;
1162 struct btrfs_trans_handle *trans;
1163 struct btrfs_device *device = NULL;
1164 char *sizestr;
1165 char *devstr = NULL;
1166 int ret = 0;
f46b5a66
CH		 1167 int mod = 0;
1168
c146afad
YZ		 1169 if (root->fs_info->sb->s_flags & MS_RDONLY)
1170 return -EROFS;
1171
e441d54d
CM		 1172 if (!capable(CAP_SYS_ADMIN))
1173 return -EPERM;
1174
dae7b665
LZ		 1175 vol_args = memdup_user(arg, sizeof(*vol_args));
1176 if (IS_ERR(vol_args))
1177 return PTR_ERR(vol_args);
5516e595
MF		 1178
1179 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 1180
7d9eb12c 1181 mutex_lock(&root->fs_info->volume_mutex);
f46b5a66
CH		 1182 sizestr = vol_args->name;
1183 devstr = strchr(sizestr, ':');
1184 if (devstr) {
1185 char *end;
1186 sizestr = devstr + 1;
1187 *devstr = '\0';
1188 devstr = vol_args->name;
1189 devid = simple_strtoull(devstr, &end, 10);
21380931
JB		 1190 printk(KERN_INFO "resizing devid %llu\n",
1191 (unsigned long long)devid);
f46b5a66 1192 }
2b82032c 1193 device = btrfs_find_device(root, devid, NULL, NULL);
f46b5a66 1194 if (!device) {
21380931
JB		 1195 printk(KERN_INFO "resizer unable to find device %llu\n",
1196 (unsigned long long)devid);
f46b5a66
CH		 1197 ret = -EINVAL;
1198 goto out_unlock;
1199 }
1200 if (!strcmp(sizestr, "max"))
1201 new_size = device->bdev->bd_inode->i_size;
1202 else {
1203 if (sizestr[0] == '-') {
1204 mod = -1;
1205 sizestr++;
1206 } else if (sizestr[0] == '+') {
1207 mod = 1;
1208 sizestr++;
1209 }
91748467 1210 new_size = memparse(sizestr, NULL);
f46b5a66
CH		 1211 if (new_size == 0) {
1212 ret = -EINVAL;
1213 goto out_unlock;
1214 }
1215 }
1216
1217 old_size = device->total_bytes;
1218
1219 if (mod < 0) {
1220 if (new_size > old_size) {
1221 ret = -EINVAL;
1222 goto out_unlock;
1223 }
1224 new_size = old_size - new_size;
1225 } else if (mod > 0) {
1226 new_size = old_size + new_size;
1227 }
1228
1229 if (new_size < 256 * 1024 * 1024) {
1230 ret = -EINVAL;
1231 goto out_unlock;
1232 }
1233 if (new_size > device->bdev->bd_inode->i_size) {
1234 ret = -EFBIG;
1235 goto out_unlock;
1236 }
1237
1238 do_div(new_size, root->sectorsize);
1239 new_size *= root->sectorsize;
1240
1241 printk(KERN_INFO "new size for %s is %llu\n",
1242 device->name, (unsigned long long)new_size);
1243
1244 if (new_size > old_size) {
a22285a6 1245 trans = btrfs_start_transaction(root, 0);
98d5dc13
TI		 1246 if (IS_ERR(trans)) {
1247 ret = PTR_ERR(trans);
1248 goto out_unlock;
1249 }
f46b5a66
CH		 1250 ret = btrfs_grow_device(trans, device, new_size);
1251 btrfs_commit_transaction(trans, root);
1252 } else {
1253 ret = btrfs_shrink_device(device, new_size);
1254 }
1255
1256out_unlock:
7d9eb12c 1257 mutex_unlock(&root->fs_info->volume_mutex);
f46b5a66
CH		 1258 kfree(vol_args);
1259 return ret;
1260}
1261
72fd032e
SW		 1262static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1263 char *name,
1264 unsigned long fd,
1265 int subvol,
b83cc969
LZ		 1266 u64 *transid,
1267 bool readonly)
f46b5a66 1268{
cb8e7090 1269 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3de4586c 1270 struct file *src_file;
f46b5a66 1271 int namelen;
3de4586c 1272 int ret = 0;
f46b5a66 1273
c146afad
YZ		 1274 if (root->fs_info->sb->s_flags & MS_RDONLY)
1275 return -EROFS;
1276
72fd032e
SW		 1277 namelen = strlen(name);
1278 if (strchr(name, '/')) {
f46b5a66
CH		 1279 ret = -EINVAL;
1280 goto out;
1281 }
1282
3de4586c 1283 if (subvol) {
72fd032e 1284 ret = btrfs_mksubvol(&file->f_path, name, namelen,
b83cc969 1285 NULL, transid, readonly);
cb8e7090 1286 } else {
3de4586c 1287 struct inode *src_inode;
72fd032e 1288 src_file = fget(fd);
3de4586c
CM		 1289 if (!src_file) {
1290 ret = -EINVAL;
1291 goto out;
1292 }
1293
1294 src_inode = src_file->f_path.dentry->d_inode;
1295 if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
d397712b
CM		 1296 printk(KERN_INFO "btrfs: Snapshot src from "
1297 "another FS\n");
3de4586c
CM		 1298 ret = -EINVAL;
1299 fput(src_file);
1300 goto out;
1301 }
72fd032e
SW		 1302 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1303 BTRFS_I(src_inode)->root,
b83cc969 1304 transid, readonly);
3de4586c 1305 fput(src_file);
cb8e7090 1306 }
f46b5a66 1307out:
72fd032e
SW		 1308 return ret;
1309}
1310
1311static noinline int btrfs_ioctl_snap_create(struct file *file,
fa0d2b9b 1312 void __user *arg, int subvol)
72fd032e 1313{
fa0d2b9b 1314 struct btrfs_ioctl_vol_args *vol_args;
72fd032e
SW		 1315 int ret;
1316
fa0d2b9b
LZ		 1317 vol_args = memdup_user(arg, sizeof(*vol_args));
1318 if (IS_ERR(vol_args))
1319 return PTR_ERR(vol_args);
1320 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
72fd032e 1321
fa0d2b9b 1322 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
b83cc969
LZ		 1323 vol_args->fd, subvol,
1324 NULL, false);
fdfb1e4f 1325
fa0d2b9b
LZ		 1326 kfree(vol_args);
1327 return ret;
1328}
fdfb1e4f 1329
fa0d2b9b
LZ		 1330static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1331 void __user *arg, int subvol)
1332{
1333 struct btrfs_ioctl_vol_args_v2 *vol_args;
1334 int ret;
1335 u64 transid = 0;
1336 u64 *ptr = NULL;
b83cc969 1337 bool readonly = false;
75eaa0e2 1338
fa0d2b9b
LZ		 1339 vol_args = memdup_user(arg, sizeof(*vol_args));
1340 if (IS_ERR(vol_args))
1341 return PTR_ERR(vol_args);
1342 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
75eaa0e2 1343
b83cc969
LZ		 1344 if (vol_args->flags &
1345 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY)) {
1346 ret = -EOPNOTSUPP;
fa0d2b9b 1347 goto out;
72fd032e 1348 }
fa0d2b9b
LZ		 1349
1350 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1351 ptr = &transid;
b83cc969
LZ		 1352 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1353 readonly = true;
fa0d2b9b
LZ		 1354
1355 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
b83cc969
LZ		 1356 vol_args->fd, subvol,
1357 ptr, readonly);
fa0d2b9b
LZ		 1358
1359 if (ret == 0 && ptr &&
1360 copy_to_user(arg +
1361 offsetof(struct btrfs_ioctl_vol_args_v2,
1362 transid), ptr, sizeof(*ptr)))
1363 ret = -EFAULT;
fdfb1e4f 1364out:
f46b5a66
CH		 1365 kfree(vol_args);
1366 return ret;
1367}
1368
0caa102d
LZ		 1369static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1370 void __user *arg)
1371{
1372 struct inode *inode = fdentry(file)->d_inode;
1373 struct btrfs_root *root = BTRFS_I(inode)->root;
1374 int ret = 0;
1375 u64 flags = 0;
1376
33345d01 1377 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
0caa102d
LZ		 1378 return -EINVAL;
1379
1380 down_read(&root->fs_info->subvol_sem);
1381 if (btrfs_root_readonly(root))
1382 flags |= BTRFS_SUBVOL_RDONLY;
1383 up_read(&root->fs_info->subvol_sem);
1384
1385 if (copy_to_user(arg, &flags, sizeof(flags)))
1386 ret = -EFAULT;
1387
1388 return ret;
1389}
1390
1391static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1392 void __user *arg)
1393{
1394 struct inode *inode = fdentry(file)->d_inode;
1395 struct btrfs_root *root = BTRFS_I(inode)->root;
1396 struct btrfs_trans_handle *trans;
1397 u64 root_flags;
1398 u64 flags;
1399 int ret = 0;
1400
1401 if (root->fs_info->sb->s_flags & MS_RDONLY)
1402 return -EROFS;
1403
33345d01 1404 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
0caa102d
LZ		 1405 return -EINVAL;
1406
1407 if (copy_from_user(&flags, arg, sizeof(flags)))
1408 return -EFAULT;
1409
b4dc2b8c 1410 if (flags & BTRFS_SUBVOL_CREATE_ASYNC)
0caa102d
LZ		 1411 return -EINVAL;
