]> git.ipfire.org Git - thirdparty/linux.git/blame - fs/btrfs/ioctl.c
projects / thirdparty / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Btrfs: calc file extent num_bytes correctly in file clone
[thirdparty/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 *
120 * Unlike extN we don't have any flags we don't want to inherit currently.
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
CH		 130
131 if (S_ISREG(inode->i_mode))
132 flags &= ~BTRFS_INODE_DIRSYNC;
133 else if (!S_ISDIR(inode->i_mode))
134 flags &= (BTRFS_INODE_NODUMP | BTRFS_INODE_NOATIME);
135
136 BTRFS_I(inode)->flags = flags;
137 btrfs_update_iflags(inode);
138}
139
140static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
141{
142 struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode);
143 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
144
145 if (copy_to_user(arg, &flags, sizeof(flags)))
146 return -EFAULT;
147 return 0;
148}
149
75e7cb7f
LB		 150static int check_flags(unsigned int flags)
151{
152 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
153 FS_NOATIME_FL | FS_NODUMP_FL | \
154 FS_SYNC_FL | FS_DIRSYNC_FL | \
e1e8fb6a
LZ		 155 FS_NOCOMP_FL | FS_COMPR_FL |
156 FS_NOCOW_FL))
75e7cb7f
LB		 157 return -EOPNOTSUPP;
158
159 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
160 return -EINVAL;
161
75e7cb7f
LB		 162 return 0;
163}
164
6cbff00f
CH		 165static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
166{
167 struct inode *inode = file->f_path.dentry->d_inode;
168 struct btrfs_inode *ip = BTRFS_I(inode);
169 struct btrfs_root *root = ip->root;
170 struct btrfs_trans_handle *trans;
171 unsigned int flags, oldflags;
172 int ret;
173
b83cc969
LZ		 174 if (btrfs_root_readonly(root))
175 return -EROFS;
176
6cbff00f
CH		 177 if (copy_from_user(&flags, arg, sizeof(flags)))
178 return -EFAULT;
179
75e7cb7f
LB		 180 ret = check_flags(flags);
181 if (ret)
182 return ret;
f46b5a66 183
2e149670 184 if (!inode_owner_or_capable(inode))
6cbff00f
CH		 185 return -EACCES;
186
187 mutex_lock(&inode->i_mutex);
188
189 flags = btrfs_mask_flags(inode->i_mode, flags);
190 oldflags = btrfs_flags_to_ioctl(ip->flags);
191 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
192 if (!capable(CAP_LINUX_IMMUTABLE)) {
193 ret = -EPERM;
194 goto out_unlock;
195 }
196 }
197
198 ret = mnt_want_write(file->f_path.mnt);
199 if (ret)
200 goto out_unlock;
201
202 if (flags & FS_SYNC_FL)
203 ip->flags |= BTRFS_INODE_SYNC;
204 else
205 ip->flags &= ~BTRFS_INODE_SYNC;
206 if (flags & FS_IMMUTABLE_FL)
207 ip->flags |= BTRFS_INODE_IMMUTABLE;
208 else
209 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
210 if (flags & FS_APPEND_FL)
211 ip->flags |= BTRFS_INODE_APPEND;
212 else
213 ip->flags &= ~BTRFS_INODE_APPEND;
214 if (flags & FS_NODUMP_FL)
215 ip->flags |= BTRFS_INODE_NODUMP;
216 else
217 ip->flags &= ~BTRFS_INODE_NODUMP;
218 if (flags & FS_NOATIME_FL)
219 ip->flags |= BTRFS_INODE_NOATIME;
220 else
221 ip->flags &= ~BTRFS_INODE_NOATIME;
222 if (flags & FS_DIRSYNC_FL)
223 ip->flags |= BTRFS_INODE_DIRSYNC;
224 else
225 ip->flags &= ~BTRFS_INODE_DIRSYNC;
e1e8fb6a
LZ		 226 if (flags & FS_NOCOW_FL)
227 ip->flags |= BTRFS_INODE_NODATACOW;
228 else
229 ip->flags &= ~BTRFS_INODE_NODATACOW;
6cbff00f 230
75e7cb7f
LB		 231 /*
232 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
233 * flag may be changed automatically if compression code won't make
234 * things smaller.
235 */
236 if (flags & FS_NOCOMP_FL) {
237 ip->flags &= ~BTRFS_INODE_COMPRESS;
238 ip->flags |= BTRFS_INODE_NOCOMPRESS;
239 } else if (flags & FS_COMPR_FL) {
240 ip->flags |= BTRFS_INODE_COMPRESS;
241 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
ebcb904d
LZ		 242 } else {
243 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
75e7cb7f 244 }
6cbff00f 245
7a7eaa40 246 trans = btrfs_join_transaction(root);
3612b495 247 BUG_ON(IS_ERR(trans));
6cbff00f
CH		 248
249 ret = btrfs_update_inode(trans, root, inode);
250 BUG_ON(ret);
251
252 btrfs_update_iflags(inode);
253 inode->i_ctime = CURRENT_TIME;
254 btrfs_end_transaction(trans, root);
255
256 mnt_drop_write(file->f_path.mnt);
2d4e6f6a 257
258 ret = 0;
6cbff00f
CH		 259 out_unlock:
260 mutex_unlock(&inode->i_mutex);
2d4e6f6a 261 return ret;
6cbff00f
CH		 262}
263
264static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
265{
266 struct inode *inode = file->f_path.dentry->d_inode;
267
268 return put_user(inode->i_generation, arg);
269}
f46b5a66 270
f7039b1d
LD		 271static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
272{
273 struct btrfs_root *root = fdentry(file)->d_sb->s_fs_info;
274 struct btrfs_fs_info *fs_info = root->fs_info;
275 struct btrfs_device *device;
276 struct request_queue *q;
277 struct fstrim_range range;
278 u64 minlen = ULLONG_MAX;
279 u64 num_devices = 0;
280 int ret;
281
282 if (!capable(CAP_SYS_ADMIN))
283 return -EPERM;
284
1f78160c
XG		 285 rcu_read_lock();
286 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
287 dev_list) {
f7039b1d
LD		 288 if (!device->bdev)
289 continue;
290 q = bdev_get_queue(device->bdev);
291 if (blk_queue_discard(q)) {
292 num_devices++;
293 minlen = min((u64)q->limits.discard_granularity,
294 minlen);
295 }
296 }
1f78160c 297 rcu_read_unlock();
f7039b1d
LD		 298 if (!num_devices)
299 return -EOPNOTSUPP;
300
301 if (copy_from_user(&range, arg, sizeof(range)))
302 return -EFAULT;
303
304 range.minlen = max(range.minlen, minlen);
305 ret = btrfs_trim_fs(root, &range);
306 if (ret < 0)
307 return ret;
308
309 if (copy_to_user(arg, &range, sizeof(range)))
310 return -EFAULT;
311
312 return 0;
313}
314
cb8e7090
CH		 315static noinline int create_subvol(struct btrfs_root *root,
316 struct dentry *dentry,
72fd032e
SW		 317 char *name, int namelen,
318 u64 *async_transid)
f46b5a66
CH		 319{
320 struct btrfs_trans_handle *trans;
321 struct btrfs_key key;
322 struct btrfs_root_item root_item;
323 struct btrfs_inode_item *inode_item;
324 struct extent_buffer *leaf;
76dda93c 325 struct btrfs_root *new_root;
6a912213
JB		 326 struct dentry *parent = dget_parent(dentry);
327 struct inode *dir;
f46b5a66
CH		 328 int ret;
329 int err;
330 u64 objectid;
331 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
3de4586c 332 u64 index = 0;
f46b5a66 333
581bb050 334 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
6a912213
JB		 335 if (ret) {
336 dput(parent);
a22285a6 337 return ret;
6a912213
JB		 338 }
339
340 dir = parent->d_inode;
341
9ed74f2d
JB		 342 /*
343 * 1 - inode item
344 * 2 - refs
345 * 1 - root item
346 * 2 - dir items
347 */
a22285a6 348 trans = btrfs_start_transaction(root, 6);
6a912213
JB		 349 if (IS_ERR(trans)) {
350 dput(parent);
a22285a6 351 return PTR_ERR(trans);
6a912213 352 }
f46b5a66 353
5d4f98a2
YZ		 354 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
355 0, objectid, NULL, 0, 0, 0);
8e8a1e31
JB		 356 if (IS_ERR(leaf)) {
357 ret = PTR_ERR(leaf);
358 goto fail;
359 }
f46b5a66 360
5d4f98a2 361 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
f46b5a66
CH		 362 btrfs_set_header_bytenr(leaf, leaf->start);
363 btrfs_set_header_generation(leaf, trans->transid);
5d4f98a2 364 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
f46b5a66
CH		 365 btrfs_set_header_owner(leaf, objectid);
366
367 write_extent_buffer(leaf, root->fs_info->fsid,
368 (unsigned long)btrfs_header_fsid(leaf),
369 BTRFS_FSID_SIZE);
5d4f98a2
YZ		 370 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
371 (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
372 BTRFS_UUID_SIZE);
f46b5a66
CH		 373 btrfs_mark_buffer_dirty(leaf);
374
375 inode_item = &root_item.inode;
376 memset(inode_item, 0, sizeof(*inode_item));
377 inode_item->generation = cpu_to_le64(1);
378 inode_item->size = cpu_to_le64(3);
379 inode_item->nlink = cpu_to_le32(1);
a76a3cd4 380 inode_item->nbytes = cpu_to_le64(root->leafsize);
f46b5a66
CH		 381 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
382
08fe4db1
LZ		 383 root_item.flags = 0;
384 root_item.byte_limit = 0;
385 inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
386
f46b5a66 387 btrfs_set_root_bytenr(&root_item, leaf->start);
84234f3a 388 btrfs_set_root_generation(&root_item, trans->transid);
f46b5a66
CH		 389 btrfs_set_root_level(&root_item, 0);
390 btrfs_set_root_refs(&root_item, 1);
86b9f2ec 391 btrfs_set_root_used(&root_item, leaf->len);
80ff3856 392 btrfs_set_root_last_snapshot(&root_item, 0);
f46b5a66
CH		 393
394 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
395 root_item.drop_level = 0;
396
925baedd 397 btrfs_tree_unlock(leaf);
f46b5a66
CH		 398 free_extent_buffer(leaf);
399 leaf = NULL;
400
401 btrfs_set_root_dirid(&root_item, new_dirid);
402
403 key.objectid = objectid;
5d4f98a2 404 key.offset = 0;
f46b5a66
CH		 405 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
406 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
407 &root_item);
408 if (ret)
409 goto fail;
410
76dda93c
YZ		 411 key.offset = (u64)-1;
412 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
413 BUG_ON(IS_ERR(new_root));
414
415 btrfs_record_root_in_trans(trans, new_root);
416
d82a6f1d 417 ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
f46b5a66
CH		 418 /*
419 * insert the directory item
420 */
3de4586c
CM		 421 ret = btrfs_set_inode_index(dir, &index);
422 BUG_ON(ret);
423
424 ret = btrfs_insert_dir_item(trans, root,
16cdcec7 425 name, namelen, dir, &key,
3de4586c 426 BTRFS_FT_DIR, index);
f46b5a66
CH		 427 if (ret)
428 goto fail;
0660b5af 429
52c26179
YZ		 430 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
431 ret = btrfs_update_inode(trans, root, dir);
432 BUG_ON(ret);
433
0660b5af 434 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
4df27c4d 435 objectid, root->root_key.objectid,
33345d01 436 btrfs_ino(dir), index, name, namelen);
0660b5af 437
76dda93c 438 BUG_ON(ret);
f46b5a66 439
76dda93c 440 d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
f46b5a66 441fail:
6a912213 442 dput(parent);
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;
6a912213 459 struct dentry *parent;
f46b5a66
CH		 460 struct btrfs_pending_snapshot *pending_snapshot;
461 struct btrfs_trans_handle *trans;
2e4bfab9 462 int ret;
f46b5a66
CH		 463
464 if (!root->ref_cows)
465 return -EINVAL;
466
3de4586c 467 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
a22285a6
YZ		 468 if (!pending_snapshot)
469 return -ENOMEM;
470
471 btrfs_init_block_rsv(&pending_snapshot->block_rsv);
3de4586c 472 pending_snapshot->dentry = dentry;
f46b5a66 473 pending_snapshot->root = root;
b83cc969 474 pending_snapshot->readonly = readonly;
a22285a6
YZ		 475
476 trans = btrfs_start_transaction(root->fs_info->extent_root, 5);
477 if (IS_ERR(trans)) {
478 ret = PTR_ERR(trans);
479 goto fail;
480 }
481
482 ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
483 BUG_ON(ret);
484
8351583e 485 spin_lock(&root->fs_info->trans_lock);
f46b5a66
CH		 486 list_add(&pending_snapshot->list,
487 &trans->transaction->pending_snapshots);
8351583e 488 spin_unlock(&root->fs_info->trans_lock);
72fd032e
SW		 489 if (async_transid) {
490 *async_transid = trans->transid;
491 ret = btrfs_commit_transaction_async(trans,
492 root->fs_info->extent_root, 1);
493 } else {
494 ret = btrfs_commit_transaction(trans,
495 root->fs_info->extent_root);
496 }
2e4bfab9 497 BUG_ON(ret);
a22285a6
YZ		 498
499 ret = pending_snapshot->error;
500 if (ret)
501 goto fail;
502
66b4ffd1
JB		 503 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
504 if (ret)
505 goto fail;
f46b5a66 506
6a912213
JB		 507 parent = dget_parent(dentry);
508 inode = btrfs_lookup_dentry(parent->d_inode, dentry);
509 dput(parent);
2e4bfab9
YZ		 510 if (IS_ERR(inode)) {
511 ret = PTR_ERR(inode);
512 goto fail;
513 }
514 BUG_ON(!inode);
515 d_instantiate(dentry, inode);
516 ret = 0;
517fail:
a22285a6 518 kfree(pending_snapshot);
f46b5a66
CH		 519 return ret;
520}
521
4260f7c7
SW		 522/* copy of check_sticky in fs/namei.c()
523* It's inline, so penalty for filesystems that don't use sticky bit is
524* minimal.
