]> git.ipfire.org Git - thirdparty/linux.git/blame - fs/btrfs/ioctl.c
projects / thirdparty / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
btrfs: don't return EINTR
[thirdparty/linux.git] / fs / btrfs / ioctl.c
CommitLineData
f46b5a66
CH		 1/*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
19#include <linux/kernel.h>
20#include <linux/bio.h>
21#include <linux/buffer_head.h>
22#include <linux/file.h>
23#include <linux/fs.h>
cb8e7090 24#include <linux/fsnotify.h>
f46b5a66
CH		 25#include <linux/pagemap.h>
26#include <linux/highmem.h>
27#include <linux/time.h>
28#include <linux/init.h>
29#include <linux/string.h>
f46b5a66 30#include <linux/backing-dev.h>
cb8e7090 31#include <linux/mount.h>
f46b5a66 32#include <linux/mpage.h>
cb8e7090 33#include <linux/namei.h>
f46b5a66
CH		 34#include <linux/swap.h>
35#include <linux/writeback.h>
36#include <linux/statfs.h>
37#include <linux/compat.h>
38#include <linux/bit_spinlock.h>
cb8e7090 39#include <linux/security.h>
f46b5a66 40#include <linux/xattr.h>
7ea394f1 41#include <linux/vmalloc.h>
5a0e3ad6 42#include <linux/slab.h>
f7039b1d 43#include <linux/blkdev.h>
4b4e25f2 44#include "compat.h"
f46b5a66
CH		 45#include "ctree.h"
46#include "disk-io.h"
47#include "transaction.h"
48#include "btrfs_inode.h"
49#include "ioctl.h"
50#include "print-tree.h"
51#include "volumes.h"
925baedd 52#include "locking.h"
581bb050 53#include "inode-map.h"
d7728c96 54#include "backref.h"
f46b5a66 55
6cbff00f
CH		 56/* Mask out flags that are inappropriate for the given type of inode. */
57static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
58{
59 if (S_ISDIR(mode))
60 return flags;
61 else if (S_ISREG(mode))
62 return flags & ~FS_DIRSYNC_FL;
63 else
64 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
65}
66
67/*
68 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
69 */
70static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
71{
72 unsigned int iflags = 0;
73
74 if (flags & BTRFS_INODE_SYNC)
75 iflags |= FS_SYNC_FL;
76 if (flags & BTRFS_INODE_IMMUTABLE)
77 iflags |= FS_IMMUTABLE_FL;
78 if (flags & BTRFS_INODE_APPEND)
79 iflags |= FS_APPEND_FL;
80 if (flags & BTRFS_INODE_NODUMP)
81 iflags |= FS_NODUMP_FL;
82 if (flags & BTRFS_INODE_NOATIME)
83 iflags |= FS_NOATIME_FL;
84 if (flags & BTRFS_INODE_DIRSYNC)
85 iflags |= FS_DIRSYNC_FL;
d0092bdd
LZ		 86 if (flags & BTRFS_INODE_NODATACOW)
87 iflags |= FS_NOCOW_FL;
88
89 if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
90 iflags |= FS_COMPR_FL;
91 else if (flags & BTRFS_INODE_NOCOMPRESS)
92 iflags |= FS_NOCOMP_FL;
6cbff00f
CH		 93
94 return iflags;
95}
96
97/*
98 * Update inode->i_flags based on the btrfs internal flags.
99 */
100void btrfs_update_iflags(struct inode *inode)
101{
102 struct btrfs_inode *ip = BTRFS_I(inode);
103
104 inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
105
106 if (ip->flags & BTRFS_INODE_SYNC)
107 inode->i_flags |= S_SYNC;
108 if (ip->flags & BTRFS_INODE_IMMUTABLE)
109 inode->i_flags |= S_IMMUTABLE;
110 if (ip->flags & BTRFS_INODE_APPEND)
111 inode->i_flags |= S_APPEND;
112 if (ip->flags & BTRFS_INODE_NOATIME)
113 inode->i_flags |= S_NOATIME;
114 if (ip->flags & BTRFS_INODE_DIRSYNC)
115 inode->i_flags |= S_DIRSYNC;
116}
117
118/*
119 * Inherit flags from the parent inode.
120 *
e27425d6 121 * Currently only the compression flags and the cow flags are inherited.
6cbff00f
CH		 122 */
123void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
124{
0b4dcea5
CM		 125 unsigned int flags;
126
127 if (!dir)
128 return;
129
130 flags = BTRFS_I(dir)->flags;
6cbff00f 131
e27425d6
JB		 132 if (flags & BTRFS_INODE_NOCOMPRESS) {
133 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
134 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
135 } else if (flags & BTRFS_INODE_COMPRESS) {
136 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
137 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
138 }
139
140 if (flags & BTRFS_INODE_NODATACOW)
141 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
6cbff00f 142
6cbff00f
CH		 143 btrfs_update_iflags(inode);
144}
145
146static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
147{
148 struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode);
149 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
150
151 if (copy_to_user(arg, &flags, sizeof(flags)))
152 return -EFAULT;
153 return 0;
154}
155
75e7cb7f
LB		 156static int check_flags(unsigned int flags)
157{
158 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
159 FS_NOATIME_FL | FS_NODUMP_FL | \
160 FS_SYNC_FL | FS_DIRSYNC_FL | \
e1e8fb6a
LZ		 161 FS_NOCOMP_FL | FS_COMPR_FL |
162 FS_NOCOW_FL))
75e7cb7f
LB		 163 return -EOPNOTSUPP;
164
165 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
166 return -EINVAL;
167
75e7cb7f
LB		 168 return 0;
169}
170
6cbff00f
CH		 171static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
172{
173 struct inode *inode = file->f_path.dentry->d_inode;
174 struct btrfs_inode *ip = BTRFS_I(inode);
175 struct btrfs_root *root = ip->root;
176 struct btrfs_trans_handle *trans;
177 unsigned int flags, oldflags;
178 int ret;
f062abf0
LZ		 179 u64 ip_oldflags;
180 unsigned int i_oldflags;
6cbff00f 181
b83cc969
LZ		 182 if (btrfs_root_readonly(root))
183 return -EROFS;
184
6cbff00f
CH		 185 if (copy_from_user(&flags, arg, sizeof(flags)))
186 return -EFAULT;
187
75e7cb7f
LB		 188 ret = check_flags(flags);
189 if (ret)
190 return ret;
f46b5a66 191
2e149670 192 if (!inode_owner_or_capable(inode))
6cbff00f
CH		 193 return -EACCES;
194
195 mutex_lock(&inode->i_mutex);
196
f062abf0
LZ		 197 ip_oldflags = ip->flags;
198 i_oldflags = inode->i_flags;
199
6cbff00f
CH		 200 flags = btrfs_mask_flags(inode->i_mode, flags);
201 oldflags = btrfs_flags_to_ioctl(ip->flags);
202 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
203 if (!capable(CAP_LINUX_IMMUTABLE)) {
204 ret = -EPERM;
205 goto out_unlock;
206 }
207 }
208
a561be71 209 ret = mnt_want_write_file(file);
6cbff00f
CH		 210 if (ret)
211 goto out_unlock;
212
213 if (flags & FS_SYNC_FL)
214 ip->flags |= BTRFS_INODE_SYNC;
215 else
216 ip->flags &= ~BTRFS_INODE_SYNC;
217 if (flags & FS_IMMUTABLE_FL)
218 ip->flags |= BTRFS_INODE_IMMUTABLE;
219 else
220 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
221 if (flags & FS_APPEND_FL)
222 ip->flags |= BTRFS_INODE_APPEND;
223 else
224 ip->flags &= ~BTRFS_INODE_APPEND;
225 if (flags & FS_NODUMP_FL)
226 ip->flags |= BTRFS_INODE_NODUMP;
227 else
228 ip->flags &= ~BTRFS_INODE_NODUMP;
229 if (flags & FS_NOATIME_FL)
230 ip->flags |= BTRFS_INODE_NOATIME;
231 else
232 ip->flags &= ~BTRFS_INODE_NOATIME;
233 if (flags & FS_DIRSYNC_FL)
234 ip->flags |= BTRFS_INODE_DIRSYNC;
235 else
236 ip->flags &= ~BTRFS_INODE_DIRSYNC;
e1e8fb6a
LZ		 237 if (flags & FS_NOCOW_FL)
238 ip->flags |= BTRFS_INODE_NODATACOW;
239 else
240 ip->flags &= ~BTRFS_INODE_NODATACOW;
6cbff00f 241
75e7cb7f
LB		 242 /*
243 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
244 * flag may be changed automatically if compression code won't make
245 * things smaller.
246 */
247 if (flags & FS_NOCOMP_FL) {
248 ip->flags &= ~BTRFS_INODE_COMPRESS;
249 ip->flags |= BTRFS_INODE_NOCOMPRESS;
250 } else if (flags & FS_COMPR_FL) {
251 ip->flags |= BTRFS_INODE_COMPRESS;
252 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
ebcb904d
LZ		 253 } else {
254 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
75e7cb7f 255 }
6cbff00f 256
4da6f1a3 257 trans = btrfs_start_transaction(root, 1);
f062abf0
LZ		 258 if (IS_ERR(trans)) {
259 ret = PTR_ERR(trans);
260 goto out_drop;
261 }
6cbff00f 262
306424cc
LZ		 263 btrfs_update_iflags(inode);
264 inode->i_ctime = CURRENT_TIME;
6cbff00f 265 ret = btrfs_update_inode(trans, root, inode);
6cbff00f 266
6cbff00f 267 btrfs_end_transaction(trans, root);
f062abf0
LZ		 268 out_drop:
269 if (ret) {
270 ip->flags = ip_oldflags;
271 inode->i_flags = i_oldflags;
272 }
6cbff00f 273
2a79f17e 274 mnt_drop_write_file(file);
6cbff00f
CH		 275 out_unlock:
276 mutex_unlock(&inode->i_mutex);
2d4e6f6a 277 return ret;
6cbff00f
CH		 278}
279
280static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
281{
282 struct inode *inode = file->f_path.dentry->d_inode;
283
284 return put_user(inode->i_generation, arg);
285}
f46b5a66 286
f7039b1d
LD		 287static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
288{
815745cf 289 struct btrfs_fs_info *fs_info = btrfs_sb(fdentry(file)->d_sb);
f7039b1d
LD		 290 struct btrfs_device *device;
291 struct request_queue *q;
292 struct fstrim_range range;
293 u64 minlen = ULLONG_MAX;
294 u64 num_devices = 0;
815745cf 295 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
f7039b1d
LD		 296 int ret;
297
298 if (!capable(CAP_SYS_ADMIN))
299 return -EPERM;
300
1f78160c
XG		 301 rcu_read_lock();
302 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
303 dev_list) {
f7039b1d
LD		 304 if (!device->bdev)
305 continue;
306 q = bdev_get_queue(device->bdev);
307 if (blk_queue_discard(q)) {
308 num_devices++;
309 minlen = min((u64)q->limits.discard_granularity,
310 minlen);
311 }
312 }
1f78160c 313 rcu_read_unlock();
f4c697e6 314
f7039b1d
LD		 315 if (!num_devices)
316 return -EOPNOTSUPP;
f7039b1d
LD		 317 if (copy_from_user(&range, arg, sizeof(range)))
318 return -EFAULT;
f4c697e6
LC		 319 if (range.start > total_bytes)
320 return -EINVAL;
f7039b1d 321
f4c697e6 322 range.len = min(range.len, total_bytes - range.start);
f7039b1d 323 range.minlen = max(range.minlen, minlen);
815745cf 324 ret = btrfs_trim_fs(fs_info->tree_root, &range);
f7039b1d
LD		 325 if (ret < 0)
326 return ret;
327
328 if (copy_to_user(arg, &range, sizeof(range)))
329 return -EFAULT;
330
331 return 0;
332}
333
cb8e7090
CH		 334static noinline int create_subvol(struct btrfs_root *root,
335 struct dentry *dentry,
72fd032e
SW		 336 char *name, int namelen,
337 u64 *async_transid)
f46b5a66
CH		 338{
339 struct btrfs_trans_handle *trans;
340 struct btrfs_key key;
341 struct btrfs_root_item root_item;
342 struct btrfs_inode_item *inode_item;
343 struct extent_buffer *leaf;
76dda93c 344 struct btrfs_root *new_root;
2fbe8c8a 345 struct dentry *parent = dentry->d_parent;
6a912213 346 struct inode *dir;
f46b5a66
CH		 347 int ret;
348 int err;
349 u64 objectid;
350 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
3de4586c 351 u64 index = 0;
f46b5a66 352
581bb050 353 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
2fbe8c8a 354 if (ret)
a22285a6 355 return ret;
6a912213
JB		 356
357 dir = parent->d_inode;
358
9ed74f2d
JB		 359 /*
360 * 1 - inode item
361 * 2 - refs
362 * 1 - root item
363 * 2 - dir items
364 */
a22285a6 365 trans = btrfs_start_transaction(root, 6);
2fbe8c8a 366 if (IS_ERR(trans))
a22285a6 367 return PTR_ERR(trans);
f46b5a66 368
5d4f98a2 369 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
66d7e7f0 370 0, objectid, NULL, 0, 0, 0, 0);
8e8a1e31
JB		 371 if (IS_ERR(leaf)) {
372 ret = PTR_ERR(leaf);
373 goto fail;
374 }
f46b5a66 375
5d4f98a2 376 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
f46b5a66
CH		 377 btrfs_set_header_bytenr(leaf, leaf->start);
378 btrfs_set_header_generation(leaf, trans->transid);
5d4f98a2 379 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
f46b5a66
CH		 380 btrfs_set_header_owner(leaf, objectid);
381
382 write_extent_buffer(leaf, root->fs_info->fsid,
383 (unsigned long)btrfs_header_fsid(leaf),
384 BTRFS_FSID_SIZE);
5d4f98a2
YZ		 385 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
386 (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
387 BTRFS_UUID_SIZE);
f46b5a66
CH		 388 btrfs_mark_buffer_dirty(leaf);
389
390 inode_item = &root_item.inode;
391 memset(inode_item, 0, sizeof(*inode_item));
392 inode_item->generation = cpu_to_le64(1);
393 inode_item->size = cpu_to_le64(3);
394 inode_item->nlink = cpu_to_le32(1);
a76a3cd4 395 inode_item->nbytes = cpu_to_le64(root->leafsize);
f46b5a66
CH		 396 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
397
08fe4db1
LZ		 398 root_item.flags = 0;
399 root_item.byte_limit = 0;
400 inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
401
f46b5a66 402 btrfs_set_root_bytenr(&root_item, leaf->start);
84234f3a 403 btrfs_set_root_generation(&root_item, trans->transid);
f46b5a66
CH		 404 btrfs_set_root_level(&root_item, 0);
405 btrfs_set_root_refs(&root_item, 1);
86b9f2ec 406 btrfs_set_root_used(&root_item, leaf->len);
80ff3856 407 btrfs_set_root_last_snapshot(&root_item, 0);
f46b5a66
CH		 408
409 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
410 root_item.drop_level = 0;
411
925baedd 412 btrfs_tree_unlock(leaf);
f46b5a66
CH		 413 free_extent_buffer(leaf);
414 leaf = NULL;
415
416 btrfs_set_root_dirid(&root_item, new_dirid);
417
418 key.objectid = objectid;
5d4f98a2 419 key.offset = 0;
f46b5a66
CH		 420 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
421 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
422 &root_item);
423 if (ret)
424 goto fail;
425
76dda93c
YZ		 426 key.offset = (u64)-1;
427 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
79787eaa
JM		 428 if (IS_ERR(new_root)) {
429 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
430 ret = PTR_ERR(new_root);
431 goto fail;
432 }
76dda93c
YZ		 433
434 btrfs_record_root_in_trans(trans, new_root);
435
d82a6f1d 436 ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
ce598979
MF		 437 if (ret) {
438 /* We potentially lose an unused inode item here */
79787eaa 439 btrfs_abort_transaction(trans, root, ret);
ce598979
MF		 440 goto fail;
441 }
442
f46b5a66
CH		 443 /*
444 * insert the directory item