1412
1413 if (flags & ~BTRFS_SUBVOL_RDONLY)
1414 return -EOPNOTSUPP;
1415
2e149670 1416 if (!inode_owner_or_capable(inode))
b4dc2b8c
LZ		 1417 return -EACCES;
1418
0caa102d
LZ		 1419 down_write(&root->fs_info->subvol_sem);
1420
1421 /* nothing to do */
1422 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1423 goto out;
1424
1425 root_flags = btrfs_root_flags(&root->root_item);
1426 if (flags & BTRFS_SUBVOL_RDONLY)
1427 btrfs_set_root_flags(&root->root_item,
1428 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1429 else
1430 btrfs_set_root_flags(&root->root_item,
1431 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1432
1433 trans = btrfs_start_transaction(root, 1);
1434 if (IS_ERR(trans)) {
1435 ret = PTR_ERR(trans);
1436 goto out_reset;
1437 }
1438
b4dc2b8c 1439 ret = btrfs_update_root(trans, root->fs_info->tree_root,
0caa102d
LZ		 1440 &root->root_key, &root->root_item);
1441
1442 btrfs_commit_transaction(trans, root);
1443out_reset:
1444 if (ret)
1445 btrfs_set_root_flags(&root->root_item, root_flags);
1446out:
1447 up_write(&root->fs_info->subvol_sem);
1448 return ret;
1449}
1450
76dda93c
YZ		 1451/*
1452 * helper to check if the subvolume references other subvolumes
1453 */
1454static noinline int may_destroy_subvol(struct btrfs_root *root)
1455{
1456 struct btrfs_path *path;
1457 struct btrfs_key key;
1458 int ret;
1459
1460 path = btrfs_alloc_path();
1461 if (!path)
1462 return -ENOMEM;
1463
1464 key.objectid = root->root_key.objectid;
1465 key.type = BTRFS_ROOT_REF_KEY;
1466 key.offset = (u64)-1;
1467
1468 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1469 &key, path, 0, 0);
1470 if (ret < 0)
1471 goto out;
1472 BUG_ON(ret == 0);
1473
1474 ret = 0;
1475 if (path->slots[0] > 0) {
1476 path->slots[0]--;
1477 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1478 if (key.objectid == root->root_key.objectid &&
1479 key.type == BTRFS_ROOT_REF_KEY)
1480 ret = -ENOTEMPTY;
1481 }
1482out:
1483 btrfs_free_path(path);
1484 return ret;
1485}
1486
ac8e9819
CM		 1487static noinline int key_in_sk(struct btrfs_key *key,
1488 struct btrfs_ioctl_search_key *sk)
1489{
abc6e134
CM		 1490 struct btrfs_key test;
1491 int ret;
1492
1493 test.objectid = sk->min_objectid;
1494 test.type = sk->min_type;
1495 test.offset = sk->min_offset;
1496
1497 ret = btrfs_comp_cpu_keys(key, &test);
1498 if (ret < 0)
ac8e9819 1499 return 0;
abc6e134
CM		 1500
1501 test.objectid = sk->max_objectid;
1502 test.type = sk->max_type;
1503 test.offset = sk->max_offset;
1504
1505 ret = btrfs_comp_cpu_keys(key, &test);
1506 if (ret > 0)
ac8e9819
CM		 1507 return 0;
1508 return 1;
1509}
1510
1511static noinline int copy_to_sk(struct btrfs_root *root,
1512 struct btrfs_path *path,
1513 struct btrfs_key *key,
1514 struct btrfs_ioctl_search_key *sk,
1515 char *buf,
1516 unsigned long *sk_offset,
1517 int *num_found)
1518{
1519 u64 found_transid;
1520 struct extent_buffer *leaf;
1521 struct btrfs_ioctl_search_header sh;
1522 unsigned long item_off;
1523 unsigned long item_len;
1524 int nritems;
1525 int i;
1526 int slot;
ac8e9819
CM		 1527 int ret = 0;
1528
1529 leaf = path->nodes[0];
1530 slot = path->slots[0];
1531 nritems = btrfs_header_nritems(leaf);
1532
1533 if (btrfs_header_generation(leaf) > sk->max_transid) {
1534 i = nritems;
1535 goto advance_key;
1536 }
1537 found_transid = btrfs_header_generation(leaf);
1538
1539 for (i = slot; i < nritems; i++) {
1540 item_off = btrfs_item_ptr_offset(leaf, i);
1541 item_len = btrfs_item_size_nr(leaf, i);
1542
1543 if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1544 item_len = 0;
1545
1546 if (sizeof(sh) + item_len + *sk_offset >
1547 BTRFS_SEARCH_ARGS_BUFSIZE) {
1548 ret = 1;
1549 goto overflow;
1550 }
1551
1552 btrfs_item_key_to_cpu(leaf, key, i);
1553 if (!key_in_sk(key, sk))
1554 continue;
1555
1556 sh.objectid = key->objectid;
1557 sh.offset = key->offset;
1558 sh.type = key->type;
1559 sh.len = item_len;
1560 sh.transid = found_transid;
1561
1562 /* copy search result header */
1563 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1564 *sk_offset += sizeof(sh);
1565
1566 if (item_len) {
1567 char *p = buf + *sk_offset;
1568 /* copy the item */
1569 read_extent_buffer(leaf, p,
1570 item_off, item_len);
1571 *sk_offset += item_len;
ac8e9819 1572 }
e2156867 1573 (*num_found)++;
ac8e9819
CM		 1574
1575 if (*num_found >= sk->nr_items)
1576 break;
1577 }
1578advance_key:
abc6e134
CM		 1579 ret = 0;
1580 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
ac8e9819 1581 key->offset++;
abc6e134
CM		 1582 else if (key->type < (u8)-1 && key->type < sk->max_type) {
1583 key->offset = 0;
ac8e9819 1584 key->type++;
abc6e134
CM		 1585 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1586 key->offset = 0;
1587 key->type = 0;
ac8e9819 1588 key->objectid++;
abc6e134
CM		 1589 } else
1590 ret = 1;
ac8e9819 1591overflow:
ac8e9819
CM		 1592 return ret;
1593}
1594
1595static noinline int search_ioctl(struct inode *inode,
1596 struct btrfs_ioctl_search_args *args)
1597{
1598 struct btrfs_root *root;
1599 struct btrfs_key key;
1600 struct btrfs_key max_key;
1601 struct btrfs_path *path;
1602 struct btrfs_ioctl_search_key *sk = &args->key;
1603 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1604 int ret;
1605 int num_found = 0;
1606 unsigned long sk_offset = 0;
1607
1608 path = btrfs_alloc_path();
1609 if (!path)
1610 return -ENOMEM;
1611
1612 if (sk->tree_id == 0) {
1613 /* search the root of the inode that was passed */
1614 root = BTRFS_I(inode)->root;
1615 } else {
1616 key.objectid = sk->tree_id;
1617 key.type = BTRFS_ROOT_ITEM_KEY;
1618 key.offset = (u64)-1;
1619 root = btrfs_read_fs_root_no_name(info, &key);
1620 if (IS_ERR(root)) {
1621 printk(KERN_ERR "could not find root %llu\n",
1622 sk->tree_id);
1623 btrfs_free_path(path);
1624 return -ENOENT;
1625 }
1626 }
1627
1628 key.objectid = sk->min_objectid;
1629 key.type = sk->min_type;
1630 key.offset = sk->min_offset;
1631
1632 max_key.objectid = sk->max_objectid;
1633 max_key.type = sk->max_type;
1634 max_key.offset = sk->max_offset;
1635
1636 path->keep_locks = 1;
1637
1638 while(1) {
1639 ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1640 sk->min_transid);
1641 if (ret != 0) {
1642 if (ret > 0)
1643 ret = 0;
1644 goto err;
1645 }
1646 ret = copy_to_sk(root, path, &key, sk, args->buf,
1647 &sk_offset, &num_found);
b3b4aa74 1648 btrfs_release_path(path);
ac8e9819
CM		 1649 if (ret || num_found >= sk->nr_items)
1650 break;
1651
1652 }
1653 ret = 0;
1654err:
1655 sk->nr_items = num_found;
1656 btrfs_free_path(path);
1657 return ret;
1658}
1659
1660static noinline int btrfs_ioctl_tree_search(struct file *file,
1661 void __user *argp)
1662{
1663 struct btrfs_ioctl_search_args *args;
1664 struct inode *inode;
1665 int ret;
1666
1667 if (!capable(CAP_SYS_ADMIN))
1668 return -EPERM;
1669
2354d08f
JL		 1670 args = memdup_user(argp, sizeof(*args));
1671 if (IS_ERR(args))
1672 return PTR_ERR(args);
ac8e9819 1673
ac8e9819
CM		 1674 inode = fdentry(file)->d_inode;
1675 ret = search_ioctl(inode, args);
1676 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1677 ret = -EFAULT;
1678 kfree(args);
1679 return ret;
1680}
1681
98d377a0 1682/*
ac8e9819
CM		 1683 * Search INODE_REFs to identify path name of 'dirid' directory
1684 * in a 'tree_id' tree. and sets path name to 'name'.