525*/
526static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
527{
528 uid_t fsuid = current_fsuid();
529
530 if (!(dir->i_mode & S_ISVTX))
531 return 0;
532 if (inode->i_uid == fsuid)
533 return 0;
534 if (dir->i_uid == fsuid)
535 return 0;
536 return !capable(CAP_FOWNER);
537}
538
539/* copy of may_delete in fs/namei.c()
540 * Check whether we can remove a link victim from directory dir, check
541 * whether the type of victim is right.
542 * 1. We can't do it if dir is read-only (done in permission())
543 * 2. We should have write and exec permissions on dir
544 * 3. We can't remove anything from append-only dir
545 * 4. We can't do anything with immutable dir (done in permission())
546 * 5. If the sticky bit on dir is set we should either
547 * a. be owner of dir, or
548 * b. be owner of victim, or
549 * c. have CAP_FOWNER capability
550 * 6. If the victim is append-only or immutable we can't do antyhing with
551 * links pointing to it.
552 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
553 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
554 * 9. We can't remove a root or mountpoint.
555 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
556 * nfs_async_unlink().
557 */
558
559static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
560{
561 int error;
562
563 if (!victim->d_inode)
564 return -ENOENT;
565
566 BUG_ON(victim->d_parent->d_inode != dir);
567 audit_inode_child(victim, dir);
568
569 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
570 if (error)
571 return error;
572 if (IS_APPEND(dir))
573 return -EPERM;
574 if (btrfs_check_sticky(dir, victim->d_inode)||
575 IS_APPEND(victim->d_inode)||
576 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
577 return -EPERM;
578 if (isdir) {
579 if (!S_ISDIR(victim->d_inode->i_mode))
580 return -ENOTDIR;
581 if (IS_ROOT(victim))
582 return -EBUSY;
583 } else if (S_ISDIR(victim->d_inode->i_mode))
584 return -EISDIR;
585 if (IS_DEADDIR(dir))
586 return -ENOENT;
587 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
588 return -EBUSY;
589 return 0;
590}
591
cb8e7090
CH		 592/* copy of may_create in fs/namei.c() */
593static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
594{
595 if (child->d_inode)
596 return -EEXIST;
597 if (IS_DEADDIR(dir))
598 return -ENOENT;
599 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
600}
601
602/*
603 * Create a new subvolume below @parent. This is largely modeled after
604 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
605 * inside this filesystem so it's quite a bit simpler.
606 */
76dda93c
YZ		 607static noinline int btrfs_mksubvol(struct path *parent,
608 char *name, int namelen,
72fd032e 609 struct btrfs_root *snap_src,
b83cc969 610 u64 *async_transid, bool readonly)
cb8e7090 611{
76dda93c 612 struct inode *dir = parent->dentry->d_inode;
cb8e7090
CH		 613 struct dentry *dentry;
614 int error;
615
76dda93c 616 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
cb8e7090
CH		 617
618 dentry = lookup_one_len(name, parent->dentry, namelen);
619 error = PTR_ERR(dentry);
620 if (IS_ERR(dentry))
621 goto out_unlock;
622
623 error = -EEXIST;
624 if (dentry->d_inode)
625 goto out_dput;
626
cb8e7090
CH		 627 error = mnt_want_write(parent->mnt);
628 if (error)
629 goto out_dput;
630
76dda93c 631 error = btrfs_may_create(dir, dentry);
cb8e7090
CH		 632 if (error)
633 goto out_drop_write;
634
76dda93c
YZ		 635 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
636
637 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
638 goto out_up_read;
639
3de4586c 640 if (snap_src) {
72fd032e 641 error = create_snapshot(snap_src, dentry,
b83cc969 642 name, namelen, async_transid, readonly);
3de4586c 643 } else {
76dda93c 644 error = create_subvol(BTRFS_I(dir)->root, dentry,
72fd032e 645 name, namelen, async_transid);
3de4586c 646 }
76dda93c
YZ		 647 if (!error)
648 fsnotify_mkdir(dir, dentry);
649out_up_read:
650 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
cb8e7090
CH		 651out_drop_write:
652 mnt_drop_write(parent->mnt);
653out_dput:
654 dput(dentry);
655out_unlock:
76dda93c 656 mutex_unlock(&dir->i_mutex);
cb8e7090
CH		 657 return error;
658}
659
4cb5300b
CM		 660/*
661 * When we're defragging a range, we don't want to kick it off again
662 * if it is really just waiting for delalloc to send it down.
663 * If we find a nice big extent or delalloc range for the bytes in the
664 * file you want to defrag, we return 0 to let you know to skip this
665 * part of the file
666 */
667static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
668{
669 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
670 struct extent_map *em = NULL;
671 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
672 u64 end;
673
674 read_lock(&em_tree->lock);
675 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
676 read_unlock(&em_tree->lock);
677
678 if (em) {
679 end = extent_map_end(em);
680 free_extent_map(em);
681 if (end - offset > thresh)
682 return 0;
683 }
684 /* if we already have a nice delalloc here, just stop */
685 thresh /= 2;
686 end = count_range_bits(io_tree, &offset, offset + thresh,
687 thresh, EXTENT_DELALLOC, 1);
688 if (end >= thresh)
689 return 0;
690 return 1;
691}
692
693/*
694 * helper function to walk through a file and find extents
695 * newer than a specific transid, and smaller than thresh.
696 *
697 * This is used by the defragging code to find new and small
698 * extents
699 */
700static int find_new_extents(struct btrfs_root *root,
701 struct inode *inode, u64 newer_than,
702 u64 *off, int thresh)
703{
704 struct btrfs_path *path;
705 struct btrfs_key min_key;
706 struct btrfs_key max_key;
707 struct extent_buffer *leaf;
708 struct btrfs_file_extent_item *extent;
709 int type;
710 int ret;
a4689d2b 711 u64 ino = btrfs_ino(inode);
4cb5300b
CM		 712
713 path = btrfs_alloc_path();
714 if (!path)
715 return -ENOMEM;
716
a4689d2b 717 min_key.objectid = ino;
4cb5300b
CM		 718 min_key.type = BTRFS_EXTENT_DATA_KEY;
719 min_key.offset = *off;
720
a4689d2b 721 max_key.objectid = ino;
4cb5300b
CM		 722 max_key.type = (u8)-1;
723 max_key.offset = (u64)-1;
724
725 path->keep_locks = 1;
726
727 while(1) {
728 ret = btrfs_search_forward(root, &min_key, &max_key,
729 path, 0, newer_than);
730 if (ret != 0)
731 goto none;
a4689d2b 732 if (min_key.objectid != ino)
4cb5300b
CM		 733 goto none;
734 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
735 goto none;
736
737 leaf = path->nodes[0];
738 extent = btrfs_item_ptr(leaf, path->slots[0],
739 struct btrfs_file_extent_item);
740
741 type = btrfs_file_extent_type(leaf, extent);
742 if (type == BTRFS_FILE_EXTENT_REG &&
743 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
744 check_defrag_in_cache(inode, min_key.offset, thresh)) {
745 *off = min_key.offset;
746 btrfs_free_path(path);
747 return 0;
748 }
749
750 if (min_key.offset == (u64)-1)
751 goto none;
752
753 min_key.offset++;
754 btrfs_release_path(path);
755 }
756none:
757 btrfs_free_path(path);
758 return -ENOENT;
759}
760
940100a4 761static int should_defrag_range(struct inode *inode, u64 start, u64 len,
1e701a32
CM		 762 int thresh, u64 *last_len, u64 *skip,
763 u64 *defrag_end)
940100a4
CM		 764{
765 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
766 struct extent_map *em = NULL;
767 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
768 int ret = 1;
769
770 /*
771 * make sure that once we start defragging and extent, we keep on
772 * defragging it
773 */
774 if (start < *defrag_end)
775 return 1;
776
777 *skip = 0;
778
779 /*
780 * hopefully we have this extent in the tree already, try without
781 * the full extent lock
782 */
783 read_lock(&em_tree->lock);
784 em = lookup_extent_mapping(em_tree, start, len);
785 read_unlock(&em_tree->lock);
786
787 if (!em) {
788 /* get the big lock and read metadata off disk */
789 lock_extent(io_tree, start, start + len - 1, GFP_NOFS);
790 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
791 unlock_extent(io_tree, start, start + len - 1, GFP_NOFS);
792
6cf8bfbf 793 if (IS_ERR(em))
940100a4
CM		 794 return 0;
795 }
796
797 /* this will cover holes, and inline extents */
798 if (em->block_start >= EXTENT_MAP_LAST_BYTE)
799 ret = 0;
800
801 /*
802 * we hit a real extent, if it is big don't bother defragging it again
803 */
1e701a32 804 if ((*last_len == 0 || *last_len >= thresh) && em->len >= thresh)
940100a4
CM		 805 ret = 0;
806
807 /*
808 * last_len ends up being a counter of how many bytes we've defragged.
809 * every time we choose not to defrag an extent, we reset *last_len
810 * so that the next tiny extent will force a defrag.
811 *
812 * The end result of this is that tiny extents before a single big
813 * extent will force at least part of that big extent to be defragged.
814 */
815 if (ret) {
816 *last_len += len;
817 *defrag_end = extent_map_end(em);
818 } else {
819 *last_len = 0;
820 *skip = extent_map_end(em);
821 *defrag_end = 0;
822 }
823
824 free_extent_map(em);
825 return ret;
826}
827
4cb5300b
CM		 828/*
829 * it doesn't do much good to defrag one or two pages
830 * at a time. This pulls in a nice chunk of pages
831 * to COW and defrag.
832 *
833 * It also makes sure the delalloc code has enough
834 * dirty data to avoid making new small extents as part
835 * of the defrag
836 *
837 * It's a good idea to start RA on this range
838 * before calling this.
