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