445 */
3de4586c 446 ret = btrfs_set_inode_index(dir, &index);
79787eaa
JM		 447 if (ret) {
448 btrfs_abort_transaction(trans, root, ret);
449 goto fail;
450 }
3de4586c
CM		 451
452 ret = btrfs_insert_dir_item(trans, root,
16cdcec7 453 name, namelen, dir, &key,
3de4586c 454 BTRFS_FT_DIR, index);
79787eaa
JM		 455 if (ret) {
456 btrfs_abort_transaction(trans, root, ret);
f46b5a66 457 goto fail;
79787eaa 458 }
0660b5af 459
52c26179
YZ		 460 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
461 ret = btrfs_update_inode(trans, root, dir);
462 BUG_ON(ret);
463
0660b5af 464 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
4df27c4d 465 objectid, root->root_key.objectid,
33345d01 466 btrfs_ino(dir), index, name, namelen);
0660b5af 467
76dda93c 468 BUG_ON(ret);
f46b5a66 469
76dda93c 470 d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
f46b5a66 471fail:
72fd032e
SW		 472 if (async_transid) {
473 *async_transid = trans->transid;
474 err = btrfs_commit_transaction_async(trans, root, 1);
475 } else {
476 err = btrfs_commit_transaction(trans, root);
477 }
f46b5a66
CH		 478 if (err && !ret)
479 ret = err;
f46b5a66
CH		 480 return ret;
481}
482
72fd032e 483static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
b83cc969
LZ		 484 char *name, int namelen, u64 *async_transid,
485 bool readonly)
f46b5a66 486{
2e4bfab9 487 struct inode *inode;
f46b5a66
CH		 488 struct btrfs_pending_snapshot *pending_snapshot;
489 struct btrfs_trans_handle *trans;
2e4bfab9 490 int ret;
f46b5a66
CH		 491
492 if (!root->ref_cows)
493 return -EINVAL;
494
3de4586c 495 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
a22285a6
YZ		 496 if (!pending_snapshot)
497 return -ENOMEM;
498
499 btrfs_init_block_rsv(&pending_snapshot->block_rsv);
3de4586c 500 pending_snapshot->dentry = dentry;
f46b5a66 501 pending_snapshot->root = root;
b83cc969 502 pending_snapshot->readonly = readonly;
a22285a6
YZ		 503
504 trans = btrfs_start_transaction(root->fs_info->extent_root, 5);
505 if (IS_ERR(trans)) {
506 ret = PTR_ERR(trans);
507 goto fail;
508 }
509
510 ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
511 BUG_ON(ret);
512
8351583e 513 spin_lock(&root->fs_info->trans_lock);
f46b5a66
CH		 514 list_add(&pending_snapshot->list,
515 &trans->transaction->pending_snapshots);
8351583e 516 spin_unlock(&root->fs_info->trans_lock);
72fd032e
SW		 517 if (async_transid) {
518 *async_transid = trans->transid;
519 ret = btrfs_commit_transaction_async(trans,
520 root->fs_info->extent_root, 1);
521 } else {
522 ret = btrfs_commit_transaction(trans,
523 root->fs_info->extent_root);
524 }
2e4bfab9 525 BUG_ON(ret);
a22285a6
YZ		 526
527 ret = pending_snapshot->error;
528 if (ret)
529 goto fail;
530
66b4ffd1
JB		 531 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
532 if (ret)
533 goto fail;
f46b5a66 534
2fbe8c8a 535 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
2e4bfab9
YZ		 536 if (IS_ERR(inode)) {
537 ret = PTR_ERR(inode);
538 goto fail;
539 }
540 BUG_ON(!inode);
541 d_instantiate(dentry, inode);
542 ret = 0;
543fail:
a22285a6 544 kfree(pending_snapshot);
f46b5a66
CH		 545 return ret;
546}
547
4260f7c7
SW		 548/* copy of check_sticky in fs/namei.c()
549* It's inline, so penalty for filesystems that don't use sticky bit is
550* minimal.
551*/
552static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
553{
554 uid_t fsuid = current_fsuid();
555
556 if (!(dir->i_mode & S_ISVTX))
557 return 0;
558 if (inode->i_uid == fsuid)
559 return 0;
560 if (dir->i_uid == fsuid)
561 return 0;
562 return !capable(CAP_FOWNER);
563}
564
565/* copy of may_delete in fs/namei.c()
566 * Check whether we can remove a link victim from directory dir, check
567 * whether the type of victim is right.
568 * 1. We can't do it if dir is read-only (done in permission())
569 * 2. We should have write and exec permissions on dir
570 * 3. We can't remove anything from append-only dir
571 * 4. We can't do anything with immutable dir (done in permission())
572 * 5. If the sticky bit on dir is set we should either
573 * a. be owner of dir, or
574 * b. be owner of victim, or
575 * c. have CAP_FOWNER capability
576 * 6. If the victim is append-only or immutable we can't do antyhing with
577 * links pointing to it.
578 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
579 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
580 * 9. We can't remove a root or mountpoint.
581 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
582 * nfs_async_unlink().
583 */
584
585static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
586{
587 int error;
588
589 if (!victim->d_inode)
590 return -ENOENT;
591
592 BUG_ON(victim->d_parent->d_inode != dir);
593 audit_inode_child(victim, dir);
594
595 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
596 if (error)
597 return error;
598 if (IS_APPEND(dir))
599 return -EPERM;
600 if (btrfs_check_sticky(dir, victim->d_inode)||
601 IS_APPEND(victim->d_inode)||
602 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
603 return -EPERM;
604 if (isdir) {
605 if (!S_ISDIR(victim->d_inode->i_mode))
606 return -ENOTDIR;
607 if (IS_ROOT(victim))
608 return -EBUSY;
609 } else if (S_ISDIR(victim->d_inode->i_mode))
610 return -EISDIR;
611 if (IS_DEADDIR(dir))
612 return -ENOENT;
613 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
614 return -EBUSY;
615 return 0;
616}
617
cb8e7090
CH		 618/* copy of may_create in fs/namei.c() */
619static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
620{
621 if (child->d_inode)
622 return -EEXIST;
623 if (IS_DEADDIR(dir))
624 return -ENOENT;
625 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
626}
627
628/*
629 * Create a new subvolume below @parent. This is largely modeled after
630 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
631 * inside this filesystem so it's quite a bit simpler.
632 */
76dda93c
YZ		 633static noinline int btrfs_mksubvol(struct path *parent,
634 char *name, int namelen,
72fd032e 635 struct btrfs_root *snap_src,
b83cc969 636 u64 *async_transid, bool readonly)
cb8e7090 637{
76dda93c 638 struct inode *dir = parent->dentry->d_inode;
cb8e7090
CH		 639 struct dentry *dentry;
640 int error;
641
76dda93c 642 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
cb8e7090
CH		 643
644 dentry = lookup_one_len(name, parent->dentry, namelen);
645 error = PTR_ERR(dentry);
646 if (IS_ERR(dentry))
647 goto out_unlock;
648
649 error = -EEXIST;
650 if (dentry->d_inode)
651 goto out_dput;
652
cb8e7090
CH		 653 error = mnt_want_write(parent->mnt);
654 if (error)
655 goto out_dput;
656
76dda93c 657 error = btrfs_may_create(dir, dentry);
cb8e7090
CH		 658 if (error)
659 goto out_drop_write;
660
76dda93c
YZ		 661 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
662
663 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
664 goto out_up_read;
665
3de4586c 666 if (snap_src) {
72fd032e 667 error = create_snapshot(snap_src, dentry,
b83cc969 668 name, namelen, async_transid, readonly);
3de4586c 669 } else {
76dda93c 670 error = create_subvol(BTRFS_I(dir)->root, dentry,
72fd032e 671 name, namelen, async_transid);
3de4586c 672 }
76dda93c
YZ		 673 if (!error)
674 fsnotify_mkdir(dir, dentry);
675out_up_read:
676 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
cb8e7090
CH		 677out_drop_write:
678 mnt_drop_write(parent->mnt);
679out_dput:
680 dput(dentry);
681out_unlock:
76dda93c 682 mutex_unlock(&dir->i_mutex);
cb8e7090
CH		 683 return error;
684}
685
4cb5300b
CM		 686/*
687 * When we're defragging a range, we don't want to kick it off again
688 * if it is really just waiting for delalloc to send it down.
689 * If we find a nice big extent or delalloc range for the bytes in the
690 * file you want to defrag, we return 0 to let you know to skip this
691 * part of the file
692 */
693static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
694{
695 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
696 struct extent_map *em = NULL;
697 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
698 u64 end;
699
700 read_lock(&em_tree->lock);
701 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
702 read_unlock(&em_tree->lock);
703
704 if (em) {
705 end = extent_map_end(em);
706 free_extent_map(em);
707 if (end - offset > thresh)
708 return 0;
709 }
710 /* if we already have a nice delalloc here, just stop */
711 thresh /= 2;
712 end = count_range_bits(io_tree, &offset, offset + thresh,
713 thresh, EXTENT_DELALLOC, 1);
714 if (end >= thresh)
715 return 0;
716 return 1;
717}
718
719/*
720 * helper function to walk through a file and find extents
721 * newer than a specific transid, and smaller than thresh.
722 *
723 * This is used by the defragging code to find new and small
724 * extents
725 */
726static int find_new_extents(struct btrfs_root *root,
727 struct inode *inode, u64 newer_than,
728 u64 *off, int thresh)
729{
730 struct btrfs_path *path;
731 struct btrfs_key min_key;
732 struct btrfs_key max_key;
733 struct extent_buffer *leaf;
734 struct btrfs_file_extent_item *extent;
735 int type;
736 int ret;
a4689d2b 737 u64 ino = btrfs_ino(inode);
4cb5300b
CM		 738
739 path = btrfs_alloc_path();
740 if (!path)
741 return -ENOMEM;
742
a4689d2b 743 min_key.objectid = ino;
4cb5300b
CM		 744 min_key.type = BTRFS_EXTENT_DATA_KEY;
745 min_key.offset = *off;
746
a4689d2b 747 max_key.objectid = ino;
4cb5300b
CM		 748 max_key.type = (u8)-1;
749 max_key.offset = (u64)-1;
750
751 path->keep_locks = 1;
752
753 while(1) {
754 ret = btrfs_search_forward(root, &min_key, &max_key,
755 path, 0, newer_than);
756 if (ret != 0)
757 goto none;
a4689d2b 758 if (min_key.objectid != ino)
4cb5300b
CM		 759 goto none;
760 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
761 goto none;
762
763 leaf = path->nodes[0];
764 extent = btrfs_item_ptr(leaf, path->slots[0],
765 struct btrfs_file_extent_item);
766
767 type = btrfs_file_extent_type(leaf, extent);
768 if (type == BTRFS_FILE_EXTENT_REG &&
769 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
770 check_defrag_in_cache(inode, min_key.offset, thresh)) {
771 *off = min_key.offset;
772 btrfs_free_path(path);
773 return 0;
774 }
775
776 if (min_key.offset == (u64)-1)
777 goto none;
778
779 min_key.offset++;
780 btrfs_release_path(path);
781 }
782none:
783 btrfs_free_path(path);
784 return -ENOENT;
785}
786
17ce6ef8
LB		 787/*
788 * Validaty check of prev em and next em:
789 * 1) no prev/next em
790 * 2) prev/next em is an hole/inline extent
791 */
792static int check_adjacent_extents(struct inode *inode, struct extent_map *em)
793{
794 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
795 struct extent_map *prev = NULL, *next = NULL;
796 int ret = 0;
797
798 read_lock(&em_tree->lock);
799 prev = lookup_extent_mapping(em_tree, em->start - 1, (u64)-1);
800 next = lookup_extent_mapping(em_tree, em->start + em->len, (u64)-1);
801 read_unlock(&em_tree->lock);
802
803 if ((!prev || prev->block_start >= EXTENT_MAP_LAST_BYTE) &&
804 (!next || next->block_start >= EXTENT_MAP_LAST_BYTE))
805 ret = 1;
806 free_extent_map(prev);
807 free_extent_map(next);
808
809 return ret;
810}
811
940100a4 812static int should_defrag_range(struct inode *inode, u64 start, u64 len,
1e701a32
CM		 813 int thresh, u64 *last_len, u64 *skip,
814 u64 *defrag_end)
940100a4
CM		 815{
816 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
817 struct extent_map *em = NULL;
818 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
819 int ret = 1;
820
821 /*
008873ea 822 * make sure that once we start defragging an extent, we keep on
940100a4
CM		 823 * defragging it
824 */
825 if (start < *defrag_end)
826 return 1;
827
828 *skip = 0;
829
830 /*
831 * hopefully we have this extent in the tree already, try without
832 * the full extent lock
833 */
834 read_lock(&em_tree->lock);
835 em = lookup_extent_mapping(em_tree, start, len);
836 read_unlock(&em_tree->lock);
837
838 if (!em) {
839 /* get the big lock and read metadata off disk */
d0082371 840 lock_extent(io_tree, start, start + len - 1);
940100a4 841 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
d0082371 842 unlock_extent(io_tree, start, start + len - 1);
940100a4 843
6cf8bfbf 844 if (IS_ERR(em))
940100a4
CM		 845 return 0;
846 }
847
848 /* this will cover holes, and inline extents */
17ce6ef8 849 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
940100a4 850 ret = 0;
17ce6ef8
LB		 851 goto out;
852 }
853
854 /* If we have nothing to merge with us, just skip. */
855 if (check_adjacent_extents(inode, em)) {
856 ret = 0;
857 goto out;
858 }
940100a4
CM		 859
860 /*
861 * we hit a real extent, if it is big don't bother defragging it again
862 */
1e701a32 863 if ((*last_len == 0 || *last_len >= thresh) && em->len >= thresh)
940100a4
CM		 864 ret = 0;
865
17ce6ef8 866out:
940100a4
CM		 867 /*
868 * last_len ends up being a counter of how many bytes we've defragged.
869 * every time we choose not to defrag an extent, we reset *last_len
870 * so that the next tiny extent will force a defrag.
871 *
872 * The end result of this is that tiny extents before a single big
873 * extent will force at least part of that big extent to be defragged.
874 */
875 if (ret) {
940100a4
CM		 876 *defrag_end = extent_map_end(em);
877 } else {
878 *last_len = 0;
879 *skip = extent_map_end(em);
880 *defrag_end = 0;
881 }
882
883 free_extent_map(em);
884 return ret;
885}
886
4cb5300b
CM		 887/*
888 * it doesn't do much good to defrag one or two pages
889 * at a time. This pulls in a nice chunk of pages
890 * to COW and defrag.