1685 */
98d377a0
TH		 1686static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1687 u64 tree_id, u64 dirid, char *name)
1688{
1689 struct btrfs_root *root;
1690 struct btrfs_key key;
ac8e9819 1691 char *ptr;
98d377a0
TH		 1692 int ret = -1;
1693 int slot;
1694 int len;
1695 int total_len = 0;
1696 struct btrfs_inode_ref *iref;
1697 struct extent_buffer *l;
1698 struct btrfs_path *path;
1699
1700 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1701 name[0]='\0';
1702 return 0;
1703 }
1704
1705 path = btrfs_alloc_path();
1706 if (!path)
1707 return -ENOMEM;
1708
ac8e9819 1709 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
98d377a0
TH		 1710
1711 key.objectid = tree_id;
1712 key.type = BTRFS_ROOT_ITEM_KEY;
1713 key.offset = (u64)-1;
1714 root = btrfs_read_fs_root_no_name(info, &key);
1715 if (IS_ERR(root)) {
1716 printk(KERN_ERR "could not find root %llu\n", tree_id);
8ad6fcab
CM		 1717 ret = -ENOENT;
1718 goto out;
98d377a0
TH		 1719 }
1720
1721 key.objectid = dirid;
1722 key.type = BTRFS_INODE_REF_KEY;
8ad6fcab 1723 key.offset = (u64)-1;
98d377a0
TH		 1724
1725 while(1) {
1726 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1727 if (ret < 0)
1728 goto out;
1729
1730 l = path->nodes[0];
1731 slot = path->slots[0];
8ad6fcab
CM		 1732 if (ret > 0 && slot > 0)
1733 slot--;
98d377a0
TH		 1734 btrfs_item_key_to_cpu(l, &key, slot);
1735
1736 if (ret > 0 && (key.objectid != dirid ||
ac8e9819
CM		 1737 key.type != BTRFS_INODE_REF_KEY)) {
1738 ret = -ENOENT;
98d377a0 1739 goto out;
ac8e9819 1740 }
98d377a0
TH		 1741
1742 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1743 len = btrfs_inode_ref_name_len(l, iref);
1744 ptr -= len + 1;
1745 total_len += len + 1;
ac8e9819 1746 if (ptr < name)
98d377a0
TH		 1747 goto out;
1748
1749 *(ptr + len) = '/';
1750 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1751
1752 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1753 break;
1754
b3b4aa74 1755 btrfs_release_path(path);
98d377a0 1756 key.objectid = key.offset;
8ad6fcab 1757 key.offset = (u64)-1;
98d377a0 1758 dirid = key.objectid;
98d377a0 1759 }
ac8e9819 1760 if (ptr < name)
98d377a0 1761 goto out;
77906a50 1762 memmove(name, ptr, total_len);
98d377a0
TH		 1763 name[total_len]='\0';
1764 ret = 0;
1765out:
1766 btrfs_free_path(path);
ac8e9819
CM		 1767 return ret;
1768}
1769
1770static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1771 void __user *argp)
1772{
1773 struct btrfs_ioctl_ino_lookup_args *args;
1774 struct inode *inode;
1775 int ret;
1776
1777 if (!capable(CAP_SYS_ADMIN))
1778 return -EPERM;
1779
2354d08f
JL		 1780 args = memdup_user(argp, sizeof(*args));
1781 if (IS_ERR(args))
1782 return PTR_ERR(args);
c2b96929 1783
ac8e9819
CM		 1784 inode = fdentry(file)->d_inode;
1785
1b53ac4d
CM		 1786 if (args->treeid == 0)
1787 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1788
ac8e9819
CM		 1789 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1790 args->treeid, args->objectid,
1791 args->name);
1792
1793 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1794 ret = -EFAULT;
1795
1796 kfree(args);
98d377a0
TH		 1797 return ret;
1798}
1799
76dda93c
YZ		 1800static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1801 void __user *arg)
1802{
1803 struct dentry *parent = fdentry(file);
1804 struct dentry *dentry;
1805 struct inode *dir = parent->d_inode;
1806 struct inode *inode;
1807 struct btrfs_root *root = BTRFS_I(dir)->root;
1808 struct btrfs_root *dest = NULL;
1809 struct btrfs_ioctl_vol_args *vol_args;
1810 struct btrfs_trans_handle *trans;
1811 int namelen;
1812 int ret;
1813 int err = 0;
1814
76dda93c
YZ		 1815 vol_args = memdup_user(arg, sizeof(*vol_args));
1816 if (IS_ERR(vol_args))
1817 return PTR_ERR(vol_args);
1818
1819 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1820 namelen = strlen(vol_args->name);
1821 if (strchr(vol_args->name, '/') ||
1822 strncmp(vol_args->name, "..", namelen) == 0) {
1823 err = -EINVAL;
1824 goto out;
1825 }
1826
1827 err = mnt_want_write(file->f_path.mnt);
1828 if (err)
1829 goto out;
1830
1831 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
1832 dentry = lookup_one_len(vol_args->name, parent, namelen);
1833 if (IS_ERR(dentry)) {
1834 err = PTR_ERR(dentry);
1835 goto out_unlock_dir;
1836 }
1837
1838 if (!dentry->d_inode) {
1839 err = -ENOENT;
1840 goto out_dput;
1841 }
1842
1843 inode = dentry->d_inode;
4260f7c7
SW		 1844 dest = BTRFS_I(inode)->root;
1845 if (!capable(CAP_SYS_ADMIN)){
1846 /*
1847 * Regular user. Only allow this with a special mount
1848 * option, when the user has write+exec access to the
1849 * subvol root, and when rmdir(2) would have been
1850 * allowed.
1851 *
1852 * Note that this is _not_ check that the subvol is
1853 * empty or doesn't contain data that we wouldn't
1854 * otherwise be able to delete.
1855 *
1856 * Users who want to delete empty subvols should try
1857 * rmdir(2).
1858 */
1859 err = -EPERM;
1860 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1861 goto out_dput;
1862
1863 /*
1864 * Do not allow deletion if the parent dir is the same
1865 * as the dir to be deleted. That means the ioctl
1866 * must be called on the dentry referencing the root
1867 * of the subvol, not a random directory contained
1868 * within it.
