projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Btrfs: don't save the inode cache if we are deleting this root
[linux-2.6-block.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
f46b5a66
CH		 485 list_add(&pending_snapshot->list,
486 &trans->transaction->pending_snapshots);
72fd032e
SW		 487 if (async_transid) {
488 *async_transid = trans->transid;
489 ret = btrfs_commit_transaction_async(trans,
490 root->fs_info->extent_root, 1);
491 } else {
492 ret = btrfs_commit_transaction(trans,
493 root->fs_info->extent_root);
494 }
2e4bfab9 495 BUG_ON(ret);
a22285a6
YZ		 496
497 ret = pending_snapshot->error;
498 if (ret)
499 goto fail;
500
66b4ffd1
JB		 501 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
502 if (ret)
503 goto fail;
f46b5a66 504
6a912213
JB		 505 parent = dget_parent(dentry);
506 inode = btrfs_lookup_dentry(parent->d_inode, dentry);
507 dput(parent);
2e4bfab9
YZ		 508 if (IS_ERR(inode)) {
509 ret = PTR_ERR(inode);
510 goto fail;
511 }
512 BUG_ON(!inode);
513 d_instantiate(dentry, inode);
514 ret = 0;
515fail:
a22285a6 516 kfree(pending_snapshot);
f46b5a66
CH		 517 return ret;
518}
519
4260f7c7
SW		 520/* copy of check_sticky in fs/namei.c()
521* It's inline, so penalty for filesystems that don't use sticky bit is
522* minimal.
523*/
524static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
525{
526 uid_t fsuid = current_fsuid();
527
528 if (!(dir->i_mode & S_ISVTX))
529 return 0;
530 if (inode->i_uid == fsuid)
531 return 0;
532 if (dir->i_uid == fsuid)
533 return 0;
534 return !capable(CAP_FOWNER);
535}
536
537/* copy of may_delete in fs/namei.c()
538 * Check whether we can remove a link victim from directory dir, check
539 * whether the type of victim is right.
540 * 1. We can't do it if dir is read-only (done in permission())
541 * 2. We should have write and exec permissions on dir
542 * 3. We can't remove anything from append-only dir
543 * 4. We can't do anything with immutable dir (done in permission())
544 * 5. If the sticky bit on dir is set we should either
545 * a. be owner of dir, or
546 * b. be owner of victim, or
547 * c. have CAP_FOWNER capability
548 * 6. If the victim is append-only or immutable we can't do antyhing with
549 * links pointing to it.
550 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
551 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
552 * 9. We can't remove a root or mountpoint.
553 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
554 * nfs_async_unlink().
555 */
556
557static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
558{
559 int error;
560
561 if (!victim->d_inode)
562 return -ENOENT;
563
564 BUG_ON(victim->d_parent->d_inode != dir);
565 audit_inode_child(victim, dir);
566
567 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
568 if (error)
569 return error;
570 if (IS_APPEND(dir))
571 return -EPERM;
572 if (btrfs_check_sticky(dir, victim->d_inode)||
573 IS_APPEND(victim->d_inode)||
574 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
575 return -EPERM;
576 if (isdir) {
577 if (!S_ISDIR(victim->d_inode->i_mode))
578 return -ENOTDIR;
579 if (IS_ROOT(victim))
580 return -EBUSY;
581 } else if (S_ISDIR(victim->d_inode->i_mode))
582 return -EISDIR;
583 if (IS_DEADDIR(dir))
584 return -ENOENT;
585 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
586 return -EBUSY;
587 return 0;
588}
589
cb8e7090
CH		 590/* copy of may_create in fs/namei.c() */
591static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
592{
593 if (child->d_inode)
594 return -EEXIST;
595 if (IS_DEADDIR(dir))
596 return -ENOENT;
597 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
598}
599
600/*
601 * Create a new subvolume below @parent. This is largely modeled after
602 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
603 * inside this filesystem so it's quite a bit simpler.
604 */
76dda93c
YZ		 605static noinline int btrfs_mksubvol(struct path *parent,
606 char *name, int namelen,
72fd032e 607 struct btrfs_root *snap_src,
b83cc969 608 u64 *async_transid, bool readonly)
cb8e7090 609{
76dda93c 610 struct inode *dir = parent->dentry->d_inode;
cb8e7090
CH		 611 struct dentry *dentry;
612 int error;
613
76dda93c 614 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
cb8e7090
CH		 615
616 dentry = lookup_one_len(name, parent->dentry, namelen);
617 error = PTR_ERR(dentry);
618 if (IS_ERR(dentry))
619 goto out_unlock;
620
621 error = -EEXIST;
622 if (dentry->d_inode)
623 goto out_dput;
624
cb8e7090
CH		 625 error = mnt_want_write(parent->mnt);
626 if (error)
627 goto out_dput;
628
76dda93c 629 error = btrfs_may_create(dir, dentry);
cb8e7090
CH		 630 if (error)
631 goto out_drop_write;
632
76dda93c
YZ		 633 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
634
635 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
636 goto out_up_read;
637
3de4586c 638 if (snap_src) {
72fd032e 639 error = create_snapshot(snap_src, dentry,
b83cc969 640 name, namelen, async_transid, readonly);
3de4586c 641 } else {
76dda93c 642 error = create_subvol(BTRFS_I(dir)->root, dentry,
72fd032e 643 name, namelen, async_transid);
3de4586c 644 }
76dda93c
YZ		 645 if (!error)
646 fsnotify_mkdir(dir, dentry);
647out_up_read:
648 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
cb8e7090
CH		 649out_drop_write:
650 mnt_drop_write(parent->mnt);
651out_dput:
652 dput(dentry);
653out_unlock:
76dda93c 654 mutex_unlock(&dir->i_mutex);
cb8e7090
CH		 655 return error;
656}
657
4cb5300b
CM		 658/*
659 * When we're defragging a range, we don't want to kick it off again
660 * if it is really just waiting for delalloc to send it down.
661 * If we find a nice big extent or delalloc range for the bytes in the
662 * file you want to defrag, we return 0 to let you know to skip this
663 * part of the file
664 */
665static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
666{
667 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
668 struct extent_map *em = NULL;
669 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
670 u64 end;
671
672 read_lock(&em_tree->lock);
673 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
674 read_unlock(&em_tree->lock);
675
676 if (em) {
677 end = extent_map_end(em);
678 free_extent_map(em);
679 if (end - offset > thresh)
680 return 0;
681 }
682 /* if we already have a nice delalloc here, just stop */
683 thresh /= 2;
684 end = count_range_bits(io_tree, &offset, offset + thresh,
685 thresh, EXTENT_DELALLOC, 1);
686 if (end >= thresh)
687 return 0;
688 return 1;
689}
690
691/*
692 * helper function to walk through a file and find extents
693 * newer than a specific transid, and smaller than thresh.
694 *
695 * This is used by the defragging code to find new and small
696 * extents
697 */
698static int find_new_extents(struct btrfs_root *root,
699 struct inode *inode, u64 newer_than,
700 u64 *off, int thresh)
701{
702 struct btrfs_path *path;
703 struct btrfs_key min_key;
704 struct btrfs_key max_key;
705 struct extent_buffer *leaf;
706 struct btrfs_file_extent_item *extent;
707 int type;
708 int ret;
709
710 path = btrfs_alloc_path();
711 if (!path)
712 return -ENOMEM;
713
714 min_key.objectid = inode->i_ino;
715 min_key.type = BTRFS_EXTENT_DATA_KEY;
716 min_key.offset = *off;
717
718 max_key.objectid = inode->i_ino;
719 max_key.type = (u8)-1;
720 max_key.offset = (u64)-1;
721
722 path->keep_locks = 1;
723
724 while(1) {
725 ret = btrfs_search_forward(root, &min_key, &max_key,
726 path, 0, newer_than);
727 if (ret != 0)
728 goto none;
729 if (min_key.objectid != inode->i_ino)
730 goto none;
731 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
732 goto none;
733
734 leaf = path->nodes[0];
735 extent = btrfs_item_ptr(leaf, path->slots[0],
736 struct btrfs_file_extent_item);
737
738 type = btrfs_file_extent_type(leaf, extent);
739 if (type == BTRFS_FILE_EXTENT_REG &&
740 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
741 check_defrag_in_cache(inode, min_key.offset, thresh)) {
742 *off = min_key.offset;
743 btrfs_free_path(path);
744 return 0;
745 }
746
747 if (min_key.offset == (u64)-1)
748 goto none;
749
750 min_key.offset++;
751 btrfs_release_path(path);
752 }
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
CM		 850 struct extent_state *cached_state = NULL;
851
852 if (isize == 0)
853 return 0;
854 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
855
856 ret = btrfs_delalloc_reserve_space(inode,
857 num_pages << PAGE_CACHE_SHIFT);
858 if (ret)
859 return ret;
860again:
861 ret = 0;
862 i_done = 0;
863
864 /* step one, lock all the pages */
865 for (i = 0; i < num_pages; i++) {
866 struct page *page;
867 page = grab_cache_page(inode->i_mapping,
868 start_index + i);
869 if (!page)
870 break;
871
872 if (!PageUptodate(page)) {
873 btrfs_readpage(NULL, page);
874 lock_page(page);
875 if (!PageUptodate(page)) {
876 unlock_page(page);
877 page_cache_release(page);
878 ret = -EIO;
879 break;
880 }
881 }
882 isize = i_size_read(inode);
883 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
884 if (!isize || page->index > file_end ||
885 page->mapping != inode->i_mapping) {
886 /* whoops, we blew past eof, skip this page */
887 unlock_page(page);