891 *
892 * It also makes sure the delalloc code has enough
893 * dirty data to avoid making new small extents as part
894 * of the defrag
895 *
896 * It's a good idea to start RA on this range
897 * before calling this.
898 */
899static int cluster_pages_for_defrag(struct inode *inode,
900 struct page **pages,
901 unsigned long start_index,
902 int num_pages)
f46b5a66 903{
4cb5300b
CM		 904 unsigned long file_end;
905 u64 isize = i_size_read(inode);
906 u64 page_start;
907 u64 page_end;
1f12bd06 908 u64 page_cnt;
4cb5300b
CM		 909 int ret;
910 int i;
911 int i_done;
3eaa2885 912 struct btrfs_ordered_extent *ordered;
4cb5300b 913 struct extent_state *cached_state = NULL;
600a45e1 914 struct extent_io_tree *tree;
3b16a4e3 915 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
4cb5300b 916
4cb5300b 917 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1f12bd06
LB		 918 if (!isize || start_index > file_end)
919 return 0;
920
921 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
4cb5300b
CM		 922
923 ret = btrfs_delalloc_reserve_space(inode,
1f12bd06 924 page_cnt << PAGE_CACHE_SHIFT);
4cb5300b
CM		 925 if (ret)
926 return ret;
4cb5300b 927 i_done = 0;
600a45e1 928 tree = &BTRFS_I(inode)->io_tree;
4cb5300b
CM		 929
930 /* step one, lock all the pages */
1f12bd06 931 for (i = 0; i < page_cnt; i++) {
4cb5300b 932 struct page *page;
600a45e1 933again:
a94733d0 934 page = find_or_create_page(inode->i_mapping,
600a45e1 935 start_index + i, mask);
4cb5300b
CM		 936 if (!page)
937 break;
938
600a45e1
MX		 939 page_start = page_offset(page);
940 page_end = page_start + PAGE_CACHE_SIZE - 1;
941 while (1) {
d0082371 942 lock_extent(tree, page_start, page_end);
600a45e1
MX		 943 ordered = btrfs_lookup_ordered_extent(inode,
944 page_start);
d0082371 945 unlock_extent(tree, page_start, page_end);
600a45e1
MX		 946 if (!ordered)
947 break;
948
949 unlock_page(page);
950 btrfs_start_ordered_extent(inode, ordered, 1);
951 btrfs_put_ordered_extent(ordered);
952 lock_page(page);
1f12bd06
LB		 953 /*
954 * we unlocked the page above, so we need check if
955 * it was released or not.
956 */
957 if (page->mapping != inode->i_mapping) {
958 unlock_page(page);
959 page_cache_release(page);
960 goto again;
961 }
600a45e1
MX		 962 }
963
4cb5300b
CM		 964 if (!PageUptodate(page)) {
965 btrfs_readpage(NULL, page);
966 lock_page(page);
967 if (!PageUptodate(page)) {
968 unlock_page(page);
969 page_cache_release(page);
970 ret = -EIO;
971 break;
972 }
973 }
600a45e1 974
600a45e1
MX		 975 if (page->mapping != inode->i_mapping) {
976 unlock_page(page);
977 page_cache_release(page);
978 goto again;
979 }
980
4cb5300b
CM		 981 pages[i] = page;
982 i_done++;
983 }
984 if (!i_done || ret)
985 goto out;
986
987 if (!(inode->i_sb->s_flags & MS_ACTIVE))
988 goto out;
989
990 /*
991 * so now we have a nice long stream of locked
992 * and up to date pages, lets wait on them
993 */
994 for (i = 0; i < i_done; i++)
995 wait_on_page_writeback(pages[i]);
996
997 page_start = page_offset(pages[0]);
998 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
999
1000 lock_extent_bits(&BTRFS_I(inode)->io_tree,
d0082371 1001 page_start, page_end - 1, 0, &cached_state);
4cb5300b
CM		 1002 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1003 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1004 EXTENT_DO_ACCOUNTING, 0, 0, &cached_state,
1005 GFP_NOFS);
1006
1f12bd06 1007 if (i_done != page_cnt) {
9e0baf60
JB		 1008 spin_lock(&BTRFS_I(inode)->lock);
1009 BTRFS_I(inode)->outstanding_extents++;
1010 spin_unlock(&BTRFS_I(inode)->lock);
4cb5300b 1011 btrfs_delalloc_release_space(inode,
1f12bd06 1012 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1013 }
1014
1015
1016 btrfs_set_extent_delalloc(inode, page_start, page_end - 1,
1017 &cached_state);
1018
1019 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1020 page_start, page_end - 1, &cached_state,
1021 GFP_NOFS);
1022
1023 for (i = 0; i < i_done; i++) {
1024 clear_page_dirty_for_io(pages[i]);
1025 ClearPageChecked(pages[i]);
1026 set_page_extent_mapped(pages[i]);
1027 set_page_dirty(pages[i]);
1028 unlock_page(pages[i]);
1029 page_cache_release(pages[i]);
1030 }
1031 return i_done;
1032out:
1033 for (i = 0; i < i_done; i++) {
1034 unlock_page(pages[i]);
1035 page_cache_release(pages[i]);
1036 }
1f12bd06 1037 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1038 return ret;
1039
1040}
1041
1042int btrfs_defrag_file(struct inode *inode, struct file *file,
1043 struct btrfs_ioctl_defrag_range_args *range,
1044 u64 newer_than, unsigned long max_to_defrag)
1045{
1046 struct btrfs_root *root = BTRFS_I(inode)->root;
1a419d85 1047 struct btrfs_super_block *disk_super;
4cb5300b 1048 struct file_ra_state *ra = NULL;
f46b5a66 1049 unsigned long last_index;
151a31b2 1050 u64 isize = i_size_read(inode);
1a419d85 1051 u64 features;
940100a4
CM		 1052 u64 last_len = 0;
1053 u64 skip = 0;
1054 u64 defrag_end = 0;
4cb5300b 1055 u64 newer_off = range->start;
f46b5a66 1056 unsigned long i;
008873ea 1057 unsigned long ra_index = 0;
f46b5a66 1058 int ret;
4cb5300b 1059 int defrag_count = 0;
1a419d85 1060 int compress_type = BTRFS_COMPRESS_ZLIB;
4cb5300b 1061 int extent_thresh = range->extent_thresh;
008873ea
LZ		 1062 int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1063 int cluster = max_cluster;
4cb5300b
CM		 1064 u64 new_align = ~((u64)128 * 1024 - 1);
1065 struct page **pages = NULL;
1066
1067 if (extent_thresh == 0)
1068 extent_thresh = 256 * 1024;
1a419d85
LZ		 1069
1070 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1071 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1072 return -EINVAL;
1073 if (range->compress_type)
1074 compress_type = range->compress_type;
1075 }
f46b5a66 1076
151a31b2 1077 if (isize == 0)
940100a4
CM		 1078 return 0;
1079
4cb5300b
CM		 1080 /*
1081 * if we were not given a file, allocate a readahead
1082 * context
1083 */
1084 if (!file) {
1085 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1086 if (!ra)
1087 return -ENOMEM;
1088 file_ra_state_init(ra, inode->i_mapping);
1089 } else {
1090 ra = &file->f_ra;
1091 }
1092
008873ea 1093 pages = kmalloc(sizeof(struct page *) * max_cluster,
4cb5300b
CM		 1094 GFP_NOFS);
1095 if (!pages) {
1096 ret = -ENOMEM;
1097 goto out_ra;
1098 }
1099
1100 /* find the last page to defrag */
1e701a32 1101 if (range->start + range->len > range->start) {
151a31b2 1102 last_index = min_t(u64, isize - 1,
1e701a32
CM		 1103 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1104 } else {
151a31b2 1105 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1e701a32
CM		 1106 }
1107
4cb5300b
CM		 1108 if (newer_than) {
1109 ret = find_new_extents(root, inode, newer_than,
1110 &newer_off, 64 * 1024);
1111 if (!ret) {
1112 range->start = newer_off;
1113 /*
1114 * we always align our defrag to help keep
1115 * the extents in the file evenly spaced
1116 */
1117 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
4cb5300b
CM		 1118 } else
1119 goto out_ra;
1120 } else {
1121 i = range->start >> PAGE_CACHE_SHIFT;
1122 }
1123 if (!max_to_defrag)
7ec31b54 1124 max_to_defrag = last_index + 1;
4cb5300b 1125
2a0f7f57
LZ		 1126 /*
1127 * make writeback starts from i, so the defrag range can be
1128 * written sequentially.
1129 */
1130 if (i < inode->i_mapping->writeback_index)
1131 inode->i_mapping->writeback_index = i;
1132
f7f43cc8
CM		 1133 while (i <= last_index && defrag_count < max_to_defrag &&
1134 (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1135 PAGE_CACHE_SHIFT)) {
4cb5300b
CM		 1136 /*
1137 * make sure we stop running if someone unmounts
1138 * the FS
1139 */
1140 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1141 break;
1142
66c26892
LB		 1143 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1144 PAGE_CACHE_SIZE, extent_thresh,
1145 &last_len, &skip, &defrag_end)) {
940100a4
CM		 1146 unsigned long next;
1147 /*
1148 * the should_defrag function tells us how much to skip
1149 * bump our counter by the suggested amount
1150 */
1151 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1152 i = max(i + 1, next);
1153 continue;
1154 }
008873ea
LZ		 1155
1156 if (!newer_than) {
1157 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1158 PAGE_CACHE_SHIFT) - i;
1159 cluster = min(cluster, max_cluster);
1160 } else {
1161 cluster = max_cluster;
1162 }
1163
1e701a32 1164 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1a419d85 1165 BTRFS_I(inode)->force_compress = compress_type;
940100a4 1166
008873ea
LZ		 1167 if (i + cluster > ra_index) {
1168 ra_index = max(i, ra_index);
1169 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1170 cluster);
1171 ra_index += max_cluster;
1172 }
940100a4 1173
ecb8bea8 1174 mutex_lock(&inode->i_mutex);
008873ea 1175 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
ecb8bea8
LB		 1176 if (ret < 0) {
1177 mutex_unlock(&inode->i_mutex);
4cb5300b 1178 goto out_ra;
ecb8bea8 1179 }
4cb5300b
CM		 1180
1181 defrag_count += ret;
1182 balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret);
ecb8bea8 1183 mutex_unlock(&inode->i_mutex);
4cb5300b
CM		 1184
1185 if (newer_than) {
1186 if (newer_off == (u64)-1)
1187 break;
1188
e1f041e1
LB		 1189 if (ret > 0)
1190 i += ret;
1191
4cb5300b
CM		 1192 newer_off = max(newer_off + 1,
1193 (u64)i << PAGE_CACHE_SHIFT);
1194
1195 ret = find_new_extents(root, inode,
1196 newer_than, &newer_off,
1197 64 * 1024);
1198 if (!ret) {
1199 range->start = newer_off;
1200 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
4cb5300b
CM		 1201 } else {
1202 break;
f46b5a66 1203 }
4cb5300b 1204 } else {
008873ea 1205 if (ret > 0) {
cbcc8326 1206 i += ret;
008873ea
LZ		 1207 last_len += ret << PAGE_CACHE_SHIFT;
1208 } else {
cbcc8326 1209 i++;
008873ea
LZ		 1210 last_len = 0;
1211 }
f46b5a66 1212 }
f46b5a66
CH		 1213 }
1214
1e701a32
CM		 1215 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1216 filemap_flush(inode->i_mapping);
1217
1218 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1219 /* the filemap_flush will queue IO into the worker threads, but
1220 * we have to make sure the IO is actually started and that
1221 * ordered extents get created before we return
1222 */
1223 atomic_inc(&root->fs_info->async_submit_draining);
1224 while (atomic_read(&root->fs_info->nr_async_submits) ||
1225 atomic_read(&root->fs_info->async_delalloc_pages)) {
1226 wait_event(root->fs_info->async_submit_wait,
1227 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1228 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1229 }
1230 atomic_dec(&root->fs_info->async_submit_draining);
1231
1232 mutex_lock(&inode->i_mutex);
261507a0 1233 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1e701a32
CM		 1234 mutex_unlock(&inode->i_mutex);
1235 }
1236
6c41761f 1237 disk_super = root->fs_info->super_copy;
1a419d85
LZ		 1238 features = btrfs_super_incompat_flags(disk_super);
1239 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1240 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
1241 btrfs_set_super_incompat_flags(disk_super, features);
1242 }
1243
60ccf82f 1244 ret = defrag_count;
940100a4 1245
4cb5300b
CM		 1246out_ra:
1247 if (!file)
1248 kfree(ra);
1249 kfree(pages);
940100a4 1250 return ret;
f46b5a66
CH		 1251}
1252
76dda93c
YZ		 1253static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
1254 void __user *arg)
f46b5a66
CH		 1255{
1256 u64 new_size;
1257 u64 old_size;
1258 u64 devid = 1;
1259 struct btrfs_ioctl_vol_args *vol_args;
1260 struct btrfs_trans_handle *trans;
1261 struct btrfs_device *device = NULL;
1262 char *sizestr;
1263 char *devstr = NULL;
1264 int ret = 0;
f46b5a66
CH		 1265 int mod = 0;
1266
c146afad
YZ		 1267 if (root->fs_info->sb->s_flags & MS_RDONLY)
1268 return -EROFS;
1269
e441d54d
CM		 1270 if (!capable(CAP_SYS_ADMIN))
1271 return -EPERM;
1272
c9e9f97b
ID		 1273 mutex_lock(&root->fs_info->volume_mutex);
1274 if (root->fs_info->balance_ctl) {
1275 printk(KERN_INFO "btrfs: balance in progress\n");
1276 ret = -EINVAL;
1277 goto out;
1278 }
1279
dae7b665 1280 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 1281 if (IS_ERR(vol_args)) {
1282 ret = PTR_ERR(vol_args);
1283 goto out;
1284 }
5516e595
MF		 1285
1286 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 1287
f46b5a66
CH		 1288 sizestr = vol_args->name;
1289 devstr = strchr(sizestr, ':');
1290 if (devstr) {
1291 char *end;
1292 sizestr = devstr + 1;
1293 *devstr = '\0';
1294 devstr = vol_args->name;
1295 devid = simple_strtoull(devstr, &end, 10);
5bb14682 1296 printk(KERN_INFO "btrfs: resizing devid %llu\n",
21380931 1297 (unsigned long long)devid);
f46b5a66 1298 }
2b82032c 1299 device = btrfs_find_device(root, devid, NULL, NULL);
f46b5a66 1300 if (!device) {
5bb14682 1301 printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
21380931 1302 (unsigned long long)devid);
f46b5a66 1303 ret = -EINVAL;
c9e9f97b 1304 goto out_free;
f46b5a66
CH		 1305 }
1306 if (!strcmp(sizestr, "max"))
1307 new_size = device->bdev->bd_inode->i_size;
1308 else {
1309 if (sizestr[0] == '-') {
1310 mod = -1;
1311 sizestr++;
1312 } else if (sizestr[0] == '+') {
1313 mod = 1;
1314 sizestr++;
1315 }
91748467 1316 new_size = memparse(sizestr, NULL);
f46b5a66
CH		 1317 if (new_size == 0) {
1318 ret = -EINVAL;
c9e9f97b 1319 goto out_free;
f46b5a66
CH		 1320 }
1321 }
1322
1323 old_size = device->total_bytes;
1324
