projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Btrfs: add mount -o auto_defrag
[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
CH		 245
246 trans = btrfs_join_transaction(root, 1);
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
417 ret = btrfs_create_subvol_root(trans, new_root, new_dirid,
418 BTRFS_I(dir)->block_group);
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:
6a912213 443 dput(parent);
72fd032e
SW		 444 if (async_transid) {
445 *async_transid = trans->transid;
446 err = btrfs_commit_transaction_async(trans, root, 1);
447 } else {
448 err = btrfs_commit_transaction(trans, root);
449 }
f46b5a66
CH		 450 if (err && !ret)
451 ret = err;
f46b5a66
CH		 452 return ret;
453}
454
72fd032e 455static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
b83cc969
LZ		 456 char *name, int namelen, u64 *async_transid,
457 bool readonly)
f46b5a66 458{
2e4bfab9 459 struct inode *inode;
6a912213 460 struct dentry *parent;
f46b5a66
CH		 461 struct btrfs_pending_snapshot *pending_snapshot;
462 struct btrfs_trans_handle *trans;
2e4bfab9 463 int ret;
f46b5a66
CH		 464
465 if (!root->ref_cows)
466 return -EINVAL;
467
3de4586c 468 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
a22285a6
YZ		 469 if (!pending_snapshot)
470 return -ENOMEM;
471
472 btrfs_init_block_rsv(&pending_snapshot->block_rsv);
3de4586c 473 pending_snapshot->dentry = dentry;
f46b5a66 474 pending_snapshot->root = root;
b83cc969 475 pending_snapshot->readonly = readonly;
a22285a6
YZ		 476
477 trans = btrfs_start_transaction(root->fs_info->extent_root, 5);
478 if (IS_ERR(trans)) {
479 ret = PTR_ERR(trans);
480 goto fail;
481 }
482
483 ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
484 BUG_ON(ret);
485
f46b5a66
CH		 486 list_add(&pending_snapshot->list,
487 &trans->transaction->pending_snapshots);
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
6a912213
JB		 506 parent = dget_parent(dentry);
507 inode = btrfs_lookup_dentry(parent->d_inode, dentry);
508 dput(parent);
2e4bfab9
YZ		 509 if (IS_ERR(inode)) {
510 ret = PTR_ERR(inode);
511 goto fail;
512 }
513 BUG_ON(!inode);
514 d_instantiate(dentry, inode);
515 ret = 0;
516fail:
a22285a6 517 kfree(pending_snapshot);
f46b5a66
CH		 518 return ret;
519}
520
4260f7c7
SW		 521/* copy of check_sticky in fs/namei.c()
522* It's inline, so penalty for filesystems that don't use sticky bit is
523* minimal.
524*/
525static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
526{
527 uid_t fsuid = current_fsuid();
528
529 if (!(dir->i_mode & S_ISVTX))
530 return 0;
531 if (inode->i_uid == fsuid)
532 return 0;
533 if (dir->i_uid == fsuid)
534 return 0;
535 return !capable(CAP_FOWNER);
536}
537
538/* copy of may_delete in fs/namei.c()
539 * Check whether we can remove a link victim from directory dir, check
540 * whether the type of victim is right.
541 * 1. We can't do it if dir is read-only (done in permission())
542 * 2. We should have write and exec permissions on dir
543 * 3. We can't remove anything from append-only dir
544 * 4. We can't do anything with immutable dir (done in permission())
545 * 5. If the sticky bit on dir is set we should either
546 * a. be owner of dir, or
547 * b. be owner of victim, or
548 * c. have CAP_FOWNER capability
549 * 6. If the victim is append-only or immutable we can't do antyhing with
550 * links pointing to it.
551 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
552 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
553 * 9. We can't remove a root or mountpoint.
554 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
555 * nfs_async_unlink().
556 */
557
558static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
559{
560 int error;
561
562 if (!victim->d_inode)
563 return -ENOENT;
564
565 BUG_ON(victim->d_parent->d_inode != dir);
566 audit_inode_child(victim, dir);
567
568 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
569 if (error)
570 return error;
571 if (IS_APPEND(dir))
572 return -EPERM;
573 if (btrfs_check_sticky(dir, victim->d_inode)||
574 IS_APPEND(victim->d_inode)||
575 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
576 return -EPERM;
577 if (isdir) {
578 if (!S_ISDIR(victim->d_inode->i_mode))
579 return -ENOTDIR;
580 if (IS_ROOT(victim))
581 return -EBUSY;
582 } else if (S_ISDIR(victim->d_inode->i_mode))
583 return -EISDIR;
584 if (IS_DEADDIR(dir))
585 return -ENOENT;
586 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
587 return -EBUSY;
588 return 0;
589}
590
cb8e7090
CH		 591/* copy of may_create in fs/namei.c() */
592static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
593{
594 if (child->d_inode)
595 return -EEXIST;
596 if (IS_DEADDIR(dir))
597 return -ENOENT;
598 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
599}
600
601/*
602 * Create a new subvolume below @parent. This is largely modeled after
603 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
604 * inside this filesystem so it's quite a bit simpler.
605 */
76dda93c
YZ		 606static noinline int btrfs_mksubvol(struct path *parent,
607 char *name, int namelen,
72fd032e 608 struct btrfs_root *snap_src,
b83cc969 609 u64 *async_transid, bool readonly)
cb8e7090 610{
76dda93c 611 struct inode *dir = parent->dentry->d_inode;
cb8e7090
CH		 612 struct dentry *dentry;
613 int error;
614
76dda93c 615 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
cb8e7090
CH		 616
617 dentry = lookup_one_len(name, parent->dentry, namelen);
618 error = PTR_ERR(dentry);
619 if (IS_ERR(dentry))
620 goto out_unlock;
621
622 error = -EEXIST;
623 if (dentry->d_inode)
624 goto out_dput;
625
cb8e7090
CH		 626 error = mnt_want_write(parent->mnt);
627 if (error)
628 goto out_dput;
629
76dda93c 630 error = btrfs_may_create(dir, dentry);
cb8e7090
CH		 631 if (error)
632 goto out_drop_write;
633
76dda93c
YZ		 634 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
635
636 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
637 goto out_up_read;
638
3de4586c 639 if (snap_src) {
72fd032e 640 error = create_snapshot(snap_src, dentry,
b83cc969 641 name, namelen, async_transid, readonly);
3de4586c 642 } else {
76dda93c 643 error = create_subvol(BTRFS_I(dir)->root, dentry,
72fd032e 644 name, namelen, async_transid);
3de4586c 645 }
76dda93c
YZ		 646 if (!error)
647 fsnotify_mkdir(dir, dentry);
648out_up_read:
649 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
cb8e7090
CH		 650out_drop_write:
651 mnt_drop_write(parent->mnt);
652out_dput:
653 dput(dentry);
654out_unlock:
76dda93c 655 mutex_unlock(&dir->i_mutex);
cb8e7090
CH		 656 return error;
657}
658
4cb5300b
CM		 659/*
660 * When we're defragging a range, we don't want to kick it off again
661 * if it is really just waiting for delalloc to send it down.
662 * If we find a nice big extent or delalloc range for the bytes in the
663 * file you want to defrag, we return 0 to let you know to skip this
664 * part of the file
665 */
666static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
667{
668 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
669 struct extent_map *em = NULL;
670 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
671 u64 end;
672
673 read_lock(&em_tree->lock);
674 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
675 read_unlock(&em_tree->lock);
676
677 if (em) {
678 end = extent_map_end(em);
679 free_extent_map(em);
680 if (end - offset > thresh)
681 return 0;
682 }
683 /* if we already have a nice delalloc here, just stop */
684 thresh /= 2;
685 end = count_range_bits(io_tree, &offset, offset + thresh,
686 thresh, EXTENT_DELALLOC, 1);
687 if (end >= thresh)
688 return 0;
689 return 1;
690}
691
692/*
693 * helper function to walk through a file and find extents
694 * newer than a specific transid, and smaller than thresh.
695 *
696 * This is used by the defragging code to find new and small
697 * extents
698 */
699static int find_new_extents(struct btrfs_root *root,
700 struct inode *inode, u64 newer_than,
701 u64 *off, int thresh)
702{
703 struct btrfs_path *path;
704 struct btrfs_key min_key;
705 struct btrfs_key max_key;
706 struct extent_buffer *leaf;
707 struct btrfs_file_extent_item *extent;
708 int type;
709 int ret;
710
711 path = btrfs_alloc_path();
712 if (!path)
713 return -ENOMEM;
714
715 min_key.objectid = inode->i_ino;
716 min_key.type = BTRFS_EXTENT_DATA_KEY;
717 min_key.offset = *off;
718
719 max_key.objectid = inode->i_ino;
720 max_key.type = (u8)-1;
721 max_key.offset = (u64)-1;
722
723 path->keep_locks = 1;
724
725 while(1) {
726 ret = btrfs_search_forward(root, &min_key, &max_key,
727 path, 0, newer_than);
728 if (ret != 0)
729 goto none;
730 if (min_key.objectid != inode->i_ino)
731 goto none;
732 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
733 goto none;
734
735 leaf = path->nodes[0];
736 extent = btrfs_item_ptr(leaf, path->slots[0],
737 struct btrfs_file_extent_item);
738
739 type = btrfs_file_extent_type(leaf, extent);
740 if (type == BTRFS_FILE_EXTENT_REG &&
741 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
742 check_defrag_in_cache(inode, min_key.offset, thresh)) {
743 *off = min_key.offset;
744 btrfs_free_path(path);
745 return 0;
746 }
747
748 if (min_key.offset == (u64)-1)
749 goto none;
750
751 min_key.offset++;
752 btrfs_release_path(path);
753 }
754none:
755 btrfs_free_path(path);
756 return -ENOENT;
757}
758
940100a4 759static int should_defrag_range(struct inode *inode, u64 start, u64 len,
1e701a32
CM		 760 int thresh, u64 *last_len, u64 *skip,
761 u64 *defrag_end)
940100a4
CM		 762{
763 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
764 struct extent_map *em = NULL;
765 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
766 int ret = 1;
767
768 /*
769 * make sure that once we start defragging and extent, we keep on
770 * defragging it
771 */
772 if (start < *defrag_end)
773 return 1;
774
775 *skip = 0;
776
777 /*
778 * hopefully we have this extent in the tree already, try without
779 * the full extent lock
780 */
781 read_lock(&em_tree->lock);
782 em = lookup_extent_mapping(em_tree, start, len);
783 read_unlock(&em_tree->lock);
784
785 if (!em) {
786 /* get the big lock and read metadata off disk */
787 lock_extent(io_tree, start, start + len - 1, GFP_NOFS);
788 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
789 unlock_extent(io_tree, start, start + len - 1, GFP_NOFS);
790
6cf8bfbf 791 if (IS_ERR(em))
940100a4
CM		 792 return 0;
793 }
794
795 /* this will cover holes, and inline extents */
796 if (em->block_start >= EXTENT_MAP_LAST_BYTE)
797 ret = 0;
798
799 /*
800 * we hit a real extent, if it is big don't bother defragging it again
801 */
1e701a32 802 if ((*last_len == 0 || *last_len >= thresh) && em->len >= thresh)
940100a4
CM		 803 ret = 0;
804
805 /*
806 * last_len ends up being a counter of how many bytes we've defragged.