839 */
840static int cluster_pages_for_defrag(struct inode *inode,
841 struct page **pages,
842 unsigned long start_index,
843 int num_pages)
f46b5a66 844{
4cb5300b
CM		 845 unsigned long file_end;
846 u64 isize = i_size_read(inode);
847 u64 page_start;
848 u64 page_end;
849 int ret;
850 int i;
851 int i_done;
3eaa2885 852 struct btrfs_ordered_extent *ordered;
4cb5300b
CM		 853 struct extent_state *cached_state = NULL;
854
855 if (isize == 0)
856 return 0;
857 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
858
859 ret = btrfs_delalloc_reserve_space(inode,
860 num_pages << PAGE_CACHE_SHIFT);
861 if (ret)
862 return ret;
863again:
864 ret = 0;
865 i_done = 0;
866
867 /* step one, lock all the pages */
868 for (i = 0; i < num_pages; i++) {
869 struct page *page;
a94733d0
JB		 870 page = find_or_create_page(inode->i_mapping,
871 start_index + i, GFP_NOFS);
4cb5300b
CM		 872 if (!page)
873 break;
874
875 if (!PageUptodate(page)) {
876 btrfs_readpage(NULL, page);
877 lock_page(page);
878 if (!PageUptodate(page)) {
879 unlock_page(page);
880 page_cache_release(page);
881 ret = -EIO;
882 break;
883 }
884 }
885 isize = i_size_read(inode);
886 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
887 if (!isize || page->index > file_end ||
888 page->mapping != inode->i_mapping) {
889 /* whoops, we blew past eof, skip this page */
890 unlock_page(page);
891 page_cache_release(page);
892 break;
893 }
894 pages[i] = page;
895 i_done++;
896 }
897 if (!i_done || ret)
898 goto out;
899
900 if (!(inode->i_sb->s_flags & MS_ACTIVE))
901 goto out;
902
903 /*
904 * so now we have a nice long stream of locked
905 * and up to date pages, lets wait on them
906 */
907 for (i = 0; i < i_done; i++)
908 wait_on_page_writeback(pages[i]);
909
910 page_start = page_offset(pages[0]);
911 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
912
913 lock_extent_bits(&BTRFS_I(inode)->io_tree,
914 page_start, page_end - 1, 0, &cached_state,
915 GFP_NOFS);
916 ordered = btrfs_lookup_first_ordered_extent(inode, page_end - 1);
917 if (ordered &&
918 ordered->file_offset + ordered->len > page_start &&
919 ordered->file_offset < page_end) {
920 btrfs_put_ordered_extent(ordered);
921 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
922 page_start, page_end - 1,
923 &cached_state, GFP_NOFS);
924 for (i = 0; i < i_done; i++) {
925 unlock_page(pages[i]);
926 page_cache_release(pages[i]);
927 }
928 btrfs_wait_ordered_range(inode, page_start,
929 page_end - page_start);
930 goto again;
931 }
932 if (ordered)
933 btrfs_put_ordered_extent(ordered);
934
935 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
936 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
937 EXTENT_DO_ACCOUNTING, 0, 0, &cached_state,
938 GFP_NOFS);
939
940 if (i_done != num_pages) {
9e0baf60
JB		 941 spin_lock(&BTRFS_I(inode)->lock);
942 BTRFS_I(inode)->outstanding_extents++;
943 spin_unlock(&BTRFS_I(inode)->lock);
4cb5300b
CM		 944 btrfs_delalloc_release_space(inode,
945 (num_pages - i_done) << PAGE_CACHE_SHIFT);
946 }
947
948
949 btrfs_set_extent_delalloc(inode, page_start, page_end - 1,
950 &cached_state);
951
952 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
953 page_start, page_end - 1, &cached_state,
954 GFP_NOFS);
955
956 for (i = 0; i < i_done; i++) {
957 clear_page_dirty_for_io(pages[i]);
958 ClearPageChecked(pages[i]);
959 set_page_extent_mapped(pages[i]);
960 set_page_dirty(pages[i]);
961 unlock_page(pages[i]);
962 page_cache_release(pages[i]);
963 }
964 return i_done;
965out:
966 for (i = 0; i < i_done; i++) {
967 unlock_page(pages[i]);
968 page_cache_release(pages[i]);
969 }
970 btrfs_delalloc_release_space(inode, num_pages << PAGE_CACHE_SHIFT);
971 return ret;
972
973}
974
975int btrfs_defrag_file(struct inode *inode, struct file *file,
976 struct btrfs_ioctl_defrag_range_args *range,
977 u64 newer_than, unsigned long max_to_defrag)
978{
979 struct btrfs_root *root = BTRFS_I(inode)->root;
1a419d85 980 struct btrfs_super_block *disk_super;
4cb5300b 981 struct file_ra_state *ra = NULL;
f46b5a66 982 unsigned long last_index;
1a419d85 983 u64 features;
940100a4
CM		 984 u64 last_len = 0;
985 u64 skip = 0;
986 u64 defrag_end = 0;
4cb5300b
CM		 987 u64 newer_off = range->start;
988 int newer_left = 0;
f46b5a66
CH		 989 unsigned long i;
990 int ret;
4cb5300b 991 int defrag_count = 0;
1a419d85 992 int compress_type = BTRFS_COMPRESS_ZLIB;
4cb5300b
CM		 993 int extent_thresh = range->extent_thresh;
994 int newer_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
995 u64 new_align = ~((u64)128 * 1024 - 1);
996 struct page **pages = NULL;
997
998 if (extent_thresh == 0)
999 extent_thresh = 256 * 1024;
1a419d85
LZ		 1000
1001 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1002 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1003 return -EINVAL;
1004 if (range->compress_type)
1005 compress_type = range->compress_type;
1006 }
f46b5a66 1007
940100a4
CM		 1008 if (inode->i_size == 0)
1009 return 0;
1010
4cb5300b
CM		 1011 /*
1012 * if we were not given a file, allocate a readahead
1013 * context
1014 */
1015 if (!file) {
1016 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1017 if (!ra)
1018 return -ENOMEM;
1019 file_ra_state_init(ra, inode->i_mapping);
1020 } else {
1021 ra = &file->f_ra;
1022 }
1023
1024 pages = kmalloc(sizeof(struct page *) * newer_cluster,
1025 GFP_NOFS);
1026 if (!pages) {
1027 ret = -ENOMEM;
1028 goto out_ra;
1029 }
1030
1031 /* find the last page to defrag */
1e701a32
CM		 1032 if (range->start + range->len > range->start) {
1033 last_index = min_t(u64, inode->i_size - 1,
1034 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1035 } else {
1036 last_index = (inode->i_size - 1) >> PAGE_CACHE_SHIFT;
1037 }
1038
4cb5300b
CM		 1039 if (newer_than) {
1040 ret = find_new_extents(root, inode, newer_than,
1041 &newer_off, 64 * 1024);
1042 if (!ret) {
1043 range->start = newer_off;
1044 /*
1045 * we always align our defrag to help keep
1046 * the extents in the file evenly spaced
1047 */
1048 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1049 newer_left = newer_cluster;
1050 } else
1051 goto out_ra;
1052 } else {
1053 i = range->start >> PAGE_CACHE_SHIFT;
1054 }
1055 if (!max_to_defrag)
1056 max_to_defrag = last_index - 1;
1057
1058 while (i <= last_index && defrag_count < max_to_defrag) {
1059 /*
1060 * make sure we stop running if someone unmounts
1061 * the FS
1062 */
1063 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1064 break;
1065
1066 if (!newer_than &&
1067 !should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1e701a32 1068 PAGE_CACHE_SIZE,
4cb5300b 1069 extent_thresh,
1e701a32 1070 &last_len, &skip,
940100a4
CM		 1071 &defrag_end)) {
1072 unsigned long next;
1073 /*
1074 * the should_defrag function tells us how much to skip
1075 * bump our counter by the suggested amount
1076 */
1077 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1078 i = max(i + 1, next);
1079 continue;
1080 }
1e701a32 1081 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1a419d85 1082 BTRFS_I(inode)->force_compress = compress_type;
940100a4 1083
4cb5300b 1084 btrfs_force_ra(inode->i_mapping, ra, file, i, newer_cluster);
940100a4 1085
4cb5300b
CM		 1086 ret = cluster_pages_for_defrag(inode, pages, i, newer_cluster);
1087 if (ret < 0)
1088 goto out_ra;
1089
1090 defrag_count += ret;
1091 balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret);
1092 i += ret;
1093
1094 if (newer_than) {
1095 if (newer_off == (u64)-1)
1096 break;
1097
1098 newer_off = max(newer_off + 1,
1099 (u64)i << PAGE_CACHE_SHIFT);
1100
1101 ret = find_new_extents(root, inode,
1102 newer_than, &newer_off,
1103 64 * 1024);
1104 if (!ret) {
1105 range->start = newer_off;
1106 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1107 newer_left = newer_cluster;
1108 } else {
1109 break;
f46b5a66 1110 }
4cb5300b
CM		 1111 } else {
1112 i++;
f46b5a66 1113 }
f46b5a66
CH		 1114 }
1115
1e701a32
CM		 1116 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1117 filemap_flush(inode->i_mapping);
1118
1119 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1120 /* the filemap_flush will queue IO into the worker threads, but
1121 * we have to make sure the IO is actually started and that
1122 * ordered extents get created before we return
1123 */
1124 atomic_inc(&root->fs_info->async_submit_draining);
1125 while (atomic_read(&root->fs_info->nr_async_submits) ||
1126 atomic_read(&root->fs_info->async_delalloc_pages)) {
1127 wait_event(root->fs_info->async_submit_wait,
1128 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1129 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1130 }
1131 atomic_dec(&root->fs_info->async_submit_draining);
1132
1133 mutex_lock(&inode->i_mutex);
261507a0 1134 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1e701a32
CM		 1135 mutex_unlock(&inode->i_mutex);
1136 }
1137
1a419d85
LZ		 1138 disk_super = &root->fs_info->super_copy;
1139 features = btrfs_super_incompat_flags(disk_super);
1140 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1141 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
1142 btrfs_set_super_incompat_flags(disk_super, features);
1143 }
1144
4cb5300b
CM		 1145 if (!file)
1146 kfree(ra);
1147 return defrag_count;
940100a4 1148
4cb5300b
CM		 1149out_ra:
1150 if (!file)
1151 kfree(ra);
1152 kfree(pages);
940100a4 1153 return ret;
f46b5a66
CH		 1154}
1155
76dda93c
YZ		 1156static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
1157 void __user *arg)
f46b5a66
CH		 1158{
1159 u64 new_size;
1160 u64 old_size;
1161 u64 devid = 1;
1162 struct btrfs_ioctl_vol_args *vol_args;
1163 struct btrfs_trans_handle *trans;
1164 struct btrfs_device *device = NULL;
1165 char *sizestr;
1166 char *devstr = NULL;
1167 int ret = 0;
f46b5a66
CH		 1168 int mod = 0;
1169
c146afad
YZ		 1170 if (root->fs_info->sb->s_flags & MS_RDONLY)
1171 return -EROFS;
1172
e441d54d
CM		 1173 if (!capable(CAP_SYS_ADMIN))
1174 return -EPERM;
1175
dae7b665
LZ		 1176 vol_args = memdup_user(arg, sizeof(*vol_args));
1177 if (IS_ERR(vol_args))
1178 return PTR_ERR(vol_args);
5516e595
MF		 1179
1180 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 1181
7d9eb12c 1182 mutex_lock(&root->fs_info->volume_mutex);
f46b5a66
CH		 1183 sizestr = vol_args->name;
1184 devstr = strchr(sizestr, ':');
1185 if (devstr) {
1186 char *end;
1187 sizestr = devstr + 1;
1188 *devstr = '\0';
1189 devstr = vol_args->name;
1190 devid = simple_strtoull(devstr, &end, 10);
21380931
JB		 1191 printk(KERN_INFO "resizing devid %llu\n",
1192 (unsigned long long)devid);
f46b5a66 1193 }
2b82032c 1194 device = btrfs_find_device(root, devid, NULL, NULL);
f46b5a66 1195 if (!device) {
21380931
JB		 1196 printk(KERN_INFO "resizer unable to find device %llu\n",
1197 (unsigned long long)devid);
f46b5a66
CH		 1198 ret = -EINVAL;
1199 goto out_unlock;
1200 }
1201 if (!strcmp(sizestr, "max"))
1202 new_size = device->bdev->bd_inode->i_size;
1203 else {
1204 if (sizestr[0] == '-') {
1205 mod = -1;
1206 sizestr++;
1207 } else if (sizestr[0] == '+') {
1208 mod = 1;
1209 sizestr++;
1210 }
91748467 1211 new_size = memparse(sizestr, NULL);
f46b5a66
CH		 1212 if (new_size == 0) {
1213 ret = -EINVAL;
1214 goto out_unlock;
1215 }
1216 }
1217
1218 old_size = device->total_bytes;
1219
1220 if (mod < 0) {
1221 if (new_size > old_size) {
1222 ret = -EINVAL;
1223 goto out_unlock;
1224 }
1225 new_size = old_size - new_size;
1226 } else if (mod > 0) {
1227 new_size = old_size + new_size;
1228 }
1229
1230 if (new_size < 256 * 1024 * 1024) {
1231 ret = -EINVAL;
1232 goto out_unlock;
1233 }
1234 if (new_size > device->bdev->bd_inode->i_size) {
1235 ret = -EFBIG;
1236 goto out_unlock;
1237 }
1238
1239 do_div(new_size, root->sectorsize);
1240 new_size *= root->sectorsize;
1241
1242 printk(KERN_INFO "new size for %s is %llu\n",
1243 device->name, (unsigned long long)new_size);
1244
1245 if (new_size > old_size) {
a22285a6 1246 trans = btrfs_start_transaction(root, 0);
98d5dc13
TI		 1247 if (IS_ERR(trans)) {
1248 ret = PTR_ERR(trans);
1249 goto out_unlock;
1250 }
f46b5a66
CH		 1251 ret = btrfs_grow_device(trans, device, new_size);