1325 if (mod < 0) {
1326 if (new_size > old_size) {
1327 ret = -EINVAL;
c9e9f97b 1328 goto out_free;
f46b5a66
CH		 1329 }
1330 new_size = old_size - new_size;
1331 } else if (mod > 0) {
1332 new_size = old_size + new_size;
1333 }
1334
1335 if (new_size < 256 * 1024 * 1024) {
1336 ret = -EINVAL;
c9e9f97b 1337 goto out_free;
f46b5a66
CH		 1338 }
1339 if (new_size > device->bdev->bd_inode->i_size) {
1340 ret = -EFBIG;
c9e9f97b 1341 goto out_free;
f46b5a66
CH		 1342 }
1343
1344 do_div(new_size, root->sectorsize);
1345 new_size *= root->sectorsize;
1346
5bb14682 1347 printk(KERN_INFO "btrfs: new size for %s is %llu\n",
f46b5a66
CH		 1348 device->name, (unsigned long long)new_size);
1349
1350 if (new_size > old_size) {
a22285a6 1351 trans = btrfs_start_transaction(root, 0);
98d5dc13
TI		 1352 if (IS_ERR(trans)) {
1353 ret = PTR_ERR(trans);
c9e9f97b 1354 goto out_free;
98d5dc13 1355 }
f46b5a66
CH		 1356 ret = btrfs_grow_device(trans, device, new_size);
1357 btrfs_commit_transaction(trans, root);
ece7d20e 1358 } else if (new_size < old_size) {
f46b5a66
CH		 1359 ret = btrfs_shrink_device(device, new_size);
1360 }
1361
c9e9f97b 1362out_free:
f46b5a66 1363 kfree(vol_args);
c9e9f97b
ID		 1364out:
1365 mutex_unlock(&root->fs_info->volume_mutex);
f46b5a66
CH		 1366 return ret;
1367}
1368
72fd032e
SW		 1369static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1370 char *name,
1371 unsigned long fd,
1372 int subvol,
b83cc969
LZ		 1373 u64 *transid,
1374 bool readonly)
f46b5a66 1375{
cb8e7090 1376 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3de4586c 1377 struct file *src_file;
f46b5a66 1378 int namelen;
3de4586c 1379 int ret = 0;
f46b5a66 1380
c146afad
YZ		 1381 if (root->fs_info->sb->s_flags & MS_RDONLY)
1382 return -EROFS;
1383
72fd032e
SW		 1384 namelen = strlen(name);
1385 if (strchr(name, '/')) {
f46b5a66
CH		 1386 ret = -EINVAL;
1387 goto out;
1388 }
1389
16780cab
CM		 1390 if (name[0] == '.' &&
1391 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1392 ret = -EEXIST;
1393 goto out;
1394 }
1395
3de4586c 1396 if (subvol) {
72fd032e 1397 ret = btrfs_mksubvol(&file->f_path, name, namelen,
b83cc969 1398 NULL, transid, readonly);
cb8e7090 1399 } else {
3de4586c 1400 struct inode *src_inode;
72fd032e 1401 src_file = fget(fd);
3de4586c
CM		 1402 if (!src_file) {
1403 ret = -EINVAL;
1404 goto out;
1405 }
1406
1407 src_inode = src_file->f_path.dentry->d_inode;
1408 if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
d397712b
CM		 1409 printk(KERN_INFO "btrfs: Snapshot src from "
1410 "another FS\n");
3de4586c
CM		 1411 ret = -EINVAL;
1412 fput(src_file);
1413 goto out;
1414 }
72fd032e
SW		 1415 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1416 BTRFS_I(src_inode)->root,
b83cc969 1417 transid, readonly);
3de4586c 1418 fput(src_file);
cb8e7090 1419 }
f46b5a66 1420out:
72fd032e
SW		 1421 return ret;
1422}
1423
1424static noinline int btrfs_ioctl_snap_create(struct file *file,
fa0d2b9b 1425 void __user *arg, int subvol)
72fd032e 1426{
fa0d2b9b 1427 struct btrfs_ioctl_vol_args *vol_args;
72fd032e
SW		 1428 int ret;
1429
fa0d2b9b
LZ		 1430 vol_args = memdup_user(arg, sizeof(*vol_args));
1431 if (IS_ERR(vol_args))
1432 return PTR_ERR(vol_args);
1433 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
72fd032e 1434
fa0d2b9b 1435 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
b83cc969
LZ		 1436 vol_args->fd, subvol,
1437 NULL, false);
fdfb1e4f 1438
fa0d2b9b
LZ		 1439 kfree(vol_args);
1440 return ret;
1441}
fdfb1e4f 1442
fa0d2b9b
LZ		 1443static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1444 void __user *arg, int subvol)
1445{
1446 struct btrfs_ioctl_vol_args_v2 *vol_args;
1447 int ret;
1448 u64 transid = 0;
1449 u64 *ptr = NULL;
b83cc969 1450 bool readonly = false;
75eaa0e2 1451
fa0d2b9b
LZ		 1452 vol_args = memdup_user(arg, sizeof(*vol_args));
1453 if (IS_ERR(vol_args))
1454 return PTR_ERR(vol_args);
1455 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
75eaa0e2 1456
b83cc969
LZ		 1457 if (vol_args->flags &
1458 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY)) {
1459 ret = -EOPNOTSUPP;
fa0d2b9b 1460 goto out;
72fd032e 1461 }
fa0d2b9b
LZ		 1462
1463 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1464 ptr = &transid;
b83cc969
LZ		 1465 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1466 readonly = true;
fa0d2b9b
LZ		 1467
1468 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
b83cc969
LZ		 1469 vol_args->fd, subvol,
1470 ptr, readonly);
fa0d2b9b
LZ		 1471
1472 if (ret == 0 && ptr &&
1473 copy_to_user(arg +
1474 offsetof(struct btrfs_ioctl_vol_args_v2,
1475 transid), ptr, sizeof(*ptr)))
1476 ret = -EFAULT;
fdfb1e4f 1477out:
f46b5a66
CH		 1478 kfree(vol_args);
1479 return ret;
1480}
1481
0caa102d
LZ		 1482static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1483 void __user *arg)
1484{
1485 struct inode *inode = fdentry(file)->d_inode;
1486 struct btrfs_root *root = BTRFS_I(inode)->root;
1487 int ret = 0;
1488 u64 flags = 0;
1489
33345d01 1490 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
0caa102d
LZ		 1491 return -EINVAL;
1492
1493 down_read(&root->fs_info->subvol_sem);
1494 if (btrfs_root_readonly(root))
1495 flags |= BTRFS_SUBVOL_RDONLY;
1496 up_read(&root->fs_info->subvol_sem);
1497
1498 if (copy_to_user(arg, &flags, sizeof(flags)))
1499 ret = -EFAULT;
1500
1501 return ret;
1502}
1503
1504static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1505 void __user *arg)
1506{
1507 struct inode *inode = fdentry(file)->d_inode;
1508 struct btrfs_root *root = BTRFS_I(inode)->root;
1509 struct btrfs_trans_handle *trans;
1510 u64 root_flags;
1511 u64 flags;
1512 int ret = 0;
1513
1514 if (root->fs_info->sb->s_flags & MS_RDONLY)
1515 return -EROFS;
1516
33345d01 1517 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
0caa102d
LZ		 1518 return -EINVAL;
1519
1520 if (copy_from_user(&flags, arg, sizeof(flags)))
1521 return -EFAULT;
1522
b4dc2b8c 1523 if (flags & BTRFS_SUBVOL_CREATE_ASYNC)
0caa102d
LZ		 1524 return -EINVAL;
1525
1526 if (flags & ~BTRFS_SUBVOL_RDONLY)
1527 return -EOPNOTSUPP;
1528
2e149670 1529 if (!inode_owner_or_capable(inode))
b4dc2b8c
LZ		 1530 return -EACCES;
1531
0caa102d
LZ		 1532 down_write(&root->fs_info->subvol_sem);
1533
1534 /* nothing to do */
1535 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1536 goto out;
1537
1538 root_flags = btrfs_root_flags(&root->root_item);
1539 if (flags & BTRFS_SUBVOL_RDONLY)
1540 btrfs_set_root_flags(&root->root_item,
1541 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1542 else
1543 btrfs_set_root_flags(&root->root_item,
1544 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1545
1546 trans = btrfs_start_transaction(root, 1);
1547 if (IS_ERR(trans)) {
1548 ret = PTR_ERR(trans);
1549 goto out_reset;
1550 }
1551
b4dc2b8c 1552 ret = btrfs_update_root(trans, root->fs_info->tree_root,
0caa102d
LZ		 1553 &root->root_key, &root->root_item);
1554
1555 btrfs_commit_transaction(trans, root);
1556out_reset:
1557 if (ret)
1558 btrfs_set_root_flags(&root->root_item, root_flags);
1559out:
1560 up_write(&root->fs_info->subvol_sem);
1561 return ret;
1562}
1563
76dda93c
YZ		 1564/*
1565 * helper to check if the subvolume references other subvolumes
1566 */
1567static noinline int may_destroy_subvol(struct btrfs_root *root)
1568{
1569 struct btrfs_path *path;
1570 struct btrfs_key key;
1571 int ret;
1572
1573 path = btrfs_alloc_path();
1574 if (!path)
1575 return -ENOMEM;
1576
1577 key.objectid = root->root_key.objectid;
1578 key.type = BTRFS_ROOT_REF_KEY;
1579 key.offset = (u64)-1;
1580
1581 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1582 &key, path, 0, 0);
1583 if (ret < 0)
1584 goto out;
1585 BUG_ON(ret == 0);
1586
1587 ret = 0;
1588 if (path->slots[0] > 0) {
1589 path->slots[0]--;
1590 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1591 if (key.objectid == root->root_key.objectid &&
1592 key.type == BTRFS_ROOT_REF_KEY)
1593 ret = -ENOTEMPTY;
1594 }
1595out:
1596 btrfs_free_path(path);
1597 return ret;
1598}
1599
ac8e9819
CM		 1600static noinline int key_in_sk(struct btrfs_key *key,
1601 struct btrfs_ioctl_search_key *sk)
1602{
abc6e134
CM		 1603 struct btrfs_key test;
1604 int ret;
1605
1606 test.objectid = sk->min_objectid;
1607 test.type = sk->min_type;
1608 test.offset = sk->min_offset;
1609
1610 ret = btrfs_comp_cpu_keys(key, &test);
1611 if (ret < 0)
ac8e9819 1612 return 0;
abc6e134
CM		 1613
1614 test.objectid = sk->max_objectid;
1615 test.type = sk->max_type;
1616 test.offset = sk->max_offset;
1617
1618 ret = btrfs_comp_cpu_keys(key, &test);
1619 if (ret > 0)
ac8e9819
CM		 1620 return 0;
1621 return 1;
1622}
1623
1624static noinline int copy_to_sk(struct btrfs_root *root,
1625 struct btrfs_path *path,
1626 struct btrfs_key *key,
1627 struct btrfs_ioctl_search_key *sk,
1628 char *buf,
1629 unsigned long *sk_offset,
1630 int *num_found)
1631{
1632 u64 found_transid;
1633 struct extent_buffer *leaf;
1634 struct btrfs_ioctl_search_header sh;
1635 unsigned long item_off;
1636 unsigned long item_len;
1637 int nritems;
1638 int i;
1639 int slot;
ac8e9819
CM		 1640 int ret = 0;
1641
1642 leaf = path->nodes[0];
1643 slot = path->slots[0];
1644 nritems = btrfs_header_nritems(leaf);
1645
1646 if (btrfs_header_generation(leaf) > sk->max_transid) {
1647 i = nritems;
1648 goto advance_key;
1649 }
1650 found_transid = btrfs_header_generation(leaf);
1651
1652 for (i = slot; i < nritems; i++) {
1653 item_off = btrfs_item_ptr_offset(leaf, i);
1654 item_len = btrfs_item_size_nr(leaf, i);
1655
1656 if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1657 item_len = 0;
1658
1659 if (sizeof(sh) + item_len + *sk_offset >
1660 BTRFS_SEARCH_ARGS_BUFSIZE) {
1661 ret = 1;
1662 goto overflow;
1663 }
1664
1665 btrfs_item_key_to_cpu(leaf, key, i);
1666 if (!key_in_sk(key, sk))
1667 continue;
1668
1669 sh.objectid = key->objectid;
1670 sh.offset = key->offset;
1671 sh.type = key->type;
1672 sh.len = item_len;
1673 sh.transid = found_transid;
1674
1675 /* copy search result header */
1676 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1677 *sk_offset += sizeof(sh);
1678
1679 if (item_len) {
1680 char *p = buf + *sk_offset;
1681 /* copy the item */
1682 read_extent_buffer(leaf, p,
1683 item_off, item_len);
1684 *sk_offset += item_len;
ac8e9819 1685 }
e2156867 1686 (*num_found)++;
ac8e9819
CM		 1687
1688 if (*num_found >= sk->nr_items)
1689 break;
1690 }
1691advance_key:
abc6e134
CM		 1692 ret = 0;
1693 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
ac8e9819 1694 key->offset++;
abc6e134
CM		 1695 else if (key->type < (u8)-1 && key->type < sk->max_type) {
1696 key->offset = 0;
ac8e9819 1697 key->type++;
abc6e134
CM		 1698 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1699 key->offset = 0;
1700 key->type = 0;
ac8e9819 1701 key->objectid++;
abc6e134
CM		 1702 } else
1703 ret = 1;
ac8e9819 1704overflow:
ac8e9819
CM		 1705 return ret;
1706}
1707
1708static noinline int search_ioctl(struct inode *inode,
1709 struct btrfs_ioctl_search_args *args)
1710{
1711 struct btrfs_root *root;
1712 struct btrfs_key key;
1713 struct btrfs_key max_key;
1714 struct btrfs_path *path;
1715 struct btrfs_ioctl_search_key *sk = &args->key;
1716 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1717 int ret;
1718 int num_found = 0;
1719 unsigned long sk_offset = 0;
1720
1721 path = btrfs_alloc_path();
1722 if (!path)
1723 return -ENOMEM;
1724
1725 if (sk->tree_id == 0) {
1726 /* search the root of the inode that was passed */
1727 root = BTRFS_I(inode)->root;
1728 } else {
1729 key.objectid = sk->tree_id;
1730 key.type = BTRFS_ROOT_ITEM_KEY;
1731 key.offset = (u64)-1;
1732 root = btrfs_read_fs_root_no_name(info, &key);
1733 if (IS_ERR(root)) {
1734 printk(KERN_ERR "could not find root %llu\n",
1735 sk->tree_id);
1736 btrfs_free_path(path);
1737 return -ENOENT;
1738 }
1739 }
1740
1741 key.objectid = sk->min_objectid;
1742 key.type = sk->min_type;
1743 key.offset = sk->min_offset;
1744
1745 max_key.objectid = sk->max_objectid;
1746 max_key.type = sk->max_type;
1747 max_key.offset = sk->max_offset;
1748
1749 path->keep_locks = 1;
1750
1751 while(1) {
1752 ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1753 sk->min_transid);
1754 if (ret != 0) {
1755 if (ret > 0)
1756 ret = 0;
1757 goto err;
1758 }
1759 ret = copy_to_sk(root, path, &key, sk, args->buf,
1760 &sk_offset, &num_found);
b3b4aa74 1761 btrfs_release_path(path);
ac8e9819
CM		 1762 if (ret || num_found >= sk->nr_items)
1763 break;
1764
1765 }
1766 ret = 0;
1767err:
1768 sk->nr_items = num_found;
1769 btrfs_free_path(path);
1770 return ret;
1771}
1772
1773static noinline int btrfs_ioctl_tree_search(struct file *file,
1774 void __user *argp)
1775{
1776 struct btrfs_ioctl_search_args *args;
1777 struct inode *inode;
1778 int ret;
1779
1780 if (!capable(CAP_SYS_ADMIN))
1781 return -EPERM;
1782
2354d08f
JL		 1783 args = memdup_user(argp, sizeof(*args));
1784 if (IS_ERR(args))
1785 return PTR_ERR(args);
ac8e9819 1786
ac8e9819
CM		 1787 inode = fdentry(file)->d_inode;
1788 ret = search_ioctl(inode, args);
1789 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1790 ret = -EFAULT;
1791 kfree(args);
1792 return ret;
1793}
1794
98d377a0 1795/*
ac8e9819
CM		 1796 * Search INODE_REFs to identify path name of 'dirid' directory
1797 * in a 'tree_id' tree. and sets path name to 'name'.