888 page_cache_release(page);
889 break;
890 }
891 pages[i] = page;
892 i_done++;
893 }
894 if (!i_done || ret)
895 goto out;
896
897 if (!(inode->i_sb->s_flags & MS_ACTIVE))
898 goto out;
899
900 /*
901 * so now we have a nice long stream of locked
902 * and up to date pages, lets wait on them
903 */
904 for (i = 0; i < i_done; i++)
905 wait_on_page_writeback(pages[i]);
906
907 page_start = page_offset(pages[0]);
908 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
909
910 lock_extent_bits(&BTRFS_I(inode)->io_tree,
911 page_start, page_end - 1, 0, &cached_state,
912 GFP_NOFS);
913 ordered = btrfs_lookup_first_ordered_extent(inode, page_end - 1);
914 if (ordered &&
915 ordered->file_offset + ordered->len > page_start &&
916 ordered->file_offset < page_end) {
917 btrfs_put_ordered_extent(ordered);
918 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
919 page_start, page_end - 1,
920 &cached_state, GFP_NOFS);
921 for (i = 0; i < i_done; i++) {
922 unlock_page(pages[i]);
923 page_cache_release(pages[i]);
924 }
925 btrfs_wait_ordered_range(inode, page_start,
926 page_end - page_start);
927 goto again;
928 }
929 if (ordered)
930 btrfs_put_ordered_extent(ordered);
931
932 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
933 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
934 EXTENT_DO_ACCOUNTING, 0, 0, &cached_state,
935 GFP_NOFS);
936
937 if (i_done != num_pages) {
938 atomic_inc(&BTRFS_I(inode)->outstanding_extents);
939 btrfs_delalloc_release_space(inode,
940 (num_pages - i_done) << PAGE_CACHE_SHIFT);
941 }
942
943
944 btrfs_set_extent_delalloc(inode, page_start, page_end - 1,
945 &cached_state);
946
947 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
948 page_start, page_end - 1, &cached_state,
949 GFP_NOFS);
950
951 for (i = 0; i < i_done; i++) {
952 clear_page_dirty_for_io(pages[i]);
953 ClearPageChecked(pages[i]);
954 set_page_extent_mapped(pages[i]);
955 set_page_dirty(pages[i]);
956 unlock_page(pages[i]);
957 page_cache_release(pages[i]);
958 }
959 return i_done;
960out:
961 for (i = 0; i < i_done; i++) {
962 unlock_page(pages[i]);
963 page_cache_release(pages[i]);
964 }
965 btrfs_delalloc_release_space(inode, num_pages << PAGE_CACHE_SHIFT);
966 return ret;
967
968}
969
970int btrfs_defrag_file(struct inode *inode, struct file *file,
971 struct btrfs_ioctl_defrag_range_args *range,
972 u64 newer_than, unsigned long max_to_defrag)
973{
974 struct btrfs_root *root = BTRFS_I(inode)->root;
1a419d85 975 struct btrfs_super_block *disk_super;
4cb5300b 976 struct file_ra_state *ra = NULL;
f46b5a66 977 unsigned long last_index;
1a419d85 978 u64 features;
940100a4
CM		 979 u64 last_len = 0;
980 u64 skip = 0;
981 u64 defrag_end = 0;
4cb5300b
CM		 982 u64 newer_off = range->start;
983 int newer_left = 0;
f46b5a66
CH		 984 unsigned long i;
985 int ret;
4cb5300b 986 int defrag_count = 0;
1a419d85 987 int compress_type = BTRFS_COMPRESS_ZLIB;
4cb5300b
CM		 988 int extent_thresh = range->extent_thresh;
989 int newer_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
990 u64 new_align = ~((u64)128 * 1024 - 1);
991 struct page **pages = NULL;
992
993 if (extent_thresh == 0)
994 extent_thresh = 256 * 1024;
1a419d85
LZ		 995
996 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
997 if (range->compress_type > BTRFS_COMPRESS_TYPES)
998 return -EINVAL;
999 if (range->compress_type)
1000 compress_type = range->compress_type;
1001 }
f46b5a66 1002
940100a4
CM		 1003 if (inode->i_size == 0)
1004 return 0;
1005
4cb5300b
CM		 1006 /*
1007 * if we were not given a file, allocate a readahead
1008 * context
1009 */
1010 if (!file) {
1011 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1012 if (!ra)
1013 return -ENOMEM;
1014 file_ra_state_init(ra, inode->i_mapping);
1015 } else {
1016 ra = &file->f_ra;
1017 }
1018
1019 pages = kmalloc(sizeof(struct page *) * newer_cluster,
1020 GFP_NOFS);
1021 if (!pages) {
1022 ret = -ENOMEM;
1023 goto out_ra;
1024 }
1025
1026 /* find the last page to defrag */
1e701a32
CM		 1027 if (range->start + range->len > range->start) {
1028 last_index = min_t(u64, inode->i_size - 1,
1029 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1030 } else {
1031 last_index = (inode->i_size - 1) >> PAGE_CACHE_SHIFT;
1032 }
1033
4cb5300b
CM		 1034 if (newer_than) {
1035 ret = find_new_extents(root, inode, newer_than,
1036 &newer_off, 64 * 1024);
1037 if (!ret) {
1038 range->start = newer_off;
1039 /*
1040 * we always align our defrag to help keep
1041 * the extents in the file evenly spaced
1042 */
1043 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1044 newer_left = newer_cluster;
1045 } else
1046 goto out_ra;
1047 } else {
1048 i = range->start >> PAGE_CACHE_SHIFT;
1049 }
1050 if (!max_to_defrag)
1051 max_to_defrag = last_index - 1;
1052
1053 while (i <= last_index && defrag_count < max_to_defrag) {
1054 /*
1055 * make sure we stop running if someone unmounts
1056 * the FS
1057 */
1058 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1059 break;
1060
1061 if (!newer_than &&
1062 !should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1e701a32 1063 PAGE_CACHE_SIZE,
4cb5300b 1064 extent_thresh,
1e701a32 1065 &last_len, &skip,
940100a4
CM		 1066 &defrag_end)) {
1067 unsigned long next;
1068 /*
1069 * the should_defrag function tells us how much to skip
1070 * bump our counter by the suggested amount
1071 */
1072 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1073 i = max(i + 1, next);
1074 continue;
1075 }
1e701a32 1076 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1a419d85 1077 BTRFS_I(inode)->force_compress = compress_type;
940100a4 1078
4cb5300b 1079 btrfs_force_ra(inode->i_mapping, ra, file, i, newer_cluster);
940100a4 1080
4cb5300b
CM		 1081 ret = cluster_pages_for_defrag(inode, pages, i, newer_cluster);
1082 if (ret < 0)
1083 goto out_ra;
1084
1085 defrag_count += ret;
1086 balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret);
1087 i += ret;
1088
1089 if (newer_than) {
1090 if (newer_off == (u64)-1)
1091 break;
1092
1093 newer_off = max(newer_off + 1,
1094 (u64)i << PAGE_CACHE_SHIFT);
1095
1096 ret = find_new_extents(root, inode,
1097 newer_than, &newer_off,
1098 64 * 1024);
1099 if (!ret) {
1100 range->start = newer_off;
1101 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1102 newer_left = newer_cluster;
1103 } else {
1104 break;
f46b5a66 1105 }
4cb5300b
CM		 1106 } else {
1107 i++;
f46b5a66 1108 }
f46b5a66
CH		 1109 }
1110
1e701a32
CM		 1111 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1112 filemap_flush(inode->i_mapping);
1113
1114 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1115 /* the filemap_flush will queue IO into the worker threads, but
1116 * we have to make sure the IO is actually started and that
1117 * ordered extents get created before we return
1118 */
1119 atomic_inc(&root->fs_info->async_submit_draining);
1120 while (atomic_read(&root->fs_info->nr_async_submits) ||
1121 atomic_read(&root->fs_info->async_delalloc_pages)) {
1122 wait_event(root->fs_info->async_submit_wait,
1123 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1124 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1125 }
1126 atomic_dec(&root->fs_info->async_submit_draining);
1127
1128 mutex_lock(&inode->i_mutex);
261507a0 1129 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1e701a32
CM		 1130 mutex_unlock(&inode->i_mutex);
1131 }
1132
1a419d85
LZ		 1133 disk_super = &root->fs_info->super_copy;
1134 features = btrfs_super_incompat_flags(disk_super);
1135 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1136 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
1137 btrfs_set_super_incompat_flags(disk_super, features);
1138 }
1139
4cb5300b
CM		 1140 if (!file)
1141 kfree(ra);
1142 return defrag_count;
940100a4 1143
4cb5300b
CM		 1144out_ra:
1145 if (!file)
1146 kfree(ra);
1147 kfree(pages);
940100a4 1148 return ret;
f46b5a66
CH		 1149}
1150
76dda93c
YZ		 1151static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
1152 void __user *arg)
f46b5a66
CH		 1153{
1154 u64 new_size;
1155 u64 old_size;
1156 u64 devid = 1;
1157 struct btrfs_ioctl_vol_args *vol_args;
1158 struct btrfs_trans_handle *trans;
1159 struct btrfs_device *device = NULL;
1160 char *sizestr;
1161 char *devstr = NULL;
1162 int ret = 0;
f46b5a66
CH		 1163 int mod = 0;
1164
c146afad
YZ		 1165 if (root->fs_info->sb->s_flags & MS_RDONLY)
1166 return -EROFS;
1167
e441d54d
CM		 1168 if (!capable(CAP_SYS_ADMIN))
1169 return -EPERM;
1170
dae7b665
LZ		 1171 vol_args = memdup_user(arg, sizeof(*vol_args));
1172 if (IS_ERR(vol_args))
1173 return PTR_ERR(vol_args);
5516e595
MF		 1174
1175 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 1176
7d9eb12c 1177 mutex_lock(&root->fs_info->volume_mutex);
f46b5a66
CH		 1178 sizestr = vol_args->name;
1179 devstr = strchr(sizestr, ':');
1180 if (devstr) {
1181 char *end;
1182 sizestr = devstr + 1;
1183 *devstr = '\0';
1184 devstr = vol_args->name;
1185 devid = simple_strtoull(devstr, &end, 10);
21380931
JB		 1186 printk(KERN_INFO "resizing devid %llu\n",
1187 (unsigned long long)devid);
f46b5a66 1188 }
2b82032c 1189 device = btrfs_find_device(root, devid, NULL, NULL);