1252 btrfs_commit_transaction(trans, root);
1253 } else {
1254 ret = btrfs_shrink_device(device, new_size);
1255 }
1256
1257out_unlock:
7d9eb12c 1258 mutex_unlock(&root->fs_info->volume_mutex);
f46b5a66
CH		 1259 kfree(vol_args);
1260 return ret;
1261}
1262
72fd032e
SW		 1263static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1264 char *name,
1265 unsigned long fd,
1266 int subvol,
b83cc969
LZ		 1267 u64 *transid,
1268 bool readonly)
f46b5a66 1269{
cb8e7090 1270 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3de4586c 1271 struct file *src_file;
f46b5a66 1272 int namelen;
3de4586c 1273 int ret = 0;
f46b5a66 1274
c146afad
YZ		 1275 if (root->fs_info->sb->s_flags & MS_RDONLY)
1276 return -EROFS;
1277
72fd032e
SW		 1278 namelen = strlen(name);
1279 if (strchr(name, '/')) {
f46b5a66
CH		 1280 ret = -EINVAL;
1281 goto out;
1282 }
1283
3de4586c 1284 if (subvol) {
72fd032e 1285 ret = btrfs_mksubvol(&file->f_path, name, namelen,
b83cc969 1286 NULL, transid, readonly);
cb8e7090 1287 } else {
3de4586c 1288 struct inode *src_inode;
72fd032e 1289 src_file = fget(fd);
3de4586c
CM		 1290 if (!src_file) {
1291 ret = -EINVAL;
1292 goto out;
1293 }
1294
1295 src_inode = src_file->f_path.dentry->d_inode;
1296 if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
d397712b
CM		 1297 printk(KERN_INFO "btrfs: Snapshot src from "
1298 "another FS\n");
3de4586c
CM		 1299 ret = -EINVAL;
1300 fput(src_file);
1301 goto out;
1302 }
72fd032e
SW		 1303 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1304 BTRFS_I(src_inode)->root,
b83cc969 1305 transid, readonly);
3de4586c 1306 fput(src_file);
cb8e7090 1307 }
f46b5a66 1308out:
72fd032e
SW		 1309 return ret;
1310}
1311
1312static noinline int btrfs_ioctl_snap_create(struct file *file,
fa0d2b9b 1313 void __user *arg, int subvol)
72fd032e 1314{
fa0d2b9b 1315 struct btrfs_ioctl_vol_args *vol_args;
72fd032e
SW		 1316 int ret;
1317
fa0d2b9b
LZ		 1318 vol_args = memdup_user(arg, sizeof(*vol_args));
1319 if (IS_ERR(vol_args))
1320 return PTR_ERR(vol_args);
1321 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
72fd032e 1322
fa0d2b9b 1323 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
b83cc969
LZ		 1324 vol_args->fd, subvol,
1325 NULL, false);
fdfb1e4f 1326
fa0d2b9b
LZ		 1327 kfree(vol_args);
1328 return ret;
1329}
fdfb1e4f 1330
fa0d2b9b
LZ		 1331static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1332 void __user *arg, int subvol)
1333{
1334 struct btrfs_ioctl_vol_args_v2 *vol_args;
1335 int ret;
1336 u64 transid = 0;
1337 u64 *ptr = NULL;
b83cc969 1338 bool readonly = false;
75eaa0e2 1339
fa0d2b9b
LZ		 1340 vol_args = memdup_user(arg, sizeof(*vol_args));
1341 if (IS_ERR(vol_args))
1342 return PTR_ERR(vol_args);
1343 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
75eaa0e2 1344
b83cc969
LZ		 1345 if (vol_args->flags &
1346 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY)) {
1347 ret = -EOPNOTSUPP;
fa0d2b9b 1348 goto out;
72fd032e 1349 }
fa0d2b9b
LZ		 1350
1351 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1352 ptr = &transid;
b83cc969
LZ		 1353 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1354 readonly = true;
fa0d2b9b
LZ		 1355
1356 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
b83cc969
LZ		 1357 vol_args->fd, subvol,
1358 ptr, readonly);
fa0d2b9b
LZ		 1359
1360 if (ret == 0 && ptr &&
1361 copy_to_user(arg +
1362 offsetof(struct btrfs_ioctl_vol_args_v2,
1363 transid), ptr, sizeof(*ptr)))
1364 ret = -EFAULT;
fdfb1e4f 1365out:
f46b5a66
CH		 1366 kfree(vol_args);
1367 return ret;
1368}
1369
0caa102d
LZ		 1370static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1371 void __user *arg)
1372{
1373 struct inode *inode = fdentry(file)->d_inode;
1374 struct btrfs_root *root = BTRFS_I(inode)->root;
1375 int ret = 0;
1376 u64 flags = 0;
1377
33345d01 1378 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
0caa102d
LZ		 1379 return -EINVAL;
1380
1381 down_read(&root->fs_info->subvol_sem);
1382 if (btrfs_root_readonly(root))
1383 flags |= BTRFS_SUBVOL_RDONLY;
1384 up_read(&root->fs_info->subvol_sem);
1385
1386 if (copy_to_user(arg, &flags, sizeof(flags)))
1387 ret = -EFAULT;
1388
1389 return ret;
1390}
1391
1392static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1393 void __user *arg)
1394{
1395 struct inode *inode = fdentry(file)->d_inode;
1396 struct btrfs_root *root = BTRFS_I(inode)->root;
1397 struct btrfs_trans_handle *trans;
1398 u64 root_flags;
1399 u64 flags;
1400 int ret = 0;
1401
1402 if (root->fs_info->sb->s_flags & MS_RDONLY)
1403 return -EROFS;
1404
33345d01 1405 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
0caa102d
LZ		 1406 return -EINVAL;
1407
1408 if (copy_from_user(&flags, arg, sizeof(flags)))
1409 return -EFAULT;
1410
b4dc2b8c 1411 if (flags & BTRFS_SUBVOL_CREATE_ASYNC)
0caa102d
LZ		 1412 return -EINVAL;
1413
1414 if (flags & ~BTRFS_SUBVOL_RDONLY)
1415 return -EOPNOTSUPP;
1416
2e149670 1417 if (!inode_owner_or_capable(inode))
b4dc2b8c
LZ		 1418 return -EACCES;
1419
0caa102d
LZ		 1420 down_write(&root->fs_info->subvol_sem);
1421
1422 /* nothing to do */
1423 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1424 goto out;
1425
1426 root_flags = btrfs_root_flags(&root->root_item);
1427 if (flags & BTRFS_SUBVOL_RDONLY)
1428 btrfs_set_root_flags(&root->root_item,
1429 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1430 else
1431 btrfs_set_root_flags(&root->root_item,
1432 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1433
1434 trans = btrfs_start_transaction(root, 1);
1435 if (IS_ERR(trans)) {
1436 ret = PTR_ERR(trans);
1437 goto out_reset;
1438 }
1439
b4dc2b8c 1440 ret = btrfs_update_root(trans, root->fs_info->tree_root,
0caa102d
LZ		 1441 &root->root_key, &root->root_item);
1442
1443 btrfs_commit_transaction(trans, root);
1444out_reset:
1445 if (ret)
1446 btrfs_set_root_flags(&root->root_item, root_flags);
1447out:
1448 up_write(&root->fs_info->subvol_sem);
1449 return ret;
1450}
1451
76dda93c
YZ		 1452/*
1453 * helper to check if the subvolume references other subvolumes
1454 */
1455static noinline int may_destroy_subvol(struct btrfs_root *root)
1456{
1457 struct btrfs_path *path;
1458 struct btrfs_key key;
1459 int ret;
1460
1461 path = btrfs_alloc_path();
1462 if (!path)
1463 return -ENOMEM;
1464
1465 key.objectid = root->root_key.objectid;
1466 key.type = BTRFS_ROOT_REF_KEY;
1467 key.offset = (u64)-1;
1468
1469 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1470 &key, path, 0, 0);
1471 if (ret < 0)
1472 goto out;
1473 BUG_ON(ret == 0);
1474
1475 ret = 0;
1476 if (path->slots[0] > 0) {
1477 path->slots[0]--;
1478 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1479 if (key.objectid == root->root_key.objectid &&
1480 key.type == BTRFS_ROOT_REF_KEY)
1481 ret = -ENOTEMPTY;
1482 }
1483out:
1484 btrfs_free_path(path);
1485 return ret;
1486}
1487
ac8e9819
CM		 1488static noinline int key_in_sk(struct btrfs_key *key,
1489 struct btrfs_ioctl_search_key *sk)
1490{
abc6e134
CM		 1491 struct btrfs_key test;
1492 int ret;
1493
1494 test.objectid = sk->min_objectid;
1495 test.type = sk->min_type;
1496 test.offset = sk->min_offset;
1497
1498 ret = btrfs_comp_cpu_keys(key, &test);
1499 if (ret < 0)
ac8e9819 1500 return 0;
abc6e134
CM		 1501
1502 test.objectid = sk->max_objectid;
1503 test.type = sk->max_type;
1504 test.offset = sk->max_offset;
1505
1506 ret = btrfs_comp_cpu_keys(key, &test);
1507 if (ret > 0)
ac8e9819
CM		 1508 return 0;
1509 return 1;
1510}
1511
1512static noinline int copy_to_sk(struct btrfs_root *root,
1513 struct btrfs_path *path,
1514 struct btrfs_key *key,
1515 struct btrfs_ioctl_search_key *sk,
1516 char *buf,
1517 unsigned long *sk_offset,
1518 int *num_found)
1519{
1520 u64 found_transid;
1521 struct extent_buffer *leaf;
1522 struct btrfs_ioctl_search_header sh;
1523 unsigned long item_off;
1524 unsigned long item_len;
1525 int nritems;
1526 int i;
1527 int slot;
ac8e9819
CM		 1528 int ret = 0;
1529
1530 leaf = path->nodes[0];
1531 slot = path->slots[0];
1532 nritems = btrfs_header_nritems(leaf);
1533
1534 if (btrfs_header_generation(leaf) > sk->max_transid) {
1535 i = nritems;
1536 goto advance_key;
1537 }
1538 found_transid = btrfs_header_generation(leaf);
1539
1540 for (i = slot; i < nritems; i++) {
1541 item_off = btrfs_item_ptr_offset(leaf, i);
1542 item_len = btrfs_item_size_nr(leaf, i);
1543
1544 if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1545 item_len = 0;
1546
1547 if (sizeof(sh) + item_len + *sk_offset >
1548 BTRFS_SEARCH_ARGS_BUFSIZE) {
1549 ret = 1;
1550 goto overflow;
1551 }
1552
1553 btrfs_item_key_to_cpu(leaf, key, i);
1554 if (!key_in_sk(key, sk))
1555 continue;
1556
1557 sh.objectid = key->objectid;
1558 sh.offset = key->offset;
1559 sh.type = key->type;
1560 sh.len = item_len;
1561 sh.transid = found_transid;
1562
1563 /* copy search result header */
1564 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1565 *sk_offset += sizeof(sh);
1566
1567 if (item_len) {
1568 char *p = buf + *sk_offset;
1569 /* copy the item */
1570 read_extent_buffer(leaf, p,
1571 item_off, item_len);
1572 *sk_offset += item_len;
ac8e9819 1573 }
e2156867 1574 (*num_found)++;
ac8e9819
CM		 1575
1576 if (*num_found >= sk->nr_items)
1577 break;
1578 }
1579advance_key:
abc6e134
CM		 1580 ret = 0;
1581 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
ac8e9819 1582 key->offset++;
abc6e134
CM		 1583 else if (key->type < (u8)-1 && key->type < sk->max_type) {
1584 key->offset = 0;
ac8e9819 1585 key->type++;
abc6e134
CM		 1586 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1587 key->offset = 0;
1588 key->type = 0;
ac8e9819 1589 key->objectid++;
abc6e134
CM		 1590 } else
1591 ret = 1;
ac8e9819 1592overflow:
ac8e9819
CM		 1593 return ret;
1594}
1595
1596static noinline int search_ioctl(struct inode *inode,
1597 struct btrfs_ioctl_search_args *args)
1598{
1599 struct btrfs_root *root;
1600 struct btrfs_key key;
1601 struct btrfs_key max_key;
1602 struct btrfs_path *path;
1603 struct btrfs_ioctl_search_key *sk = &args->key;
1604 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1605 int ret;
1606 int num_found = 0;
1607 unsigned long sk_offset = 0;
1608
1609 path = btrfs_alloc_path();
1610 if (!path)
1611 return -ENOMEM;
1612
1613 if (sk->tree_id == 0) {
1614 /* search the root of the inode that was passed */
1615 root = BTRFS_I(inode)->root;
1616 } else {
1617 key.objectid = sk->tree_id;
1618 key.type = BTRFS_ROOT_ITEM_KEY;
1619 key.offset = (u64)-1;
1620 root = btrfs_read_fs_root_no_name(info, &key);
1621 if (IS_ERR(root)) {
1622 printk(KERN_ERR "could not find root %llu\n",
1623 sk->tree_id);
1624 btrfs_free_path(path);
1625 return -ENOENT;
1626 }
1627 }
1628
1629 key.objectid = sk->min_objectid;
1630 key.type = sk->min_type;
1631 key.offset = sk->min_offset;
1632
1633 max_key.objectid = sk->max_objectid;
1634 max_key.type = sk->max_type;
1635 max_key.offset = sk->max_offset;
1636
1637 path->keep_locks = 1;
1638
1639 while(1) {
1640 ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1641 sk->min_transid);
1642 if (ret != 0) {
1643 if (ret > 0)
1644 ret = 0;
1645 goto err;
1646 }
1647 ret = copy_to_sk(root, path, &key, sk, args->buf,
1648 &sk_offset, &num_found);
b3b4aa74 1649 btrfs_release_path(path);
ac8e9819
CM		 1650 if (ret || num_found >= sk->nr_items)
1651 break;
1652
1653 }
1654 ret = 0;
1655err:
1656 sk->nr_items = num_found;
1657 btrfs_free_path(path);
1658 return ret;
1659}
1660
1661static noinline int btrfs_ioctl_tree_search(struct file *file,
1662 void __user *argp)
1663{
1664 struct btrfs_ioctl_search_args *args;
1665 struct inode *inode;
1666 int ret;
1667
1668 if (!capable(CAP_SYS_ADMIN))
1669 return -EPERM;
1670
2354d08f
JL		 1671 args = memdup_user(argp, sizeof(*args));
1672 if (IS_ERR(args))
1673 return PTR_ERR(args);
ac8e9819 1674
ac8e9819
CM		 1675 inode = fdentry(file)->d_inode;
1676 ret = search_ioctl(inode, args);
1677 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1678 ret = -EFAULT;
1679 kfree(args);
1680 return ret;
1681}
1682
98d377a0 1683/*
ac8e9819
CM		 1684 * Search INODE_REFs to identify path name of 'dirid' directory
1685 * in a 'tree_id' tree. and sets path name to 'name'.