1798 */
98d377a0
TH		 1799static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1800 u64 tree_id, u64 dirid, char *name)
1801{
1802 struct btrfs_root *root;
1803 struct btrfs_key key;
ac8e9819 1804 char *ptr;
98d377a0
TH		 1805 int ret = -1;
1806 int slot;
1807 int len;
1808 int total_len = 0;
1809 struct btrfs_inode_ref *iref;
1810 struct extent_buffer *l;
1811 struct btrfs_path *path;
1812
1813 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1814 name[0]='\0';
1815 return 0;
1816 }
1817
1818 path = btrfs_alloc_path();
1819 if (!path)
1820 return -ENOMEM;
1821
ac8e9819 1822 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
98d377a0
TH		 1823
1824 key.objectid = tree_id;
1825 key.type = BTRFS_ROOT_ITEM_KEY;
1826 key.offset = (u64)-1;
1827 root = btrfs_read_fs_root_no_name(info, &key);
1828 if (IS_ERR(root)) {
1829 printk(KERN_ERR "could not find root %llu\n", tree_id);
8ad6fcab
CM		 1830 ret = -ENOENT;
1831 goto out;
98d377a0
TH		 1832 }
1833
1834 key.objectid = dirid;
1835 key.type = BTRFS_INODE_REF_KEY;
8ad6fcab 1836 key.offset = (u64)-1;
98d377a0
TH		 1837
1838 while(1) {
1839 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1840 if (ret < 0)
1841 goto out;
1842
1843 l = path->nodes[0];
1844 slot = path->slots[0];
8ad6fcab
CM		 1845 if (ret > 0 && slot > 0)
1846 slot--;
98d377a0
TH		 1847 btrfs_item_key_to_cpu(l, &key, slot);
1848
1849 if (ret > 0 && (key.objectid != dirid ||
ac8e9819
CM		 1850 key.type != BTRFS_INODE_REF_KEY)) {
1851 ret = -ENOENT;
98d377a0 1852 goto out;
ac8e9819 1853 }
98d377a0
TH		 1854
1855 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1856 len = btrfs_inode_ref_name_len(l, iref);
1857 ptr -= len + 1;
1858 total_len += len + 1;
ac8e9819 1859 if (ptr < name)
98d377a0
TH		 1860 goto out;
1861
1862 *(ptr + len) = '/';
1863 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1864
1865 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1866 break;
1867
b3b4aa74 1868 btrfs_release_path(path);
98d377a0 1869 key.objectid = key.offset;
8ad6fcab 1870 key.offset = (u64)-1;
98d377a0 1871 dirid = key.objectid;
98d377a0 1872 }
ac8e9819 1873 if (ptr < name)
98d377a0 1874 goto out;
77906a50 1875 memmove(name, ptr, total_len);
98d377a0
TH		 1876 name[total_len]='\0';
1877 ret = 0;
1878out:
1879 btrfs_free_path(path);
ac8e9819
CM		 1880 return ret;
1881}
1882
1883static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1884 void __user *argp)
1885{
1886 struct btrfs_ioctl_ino_lookup_args *args;
1887 struct inode *inode;
1888 int ret;
1889
1890 if (!capable(CAP_SYS_ADMIN))
1891 return -EPERM;
1892
2354d08f
JL		 1893 args = memdup_user(argp, sizeof(*args));
1894 if (IS_ERR(args))
1895 return PTR_ERR(args);
c2b96929 1896
ac8e9819
CM		 1897 inode = fdentry(file)->d_inode;
1898
1b53ac4d
CM		 1899 if (args->treeid == 0)
1900 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1901
ac8e9819
CM		 1902 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1903 args->treeid, args->objectid,
1904 args->name);
1905
1906 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1907 ret = -EFAULT;
1908
1909 kfree(args);
98d377a0
TH		 1910 return ret;
1911}
1912
76dda93c
YZ		 1913static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1914 void __user *arg)
1915{
1916 struct dentry *parent = fdentry(file);
1917 struct dentry *dentry;
1918 struct inode *dir = parent->d_inode;
1919 struct inode *inode;
1920 struct btrfs_root *root = BTRFS_I(dir)->root;
1921 struct btrfs_root *dest = NULL;
1922 struct btrfs_ioctl_vol_args *vol_args;
1923 struct btrfs_trans_handle *trans;
1924 int namelen;
1925 int ret;
1926 int err = 0;
1927
76dda93c
YZ		 1928 vol_args = memdup_user(arg, sizeof(*vol_args));
1929 if (IS_ERR(vol_args))
1930 return PTR_ERR(vol_args);
1931
1932 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1933 namelen = strlen(vol_args->name);
1934 if (strchr(vol_args->name, '/') ||
1935 strncmp(vol_args->name, "..", namelen) == 0) {
1936 err = -EINVAL;
1937 goto out;
1938 }
1939
a561be71 1940 err = mnt_want_write_file(file);
76dda93c
YZ		 1941 if (err)
1942 goto out;
1943
1944 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
1945 dentry = lookup_one_len(vol_args->name, parent, namelen);
1946 if (IS_ERR(dentry)) {
1947 err = PTR_ERR(dentry);
1948 goto out_unlock_dir;
1949 }
1950
1951 if (!dentry->d_inode) {
1952 err = -ENOENT;
1953 goto out_dput;
1954 }
1955
1956 inode = dentry->d_inode;
4260f7c7
SW		 1957 dest = BTRFS_I(inode)->root;
1958 if (!capable(CAP_SYS_ADMIN)){
1959 /*
1960 * Regular user. Only allow this with a special mount
1961 * option, when the user has write+exec access to the
1962 * subvol root, and when rmdir(2) would have been
1963 * allowed.
1964 *
1965 * Note that this is _not_ check that the subvol is
1966 * empty or doesn't contain data that we wouldn't
1967 * otherwise be able to delete.
1968 *
1969 * Users who want to delete empty subvols should try
1970 * rmdir(2).
1971 */
1972 err = -EPERM;
1973 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1974 goto out_dput;
1975
1976 /*
1977 * Do not allow deletion if the parent dir is the same
1978 * as the dir to be deleted. That means the ioctl
1979 * must be called on the dentry referencing the root
1980 * of the subvol, not a random directory contained
1981 * within it.
1982 */
1983 err = -EINVAL;
1984 if (root == dest)
1985 goto out_dput;
1986
1987 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
1988 if (err)
1989 goto out_dput;
1990
1991 /* check if subvolume may be deleted by a non-root user */
1992 err = btrfs_may_delete(dir, dentry, 1);
1993 if (err)
1994 goto out_dput;
1995 }
1996
33345d01 1997 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
76dda93c
YZ		 1998 err = -EINVAL;
1999 goto out_dput;
2000 }
2001
76dda93c
YZ		 2002 mutex_lock(&inode->i_mutex);
2003 err = d_invalidate(dentry);
2004 if (err)
2005 goto out_unlock;
2006
2007 down_write(&root->fs_info->subvol_sem);
2008
2009 err = may_destroy_subvol(dest);
2010 if (err)
2011 goto out_up_write;
2012
a22285a6
YZ		 2013 trans = btrfs_start_transaction(root, 0);
2014 if (IS_ERR(trans)) {
2015 err = PTR_ERR(trans);
d327099a 2016 goto out_up_write;
a22285a6
YZ		 2017 }
2018 trans->block_rsv = &root->fs_info->global_block_rsv;
2019
76dda93c
YZ		 2020 ret = btrfs_unlink_subvol(trans, root, dir,
2021 dest->root_key.objectid,
2022 dentry->d_name.name,
2023 dentry->d_name.len);
79787eaa
JM		 2024 if (ret) {
2025 err = ret;
2026 btrfs_abort_transaction(trans, root, ret);
2027 goto out_end_trans;
2028 }
76dda93c
YZ		 2029
2030 btrfs_record_root_in_trans(trans, dest);
2031
2032 memset(&dest->root_item.drop_progress, 0,
2033 sizeof(dest->root_item.drop_progress));
2034 dest->root_item.drop_level = 0;
2035 btrfs_set_root_refs(&dest->root_item, 0);
2036
d68fc57b
YZ		 2037 if (!xchg(&dest->orphan_item_inserted, 1)) {
2038 ret = btrfs_insert_orphan_item(trans,
2039 root->fs_info->tree_root,
2040 dest->root_key.objectid);
79787eaa
JM		 2041 if (ret) {
2042 btrfs_abort_transaction(trans, root, ret);
2043 err = ret;
2044 goto out_end_trans;
2045 }
d68fc57b 2046 }
79787eaa 2047out_end_trans:
531cb13f 2048 ret = btrfs_end_transaction(trans, root);
79787eaa
JM		 2049 if (ret && !err)
2050 err = ret;
76dda93c
YZ		 2051 inode->i_flags |= S_DEAD;
2052out_up_write:
2053 up_write(&root->fs_info->subvol_sem);
2054out_unlock:
2055 mutex_unlock(&inode->i_mutex);
2056 if (!err) {
efefb143 2057 shrink_dcache_sb(root->fs_info->sb);
76dda93c
YZ		 2058 btrfs_invalidate_inodes(dest);
2059 d_delete(dentry);
2060 }
2061out_dput:
2062 dput(dentry);
2063out_unlock_dir:
2064 mutex_unlock(&dir->i_mutex);
2a79f17e 2065 mnt_drop_write_file(file);
76dda93c
YZ		 2066out:
2067 kfree(vol_args);
2068 return err;
2069}
2070
1e701a32 2071static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
f46b5a66
CH		 2072{
2073 struct inode *inode = fdentry(file)->d_inode;
2074 struct btrfs_root *root = BTRFS_I(inode)->root;
1e701a32 2075 struct btrfs_ioctl_defrag_range_args *range;
c146afad
YZ		 2076 int ret;
2077
b83cc969
LZ		 2078 if (btrfs_root_readonly(root))
2079 return -EROFS;
2080
a561be71 2081 ret = mnt_want_write_file(file);
c146afad
YZ		 2082 if (ret)
2083 return ret;
f46b5a66
CH		 2084
2085 switch (inode->i_mode & S_IFMT) {
2086 case S_IFDIR:
e441d54d
CM		 2087 if (!capable(CAP_SYS_ADMIN)) {
2088 ret = -EPERM;
2089 goto out;
2090 }
8929ecfa
YZ		 2091 ret = btrfs_defrag_root(root, 0);
2092 if (ret)
2093 goto out;
2094 ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
f46b5a66
CH		 2095 break;
2096 case S_IFREG:
e441d54d
CM		 2097 if (!(file->f_mode & FMODE_WRITE)) {
2098 ret = -EINVAL;
2099 goto out;
2100 }
1e701a32
CM		 2101
2102 range = kzalloc(sizeof(*range), GFP_KERNEL);
2103 if (!range) {
2104 ret = -ENOMEM;
2105 goto out;
2106 }
2107
2108 if (argp) {
2109 if (copy_from_user(range, argp,
2110 sizeof(*range))) {
2111 ret = -EFAULT;
2112 kfree(range);
683be16e 2113 goto out;
1e701a32
CM		 2114 }
2115 /* compression requires us to start the IO */
2116 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2117 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2118 range->extent_thresh = (u32)-1;
2119 }
2120 } else {
2121 /* the rest are all set to zero by kzalloc */
2122 range->len = (u64)-1;
2123 }
4cb5300b
CM		 2124 ret = btrfs_defrag_file(fdentry(file)->d_inode, file,
2125 range, 0, 0);
2126 if (ret > 0)
2127 ret = 0;
1e701a32 2128 kfree(range);
f46b5a66 2129 break;
8929ecfa
YZ		 2130 default:
2131 ret = -EINVAL;
f46b5a66 2132 }
e441d54d 2133out:
2a79f17e 2134 mnt_drop_write_file(file);
e441d54d 2135 return ret;
f46b5a66
CH		 2136}
2137
b2950863 2138static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
f46b5a66
CH		 2139{
2140 struct btrfs_ioctl_vol_args *vol_args;
2141 int ret;
2142
e441d54d
CM		 2143 if (!capable(CAP_SYS_ADMIN))
2144 return -EPERM;
2145
c9e9f97b
ID		 2146 mutex_lock(&root->fs_info->volume_mutex);
2147 if (root->fs_info->balance_ctl) {
2148 printk(KERN_INFO "btrfs: balance in progress\n");
2149 ret = -EINVAL;
2150 goto out;
2151 }
2152
dae7b665 2153 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 2154 if (IS_ERR(vol_args)) {
2155 ret = PTR_ERR(vol_args);
2156 goto out;
2157 }
f46b5a66 2158
5516e595 2159 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66
CH		 2160 ret = btrfs_init_new_device(root, vol_args->name);
2161
f46b5a66 2162 kfree(vol_args);
c9e9f97b
ID		 2163out:
2164 mutex_unlock(&root->fs_info->volume_mutex);
f46b5a66
CH		 2165 return ret;
2166}
2167
b2950863 2168static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
f46b5a66
CH		 2169{
2170 struct btrfs_ioctl_vol_args *vol_args;
2171 int ret;
2172
e441d54d
CM		 2173 if (!capable(CAP_SYS_ADMIN))
2174 return -EPERM;
2175
c146afad
YZ		 2176 if (root->fs_info->sb->s_flags & MS_RDONLY)
2177 return -EROFS;
2178
c9e9f97b
ID		 2179 mutex_lock(&root->fs_info->volume_mutex);
2180 if (root->fs_info->balance_ctl) {
2181 printk(KERN_INFO "btrfs: balance in progress\n");
2182 ret = -EINVAL;
2183 goto out;
2184 }
2185
dae7b665 2186 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 2187 if (IS_ERR(vol_args)) {
2188 ret = PTR_ERR(vol_args);
2189 goto out;
2190 }
f46b5a66 2191
5516e595 2192 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66
CH		 2193 ret = btrfs_rm_device(root, vol_args->name);
2194
f46b5a66 2195 kfree(vol_args);
c9e9f97b
ID		 2196out:
2197 mutex_unlock(&root->fs_info->volume_mutex);
f46b5a66
CH		 2198 return ret;
2199}
2200
475f6387
JS		 2201static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2202{
027ed2f0 2203 struct btrfs_ioctl_fs_info_args *fi_args;
475f6387
JS		 2204 struct btrfs_device *device;
2205 struct btrfs_device *next;
2206 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
027ed2f0 2207 int ret = 0;
475f6387
JS		 2208
2209 if (!capable(CAP_SYS_ADMIN))
2210 return -EPERM;
2211
027ed2f0
LZ		 2212 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2213 if (!fi_args)
2214 return -ENOMEM;
2215
2216 fi_args->num_devices = fs_devices->num_devices;
2217 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
475f6387
JS		 2218
2219 mutex_lock(&fs_devices->device_list_mutex);
2220 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
027ed2f0
LZ		 2221 if (device->devid > fi_args->max_id)
2222 fi_args->max_id = device->devid;
475f6387
JS		 2223 }
2224 mutex_unlock(&fs_devices->device_list_mutex);
2225
027ed2f0
LZ		 2226 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2227 ret = -EFAULT;
475f6387 2228
027ed2f0
LZ		 2229 kfree(fi_args);
2230 return ret;
475f6387
JS		 2231}
2232
2233static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2234{
2235 struct btrfs_ioctl_dev_info_args *di_args;
2236 struct btrfs_device *dev;
2237 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2238 int ret = 0;
2239 char *s_uuid = NULL;
2240 char empty_uuid[BTRFS_UUID_SIZE] = {0};
2241
2242 if (!capable(CAP_SYS_ADMIN))
2243 return -EPERM;
2244
2245 di_args = memdup_user(arg, sizeof(*di_args));
2246 if (IS_ERR(di_args))
2247 return PTR_ERR(di_args);
2248
2249 if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2250 s_uuid = di_args->uuid;
2251
2252 mutex_lock(&fs_devices->device_list_mutex);
2253 dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL);
2254 mutex_unlock(&fs_devices->device_list_mutex);
2255
2256 if (!dev) {
2257 ret = -ENODEV;
2258 goto out;
2259 }
2260
2261 di_args->devid = dev->devid;
2262 di_args->bytes_used = dev->bytes_used;
2263 di_args->total_bytes = dev->total_bytes;
2264 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2265 strncpy(di_args->path, dev->name, sizeof(di_args->path));
2266
2267out:
2268 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2269 ret = -EFAULT;
2270
2271 kfree(di_args);
2272 return ret;
2273}
2274
76dda93c
YZ		 2275static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2276 u64 off, u64 olen, u64 destoff)
f46b5a66
CH		 2277{
2278 struct inode *inode = fdentry(file)->d_inode;
2279 struct btrfs_root *root = BTRFS_I(inode)->root;
2280 struct file *src_file;
2281 struct inode *src;
2282 struct btrfs_trans_handle *trans;
f46b5a66 2283 struct btrfs_path *path;
f46b5a66 2284 struct extent_buffer *leaf;
ae01a0ab
YZ		 2285 char *buf;
2286 struct btrfs_key key;
f46b5a66
CH		 2287 u32 nritems;
2288 int slot;
ae01a0ab 2289 int ret;
c5c9cd4d
SW		 2290 u64 len = olen;
2291 u64 bs = root->fs_info->sb->s_blocksize;
2292 u64 hint_byte;
d20f7043 2293
c5c9cd4d
SW		 2294 /*
2295 * TODO:
2296 * - split compressed inline extents. annoying: we need to
2297 * decompress into destination's address_space (the file offset
2298 * may change, so source mapping won't do), then recompress (or
2299 * otherwise reinsert) a subrange.