1869 */
1870 err = -EINVAL;
1871 if (root == dest)
1872 goto out_dput;
1873
1874 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
1875 if (err)
1876 goto out_dput;
1877
1878 /* check if subvolume may be deleted by a non-root user */
1879 err = btrfs_may_delete(dir, dentry, 1);
1880 if (err)
1881 goto out_dput;
1882 }
1883
33345d01 1884 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
76dda93c
YZ		 1885 err = -EINVAL;
1886 goto out_dput;
1887 }
1888
76dda93c
YZ		 1889 mutex_lock(&inode->i_mutex);
1890 err = d_invalidate(dentry);
1891 if (err)
1892 goto out_unlock;
1893
1894 down_write(&root->fs_info->subvol_sem);
1895
1896 err = may_destroy_subvol(dest);
1897 if (err)
1898 goto out_up_write;
1899
a22285a6
YZ		 1900 trans = btrfs_start_transaction(root, 0);
1901 if (IS_ERR(trans)) {
1902 err = PTR_ERR(trans);
d327099a 1903 goto out_up_write;
a22285a6
YZ		 1904 }
1905 trans->block_rsv = &root->fs_info->global_block_rsv;
1906
76dda93c
YZ		 1907 ret = btrfs_unlink_subvol(trans, root, dir,
1908 dest->root_key.objectid,
1909 dentry->d_name.name,
1910 dentry->d_name.len);
1911 BUG_ON(ret);
1912
1913 btrfs_record_root_in_trans(trans, dest);
1914
1915 memset(&dest->root_item.drop_progress, 0,
1916 sizeof(dest->root_item.drop_progress));
1917 dest->root_item.drop_level = 0;
1918 btrfs_set_root_refs(&dest->root_item, 0);
1919
d68fc57b
YZ		 1920 if (!xchg(&dest->orphan_item_inserted, 1)) {
1921 ret = btrfs_insert_orphan_item(trans,
1922 root->fs_info->tree_root,
1923 dest->root_key.objectid);
1924 BUG_ON(ret);
1925 }
76dda93c 1926
531cb13f 1927 ret = btrfs_end_transaction(trans, root);
76dda93c
YZ		 1928 BUG_ON(ret);
1929 inode->i_flags |= S_DEAD;
1930out_up_write:
1931 up_write(&root->fs_info->subvol_sem);
1932out_unlock:
1933 mutex_unlock(&inode->i_mutex);
1934 if (!err) {
efefb143 1935 shrink_dcache_sb(root->fs_info->sb);
76dda93c
YZ		 1936 btrfs_invalidate_inodes(dest);
1937 d_delete(dentry);
1938 }
1939out_dput:
1940 dput(dentry);
1941out_unlock_dir:
1942 mutex_unlock(&dir->i_mutex);
1943 mnt_drop_write(file->f_path.mnt);
1944out:
1945 kfree(vol_args);
1946 return err;
1947}
1948
1e701a32 1949static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
f46b5a66
CH		 1950{
1951 struct inode *inode = fdentry(file)->d_inode;
1952 struct btrfs_root *root = BTRFS_I(inode)->root;
1e701a32 1953 struct btrfs_ioctl_defrag_range_args *range;
c146afad
YZ		 1954 int ret;
1955
b83cc969
LZ		 1956 if (btrfs_root_readonly(root))
1957 return -EROFS;
1958
c146afad
YZ		 1959 ret = mnt_want_write(file->f_path.mnt);
1960 if (ret)
1961 return ret;
f46b5a66
CH		 1962
1963 switch (inode->i_mode & S_IFMT) {
1964 case S_IFDIR:
e441d54d
CM		 1965 if (!capable(CAP_SYS_ADMIN)) {
1966 ret = -EPERM;
1967 goto out;
1968 }
8929ecfa
YZ		 1969 ret = btrfs_defrag_root(root, 0);
1970 if (ret)
1971 goto out;
1972 ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
f46b5a66
CH		 1973 break;
1974 case S_IFREG:
e441d54d
CM		 1975 if (!(file->f_mode & FMODE_WRITE)) {
1976 ret = -EINVAL;
1977 goto out;
1978 }
1e701a32
CM		 1979
1980 range = kzalloc(sizeof(*range), GFP_KERNEL);
1981 if (!range) {
1982 ret = -ENOMEM;
1983 goto out;
1984 }
1985
1986 if (argp) {
1987 if (copy_from_user(range, argp,
1988 sizeof(*range))) {
1989 ret = -EFAULT;
1990 kfree(range);
683be16e 1991 goto out;
1e701a32
CM		 1992 }
1993 /* compression requires us to start the IO */
1994 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1995 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
1996 range->extent_thresh = (u32)-1;
1997 }
1998 } else {
1999 /* the rest are all set to zero by kzalloc */
2000 range->len = (u64)-1;
2001 }
4cb5300b
CM		 2002 ret = btrfs_defrag_file(fdentry(file)->d_inode, file,
2003 range, 0, 0);
2004 if (ret > 0)
2005 ret = 0;
1e701a32 2006 kfree(range);
f46b5a66 2007 break;
8929ecfa
YZ		 2008 default:
2009 ret = -EINVAL;
f46b5a66 2010 }
e441d54d 2011out:
ab67b7c1 2012 mnt_drop_write(file->f_path.mnt);
e441d54d 2013 return ret;
f46b5a66
CH		 2014}
2015
b2950863 2016static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
f46b5a66
CH		 2017{
2018 struct btrfs_ioctl_vol_args *vol_args;
2019 int ret;
2020
e441d54d
CM		 2021 if (!capable(CAP_SYS_ADMIN))
2022 return -EPERM;
2023
dae7b665
LZ		 2024 vol_args = memdup_user(arg, sizeof(*vol_args));
2025 if (IS_ERR(vol_args))
2026 return PTR_ERR(vol_args);
f46b5a66 2027
5516e595 2028 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66
CH		 2029 ret = btrfs_init_new_device(root, vol_args->name);
2030
f46b5a66
CH		 2031 kfree(vol_args);
2032 return ret;
2033}
2034
b2950863 2035static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
f46b5a66
CH		 2036{
2037 struct btrfs_ioctl_vol_args *vol_args;
2038 int ret;
2039
e441d54d
CM		 2040 if (!capable(CAP_SYS_ADMIN))
2041 return -EPERM;
2042
c146afad
YZ		 2043 if (root->fs_info->sb->s_flags & MS_RDONLY)
2044 return -EROFS;
2045
dae7b665
LZ		 2046 vol_args = memdup_user(arg, sizeof(*vol_args));
2047 if (IS_ERR(vol_args))
2048 return PTR_ERR(vol_args);
f46b5a66 2049
5516e595 2050 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66
CH		 2051 ret = btrfs_rm_device(root, vol_args->name);
2052
f46b5a66
CH		 2053 kfree(vol_args);
2054 return ret;
2055}
2056
475f6387
JS		 2057static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2058{
027ed2f0 2059 struct btrfs_ioctl_fs_info_args *fi_args;
475f6387
JS		 2060 struct btrfs_device *device;
2061 struct btrfs_device *next;
2062 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
027ed2f0 2063 int ret = 0;
475f6387
JS		 2064
2065 if (!capable(CAP_SYS_ADMIN))
2066 return -EPERM;
2067
027ed2f0
LZ		 2068 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2069 if (!fi_args)
2070 return -ENOMEM;
2071
2072 fi_args->num_devices = fs_devices->num_devices;
2073 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
475f6387
JS		 2074
2075 mutex_lock(&fs_devices->device_list_mutex);
2076 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
027ed2f0
LZ		 2077 if (device->devid > fi_args->max_id)
2078 fi_args->max_id = device->devid;
475f6387
JS		 2079 }
2080 mutex_unlock(&fs_devices->device_list_mutex);
2081
027ed2f0
LZ		 2082 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2083 ret = -EFAULT;
475f6387 2084
027ed2f0
LZ		 2085 kfree(fi_args);
2086 return ret;
475f6387
JS		 2087}
2088
2089static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2090{
2091 struct btrfs_ioctl_dev_info_args *di_args;
2092 struct btrfs_device *dev;
2093 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2094 int ret = 0;
2095 char *s_uuid = NULL;
2096 char empty_uuid[BTRFS_UUID_SIZE] = {0};
2097
2098 if (!capable(CAP_SYS_ADMIN))
2099 return -EPERM;
2100
2101 di_args = memdup_user(arg, sizeof(*di_args));
2102 if (IS_ERR(di_args))
2103 return PTR_ERR(di_args);
2104
2105 if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2106 s_uuid = di_args->uuid;
2107
2108 mutex_lock(&fs_devices->device_list_mutex);
2109 dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL);
2110 mutex_unlock(&fs_devices->device_list_mutex);
2111
2112 if (!dev) {
2113 ret = -ENODEV;
2114 goto out;
2115 }
2116
2117 di_args->devid = dev->devid;
2118 di_args->bytes_used = dev->bytes_used;
2119 di_args->total_bytes = dev->total_bytes;
2120 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2121 strncpy(di_args->path, dev->name, sizeof(di_args->path));
2122
2123out:
2124 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2125 ret = -EFAULT;
2126
2127 kfree(di_args);
2128 return ret;
2129}
2130
76dda93c
YZ		 2131static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2132 u64 off, u64 olen, u64 destoff)
f46b5a66
CH		 2133{
2134 struct inode *inode = fdentry(file)->d_inode;
2135 struct btrfs_root *root = BTRFS_I(inode)->root;
2136 struct file *src_file;
2137 struct inode *src;
2138 struct btrfs_trans_handle *trans;
f46b5a66 2139 struct btrfs_path *path;
f46b5a66 2140 struct extent_buffer *leaf;
ae01a0ab
YZ		 2141 char *buf;
2142 struct btrfs_key key;
f46b5a66
CH		 2143 u32 nritems;
2144 int slot;
ae01a0ab 2145 int ret;
c5c9cd4d
SW		 2146 u64 len = olen;
2147 u64 bs = root->fs_info->sb->s_blocksize;
2148 u64 hint_byte;
d20f7043 2149
c5c9cd4d
SW		 2150 /*
2151 * TODO:
2152 * - split compressed inline extents. annoying: we need to
2153 * decompress into destination's address_space (the file offset
2154 * may change, so source mapping won't do), then recompress (or
2155 * otherwise reinsert) a subrange.