1686 */
98d377a0
TH		 1687static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1688 u64 tree_id, u64 dirid, char *name)
1689{
1690 struct btrfs_root *root;
1691 struct btrfs_key key;
ac8e9819 1692 char *ptr;
98d377a0
TH		 1693 int ret = -1;
1694 int slot;
1695 int len;
1696 int total_len = 0;
1697 struct btrfs_inode_ref *iref;
1698 struct extent_buffer *l;
1699 struct btrfs_path *path;
1700
1701 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1702 name[0]='\0';
1703 return 0;
1704 }
1705
1706 path = btrfs_alloc_path();
1707 if (!path)
1708 return -ENOMEM;
1709
ac8e9819 1710 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
98d377a0
TH		 1711
1712 key.objectid = tree_id;
1713 key.type = BTRFS_ROOT_ITEM_KEY;
1714 key.offset = (u64)-1;
1715 root = btrfs_read_fs_root_no_name(info, &key);
1716 if (IS_ERR(root)) {
1717 printk(KERN_ERR "could not find root %llu\n", tree_id);
8ad6fcab
CM		 1718 ret = -ENOENT;
1719 goto out;
98d377a0
TH		 1720 }
1721
1722 key.objectid = dirid;
1723 key.type = BTRFS_INODE_REF_KEY;
8ad6fcab 1724 key.offset = (u64)-1;
98d377a0
TH		 1725
1726 while(1) {
1727 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1728 if (ret < 0)
1729 goto out;
1730
1731 l = path->nodes[0];
1732 slot = path->slots[0];
8ad6fcab
CM		 1733 if (ret > 0 && slot > 0)
1734 slot--;
98d377a0
TH		 1735 btrfs_item_key_to_cpu(l, &key, slot);
1736
1737 if (ret > 0 && (key.objectid != dirid ||
ac8e9819
CM		 1738 key.type != BTRFS_INODE_REF_KEY)) {
1739 ret = -ENOENT;
98d377a0 1740 goto out;
ac8e9819 1741 }
98d377a0
TH		 1742
1743 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1744 len = btrfs_inode_ref_name_len(l, iref);
1745 ptr -= len + 1;
1746 total_len += len + 1;
ac8e9819 1747 if (ptr < name)
98d377a0
TH		 1748 goto out;
1749
1750 *(ptr + len) = '/';
1751 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1752
1753 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1754 break;
1755
b3b4aa74 1756 btrfs_release_path(path);
98d377a0 1757 key.objectid = key.offset;
8ad6fcab 1758 key.offset = (u64)-1;
98d377a0 1759 dirid = key.objectid;
98d377a0 1760 }
ac8e9819 1761 if (ptr < name)
98d377a0 1762 goto out;
77906a50 1763 memmove(name, ptr, total_len);
98d377a0
TH		 1764 name[total_len]='\0';
1765 ret = 0;
1766out:
1767 btrfs_free_path(path);
ac8e9819
CM		 1768 return ret;
1769}
1770
1771static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1772 void __user *argp)
1773{
1774 struct btrfs_ioctl_ino_lookup_args *args;
1775 struct inode *inode;
1776 int ret;
1777
1778 if (!capable(CAP_SYS_ADMIN))
1779 return -EPERM;
1780
2354d08f
JL		 1781 args = memdup_user(argp, sizeof(*args));
1782 if (IS_ERR(args))
1783 return PTR_ERR(args);
c2b96929 1784
ac8e9819
CM		 1785 inode = fdentry(file)->d_inode;
1786
1b53ac4d
CM		 1787 if (args->treeid == 0)
1788 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1789
ac8e9819
CM		 1790 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1791 args->treeid, args->objectid,
1792 args->name);
1793
1794 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1795 ret = -EFAULT;
1796
1797 kfree(args);
98d377a0
TH		 1798 return ret;
1799}
1800
76dda93c
YZ		 1801static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1802 void __user *arg)
1803{
1804 struct dentry *parent = fdentry(file);
1805 struct dentry *dentry;
1806 struct inode *dir = parent->d_inode;
1807 struct inode *inode;
1808 struct btrfs_root *root = BTRFS_I(dir)->root;
1809 struct btrfs_root *dest = NULL;
1810 struct btrfs_ioctl_vol_args *vol_args;
1811 struct btrfs_trans_handle *trans;
1812 int namelen;
1813 int ret;
1814 int err = 0;
1815
76dda93c
YZ		 1816 vol_args = memdup_user(arg, sizeof(*vol_args));
1817 if (IS_ERR(vol_args))
1818 return PTR_ERR(vol_args);
1819
1820 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1821 namelen = strlen(vol_args->name);
1822 if (strchr(vol_args->name, '/') ||
1823 strncmp(vol_args->name, "..", namelen) == 0) {
1824 err = -EINVAL;
1825 goto out;
1826 }
1827
1828 err = mnt_want_write(file->f_path.mnt);
1829 if (err)
1830 goto out;
1831
1832 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
1833 dentry = lookup_one_len(vol_args->name, parent, namelen);
1834 if (IS_ERR(dentry)) {
1835 err = PTR_ERR(dentry);
1836 goto out_unlock_dir;
1837 }
1838
1839 if (!dentry->d_inode) {
1840 err = -ENOENT;
1841 goto out_dput;
1842 }
1843
1844 inode = dentry->d_inode;
4260f7c7
SW		 1845 dest = BTRFS_I(inode)->root;
1846 if (!capable(CAP_SYS_ADMIN)){
1847 /*
1848 * Regular user. Only allow this with a special mount
1849 * option, when the user has write+exec access to the
1850 * subvol root, and when rmdir(2) would have been
1851 * allowed.
1852 *
1853 * Note that this is _not_ check that the subvol is
1854 * empty or doesn't contain data that we wouldn't
1855 * otherwise be able to delete.
1856 *
1857 * Users who want to delete empty subvols should try
1858 * rmdir(2).
1859 */
1860 err = -EPERM;
1861 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1862 goto out_dput;
1863
1864 /*
1865 * Do not allow deletion if the parent dir is the same
1866 * as the dir to be deleted. That means the ioctl
1867 * must be called on the dentry referencing the root
1868 * of the subvol, not a random directory contained
1869 * within it.
1870 */
1871 err = -EINVAL;
1872 if (root == dest)
1873 goto out_dput;
1874
1875 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
1876 if (err)
1877 goto out_dput;
1878
1879 /* check if subvolume may be deleted by a non-root user */
1880 err = btrfs_may_delete(dir, dentry, 1);
1881 if (err)
1882 goto out_dput;
1883 }
1884
33345d01 1885 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
76dda93c
YZ		 1886 err = -EINVAL;
1887 goto out_dput;
1888 }
1889
76dda93c
YZ		 1890 mutex_lock(&inode->i_mutex);
1891 err = d_invalidate(dentry);
1892 if (err)
1893 goto out_unlock;
1894
1895 down_write(&root->fs_info->subvol_sem);
1896
1897 err = may_destroy_subvol(dest);
1898 if (err)
1899 goto out_up_write;
1900
a22285a6
YZ		 1901 trans = btrfs_start_transaction(root, 0);
1902 if (IS_ERR(trans)) {
1903 err = PTR_ERR(trans);
d327099a 1904 goto out_up_write;
a22285a6
YZ		 1905 }
1906 trans->block_rsv = &root->fs_info->global_block_rsv;
1907
76dda93c
YZ		 1908 ret = btrfs_unlink_subvol(trans, root, dir,
1909 dest->root_key.objectid,
1910 dentry->d_name.name,
1911 dentry->d_name.len);
1912 BUG_ON(ret);
1913
1914 btrfs_record_root_in_trans(trans, dest);
1915
1916 memset(&dest->root_item.drop_progress, 0,
1917 sizeof(dest->root_item.drop_progress));
1918 dest->root_item.drop_level = 0;
1919 btrfs_set_root_refs(&dest->root_item, 0);
1920
d68fc57b
YZ		 1921 if (!xchg(&dest->orphan_item_inserted, 1)) {
1922 ret = btrfs_insert_orphan_item(trans,
1923 root->fs_info->tree_root,
1924 dest->root_key.objectid);
1925 BUG_ON(ret);
1926 }
76dda93c 1927
531cb13f 1928 ret = btrfs_end_transaction(trans, root);
76dda93c
YZ		 1929 BUG_ON(ret);
1930 inode->i_flags |= S_DEAD;
1931out_up_write:
1932 up_write(&root->fs_info->subvol_sem);
1933out_unlock:
1934 mutex_unlock(&inode->i_mutex);
1935 if (!err) {
efefb143 1936 shrink_dcache_sb(root->fs_info->sb);
76dda93c
YZ		 1937 btrfs_invalidate_inodes(dest);
1938 d_delete(dentry);
1939 }
1940out_dput:
1941 dput(dentry);
1942out_unlock_dir:
1943 mutex_unlock(&dir->i_mutex);
1944 mnt_drop_write(file->f_path.mnt);
1945out:
1946 kfree(vol_args);
1947 return err;
1948}
1949
1e701a32 1950static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
f46b5a66
CH		 1951{
1952 struct inode *inode = fdentry(file)->d_inode;
1953 struct btrfs_root *root = BTRFS_I(inode)->root;
1e701a32 1954 struct btrfs_ioctl_defrag_range_args *range;
c146afad
YZ		 1955 int ret;
1956
b83cc969
LZ		 1957 if (btrfs_root_readonly(root))
1958 return -EROFS;
1959
c146afad
YZ		 1960 ret = mnt_want_write(file->f_path.mnt);
1961 if (ret)
1962 return ret;
f46b5a66
CH		 1963
1964 switch (inode->i_mode & S_IFMT) {
1965 case S_IFDIR:
e441d54d
CM		 1966 if (!capable(CAP_SYS_ADMIN)) {
1967 ret = -EPERM;
1968 goto out;
1969 }
8929ecfa
YZ		 1970 ret = btrfs_defrag_root(root, 0);
1971 if (ret)
1972 goto out;
1973 ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
f46b5a66
CH		 1974 break;
1975 case S_IFREG:
e441d54d
CM		 1976 if (!(file->f_mode & FMODE_WRITE)) {
1977 ret = -EINVAL;
1978 goto out;
1979 }
1e701a32
CM		 1980
1981 range = kzalloc(sizeof(*range), GFP_KERNEL);
1982 if (!range) {
1983 ret = -ENOMEM;
1984 goto out;
1985 }
1986
1987 if (argp) {
1988 if (copy_from_user(range, argp,
1989 sizeof(*range))) {
1990 ret = -EFAULT;
1991 kfree(range);
683be16e 1992 goto out;