2300 * - allow ranges within the same file to be cloned (provided
2301 * they don't overlap)?
2302 */
2303
e441d54d 2304 /* the destination must be opened for writing */
2ebc3464 2305 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
e441d54d
CM		 2306 return -EINVAL;
2307
b83cc969
LZ		 2308 if (btrfs_root_readonly(root))
2309 return -EROFS;
2310
a561be71 2311 ret = mnt_want_write_file(file);
c146afad
YZ		 2312 if (ret)
2313 return ret;
2314
c5c9cd4d 2315 src_file = fget(srcfd);
ab67b7c1
YZ		 2316 if (!src_file) {
2317 ret = -EBADF;
2318 goto out_drop_write;
2319 }
5dc64164 2320
f46b5a66
CH		 2321 src = src_file->f_dentry->d_inode;
2322
c5c9cd4d
SW		 2323 ret = -EINVAL;
2324 if (src == inode)
2325 goto out_fput;
2326
5dc64164
DR		 2327 /* the src must be open for reading */
2328 if (!(src_file->f_mode & FMODE_READ))
2329 goto out_fput;
2330
0e7b824c
LZ		 2331 /* don't make the dst file partly checksummed */
2332 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2333 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2334 goto out_fput;
2335
ae01a0ab
YZ		 2336 ret = -EISDIR;
2337 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2338 goto out_fput;
2339
f46b5a66 2340 ret = -EXDEV;
ae01a0ab
YZ		 2341 if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
2342 goto out_fput;
2343
2344 ret = -ENOMEM;
2345 buf = vmalloc(btrfs_level_size(root, 0));
2346 if (!buf)
2347 goto out_fput;
2348
2349 path = btrfs_alloc_path();
2350 if (!path) {
2351 vfree(buf);
f46b5a66 2352 goto out_fput;
ae01a0ab
YZ		 2353 }
2354 path->reada = 2;
f46b5a66
CH		 2355
2356 if (inode < src) {
fccdae43
SW		 2357 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2358 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
f46b5a66 2359 } else {
fccdae43
SW		 2360 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2361 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
f46b5a66
CH		 2362 }
2363
c5c9cd4d
SW		 2364 /* determine range to clone */
2365 ret = -EINVAL;
2ebc3464 2366 if (off + len > src->i_size || off + len < off)
f46b5a66 2367 goto out_unlock;
c5c9cd4d
SW		 2368 if (len == 0)
2369 olen = len = src->i_size - off;
2370 /* if we extend to eof, continue to block boundary */
2371 if (off + len == src->i_size)
2a6b8dae 2372 len = ALIGN(src->i_size, bs) - off;
c5c9cd4d
SW		 2373
2374 /* verify the end result is block aligned */
2a6b8dae
LZ		 2375 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2376 !IS_ALIGNED(destoff, bs))
c5c9cd4d
SW		 2377 goto out_unlock;
2378
d525e8ab
LZ		 2379 if (destoff > inode->i_size) {
2380 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2381 if (ret)
2382 goto out_unlock;
2383 }
2384
71ef0786
LZ		 2385 /* truncate page cache pages from target inode range */
2386 truncate_inode_pages_range(&inode->i_data, destoff,
2387 PAGE_CACHE_ALIGN(destoff + len) - 1);
2388
f46b5a66
CH		 2389 /* do any pending delalloc/csum calc on src, one way or
2390 another, and lock file content */
2391 while (1) {
31840ae1 2392 struct btrfs_ordered_extent *ordered;
d0082371 2393 lock_extent(&BTRFS_I(src)->io_tree, off, off+len);
9a019196
SW		 2394 ordered = btrfs_lookup_first_ordered_extent(src, off+len);
2395 if (!ordered &&
2396 !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len,
2397 EXTENT_DELALLOC, 0, NULL))
f46b5a66 2398 break;
d0082371 2399 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len);
ae01a0ab
YZ		 2400 if (ordered)
2401 btrfs_put_ordered_extent(ordered);
9a019196 2402 btrfs_wait_ordered_range(src, off, len);
f46b5a66
CH		 2403 }
2404
c5c9cd4d 2405 /* clone data */
33345d01 2406 key.objectid = btrfs_ino(src);
ae01a0ab
YZ		 2407 key.type = BTRFS_EXTENT_DATA_KEY;
2408 key.offset = 0;
f46b5a66
CH		 2409
2410 while (1) {
2411 /*
2412 * note the key will change type as we walk through the
2413 * tree.
2414 */
a22285a6 2415 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
f46b5a66
CH		 2416 if (ret < 0)
2417 goto out;
2418
ae01a0ab
YZ		 2419 nritems = btrfs_header_nritems(path->nodes[0]);
2420 if (path->slots[0] >= nritems) {
f46b5a66
CH		 2421 ret = btrfs_next_leaf(root, path);
2422 if (ret < 0)
2423 goto out;
2424 if (ret > 0)
2425 break;
ae01a0ab 2426 nritems = btrfs_header_nritems(path->nodes[0]);
f46b5a66
CH		 2427 }
2428 leaf = path->nodes[0];
2429 slot = path->slots[0];
f46b5a66 2430
ae01a0ab 2431 btrfs_item_key_to_cpu(leaf, &key, slot);
d20f7043 2432 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
33345d01 2433 key.objectid != btrfs_ino(src))
f46b5a66
CH		 2434 break;
2435
c5c9cd4d
SW		 2436 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2437 struct btrfs_file_extent_item *extent;
2438 int type;
31840ae1
ZY		 2439 u32 size;
2440 struct btrfs_key new_key;
c5c9cd4d
SW		 2441 u64 disko = 0, diskl = 0;
2442 u64 datao = 0, datal = 0;
2443 u8 comp;
b5384d48 2444 u64 endoff;
31840ae1
ZY		 2445
2446 size = btrfs_item_size_nr(leaf, slot);
2447 read_extent_buffer(leaf, buf,
2448 btrfs_item_ptr_offset(leaf, slot),
2449 size);
c5c9cd4d
SW		 2450
2451 extent = btrfs_item_ptr(leaf, slot,
2452 struct btrfs_file_extent_item);
2453 comp = btrfs_file_extent_compression(leaf, extent);
2454 type = btrfs_file_extent_type(leaf, extent);
c8a894d7
CM		 2455 if (type == BTRFS_FILE_EXTENT_REG ||
2456 type == BTRFS_FILE_EXTENT_PREALLOC) {
d397712b
CM		 2457 disko = btrfs_file_extent_disk_bytenr(leaf,
2458 extent);
2459 diskl = btrfs_file_extent_disk_num_bytes(leaf,
2460 extent);
c5c9cd4d 2461 datao = btrfs_file_extent_offset(leaf, extent);
d397712b
CM		 2462 datal = btrfs_file_extent_num_bytes(leaf,
2463 extent);
c5c9cd4d
SW		 2464 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2465 /* take upper bound, may be compressed */
2466 datal = btrfs_file_extent_ram_bytes(leaf,
2467 extent);
2468 }
b3b4aa74 2469 btrfs_release_path(path);
31840ae1 2470
050006a7 2471 if (key.offset + datal <= off ||
c5c9cd4d
SW		 2472 key.offset >= off+len)
2473 goto next;
2474
31840ae1 2475 memcpy(&new_key, &key, sizeof(new_key));
33345d01 2476 new_key.objectid = btrfs_ino(inode);
4d728ec7
LZ		 2477 if (off <= key.offset)
2478 new_key.offset = key.offset + destoff - off;
2479 else
2480 new_key.offset = destoff;
31840ae1 2481
b6f3409b
SW		 2482 /*
2483 * 1 - adjusting old extent (we may have to split it)
2484 * 1 - add new extent
2485 * 1 - inode update
2486 */
2487 trans = btrfs_start_transaction(root, 3);
a22285a6
YZ		 2488 if (IS_ERR(trans)) {
2489 ret = PTR_ERR(trans);
2490 goto out;
2491 }
2492
c8a894d7
CM		 2493 if (type == BTRFS_FILE_EXTENT_REG ||
2494 type == BTRFS_FILE_EXTENT_PREALLOC) {
d72c0842
LZ		 2495 /*
2496 * a | --- range to clone ---| b
2497 * | ------------- extent ------------- |
2498 */
2499
2500 /* substract range b */
2501 if (key.offset + datal > off + len)
2502 datal = off + len - key.offset;
2503
2504 /* substract range a */
a22285a6
YZ		 2505 if (off > key.offset) {
2506 datao += off - key.offset;
2507 datal -= off - key.offset;
2508 }
2509
a22285a6
YZ		 2510 ret = btrfs_drop_extents(trans, inode,
2511 new_key.offset,
2512 new_key.offset + datal,
2513 &hint_byte, 1);
79787eaa
JM		 2514 if (ret) {
2515 btrfs_abort_transaction(trans, root,
2516 ret);
2517 btrfs_end_transaction(trans, root);
2518 goto out;
2519 }
a22285a6 2520
c5c9cd4d
SW		 2521 ret = btrfs_insert_empty_item(trans, root, path,
2522 &new_key, size);
79787eaa
JM		 2523 if (ret) {
2524 btrfs_abort_transaction(trans, root,
2525 ret);
2526 btrfs_end_transaction(trans, root);
2527 goto out;
2528 }
c5c9cd4d
SW		 2529
2530 leaf = path->nodes[0];
2531 slot = path->slots[0];
2532 write_extent_buffer(leaf, buf,
31840ae1
ZY		 2533 btrfs_item_ptr_offset(leaf, slot),
2534 size);
ae01a0ab 2535
c5c9cd4d 2536 extent = btrfs_item_ptr(leaf, slot,
f46b5a66 2537 struct btrfs_file_extent_item);
c5c9cd4d 2538
c5c9cd4d
SW		 2539 /* disko == 0 means it's a hole */
2540 if (!disko)
2541 datao = 0;
c5c9cd4d
SW		 2542
2543 btrfs_set_file_extent_offset(leaf, extent,
2544 datao);
2545 btrfs_set_file_extent_num_bytes(leaf, extent,
2546 datal);
2547 if (disko) {
2548 inode_add_bytes(inode, datal);
ae01a0ab 2549 ret = btrfs_inc_extent_ref(trans, root,
5d4f98a2
YZ		 2550 disko, diskl, 0,
2551 root->root_key.objectid,
33345d01 2552 btrfs_ino(inode),
66d7e7f0
AJ		 2553 new_key.offset - datao,
2554 0);
79787eaa
JM		 2555 if (ret) {
2556 btrfs_abort_transaction(trans,
2557 root,
2558 ret);
2559 btrfs_end_transaction(trans,
2560 root);
2561 goto out;
2562
2563 }
f46b5a66 2564 }
c5c9cd4d
SW		 2565 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2566 u64 skip = 0;
2567 u64 trim = 0;
2568 if (off > key.offset) {
2569 skip = off - key.offset;
2570 new_key.offset += skip;
2571 }
d397712b 2572
c5c9cd4d
SW		 2573 if (key.offset + datal > off+len)
2574 trim = key.offset + datal - (off+len);
d397712b 2575
c5c9cd4d 2576 if (comp && (skip || trim)) {
c5c9cd4d 2577 ret = -EINVAL;
a22285a6 2578 btrfs_end_transaction(trans, root);
c5c9cd4d
SW		 2579 goto out;
2580 }
2581 size -= skip + trim;
2582 datal -= skip + trim;
a22285a6
YZ		 2583
2584 ret = btrfs_drop_extents(trans, inode,
2585 new_key.offset,
2586 new_key.offset + datal,
2587 &hint_byte, 1);
79787eaa
JM		 2588 if (ret) {
2589 btrfs_abort_transaction(trans, root,
2590 ret);
2591 btrfs_end_transaction(trans, root);
2592 goto out;
2593 }
a22285a6 2594
c5c9cd4d
SW		 2595 ret = btrfs_insert_empty_item(trans, root, path,
2596 &new_key, size);
79787eaa
JM		 2597 if (ret) {
2598 btrfs_abort_transaction(trans, root,
2599 ret);
2600 btrfs_end_transaction(trans, root);
2601 goto out;
2602 }
c5c9cd4d
SW		 2603
2604 if (skip) {
d397712b
CM		 2605 u32 start =
2606 btrfs_file_extent_calc_inline_size(0);
c5c9cd4d
SW		 2607 memmove(buf+start, buf+start+skip,
2608 datal);
2609 }
2610
2611 leaf = path->nodes[0];
2612 slot = path->slots[0];
2613 write_extent_buffer(leaf, buf,
2614 btrfs_item_ptr_offset(leaf, slot),
2615 size);
2616 inode_add_bytes(inode, datal);
f46b5a66 2617 }
c5c9cd4d
SW		 2618
2619 btrfs_mark_buffer_dirty(leaf);
b3b4aa74 2620 btrfs_release_path(path);
c5c9cd4d 2621
a22285a6 2622 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
b5384d48
SW		 2623
2624 /*
2625 * we round up to the block size at eof when
2626 * determining which extents to clone above,
2627 * but shouldn't round up the file size
2628 */
2629 endoff = new_key.offset + datal;
5f3888ff
LZ		 2630 if (endoff > destoff+olen)
2631 endoff = destoff+olen;
b5384d48
SW		 2632 if (endoff > inode->i_size)
2633 btrfs_i_size_write(inode, endoff);
2634
a22285a6 2635 ret = btrfs_update_inode(trans, root, inode);
79787eaa
JM		 2636 if (ret) {
2637 btrfs_abort_transaction(trans, root, ret);
2638 btrfs_end_transaction(trans, root);
2639 goto out;
2640 }
2641 ret = btrfs_end_transaction(trans, root);
a22285a6 2642 }
d397712b 2643next:
b3b4aa74 2644 btrfs_release_path(path);
f46b5a66 2645 key.offset++;
f46b5a66 2646 }
f46b5a66
CH		 2647 ret = 0;
2648out:
b3b4aa74 2649 btrfs_release_path(path);
d0082371 2650 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len);
f46b5a66
CH		 2651out_unlock:
2652 mutex_unlock(&src->i_mutex);
2653 mutex_unlock(&inode->i_mutex);
ae01a0ab
YZ		 2654 vfree(buf);
2655 btrfs_free_path(path);
f46b5a66
CH		 2656out_fput:
2657 fput(src_file);
ab67b7c1 2658out_drop_write:
2a79f17e 2659 mnt_drop_write_file(file);
f46b5a66
CH		 2660 return ret;
2661}
2662
7a865e8a 2663static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
c5c9cd4d
SW		 2664{
2665 struct btrfs_ioctl_clone_range_args args;
2666
7a865e8a 2667 if (copy_from_user(&args, argp, sizeof(args)))
c5c9cd4d
SW		 2668 return -EFAULT;
2669 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2670 args.src_length, args.dest_offset);
2671}
2672
f46b5a66
CH		 2673/*
2674 * there are many ways the trans_start and trans_end ioctls can lead
2675 * to deadlocks. They should only be used by applications that
2676 * basically own the machine, and have a very in depth understanding
2677 * of all the possible deadlocks and enospc problems.