2156 * - allow ranges within the same file to be cloned (provided
2157 * they don't overlap)?
2158 */
2159
e441d54d 2160 /* the destination must be opened for writing */
2ebc3464 2161 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
e441d54d
CM		 2162 return -EINVAL;
2163
b83cc969
LZ		 2164 if (btrfs_root_readonly(root))
2165 return -EROFS;
2166
c146afad
YZ		 2167 ret = mnt_want_write(file->f_path.mnt);
2168 if (ret)
2169 return ret;
2170
c5c9cd4d 2171 src_file = fget(srcfd);
ab67b7c1
YZ		 2172 if (!src_file) {
2173 ret = -EBADF;
2174 goto out_drop_write;
2175 }
5dc64164 2176
f46b5a66
CH		 2177 src = src_file->f_dentry->d_inode;
2178
c5c9cd4d
SW		 2179 ret = -EINVAL;
2180 if (src == inode)
2181 goto out_fput;
2182
5dc64164
DR		 2183 /* the src must be open for reading */
2184 if (!(src_file->f_mode & FMODE_READ))
2185 goto out_fput;
2186
0e7b824c
LZ		 2187 /* don't make the dst file partly checksummed */
2188 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2189 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2190 goto out_fput;
2191
ae01a0ab
YZ		 2192 ret = -EISDIR;
2193 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2194 goto out_fput;
2195
f46b5a66 2196 ret = -EXDEV;
ae01a0ab
YZ		 2197 if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
2198 goto out_fput;
2199
2200 ret = -ENOMEM;
2201 buf = vmalloc(btrfs_level_size(root, 0));
2202 if (!buf)
2203 goto out_fput;
2204
2205 path = btrfs_alloc_path();
2206 if (!path) {
2207 vfree(buf);
f46b5a66 2208 goto out_fput;
ae01a0ab
YZ		 2209 }
2210 path->reada = 2;
f46b5a66
CH		 2211
2212 if (inode < src) {
fccdae43
SW		 2213 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2214 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
f46b5a66 2215 } else {
fccdae43
SW		 2216 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2217 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
f46b5a66
CH		 2218 }
2219
c5c9cd4d
SW		 2220 /* determine range to clone */
2221 ret = -EINVAL;
2ebc3464 2222 if (off + len > src->i_size || off + len < off)
f46b5a66 2223 goto out_unlock;
c5c9cd4d
SW		 2224 if (len == 0)
2225 olen = len = src->i_size - off;
2226 /* if we extend to eof, continue to block boundary */
2227 if (off + len == src->i_size)
2a6b8dae 2228 len = ALIGN(src->i_size, bs) - off;
c5c9cd4d
SW		 2229
2230 /* verify the end result is block aligned */
2a6b8dae
LZ		 2231 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2232 !IS_ALIGNED(destoff, bs))
c5c9cd4d
SW		 2233 goto out_unlock;
2234
d525e8ab
LZ		 2235 if (destoff > inode->i_size) {
2236 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2237 if (ret)
2238 goto out_unlock;
2239 }
2240
71ef0786
LZ		 2241 /* truncate page cache pages from target inode range */
2242 truncate_inode_pages_range(&inode->i_data, destoff,
2243 PAGE_CACHE_ALIGN(destoff + len) - 1);
2244
f46b5a66
CH		 2245 /* do any pending delalloc/csum calc on src, one way or
2246 another, and lock file content */
2247 while (1) {
31840ae1 2248 struct btrfs_ordered_extent *ordered;
c5c9cd4d 2249 lock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
9a019196
SW		 2250 ordered = btrfs_lookup_first_ordered_extent(src, off+len);
2251 if (!ordered &&
2252 !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len,
2253 EXTENT_DELALLOC, 0, NULL))
f46b5a66 2254 break;
c5c9cd4d 2255 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
ae01a0ab
YZ		 2256 if (ordered)
2257 btrfs_put_ordered_extent(ordered);
9a019196 2258 btrfs_wait_ordered_range(src, off, len);
f46b5a66
CH		 2259 }
2260
c5c9cd4d 2261 /* clone data */
33345d01 2262 key.objectid = btrfs_ino(src);
ae01a0ab
YZ		 2263 key.type = BTRFS_EXTENT_DATA_KEY;
2264 key.offset = 0;
f46b5a66
CH		 2265
2266 while (1) {
2267 /*
2268 * note the key will change type as we walk through the
2269 * tree.