807 * every time we choose not to defrag an extent, we reset *last_len
808 * so that the next tiny extent will force a defrag.
809 *
810 * The end result of this is that tiny extents before a single big
811 * extent will force at least part of that big extent to be defragged.
812 */
813 if (ret) {
814 *last_len += len;
815 *defrag_end = extent_map_end(em);
816 } else {
817 *last_len = 0;
818 *skip = extent_map_end(em);
819 *defrag_end = 0;
820 }
821
822 free_extent_map(em);
823 return ret;
824}
825
4cb5300b
CM		 826/*
827 * it doesn't do much good to defrag one or two pages
828 * at a time. This pulls in a nice chunk of pages
829 * to COW and defrag.
830 *
831 * It also makes sure the delalloc code has enough
832 * dirty data to avoid making new small extents as part
833 * of the defrag
834 *
835 * It's a good idea to start RA on this range
836 * before calling this.
837 */
838static int cluster_pages_for_defrag(struct inode *inode,
839 struct page **pages,
840 unsigned long start_index,
841 int num_pages)
f46b5a66 842{
4cb5300b
CM		 843 unsigned long file_end;
844 u64 isize = i_size_read(inode);
845 u64 page_start;
846 u64 page_end;
847 int ret;
848 int i;
849 int i_done;
3eaa2885 850 struct btrfs_ordered_extent *ordered;
4cb5300b
CM		 851 struct extent_state *cached_state = NULL;
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;
868 page = grab_cache_page(inode->i_mapping,
869 start_index + i);
870 if (!page)
871 break;
872
873 if (!PageUptodate(page)) {
874 btrfs_readpage(NULL, page);
875 lock_page(page);
876 if (!PageUptodate(page)) {
877 unlock_page(page);
878 page_cache_release(page);
879 ret = -EIO;
880 break;
881 }
882 }
883 isize = i_size_read(inode);
884 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
885 if (!isize || page->index > file_end ||
886 page->mapping != inode->i_mapping) {
887 /* whoops, we blew past eof, skip this page */
888 unlock_page(page);
889 page_cache_release(page);
890 break;
891 }
892 pages[i] = page;
893 i_done++;
894 }
895 if (!i_done || ret)
896 goto out;
897
898 if (!(inode->i_sb->s_flags & MS_ACTIVE))
899 goto out;
900
901 /*
902 * so now we have a nice long stream of locked
903 * and up to date pages, lets wait on them
904 */
905 for (i = 0; i < i_done; i++)
906 wait_on_page_writeback(pages[i]);
907
908 page_start = page_offset(pages[0]);
909 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
910
911 lock_extent_bits(&BTRFS_I(inode)->io_tree,
912 page_start, page_end - 1, 0, &cached_state,
913 GFP_NOFS);
914 ordered = btrfs_lookup_first_ordered_extent(inode, page_end - 1);
915 if (ordered &&
916 ordered->file_offset + ordered->len > page_start &&
917 ordered->file_offset < page_end) {
918 btrfs_put_ordered_extent(ordered);
919 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
920 page_start, page_end - 1,
921 &cached_state, GFP_NOFS);
922 for (i = 0; i < i_done; i++) {
923 unlock_page(pages[i]);
924 page_cache_release(pages[i]);
925 }
926 btrfs_wait_ordered_range(inode, page_start,
927 page_end - page_start);
928 goto again;
929 }
930 if (ordered)
931 btrfs_put_ordered_extent(ordered);
932
933 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
934 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
935 EXTENT_DO_ACCOUNTING, 0, 0, &cached_state,
936 GFP_NOFS);
937
938 if (i_done != num_pages) {
939 atomic_inc(&BTRFS_I(inode)->outstanding_extents);
940 btrfs_delalloc_release_space(inode,
941 (num_pages - i_done) << PAGE_CACHE_SHIFT);
942 }
943
944
945 btrfs_set_extent_delalloc(inode, page_start, page_end - 1,
946 &cached_state);
947
948 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
949 page_start, page_end - 1, &cached_state,
950 GFP_NOFS);
951
952 for (i = 0; i < i_done; i++) {
953 clear_page_dirty_for_io(pages[i]);
954 ClearPageChecked(pages[i]);
955 set_page_extent_mapped(pages[i]);
956 set_page_dirty(pages[i]);
957 unlock_page(pages[i]);
958 page_cache_release(pages[i]);
959 }
960 return i_done;
961out:
962 for (i = 0; i < i_done; i++) {
963 unlock_page(pages[i]);
964 page_cache_release(pages[i]);
965 }
966 btrfs_delalloc_release_space(inode, num_pages << PAGE_CACHE_SHIFT);
967 return ret;
968
969}
970
971int btrfs_defrag_file(struct inode *inode, struct file *file,
972 struct btrfs_ioctl_defrag_range_args *range,
973 u64 newer_than, unsigned long max_to_defrag)
974{
975 struct btrfs_root *root = BTRFS_I(inode)->root;
1a419d85 976 struct btrfs_super_block *disk_super;
4cb5300b 977 struct file_ra_state *ra = NULL;
f46b5a66 978 unsigned long last_index;
1a419d85 979 u64 features;
940100a4
CM		 980 u64 last_len = 0;
981 u64 skip = 0;
982 u64 defrag_end = 0;
4cb5300b
CM		 983 u64 newer_off = range->start;
984 int newer_left = 0;
f46b5a66
CH		 985 unsigned long i;
986 int ret;
4cb5300b 987 int defrag_count = 0;
1a419d85 988 int compress_type = BTRFS_COMPRESS_ZLIB;
4cb5300b
CM		 989 int extent_thresh = range->extent_thresh;
990 int newer_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
991 u64 new_align = ~((u64)128 * 1024 - 1);
992 struct page **pages = NULL;
993
994 if (extent_thresh == 0)
995 extent_thresh = 256 * 1024;
1a419d85
LZ		 996
997 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
998 if (range->compress_type > BTRFS_COMPRESS_TYPES)
999 return -EINVAL;
1000 if (range->compress_type)
1001 compress_type = range->compress_type;
1002 }
f46b5a66 1003
940100a4
CM		 1004 if (inode->i_size == 0)
1005 return 0;
1006
4cb5300b
CM		 1007 /*
1008 * if we were not given a file, allocate a readahead
1009 * context
1010 */
1011 if (!file) {
1012 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1013 if (!ra)
1014 return -ENOMEM;
1015 file_ra_state_init(ra, inode->i_mapping);
1016 } else {
1017 ra = &file->f_ra;
1018 }
1019
1020 pages = kmalloc(sizeof(struct page *) * newer_cluster,
1021 GFP_NOFS);
1022 if (!pages) {
1023 ret = -ENOMEM;
1024 goto out_ra;
1025 }
1026
1027 /* find the last page to defrag */
1e701a32
CM		 1028 if (range->start + range->len > range->start) {
1029 last_index = min_t(u64, inode->i_size - 1,
1030 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1031 } else {
1032 last_index = (inode->i_size - 1) >> PAGE_CACHE_SHIFT;
1033 }
1034
4cb5300b
CM		 1035 if (newer_than) {
1036 ret = find_new_extents(root, inode, newer_than,
1037 &newer_off, 64 * 1024);
1038 if (!ret) {
1039 range->start = newer_off;
1040 /*
1041 * we always align our defrag to help keep
1042 * the extents in the file evenly spaced
1043 */
1044 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1045 newer_left = newer_cluster;
1046 } else
1047 goto out_ra;
1048 } else {
1049 i = range->start >> PAGE_CACHE_SHIFT;
1050 }
1051 if (!max_to_defrag)
1052 max_to_defrag = last_index - 1;
1053
1054 while (i <= last_index && defrag_count < max_to_defrag) {
1055 /*
1056 * make sure we stop running if someone unmounts
1057 * the FS
1058 */
1059 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1060 break;
1061
1062 if (!newer_than &&
1063 !should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1e701a32 1064 PAGE_CACHE_SIZE,
4cb5300b 1065 extent_thresh,
1e701a32 1066 &last_len, &skip,
940100a4
CM		 1067 &defrag_end)) {
1068 unsigned long next;
1069 /*
1070 * the should_defrag function tells us how much to skip
1071 * bump our counter by the suggested amount
1072 */
1073 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1074 i = max(i + 1, next);
1075 continue;
1076 }
1e701a32 1077 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1a419d85 1078 BTRFS_I(inode)->force_compress = compress_type;
940100a4 1079
4cb5300b 1080 btrfs_force_ra(inode->i_mapping, ra, file, i, newer_cluster);
940100a4 1081
4cb5300b
CM		 1082 ret = cluster_pages_for_defrag(inode, pages, i, newer_cluster);
1083 if (ret < 0)
1084 goto out_ra;
1085
1086 defrag_count += ret;
1087 balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret);
1088 i += ret;
1089
1090 if (newer_than) {
1091 if (newer_off == (u64)-1)
1092 break;
1093
1094 newer_off = max(newer_off + 1,
1095 (u64)i << PAGE_CACHE_SHIFT);
1096
1097 ret = find_new_extents(root, inode,
1098 newer_than, &newer_off,
1099 64 * 1024);
1100 if (!ret) {
1101 range->start = newer_off;
1102 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
1103 newer_left = newer_cluster;
1104 } else {
1105 break;
f46b5a66 1106 }
4cb5300b
CM		 1107 } else {
1108 i++;
f46b5a66 1109 }
f46b5a66
CH		 1110 }
1111
1e701a32
CM		 1112 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1113 filemap_flush(inode->i_mapping);
1114
1115 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1116 /* the filemap_flush will queue IO into the worker threads, but
1117 * we have to make sure the IO is actually started and that
1118 * ordered extents get created before we return
1119 */
1120 atomic_inc(&root->fs_info->async_submit_draining);
1121 while (atomic_read(&root->fs_info->nr_async_submits) ||
1122 atomic_read(&root->fs_info->async_delalloc_pages)) {
1123 wait_event(root->fs_info->async_submit_wait,
1124 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1125 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1126 }
1127 atomic_dec(&root->fs_info->async_submit_draining);
1128
1129 mutex_lock(&inode->i_mutex);
261507a0 1130 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1e701a32
CM		 1131 mutex_unlock(&inode->i_mutex);
1132 }
1133
1a419d85
LZ		 1134 disk_super = &root->fs_info->super_copy;
1135 features = btrfs_super_incompat_flags(disk_super);
1136 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1137 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
1138 btrfs_set_super_incompat_flags(disk_super, features);
1139 }
1140
4cb5300b
CM		 1141 if (!file)
1142 kfree(ra);
1143 return defrag_count;
940100a4 1144
4cb5300b
CM		 1145out_ra:
1146 if (!file)
1147 kfree(ra);
1148 kfree(pages);
940100a4 1149 return ret;