2678 */
b2950863 2679static long btrfs_ioctl_trans_start(struct file *file)
f46b5a66
CH		 2680{
2681 struct inode *inode = fdentry(file)->d_inode;
2682 struct btrfs_root *root = BTRFS_I(inode)->root;
2683 struct btrfs_trans_handle *trans;
1ab86aed 2684 int ret;
f46b5a66 2685
1ab86aed 2686 ret = -EPERM;
df5b5520 2687 if (!capable(CAP_SYS_ADMIN))
1ab86aed 2688 goto out;
df5b5520 2689
1ab86aed
SW		 2690 ret = -EINPROGRESS;
2691 if (file->private_data)
f46b5a66 2692 goto out;
9ca9ee09 2693
b83cc969
LZ		 2694 ret = -EROFS;
2695 if (btrfs_root_readonly(root))
2696 goto out;
2697
a561be71 2698 ret = mnt_want_write_file(file);
c146afad
YZ		 2699 if (ret)
2700 goto out;
2701
a4abeea4 2702 atomic_inc(&root->fs_info->open_ioctl_trans);
9ca9ee09 2703
1ab86aed 2704 ret = -ENOMEM;
7a7eaa40 2705 trans = btrfs_start_ioctl_transaction(root);
abd30bb0 2706 if (IS_ERR(trans))
1ab86aed
SW		 2707 goto out_drop;
2708
2709 file->private_data = trans;
2710 return 0;
2711
2712out_drop:
a4abeea4 2713 atomic_dec(&root->fs_info->open_ioctl_trans);
2a79f17e 2714 mnt_drop_write_file(file);
f46b5a66 2715out:
f46b5a66
CH		 2716 return ret;
2717}
2718
6ef5ed0d
JB		 2719static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2720{
2721 struct inode *inode = fdentry(file)->d_inode;
2722 struct btrfs_root *root = BTRFS_I(inode)->root;
2723 struct btrfs_root *new_root;
2724 struct btrfs_dir_item *di;
2725 struct btrfs_trans_handle *trans;
2726 struct btrfs_path *path;
2727 struct btrfs_key location;
2728 struct btrfs_disk_key disk_key;
2729 struct btrfs_super_block *disk_super;
2730 u64 features;
2731 u64 objectid = 0;
2732 u64 dir_id;
2733
2734 if (!capable(CAP_SYS_ADMIN))
2735 return -EPERM;
2736
2737 if (copy_from_user(&objectid, argp, sizeof(objectid)))
2738 return -EFAULT;
2739
2740 if (!objectid)
2741 objectid = root->root_key.objectid;
2742
2743 location.objectid = objectid;
2744 location.type = BTRFS_ROOT_ITEM_KEY;
2745 location.offset = (u64)-1;
2746
2747 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2748 if (IS_ERR(new_root))
2749 return PTR_ERR(new_root);
2750
2751 if (btrfs_root_refs(&new_root->root_item) == 0)
2752 return -ENOENT;
2753
2754 path = btrfs_alloc_path();
2755 if (!path)
2756 return -ENOMEM;
2757 path->leave_spinning = 1;
2758
2759 trans = btrfs_start_transaction(root, 1);
98d5dc13 2760 if (IS_ERR(trans)) {
6ef5ed0d 2761 btrfs_free_path(path);
98d5dc13 2762 return PTR_ERR(trans);
6ef5ed0d
JB		 2763 }
2764
6c41761f 2765 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
6ef5ed0d
JB		 2766 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2767 dir_id, "default", 7, 1);
cf1e99a4 2768 if (IS_ERR_OR_NULL(di)) {
6ef5ed0d
JB		 2769 btrfs_free_path(path);
2770 btrfs_end_transaction(trans, root);
2771 printk(KERN_ERR "Umm, you don't have the default dir item, "
2772 "this isn't going to work\n");
2773 return -ENOENT;
2774 }
2775
2776 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2777 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2778 btrfs_mark_buffer_dirty(path->nodes[0]);
2779 btrfs_free_path(path);
2780
6c41761f 2781 disk_super = root->fs_info->super_copy;
6ef5ed0d
JB		 2782 features = btrfs_super_incompat_flags(disk_super);
2783 if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) {
2784 features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL;
2785 btrfs_set_super_incompat_flags(disk_super, features);
2786 }
2787 btrfs_end_transaction(trans, root);
2788
2789 return 0;
2790}
2791
bf5fc093
JB		 2792static void get_block_group_info(struct list_head *groups_list,
2793 struct btrfs_ioctl_space_info *space)
2794{
2795 struct btrfs_block_group_cache *block_group;
2796
2797 space->total_bytes = 0;
2798 space->used_bytes = 0;
2799 space->flags = 0;
2800 list_for_each_entry(block_group, groups_list, list) {
2801 space->flags = block_group->flags;
2802 space->total_bytes += block_group->key.offset;
2803 space->used_bytes +=
2804 btrfs_block_group_used(&block_group->item);
2805 }
2806}
2807
1406e432
JB		 2808long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2809{
2810 struct btrfs_ioctl_space_args space_args;
2811 struct btrfs_ioctl_space_info space;
2812 struct btrfs_ioctl_space_info *dest;
7fde62bf 2813 struct btrfs_ioctl_space_info *dest_orig;
13f2696f 2814 struct btrfs_ioctl_space_info __user *user_dest;
1406e432 2815 struct btrfs_space_info *info;
bf5fc093
JB		 2816 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2817 BTRFS_BLOCK_GROUP_SYSTEM,
2818 BTRFS_BLOCK_GROUP_METADATA,
2819 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2820 int num_types = 4;
7fde62bf 2821 int alloc_size;
1406e432 2822 int ret = 0;
51788b1b 2823 u64 slot_count = 0;
bf5fc093 2824 int i, c;
1406e432
JB		 2825
2826 if (copy_from_user(&space_args,
2827 (struct btrfs_ioctl_space_args __user *)arg,
2828 sizeof(space_args)))
2829 return -EFAULT;
2830
bf5fc093
JB		 2831 for (i = 0; i < num_types; i++) {
2832 struct btrfs_space_info *tmp;
2833
2834 info = NULL;
2835 rcu_read_lock();
2836 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2837 list) {
2838 if (tmp->flags == types[i]) {
2839 info = tmp;
2840 break;
2841 }
2842 }
2843 rcu_read_unlock();
2844
2845 if (!info)
2846 continue;
2847
2848 down_read(&info->groups_sem);
2849 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2850 if (!list_empty(&info->block_groups[c]))
2851 slot_count++;
2852 }
2853 up_read(&info->groups_sem);
2854 }
7fde62bf
CM		 2855
2856 /* space_slots == 0 means they are asking for a count */
2857 if (space_args.space_slots == 0) {
2858 space_args.total_spaces = slot_count;
2859 goto out;
2860 }
bf5fc093 2861
51788b1b 2862 slot_count = min_t(u64, space_args.space_slots, slot_count);
bf5fc093 2863
7fde62bf 2864 alloc_size = sizeof(*dest) * slot_count;
bf5fc093 2865
7fde62bf
CM		 2866 /* we generally have at most 6 or so space infos, one for each raid
2867 * level. So, a whole page should be more than enough for everyone
2868 */
2869 if (alloc_size > PAGE_CACHE_SIZE)
2870 return -ENOMEM;
2871
1406e432 2872 space_args.total_spaces = 0;
7fde62bf
CM		 2873 dest = kmalloc(alloc_size, GFP_NOFS);
2874 if (!dest)
2875 return -ENOMEM;
2876 dest_orig = dest;
1406e432 2877
7fde62bf 2878 /* now we have a buffer to copy into */
bf5fc093
JB		 2879 for (i = 0; i < num_types; i++) {
2880 struct btrfs_space_info *tmp;
2881
51788b1b
DR		 2882 if (!slot_count)
2883 break;
2884
bf5fc093
JB		 2885 info = NULL;
2886 rcu_read_lock();
2887 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2888 list) {
2889 if (tmp->flags == types[i]) {
2890 info = tmp;
2891 break;
2892 }
2893 }
2894 rcu_read_unlock();
7fde62bf 2895
bf5fc093
JB		 2896 if (!info)
2897 continue;
2898 down_read(&info->groups_sem);
2899 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2900 if (!list_empty(&info->block_groups[c])) {
2901 get_block_group_info(&info->block_groups[c],
2902 &space);
2903 memcpy(dest, &space, sizeof(space));
2904 dest++;
2905 space_args.total_spaces++;
51788b1b 2906 slot_count--;
bf5fc093 2907 }
51788b1b
DR		 2908 if (!slot_count)
2909 break;
bf5fc093
JB		 2910 }
2911 up_read(&info->groups_sem);
1406e432 2912 }
1406e432 2913
7fde62bf
CM		 2914 user_dest = (struct btrfs_ioctl_space_info *)
2915 (arg + sizeof(struct btrfs_ioctl_space_args));
2916
2917 if (copy_to_user(user_dest, dest_orig, alloc_size))
2918 ret = -EFAULT;
2919
2920 kfree(dest_orig);
2921out:
2922 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
1406e432
JB		 2923 ret = -EFAULT;
2924
2925 return ret;
2926}
2927
f46b5a66
CH		 2928/*
2929 * there are many ways the trans_start and trans_end ioctls can lead
2930 * to deadlocks. They should only be used by applications that
2931 * basically own the machine, and have a very in depth understanding
2932 * of all the possible deadlocks and enospc problems.