f46b5a66 1190 if (!device) {
21380931
JB		 1191 printk(KERN_INFO "resizer unable to find device %llu\n",
1192 (unsigned long long)devid);
f46b5a66
CH		 1193 ret = -EINVAL;
1194 goto out_unlock;
1195 }
1196 if (!strcmp(sizestr, "max"))
1197 new_size = device->bdev->bd_inode->i_size;
1198 else {
1199 if (sizestr[0] == '-') {
1200 mod = -1;
1201 sizestr++;
1202 } else if (sizestr[0] == '+') {
1203 mod = 1;
1204 sizestr++;
1205 }
91748467 1206 new_size = memparse(sizestr, NULL);
f46b5a66
CH		 1207 if (new_size == 0) {
1208 ret = -EINVAL;
1209 goto out_unlock;
1210 }
1211 }
1212
1213 old_size = device->total_bytes;
1214
1215 if (mod < 0) {
1216 if (new_size > old_size) {
1217 ret = -EINVAL;
1218 goto out_unlock;
1219 }
1220 new_size = old_size - new_size;
1221 } else if (mod > 0) {
1222 new_size = old_size + new_size;
1223 }
1224
1225 if (new_size < 256 * 1024 * 1024) {
1226 ret = -EINVAL;
1227 goto out_unlock;
1228 }
1229 if (new_size > device->bdev->bd_inode->i_size) {
1230 ret = -EFBIG;
1231 goto out_unlock;
1232 }
1233
1234 do_div(new_size, root->sectorsize);
1235 new_size *= root->sectorsize;
1236
1237 printk(KERN_INFO "new size for %s is %llu\n",
1238 device->name, (unsigned long long)new_size);
1239
1240 if (new_size > old_size) {
a22285a6 1241 trans = btrfs_start_transaction(root, 0);
98d5dc13
TI		 1242 if (IS_ERR(trans)) {
1243 ret = PTR_ERR(trans);
1244 goto out_unlock;
1245 }
f46b5a66
CH		 1246 ret = btrfs_grow_device(trans, device, new_size);
1247 btrfs_commit_transaction(trans, root);
1248 } else {
1249 ret = btrfs_shrink_device(device, new_size);
1250 }
1251
1252out_unlock:
7d9eb12c 1253 mutex_unlock(&root->fs_info->volume_mutex);
f46b5a66
CH		 1254 kfree(vol_args);
1255 return ret;
1256}
1257
72fd032e
SW		 1258static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1259 char *name,
1260 unsigned long fd,
1261 int subvol,
b83cc969
LZ		 1262 u64 *transid,
1263 bool readonly)
f46b5a66 1264{
cb8e7090 1265 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3de4586c 1266 struct file *src_file;
f46b5a66 1267 int namelen;
3de4586c 1268 int ret = 0;
f46b5a66 1269
c146afad
YZ		 1270 if (root->fs_info->sb->s_flags & MS_RDONLY)
1271 return -EROFS;
1272
72fd032e
SW		 1273 namelen = strlen(name);
1274 if (strchr(name, '/')) {
f46b5a66
CH		 1275 ret = -EINVAL;
1276 goto out;
1277 }
1278
3de4586c 1279 if (subvol) {
72fd032e 1280 ret = btrfs_mksubvol(&file->f_path, name, namelen,
b83cc969 1281 NULL, transid, readonly);
cb8e7090 1282 } else {
3de4586c 1283 struct inode *src_inode;
72fd032e 1284 src_file = fget(fd);
3de4586c
CM		 1285 if (!src_file) {
1286 ret = -EINVAL;
1287 goto out;
1288 }
1289
1290 src_inode = src_file->f_path.dentry->d_inode;
1291 if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
d397712b
CM		 1292 printk(KERN_INFO "btrfs: Snapshot src from "
1293 "another FS\n");
3de4586c
CM		 1294 ret = -EINVAL;
1295 fput(src_file);
1296 goto out;
1297 }
72fd032e
SW		 1298 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1299 BTRFS_I(src_inode)->root,
b83cc969 1300 transid, readonly);
3de4586c 1301 fput(src_file);
cb8e7090 1302 }
f46b5a66 1303out:
72fd032e
SW		 1304 return ret;
1305}
1306
1307static noinline int btrfs_ioctl_snap_create(struct file *file,
fa0d2b9b 1308 void __user *arg, int subvol)
72fd032e 1309{
fa0d2b9b 1310 struct btrfs_ioctl_vol_args *vol_args;
72fd032e
SW		 1311 int ret;
1312
fa0d2b9b
LZ		 1313 vol_args = memdup_user(arg, sizeof(*vol_args));
1314 if (IS_ERR(vol_args))
1315 return PTR_ERR(vol_args);
1316 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
72fd032e 1317
fa0d2b9b 1318 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
b83cc969
LZ		 1319 vol_args->fd, subvol,
1320 NULL, false);
fdfb1e4f 1321
fa0d2b9b
LZ		 1322 kfree(vol_args);
1323 return ret;
1324}
fdfb1e4f 1325
fa0d2b9b
LZ		 1326static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1327 void __user *arg, int subvol)
1328{
1329 struct btrfs_ioctl_vol_args_v2 *vol_args;
1330 int ret;
1331 u64 transid = 0;
1332 u64 *ptr = NULL;
b83cc969 1333 bool readonly = false;
75eaa0e2 1334
fa0d2b9b
LZ		 1335 vol_args = memdup_user(arg, sizeof(*vol_args));
1336 if (IS_ERR(vol_args))
1337 return PTR_ERR(vol_args);
1338 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
75eaa0e2 1339
b83cc969
LZ		 1340 if (vol_args->flags &
1341 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY)) {
1342 ret = -EOPNOTSUPP;
fa0d2b9b 1343 goto out;
72fd032e 1344 }
fa0d2b9b
LZ		 1345
1346 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1347 ptr = &transid;
b83cc969
LZ		 1348 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1349 readonly = true;
fa0d2b9b
LZ		 1350
1351 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
b83cc969
LZ		 1352 vol_args->fd, subvol,
1353 ptr, readonly);
fa0d2b9b
LZ		 1354
1355 if (ret == 0 && ptr &&
1356 copy_to_user(arg +
1357 offsetof(struct btrfs_ioctl_vol_args_v2,
1358 transid), ptr, sizeof(*ptr)))
1359 ret = -EFAULT;
fdfb1e4f 1360out:
f46b5a66
CH		 1361 kfree(vol_args);
1362 return ret;
1363}
1364
0caa102d
LZ		 1365static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1366 void __user *arg)
1367{
1368 struct inode *inode = fdentry(file)->d_inode;
1369 struct btrfs_root *root = BTRFS_I(inode)->root;
1370 int ret = 0;
1371 u64 flags = 0;
1372
33345d01 1373 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
0caa102d
LZ		 1374 return -EINVAL;
1375
1376 down_read(&root->fs_info->subvol_sem);
1377 if (btrfs_root_readonly(root))
1378 flags |= BTRFS_SUBVOL_RDONLY;
1379 up_read(&root->fs_info->subvol_sem);
1380
1381 if (copy_to_user(arg, &flags, sizeof(flags)))
1382 ret = -EFAULT;
1383
1384 return ret;
1385}
1386
1387static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1388 void __user *arg)
1389{
1390 struct inode *inode = fdentry(file)->d_inode;
1391 struct btrfs_root *root = BTRFS_I(inode)->root;
1392 struct btrfs_trans_handle *trans;
1393 u64 root_flags;
1394 u64 flags;
1395 int ret = 0;
1396
1397 if (root->fs_info->sb->s_flags & MS_RDONLY)
1398 return -EROFS;
1399
33345d01 1400 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
0caa102d
LZ		 1401 return -EINVAL;
1402
1403 if (copy_from_user(&flags, arg, sizeof(flags)))
1404 return -EFAULT;
1405
b4dc2b8c 1406 if (flags & BTRFS_SUBVOL_CREATE_ASYNC)
0caa102d
LZ		 1407 return -EINVAL;
1408
1409 if (flags & ~BTRFS_SUBVOL_RDONLY)
1410 return -EOPNOTSUPP;
1411
2e149670 1412 if (!inode_owner_or_capable(inode))
b4dc2b8c
LZ		 1413 return -EACCES;
1414
0caa102d
LZ		 1415 down_write(&root->fs_info->subvol_sem);
1416
1417 /* nothing to do */
1418 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1419 goto out;
1420
1421 root_flags = btrfs_root_flags(&root->root_item);
1422 if (flags & BTRFS_SUBVOL_RDONLY)
1423 btrfs_set_root_flags(&root->root_item,
1424 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1425 else
1426 btrfs_set_root_flags(&root->root_item,
1427 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1428
1429 trans = btrfs_start_transaction(root, 1);
1430 if (IS_ERR(trans)) {
1431 ret = PTR_ERR(trans);
1432 goto out_reset;
1433 }
1434
b4dc2b8c 1435 ret = btrfs_update_root(trans, root->fs_info->tree_root,
0caa102d
LZ		 1436 &root->root_key, &root->root_item);
1437
1438 btrfs_commit_transaction(trans, root);
1439out_reset:
1440 if (ret)
1441 btrfs_set_root_flags(&root->root_item, root_flags);
1442out:
1443 up_write(&root->fs_info->subvol_sem);
1444 return ret;
1445}
1446
76dda93c
YZ		 1447/*
1448 * helper to check if the subvolume references other subvolumes
1449 */
1450static noinline int may_destroy_subvol(struct btrfs_root *root)
1451{
1452 struct btrfs_path *path;
1453 struct btrfs_key key;
1454 int ret;
1455
1456 path = btrfs_alloc_path();
1457 if (!path)
1458 return -ENOMEM;
1459
1460 key.objectid = root->root_key.objectid;
1461 key.type = BTRFS_ROOT_REF_KEY;
1462 key.offset = (u64)-1;
1463
1464 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1465 &key, path, 0, 0);
1466 if (ret < 0)
1467 goto out;
1468 BUG_ON(ret == 0);
1469
1470 ret = 0;
1471 if (path->slots[0] > 0) {
1472 path->slots[0]--;
1473 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1474 if (key.objectid == root->root_key.objectid &&
1475 key.type == BTRFS_ROOT_REF_KEY)
1476 ret = -ENOTEMPTY;
1477 }
1478out:
1479 btrfs_free_path(path);
1480 return ret;
1481}
1482
ac8e9819
CM		 1483static noinline int key_in_sk(struct btrfs_key *key,
1484 struct btrfs_ioctl_search_key *sk)
1485{
abc6e134
CM		 1486 struct btrfs_key test;
1487 int ret;
1488
1489 test.objectid = sk->min_objectid;
1490 test.type = sk->min_type;
1491 test.offset = sk->min_offset;
1492
1493 ret = btrfs_comp_cpu_keys(key, &test);
1494 if (ret < 0)
ac8e9819 1495 return 0;
abc6e134
CM		 1496
1497 test.objectid = sk->max_objectid;
1498 test.type = sk->max_type;
1499 test.offset = sk->max_offset;
1500
1501 ret = btrfs_comp_cpu_keys(key, &test);
1502 if (ret > 0)
ac8e9819
CM		 1503 return 0;
1504 return 1;
1505}
1506
1507static noinline int copy_to_sk(struct btrfs_root *root,
1508 struct btrfs_path *path,
1509 struct btrfs_key *key,
1510 struct btrfs_ioctl_search_key *sk,
1511 char *buf,
1512 unsigned long *sk_offset,