f46b5a66
CH		 1150}
1151
76dda93c
YZ		 1152static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
1153 void __user *arg)
f46b5a66
CH		 1154{
1155 u64 new_size;
1156 u64 old_size;
1157 u64 devid = 1;
1158 struct btrfs_ioctl_vol_args *vol_args;
1159 struct btrfs_trans_handle *trans;
1160 struct btrfs_device *device = NULL;
1161 char *sizestr;
1162 char *devstr = NULL;
1163 int ret = 0;
f46b5a66
CH		 1164 int mod = 0;
1165
c146afad
YZ		 1166 if (root->fs_info->sb->s_flags & MS_RDONLY)
1167 return -EROFS;
1168
e441d54d
CM		 1169 if (!capable(CAP_SYS_ADMIN))
1170 return -EPERM;
1171
dae7b665
LZ		 1172 vol_args = memdup_user(arg, sizeof(*vol_args));
1173 if (IS_ERR(vol_args))
1174 return PTR_ERR(vol_args);
5516e595
MF		 1175
1176 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 1177
7d9eb12c 1178 mutex_lock(&root->fs_info->volume_mutex);
f46b5a66
CH		 1179 sizestr = vol_args->name;
1180 devstr = strchr(sizestr, ':');
1181 if (devstr) {
1182 char *end;
1183 sizestr = devstr + 1;
1184 *devstr = '\0';
1185 devstr = vol_args->name;
1186 devid = simple_strtoull(devstr, &end, 10);
21380931
JB		 1187 printk(KERN_INFO "resizing devid %llu\n",
1188 (unsigned long long)devid);
f46b5a66 1189 }
2b82032c 1190 device = btrfs_find_device(root, devid, NULL, NULL);
f46b5a66 1191 if (!device) {
21380931
JB		 1192 printk(KERN_INFO "resizer unable to find device %llu\n",
1193 (unsigned long long)devid);
f46b5a66
CH		 1194 ret = -EINVAL;
1195 goto out_unlock;
1196 }
1197 if (!strcmp(sizestr, "max"))
1198 new_size = device->bdev->bd_inode->i_size;
1199 else {
1200 if (sizestr[0] == '-') {
1201 mod = -1;
1202 sizestr++;
1203 } else if (sizestr[0] == '+') {
1204 mod = 1;
1205 sizestr++;
1206 }
91748467 1207 new_size = memparse(sizestr, NULL);
f46b5a66
CH		 1208 if (new_size == 0) {
1209 ret = -EINVAL;
1210 goto out_unlock;
1211 }
1212 }
1213
1214 old_size = device->total_bytes;
1215
1216 if (mod < 0) {
1217 if (new_size > old_size) {
1218 ret = -EINVAL;
1219 goto out_unlock;
1220 }
1221 new_size = old_size - new_size;
1222 } else if (mod > 0) {
1223 new_size = old_size + new_size;
1224 }
1225
1226 if (new_size < 256 * 1024 * 1024) {
1227 ret = -EINVAL;
1228 goto out_unlock;
1229 }
1230 if (new_size > device->bdev->bd_inode->i_size) {
1231 ret = -EFBIG;
1232 goto out_unlock;
1233 }
1234
1235 do_div(new_size, root->sectorsize);
1236 new_size *= root->sectorsize;
1237
1238 printk(KERN_INFO "new size for %s is %llu\n",
1239 device->name, (unsigned long long)new_size);
1240
1241 if (new_size > old_size) {
a22285a6 1242 trans = btrfs_start_transaction(root, 0);
98d5dc13
TI		 1243 if (IS_ERR(trans)) {
1244 ret = PTR_ERR(trans);
1245 goto out_unlock;
1246 }
f46b5a66
CH		 1247 ret = btrfs_grow_device(trans, device, new_size);
1248 btrfs_commit_transaction(trans, root);
1249 } else {
1250 ret = btrfs_shrink_device(device, new_size);
1251 }
1252
1253out_unlock:
7d9eb12c 1254 mutex_unlock(&root->fs_info->volume_mutex);
f46b5a66
CH		 1255 kfree(vol_args);
1256 return ret;
1257}
1258
72fd032e
SW		 1259static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1260 char *name,
1261 unsigned long fd,
1262 int subvol,
b83cc969
LZ		 1263 u64 *transid,
1264 bool readonly)
f46b5a66 1265{
cb8e7090 1266 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3de4586c 1267 struct file *src_file;
f46b5a66 1268 int namelen;
3de4586c 1269 int ret = 0;
f46b5a66 1270
c146afad
YZ		 1271 if (root->fs_info->sb->s_flags & MS_RDONLY)
1272 return -EROFS;
1273
72fd032e
SW		 1274 namelen = strlen(name);
1275 if (strchr(name, '/')) {
f46b5a66
CH		 1276 ret = -EINVAL;
1277 goto out;
1278 }
1279
3de4586c 1280 if (subvol) {
72fd032e 1281 ret = btrfs_mksubvol(&file->f_path, name, namelen,
b83cc969 1282 NULL, transid, readonly);
cb8e7090 1283 } else {
3de4586c 1284 struct inode *src_inode;
72fd032e 1285 src_file = fget(fd);
3de4586c
CM		 1286 if (!src_file) {
1287 ret = -EINVAL;
1288 goto out;
1289 }
1290
1291 src_inode = src_file->f_path.dentry->d_inode;
1292 if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
d397712b
CM		 1293 printk(KERN_INFO "btrfs: Snapshot src from "
1294 "another FS\n");
3de4586c
CM		 1295 ret = -EINVAL;
1296 fput(src_file);
1297 goto out;
1298 }
72fd032e
SW		 1299 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1300 BTRFS_I(src_inode)->root,
b83cc969 1301 transid, readonly);
3de4586c 1302 fput(src_file);
cb8e7090 1303 }
f46b5a66 1304out:
72fd032e
SW		 1305 return ret;
1306}
1307
1308static noinline int btrfs_ioctl_snap_create(struct file *file,
fa0d2b9b 1309 void __user *arg, int subvol)
72fd032e 1310{
fa0d2b9b 1311 struct btrfs_ioctl_vol_args *vol_args;
72fd032e
SW		 1312 int ret;
1313
fa0d2b9b
LZ		 1314 vol_args = memdup_user(arg, sizeof(*vol_args));
1315 if (IS_ERR(vol_args))
1316 return PTR_ERR(vol_args);
1317 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
72fd032e 1318
fa0d2b9b 1319 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
b83cc969
LZ		 1320 vol_args->fd, subvol,
1321 NULL, false);
fdfb1e4f 1322
fa0d2b9b
LZ		 1323 kfree(vol_args);
1324 return ret;
1325}
fdfb1e4f 1326
fa0d2b9b
LZ		 1327static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1328 void __user *arg, int subvol)
1329{
1330 struct btrfs_ioctl_vol_args_v2 *vol_args;
1331 int ret;
1332 u64 transid = 0;
1333 u64 *ptr = NULL;
b83cc969 1334 bool readonly = false;
75eaa0e2 1335
fa0d2b9b
LZ		 1336 vol_args = memdup_user(arg, sizeof(*vol_args));
1337 if (IS_ERR(vol_args))
1338 return PTR_ERR(vol_args);
1339 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
75eaa0e2 1340
b83cc969
LZ		 1341 if (vol_args->flags &
1342 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY)) {
1343 ret = -EOPNOTSUPP;
fa0d2b9b 1344 goto out;
72fd032e 1345 }
fa0d2b9b
LZ		 1346
1347 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1348 ptr = &transid;
b83cc969
LZ		 1349 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1350 readonly = true;
fa0d2b9b
LZ		 1351
1352 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
b83cc969
LZ		 1353 vol_args->fd, subvol,
1354 ptr, readonly);
fa0d2b9b
LZ		 1355
1356 if (ret == 0 && ptr &&
1357 copy_to_user(arg +
1358 offsetof(struct btrfs_ioctl_vol_args_v2,
1359 transid), ptr, sizeof(*ptr)))
1360 ret = -EFAULT;
fdfb1e4f 1361out:
f46b5a66
CH		 1362 kfree(vol_args);
1363 return ret;
1364}
1365
0caa102d
LZ		 1366static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1367 void __user *arg)
1368{
1369 struct inode *inode = fdentry(file)->d_inode;
1370 struct btrfs_root *root = BTRFS_I(inode)->root;
1371 int ret = 0;
1372 u64 flags = 0;
1373
33345d01 1374 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
0caa102d
LZ		 1375 return -EINVAL;
1376
1377 down_read(&root->fs_info->subvol_sem);
1378 if (btrfs_root_readonly(root))
1379 flags |= BTRFS_SUBVOL_RDONLY;
1380 up_read(&root->fs_info->subvol_sem);
1381
1382 if (copy_to_user(arg, &flags, sizeof(flags)))
1383 ret = -EFAULT;
1384
1385 return ret;
1386}
1387
1388static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1389 void __user *arg)
1390{
1391 struct inode *inode = fdentry(file)->d_inode;
1392 struct btrfs_root *root = BTRFS_I(inode)->root;
1393 struct btrfs_trans_handle *trans;
1394 u64 root_flags;
1395 u64 flags;
1396 int ret = 0;
1397
1398 if (root->fs_info->sb->s_flags & MS_RDONLY)
1399 return -EROFS;
1400
33345d01 1401 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
0caa102d
LZ		 1402 return -EINVAL;
1403
1404 if (copy_from_user(&flags, arg, sizeof(flags)))
1405 return -EFAULT;
1406
b4dc2b8c 1407 if (flags & BTRFS_SUBVOL_CREATE_ASYNC)
0caa102d
LZ		 1408 return -EINVAL;
1409
1410 if (flags & ~BTRFS_SUBVOL_RDONLY)
1411 return -EOPNOTSUPP;
1412
2e149670 1413 if (!inode_owner_or_capable(inode))
b4dc2b8c
LZ		 1414 return -EACCES;
1415
0caa102d
LZ		 1416 down_write(&root->fs_info->subvol_sem);
1417
1418 /* nothing to do */
1419 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1420 goto out;
1421
1422 root_flags = btrfs_root_flags(&root->root_item);
1423 if (flags & BTRFS_SUBVOL_RDONLY)
1424 btrfs_set_root_flags(&root->root_item,
1425 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1426 else
1427 btrfs_set_root_flags(&root->root_item,
1428 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1429
1430 trans = btrfs_start_transaction(root, 1);
1431 if (IS_ERR(trans)) {
1432 ret = PTR_ERR(trans);
1433 goto out_reset;
1434 }
1435
b4dc2b8c 1436 ret = btrfs_update_root(trans, root->fs_info->tree_root,
0caa102d
LZ		 1437 &root->root_key, &root->root_item);
1438
1439 btrfs_commit_transaction(trans, root);
1440out_reset:
1441 if (ret)
1442 btrfs_set_root_flags(&root->root_item, root_flags);
1443out:
1444 up_write(&root->fs_info->subvol_sem);
1445 return ret;
1446}
1447
76dda93c
YZ		 1448/*
1449 * helper to check if the subvolume references other subvolumes
1450 */
1451static noinline int may_destroy_subvol(struct btrfs_root *root)
1452{
1453 struct btrfs_path *path;
1454 struct btrfs_key key;
1455 int ret;
1456
1457 path = btrfs_alloc_path();
1458 if (!path)
1459 return -ENOMEM;
1460
1461 key.objectid = root->root_key.objectid;
1462 key.type = BTRFS_ROOT_REF_KEY;
1463 key.offset = (u64)-1;
1464
1465 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1466 &key, path, 0, 0);
1467 if (ret < 0)
1468 goto out;
1469 BUG_ON(ret == 0);
1470
1471 ret = 0;
1472 if (path->slots[0] > 0) {
1473 path->slots[0]--;
1474 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1475 if (key.objectid == root->root_key.objectid &&
1476 key.type == BTRFS_ROOT_REF_KEY)
1477 ret = -ENOTEMPTY;
1478 }
1479out:
1480 btrfs_free_path(path);
1481 return ret;
1482}
1483
ac8e9819
CM		 1484static noinline int key_in_sk(struct btrfs_key *key,
1485 struct btrfs_ioctl_search_key *sk)
1486{
abc6e134
CM		 1487 struct btrfs_key test;
1488 int ret;
1489