2933 */
2934long btrfs_ioctl_trans_end(struct file *file)
2935{
2936 struct inode *inode = fdentry(file)->d_inode;
2937 struct btrfs_root *root = BTRFS_I(inode)->root;
2938 struct btrfs_trans_handle *trans;
f46b5a66 2939
f46b5a66 2940 trans = file->private_data;
1ab86aed
SW		 2941 if (!trans)
2942 return -EINVAL;
b214107e 2943 file->private_data = NULL;
9ca9ee09 2944
1ab86aed
SW		 2945 btrfs_end_transaction(trans, root);
2946
a4abeea4 2947 atomic_dec(&root->fs_info->open_ioctl_trans);
9ca9ee09 2948
2a79f17e 2949 mnt_drop_write_file(file);
1ab86aed 2950 return 0;
f46b5a66
CH		 2951}
2952
46204592
SW		 2953static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
2954{
2955 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2956 struct btrfs_trans_handle *trans;
2957 u64 transid;
db5b493a 2958 int ret;
46204592
SW		 2959
2960 trans = btrfs_start_transaction(root, 0);
98d5dc13
TI		 2961 if (IS_ERR(trans))
2962 return PTR_ERR(trans);
46204592 2963 transid = trans->transid;
db5b493a 2964 ret = btrfs_commit_transaction_async(trans, root, 0);
8b2b2d3c
TI		 2965 if (ret) {
2966 btrfs_end_transaction(trans, root);
db5b493a 2967 return ret;
8b2b2d3c 2968 }
46204592
SW		 2969
2970 if (argp)
2971 if (copy_to_user(argp, &transid, sizeof(transid)))
2972 return -EFAULT;
2973 return 0;
2974}
2975
2976static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
2977{
2978 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2979 u64 transid;
2980
2981 if (argp) {
2982 if (copy_from_user(&transid, argp, sizeof(transid)))
2983 return -EFAULT;
2984 } else {
2985 transid = 0; /* current trans */
2986 }
2987 return btrfs_wait_for_commit(root, transid);
2988}
2989
475f6387
JS		 2990static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg)
2991{
2992 int ret;
2993 struct btrfs_ioctl_scrub_args *sa;
2994
2995 if (!capable(CAP_SYS_ADMIN))
2996 return -EPERM;
2997
2998 sa = memdup_user(arg, sizeof(*sa));
2999 if (IS_ERR(sa))
3000 return PTR_ERR(sa);
3001
3002 ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end,
8628764e 3003 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY);
475f6387
JS		 3004
3005 if (copy_to_user(arg, sa, sizeof(*sa)))
3006 ret = -EFAULT;
3007
3008 kfree(sa);
3009 return ret;
3010}
3011
3012static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3013{
3014 if (!capable(CAP_SYS_ADMIN))
3015 return -EPERM;
3016
3017 return btrfs_scrub_cancel(root);
3018}
3019
3020static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3021 void __user *arg)
3022{
3023 struct btrfs_ioctl_scrub_args *sa;
3024 int ret;
3025
3026 if (!capable(CAP_SYS_ADMIN))
3027 return -EPERM;
3028
3029 sa = memdup_user(arg, sizeof(*sa));
3030 if (IS_ERR(sa))
3031 return PTR_ERR(sa);
3032
3033 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3034
3035 if (copy_to_user(arg, sa, sizeof(*sa)))
3036 ret = -EFAULT;
3037
3038 kfree(sa);
3039 return ret;
3040}
3041
d7728c96
JS		 3042static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3043{
3044 int ret = 0;
3045 int i;
740c3d22 3046 u64 rel_ptr;
d7728c96 3047 int size;
806468f8 3048 struct btrfs_ioctl_ino_path_args *ipa = NULL;
d7728c96
JS		 3049 struct inode_fs_paths *ipath = NULL;
3050 struct btrfs_path *path;
3051
3052 if (!capable(CAP_SYS_ADMIN))
3053 return -EPERM;
3054
3055 path = btrfs_alloc_path();
3056 if (!path) {
3057 ret = -ENOMEM;
3058 goto out;
3059 }
3060
3061 ipa = memdup_user(arg, sizeof(*ipa));
3062 if (IS_ERR(ipa)) {
3063 ret = PTR_ERR(ipa);
3064 ipa = NULL;
3065 goto out;
3066 }
3067
3068 size = min_t(u32, ipa->size, 4096);
3069 ipath = init_ipath(size, root, path);
3070 if (IS_ERR(ipath)) {
3071 ret = PTR_ERR(ipath);
3072 ipath = NULL;
3073 goto out;
3074 }
3075
3076 ret = paths_from_inode(ipa->inum, ipath);
3077 if (ret < 0)
3078 goto out;
3079
3080 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
745c4d8e
JM		 3081 rel_ptr = ipath->fspath->val[i] -
3082 (u64)(unsigned long)ipath->fspath->val;
740c3d22 3083 ipath->fspath->val[i] = rel_ptr;
d7728c96
JS		 3084 }
3085
745c4d8e
JM		 3086 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3087 (void *)(unsigned long)ipath->fspath, size);
d7728c96
JS		 3088 if (ret) {
3089 ret = -EFAULT;
3090 goto out;
3091 }
3092
3093out:
3094 btrfs_free_path(path);
3095 free_ipath(ipath);
3096 kfree(ipa);
3097
3098 return ret;
3099}
3100
3101static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3102{
3103 struct btrfs_data_container *inodes = ctx;
3104 const size_t c = 3 * sizeof(u64);
3105
3106 if (inodes->bytes_left >= c) {
3107 inodes->bytes_left -= c;
3108 inodes->val[inodes->elem_cnt] = inum;
3109 inodes->val[inodes->elem_cnt + 1] = offset;
3110 inodes->val[inodes->elem_cnt + 2] = root;
3111 inodes->elem_cnt += 3;
3112 } else {
3113 inodes->bytes_missing += c - inodes->bytes_left;
3114 inodes->bytes_left = 0;
3115 inodes->elem_missed += 3;
3116 }
3117
3118 return 0;
3119}
3120
3121static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3122 void __user *arg)
3123{
3124 int ret = 0;
3125 int size;
4692cf58 3126 u64 extent_item_pos;
d7728c96
JS		 3127 struct btrfs_ioctl_logical_ino_args *loi;
3128 struct btrfs_data_container *inodes = NULL;
3129 struct btrfs_path *path = NULL;
3130 struct btrfs_key key;
3131
3132 if (!capable(CAP_SYS_ADMIN))
3133 return -EPERM;
3134
3135 loi = memdup_user(arg, sizeof(*loi));
3136 if (IS_ERR(loi)) {
3137 ret = PTR_ERR(loi);
3138 loi = NULL;
3139 goto out;
3140 }
3141
3142 path = btrfs_alloc_path();
3143 if (!path) {
3144 ret = -ENOMEM;
3145 goto out;
3146 }
3147
3148 size = min_t(u32, loi->size, 4096);
3149 inodes = init_data_container(size);
3150 if (IS_ERR(inodes)) {
3151 ret = PTR_ERR(inodes);
3152 inodes = NULL;
3153 goto out;
3154 }
3155
3156 ret = extent_from_logical(root->fs_info, loi->logical, path, &key);
4692cf58 3157 btrfs_release_path(path);
d7728c96
JS		 3158
3159 if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK)
3160 ret = -ENOENT;
3161 if (ret < 0)
3162 goto out;
3163
4692cf58 3164 extent_item_pos = loi->logical - key.objectid;
7a3ae2f8
JS		 3165 ret = iterate_extent_inodes(root->fs_info, key.objectid,
3166 extent_item_pos, 0, build_ino_list,
4692cf58 3167 inodes);
d7728c96
JS		 3168
3169 if (ret < 0)
3170 goto out;
3171
745c4d8e
JM		 3172 ret = copy_to_user((void *)(unsigned long)loi->inodes,
3173 (void *)(unsigned long)inodes, size);
d7728c96
JS		 3174 if (ret)
3175 ret = -EFAULT;
3176
3177out:
3178 btrfs_free_path(path);
3179 kfree(inodes);
3180 kfree(loi);
3181
3182 return ret;
3183}
3184
19a39dce 3185void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
c9e9f97b
ID		 3186 struct btrfs_ioctl_balance_args *bargs)
3187{
3188 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3189
3190 bargs->flags = bctl->flags;
3191
837d5b6e
ID		 3192 if (atomic_read(&fs_info->balance_running))
3193 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3194 if (atomic_read(&fs_info->balance_pause_req))
3195 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
a7e99c69
ID		 3196 if (atomic_read(&fs_info->balance_cancel_req))
3197 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
837d5b6e 3198
c9e9f97b
ID		 3199 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3200 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3201 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
19a39dce
ID		 3202
3203 if (lock) {
3204 spin_lock(&fs_info->balance_lock);
3205 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3206 spin_unlock(&fs_info->balance_lock);
3207 } else {
3208 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3209 }
c9e9f97b
ID		 3210}
3211
3212static long btrfs_ioctl_balance(struct btrfs_root *root, void __user *arg)
3213{
3214 struct btrfs_fs_info *fs_info = root->fs_info;
3215 struct btrfs_ioctl_balance_args *bargs;
3216 struct btrfs_balance_control *bctl;
3217 int ret;
3218
3219 if (!capable(CAP_SYS_ADMIN))
3220 return -EPERM;
3221
3222 if (fs_info->sb->s_flags & MS_RDONLY)
3223 return -EROFS;
3224
3225 mutex_lock(&fs_info->volume_mutex);
3226 mutex_lock(&fs_info->balance_mutex);
3227
3228 if (arg) {
3229 bargs = memdup_user(arg, sizeof(*bargs));
3230 if (IS_ERR(bargs)) {
3231 ret = PTR_ERR(bargs);
3232 goto out;
3233 }
de322263
ID		 3234
3235 if (bargs->flags & BTRFS_BALANCE_RESUME) {
3236 if (!fs_info->balance_ctl) {
3237 ret = -ENOTCONN;
3238 goto out_bargs;
3239 }
3240
3241 bctl = fs_info->balance_ctl;
3242 spin_lock(&fs_info->balance_lock);
3243 bctl->flags |= BTRFS_BALANCE_RESUME;
3244 spin_unlock(&fs_info->balance_lock);
3245
3246 goto do_balance;
3247 }
c9e9f97b
ID		 3248 } else {
3249 bargs = NULL;
3250 }
3251
837d5b6e
ID		 3252 if (fs_info->balance_ctl) {
3253 ret = -EINPROGRESS;
3254 goto out_bargs;
3255 }
3256
c9e9f97b
ID		 3257 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3258 if (!bctl) {
3259 ret = -ENOMEM;
3260 goto out_bargs;
3261 }
3262
3263 bctl->fs_info = fs_info;
3264 if (arg) {
3265 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3266 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3267 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3268
3269 bctl->flags = bargs->flags;
f43ffb60
ID		 3270 } else {
3271 /* balance everything - no filters */
3272 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
c9e9f97b
ID		 3273 }
3274
de322263 3275do_balance:
c9e9f97b
ID		 3276 ret = btrfs_balance(bctl, bargs);
3277 /*
837d5b6e
ID		 3278 * bctl is freed in __cancel_balance or in free_fs_info if
3279 * restriper was paused all the way until unmount
c9e9f97b
ID		 3280 */
3281 if (arg) {
3282 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3283 ret = -EFAULT;
3284 }
3285
3286out_bargs:
3287 kfree(bargs);
3288out:
3289 mutex_unlock(&fs_info->balance_mutex);
3290 mutex_unlock(&fs_info->volume_mutex);
3291 return ret;
3292}
3293
837d5b6e
ID		 3294static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
3295{
3296 if (!capable(CAP_SYS_ADMIN))
3297 return -EPERM;
3298
3299 switch (cmd) {
3300 case BTRFS_BALANCE_CTL_PAUSE:
3301 return btrfs_pause_balance(root->fs_info);
a7e99c69
ID		 3302 case BTRFS_BALANCE_CTL_CANCEL:
3303 return btrfs_cancel_balance(root->fs_info);
837d5b6e
ID		 3304 }
3305
3306 return -EINVAL;
3307}
3308
19a39dce
ID		 3309static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
3310 void __user *arg)
3311{
3312 struct btrfs_fs_info *fs_info = root->fs_info;
3313 struct btrfs_ioctl_balance_args *bargs;
3314 int ret = 0;
3315
3316 if (!capable(CAP_SYS_ADMIN))
3317 return -EPERM;
3318
3319 mutex_lock(&fs_info->balance_mutex);
3320 if (!fs_info->balance_ctl) {
3321 ret = -ENOTCONN;
3322 goto out;
3323 }
3324
3325 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
3326 if (!bargs) {
3327 ret = -ENOMEM;
3328 goto out;
3329 }
3330
3331 update_ioctl_balance_args(fs_info, 1, bargs);
3332
3333 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3334 ret = -EFAULT;
3335
3336 kfree(bargs);
3337out:
3338 mutex_unlock(&fs_info->balance_mutex);
3339 return ret;
3340}
3341
f46b5a66
CH		 3342long btrfs_ioctl(struct file *file, unsigned int
3343 cmd, unsigned long arg)
3344{
3345 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
4bcabaa3 3346 void __user *argp = (void __user *)arg;
f46b5a66
CH		 3347
3348 switch (cmd) {
6cbff00f
CH		 3349 case FS_IOC_GETFLAGS:
3350 return btrfs_ioctl_getflags(file, argp);
3351 case FS_IOC_SETFLAGS:
3352 return btrfs_ioctl_setflags(file, argp);
3353 case FS_IOC_GETVERSION:
3354 return btrfs_ioctl_getversion(file, argp);
f7039b1d
LD		 3355 case FITRIM:
3356 return btrfs_ioctl_fitrim(file, argp);
f46b5a66 3357 case BTRFS_IOC_SNAP_CREATE:
fa0d2b9b 3358 return btrfs_ioctl_snap_create(file, argp, 0);
fdfb1e4f 3359 case BTRFS_IOC_SNAP_CREATE_V2:
fa0d2b9b 3360 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3de4586c 3361 case BTRFS_IOC_SUBVOL_CREATE:
fa0d2b9b 3362 return btrfs_ioctl_snap_create(file, argp, 1);
76dda93c
YZ		 3363 case BTRFS_IOC_SNAP_DESTROY:
3364 return btrfs_ioctl_snap_destroy(file, argp);
0caa102d
LZ		 3365 case BTRFS_IOC_SUBVOL_GETFLAGS:
3366 return btrfs_ioctl_subvol_getflags(file, argp);
3367 case BTRFS_IOC_SUBVOL_SETFLAGS:
3368 return btrfs_ioctl_subvol_setflags(file, argp);
6ef5ed0d
JB		 3369 case BTRFS_IOC_DEFAULT_SUBVOL:
3370 return btrfs_ioctl_default_subvol(file, argp);
f46b5a66 3371 case BTRFS_IOC_DEFRAG:
1e701a32
CM		 3372 return btrfs_ioctl_defrag(file, NULL);
3373 case BTRFS_IOC_DEFRAG_RANGE:
3374 return btrfs_ioctl_defrag(file, argp);
f46b5a66 3375 case BTRFS_IOC_RESIZE:
4bcabaa3 3376 return btrfs_ioctl_resize(root, argp);
f46b5a66 3377 case BTRFS_IOC_ADD_DEV:
4bcabaa3 3378 return btrfs_ioctl_add_dev(root, argp);
f46b5a66 3379 case BTRFS_IOC_RM_DEV:
4bcabaa3 3380 return btrfs_ioctl_rm_dev(root, argp);
475f6387
JS		 3381 case BTRFS_IOC_FS_INFO:
3382 return btrfs_ioctl_fs_info(root, argp);
3383 case BTRFS_IOC_DEV_INFO:
3384 return btrfs_ioctl_dev_info(root, argp);
f46b5a66 3385 case BTRFS_IOC_BALANCE:
c9e9f97b 3386 return btrfs_ioctl_balance(root, NULL);
f46b5a66 3387 case BTRFS_IOC_CLONE:
c5c9cd4d
SW		 3388 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
3389 case BTRFS_IOC_CLONE_RANGE:
7a865e8a 3390 return btrfs_ioctl_clone_range(file, argp);
f46b5a66
CH		 3391 case BTRFS_IOC_TRANS_START:
3392 return btrfs_ioctl_trans_start(file);
3393 case BTRFS_IOC_TRANS_END:
3394 return btrfs_ioctl_trans_end(file);
ac8e9819
CM		 3395 case BTRFS_IOC_TREE_SEARCH:
3396 return btrfs_ioctl_tree_search(file, argp);
3397 case BTRFS_IOC_INO_LOOKUP:
3398 return btrfs_ioctl_ino_lookup(file, argp);
d7728c96
JS		 3399 case BTRFS_IOC_INO_PATHS:
3400 return btrfs_ioctl_ino_to_path(root, argp);
3401 case BTRFS_IOC_LOGICAL_INO:
3402 return btrfs_ioctl_logical_to_ino(root, argp);
1406e432
JB		 3403 case BTRFS_IOC_SPACE_INFO:
3404 return btrfs_ioctl_space_info(root, argp);
f46b5a66
CH		 3405 case BTRFS_IOC_SYNC:
3406 btrfs_sync_fs(file->f_dentry->d_sb, 1);
3407 return 0;
46204592
SW		 3408 case BTRFS_IOC_START_SYNC:
3409 return btrfs_ioctl_start_sync(file, argp);
3410 case BTRFS_IOC_WAIT_SYNC:
3411 return btrfs_ioctl_wait_sync(file, argp);
475f6387
JS		 3412 case BTRFS_IOC_SCRUB:
3413 return btrfs_ioctl_scrub(root, argp);
3414 case BTRFS_IOC_SCRUB_CANCEL:
3415 return btrfs_ioctl_scrub_cancel(root, argp);
3416 case BTRFS_IOC_SCRUB_PROGRESS:
3417 return btrfs_ioctl_scrub_progress(root, argp);
c9e9f97b
ID		 3418 case BTRFS_IOC_BALANCE_V2:
3419 return btrfs_ioctl_balance(root, argp);
837d5b6e
ID		 3420 case BTRFS_IOC_BALANCE_CTL:
3421 return btrfs_ioctl_balance_ctl(root, arg);
19a39dce
ID		 3422 case BTRFS_IOC_BALANCE_PROGRESS:
3423 return btrfs_ioctl_balance_progress(root, argp);
f46b5a66
CH		 3424 }
3425
3426 return -ENOTTY;
3427}
A mirror of Linus' kernel repository