1513 int *num_found)
1514{
1515 u64 found_transid;
1516 struct extent_buffer *leaf;
1517 struct btrfs_ioctl_search_header sh;
1518 unsigned long item_off;
1519 unsigned long item_len;
1520 int nritems;
1521 int i;
1522 int slot;
ac8e9819
CM		 1523 int ret = 0;
1524
1525 leaf = path->nodes[0];
1526 slot = path->slots[0];
1527 nritems = btrfs_header_nritems(leaf);
1528
1529 if (btrfs_header_generation(leaf) > sk->max_transid) {
1530 i = nritems;
1531 goto advance_key;
1532 }
1533 found_transid = btrfs_header_generation(leaf);
1534
1535 for (i = slot; i < nritems; i++) {
1536 item_off = btrfs_item_ptr_offset(leaf, i);
1537 item_len = btrfs_item_size_nr(leaf, i);
1538
1539 if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1540 item_len = 0;
1541
1542 if (sizeof(sh) + item_len + *sk_offset >
1543 BTRFS_SEARCH_ARGS_BUFSIZE) {
1544 ret = 1;
1545 goto overflow;
1546 }
1547
1548 btrfs_item_key_to_cpu(leaf, key, i);
1549 if (!key_in_sk(key, sk))
1550 continue;
1551
1552 sh.objectid = key->objectid;
1553 sh.offset = key->offset;
1554 sh.type = key->type;
1555 sh.len = item_len;
1556 sh.transid = found_transid;
1557
1558 /* copy search result header */
1559 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1560 *sk_offset += sizeof(sh);
1561
1562 if (item_len) {
1563 char *p = buf + *sk_offset;
1564 /* copy the item */
1565 read_extent_buffer(leaf, p,
1566 item_off, item_len);
1567 *sk_offset += item_len;
ac8e9819 1568 }
e2156867 1569 (*num_found)++;
ac8e9819
CM		 1570
1571 if (*num_found >= sk->nr_items)
1572 break;
1573 }
1574advance_key:
abc6e134
CM		 1575 ret = 0;
1576 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
ac8e9819 1577 key->offset++;
abc6e134
CM		 1578 else if (key->type < (u8)-1 && key->type < sk->max_type) {
1579 key->offset = 0;
ac8e9819 1580 key->type++;
abc6e134
CM		 1581 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1582 key->offset = 0;
1583 key->type = 0;
ac8e9819 1584 key->objectid++;
abc6e134
CM		 1585 } else
1586 ret = 1;
ac8e9819 1587overflow:
ac8e9819
CM		 1588 return ret;
1589}
1590
1591static noinline int search_ioctl(struct inode *inode,
1592 struct btrfs_ioctl_search_args *args)
1593{
1594 struct btrfs_root *root;
1595 struct btrfs_key key;
1596 struct btrfs_key max_key;
1597 struct btrfs_path *path;
1598 struct btrfs_ioctl_search_key *sk = &args->key;
1599 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1600 int ret;
1601 int num_found = 0;
1602 unsigned long sk_offset = 0;
1603
1604 path = btrfs_alloc_path();
1605 if (!path)
1606 return -ENOMEM;
1607
1608 if (sk->tree_id == 0) {
1609 /* search the root of the inode that was passed */
1610 root = BTRFS_I(inode)->root;
1611 } else {
1612 key.objectid = sk->tree_id;
1613 key.type = BTRFS_ROOT_ITEM_KEY;
1614 key.offset = (u64)-1;
1615 root = btrfs_read_fs_root_no_name(info, &key);
1616 if (IS_ERR(root)) {
1617 printk(KERN_ERR "could not find root %llu\n",
1618 sk->tree_id);
1619 btrfs_free_path(path);
1620 return -ENOENT;
1621 }
1622 }
1623
1624 key.objectid = sk->min_objectid;
1625 key.type = sk->min_type;
1626 key.offset = sk->min_offset;
1627
1628 max_key.objectid = sk->max_objectid;
1629 max_key.type = sk->max_type;
1630 max_key.offset = sk->max_offset;
1631
1632 path->keep_locks = 1;
1633
1634 while(1) {
1635 ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1636 sk->min_transid);
1637 if (ret != 0) {
1638 if (ret > 0)
1639 ret = 0;
1640 goto err;
1641 }
1642 ret = copy_to_sk(root, path, &key, sk, args->buf,
1643 &sk_offset, &num_found);
b3b4aa74 1644 btrfs_release_path(path);
ac8e9819
CM		 1645 if (ret || num_found >= sk->nr_items)
1646 break;
1647
1648 }
1649 ret = 0;
1650err:
1651 sk->nr_items = num_found;
1652 btrfs_free_path(path);
1653 return ret;
1654}
1655
1656static noinline int btrfs_ioctl_tree_search(struct file *file,
1657 void __user *argp)
1658{
1659 struct btrfs_ioctl_search_args *args;
1660 struct inode *inode;
1661 int ret;
1662
1663 if (!capable(CAP_SYS_ADMIN))
1664 return -EPERM;
1665
2354d08f
JL		 1666 args = memdup_user(argp, sizeof(*args));
1667 if (IS_ERR(args))
1668 return PTR_ERR(args);
ac8e9819 1669
ac8e9819
CM		 1670 inode = fdentry(file)->d_inode;
1671 ret = search_ioctl(inode, args);
1672 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1673 ret = -EFAULT;
1674 kfree(args);
1675 return ret;
1676}
1677
98d377a0 1678/*
ac8e9819
CM		 1679 * Search INODE_REFs to identify path name of 'dirid' directory
1680 * in a 'tree_id' tree. and sets path name to 'name'.
1681 */
98d377a0
TH		 1682static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1683 u64 tree_id, u64 dirid, char *name)
1684{
1685 struct btrfs_root *root;
1686 struct btrfs_key key;
ac8e9819 1687 char *ptr;
98d377a0
TH		 1688 int ret = -1;
1689 int slot;
1690 int len;
1691 int total_len = 0;
1692 struct btrfs_inode_ref *iref;
1693 struct extent_buffer *l;
1694 struct btrfs_path *path;
1695
1696 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1697 name[0]='\0';
1698 return 0;
1699 }
1700
1701 path = btrfs_alloc_path();
1702 if (!path)
1703 return -ENOMEM;
1704
ac8e9819 1705 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
98d377a0
TH		 1706
1707 key.objectid = tree_id;
1708 key.type = BTRFS_ROOT_ITEM_KEY;
1709 key.offset = (u64)-1;
1710 root = btrfs_read_fs_root_no_name(info, &key);
1711 if (IS_ERR(root)) {
1712 printk(KERN_ERR "could not find root %llu\n", tree_id);
8ad6fcab
CM		 1713 ret = -ENOENT;
1714 goto out;
98d377a0
TH		 1715 }
1716
1717 key.objectid = dirid;
1718 key.type = BTRFS_INODE_REF_KEY;
8ad6fcab 1719 key.offset = (u64)-1;
98d377a0
TH		 1720
1721 while(1) {
1722 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1723 if (ret < 0)
1724 goto out;
1725
1726 l = path->nodes[0];
1727 slot = path->slots[0];
8ad6fcab
CM		 1728 if (ret > 0 && slot > 0)
1729 slot--;
98d377a0
TH		 1730 btrfs_item_key_to_cpu(l, &key, slot);
1731
1732 if (ret > 0 && (key.objectid != dirid ||
ac8e9819
CM		 1733 key.type != BTRFS_INODE_REF_KEY)) {
1734 ret = -ENOENT;
98d377a0 1735 goto out;
ac8e9819 1736 }
98d377a0
TH		 1737
1738 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1739 len = btrfs_inode_ref_name_len(l, iref);
1740 ptr -= len + 1;
1741 total_len += len + 1;
ac8e9819 1742 if (ptr < name)
98d377a0
TH		 1743 goto out;
1744
1745 *(ptr + len) = '/';
1746 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1747
1748 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1749 break;
1750
b3b4aa74 1751 btrfs_release_path(path);
98d377a0 1752 key.objectid = key.offset;
8ad6fcab 1753 key.offset = (u64)-1;
98d377a0
TH		 1754 dirid = key.objectid;
1755
1756 }
ac8e9819 1757 if (ptr < name)
98d377a0 1758 goto out;
ac8e9819 1759 memcpy(name, ptr, total_len);
98d377a0
TH		 1760 name[total_len]='\0';
1761 ret = 0;
1762out:
1763 btrfs_free_path(path);
ac8e9819
CM		 1764 return ret;
1765}
1766
1767static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1768 void __user *argp)
1769{
1770 struct btrfs_ioctl_ino_lookup_args *args;
1771 struct inode *inode;
1772 int ret;
1773
1774 if (!capable(CAP_SYS_ADMIN))
1775 return -EPERM;
1776
2354d08f
JL		 1777 args = memdup_user(argp, sizeof(*args));
1778 if (IS_ERR(args))
1779 return PTR_ERR(args);
c2b96929 1780
ac8e9819
CM		 1781 inode = fdentry(file)->d_inode;
1782
1b53ac4d
CM		 1783 if (args->treeid == 0)
1784 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1785
ac8e9819
CM		 1786 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1787 args->treeid, args->objectid,
1788 args->name);
1789
1790 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1791 ret = -EFAULT;
1792
1793 kfree(args);
98d377a0
TH		 1794 return ret;
1795}
1796
76dda93c
YZ		 1797static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1798 void __user *arg)
1799{
1800 struct dentry *parent = fdentry(file);
1801 struct dentry *dentry;
1802 struct inode *dir = parent->d_inode;
1803 struct inode *inode;
1804 struct btrfs_root *root = BTRFS_I(dir)->root;
1805 struct btrfs_root *dest = NULL;
1806 struct btrfs_ioctl_vol_args *vol_args;
1807 struct btrfs_trans_handle *trans;
1808 int namelen;
1809 int ret;
1810 int err = 0;
1811
76dda93c
YZ		 1812 vol_args = memdup_user(arg, sizeof(*vol_args));
1813 if (IS_ERR(vol_args))
1814 return PTR_ERR(vol_args);
1815
1816 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1817 namelen = strlen(vol_args->name);
1818 if (strchr(vol_args->name, '/') ||
1819 strncmp(vol_args->name, "..", namelen) == 0) {
1820 err = -EINVAL;
1821 goto out;
1822 }
1823
1824 err = mnt_want_write(file->f_path.mnt);
1825 if (err)
1826 goto out;
1827
1828 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
1829 dentry = lookup_one_len(vol_args->name, parent, namelen);
1830 if (IS_ERR(dentry)) {
1831 err = PTR_ERR(dentry);
1832 goto out_unlock_dir;
1833 }
1834
1835 if (!dentry->d_inode) {
1836 err = -ENOENT;
1837 goto out_dput;
1838 }
1839
1840 inode = dentry->d_inode;
4260f7c7
SW		 1841 dest = BTRFS_I(inode)->root;
1842 if (!capable(CAP_SYS_ADMIN)){
1843 /*
1844 * Regular user. Only allow this with a special mount
1845 * option, when the user has write+exec access to the
1846 * subvol root, and when rmdir(2) would have been
1847 * allowed.