1490 test.objectid = sk->min_objectid;
1491 test.type = sk->min_type;
1492 test.offset = sk->min_offset;
1493
1494 ret = btrfs_comp_cpu_keys(key, &test);
1495 if (ret < 0)
ac8e9819 1496 return 0;
abc6e134
CM		 1497
1498 test.objectid = sk->max_objectid;
1499 test.type = sk->max_type;
1500 test.offset = sk->max_offset;
1501
1502 ret = btrfs_comp_cpu_keys(key, &test);
1503 if (ret > 0)
ac8e9819
CM		 1504 return 0;
1505 return 1;
1506}
1507
1508static noinline int copy_to_sk(struct btrfs_root *root,
1509 struct btrfs_path *path,
1510 struct btrfs_key *key,
1511 struct btrfs_ioctl_search_key *sk,
1512 char *buf,
1513 unsigned long *sk_offset,
1514 int *num_found)
1515{
1516 u64 found_transid;
1517 struct extent_buffer *leaf;
1518 struct btrfs_ioctl_search_header sh;
1519 unsigned long item_off;
1520 unsigned long item_len;
1521 int nritems;
1522 int i;
1523 int slot;
ac8e9819
CM		 1524 int ret = 0;
1525
1526 leaf = path->nodes[0];
1527 slot = path->slots[0];
1528 nritems = btrfs_header_nritems(leaf);
1529
1530 if (btrfs_header_generation(leaf) > sk->max_transid) {
1531 i = nritems;
1532 goto advance_key;
1533 }
1534 found_transid = btrfs_header_generation(leaf);
1535
1536 for (i = slot; i < nritems; i++) {
1537 item_off = btrfs_item_ptr_offset(leaf, i);
1538 item_len = btrfs_item_size_nr(leaf, i);
1539
1540 if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1541 item_len = 0;
1542
1543 if (sizeof(sh) + item_len + *sk_offset >
1544 BTRFS_SEARCH_ARGS_BUFSIZE) {
1545 ret = 1;
1546 goto overflow;
1547 }
1548
1549 btrfs_item_key_to_cpu(leaf, key, i);
1550 if (!key_in_sk(key, sk))
1551 continue;
1552
1553 sh.objectid = key->objectid;
1554 sh.offset = key->offset;
1555 sh.type = key->type;
1556 sh.len = item_len;
1557 sh.transid = found_transid;
1558
1559 /* copy search result header */
1560 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1561 *sk_offset += sizeof(sh);
1562
1563 if (item_len) {
1564 char *p = buf + *sk_offset;
1565 /* copy the item */
1566 read_extent_buffer(leaf, p,
1567 item_off, item_len);
1568 *sk_offset += item_len;
ac8e9819 1569 }
e2156867 1570 (*num_found)++;
ac8e9819
CM		 1571
1572 if (*num_found >= sk->nr_items)
1573 break;
1574 }
1575advance_key:
abc6e134
CM		 1576 ret = 0;
1577 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
ac8e9819 1578 key->offset++;
abc6e134
CM		 1579 else if (key->type < (u8)-1 && key->type < sk->max_type) {
1580 key->offset = 0;
ac8e9819 1581 key->type++;
abc6e134
CM		 1582 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1583 key->offset = 0;
1584 key->type = 0;
ac8e9819 1585 key->objectid++;
abc6e134
CM		 1586 } else
1587 ret = 1;
ac8e9819 1588overflow:
ac8e9819
CM		 1589 return ret;
1590}
1591
1592static noinline int search_ioctl(struct inode *inode,
1593 struct btrfs_ioctl_search_args *args)
1594{
1595 struct btrfs_root *root;
1596 struct btrfs_key key;
1597 struct btrfs_key max_key;
1598 struct btrfs_path *path;
1599 struct btrfs_ioctl_search_key *sk = &args->key;
1600 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1601 int ret;
1602 int num_found = 0;
1603 unsigned long sk_offset = 0;
1604
1605 path = btrfs_alloc_path();
1606 if (!path)
1607 return -ENOMEM;
1608
1609 if (sk->tree_id == 0) {
1610 /* search the root of the inode that was passed */
1611 root = BTRFS_I(inode)->root;
1612 } else {
1613 key.objectid = sk->tree_id;
1614 key.type = BTRFS_ROOT_ITEM_KEY;
1615 key.offset = (u64)-1;
1616 root = btrfs_read_fs_root_no_name(info, &key);
1617 if (IS_ERR(root)) {
1618 printk(KERN_ERR "could not find root %llu\n",
1619 sk->tree_id);
1620 btrfs_free_path(path);
1621 return -ENOENT;
1622 }
1623 }
1624
1625 key.objectid = sk->min_objectid;
1626 key.type = sk->min_type;
1627 key.offset = sk->min_offset;
1628
1629 max_key.objectid = sk->max_objectid;
1630 max_key.type = sk->max_type;
1631 max_key.offset = sk->max_offset;
1632
1633 path->keep_locks = 1;
1634
1635 while(1) {
1636 ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1637 sk->min_transid);
1638 if (ret != 0) {
1639 if (ret > 0)
1640 ret = 0;
1641 goto err;
1642 }
1643 ret = copy_to_sk(root, path, &key, sk, args->buf,
1644 &sk_offset, &num_found);
b3b4aa74 1645 btrfs_release_path(path);
ac8e9819
CM		 1646 if (ret || num_found >= sk->nr_items)
1647 break;
1648
1649 }
1650 ret = 0;
1651err:
1652 sk->nr_items = num_found;
1653 btrfs_free_path(path);
1654 return ret;
1655}
1656
1657static noinline int btrfs_ioctl_tree_search(struct file *file,
1658 void __user *argp)
1659{
1660 struct btrfs_ioctl_search_args *args;
1661 struct inode *inode;
1662 int ret;
1663
1664 if (!capable(CAP_SYS_ADMIN))
1665 return -EPERM;
1666
2354d08f
JL		 1667 args = memdup_user(argp, sizeof(*args));
1668 if (IS_ERR(args))
1669 return PTR_ERR(args);
ac8e9819 1670
ac8e9819
CM		 1671 inode = fdentry(file)->d_inode;
1672 ret = search_ioctl(inode, args);
1673 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1674 ret = -EFAULT;
1675 kfree(args);
1676 return ret;
1677}
1678
98d377a0 1679/*
ac8e9819
CM		 1680 * Search INODE_REFs to identify path name of 'dirid' directory
1681 * in a 'tree_id' tree. and sets path name to 'name'.
1682 */
98d377a0
TH		 1683static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1684 u64 tree_id, u64 dirid, char *name)
1685{
1686 struct btrfs_root *root;
1687 struct btrfs_key key;
ac8e9819 1688 char *ptr;
98d377a0
TH		 1689 int ret = -1;
1690 int slot;
1691 int len;
1692 int total_len = 0;
1693 struct btrfs_inode_ref *iref;
1694 struct extent_buffer *l;
1695 struct btrfs_path *path;
1696
1697 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1698 name[0]='\0';
1699 return 0;
1700 }
1701
1702 path = btrfs_alloc_path();
1703 if (!path)
1704 return -ENOMEM;
1705
ac8e9819 1706 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
98d377a0
TH		 1707
1708 key.objectid = tree_id;
1709 key.type = BTRFS_ROOT_ITEM_KEY;
1710 key.offset = (u64)-1;
1711 root = btrfs_read_fs_root_no_name(info, &key);
1712 if (IS_ERR(root)) {
1713 printk(KERN_ERR "could not find root %llu\n", tree_id);
8ad6fcab
CM		 1714 ret = -ENOENT;
1715 goto out;
98d377a0
TH		 1716 }
1717
1718 key.objectid = dirid;
1719 key.type = BTRFS_INODE_REF_KEY;
8ad6fcab 1720 key.offset = (u64)-1;
98d377a0
TH		 1721
1722 while(1) {
1723 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1724 if (ret < 0)
1725 goto out;
1726
1727 l = path->nodes[0];
1728 slot = path->slots[0];
8ad6fcab
CM		 1729 if (ret > 0 && slot > 0)
1730 slot--;
98d377a0
TH		 1731 btrfs_item_key_to_cpu(l, &key, slot);
1732
1733 if (ret > 0 && (key.objectid != dirid ||
ac8e9819
CM		 1734 key.type != BTRFS_INODE_REF_KEY)) {
1735 ret = -ENOENT;
98d377a0 1736 goto out;
ac8e9819 1737 }
98d377a0
TH		 1738
1739 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1740 len = btrfs_inode_ref_name_len(l, iref);
1741 ptr -= len + 1;
1742 total_len += len + 1;
ac8e9819 1743 if (ptr < name)
98d377a0
TH		 1744 goto out;
1745
1746 *(ptr + len) = '/';
1747 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1748
1749 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1750 break;
1751
b3b4aa74 1752 btrfs_release_path(path);
98d377a0 1753 key.objectid = key.offset;
8ad6fcab 1754 key.offset = (u64)-1;
98d377a0
TH		 1755 dirid = key.objectid;
1756
1757 }
ac8e9819 1758 if (ptr < name)
98d377a0 1759 goto out;
ac8e9819 1760 memcpy(name, ptr, total_len);
98d377a0
TH		 1761 name[total_len]='\0';
1762 ret = 0;
1763out:
1764 btrfs_free_path(path);
ac8e9819
CM		 1765 return ret;
1766}
1767
1768static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1769 void __user *argp)
1770{
1771 struct btrfs_ioctl_ino_lookup_args *args;
1772 struct inode *inode;
1773 int ret;
1774
1775 if (!capable(CAP_SYS_ADMIN))
1776 return -EPERM;
1777
2354d08f
JL		 1778 args = memdup_user(argp, sizeof(*args));
1779 if (IS_ERR(args))
1780 return PTR_ERR(args);
c2b96929 1781
ac8e9819
CM		 1782 inode = fdentry(file)->d_inode;
1783
1b53ac4d
CM		 1784 if (args->treeid == 0)
1785 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1786
ac8e9819
CM		 1787 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1788 args->treeid, args->objectid,
1789 args->name);
1790
1791 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1792 ret = -EFAULT;
1793
1794 kfree(args);
98d377a0
TH		 1795 return ret;
1796}
1797
76dda93c
YZ		 1798static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1799 void __user *arg)
1800{
1801 struct dentry *parent = fdentry(file);
1802 struct dentry *dentry;
1803 struct inode *dir = parent->d_inode;
1804 struct inode *inode;
1805 struct btrfs_root *root = BTRFS_I(dir)->root;
1806 struct btrfs_root *dest = NULL;
1807 struct btrfs_ioctl_vol_args *vol_args;
1808 struct btrfs_trans_handle *trans;
1809 int namelen;
1810 int ret;
1811 int err = 0;
1812
76dda93c
YZ		 1813 vol_args = memdup_user(arg, sizeof(*vol_args));
1814 if (IS_ERR(vol_args))
1815 return PTR_ERR(vol_args);
1816
1817 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1818 namelen = strlen(vol_args->name);
1819 if (strchr(vol_args->name, '/') ||
1820 strncmp(vol_args->name, "..", namelen) == 0) {
1821 err = -EINVAL;
1822 goto out;
1823 }
1824
1825 err = mnt_want_write(file->f_path.mnt);
1826 if (err)
1827 goto out;
1828
1829 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
1830 dentry = lookup_one_len(vol_args->name, parent, namelen);
1831 if (IS_ERR(dentry)) {
1832 err = PTR_ERR(dentry);
1833 goto out_unlock_dir;
1834 }
1835
1836 if (!dentry->d_inode) {
1837 err = -ENOENT;
1838 goto out_dput;
1839 }
1840
1841 inode = dentry->d_inode;
4260f7c7
SW		 1842 dest = BTRFS_I(inode)->root;
1843 if (!capable(CAP_SYS_ADMIN)){
1844 /*
1845 * Regular user. Only allow this with a special mount
1846 * option, when the user has write+exec access to the
1847 * subvol root, and when rmdir(2) would have been
1848 * allowed.