1e701a32
CM		 1993 }
1994 /* compression requires us to start the IO */
1995 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1996 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
1997 range->extent_thresh = (u32)-1;
1998 }
1999 } else {
2000 /* the rest are all set to zero by kzalloc */
2001 range->len = (u64)-1;
2002 }
4cb5300b
CM		 2003 ret = btrfs_defrag_file(fdentry(file)->d_inode, file,
2004 range, 0, 0);
2005 if (ret > 0)
2006 ret = 0;
1e701a32 2007 kfree(range);
f46b5a66 2008 break;
8929ecfa
YZ		 2009 default:
2010 ret = -EINVAL;
f46b5a66 2011 }
e441d54d 2012out:
ab67b7c1 2013 mnt_drop_write(file->f_path.mnt);
e441d54d 2014 return ret;
f46b5a66
CH		 2015}
2016
b2950863 2017static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
f46b5a66
CH		 2018{
2019 struct btrfs_ioctl_vol_args *vol_args;
2020 int ret;
2021
e441d54d
CM		 2022 if (!capable(CAP_SYS_ADMIN))
2023 return -EPERM;
2024
dae7b665
LZ		 2025 vol_args = memdup_user(arg, sizeof(*vol_args));
2026 if (IS_ERR(vol_args))
2027 return PTR_ERR(vol_args);
f46b5a66 2028
5516e595 2029 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66
CH		 2030 ret = btrfs_init_new_device(root, vol_args->name);
2031
f46b5a66
CH		 2032 kfree(vol_args);
2033 return ret;
2034}
2035
b2950863 2036static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
f46b5a66
CH		 2037{
2038 struct btrfs_ioctl_vol_args *vol_args;
2039 int ret;
2040
e441d54d
CM		 2041 if (!capable(CAP_SYS_ADMIN))
2042 return -EPERM;
2043
c146afad
YZ		 2044 if (root->fs_info->sb->s_flags & MS_RDONLY)
2045 return -EROFS;
2046
dae7b665
LZ		 2047 vol_args = memdup_user(arg, sizeof(*vol_args));
2048 if (IS_ERR(vol_args))
2049 return PTR_ERR(vol_args);
f46b5a66 2050
5516e595 2051 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66
CH		 2052 ret = btrfs_rm_device(root, vol_args->name);
2053
f46b5a66
CH		 2054 kfree(vol_args);
2055 return ret;
2056}
2057
475f6387
JS		 2058static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2059{
027ed2f0 2060 struct btrfs_ioctl_fs_info_args *fi_args;
475f6387
JS		 2061 struct btrfs_device *device;
2062 struct btrfs_device *next;
2063 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
027ed2f0 2064 int ret = 0;
475f6387
JS		 2065
2066 if (!capable(CAP_SYS_ADMIN))
2067 return -EPERM;
2068
027ed2f0
LZ		 2069 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2070 if (!fi_args)
2071 return -ENOMEM;
2072
2073 fi_args->num_devices = fs_devices->num_devices;
2074 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
475f6387
JS		 2075
2076 mutex_lock(&fs_devices->device_list_mutex);
2077 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
027ed2f0
LZ		 2078 if (device->devid > fi_args->max_id)
2079 fi_args->max_id = device->devid;
475f6387
JS		 2080 }
2081 mutex_unlock(&fs_devices->device_list_mutex);
2082
027ed2f0
LZ		 2083 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2084 ret = -EFAULT;
475f6387 2085
027ed2f0
LZ		 2086 kfree(fi_args);
2087 return ret;
475f6387
JS		 2088}
2089
2090static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2091{
2092 struct btrfs_ioctl_dev_info_args *di_args;
2093 struct btrfs_device *dev;
2094 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2095 int ret = 0;
2096 char *s_uuid = NULL;
2097 char empty_uuid[BTRFS_UUID_SIZE] = {0};
2098
2099 if (!capable(CAP_SYS_ADMIN))
2100 return -EPERM;
2101
2102 di_args = memdup_user(arg, sizeof(*di_args));
2103 if (IS_ERR(di_args))
2104 return PTR_ERR(di_args);
2105
2106 if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2107 s_uuid = di_args->uuid;
2108
2109 mutex_lock(&fs_devices->device_list_mutex);
2110 dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL);
2111 mutex_unlock(&fs_devices->device_list_mutex);
2112
2113 if (!dev) {
2114 ret = -ENODEV;
2115 goto out;
2116 }
2117
2118 di_args->devid = dev->devid;
2119 di_args->bytes_used = dev->bytes_used;
2120 di_args->total_bytes = dev->total_bytes;
2121 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2122 strncpy(di_args->path, dev->name, sizeof(di_args->path));
2123
2124out:
2125 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2126 ret = -EFAULT;
2127
2128 kfree(di_args);
2129 return ret;
2130}
2131
76dda93c
YZ		 2132static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2133 u64 off, u64 olen, u64 destoff)
f46b5a66
CH		 2134{
2135 struct inode *inode = fdentry(file)->d_inode;
2136 struct btrfs_root *root = BTRFS_I(inode)->root;
2137 struct file *src_file;
2138 struct inode *src;
2139 struct btrfs_trans_handle *trans;
f46b5a66 2140 struct btrfs_path *path;
f46b5a66 2141 struct extent_buffer *leaf;
ae01a0ab
YZ		 2142 char *buf;
2143 struct btrfs_key key;
f46b5a66
CH		 2144 u32 nritems;
2145 int slot;
ae01a0ab 2146 int ret;
c5c9cd4d
SW		 2147 u64 len = olen;
2148 u64 bs = root->fs_info->sb->s_blocksize;
2149 u64 hint_byte;
d20f7043 2150
c5c9cd4d
SW		 2151 /*
2152 * TODO:
2153 * - split compressed inline extents. annoying: we need to
2154 * decompress into destination's address_space (the file offset
2155 * may change, so source mapping won't do), then recompress (or
2156 * otherwise reinsert) a subrange.
2157 * - allow ranges within the same file to be cloned (provided
2158 * they don't overlap)?
2159 */
2160
e441d54d 2161 /* the destination must be opened for writing */
2ebc3464 2162 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
e441d54d
CM		 2163 return -EINVAL;
2164
b83cc969
LZ		 2165 if (btrfs_root_readonly(root))
2166 return -EROFS;
2167
c146afad
YZ		 2168 ret = mnt_want_write(file->f_path.mnt);
2169 if (ret)
2170 return ret;
2171
c5c9cd4d 2172 src_file = fget(srcfd);
ab67b7c1
YZ		 2173 if (!src_file) {
2174 ret = -EBADF;
2175 goto out_drop_write;
2176 }
5dc64164 2177
f46b5a66
CH		 2178 src = src_file->f_dentry->d_inode;
2179
c5c9cd4d
SW		 2180 ret = -EINVAL;
2181 if (src == inode)
2182 goto out_fput;
2183
5dc64164
DR		 2184 /* the src must be open for reading */
2185 if (!(src_file->f_mode & FMODE_READ))
2186 goto out_fput;
2187
ae01a0ab
YZ		 2188 ret = -EISDIR;
2189 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2190 goto out_fput;
2191
f46b5a66 2192 ret = -EXDEV;
ae01a0ab
YZ		 2193 if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
2194 goto out_fput;
2195
2196 ret = -ENOMEM;
2197 buf = vmalloc(btrfs_level_size(root, 0));
2198 if (!buf)
2199 goto out_fput;
2200
2201 path = btrfs_alloc_path();
2202 if (!path) {
2203 vfree(buf);
f46b5a66 2204 goto out_fput;
ae01a0ab
YZ		 2205 }
2206 path->reada = 2;
f46b5a66
CH		 2207
2208 if (inode < src) {
fccdae43
SW		 2209 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2210 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
f46b5a66 2211 } else {
fccdae43
SW		 2212 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2213 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
f46b5a66
CH		 2214 }
2215
c5c9cd4d
SW		 2216 /* determine range to clone */
2217 ret = -EINVAL;
2ebc3464 2218 if (off + len > src->i_size || off + len < off)
f46b5a66 2219 goto out_unlock;
c5c9cd4d
SW		 2220 if (len == 0)
2221 olen = len = src->i_size - off;
2222 /* if we extend to eof, continue to block boundary */
2223 if (off + len == src->i_size)
2a6b8dae 2224 len = ALIGN(src->i_size, bs) - off;
c5c9cd4d
SW		 2225
2226 /* verify the end result is block aligned */
2a6b8dae
LZ		 2227 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2228 !IS_ALIGNED(destoff, bs))
c5c9cd4d
SW		 2229 goto out_unlock;
2230
f46b5a66
CH		 2231 /* do any pending delalloc/csum calc on src, one way or
2232 another, and lock file content */
2233 while (1) {
31840ae1 2234 struct btrfs_ordered_extent *ordered;
c5c9cd4d 2235 lock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
9a019196
SW		 2236 ordered = btrfs_lookup_first_ordered_extent(src, off+len);
2237 if (!ordered &&
2238 !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len,
2239 EXTENT_DELALLOC, 0, NULL))
f46b5a66 2240 break;
c5c9cd4d 2241 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
ae01a0ab
YZ		 2242 if (ordered)
2243 btrfs_put_ordered_extent(ordered);
9a019196 2244 btrfs_wait_ordered_range(src, off, len);
f46b5a66
CH		 2245 }
2246
f81c9cdc
SW		 2247 /* truncate page cache pages from target inode range */
2248 truncate_inode_pages_range(&inode->i_data, off,
2249 ALIGN(off + len, PAGE_CACHE_SIZE) - 1);
2250
c5c9cd4d 2251 /* clone data */
33345d01 2252 key.objectid = btrfs_ino(src);
ae01a0ab
YZ		 2253 key.type = BTRFS_EXTENT_DATA_KEY;
2254 key.offset = 0;
f46b5a66
CH		 2255
2256 while (1) {
2257 /*
2258 * note the key will change type as we walk through the
2259 * tree.