1848 *
1849 * Note that this is _not_ check that the subvol is
1850 * empty or doesn't contain data that we wouldn't
1851 * otherwise be able to delete.
1852 *
1853 * Users who want to delete empty subvols should try
1854 * rmdir(2).
1855 */
1856 err = -EPERM;
1857 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1858 goto out_dput;
1859
1860 /*
1861 * Do not allow deletion if the parent dir is the same
1862 * as the dir to be deleted. That means the ioctl
1863 * must be called on the dentry referencing the root
1864 * of the subvol, not a random directory contained
1865 * within it.
1866 */
1867 err = -EINVAL;
1868 if (root == dest)
1869 goto out_dput;
1870
1871 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
1872 if (err)
1873 goto out_dput;
1874
1875 /* check if subvolume may be deleted by a non-root user */
1876 err = btrfs_may_delete(dir, dentry, 1);
1877 if (err)
1878 goto out_dput;
1879 }
1880
33345d01 1881 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
76dda93c
YZ		 1882 err = -EINVAL;
1883 goto out_dput;
1884 }
1885
76dda93c
YZ		 1886 mutex_lock(&inode->i_mutex);
1887 err = d_invalidate(dentry);
1888 if (err)
1889 goto out_unlock;
1890
1891 down_write(&root->fs_info->subvol_sem);
1892
1893 err = may_destroy_subvol(dest);
1894 if (err)
1895 goto out_up_write;
1896
a22285a6
YZ		 1897 trans = btrfs_start_transaction(root, 0);
1898 if (IS_ERR(trans)) {
1899 err = PTR_ERR(trans);
d327099a 1900 goto out_up_write;
a22285a6
YZ		 1901 }
1902 trans->block_rsv = &root->fs_info->global_block_rsv;
1903
76dda93c
YZ		 1904 ret = btrfs_unlink_subvol(trans, root, dir,
1905 dest->root_key.objectid,
1906 dentry->d_name.name,
1907 dentry->d_name.len);
1908 BUG_ON(ret);
1909
1910 btrfs_record_root_in_trans(trans, dest);
1911
1912 memset(&dest->root_item.drop_progress, 0,
1913 sizeof(dest->root_item.drop_progress));
1914 dest->root_item.drop_level = 0;
1915 btrfs_set_root_refs(&dest->root_item, 0);
1916
d68fc57b
YZ		 1917 if (!xchg(&dest->orphan_item_inserted, 1)) {
1918 ret = btrfs_insert_orphan_item(trans,
1919 root->fs_info->tree_root,
1920 dest->root_key.objectid);
1921 BUG_ON(ret);
1922 }
76dda93c 1923
531cb13f 1924 ret = btrfs_end_transaction(trans, root);
76dda93c
YZ		 1925 BUG_ON(ret);
1926 inode->i_flags |= S_DEAD;
1927out_up_write:
1928 up_write(&root->fs_info->subvol_sem);
1929out_unlock:
1930 mutex_unlock(&inode->i_mutex);
1931 if (!err) {
efefb143 1932 shrink_dcache_sb(root->fs_info->sb);
76dda93c
YZ		 1933 btrfs_invalidate_inodes(dest);
1934 d_delete(dentry);
1935 }
1936out_dput:
1937 dput(dentry);
1938out_unlock_dir:
1939 mutex_unlock(&dir->i_mutex);
1940 mnt_drop_write(file->f_path.mnt);
1941out:
1942 kfree(vol_args);
1943 return err;
1944}
1945
1e701a32 1946static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
f46b5a66
CH		 1947{
1948 struct inode *inode = fdentry(file)->d_inode;
1949 struct btrfs_root *root = BTRFS_I(inode)->root;
1e701a32 1950 struct btrfs_ioctl_defrag_range_args *range;
c146afad
YZ		 1951 int ret;
1952
b83cc969
LZ		 1953 if (btrfs_root_readonly(root))
1954 return -EROFS;
1955
c146afad
YZ		 1956 ret = mnt_want_write(file->f_path.mnt);
1957 if (ret)
1958 return ret;
f46b5a66
CH		 1959
1960 switch (inode->i_mode & S_IFMT) {
1961 case S_IFDIR:
e441d54d
CM		 1962 if (!capable(CAP_SYS_ADMIN)) {
1963 ret = -EPERM;
1964 goto out;
1965 }
8929ecfa
YZ		 1966 ret = btrfs_defrag_root(root, 0);
1967 if (ret)
1968 goto out;
1969 ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
f46b5a66
CH		 1970 break;
1971 case S_IFREG:
e441d54d
CM		 1972 if (!(file->f_mode & FMODE_WRITE)) {
1973 ret = -EINVAL;
1974 goto out;
1975 }
1e701a32
CM		 1976
1977 range = kzalloc(sizeof(*range), GFP_KERNEL);
1978 if (!range) {
1979 ret = -ENOMEM;
1980 goto out;
1981 }
1982
1983 if (argp) {
1984 if (copy_from_user(range, argp,
1985 sizeof(*range))) {
1986 ret = -EFAULT;
1987 kfree(range);
683be16e 1988 goto out;
1e701a32
CM		 1989 }
1990 /* compression requires us to start the IO */
1991 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1992 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
1993 range->extent_thresh = (u32)-1;
1994 }
1995 } else {
1996 /* the rest are all set to zero by kzalloc */
1997 range->len = (u64)-1;
1998 }
4cb5300b
CM		 1999 ret = btrfs_defrag_file(fdentry(file)->d_inode, file,
2000 range, 0, 0);
2001 if (ret > 0)
2002 ret = 0;
1e701a32 2003 kfree(range);
f46b5a66 2004 break;
8929ecfa
YZ		 2005 default:
2006 ret = -EINVAL;
f46b5a66 2007 }
e441d54d 2008out:
ab67b7c1 2009 mnt_drop_write(file->f_path.mnt);
e441d54d 2010 return ret;
f46b5a66
CH		 2011}
2012
b2950863 2013static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
f46b5a66
CH		 2014{
2015 struct btrfs_ioctl_vol_args *vol_args;
2016 int ret;
2017
e441d54d
CM		 2018 if (!capable(CAP_SYS_ADMIN))
2019 return -EPERM;
2020
dae7b665
LZ		 2021 vol_args = memdup_user(arg, sizeof(*vol_args));
2022 if (IS_ERR(vol_args))
2023 return PTR_ERR(vol_args);
f46b5a66 2024
5516e595 2025 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66
CH		 2026 ret = btrfs_init_new_device(root, vol_args->name);
2027
f46b5a66
CH		 2028 kfree(vol_args);
2029 return ret;
2030}
2031
b2950863 2032static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
f46b5a66
CH		 2033{
2034 struct btrfs_ioctl_vol_args *vol_args;
2035 int ret;
2036
e441d54d
CM		 2037 if (!capable(CAP_SYS_ADMIN))
2038 return -EPERM;
2039
c146afad
YZ		 2040 if (root->fs_info->sb->s_flags & MS_RDONLY)
2041 return -EROFS;
2042
dae7b665
LZ		 2043 vol_args = memdup_user(arg, sizeof(*vol_args));
2044 if (IS_ERR(vol_args))
2045 return PTR_ERR(vol_args);
f46b5a66 2046
5516e595 2047 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66
CH		 2048 ret = btrfs_rm_device(root, vol_args->name);
2049
f46b5a66
CH		 2050 kfree(vol_args);
2051 return ret;
2052}
2053
475f6387
JS		 2054static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2055{
2056 struct btrfs_ioctl_fs_info_args fi_args;
2057 struct btrfs_device *device;
2058 struct btrfs_device *next;
2059 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2060
2061 if (!capable(CAP_SYS_ADMIN))
2062 return -EPERM;
2063
2064 fi_args.num_devices = fs_devices->num_devices;
2065 fi_args.max_id = 0;
2066 memcpy(&fi_args.fsid, root->fs_info->fsid, sizeof(fi_args.fsid));
2067
2068 mutex_lock(&fs_devices->device_list_mutex);
2069 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2070 if (device->devid > fi_args.max_id)
2071 fi_args.max_id = device->devid;
2072 }
2073 mutex_unlock(&fs_devices->device_list_mutex);
2074
2075 if (copy_to_user(arg, &fi_args, sizeof(fi_args)))
2076 return -EFAULT;
2077
2078 return 0;
2079}
2080
2081static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2082{
2083 struct btrfs_ioctl_dev_info_args *di_args;
2084 struct btrfs_device *dev;
2085 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2086 int ret = 0;
2087 char *s_uuid = NULL;
2088 char empty_uuid[BTRFS_UUID_SIZE] = {0};
2089
2090 if (!capable(CAP_SYS_ADMIN))
2091 return -EPERM;
2092
2093 di_args = memdup_user(arg, sizeof(*di_args));
2094 if (IS_ERR(di_args))
2095 return PTR_ERR(di_args);
2096
2097 if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2098 s_uuid = di_args->uuid;
2099
2100 mutex_lock(&fs_devices->device_list_mutex);
2101 dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL);
2102 mutex_unlock(&fs_devices->device_list_mutex);
2103
2104 if (!dev) {
2105 ret = -ENODEV;
2106 goto out;
2107 }
2108
2109 di_args->devid = dev->devid;
2110 di_args->bytes_used = dev->bytes_used;
2111 di_args->total_bytes = dev->total_bytes;
2112 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2113 strncpy(di_args->path, dev->name, sizeof(di_args->path));
2114
2115out:
2116 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2117 ret = -EFAULT;
2118
2119 kfree(di_args);
2120 return ret;
2121}
2122
76dda93c
YZ		 2123static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2124 u64 off, u64 olen, u64 destoff)
f46b5a66
CH		 2125{
2126 struct inode *inode = fdentry(file)->d_inode;
2127 struct btrfs_root *root = BTRFS_I(inode)->root;
2128 struct file *src_file;
2129 struct inode *src;
2130 struct btrfs_trans_handle *trans;
f46b5a66 2131 struct btrfs_path *path;
f46b5a66 2132 struct extent_buffer *leaf;
ae01a0ab
YZ		 2133 char *buf;
2134 struct btrfs_key key;
f46b5a66
CH		 2135 u32 nritems;
2136 int slot;
ae01a0ab 2137 int ret;
c5c9cd4d
SW		 2138 u64 len = olen;
2139 u64 bs = root->fs_info->sb->s_blocksize;
2140 u64 hint_byte;
d20f7043 2141
c5c9cd4d
SW		 2142 /*
2143 * TODO:
2144 * - split compressed inline extents. annoying: we need to
2145 * decompress into destination's address_space (the file offset
2146 * may change, so source mapping won't do), then recompress (or
2147 * otherwise reinsert) a subrange.