1849 *
1850 * Note that this is _not_ check that the subvol is
1851 * empty or doesn't contain data that we wouldn't
1852 * otherwise be able to delete.
1853 *
1854 * Users who want to delete empty subvols should try
1855 * rmdir(2).
1856 */
1857 err = -EPERM;
1858 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1859 goto out_dput;
1860
1861 /*
1862 * Do not allow deletion if the parent dir is the same
1863 * as the dir to be deleted. That means the ioctl
1864 * must be called on the dentry referencing the root
1865 * of the subvol, not a random directory contained
1866 * within it.
1867 */
1868 err = -EINVAL;
1869 if (root == dest)
1870 goto out_dput;
1871
1872 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
1873 if (err)
1874 goto out_dput;
1875
1876 /* check if subvolume may be deleted by a non-root user */
1877 err = btrfs_may_delete(dir, dentry, 1);
1878 if (err)
1879 goto out_dput;
1880 }
1881
33345d01 1882 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
76dda93c
YZ		 1883 err = -EINVAL;
1884 goto out_dput;
1885 }
1886
76dda93c
YZ		 1887 mutex_lock(&inode->i_mutex);
1888 err = d_invalidate(dentry);
1889 if (err)
1890 goto out_unlock;
1891
1892 down_write(&root->fs_info->subvol_sem);
1893
1894 err = may_destroy_subvol(dest);
1895 if (err)
1896 goto out_up_write;
1897
a22285a6
YZ		 1898 trans = btrfs_start_transaction(root, 0);
1899 if (IS_ERR(trans)) {
1900 err = PTR_ERR(trans);
d327099a 1901 goto out_up_write;
a22285a6
YZ		 1902 }
1903 trans->block_rsv = &root->fs_info->global_block_rsv;
1904
76dda93c
YZ		 1905 ret = btrfs_unlink_subvol(trans, root, dir,
1906 dest->root_key.objectid,
1907 dentry->d_name.name,
1908 dentry->d_name.len);
1909 BUG_ON(ret);
1910
1911 btrfs_record_root_in_trans(trans, dest);
1912
1913 memset(&dest->root_item.drop_progress, 0,
1914 sizeof(dest->root_item.drop_progress));
1915 dest->root_item.drop_level = 0;
1916 btrfs_set_root_refs(&dest->root_item, 0);
1917
d68fc57b
YZ		 1918 if (!xchg(&dest->orphan_item_inserted, 1)) {
1919 ret = btrfs_insert_orphan_item(trans,
1920 root->fs_info->tree_root,
1921 dest->root_key.objectid);
1922 BUG_ON(ret);
1923 }
76dda93c 1924
531cb13f 1925 ret = btrfs_end_transaction(trans, root);
76dda93c
YZ		 1926 BUG_ON(ret);
1927 inode->i_flags |= S_DEAD;
1928out_up_write:
1929 up_write(&root->fs_info->subvol_sem);
1930out_unlock:
1931 mutex_unlock(&inode->i_mutex);
1932 if (!err) {
efefb143 1933 shrink_dcache_sb(root->fs_info->sb);
76dda93c
YZ		 1934 btrfs_invalidate_inodes(dest);
1935 d_delete(dentry);
1936 }
1937out_dput:
1938 dput(dentry);
1939out_unlock_dir:
1940 mutex_unlock(&dir->i_mutex);
1941 mnt_drop_write(file->f_path.mnt);
1942out:
1943 kfree(vol_args);
1944 return err;
1945}
1946
1e701a32 1947static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
f46b5a66
CH		 1948{
1949 struct inode *inode = fdentry(file)->d_inode;
1950 struct btrfs_root *root = BTRFS_I(inode)->root;
1e701a32 1951 struct btrfs_ioctl_defrag_range_args *range;
c146afad
YZ		 1952 int ret;
1953
b83cc969
LZ		 1954 if (btrfs_root_readonly(root))
1955 return -EROFS;
1956
c146afad
YZ		 1957 ret = mnt_want_write(file->f_path.mnt);
1958 if (ret)
1959 return ret;
f46b5a66
CH		 1960
1961 switch (inode->i_mode & S_IFMT) {
1962 case S_IFDIR:
e441d54d
CM		 1963 if (!capable(CAP_SYS_ADMIN)) {
1964 ret = -EPERM;
1965 goto out;
1966 }
8929ecfa
YZ		 1967 ret = btrfs_defrag_root(root, 0);
1968 if (ret)
1969 goto out;
1970 ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
f46b5a66
CH		 1971 break;
1972 case S_IFREG:
e441d54d
CM		 1973 if (!(file->f_mode & FMODE_WRITE)) {
1974 ret = -EINVAL;
1975 goto out;
1976 }
1e701a32
CM		 1977
1978 range = kzalloc(sizeof(*range), GFP_KERNEL);
1979 if (!range) {
1980 ret = -ENOMEM;
1981 goto out;
1982 }
1983
1984 if (argp) {
1985 if (copy_from_user(range, argp,
1986 sizeof(*range))) {
1987 ret = -EFAULT;
1988 kfree(range);
683be16e 1989 goto out;
1e701a32
CM		 1990 }
1991 /* compression requires us to start the IO */
1992 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1993 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
1994 range->extent_thresh = (u32)-1;
1995 }
1996 } else {
1997 /* the rest are all set to zero by kzalloc */
1998 range->len = (u64)-1;
1999 }
4cb5300b
CM		 2000 ret = btrfs_defrag_file(fdentry(file)->d_inode, file,
2001 range, 0, 0);
2002 if (ret > 0)
2003 ret = 0;
1e701a32 2004 kfree(range);
f46b5a66 2005 break;
8929ecfa
YZ		 2006 default:
2007 ret = -EINVAL;
f46b5a66 2008 }
e441d54d 2009out:
ab67b7c1 2010 mnt_drop_write(file->f_path.mnt);
e441d54d 2011 return ret;
f46b5a66
CH		 2012}
2013
b2950863 2014static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
f46b5a66
CH		 2015{
2016 struct btrfs_ioctl_vol_args *vol_args;
2017 int ret;
2018
e441d54d
CM		 2019 if (!capable(CAP_SYS_ADMIN))
2020 return -EPERM;
2021
dae7b665
LZ		 2022 vol_args = memdup_user(arg, sizeof(*vol_args));
2023 if (IS_ERR(vol_args))
2024 return PTR_ERR(vol_args);
f46b5a66 2025
5516e595 2026 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66
CH		 2027 ret = btrfs_init_new_device(root, vol_args->name);
2028
f46b5a66
CH		 2029 kfree(vol_args);
2030 return ret;
2031}
2032
b2950863 2033static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
f46b5a66
CH		 2034{
2035 struct btrfs_ioctl_vol_args *vol_args;
2036 int ret;
2037
e441d54d
CM		 2038 if (!capable(CAP_SYS_ADMIN))
2039 return -EPERM;
2040
c146afad
YZ		 2041 if (root->fs_info->sb->s_flags & MS_RDONLY)
2042 return -EROFS;
2043
dae7b665
LZ		 2044 vol_args = memdup_user(arg, sizeof(*vol_args));
2045 if (IS_ERR(vol_args))
2046 return PTR_ERR(vol_args);
f46b5a66 2047
5516e595 2048 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66
CH		 2049 ret = btrfs_rm_device(root, vol_args->name);
2050
f46b5a66
CH		 2051 kfree(vol_args);
2052 return ret;
2053}
2054
475f6387
JS		 2055static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2056{
2057 struct btrfs_ioctl_fs_info_args fi_args;
2058 struct btrfs_device *device;
2059 struct btrfs_device *next;
2060 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2061
2062 if (!capable(CAP_SYS_ADMIN))
2063 return -EPERM;
2064
2065 fi_args.num_devices = fs_devices->num_devices;
2066 fi_args.max_id = 0;
2067 memcpy(&fi_args.fsid, root->fs_info->fsid, sizeof(fi_args.fsid));
2068
2069 mutex_lock(&fs_devices->device_list_mutex);
2070 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
2071 if (device->devid > fi_args.max_id)
2072 fi_args.max_id = device->devid;
2073 }
2074 mutex_unlock(&fs_devices->device_list_mutex);
2075
2076 if (copy_to_user(arg, &fi_args, sizeof(fi_args)))
2077 return -EFAULT;
2078
2079 return 0;
2080}
2081
2082static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2083{
2084 struct btrfs_ioctl_dev_info_args *di_args;
2085 struct btrfs_device *dev;
2086 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2087 int ret = 0;
2088 char *s_uuid = NULL;
2089 char empty_uuid[BTRFS_UUID_SIZE] = {0};
2090
2091 if (!capable(CAP_SYS_ADMIN))
2092 return -EPERM;
2093
2094 di_args = memdup_user(arg, sizeof(*di_args));
2095 if (IS_ERR(di_args))
2096 return PTR_ERR(di_args);
2097
2098 if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2099 s_uuid = di_args->uuid;
2100
2101 mutex_lock(&fs_devices->device_list_mutex);
2102 dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL);
2103 mutex_unlock(&fs_devices->device_list_mutex);
2104
2105 if (!dev) {
2106 ret = -ENODEV;
2107 goto out;
2108 }
2109
2110 di_args->devid = dev->devid;
2111 di_args->bytes_used = dev->bytes_used;
2112 di_args->total_bytes = dev->total_bytes;
2113 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2114 strncpy(di_args->path, dev->name, sizeof(di_args->path));
2115
2116out:
2117 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2118 ret = -EFAULT;
2119
2120 kfree(di_args);
2121 return ret;
2122}
2123
76dda93c
YZ		 2124static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2125 u64 off, u64 olen, u64 destoff)
f46b5a66
CH		 2126{
2127 struct inode *inode = fdentry(file)->d_inode;
2128 struct btrfs_root *root = BTRFS_I(inode)->root;
2129 struct file *src_file;
2130 struct inode *src;
2131 struct btrfs_trans_handle *trans;
f46b5a66 2132 struct btrfs_path *path;
f46b5a66 2133 struct extent_buffer *leaf;
ae01a0ab
YZ		 2134 char *buf;
2135 struct btrfs_key key;
f46b5a66
CH		 2136 u32 nritems;
2137 int slot;
ae01a0ab 2138 int ret;
c5c9cd4d
SW		 2139 u64 len = olen;
2140 u64 bs = root->fs_info->sb->s_blocksize;
2141 u64 hint_byte;
d20f7043 2142
c5c9cd4d
SW		 2143 /*
2144 * TODO:
2145 * - split compressed inline extents. annoying: we need to
2146 * decompress into destination's address_space (the file offset
2147 * may change, so source mapping won't do), then recompress (or
2148 * otherwise reinsert) a subrange.