2260 */
a22285a6 2261 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
f46b5a66
CH		 2262 if (ret < 0)
2263 goto out;
2264
ae01a0ab
YZ		 2265 nritems = btrfs_header_nritems(path->nodes[0]);
2266 if (path->slots[0] >= nritems) {
f46b5a66
CH		 2267 ret = btrfs_next_leaf(root, path);
2268 if (ret < 0)
2269 goto out;
2270 if (ret > 0)
2271 break;
ae01a0ab 2272 nritems = btrfs_header_nritems(path->nodes[0]);
f46b5a66
CH		 2273 }
2274 leaf = path->nodes[0];
2275 slot = path->slots[0];
f46b5a66 2276
ae01a0ab 2277 btrfs_item_key_to_cpu(leaf, &key, slot);
d20f7043 2278 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
33345d01 2279 key.objectid != btrfs_ino(src))
f46b5a66
CH		 2280 break;
2281
c5c9cd4d
SW		 2282 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2283 struct btrfs_file_extent_item *extent;
2284 int type;
31840ae1
ZY		 2285 u32 size;
2286 struct btrfs_key new_key;
c5c9cd4d
SW		 2287 u64 disko = 0, diskl = 0;
2288 u64 datao = 0, datal = 0;
2289 u8 comp;
b5384d48 2290 u64 endoff;
31840ae1
ZY		 2291
2292 size = btrfs_item_size_nr(leaf, slot);
2293 read_extent_buffer(leaf, buf,
2294 btrfs_item_ptr_offset(leaf, slot),
2295 size);
c5c9cd4d
SW		 2296
2297 extent = btrfs_item_ptr(leaf, slot,
2298 struct btrfs_file_extent_item);
2299 comp = btrfs_file_extent_compression(leaf, extent);
2300 type = btrfs_file_extent_type(leaf, extent);
c8a894d7
CM		 2301 if (type == BTRFS_FILE_EXTENT_REG ||
2302 type == BTRFS_FILE_EXTENT_PREALLOC) {
d397712b
CM		 2303 disko = btrfs_file_extent_disk_bytenr(leaf,
2304 extent);
2305 diskl = btrfs_file_extent_disk_num_bytes(leaf,
2306 extent);
c5c9cd4d 2307 datao = btrfs_file_extent_offset(leaf, extent);
d397712b
CM		 2308 datal = btrfs_file_extent_num_bytes(leaf,
2309 extent);
c5c9cd4d
SW		 2310 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2311 /* take upper bound, may be compressed */
2312 datal = btrfs_file_extent_ram_bytes(leaf,
2313 extent);
2314 }
b3b4aa74 2315 btrfs_release_path(path);
31840ae1 2316
050006a7 2317 if (key.offset + datal <= off ||
c5c9cd4d
SW		 2318 key.offset >= off+len)
2319 goto next;
2320
31840ae1 2321 memcpy(&new_key, &key, sizeof(new_key));
33345d01 2322 new_key.objectid = btrfs_ino(inode);
4d728ec7
LZ		 2323 if (off <= key.offset)
2324 new_key.offset = key.offset + destoff - off;
2325 else
2326 new_key.offset = destoff;
31840ae1 2327
a22285a6
YZ		 2328 trans = btrfs_start_transaction(root, 1);
2329 if (IS_ERR(trans)) {
2330 ret = PTR_ERR(trans);
2331 goto out;
2332 }
2333
c8a894d7
CM		 2334 if (type == BTRFS_FILE_EXTENT_REG ||
2335 type == BTRFS_FILE_EXTENT_PREALLOC) {
d72c0842
LZ		 2336 /*
2337 * a | --- range to clone ---| b
2338 * | ------------- extent ------------- |
2339 */
2340
2341 /* substract range b */
2342 if (key.offset + datal > off + len)
2343 datal = off + len - key.offset;
2344
2345 /* substract range a */
a22285a6
YZ		 2346 if (off > key.offset) {
2347 datao += off - key.offset;
2348 datal -= off - key.offset;
2349 }
2350
a22285a6
YZ		 2351 ret = btrfs_drop_extents(trans, inode,
2352 new_key.offset,
2353 new_key.offset + datal,
2354 &hint_byte, 1);
2355 BUG_ON(ret);
2356
c5c9cd4d
SW		 2357 ret = btrfs_insert_empty_item(trans, root, path,
2358 &new_key, size);
a22285a6 2359 BUG_ON(ret);
c5c9cd4d
SW		 2360
2361 leaf = path->nodes[0];
2362 slot = path->slots[0];
2363 write_extent_buffer(leaf, buf,
31840ae1
ZY		 2364 btrfs_item_ptr_offset(leaf, slot),
2365 size);
ae01a0ab 2366
c5c9cd4d 2367 extent = btrfs_item_ptr(leaf, slot,
f46b5a66 2368 struct btrfs_file_extent_item);
c5c9cd4d 2369
c5c9cd4d
SW		 2370 /* disko == 0 means it's a hole */
2371 if (!disko)
2372 datao = 0;
c5c9cd4d
SW		 2373
2374 btrfs_set_file_extent_offset(leaf, extent,
2375 datao);
2376 btrfs_set_file_extent_num_bytes(leaf, extent,
2377 datal);
2378 if (disko) {
2379 inode_add_bytes(inode, datal);
ae01a0ab 2380 ret = btrfs_inc_extent_ref(trans, root,
5d4f98a2
YZ		 2381 disko, diskl, 0,
2382 root->root_key.objectid,
33345d01 2383 btrfs_ino(inode),
5d4f98a2 2384 new_key.offset - datao);
31840ae1 2385 BUG_ON(ret);
f46b5a66 2386 }
c5c9cd4d
SW		 2387 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2388 u64 skip = 0;
2389 u64 trim = 0;
2390 if (off > key.offset) {
2391 skip = off - key.offset;
2392 new_key.offset += skip;
2393 }
d397712b 2394
c5c9cd4d
SW		 2395 if (key.offset + datal > off+len)
2396 trim = key.offset + datal - (off+len);
d397712b 2397
c5c9cd4d 2398 if (comp && (skip || trim)) {
c5c9cd4d 2399 ret = -EINVAL;
a22285a6 2400 btrfs_end_transaction(trans, root);
c5c9cd4d
SW		 2401 goto out;
2402 }
2403 size -= skip + trim;
2404 datal -= skip + trim;
a22285a6
YZ		 2405
2406 ret = btrfs_drop_extents(trans, inode,
2407 new_key.offset,
2408 new_key.offset + datal,
2409 &hint_byte, 1);
2410 BUG_ON(ret);
2411
c5c9cd4d
SW		 2412 ret = btrfs_insert_empty_item(trans, root, path,
2413 &new_key, size);
a22285a6 2414 BUG_ON(ret);
c5c9cd4d
SW		 2415
2416 if (skip) {
d397712b
CM		 2417 u32 start =
2418 btrfs_file_extent_calc_inline_size(0);
c5c9cd4d
SW		 2419 memmove(buf+start, buf+start+skip,
2420 datal);
2421 }
2422
2423 leaf = path->nodes[0];
2424 slot = path->slots[0];
2425 write_extent_buffer(leaf, buf,
2426 btrfs_item_ptr_offset(leaf, slot),
2427 size);
2428 inode_add_bytes(inode, datal);
f46b5a66 2429 }
c5c9cd4d
SW		 2430
2431 btrfs_mark_buffer_dirty(leaf);
b3b4aa74 2432 btrfs_release_path(path);
c5c9cd4d 2433
a22285a6 2434 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
b5384d48
SW		 2435
2436 /*
2437 * we round up to the block size at eof when
2438 * determining which extents to clone above,
2439 * but shouldn't round up the file size
2440 */
2441 endoff = new_key.offset + datal;
5f3888ff
LZ		 2442 if (endoff > destoff+olen)
2443 endoff = destoff+olen;
b5384d48
SW		 2444 if (endoff > inode->i_size)
2445 btrfs_i_size_write(inode, endoff);
2446
a22285a6
YZ		 2447 BTRFS_I(inode)->flags = BTRFS_I(src)->flags;
2448 ret = btrfs_update_inode(trans, root, inode);
2449 BUG_ON(ret);
2450 btrfs_end_transaction(trans, root);
2451 }
d397712b 2452next:
b3b4aa74 2453 btrfs_release_path(path);
f46b5a66 2454 key.offset++;
f46b5a66 2455 }
f46b5a66
CH		 2456 ret = 0;
2457out:
b3b4aa74 2458 btrfs_release_path(path);
c5c9cd4d 2459 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
f46b5a66
CH		 2460out_unlock:
2461 mutex_unlock(&src->i_mutex);
2462 mutex_unlock(&inode->i_mutex);
ae01a0ab
YZ		 2463 vfree(buf);
2464 btrfs_free_path(path);
f46b5a66
CH		 2465out_fput:
2466 fput(src_file);
ab67b7c1
YZ		 2467out_drop_write:
2468 mnt_drop_write(file->f_path.mnt);
f46b5a66
CH		 2469 return ret;
2470}
2471
7a865e8a 2472static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
c5c9cd4d
SW		 2473{
2474 struct btrfs_ioctl_clone_range_args args;
2475
7a865e8a 2476 if (copy_from_user(&args, argp, sizeof(args)))
c5c9cd4d
SW		 2477 return -EFAULT;
2478 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2479 args.src_length, args.dest_offset);
2480}
2481
f46b5a66
CH		 2482/*
2483 * there are many ways the trans_start and trans_end ioctls can lead
2484 * to deadlocks. They should only be used by applications that
2485 * basically own the machine, and have a very in depth understanding
2486 * of all the possible deadlocks and enospc problems.
2487 */
b2950863 2488static long btrfs_ioctl_trans_start(struct file *file)
f46b5a66
CH		 2489{
2490 struct inode *inode = fdentry(file)->d_inode;
2491 struct btrfs_root *root = BTRFS_I(inode)->root;
2492 struct btrfs_trans_handle *trans;
1ab86aed 2493 int ret;
f46b5a66 2494
1ab86aed 2495 ret = -EPERM;
df5b5520 2496 if (!capable(CAP_SYS_ADMIN))
1ab86aed 2497 goto out;
df5b5520 2498
1ab86aed
SW		 2499 ret = -EINPROGRESS;
2500 if (file->private_data)
f46b5a66 2501 goto out;
9ca9ee09 2502
b83cc969
LZ		 2503 ret = -EROFS;
2504 if (btrfs_root_readonly(root))
2505 goto out;
2506
c146afad
YZ		 2507 ret = mnt_want_write(file->f_path.mnt);
2508 if (ret)
2509 goto out;
2510
a4abeea4 2511 atomic_inc(&root->fs_info->open_ioctl_trans);
9ca9ee09 2512
1ab86aed 2513 ret = -ENOMEM;
7a7eaa40 2514 trans = btrfs_start_ioctl_transaction(root);
abd30bb0 2515 if (IS_ERR(trans))
1ab86aed
SW		 2516 goto out_drop;
2517
2518 file->private_data = trans;
2519 return 0;
2520
2521out_drop:
a4abeea4 2522 atomic_dec(&root->fs_info->open_ioctl_trans);
1ab86aed 2523 mnt_drop_write(file->f_path.mnt);
f46b5a66 2524out:
f46b5a66
CH		 2525 return ret;
2526}
2527
6ef5ed0d
JB		 2528static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2529{
2530 struct inode *inode = fdentry(file)->d_inode;
2531 struct btrfs_root *root = BTRFS_I(inode)->root;
2532 struct btrfs_root *new_root;
2533 struct btrfs_dir_item *di;
2534 struct btrfs_trans_handle *trans;
2535 struct btrfs_path *path;
2536 struct btrfs_key location;
2537 struct btrfs_disk_key disk_key;
2538 struct btrfs_super_block *disk_super;
2539 u64 features;
2540 u64 objectid = 0;
2541 u64 dir_id;
2542
2543 if (!capable(CAP_SYS_ADMIN))
2544 return -EPERM;
2545
2546 if (copy_from_user(&objectid, argp, sizeof(objectid)))
2547 return -EFAULT;
2548
2549 if (!objectid)
2550 objectid = root->root_key.objectid;
2551
2552 location.objectid = objectid;
2553 location.type = BTRFS_ROOT_ITEM_KEY;
2554 location.offset = (u64)-1;
2555
2556 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2557 if (IS_ERR(new_root))
2558 return PTR_ERR(new_root);
2559
2560 if (btrfs_root_refs(&new_root->root_item) == 0)
2561 return -ENOENT;
2562
2563 path = btrfs_alloc_path();
2564 if (!path)
2565 return -ENOMEM;
2566 path->leave_spinning = 1;
2567
2568 trans = btrfs_start_transaction(root, 1);
98d5dc13 2569 if (IS_ERR(trans)) {
6ef5ed0d 2570 btrfs_free_path(path);
98d5dc13 2571 return PTR_ERR(trans);
6ef5ed0d
JB		 2572 }
2573
2574 dir_id = btrfs_super_root_dir(&root->fs_info->super_copy);
2575 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2576 dir_id, "default", 7, 1);
cf1e99a4 2577 if (IS_ERR_OR_NULL(di)) {
6ef5ed0d
JB		 2578 btrfs_free_path(path);
2579 btrfs_end_transaction(trans, root);
2580 printk(KERN_ERR "Umm, you don't have the default dir item, "
2581 "this isn't going to work\n");
2582 return -ENOENT;
2583 }
2584
2585 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2586 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2587 btrfs_mark_buffer_dirty(path->nodes[0]);
2588 btrfs_free_path(path);
2589
2590 disk_super = &root->fs_info->super_copy;
2591 features = btrfs_super_incompat_flags(disk_super);
2592 if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) {
2593 features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL;
2594 btrfs_set_super_incompat_flags(disk_super, features);