2270 */
a22285a6 2271 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
f46b5a66
CH		 2272 if (ret < 0)
2273 goto out;
2274
ae01a0ab
YZ		 2275 nritems = btrfs_header_nritems(path->nodes[0]);
2276 if (path->slots[0] >= nritems) {
f46b5a66
CH		 2277 ret = btrfs_next_leaf(root, path);
2278 if (ret < 0)
2279 goto out;
2280 if (ret > 0)
2281 break;
ae01a0ab 2282 nritems = btrfs_header_nritems(path->nodes[0]);
f46b5a66
CH		 2283 }
2284 leaf = path->nodes[0];
2285 slot = path->slots[0];
f46b5a66 2286
ae01a0ab 2287 btrfs_item_key_to_cpu(leaf, &key, slot);
d20f7043 2288 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
33345d01 2289 key.objectid != btrfs_ino(src))
f46b5a66
CH		 2290 break;
2291
c5c9cd4d
SW		 2292 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2293 struct btrfs_file_extent_item *extent;
2294 int type;
31840ae1
ZY		 2295 u32 size;
2296 struct btrfs_key new_key;
c5c9cd4d
SW		 2297 u64 disko = 0, diskl = 0;
2298 u64 datao = 0, datal = 0;
2299 u8 comp;
b5384d48 2300 u64 endoff;
31840ae1
ZY		 2301
2302 size = btrfs_item_size_nr(leaf, slot);
2303 read_extent_buffer(leaf, buf,
2304 btrfs_item_ptr_offset(leaf, slot),
2305 size);
c5c9cd4d
SW		 2306
2307 extent = btrfs_item_ptr(leaf, slot,
2308 struct btrfs_file_extent_item);
2309 comp = btrfs_file_extent_compression(leaf, extent);
2310 type = btrfs_file_extent_type(leaf, extent);
c8a894d7
CM		 2311 if (type == BTRFS_FILE_EXTENT_REG ||
2312 type == BTRFS_FILE_EXTENT_PREALLOC) {
d397712b
CM		 2313 disko = btrfs_file_extent_disk_bytenr(leaf,
2314 extent);
2315 diskl = btrfs_file_extent_disk_num_bytes(leaf,
2316 extent);
c5c9cd4d 2317 datao = btrfs_file_extent_offset(leaf, extent);
d397712b
CM		 2318 datal = btrfs_file_extent_num_bytes(leaf,
2319 extent);
c5c9cd4d
SW		 2320 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2321 /* take upper bound, may be compressed */
2322 datal = btrfs_file_extent_ram_bytes(leaf,
2323 extent);
2324 }
b3b4aa74 2325 btrfs_release_path(path);
31840ae1 2326
050006a7 2327 if (key.offset + datal <= off ||
c5c9cd4d
SW		 2328 key.offset >= off+len)
2329 goto next;
2330
31840ae1 2331 memcpy(&new_key, &key, sizeof(new_key));
33345d01 2332 new_key.objectid = btrfs_ino(inode);
4d728ec7
LZ		 2333 if (off <= key.offset)
2334 new_key.offset = key.offset + destoff - off;
2335 else
2336 new_key.offset = destoff;
31840ae1 2337
b6f3409b
SW		 2338 /*
2339 * 1 - adjusting old extent (we may have to split it)
2340 * 1 - add new extent
2341 * 1 - inode update
2342 */
2343 trans = btrfs_start_transaction(root, 3);
a22285a6
YZ		 2344 if (IS_ERR(trans)) {
2345 ret = PTR_ERR(trans);
2346 goto out;
2347 }
2348
c8a894d7
CM		 2349 if (type == BTRFS_FILE_EXTENT_REG ||
2350 type == BTRFS_FILE_EXTENT_PREALLOC) {
d72c0842
LZ		 2351 /*
2352 * a | --- range to clone ---| b
2353 * | ------------- extent ------------- |
2354 */
2355
2356 /* substract range b */
2357 if (key.offset + datal > off + len)
2358 datal = off + len - key.offset;
2359
2360 /* substract range a */
a22285a6
YZ		 2361 if (off > key.offset) {
2362 datao += off - key.offset;
2363 datal -= off - key.offset;
2364 }
2365
a22285a6
YZ		 2366 ret = btrfs_drop_extents(trans, inode,
2367 new_key.offset,
2368 new_key.offset + datal,
2369 &hint_byte, 1);
2370 BUG_ON(ret);
2371
c5c9cd4d
SW		 2372 ret = btrfs_insert_empty_item(trans, root, path,
2373 &new_key, size);
a22285a6 2374 BUG_ON(ret);
c5c9cd4d
SW		 2375
2376 leaf = path->nodes[0];
2377 slot = path->slots[0];
2378 write_extent_buffer(leaf, buf,
31840ae1
ZY		 2379 btrfs_item_ptr_offset(leaf, slot),
2380 size);
ae01a0ab 2381
c5c9cd4d 2382 extent = btrfs_item_ptr(leaf, slot,
f46b5a66 2383 struct btrfs_file_extent_item);
c5c9cd4d 2384
c5c9cd4d
SW		 2385 /* disko == 0 means it's a hole */
2386 if (!disko)
2387 datao = 0;
c5c9cd4d
SW		 2388
2389 btrfs_set_file_extent_offset(leaf, extent,
2390 datao);
2391 btrfs_set_file_extent_num_bytes(leaf, extent,
2392 datal);
2393 if (disko) {
2394 inode_add_bytes(inode, datal);
ae01a0ab 2395 ret = btrfs_inc_extent_ref(trans, root,
5d4f98a2
YZ		 2396 disko, diskl, 0,
2397 root->root_key.objectid,
33345d01 2398 btrfs_ino(inode),
5d4f98a2 2399 new_key.offset - datao);
31840ae1 2400 BUG_ON(ret);
f46b5a66 2401 }
c5c9cd4d
SW		 2402 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2403 u64 skip = 0;
2404 u64 trim = 0;
2405 if (off > key.offset) {
2406 skip = off - key.offset;
2407 new_key.offset += skip;
2408 }
d397712b 2409
c5c9cd4d
SW		 2410 if (key.offset + datal > off+len)
2411 trim = key.offset + datal - (off+len);
d397712b 2412
c5c9cd4d 2413 if (comp && (skip || trim)) {
c5c9cd4d 2414 ret = -EINVAL;
a22285a6 2415 btrfs_end_transaction(trans, root);
c5c9cd4d
SW		 2416 goto out;
2417 }
2418 size -= skip + trim;
2419 datal -= skip + trim;
a22285a6
YZ		 2420
2421 ret = btrfs_drop_extents(trans, inode,
2422 new_key.offset,
2423 new_key.offset + datal,
2424 &hint_byte, 1);
2425 BUG_ON(ret);
2426
c5c9cd4d
SW		 2427 ret = btrfs_insert_empty_item(trans, root, path,
2428 &new_key, size);
a22285a6 2429 BUG_ON(ret);
c5c9cd4d
SW		 2430
2431 if (skip) {
d397712b
CM		 2432 u32 start =
2433 btrfs_file_extent_calc_inline_size(0);
c5c9cd4d
SW		 2434 memmove(buf+start, buf+start+skip,
2435 datal);
2436 }
2437
2438 leaf = path->nodes[0];
2439 slot = path->slots[0];
2440 write_extent_buffer(leaf, buf,
2441 btrfs_item_ptr_offset(leaf, slot),
2442 size);
2443 inode_add_bytes(inode, datal);
f46b5a66 2444 }
c5c9cd4d
SW		 2445
2446 btrfs_mark_buffer_dirty(leaf);
b3b4aa74 2447 btrfs_release_path(path);
c5c9cd4d 2448
a22285a6 2449 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
b5384d48
SW		 2450
2451 /*
2452 * we round up to the block size at eof when
2453 * determining which extents to clone above,
2454 * but shouldn't round up the file size
2455 */
2456 endoff = new_key.offset + datal;
5f3888ff
LZ		 2457 if (endoff > destoff+olen)
2458 endoff = destoff+olen;
b5384d48
SW		 2459 if (endoff > inode->i_size)
2460 btrfs_i_size_write(inode, endoff);
2461
a22285a6
YZ		 2462 ret = btrfs_update_inode(trans, root, inode);
2463 BUG_ON(ret);
2464 btrfs_end_transaction(trans, root);
2465 }
d397712b 2466next:
b3b4aa74 2467 btrfs_release_path(path);
f46b5a66 2468 key.offset++;
f46b5a66 2469 }
f46b5a66
CH		 2470 ret = 0;
2471out:
b3b4aa74 2472 btrfs_release_path(path);
c5c9cd4d 2473 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
f46b5a66
CH		 2474out_unlock:
2475 mutex_unlock(&src->i_mutex);
2476 mutex_unlock(&inode->i_mutex);
ae01a0ab
YZ		 2477 vfree(buf);
2478 btrfs_free_path(path);
f46b5a66
CH		 2479out_fput:
2480 fput(src_file);
ab67b7c1
YZ		 2481out_drop_write:
2482 mnt_drop_write(file->f_path.mnt);
f46b5a66
CH		 2483 return ret;
2484}
2485
7a865e8a 2486static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
c5c9cd4d
SW		 2487{
2488 struct btrfs_ioctl_clone_range_args args;
2489
7a865e8a 2490 if (copy_from_user(&args, argp, sizeof(args)))
c5c9cd4d
SW		 2491 return -EFAULT;
2492 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2493 args.src_length, args.dest_offset);
2494}
2495
f46b5a66
CH		 2496/*
2497 * there are many ways the trans_start and trans_end ioctls can lead
2498 * to deadlocks. They should only be used by applications that
2499 * basically own the machine, and have a very in depth understanding
2500 * of all the possible deadlocks and enospc problems.