2148 * - allow ranges within the same file to be cloned (provided
2149 * they don't overlap)?
2150 */
2151
e441d54d 2152 /* the destination must be opened for writing */
2ebc3464 2153 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
e441d54d
CM		 2154 return -EINVAL;
2155
b83cc969
LZ		 2156 if (btrfs_root_readonly(root))
2157 return -EROFS;
2158
c146afad
YZ		 2159 ret = mnt_want_write(file->f_path.mnt);
2160 if (ret)
2161 return ret;
2162
c5c9cd4d 2163 src_file = fget(srcfd);
ab67b7c1
YZ		 2164 if (!src_file) {
2165 ret = -EBADF;
2166 goto out_drop_write;
2167 }
5dc64164 2168
f46b5a66
CH		 2169 src = src_file->f_dentry->d_inode;
2170
c5c9cd4d
SW		 2171 ret = -EINVAL;
2172 if (src == inode)
2173 goto out_fput;
2174
5dc64164
DR		 2175 /* the src must be open for reading */
2176 if (!(src_file->f_mode & FMODE_READ))
2177 goto out_fput;
2178
ae01a0ab
YZ		 2179 ret = -EISDIR;
2180 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2181 goto out_fput;
2182
f46b5a66 2183 ret = -EXDEV;
ae01a0ab
YZ		 2184 if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
2185 goto out_fput;
2186
2187 ret = -ENOMEM;
2188 buf = vmalloc(btrfs_level_size(root, 0));
2189 if (!buf)
2190 goto out_fput;
2191
2192 path = btrfs_alloc_path();
2193 if (!path) {
2194 vfree(buf);
f46b5a66 2195 goto out_fput;
ae01a0ab
YZ		 2196 }
2197 path->reada = 2;
f46b5a66
CH		 2198
2199 if (inode < src) {
fccdae43
SW		 2200 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2201 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
f46b5a66 2202 } else {
fccdae43
SW		 2203 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2204 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
f46b5a66
CH		 2205 }
2206
c5c9cd4d
SW		 2207 /* determine range to clone */
2208 ret = -EINVAL;
2ebc3464 2209 if (off + len > src->i_size || off + len < off)
f46b5a66 2210 goto out_unlock;
c5c9cd4d
SW		 2211 if (len == 0)
2212 olen = len = src->i_size - off;
2213 /* if we extend to eof, continue to block boundary */
2214 if (off + len == src->i_size)
2a6b8dae 2215 len = ALIGN(src->i_size, bs) - off;
c5c9cd4d
SW		 2216
2217 /* verify the end result is block aligned */
2a6b8dae
LZ		 2218 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2219 !IS_ALIGNED(destoff, bs))
c5c9cd4d
SW		 2220 goto out_unlock;
2221
f46b5a66
CH		 2222 /* do any pending delalloc/csum calc on src, one way or
2223 another, and lock file content */
2224 while (1) {
31840ae1 2225 struct btrfs_ordered_extent *ordered;
c5c9cd4d 2226 lock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
9a019196
SW		 2227 ordered = btrfs_lookup_first_ordered_extent(src, off+len);
2228 if (!ordered &&
2229 !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len,
2230 EXTENT_DELALLOC, 0, NULL))
f46b5a66 2231 break;
c5c9cd4d 2232 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
ae01a0ab
YZ		 2233 if (ordered)
2234 btrfs_put_ordered_extent(ordered);
9a019196 2235 btrfs_wait_ordered_range(src, off, len);
f46b5a66
CH		 2236 }
2237
c5c9cd4d 2238 /* clone data */
33345d01 2239 key.objectid = btrfs_ino(src);
ae01a0ab
YZ		 2240 key.type = BTRFS_EXTENT_DATA_KEY;
2241 key.offset = 0;
f46b5a66
CH		 2242
2243 while (1) {
2244 /*
2245 * note the key will change type as we walk through the
2246 * tree.
2247 */
a22285a6 2248 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
f46b5a66
CH		 2249 if (ret < 0)
2250 goto out;
2251
ae01a0ab
YZ		 2252 nritems = btrfs_header_nritems(path->nodes[0]);
2253 if (path->slots[0] >= nritems) {
f46b5a66
CH		 2254 ret = btrfs_next_leaf(root, path);
2255 if (ret < 0)
2256 goto out;
2257 if (ret > 0)
2258 break;
ae01a0ab 2259 nritems = btrfs_header_nritems(path->nodes[0]);
f46b5a66
CH		 2260 }
2261 leaf = path->nodes[0];
2262 slot = path->slots[0];
f46b5a66 2263
ae01a0ab 2264 btrfs_item_key_to_cpu(leaf, &key, slot);
d20f7043 2265 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
33345d01 2266 key.objectid != btrfs_ino(src))
f46b5a66
CH		 2267 break;
2268
c5c9cd4d
SW		 2269 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2270 struct btrfs_file_extent_item *extent;
2271 int type;
31840ae1
ZY		 2272 u32 size;
2273 struct btrfs_key new_key;
c5c9cd4d
SW		 2274 u64 disko = 0, diskl = 0;
2275 u64 datao = 0, datal = 0;
2276 u8 comp;
b5384d48 2277 u64 endoff;
31840ae1
ZY		 2278
2279 size = btrfs_item_size_nr(leaf, slot);
2280 read_extent_buffer(leaf, buf,
2281 btrfs_item_ptr_offset(leaf, slot),
2282 size);
c5c9cd4d
SW		 2283
2284 extent = btrfs_item_ptr(leaf, slot,
2285 struct btrfs_file_extent_item);
2286 comp = btrfs_file_extent_compression(leaf, extent);
2287 type = btrfs_file_extent_type(leaf, extent);
c8a894d7
CM		 2288 if (type == BTRFS_FILE_EXTENT_REG ||
2289 type == BTRFS_FILE_EXTENT_PREALLOC) {
d397712b
CM		 2290 disko = btrfs_file_extent_disk_bytenr(leaf,
2291 extent);
2292 diskl = btrfs_file_extent_disk_num_bytes(leaf,
2293 extent);
c5c9cd4d 2294 datao = btrfs_file_extent_offset(leaf, extent);
d397712b
CM		 2295 datal = btrfs_file_extent_num_bytes(leaf,
2296 extent);
c5c9cd4d
SW		 2297 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2298 /* take upper bound, may be compressed */
2299 datal = btrfs_file_extent_ram_bytes(leaf,
2300 extent);
2301 }
b3b4aa74 2302 btrfs_release_path(path);
31840ae1 2303
050006a7 2304 if (key.offset + datal <= off ||
c5c9cd4d
SW		 2305 key.offset >= off+len)
2306 goto next;
2307
31840ae1 2308 memcpy(&new_key, &key, sizeof(new_key));
33345d01 2309 new_key.objectid = btrfs_ino(inode);
4d728ec7
LZ		 2310 if (off <= key.offset)
2311 new_key.offset = key.offset + destoff - off;
2312 else
2313 new_key.offset = destoff;
31840ae1 2314
a22285a6
YZ		 2315 trans = btrfs_start_transaction(root, 1);
2316 if (IS_ERR(trans)) {
2317 ret = PTR_ERR(trans);
2318 goto out;
2319 }
2320
c8a894d7
CM		 2321 if (type == BTRFS_FILE_EXTENT_REG ||
2322 type == BTRFS_FILE_EXTENT_PREALLOC) {
a22285a6
YZ		 2323 if (off > key.offset) {
2324 datao += off - key.offset;
2325 datal -= off - key.offset;
2326 }
2327
2328 if (key.offset + datal > off + len)
2329 datal = off + len - key.offset;
2330
2331 ret = btrfs_drop_extents(trans, inode,
2332 new_key.offset,
2333 new_key.offset + datal,
2334 &hint_byte, 1);
2335 BUG_ON(ret);
2336
c5c9cd4d
SW		 2337 ret = btrfs_insert_empty_item(trans, root, path,
2338 &new_key, size);
a22285a6 2339 BUG_ON(ret);
c5c9cd4d
SW		 2340
2341 leaf = path->nodes[0];
2342 slot = path->slots[0];
2343 write_extent_buffer(leaf, buf,
31840ae1
ZY		 2344 btrfs_item_ptr_offset(leaf, slot),
2345 size);
ae01a0ab 2346
c5c9cd4d 2347 extent = btrfs_item_ptr(leaf, slot,
f46b5a66 2348 struct btrfs_file_extent_item);
c5c9cd4d 2349
c5c9cd4d
SW		 2350 /* disko == 0 means it's a hole */
2351 if (!disko)
2352 datao = 0;
c5c9cd4d
SW		 2353
2354 btrfs_set_file_extent_offset(leaf, extent,
2355 datao);
2356 btrfs_set_file_extent_num_bytes(leaf, extent,
2357 datal);
2358 if (disko) {
2359 inode_add_bytes(inode, datal);
ae01a0ab 2360 ret = btrfs_inc_extent_ref(trans, root,
5d4f98a2
YZ		 2361 disko, diskl, 0,
2362 root->root_key.objectid,
33345d01 2363 btrfs_ino(inode),
5d4f98a2 2364 new_key.offset - datao);
31840ae1 2365 BUG_ON(ret);
f46b5a66 2366 }
c5c9cd4d
SW		 2367 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2368 u64 skip = 0;
2369 u64 trim = 0;
2370 if (off > key.offset) {
2371 skip = off - key.offset;
2372 new_key.offset += skip;
2373 }
d397712b 2374
c5c9cd4d
SW		 2375 if (key.offset + datal > off+len)
2376 trim = key.offset + datal - (off+len);
d397712b 2377
c5c9cd4d 2378 if (comp && (skip || trim)) {
c5c9cd4d 2379 ret = -EINVAL;
a22285a6 2380 btrfs_end_transaction(trans, root);
c5c9cd4d
SW		 2381 goto out;
2382 }
2383 size -= skip + trim;
2384 datal -= skip + trim;
a22285a6
YZ		 2385
2386 ret = btrfs_drop_extents(trans, inode,
2387 new_key.offset,
2388 new_key.offset + datal,
2389 &hint_byte, 1);
2390 BUG_ON(ret);
2391
c5c9cd4d
SW		 2392 ret = btrfs_insert_empty_item(trans, root, path,
2393 &new_key, size);
a22285a6 2394 BUG_ON(ret);
c5c9cd4d
SW		 2395
2396 if (skip) {
d397712b
CM		 2397 u32 start =
2398 btrfs_file_extent_calc_inline_size(0);
c5c9cd4d
SW		 2399 memmove(buf+start, buf+start+skip,
2400 datal);
2401 }
2402
2403 leaf = path->nodes[0];
2404 slot = path->slots[0];
2405 write_extent_buffer(leaf, buf,
2406 btrfs_item_ptr_offset(leaf, slot),
2407 size);
2408 inode_add_bytes(inode, datal);
f46b5a66 2409 }
c5c9cd4d
SW		 2410
2411 btrfs_mark_buffer_dirty(leaf);
b3b4aa74 2412 btrfs_release_path(path);
c5c9cd4d 2413
a22285a6 2414 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
b5384d48
SW		 2415
2416 /*
2417 * we round up to the block size at eof when
2418 * determining which extents to clone above,
2419 * but shouldn't round up the file size
2420 */
2421 endoff = new_key.offset + datal;
5f3888ff
LZ		 2422 if (endoff > destoff+olen)
2423 endoff = destoff+olen;
b5384d48
SW		 2424 if (endoff > inode->i_size)
2425 btrfs_i_size_write(inode, endoff);
2426
a22285a6
YZ		 2427 BTRFS_I(inode)->flags = BTRFS_I(src)->flags;
2428 ret = btrfs_update_inode(trans, root, inode);
2429 BUG_ON(ret);