2149 * - allow ranges within the same file to be cloned (provided
2150 * they don't overlap)?
2151 */
2152
e441d54d 2153 /* the destination must be opened for writing */
2ebc3464 2154 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
e441d54d
CM		 2155 return -EINVAL;
2156
b83cc969
LZ		 2157 if (btrfs_root_readonly(root))
2158 return -EROFS;
2159
c146afad
YZ		 2160 ret = mnt_want_write(file->f_path.mnt);
2161 if (ret)
2162 return ret;
2163
c5c9cd4d 2164 src_file = fget(srcfd);
ab67b7c1
YZ		 2165 if (!src_file) {
2166 ret = -EBADF;
2167 goto out_drop_write;
2168 }
5dc64164 2169
f46b5a66
CH		 2170 src = src_file->f_dentry->d_inode;
2171
c5c9cd4d
SW		 2172 ret = -EINVAL;
2173 if (src == inode)
2174 goto out_fput;
2175
5dc64164
DR		 2176 /* the src must be open for reading */
2177 if (!(src_file->f_mode & FMODE_READ))
2178 goto out_fput;
2179
ae01a0ab
YZ		 2180 ret = -EISDIR;
2181 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2182 goto out_fput;
2183
f46b5a66 2184 ret = -EXDEV;
ae01a0ab
YZ		 2185 if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
2186 goto out_fput;
2187
2188 ret = -ENOMEM;
2189 buf = vmalloc(btrfs_level_size(root, 0));
2190 if (!buf)
2191 goto out_fput;
2192
2193 path = btrfs_alloc_path();
2194 if (!path) {
2195 vfree(buf);
f46b5a66 2196 goto out_fput;
ae01a0ab
YZ		 2197 }
2198 path->reada = 2;
f46b5a66
CH		 2199
2200 if (inode < src) {
fccdae43
SW		 2201 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2202 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
f46b5a66 2203 } else {
fccdae43
SW		 2204 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2205 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
f46b5a66
CH		 2206 }
2207
c5c9cd4d
SW		 2208 /* determine range to clone */
2209 ret = -EINVAL;
2ebc3464 2210 if (off + len > src->i_size || off + len < off)
f46b5a66 2211 goto out_unlock;
c5c9cd4d
SW		 2212 if (len == 0)
2213 olen = len = src->i_size - off;
2214 /* if we extend to eof, continue to block boundary */
2215 if (off + len == src->i_size)
2a6b8dae 2216 len = ALIGN(src->i_size, bs) - off;
c5c9cd4d
SW		 2217
2218 /* verify the end result is block aligned */
2a6b8dae
LZ		 2219 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2220 !IS_ALIGNED(destoff, bs))
c5c9cd4d
SW		 2221 goto out_unlock;
2222
f46b5a66
CH		 2223 /* do any pending delalloc/csum calc on src, one way or
2224 another, and lock file content */
2225 while (1) {
31840ae1 2226 struct btrfs_ordered_extent *ordered;
c5c9cd4d 2227 lock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
9a019196
SW		 2228 ordered = btrfs_lookup_first_ordered_extent(src, off+len);
2229 if (!ordered &&
2230 !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len,
2231 EXTENT_DELALLOC, 0, NULL))
f46b5a66 2232 break;
c5c9cd4d 2233 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
ae01a0ab
YZ		 2234 if (ordered)
2235 btrfs_put_ordered_extent(ordered);
9a019196 2236 btrfs_wait_ordered_range(src, off, len);
f46b5a66
CH		 2237 }
2238
c5c9cd4d 2239 /* clone data */
33345d01 2240 key.objectid = btrfs_ino(src);
ae01a0ab
YZ		 2241 key.type = BTRFS_EXTENT_DATA_KEY;
2242 key.offset = 0;
f46b5a66
CH		 2243
2244 while (1) {
2245 /*
2246 * note the key will change type as we walk through the
2247 * tree.
2248 */
a22285a6 2249 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
f46b5a66
CH		 2250 if (ret < 0)
2251 goto out;
2252
ae01a0ab
YZ		 2253 nritems = btrfs_header_nritems(path->nodes[0]);
2254 if (path->slots[0] >= nritems) {
f46b5a66
CH		 2255 ret = btrfs_next_leaf(root, path);
2256 if (ret < 0)
2257 goto out;
2258 if (ret > 0)
2259 break;
ae01a0ab 2260 nritems = btrfs_header_nritems(path->nodes[0]);
f46b5a66
CH		 2261 }
2262 leaf = path->nodes[0];
2263 slot = path->slots[0];
f46b5a66 2264
ae01a0ab 2265 btrfs_item_key_to_cpu(leaf, &key, slot);
d20f7043 2266 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
33345d01 2267 key.objectid != btrfs_ino(src))
f46b5a66
CH		 2268 break;
2269
c5c9cd4d
SW		 2270 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2271 struct btrfs_file_extent_item *extent;
2272 int type;
31840ae1
ZY		 2273 u32 size;
2274 struct btrfs_key new_key;
c5c9cd4d
SW		 2275 u64 disko = 0, diskl = 0;
2276 u64 datao = 0, datal = 0;
2277 u8 comp;
b5384d48 2278 u64 endoff;
31840ae1
ZY		 2279
2280 size = btrfs_item_size_nr(leaf, slot);
2281 read_extent_buffer(leaf, buf,
2282 btrfs_item_ptr_offset(leaf, slot),
2283 size);
c5c9cd4d
SW		 2284
2285 extent = btrfs_item_ptr(leaf, slot,
2286 struct btrfs_file_extent_item);
2287 comp = btrfs_file_extent_compression(leaf, extent);
2288 type = btrfs_file_extent_type(leaf, extent);
c8a894d7
CM		 2289 if (type == BTRFS_FILE_EXTENT_REG ||
2290 type == BTRFS_FILE_EXTENT_PREALLOC) {
d397712b
CM		 2291 disko = btrfs_file_extent_disk_bytenr(leaf,
2292 extent);
2293 diskl = btrfs_file_extent_disk_num_bytes(leaf,
2294 extent);
c5c9cd4d 2295 datao = btrfs_file_extent_offset(leaf, extent);
d397712b
CM		 2296 datal = btrfs_file_extent_num_bytes(leaf,
2297 extent);
c5c9cd4d
SW		 2298 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2299 /* take upper bound, may be compressed */
2300 datal = btrfs_file_extent_ram_bytes(leaf,
2301 extent);
2302 }
b3b4aa74 2303 btrfs_release_path(path);
31840ae1 2304
050006a7 2305 if (key.offset + datal <= off ||
c5c9cd4d
SW		 2306 key.offset >= off+len)
2307 goto next;
2308
31840ae1 2309 memcpy(&new_key, &key, sizeof(new_key));
33345d01 2310 new_key.objectid = btrfs_ino(inode);
4d728ec7
LZ		 2311 if (off <= key.offset)
2312 new_key.offset = key.offset + destoff - off;
2313 else
2314 new_key.offset = destoff;
31840ae1 2315
a22285a6
YZ		 2316 trans = btrfs_start_transaction(root, 1);
2317 if (IS_ERR(trans)) {
2318 ret = PTR_ERR(trans);
2319 goto out;
2320 }
2321
c8a894d7
CM		 2322 if (type == BTRFS_FILE_EXTENT_REG ||
2323 type == BTRFS_FILE_EXTENT_PREALLOC) {
a22285a6
YZ		 2324 if (off > key.offset) {
2325 datao += off - key.offset;
2326 datal -= off - key.offset;
2327 }
2328
2329 if (key.offset + datal > off + len)
2330 datal = off + len - key.offset;
2331
2332 ret = btrfs_drop_extents(trans, inode,
2333 new_key.offset,
2334 new_key.offset + datal,
2335 &hint_byte, 1);
2336 BUG_ON(ret);
2337
c5c9cd4d
SW		 2338 ret = btrfs_insert_empty_item(trans, root, path,
2339 &new_key, size);
a22285a6 2340 BUG_ON(ret);
c5c9cd4d
SW		 2341
2342 leaf = path->nodes[0];
2343 slot = path->slots[0];
2344 write_extent_buffer(leaf, buf,
31840ae1
ZY		 2345 btrfs_item_ptr_offset(leaf, slot),
2346 size);
ae01a0ab 2347
c5c9cd4d 2348 extent = btrfs_item_ptr(leaf, slot,
f46b5a66 2349 struct btrfs_file_extent_item);
c5c9cd4d 2350
c5c9cd4d
SW		 2351 /* disko == 0 means it's a hole */
2352 if (!disko)
2353 datao = 0;
c5c9cd4d
SW		 2354
2355 btrfs_set_file_extent_offset(leaf, extent,
2356 datao);
2357 btrfs_set_file_extent_num_bytes(leaf, extent,
2358 datal);
2359 if (disko) {
2360 inode_add_bytes(inode, datal);
ae01a0ab 2361 ret = btrfs_inc_extent_ref(trans, root,
5d4f98a2
YZ		 2362 disko, diskl, 0,
2363 root->root_key.objectid,
33345d01 2364 btrfs_ino(inode),
5d4f98a2 2365 new_key.offset - datao);
31840ae1 2366 BUG_ON(ret);
f46b5a66 2367 }
c5c9cd4d
SW		 2368 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2369 u64 skip = 0;
2370 u64 trim = 0;
2371 if (off > key.offset) {
2372 skip = off - key.offset;
2373 new_key.offset += skip;
2374 }
d397712b 2375
c5c9cd4d
SW		 2376 if (key.offset + datal > off+len)
2377 trim = key.offset + datal - (off+len);
d397712b 2378
c5c9cd4d 2379 if (comp && (skip || trim)) {
c5c9cd4d 2380 ret = -EINVAL;
a22285a6 2381 btrfs_end_transaction(trans, root);
c5c9cd4d
SW		 2382 goto out;
2383 }
2384 size -= skip + trim;
2385 datal -= skip + trim;
a22285a6
YZ		 2386
2387 ret = btrfs_drop_extents(trans, inode,
2388 new_key.offset,
2389 new_key.offset + datal,
2390 &hint_byte, 1);
2391 BUG_ON(ret);
2392
c5c9cd4d
SW		 2393 ret = btrfs_insert_empty_item(trans, root, path,
2394 &new_key, size);
a22285a6 2395 BUG_ON(ret);
c5c9cd4d
SW		 2396
2397 if (skip) {
d397712b
CM		 2398 u32 start =
2399 btrfs_file_extent_calc_inline_size(0);
c5c9cd4d
SW		 2400 memmove(buf+start, buf+start+skip,
2401 datal);
2402 }
2403
2404 leaf = path->nodes[0];
2405 slot = path->slots[0];
2406 write_extent_buffer(leaf, buf,
2407 btrfs_item_ptr_offset(leaf, slot),
2408 size);
2409 inode_add_bytes(inode, datal);
f46b5a66 2410 }
c5c9cd4d
SW		 2411
2412 btrfs_mark_buffer_dirty(leaf);
b3b4aa74 2413 btrfs_release_path(path);
c5c9cd4d 2414
a22285a6 2415 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
b5384d48
SW		 2416
2417 /*
2418 * we round up to the block size at eof when
2419 * determining which extents to clone above,
2420 * but shouldn't round up the file size
2421 */
2422 endoff = new_key.offset + datal;
5f3888ff
LZ		 2423 if (endoff > destoff+olen)
2424 endoff = destoff+olen;
b5384d48
SW		 2425 if (endoff > inode->i_size)
2426 btrfs_i_size_write(inode, endoff);
2427
a22285a6
YZ		 2428 BTRFS_I(inode)->flags = BTRFS_I(src)->flags;
2429 ret = btrfs_update_inode(trans, root, inode);
2430 BUG_ON(ret);
2431 btrfs_end_transaction(trans, root);
2432 }
d397712b 2433next:
b3b4aa74 2434 btrfs_release_path(path);
f46b5a66 2435 key.offset++;
f46b5a66 2436 }
f46b5a66
CH		 2437 ret = 0;
2438out:
b3b4aa74 2439 btrfs_release_path(path);
c5c9cd4d 2440 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len, GFP_NOFS);
f46b5a66
CH		 2441out_unlock:
2442 mutex_unlock(&src->i_mutex);
2443 mutex_unlock(&inode->i_mutex);
ae01a0ab
YZ		 2444 vfree(buf);
2445 btrfs_free_path(path);
f46b5a66
CH		 2446out_fput:
2447 fput(src_file);
ab67b7c1
YZ		 2448out_drop_write:
2449 mnt_drop_write(file->f_path.mnt);
f46b5a66
CH		 2450 return ret;
2451}
2452
7a865e8a 2453static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
c5c9cd4d
SW		 2454{
2455 struct btrfs_ioctl_clone_range_args args;
2456
7a865e8a 2457 if (copy_from_user(&args, argp, sizeof(args)))
c5c9cd4d
SW		 2458 return -EFAULT;
2459 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2460 args.src_length, args.dest_offset);
2461}
2462
f46b5a66
CH		 2463/*
2464 * there are many ways the trans_start and trans_end ioctls can lead
2465 * to deadlocks. They should only be used by applications that
2466 * basically own the machine, and have a very in depth understanding
2467 * of all the possible deadlocks and enospc problems.
2468 */
b2950863 2469static long btrfs_ioctl_trans_start(struct file *file)
f46b5a66
CH		 2470{
2471 struct inode *inode = fdentry(file)->d_inode;
2472 struct btrfs_root *root = BTRFS_I(inode)->root;
2473 struct btrfs_trans_handle *trans;
1ab86aed 2474 int ret;
f46b5a66 2475
1ab86aed 2476 ret = -EPERM;
df5b5520 2477 if (!capable(CAP_SYS_ADMIN))
1ab86aed 2478 goto out;
df5b5520 2479
1ab86aed
SW		 2480 ret = -EINPROGRESS;
2481 if (file->private_data)
f46b5a66 2482 goto out;
9ca9ee09 2483
b83cc969
LZ		 2484 ret = -EROFS;
2485 if (btrfs_root_readonly(root))
2486 goto out;
2487
c146afad
YZ		 2488 ret = mnt_want_write(file->f_path.mnt);
2489 if (ret)
2490 goto out;
2491
9ca9ee09
SW		 2492 mutex_lock(&root->fs_info->trans_mutex);
2493 root->fs_info->open_ioctl_trans++;
2494 mutex_unlock(&root->fs_info->trans_mutex);
2495
1ab86aed 2496 ret = -ENOMEM;
9ca9ee09 2497 trans = btrfs_start_ioctl_transaction(root, 0);
abd30bb0 2498 if (IS_ERR(trans))
1ab86aed
SW		 2499 goto out_drop;
2500
2501 file->private_data = trans;
2502 return 0;
2503
2504out_drop:
2505 mutex_lock(&root->fs_info->trans_mutex);
2506 root->fs_info->open_ioctl_trans--;
2507 mutex_unlock(&root->fs_info->trans_mutex);
2508 mnt_drop_write(file->f_path.mnt);
f46b5a66 2509out:
f46b5a66
CH		 2510 return ret;
2511}
2512
6ef5ed0d
JB		 2513static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2514{
2515 struct inode *inode = fdentry(file)->d_inode;
2516 struct btrfs_root *root = BTRFS_I(inode)->root;
2517 struct btrfs_root *new_root;
2518 struct btrfs_dir_item *di;
2519 struct btrfs_trans_handle *trans;
2520 struct btrfs_path *path;
2521 struct btrfs_key location;
2522 struct btrfs_disk_key disk_key;
2523 struct btrfs_super_block *disk_super;
2524 u64 features;
2525 u64 objectid = 0;
2526 u64 dir_id;
2527
2528 if (!capable(CAP_SYS_ADMIN))
2529 return -EPERM;
2530
2531 if (copy_from_user(&objectid, argp, sizeof(objectid)))
2532 return -EFAULT;
2533
2534 if (!objectid)
2535 objectid = root->root_key.objectid;
2536
2537 location.objectid = objectid;
2538 location.type = BTRFS_ROOT_ITEM_KEY;
2539 location.offset = (u64)-1;
2540
2541 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2542 if (IS_ERR(new_root))
2543 return PTR_ERR(new_root);
2544
2545 if (btrfs_root_refs(&new_root->root_item) == 0)
2546 return -ENOENT;
2547
2548 path = btrfs_alloc_path();
2549 if (!path)
2550 return -ENOMEM;
2551 path->leave_spinning = 1;
2552
2553 trans = btrfs_start_transaction(root, 1);
98d5dc13 2554 if (IS_ERR(trans)) {
6ef5ed0d 2555 btrfs_free_path(path);
98d5dc13 2556 return PTR_ERR(trans);
6ef5ed0d
JB		 2557 }
2558
2559 dir_id = btrfs_super_root_dir(&root->fs_info->super_copy);
2560 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2561 dir_id, "default", 7, 1);
cf1e99a4 2562 if (IS_ERR_OR_NULL(di)) {
6ef5ed0d
JB		 2563 btrfs_free_path(path);
2564 btrfs_end_transaction(trans, root);
2565 printk(KERN_ERR "Umm, you don't have the default dir item, "
2566 "this isn't going to work\n");
2567 return -ENOENT;
2568 }
2569
2570 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2571 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2572 btrfs_mark_buffer_dirty(path->nodes[0]);
2573 btrfs_free_path(path);
2574
2575 disk_super = &root->fs_info->super_copy;
2576 features = btrfs_super_incompat_flags(disk_super);
2577 if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) {
2578 features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL;
2579 btrfs_set_super_incompat_flags(disk_super, features);
2580 }
2581 btrfs_end_transaction(trans, root);
2582
2583 return 0;
2584}
2585
bf5fc093
JB		 2586static void get_block_group_info(struct list_head *groups_list,
2587 struct btrfs_ioctl_space_info *space)
2588{
2589 struct btrfs_block_group_cache *block_group;
2590
2591 space->total_bytes = 0;
2592 space->used_bytes = 0;
2593 space->flags = 0;
2594 list_for_each_entry(block_group, groups_list, list) {
2595 space->flags = block_group->flags;
2596 space->total_bytes += block_group->key.offset;
2597 space->used_bytes +=
2598 btrfs_block_group_used(&block_group->item);
2599 }
2600}
2601
1406e432
JB		 2602long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2603{
2604 struct btrfs_ioctl_space_args space_args;
2605 struct btrfs_ioctl_space_info space;
2606 struct btrfs_ioctl_space_info *dest;
7fde62bf 2607 struct btrfs_ioctl_space_info *dest_orig;
13f2696f 2608 struct btrfs_ioctl_space_info __user *user_dest;
1406e432 2609 struct btrfs_space_info *info;
bf5fc093
JB		 2610 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2611 BTRFS_BLOCK_GROUP_SYSTEM,
2612 BTRFS_BLOCK_GROUP_METADATA,
2613 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2614 int num_types = 4;
7fde62bf 2615 int alloc_size;
1406e432 2616 int ret = 0;
51788b1b 2617 u64 slot_count = 0;
bf5fc093 2618 int i, c;
1406e432
JB		 2619
2620 if (copy_from_user(&space_args,
2621 (struct btrfs_ioctl_space_args __user *)arg,
2622 sizeof(space_args)))
2623 return -EFAULT;
2624
bf5fc093
JB		 2625 for (i = 0; i < num_types; i++) {
2626 struct btrfs_space_info *tmp;
2627
2628 info = NULL;
2629 rcu_read_lock();
2630 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2631 list) {
2632 if (tmp->flags == types[i]) {
2633 info = tmp;
2634 break;
2635 }
2636 }
2637 rcu_read_unlock();
2638
2639 if (!info)
2640 continue;
2641
2642 down_read(&info->groups_sem);
2643 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2644 if (!list_empty(&info->block_groups[c]))
2645 slot_count++;
2646 }
2647 up_read(&info->groups_sem);
2648 }
7fde62bf
CM		 2649
2650 /* space_slots == 0 means they are asking for a count */
2651 if (space_args.space_slots == 0) {
2652 space_args.total_spaces = slot_count;
2653 goto out;
2654 }
bf5fc093 2655
51788b1b 2656 slot_count = min_t(u64, space_args.space_slots, slot_count);
bf5fc093 2657
7fde62bf 2658 alloc_size = sizeof(*dest) * slot_count;
bf5fc093 2659
7fde62bf
CM		 2660 /* we generally have at most 6 or so space infos, one for each raid
2661 * level. So, a whole page should be more than enough for everyone
2662 */
2663 if (alloc_size > PAGE_CACHE_SIZE)
2664 return -ENOMEM;
2665
1406e432 2666 space_args.total_spaces = 0;
7fde62bf
CM		 2667 dest = kmalloc(alloc_size, GFP_NOFS);
2668 if (!dest)
2669 return -ENOMEM;
2670 dest_orig = dest;
1406e432 2671
7fde62bf 2672 /* now we have a buffer to copy into */
bf5fc093
JB		 2673 for (i = 0; i < num_types; i++) {
2674 struct btrfs_space_info *tmp;
2675
51788b1b
DR		 2676 if (!slot_count)
2677 break;
2678
bf5fc093
JB		 2679 info = NULL;
2680 rcu_read_lock();
2681 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2682 list) {
2683 if (tmp->flags == types[i]) {
2684 info = tmp;
2685 break;
2686 }
2687 }
2688 rcu_read_unlock();
7fde62bf 2689
bf5fc093
JB		 2690 if (!info)
2691 continue;
2692 down_read(&info->groups_sem);
2693 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2694 if (!list_empty(&info->block_groups[c])) {
2695 get_block_group_info(&info->block_groups[c],
2696 &space);
2697 memcpy(dest, &space, sizeof(space));
2698 dest++;
2699 space_args.total_spaces++;
51788b1b 2700 slot_count--;
bf5fc093 2701 }
51788b1b
DR		 2702 if (!slot_count)
2703 break;
bf5fc093
JB		 2704 }
2705 up_read(&info->groups_sem);
1406e432 2706 }
1406e432 2707
7fde62bf
CM		 2708 user_dest = (struct btrfs_ioctl_space_info *)
2709 (arg + sizeof(struct btrfs_ioctl_space_args));
2710
2711 if (copy_to_user(user_dest, dest_orig, alloc_size))
2712 ret = -EFAULT;
2713
2714 kfree(dest_orig);
2715out:
2716 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
1406e432
JB		 2717 ret = -EFAULT;
2718
2719 return ret;
2720}
2721
f46b5a66
CH		 2722/*
2723 * there are many ways the trans_start and trans_end ioctls can lead
2724 * to deadlocks. They should only be used by applications that
2725 * basically own the machine, and have a very in depth understanding
2726 * of all the possible deadlocks and enospc problems.