2595 }
2596 btrfs_end_transaction(trans, root);
2597
2598 return 0;
2599}
2600
bf5fc093
JB		 2601static void get_block_group_info(struct list_head *groups_list,
2602 struct btrfs_ioctl_space_info *space)
2603{
2604 struct btrfs_block_group_cache *block_group;
2605
2606 space->total_bytes = 0;
2607 space->used_bytes = 0;
2608 space->flags = 0;
2609 list_for_each_entry(block_group, groups_list, list) {
2610 space->flags = block_group->flags;
2611 space->total_bytes += block_group->key.offset;
2612 space->used_bytes +=
2613 btrfs_block_group_used(&block_group->item);
2614 }
2615}
2616
1406e432
JB		 2617long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2618{
2619 struct btrfs_ioctl_space_args space_args;
2620 struct btrfs_ioctl_space_info space;
2621 struct btrfs_ioctl_space_info *dest;
7fde62bf 2622 struct btrfs_ioctl_space_info *dest_orig;
13f2696f 2623 struct btrfs_ioctl_space_info __user *user_dest;
1406e432 2624 struct btrfs_space_info *info;
bf5fc093
JB		 2625 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2626 BTRFS_BLOCK_GROUP_SYSTEM,
2627 BTRFS_BLOCK_GROUP_METADATA,
2628 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2629 int num_types = 4;
7fde62bf 2630 int alloc_size;
1406e432 2631 int ret = 0;
51788b1b 2632 u64 slot_count = 0;
bf5fc093 2633 int i, c;
1406e432
JB		 2634
2635 if (copy_from_user(&space_args,
2636 (struct btrfs_ioctl_space_args __user *)arg,
2637 sizeof(space_args)))
2638 return -EFAULT;
2639
bf5fc093
JB		 2640 for (i = 0; i < num_types; i++) {
2641 struct btrfs_space_info *tmp;
2642
2643 info = NULL;
2644 rcu_read_lock();
2645 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2646 list) {
2647 if (tmp->flags == types[i]) {
2648 info = tmp;
2649 break;
2650 }
2651 }
2652 rcu_read_unlock();
2653
2654 if (!info)
2655 continue;
2656
2657 down_read(&info->groups_sem);
2658 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2659 if (!list_empty(&info->block_groups[c]))
2660 slot_count++;
2661 }
2662 up_read(&info->groups_sem);
2663 }
7fde62bf
CM		 2664
2665 /* space_slots == 0 means they are asking for a count */
2666 if (space_args.space_slots == 0) {
2667 space_args.total_spaces = slot_count;
2668 goto out;
2669 }
bf5fc093 2670
51788b1b 2671 slot_count = min_t(u64, space_args.space_slots, slot_count);
bf5fc093 2672
7fde62bf 2673 alloc_size = sizeof(*dest) * slot_count;
bf5fc093 2674
7fde62bf
CM		 2675 /* we generally have at most 6 or so space infos, one for each raid
2676 * level. So, a whole page should be more than enough for everyone
2677 */
2678 if (alloc_size > PAGE_CACHE_SIZE)
2679 return -ENOMEM;
2680
1406e432 2681 space_args.total_spaces = 0;
7fde62bf
CM		 2682 dest = kmalloc(alloc_size, GFP_NOFS);
2683 if (!dest)
2684 return -ENOMEM;
2685 dest_orig = dest;
1406e432 2686
7fde62bf 2687 /* now we have a buffer to copy into */
bf5fc093
JB		 2688 for (i = 0; i < num_types; i++) {
2689 struct btrfs_space_info *tmp;
2690
51788b1b
DR		 2691 if (!slot_count)
2692 break;
2693
bf5fc093
JB		 2694 info = NULL;
2695 rcu_read_lock();
2696 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2697 list) {
2698 if (tmp->flags == types[i]) {
2699 info = tmp;
2700 break;
2701 }
2702 }
2703 rcu_read_unlock();
7fde62bf 2704
bf5fc093
JB		 2705 if (!info)
2706 continue;
2707 down_read(&info->groups_sem);
2708 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2709 if (!list_empty(&info->block_groups[c])) {
2710 get_block_group_info(&info->block_groups[c],
2711 &space);
2712 memcpy(dest, &space, sizeof(space));
2713 dest++;
2714 space_args.total_spaces++;
51788b1b 2715 slot_count--;
bf5fc093 2716 }
51788b1b
DR		 2717 if (!slot_count)
2718 break;
bf5fc093
JB		 2719 }
2720 up_read(&info->groups_sem);
1406e432 2721 }
1406e432 2722
7fde62bf
CM		 2723 user_dest = (struct btrfs_ioctl_space_info *)
2724 (arg + sizeof(struct btrfs_ioctl_space_args));
2725
2726 if (copy_to_user(user_dest, dest_orig, alloc_size))
2727 ret = -EFAULT;
2728
2729 kfree(dest_orig);
2730out:
2731 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
1406e432
JB		 2732 ret = -EFAULT;
2733
2734 return ret;
2735}
2736
f46b5a66
CH		 2737/*
2738 * there are many ways the trans_start and trans_end ioctls can lead
2739 * to deadlocks. They should only be used by applications that
2740 * basically own the machine, and have a very in depth understanding
2741 * of all the possible deadlocks and enospc problems.
2742 */
2743long btrfs_ioctl_trans_end(struct file *file)
2744{
2745 struct inode *inode = fdentry(file)->d_inode;
2746 struct btrfs_root *root = BTRFS_I(inode)->root;
2747 struct btrfs_trans_handle *trans;
f46b5a66 2748
f46b5a66 2749 trans = file->private_data;
1ab86aed
SW		 2750 if (!trans)
2751 return -EINVAL;
b214107e 2752 file->private_data = NULL;
9ca9ee09 2753
1ab86aed
SW		 2754 btrfs_end_transaction(trans, root);
2755
a4abeea4 2756 atomic_dec(&root->fs_info->open_ioctl_trans);
9ca9ee09 2757
cfc8ea87 2758 mnt_drop_write(file->f_path.mnt);
1ab86aed 2759 return 0;
f46b5a66
CH		 2760}
2761
46204592
SW		 2762static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
2763{
2764 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2765 struct btrfs_trans_handle *trans;
2766 u64 transid;
db5b493a 2767 int ret;
46204592
SW		 2768
2769 trans = btrfs_start_transaction(root, 0);
98d5dc13
TI		 2770 if (IS_ERR(trans))
2771 return PTR_ERR(trans);
46204592 2772 transid = trans->transid;
db5b493a 2773 ret = btrfs_commit_transaction_async(trans, root, 0);
8b2b2d3c
TI		 2774 if (ret) {
2775 btrfs_end_transaction(trans, root);
db5b493a 2776 return ret;
8b2b2d3c 2777 }
46204592
SW		 2778
2779 if (argp)
2780 if (copy_to_user(argp, &transid, sizeof(transid)))
2781 return -EFAULT;
2782 return 0;
2783}
2784
2785static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
2786{
2787 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2788 u64 transid;
2789
2790 if (argp) {
2791 if (copy_from_user(&transid, argp, sizeof(transid)))
2792 return -EFAULT;
2793 } else {
2794 transid = 0; /* current trans */
2795 }
2796 return btrfs_wait_for_commit(root, transid);
2797}
2798
475f6387
JS		 2799static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg)
2800{
2801 int ret;
2802 struct btrfs_ioctl_scrub_args *sa;
2803
2804 if (!capable(CAP_SYS_ADMIN))
2805 return -EPERM;
2806
2807 sa = memdup_user(arg, sizeof(*sa));
2808 if (IS_ERR(sa))
2809 return PTR_ERR(sa);
2810
2811 ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end,
8628764e 2812 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY);
475f6387
JS		 2813
2814 if (copy_to_user(arg, sa, sizeof(*sa)))
2815 ret = -EFAULT;
2816
2817 kfree(sa);
2818 return ret;
2819}
2820
2821static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
2822{
2823 if (!capable(CAP_SYS_ADMIN))
2824 return -EPERM;
2825
2826 return btrfs_scrub_cancel(root);
2827}
2828
2829static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
2830 void __user *arg)
2831{
2832 struct btrfs_ioctl_scrub_args *sa;
2833 int ret;
2834
2835 if (!capable(CAP_SYS_ADMIN))
2836 return -EPERM;
2837
2838 sa = memdup_user(arg, sizeof(*sa));
2839 if (IS_ERR(sa))
2840 return PTR_ERR(sa);
2841
2842 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
2843
2844 if (copy_to_user(arg, sa, sizeof(*sa)))
2845 ret = -EFAULT;
2846
2847 kfree(sa);
2848 return ret;
2849}
2850
f46b5a66
CH		 2851long btrfs_ioctl(struct file *file, unsigned int
2852 cmd, unsigned long arg)
2853{
2854 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
4bcabaa3 2855 void __user *argp = (void __user *)arg;
f46b5a66
CH		 2856
2857 switch (cmd) {
6cbff00f
CH		 2858 case FS_IOC_GETFLAGS:
2859 return btrfs_ioctl_getflags(file, argp);
2860 case FS_IOC_SETFLAGS:
2861 return btrfs_ioctl_setflags(file, argp);
2862 case FS_IOC_GETVERSION:
2863 return btrfs_ioctl_getversion(file, argp);
f7039b1d
LD		 2864 case FITRIM:
2865 return btrfs_ioctl_fitrim(file, argp);
f46b5a66 2866 case BTRFS_IOC_SNAP_CREATE:
fa0d2b9b 2867 return btrfs_ioctl_snap_create(file, argp, 0);
fdfb1e4f 2868 case BTRFS_IOC_SNAP_CREATE_V2:
fa0d2b9b 2869 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3de4586c 2870 case BTRFS_IOC_SUBVOL_CREATE:
fa0d2b9b 2871 return btrfs_ioctl_snap_create(file, argp, 1);
76dda93c
YZ		 2872 case BTRFS_IOC_SNAP_DESTROY:
2873 return btrfs_ioctl_snap_destroy(file, argp);
0caa102d
LZ		 2874 case BTRFS_IOC_SUBVOL_GETFLAGS:
2875 return btrfs_ioctl_subvol_getflags(file, argp);
2876 case BTRFS_IOC_SUBVOL_SETFLAGS:
2877 return btrfs_ioctl_subvol_setflags(file, argp);
6ef5ed0d
JB		 2878 case BTRFS_IOC_DEFAULT_SUBVOL:
2879 return btrfs_ioctl_default_subvol(file, argp);
f46b5a66 2880 case BTRFS_IOC_DEFRAG:
1e701a32
CM		 2881 return btrfs_ioctl_defrag(file, NULL);
2882 case BTRFS_IOC_DEFRAG_RANGE:
2883 return btrfs_ioctl_defrag(file, argp);
f46b5a66 2884 case BTRFS_IOC_RESIZE:
4bcabaa3 2885 return btrfs_ioctl_resize(root, argp);
f46b5a66 2886 case BTRFS_IOC_ADD_DEV:
4bcabaa3 2887 return btrfs_ioctl_add_dev(root, argp);
f46b5a66 2888 case BTRFS_IOC_RM_DEV:
4bcabaa3 2889 return btrfs_ioctl_rm_dev(root, argp);
475f6387
JS		 2890 case BTRFS_IOC_FS_INFO:
2891 return btrfs_ioctl_fs_info(root, argp);
2892 case BTRFS_IOC_DEV_INFO:
2893 return btrfs_ioctl_dev_info(root, argp);
f46b5a66
CH		 2894 case BTRFS_IOC_BALANCE:
2895 return btrfs_balance(root->fs_info->dev_root);
2896 case BTRFS_IOC_CLONE:
c5c9cd4d
SW		 2897 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
2898 case BTRFS_IOC_CLONE_RANGE:
7a865e8a 2899 return btrfs_ioctl_clone_range(file, argp);
f46b5a66
CH		 2900 case BTRFS_IOC_TRANS_START:
2901 return btrfs_ioctl_trans_start(file);
2902 case BTRFS_IOC_TRANS_END:
2903 return btrfs_ioctl_trans_end(file);
ac8e9819
CM		 2904 case BTRFS_IOC_TREE_SEARCH:
2905 return btrfs_ioctl_tree_search(file, argp);
2906 case BTRFS_IOC_INO_LOOKUP:
2907 return btrfs_ioctl_ino_lookup(file, argp);
1406e432
JB		 2908 case BTRFS_IOC_SPACE_INFO:
2909 return btrfs_ioctl_space_info(root, argp);
f46b5a66
CH		 2910 case BTRFS_IOC_SYNC:
2911 btrfs_sync_fs(file->f_dentry->d_sb, 1);
2912 return 0;
46204592
SW		 2913 case BTRFS_IOC_START_SYNC:
2914 return btrfs_ioctl_start_sync(file, argp);
2915 case BTRFS_IOC_WAIT_SYNC:
2916 return btrfs_ioctl_wait_sync(file, argp);
475f6387
JS		 2917 case BTRFS_IOC_SCRUB:
2918 return btrfs_ioctl_scrub(root, argp);
2919 case BTRFS_IOC_SCRUB_CANCEL:
2920 return btrfs_ioctl_scrub_cancel(root, argp);
2921 case BTRFS_IOC_SCRUB_PROGRESS:
2922 return btrfs_ioctl_scrub_progress(root, argp);
f46b5a66
CH		 2923 }
2924
2925 return -ENOTTY;
2926}
A mirror of Linus' kernel repository