2501 */
b2950863 2502static long btrfs_ioctl_trans_start(struct file *file)
f46b5a66
CH		 2503{
2504 struct inode *inode = fdentry(file)->d_inode;
2505 struct btrfs_root *root = BTRFS_I(inode)->root;
2506 struct btrfs_trans_handle *trans;
1ab86aed 2507 int ret;
f46b5a66 2508
1ab86aed 2509 ret = -EPERM;
df5b5520 2510 if (!capable(CAP_SYS_ADMIN))
1ab86aed 2511 goto out;
df5b5520 2512
1ab86aed
SW		 2513 ret = -EINPROGRESS;
2514 if (file->private_data)
f46b5a66 2515 goto out;
9ca9ee09 2516
b83cc969
LZ		 2517 ret = -EROFS;
2518 if (btrfs_root_readonly(root))
2519 goto out;
2520
c146afad
YZ		 2521 ret = mnt_want_write(file->f_path.mnt);
2522 if (ret)
2523 goto out;
2524
a4abeea4 2525 atomic_inc(&root->fs_info->open_ioctl_trans);
9ca9ee09 2526
1ab86aed 2527 ret = -ENOMEM;
7a7eaa40 2528 trans = btrfs_start_ioctl_transaction(root);
abd30bb0 2529 if (IS_ERR(trans))
1ab86aed
SW		 2530 goto out_drop;
2531
2532 file->private_data = trans;
2533 return 0;
2534
2535out_drop:
a4abeea4 2536 atomic_dec(&root->fs_info->open_ioctl_trans);
1ab86aed 2537 mnt_drop_write(file->f_path.mnt);
f46b5a66 2538out:
f46b5a66
CH		 2539 return ret;
2540}
2541
6ef5ed0d
JB		 2542static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2543{
2544 struct inode *inode = fdentry(file)->d_inode;
2545 struct btrfs_root *root = BTRFS_I(inode)->root;
2546 struct btrfs_root *new_root;
2547 struct btrfs_dir_item *di;
2548 struct btrfs_trans_handle *trans;
2549 struct btrfs_path *path;
2550 struct btrfs_key location;
2551 struct btrfs_disk_key disk_key;
2552 struct btrfs_super_block *disk_super;
2553 u64 features;
2554 u64 objectid = 0;
2555 u64 dir_id;
2556
2557 if (!capable(CAP_SYS_ADMIN))
2558 return -EPERM;
2559
2560 if (copy_from_user(&objectid, argp, sizeof(objectid)))
2561 return -EFAULT;
2562
2563 if (!objectid)
2564 objectid = root->root_key.objectid;
2565
2566 location.objectid = objectid;
2567 location.type = BTRFS_ROOT_ITEM_KEY;
2568 location.offset = (u64)-1;
2569
2570 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2571 if (IS_ERR(new_root))
2572 return PTR_ERR(new_root);
2573
2574 if (btrfs_root_refs(&new_root->root_item) == 0)
2575 return -ENOENT;
2576
2577 path = btrfs_alloc_path();
2578 if (!path)
2579 return -ENOMEM;
2580 path->leave_spinning = 1;
2581
2582 trans = btrfs_start_transaction(root, 1);
98d5dc13 2583 if (IS_ERR(trans)) {
6ef5ed0d 2584 btrfs_free_path(path);
98d5dc13 2585 return PTR_ERR(trans);
6ef5ed0d
JB		 2586 }
2587
2588 dir_id = btrfs_super_root_dir(&root->fs_info->super_copy);
2589 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2590 dir_id, "default", 7, 1);
cf1e99a4 2591 if (IS_ERR_OR_NULL(di)) {
6ef5ed0d
JB		 2592 btrfs_free_path(path);
2593 btrfs_end_transaction(trans, root);
2594 printk(KERN_ERR "Umm, you don't have the default dir item, "
2595 "this isn't going to work\n");
2596 return -ENOENT;
2597 }
2598
2599 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2600 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2601 btrfs_mark_buffer_dirty(path->nodes[0]);
2602 btrfs_free_path(path);
2603
2604 disk_super = &root->fs_info->super_copy;
2605 features = btrfs_super_incompat_flags(disk_super);
2606 if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) {
2607 features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL;
2608 btrfs_set_super_incompat_flags(disk_super, features);
2609 }
2610 btrfs_end_transaction(trans, root);
2611
2612 return 0;
2613}
2614
bf5fc093
JB		 2615static void get_block_group_info(struct list_head *groups_list,
2616 struct btrfs_ioctl_space_info *space)
2617{
2618 struct btrfs_block_group_cache *block_group;
2619
2620 space->total_bytes = 0;
2621 space->used_bytes = 0;
2622 space->flags = 0;
2623 list_for_each_entry(block_group, groups_list, list) {
2624 space->flags = block_group->flags;
2625 space->total_bytes += block_group->key.offset;
2626 space->used_bytes +=
2627 btrfs_block_group_used(&block_group->item);
2628 }
2629}
2630
1406e432
JB		 2631long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2632{
2633 struct btrfs_ioctl_space_args space_args;
2634 struct btrfs_ioctl_space_info space;
2635 struct btrfs_ioctl_space_info *dest;
7fde62bf 2636 struct btrfs_ioctl_space_info *dest_orig;
13f2696f 2637 struct btrfs_ioctl_space_info __user *user_dest;
1406e432 2638 struct btrfs_space_info *info;
bf5fc093
JB		 2639 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2640 BTRFS_BLOCK_GROUP_SYSTEM,
2641 BTRFS_BLOCK_GROUP_METADATA,
2642 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2643 int num_types = 4;
7fde62bf 2644 int alloc_size;
1406e432 2645 int ret = 0;
51788b1b 2646 u64 slot_count = 0;
bf5fc093 2647 int i, c;
1406e432
JB		 2648
2649 if (copy_from_user(&space_args,
2650 (struct btrfs_ioctl_space_args __user *)arg,
2651 sizeof(space_args)))
2652 return -EFAULT;
2653
bf5fc093
JB		 2654 for (i = 0; i < num_types; i++) {
2655 struct btrfs_space_info *tmp;
2656
2657 info = NULL;
2658 rcu_read_lock();
2659 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2660 list) {
2661 if (tmp->flags == types[i]) {
2662 info = tmp;
2663 break;
2664 }
2665 }
2666 rcu_read_unlock();
2667
2668 if (!info)
2669 continue;
2670
2671 down_read(&info->groups_sem);
2672 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2673 if (!list_empty(&info->block_groups[c]))
2674 slot_count++;
2675 }
2676 up_read(&info->groups_sem);
2677 }
7fde62bf
CM		 2678
2679 /* space_slots == 0 means they are asking for a count */
2680 if (space_args.space_slots == 0) {
2681 space_args.total_spaces = slot_count;
2682 goto out;
2683 }
bf5fc093 2684
51788b1b 2685 slot_count = min_t(u64, space_args.space_slots, slot_count);
bf5fc093 2686
7fde62bf 2687 alloc_size = sizeof(*dest) * slot_count;
bf5fc093 2688
7fde62bf
CM		 2689 /* we generally have at most 6 or so space infos, one for each raid
2690 * level. So, a whole page should be more than enough for everyone
2691 */
2692 if (alloc_size > PAGE_CACHE_SIZE)
2693 return -ENOMEM;
2694
1406e432 2695 space_args.total_spaces = 0;
7fde62bf
CM		 2696 dest = kmalloc(alloc_size, GFP_NOFS);
2697 if (!dest)
2698 return -ENOMEM;
2699 dest_orig = dest;
1406e432 2700
7fde62bf 2701 /* now we have a buffer to copy into */
bf5fc093
JB		 2702 for (i = 0; i < num_types; i++) {
2703 struct btrfs_space_info *tmp;
2704
51788b1b
DR		 2705 if (!slot_count)
2706 break;
2707
bf5fc093
JB		 2708 info = NULL;
2709 rcu_read_lock();
2710 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2711 list) {
2712 if (tmp->flags == types[i]) {
2713 info = tmp;
2714 break;
2715 }
2716 }
2717 rcu_read_unlock();
7fde62bf 2718
bf5fc093
JB		 2719 if (!info)
2720 continue;
2721 down_read(&info->groups_sem);
2722 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2723 if (!list_empty(&info->block_groups[c])) {
2724 get_block_group_info(&info->block_groups[c],
2725 &space);
2726 memcpy(dest, &space, sizeof(space));
2727 dest++;
2728 space_args.total_spaces++;
51788b1b 2729 slot_count--;
bf5fc093 2730 }
51788b1b
DR		 2731 if (!slot_count)
2732 break;
bf5fc093
JB		 2733 }
2734 up_read(&info->groups_sem);
1406e432 2735 }
1406e432 2736
7fde62bf
CM		 2737 user_dest = (struct btrfs_ioctl_space_info *)
2738 (arg + sizeof(struct btrfs_ioctl_space_args));
2739
2740 if (copy_to_user(user_dest, dest_orig, alloc_size))
2741 ret = -EFAULT;
2742
2743 kfree(dest_orig);
2744out:
2745 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
1406e432
JB		 2746 ret = -EFAULT;
2747
2748 return ret;
2749}
2750
f46b5a66
CH		 2751/*
2752 * there are many ways the trans_start and trans_end ioctls can lead
2753 * to deadlocks. They should only be used by applications that
2754 * basically own the machine, and have a very in depth understanding
2755 * of all the possible deadlocks and enospc problems.