2727 */
2728long btrfs_ioctl_trans_end(struct file *file)
2729{
2730 struct inode *inode = fdentry(file)->d_inode;
2731 struct btrfs_root *root = BTRFS_I(inode)->root;
2732 struct btrfs_trans_handle *trans;
f46b5a66 2733
f46b5a66 2734 trans = file->private_data;
1ab86aed
SW		 2735 if (!trans)
2736 return -EINVAL;
b214107e 2737 file->private_data = NULL;
9ca9ee09 2738
1ab86aed
SW		 2739 btrfs_end_transaction(trans, root);
2740
9ca9ee09
SW		 2741 mutex_lock(&root->fs_info->trans_mutex);
2742 root->fs_info->open_ioctl_trans--;
2743 mutex_unlock(&root->fs_info->trans_mutex);
2744
cfc8ea87 2745 mnt_drop_write(file->f_path.mnt);
1ab86aed 2746 return 0;
f46b5a66
CH		 2747}
2748
46204592
SW		 2749static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
2750{
2751 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2752 struct btrfs_trans_handle *trans;
2753 u64 transid;
db5b493a 2754 int ret;
46204592
SW		 2755
2756 trans = btrfs_start_transaction(root, 0);
98d5dc13
TI		 2757 if (IS_ERR(trans))
2758 return PTR_ERR(trans);
46204592 2759 transid = trans->transid;
db5b493a 2760 ret = btrfs_commit_transaction_async(trans, root, 0);
8b2b2d3c
TI		 2761 if (ret) {
2762 btrfs_end_transaction(trans, root);
db5b493a 2763 return ret;
8b2b2d3c 2764 }
46204592
SW		 2765
2766 if (argp)
2767 if (copy_to_user(argp, &transid, sizeof(transid)))
2768 return -EFAULT;
2769 return 0;
2770}
2771
2772static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
2773{
2774 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2775 u64 transid;
2776
2777 if (argp) {
2778 if (copy_from_user(&transid, argp, sizeof(transid)))
2779 return -EFAULT;
2780 } else {
2781 transid = 0; /* current trans */
2782 }
2783 return btrfs_wait_for_commit(root, transid);
2784}
2785
475f6387
JS		 2786static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg)
2787{
2788 int ret;
2789 struct btrfs_ioctl_scrub_args *sa;
2790
2791 if (!capable(CAP_SYS_ADMIN))
2792 return -EPERM;
2793
2794 sa = memdup_user(arg, sizeof(*sa));
2795 if (IS_ERR(sa))
2796 return PTR_ERR(sa);
2797
2798 ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end,
8628764e 2799 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY);
475f6387
JS		 2800
2801 if (copy_to_user(arg, sa, sizeof(*sa)))
2802 ret = -EFAULT;
2803
2804 kfree(sa);
2805 return ret;
2806}
2807
2808static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
2809{
2810 if (!capable(CAP_SYS_ADMIN))
2811 return -EPERM;
2812
2813 return btrfs_scrub_cancel(root);
2814}
2815
2816static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
2817 void __user *arg)
2818{
2819 struct btrfs_ioctl_scrub_args *sa;
2820 int ret;
2821
2822 if (!capable(CAP_SYS_ADMIN))
2823 return -EPERM;
2824
2825 sa = memdup_user(arg, sizeof(*sa));
2826 if (IS_ERR(sa))
2827 return PTR_ERR(sa);
2828
2829 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
2830
2831 if (copy_to_user(arg, sa, sizeof(*sa)))
2832 ret = -EFAULT;
2833
2834 kfree(sa);
2835 return ret;
2836}
2837
f46b5a66
CH		 2838long btrfs_ioctl(struct file *file, unsigned int
2839 cmd, unsigned long arg)
2840{
2841 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
4bcabaa3 2842 void __user *argp = (void __user *)arg;
f46b5a66
CH		 2843
2844 switch (cmd) {
6cbff00f
CH		 2845 case FS_IOC_GETFLAGS:
2846 return btrfs_ioctl_getflags(file, argp);
2847 case FS_IOC_SETFLAGS:
2848 return btrfs_ioctl_setflags(file, argp);
2849 case FS_IOC_GETVERSION:
2850 return btrfs_ioctl_getversion(file, argp);
f7039b1d
LD		 2851 case FITRIM:
2852 return btrfs_ioctl_fitrim(file, argp);
f46b5a66 2853 case BTRFS_IOC_SNAP_CREATE:
fa0d2b9b 2854 return btrfs_ioctl_snap_create(file, argp, 0);
fdfb1e4f 2855 case BTRFS_IOC_SNAP_CREATE_V2:
fa0d2b9b 2856 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3de4586c 2857 case BTRFS_IOC_SUBVOL_CREATE:
fa0d2b9b 2858 return btrfs_ioctl_snap_create(file, argp, 1);
76dda93c
YZ		 2859 case BTRFS_IOC_SNAP_DESTROY:
2860 return btrfs_ioctl_snap_destroy(file, argp);
0caa102d
LZ		 2861 case BTRFS_IOC_SUBVOL_GETFLAGS:
2862 return btrfs_ioctl_subvol_getflags(file, argp);
2863 case BTRFS_IOC_SUBVOL_SETFLAGS:
2864 return btrfs_ioctl_subvol_setflags(file, argp);
6ef5ed0d
JB		 2865 case BTRFS_IOC_DEFAULT_SUBVOL:
2866 return btrfs_ioctl_default_subvol(file, argp);
f46b5a66 2867 case BTRFS_IOC_DEFRAG:
1e701a32
CM		 2868 return btrfs_ioctl_defrag(file, NULL);
2869 case BTRFS_IOC_DEFRAG_RANGE:
2870 return btrfs_ioctl_defrag(file, argp);
f46b5a66 2871 case BTRFS_IOC_RESIZE:
4bcabaa3 2872 return btrfs_ioctl_resize(root, argp);
f46b5a66 2873 case BTRFS_IOC_ADD_DEV:
4bcabaa3 2874 return btrfs_ioctl_add_dev(root, argp);
f46b5a66 2875 case BTRFS_IOC_RM_DEV:
4bcabaa3 2876 return btrfs_ioctl_rm_dev(root, argp);
475f6387
JS		 2877 case BTRFS_IOC_FS_INFO:
2878 return btrfs_ioctl_fs_info(root, argp);
2879 case BTRFS_IOC_DEV_INFO:
2880 return btrfs_ioctl_dev_info(root, argp);
f46b5a66
CH		 2881 case BTRFS_IOC_BALANCE:
2882 return btrfs_balance(root->fs_info->dev_root);
2883 case BTRFS_IOC_CLONE:
c5c9cd4d
SW		 2884 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
2885 case BTRFS_IOC_CLONE_RANGE:
7a865e8a 2886 return btrfs_ioctl_clone_range(file, argp);
f46b5a66
CH		 2887 case BTRFS_IOC_TRANS_START:
2888 return btrfs_ioctl_trans_start(file);
2889 case BTRFS_IOC_TRANS_END:
2890 return btrfs_ioctl_trans_end(file);
ac8e9819
CM		 2891 case BTRFS_IOC_TREE_SEARCH:
2892 return btrfs_ioctl_tree_search(file, argp);
2893 case BTRFS_IOC_INO_LOOKUP:
2894 return btrfs_ioctl_ino_lookup(file, argp);
1406e432
JB		 2895 case BTRFS_IOC_SPACE_INFO:
2896 return btrfs_ioctl_space_info(root, argp);
f46b5a66
CH		 2897 case BTRFS_IOC_SYNC:
2898 btrfs_sync_fs(file->f_dentry->d_sb, 1);
2899 return 0;
46204592
SW		 2900 case BTRFS_IOC_START_SYNC:
2901 return btrfs_ioctl_start_sync(file, argp);
2902 case BTRFS_IOC_WAIT_SYNC:
2903 return btrfs_ioctl_wait_sync(file, argp);
475f6387
JS		 2904 case BTRFS_IOC_SCRUB:
2905 return btrfs_ioctl_scrub(root, argp);
2906 case BTRFS_IOC_SCRUB_CANCEL:
2907 return btrfs_ioctl_scrub_cancel(root, argp);
2908 case BTRFS_IOC_SCRUB_PROGRESS:
2909 return btrfs_ioctl_scrub_progress(root, argp);
f46b5a66
CH		 2910 }
2911
2912 return -ENOTTY;
2913}
Linux 6.x block layer and io_uring tree(s)