2430 btrfs_end_transaction(trans, root);
2431 }
d397712b 2432next:
b3b4aa74 2433 btrfs_release_path(path);
f46b5a66 2434 key.offset++;
f46b5a66 2435 }
f46b5a66
CH		 2436 ret = 0;
2437out:
b3b4aa74 2438 btrfs_release_path(path);
c5c9cd4d 2439 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
f46b5a66
CH		 2440out_unlock:
2441 mutex_unlock(&src->i_mutex);
2442 mutex_unlock(&inode->i_mutex);
ae01a0ab
YZ		 2443 vfree(buf);
2444 btrfs_free_path(path);
f46b5a66
CH		 2445out_fput:
2446 fput(src_file);
ab67b7c1
YZ		 2447out_drop_write:
2448 mnt_drop_write(file->f_path.mnt);
f46b5a66
CH		 2449 return ret;
2450}
2451
7a865e8a 2452static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
c5c9cd4d
SW		 2453{
2454 struct btrfs_ioctl_clone_range_args args;
2455
7a865e8a 2456 if (copy_from_user(&args, argp, sizeof(args)))
c5c9cd4d
SW		 2457 return -EFAULT;
2458 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2459 args.src_length, args.dest_offset);
2460}
2461
f46b5a66
CH		 2462/*
2463 * there are many ways the trans_start and trans_end ioctls can lead
2464 * to deadlocks. They should only be used by applications that
2465 * basically own the machine, and have a very in depth understanding
2466 * of all the possible deadlocks and enospc problems.
2467 */
b2950863 2468static long btrfs_ioctl_trans_start(struct file *file)
f46b5a66
CH		 2469{
2470 struct inode *inode = fdentry(file)->d_inode;
2471 struct btrfs_root *root = BTRFS_I(inode)->root;
2472 struct btrfs_trans_handle *trans;
1ab86aed 2473 int ret;
f46b5a66 2474
1ab86aed 2475 ret = -EPERM;
df5b5520 2476 if (!capable(CAP_SYS_ADMIN))
1ab86aed 2477 goto out;
df5b5520 2478
1ab86aed
SW		 2479 ret = -EINPROGRESS;
2480 if (file->private_data)
f46b5a66 2481 goto out;
9ca9ee09 2482
b83cc969
LZ		 2483 ret = -EROFS;
2484 if (btrfs_root_readonly(root))
2485 goto out;
2486
c146afad
YZ		 2487 ret = mnt_want_write(file->f_path.mnt);
2488 if (ret)
2489 goto out;
2490
a4abeea4 2491 atomic_inc(&root->fs_info->open_ioctl_trans);
9ca9ee09 2492
1ab86aed 2493 ret = -ENOMEM;
7a7eaa40 2494 trans = btrfs_start_ioctl_transaction(root);
abd30bb0 2495 if (IS_ERR(trans))
1ab86aed
SW		 2496 goto out_drop;
2497
2498 file->private_data = trans;
2499 return 0;
2500
2501out_drop:
a4abeea4 2502 atomic_dec(&root->fs_info->open_ioctl_trans);
1ab86aed 2503 mnt_drop_write(file->f_path.mnt);
f46b5a66 2504out:
f46b5a66
CH		 2505 return ret;
2506}
2507
6ef5ed0d
JB		 2508static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2509{
2510 struct inode *inode = fdentry(file)->d_inode;
2511 struct btrfs_root *root = BTRFS_I(inode)->root;
2512 struct btrfs_root *new_root;
2513 struct btrfs_dir_item *di;
2514 struct btrfs_trans_handle *trans;
2515 struct btrfs_path *path;
2516 struct btrfs_key location;
2517 struct btrfs_disk_key disk_key;
2518 struct btrfs_super_block *disk_super;
2519 u64 features;
2520 u64 objectid = 0;
2521 u64 dir_id;
2522
2523 if (!capable(CAP_SYS_ADMIN))
2524 return -EPERM;
2525
2526 if (copy_from_user(&objectid, argp, sizeof(objectid)))
2527 return -EFAULT;
2528
2529 if (!objectid)
2530 objectid = root->root_key.objectid;
2531
2532 location.objectid = objectid;
2533 location.type = BTRFS_ROOT_ITEM_KEY;
2534 location.offset = (u64)-1;
2535
2536 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2537 if (IS_ERR(new_root))
2538 return PTR_ERR(new_root);
2539
2540 if (btrfs_root_refs(&new_root->root_item) == 0)
2541 return -ENOENT;
2542
2543 path = btrfs_alloc_path();
2544 if (!path)
2545 return -ENOMEM;
2546 path->leave_spinning = 1;
2547
2548 trans = btrfs_start_transaction(root, 1);
98d5dc13 2549 if (IS_ERR(trans)) {
6ef5ed0d 2550 btrfs_free_path(path);
98d5dc13 2551 return PTR_ERR(trans);
6ef5ed0d
JB		 2552 }
2553
2554 dir_id = btrfs_super_root_dir(&root->fs_info->super_copy);
2555 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2556 dir_id, "default", 7, 1);
cf1e99a4 2557 if (IS_ERR_OR_NULL(di)) {
6ef5ed0d
JB		 2558 btrfs_free_path(path);
2559 btrfs_end_transaction(trans, root);
2560 printk(KERN_ERR "Umm, you don't have the default dir item, "
2561 "this isn't going to work\n");
2562 return -ENOENT;
2563 }
2564
2565 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2566 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2567 btrfs_mark_buffer_dirty(path->nodes[0]);
2568 btrfs_free_path(path);
2569
2570 disk_super = &root->fs_info->super_copy;
2571 features = btrfs_super_incompat_flags(disk_super);
2572 if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) {
2573 features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL;
2574 btrfs_set_super_incompat_flags(disk_super, features);
2575 }
2576 btrfs_end_transaction(trans, root);
2577
2578 return 0;
2579}
2580
bf5fc093
JB		 2581static void get_block_group_info(struct list_head *groups_list,
2582 struct btrfs_ioctl_space_info *space)
2583{
2584 struct btrfs_block_group_cache *block_group;
2585
2586 space->total_bytes = 0;
2587 space->used_bytes = 0;
2588 space->flags = 0;
2589 list_for_each_entry(block_group, groups_list, list) {
2590 space->flags = block_group->flags;
2591 space->total_bytes += block_group->key.offset;
2592 space->used_bytes +=
2593 btrfs_block_group_used(&block_group->item);
2594 }
2595}
2596
1406e432
JB		 2597long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2598{
2599 struct btrfs_ioctl_space_args space_args;
2600 struct btrfs_ioctl_space_info space;
2601 struct btrfs_ioctl_space_info *dest;
7fde62bf 2602 struct btrfs_ioctl_space_info *dest_orig;
13f2696f 2603 struct btrfs_ioctl_space_info __user *user_dest;
1406e432 2604 struct btrfs_space_info *info;
bf5fc093
JB		 2605 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2606 BTRFS_BLOCK_GROUP_SYSTEM,
2607 BTRFS_BLOCK_GROUP_METADATA,
2608 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2609 int num_types = 4;
7fde62bf 2610 int alloc_size;
1406e432 2611 int ret = 0;
51788b1b 2612 u64 slot_count = 0;
bf5fc093 2613 int i, c;
1406e432
JB		 2614
2615 if (copy_from_user(&space_args,
2616 (struct btrfs_ioctl_space_args __user *)arg,
2617 sizeof(space_args)))
2618 return -EFAULT;
2619
bf5fc093
JB		 2620 for (i = 0; i < num_types; i++) {
2621 struct btrfs_space_info *tmp;
2622
2623 info = NULL;
2624 rcu_read_lock();
2625 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2626 list) {
2627 if (tmp->flags == types[i]) {
2628 info = tmp;
2629 break;
2630 }
2631 }
2632 rcu_read_unlock();
2633
2634 if (!info)
2635 continue;
2636
2637 down_read(&info->groups_sem);
2638 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2639 if (!list_empty(&info->block_groups[c]))
2640 slot_count++;
2641 }
2642 up_read(&info->groups_sem);
2643 }
7fde62bf
CM		 2644
2645 /* space_slots == 0 means they are asking for a count */
2646 if (space_args.space_slots == 0) {
2647 space_args.total_spaces = slot_count;
2648 goto out;
2649 }
bf5fc093 2650
51788b1b 2651 slot_count = min_t(u64, space_args.space_slots, slot_count);
bf5fc093 2652
7fde62bf 2653 alloc_size = sizeof(*dest) * slot_count;
bf5fc093 2654
7fde62bf
CM		 2655 /* we generally have at most 6 or so space infos, one for each raid
2656 * level. So, a whole page should be more than enough for everyone
2657 */
2658 if (alloc_size > PAGE_CACHE_SIZE)
2659 return -ENOMEM;
2660
1406e432 2661 space_args.total_spaces = 0;
7fde62bf
CM		 2662 dest = kmalloc(alloc_size, GFP_NOFS);
2663 if (!dest)
2664 return -ENOMEM;
2665 dest_orig = dest;
1406e432 2666
7fde62bf 2667 /* now we have a buffer to copy into */
bf5fc093
JB		 2668 for (i = 0; i < num_types; i++) {
2669 struct btrfs_space_info *tmp;
2670
51788b1b
DR		 2671 if (!slot_count)
2672 break;
2673
bf5fc093
JB		 2674 info = NULL;
2675 rcu_read_lock();
2676 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2677 list) {
2678 if (tmp->flags == types[i]) {
2679 info = tmp;
2680 break;
2681 }
2682 }
2683 rcu_read_unlock();
7fde62bf 2684
bf5fc093
JB		 2685 if (!info)
2686 continue;
2687 down_read(&info->groups_sem);
2688 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2689 if (!list_empty(&info->block_groups[c])) {
2690 get_block_group_info(&info->block_groups[c],
2691 &space);
2692 memcpy(dest, &space, sizeof(space));
2693 dest++;
2694 space_args.total_spaces++;
51788b1b 2695 slot_count--;
bf5fc093 2696 }
51788b1b
DR		 2697 if (!slot_count)
2698 break;
bf5fc093
JB		 2699 }
2700 up_read(&info->groups_sem);
1406e432 2701 }
1406e432 2702
7fde62bf
CM		 2703 user_dest = (struct btrfs_ioctl_space_info *)
2704 (arg + sizeof(struct btrfs_ioctl_space_args));
2705
2706 if (copy_to_user(user_dest, dest_orig, alloc_size))
2707 ret = -EFAULT;
2708
2709 kfree(dest_orig);
2710out:
2711 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
1406e432
JB		 2712 ret = -EFAULT;
2713
2714 return ret;
2715}
2716
f46b5a66
CH		 2717/*
2718 * there are many ways the trans_start and trans_end ioctls can lead
2719 * to deadlocks. They should only be used by applications that
2720 * basically own the machine, and have a very in depth understanding
2721 * of all the possible deadlocks and enospc problems.