2756 */
2757long btrfs_ioctl_trans_end(struct file *file)
2758{
2759 struct inode *inode = fdentry(file)->d_inode;
2760 struct btrfs_root *root = BTRFS_I(inode)->root;
2761 struct btrfs_trans_handle *trans;
f46b5a66 2762
f46b5a66 2763 trans = file->private_data;
1ab86aed
SW		 2764 if (!trans)
2765 return -EINVAL;
b214107e 2766 file->private_data = NULL;
9ca9ee09 2767
1ab86aed
SW		 2768 btrfs_end_transaction(trans, root);
2769
a4abeea4 2770 atomic_dec(&root->fs_info->open_ioctl_trans);
9ca9ee09 2771
cfc8ea87 2772 mnt_drop_write(file->f_path.mnt);
1ab86aed 2773 return 0;
f46b5a66
CH		 2774}
2775
46204592
SW		 2776static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
2777{
2778 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2779 struct btrfs_trans_handle *trans;
2780 u64 transid;
db5b493a 2781 int ret;
46204592
SW		 2782
2783 trans = btrfs_start_transaction(root, 0);
98d5dc13
TI		 2784 if (IS_ERR(trans))
2785 return PTR_ERR(trans);
46204592 2786 transid = trans->transid;
db5b493a 2787 ret = btrfs_commit_transaction_async(trans, root, 0);
8b2b2d3c
TI		 2788 if (ret) {
2789 btrfs_end_transaction(trans, root);
db5b493a 2790 return ret;
8b2b2d3c 2791 }
46204592
SW		 2792
2793 if (argp)
2794 if (copy_to_user(argp, &transid, sizeof(transid)))
2795 return -EFAULT;
2796 return 0;
2797}
2798
2799static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
2800{
2801 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2802 u64 transid;
2803
2804 if (argp) {
2805 if (copy_from_user(&transid, argp, sizeof(transid)))
2806 return -EFAULT;
2807 } else {
2808 transid = 0; /* current trans */
2809 }
2810 return btrfs_wait_for_commit(root, transid);
2811}
2812
475f6387
JS		 2813static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg)
2814{
2815 int ret;
2816 struct btrfs_ioctl_scrub_args *sa;
2817
2818 if (!capable(CAP_SYS_ADMIN))
2819 return -EPERM;
2820
2821 sa = memdup_user(arg, sizeof(*sa));
2822 if (IS_ERR(sa))
2823 return PTR_ERR(sa);
2824
2825 ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end,
8628764e 2826 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY);
475f6387
JS		 2827
2828 if (copy_to_user(arg, sa, sizeof(*sa)))
2829 ret = -EFAULT;
2830
2831 kfree(sa);
2832 return ret;
2833}
2834
2835static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
2836{
2837 if (!capable(CAP_SYS_ADMIN))
2838 return -EPERM;
2839
2840 return btrfs_scrub_cancel(root);
2841}
2842
2843static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
2844 void __user *arg)
2845{
2846 struct btrfs_ioctl_scrub_args *sa;
2847 int ret;
2848
2849 if (!capable(CAP_SYS_ADMIN))
2850 return -EPERM;
2851
2852 sa = memdup_user(arg, sizeof(*sa));
2853 if (IS_ERR(sa))
2854 return PTR_ERR(sa);
2855
2856 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
2857
2858 if (copy_to_user(arg, sa, sizeof(*sa)))
2859 ret = -EFAULT;
2860
2861 kfree(sa);
2862 return ret;
2863}
2864
f46b5a66
CH		 2865long btrfs_ioctl(struct file *file, unsigned int
2866 cmd, unsigned long arg)
2867{
2868 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
4bcabaa3 2869 void __user *argp = (void __user *)arg;
f46b5a66
CH		 2870
2871 switch (cmd) {
6cbff00f
CH		 2872 case FS_IOC_GETFLAGS:
2873 return btrfs_ioctl_getflags(file, argp);
2874 case FS_IOC_SETFLAGS:
2875 return btrfs_ioctl_setflags(file, argp);
2876 case FS_IOC_GETVERSION:
2877 return btrfs_ioctl_getversion(file, argp);
f7039b1d
LD		 2878 case FITRIM:
2879 return btrfs_ioctl_fitrim(file, argp);
f46b5a66 2880 case BTRFS_IOC_SNAP_CREATE:
fa0d2b9b 2881 return btrfs_ioctl_snap_create(file, argp, 0);
fdfb1e4f 2882 case BTRFS_IOC_SNAP_CREATE_V2:
fa0d2b9b 2883 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3de4586c 2884 case BTRFS_IOC_SUBVOL_CREATE:
fa0d2b9b 2885 return btrfs_ioctl_snap_create(file, argp, 1);
76dda93c
YZ		 2886 case BTRFS_IOC_SNAP_DESTROY:
2887 return btrfs_ioctl_snap_destroy(file, argp);
0caa102d
LZ		 2888 case BTRFS_IOC_SUBVOL_GETFLAGS:
2889 return btrfs_ioctl_subvol_getflags(file, argp);
2890 case BTRFS_IOC_SUBVOL_SETFLAGS:
2891 return btrfs_ioctl_subvol_setflags(file, argp);
6ef5ed0d
JB		 2892 case BTRFS_IOC_DEFAULT_SUBVOL:
2893 return btrfs_ioctl_default_subvol(file, argp);
f46b5a66 2894 case BTRFS_IOC_DEFRAG:
1e701a32
CM		 2895 return btrfs_ioctl_defrag(file, NULL);
2896 case BTRFS_IOC_DEFRAG_RANGE:
2897 return btrfs_ioctl_defrag(file, argp);
f46b5a66 2898 case BTRFS_IOC_RESIZE:
4bcabaa3 2899 return btrfs_ioctl_resize(root, argp);
f46b5a66 2900 case BTRFS_IOC_ADD_DEV:
4bcabaa3 2901 return btrfs_ioctl_add_dev(root, argp);
f46b5a66 2902 case BTRFS_IOC_RM_DEV:
4bcabaa3 2903 return btrfs_ioctl_rm_dev(root, argp);
475f6387
JS		 2904 case BTRFS_IOC_FS_INFO:
2905 return btrfs_ioctl_fs_info(root, argp);
2906 case BTRFS_IOC_DEV_INFO:
2907 return btrfs_ioctl_dev_info(root, argp);
f46b5a66
CH		 2908 case BTRFS_IOC_BALANCE:
2909 return btrfs_balance(root->fs_info->dev_root);
2910 case BTRFS_IOC_CLONE:
c5c9cd4d
SW		 2911 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
2912 case BTRFS_IOC_CLONE_RANGE:
7a865e8a 2913 return btrfs_ioctl_clone_range(file, argp);
f46b5a66
CH		 2914 case BTRFS_IOC_TRANS_START:
2915 return btrfs_ioctl_trans_start(file);
2916 case BTRFS_IOC_TRANS_END:
2917 return btrfs_ioctl_trans_end(file);
ac8e9819
CM		 2918 case BTRFS_IOC_TREE_SEARCH:
2919 return btrfs_ioctl_tree_search(file, argp);
2920 case BTRFS_IOC_INO_LOOKUP:
2921 return btrfs_ioctl_ino_lookup(file, argp);
1406e432
JB		 2922 case BTRFS_IOC_SPACE_INFO:
2923 return btrfs_ioctl_space_info(root, argp);
f46b5a66
CH		 2924 case BTRFS_IOC_SYNC:
2925 btrfs_sync_fs(file->f_dentry->d_sb, 1);
2926 return 0;
46204592
SW		 2927 case BTRFS_IOC_START_SYNC:
2928 return btrfs_ioctl_start_sync(file, argp);
2929 case BTRFS_IOC_WAIT_SYNC:
2930 return btrfs_ioctl_wait_sync(file, argp);
475f6387
JS		 2931 case BTRFS_IOC_SCRUB:
2932 return btrfs_ioctl_scrub(root, argp);
2933 case BTRFS_IOC_SCRUB_CANCEL:
2934 return btrfs_ioctl_scrub_cancel(root, argp);
2935 case BTRFS_IOC_SCRUB_PROGRESS:
2936 return btrfs_ioctl_scrub_progress(root, argp);
f46b5a66
CH		 2937 }
2938
2939 return -ENOTTY;
2940}
Michael's Linux tree.