2722 */
2723long btrfs_ioctl_trans_end(struct file *file)
2724{
2725 struct inode *inode = fdentry(file)->d_inode;
2726 struct btrfs_root *root = BTRFS_I(inode)->root;
2727 struct btrfs_trans_handle *trans;
f46b5a66 2728
f46b5a66 2729 trans = file->private_data;
1ab86aed
SW		 2730 if (!trans)
2731 return -EINVAL;
b214107e 2732 file->private_data = NULL;
9ca9ee09 2733
1ab86aed
SW		 2734 btrfs_end_transaction(trans, root);
2735
a4abeea4 2736 atomic_dec(&root->fs_info->open_ioctl_trans);
9ca9ee09 2737
cfc8ea87 2738 mnt_drop_write(file->f_path.mnt);
1ab86aed 2739 return 0;
f46b5a66
CH		 2740}
2741
46204592
SW		 2742static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
2743{
2744 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2745 struct btrfs_trans_handle *trans;
2746 u64 transid;
db5b493a 2747 int ret;
46204592
SW		 2748
2749 trans = btrfs_start_transaction(root, 0);
98d5dc13
TI		 2750 if (IS_ERR(trans))
2751 return PTR_ERR(trans);
46204592 2752 transid = trans->transid;
db5b493a 2753 ret = btrfs_commit_transaction_async(trans, root, 0);
8b2b2d3c
TI		 2754 if (ret) {
2755 btrfs_end_transaction(trans, root);
db5b493a 2756 return ret;
8b2b2d3c 2757 }
46204592
SW		 2758
2759 if (argp)
2760 if (copy_to_user(argp, &transid, sizeof(transid)))
2761 return -EFAULT;
2762 return 0;
2763}
2764
2765static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
2766{
2767 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2768 u64 transid;
2769
2770 if (argp) {
2771 if (copy_from_user(&transid, argp, sizeof(transid)))
2772 return -EFAULT;
2773 } else {
2774 transid = 0; /* current trans */
2775 }
2776 return btrfs_wait_for_commit(root, transid);
2777}
2778
475f6387
JS		 2779static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg)
2780{
2781 int ret;
2782 struct btrfs_ioctl_scrub_args *sa;
2783
2784 if (!capable(CAP_SYS_ADMIN))
2785 return -EPERM;
2786
2787 sa = memdup_user(arg, sizeof(*sa));
2788 if (IS_ERR(sa))
2789 return PTR_ERR(sa);
2790
2791 ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end,
8628764e 2792 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY);
475f6387
JS		 2793
2794 if (copy_to_user(arg, sa, sizeof(*sa)))
2795 ret = -EFAULT;
2796
2797 kfree(sa);
2798 return ret;
2799}
2800
2801static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
2802{
2803 if (!capable(CAP_SYS_ADMIN))
2804 return -EPERM;
2805
2806 return btrfs_scrub_cancel(root);
2807}
2808
2809static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
2810 void __user *arg)
2811{
2812 struct btrfs_ioctl_scrub_args *sa;
2813 int ret;
2814
2815 if (!capable(CAP_SYS_ADMIN))
2816 return -EPERM;
2817
2818 sa = memdup_user(arg, sizeof(*sa));
2819 if (IS_ERR(sa))
2820 return PTR_ERR(sa);
2821
2822 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
2823
2824 if (copy_to_user(arg, sa, sizeof(*sa)))
2825 ret = -EFAULT;
2826
2827 kfree(sa);
2828 return ret;
2829}
2830
f46b5a66
CH		 2831long btrfs_ioctl(struct file *file, unsigned int
2832 cmd, unsigned long arg)
2833{
2834 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
4bcabaa3 2835 void __user *argp = (void __user *)arg;
f46b5a66
CH		 2836
2837 switch (cmd) {
6cbff00f
CH		 2838 case FS_IOC_GETFLAGS:
2839 return btrfs_ioctl_getflags(file, argp);
2840 case FS_IOC_SETFLAGS:
2841 return btrfs_ioctl_setflags(file, argp);
2842 case FS_IOC_GETVERSION:
2843 return btrfs_ioctl_getversion(file, argp);
f7039b1d
LD		 2844 case FITRIM:
2845 return btrfs_ioctl_fitrim(file, argp);
f46b5a66 2846 case BTRFS_IOC_SNAP_CREATE:
fa0d2b9b 2847 return btrfs_ioctl_snap_create(file, argp, 0);
fdfb1e4f 2848 case BTRFS_IOC_SNAP_CREATE_V2:
fa0d2b9b 2849 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3de4586c 2850 case BTRFS_IOC_SUBVOL_CREATE:
fa0d2b9b 2851 return btrfs_ioctl_snap_create(file, argp, 1);
76dda93c
YZ		 2852 case BTRFS_IOC_SNAP_DESTROY:
2853 return btrfs_ioctl_snap_destroy(file, argp);
0caa102d
LZ		 2854 case BTRFS_IOC_SUBVOL_GETFLAGS:
2855 return btrfs_ioctl_subvol_getflags(file, argp);
2856 case BTRFS_IOC_SUBVOL_SETFLAGS:
2857 return btrfs_ioctl_subvol_setflags(file, argp);
6ef5ed0d
JB		 2858 case BTRFS_IOC_DEFAULT_SUBVOL:
2859 return btrfs_ioctl_default_subvol(file, argp);
f46b5a66 2860 case BTRFS_IOC_DEFRAG:
1e701a32
CM		 2861 return btrfs_ioctl_defrag(file, NULL);
2862 case BTRFS_IOC_DEFRAG_RANGE:
2863 return btrfs_ioctl_defrag(file, argp);
f46b5a66 2864 case BTRFS_IOC_RESIZE:
4bcabaa3 2865 return btrfs_ioctl_resize(root, argp);
f46b5a66 2866 case BTRFS_IOC_ADD_DEV:
4bcabaa3 2867 return btrfs_ioctl_add_dev(root, argp);
f46b5a66 2868 case BTRFS_IOC_RM_DEV:
4bcabaa3 2869 return btrfs_ioctl_rm_dev(root, argp);
475f6387
JS		 2870 case BTRFS_IOC_FS_INFO:
2871 return btrfs_ioctl_fs_info(root, argp);
2872 case BTRFS_IOC_DEV_INFO:
2873 return btrfs_ioctl_dev_info(root, argp);
f46b5a66
CH		 2874 case BTRFS_IOC_BALANCE:
2875 return btrfs_balance(root->fs_info->dev_root);
2876 case BTRFS_IOC_CLONE:
c5c9cd4d
SW		 2877 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
2878 case BTRFS_IOC_CLONE_RANGE:
7a865e8a 2879 return btrfs_ioctl_clone_range(file, argp);
f46b5a66
CH		 2880 case BTRFS_IOC_TRANS_START:
2881 return btrfs_ioctl_trans_start(file);
2882 case BTRFS_IOC_TRANS_END:
2883 return btrfs_ioctl_trans_end(file);
ac8e9819
CM		 2884 case BTRFS_IOC_TREE_SEARCH:
2885 return btrfs_ioctl_tree_search(file, argp);
2886 case BTRFS_IOC_INO_LOOKUP:
2887 return btrfs_ioctl_ino_lookup(file, argp);
1406e432
JB		 2888 case BTRFS_IOC_SPACE_INFO:
2889 return btrfs_ioctl_space_info(root, argp);
f46b5a66
CH		 2890 case BTRFS_IOC_SYNC:
2891 btrfs_sync_fs(file->f_dentry->d_sb, 1);
2892 return 0;
46204592
SW		 2893 case BTRFS_IOC_START_SYNC:
2894 return btrfs_ioctl_start_sync(file, argp);
2895 case BTRFS_IOC_WAIT_SYNC:
2896 return btrfs_ioctl_wait_sync(file, argp);
475f6387
JS		 2897 case BTRFS_IOC_SCRUB:
2898 return btrfs_ioctl_scrub(root, argp);
2899 case BTRFS_IOC_SCRUB_CANCEL:
2900 return btrfs_ioctl_scrub_cancel(root, argp);
2901 case BTRFS_IOC_SCRUB_PROGRESS:
2902 return btrfs_ioctl_scrub_progress(root, argp);
f46b5a66
CH		 2903 }
2904
2905 return -ENOTTY;
2906}
Linux 6.x block layer and io_uring tree(s)