]> git.ipfire.org Git - people/arne_f/kernel.git/blame - fs/btrfs/ioctl.c
projects / people / arne_f / kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Btrfs: avoid buffer overrun in btrfs_printk
[people/arne_f/kernel.git] / fs / btrfs / ioctl.c
CommitLineData
f46b5a66
CH		 1/*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
19#include <linux/kernel.h>
20#include <linux/bio.h>
21#include <linux/buffer_head.h>
22#include <linux/file.h>
23#include <linux/fs.h>
cb8e7090 24#include <linux/fsnotify.h>
f46b5a66
CH		 25#include <linux/pagemap.h>
26#include <linux/highmem.h>
27#include <linux/time.h>
28#include <linux/init.h>
29#include <linux/string.h>
f46b5a66 30#include <linux/backing-dev.h>
cb8e7090 31#include <linux/mount.h>
f46b5a66 32#include <linux/mpage.h>
cb8e7090 33#include <linux/namei.h>
f46b5a66
CH		 34#include <linux/swap.h>
35#include <linux/writeback.h>
36#include <linux/statfs.h>
37#include <linux/compat.h>
38#include <linux/bit_spinlock.h>
cb8e7090 39#include <linux/security.h>
f46b5a66 40#include <linux/xattr.h>
7ea394f1 41#include <linux/vmalloc.h>
5a0e3ad6 42#include <linux/slab.h>
f7039b1d 43#include <linux/blkdev.h>
4b4e25f2 44#include "compat.h"
f46b5a66
CH		 45#include "ctree.h"
46#include "disk-io.h"
47#include "transaction.h"
48#include "btrfs_inode.h"
49#include "ioctl.h"
50#include "print-tree.h"
51#include "volumes.h"
925baedd 52#include "locking.h"
581bb050 53#include "inode-map.h"
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 263 btrfs_update_iflags(inode);
0c4d2d95 264 inode_inc_iversion(inode);
306424cc 265 inode->i_ctime = CURRENT_TIME;
6cbff00f 266 ret = btrfs_update_inode(trans, root, inode);
6cbff00f 267
6cbff00f 268 btrfs_end_transaction(trans, root);
f062abf0
LZ		 269 out_drop:
270 if (ret) {
271 ip->flags = ip_oldflags;
272 inode->i_flags = i_oldflags;
273 }
6cbff00f 274
2a79f17e 275 mnt_drop_write_file(file);
6cbff00f
CH		 276 out_unlock:
277 mutex_unlock(&inode->i_mutex);
2d4e6f6a 278 return ret;
6cbff00f
CH		 279}
280
281static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
282{
283 struct inode *inode = file->f_path.dentry->d_inode;
284
285 return put_user(inode->i_generation, arg);
286}
f46b5a66 287
f7039b1d
LD		 288static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
289{
815745cf 290 struct btrfs_fs_info *fs_info = btrfs_sb(fdentry(file)->d_sb);
f7039b1d
LD		 291 struct btrfs_device *device;
292 struct request_queue *q;
293 struct fstrim_range range;
294 u64 minlen = ULLONG_MAX;
295 u64 num_devices = 0;
815745cf 296 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
f7039b1d
LD		 297 int ret;
298
299 if (!capable(CAP_SYS_ADMIN))
300 return -EPERM;
301
1f78160c
XG		 302 rcu_read_lock();
303 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
304 dev_list) {
f7039b1d
LD		 305 if (!device->bdev)
306 continue;
307 q = bdev_get_queue(device->bdev);
308 if (blk_queue_discard(q)) {
309 num_devices++;
310 minlen = min((u64)q->limits.discard_granularity,
311 minlen);
312 }
313 }
1f78160c 314 rcu_read_unlock();
f4c697e6 315
f7039b1d
LD		 316 if (!num_devices)
317 return -EOPNOTSUPP;
f7039b1d
LD		 318 if (copy_from_user(&range, arg, sizeof(range)))
319 return -EFAULT;
f4c697e6
LC		 320 if (range.start > total_bytes)
321 return -EINVAL;
f7039b1d 322
f4c697e6 323 range.len = min(range.len, total_bytes - range.start);
f7039b1d 324 range.minlen = max(range.minlen, minlen);
815745cf 325 ret = btrfs_trim_fs(fs_info->tree_root, &range);
f7039b1d
LD		 326 if (ret < 0)
327 return ret;
328
329 if (copy_to_user(arg, &range, sizeof(range)))
330 return -EFAULT;
331
332 return 0;
333}
334
cb8e7090
CH		 335static noinline int create_subvol(struct btrfs_root *root,
336 struct dentry *dentry,
72fd032e
SW		 337 char *name, int namelen,
338 u64 *async_transid)
f46b5a66
CH		 339{
340 struct btrfs_trans_handle *trans;
341 struct btrfs_key key;
342 struct btrfs_root_item root_item;
343 struct btrfs_inode_item *inode_item;
344 struct extent_buffer *leaf;
76dda93c 345 struct btrfs_root *new_root;
2fbe8c8a 346 struct dentry *parent = dentry->d_parent;
6a912213 347 struct inode *dir;
f46b5a66
CH		 348 int ret;
349 int err;
350 u64 objectid;
351 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
3de4586c 352 u64 index = 0;
f46b5a66 353
581bb050 354 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
2fbe8c8a 355 if (ret)
a22285a6 356 return ret;
6a912213
JB		 357
358 dir = parent->d_inode;
359
9ed74f2d
JB		 360 /*
361 * 1 - inode item
362 * 2 - refs
363 * 1 - root item
364 * 2 - dir items
365 */
a22285a6 366 trans = btrfs_start_transaction(root, 6);
2fbe8c8a 367 if (IS_ERR(trans))
a22285a6 368 return PTR_ERR(trans);
f46b5a66 369
5d4f98a2 370 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
66d7e7f0 371 0, objectid, NULL, 0, 0, 0, 0);
8e8a1e31
JB		 372 if (IS_ERR(leaf)) {
373 ret = PTR_ERR(leaf);
374 goto fail;
375 }
f46b5a66 376
5d4f98a2 377 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
f46b5a66
CH		 378 btrfs_set_header_bytenr(leaf, leaf->start);
379 btrfs_set_header_generation(leaf, trans->transid);
5d4f98a2 380 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
f46b5a66
CH		 381 btrfs_set_header_owner(leaf, objectid);
382
383 write_extent_buffer(leaf, root->fs_info->fsid,
384 (unsigned long)btrfs_header_fsid(leaf),
385 BTRFS_FSID_SIZE);
5d4f98a2
YZ		 386 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
387 (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
388 BTRFS_UUID_SIZE);
f46b5a66
CH		 389 btrfs_mark_buffer_dirty(leaf);
390
391 inode_item = &root_item.inode;
392 memset(inode_item, 0, sizeof(*inode_item));
393 inode_item->generation = cpu_to_le64(1);
394 inode_item->size = cpu_to_le64(3);
395 inode_item->nlink = cpu_to_le32(1);
a76a3cd4 396 inode_item->nbytes = cpu_to_le64(root->leafsize);
f46b5a66
CH		 397 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
398
08fe4db1
LZ		 399 root_item.flags = 0;
400 root_item.byte_limit = 0;
401 inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
402
f46b5a66 403 btrfs_set_root_bytenr(&root_item, leaf->start);
84234f3a 404 btrfs_set_root_generation(&root_item, trans->transid);
f46b5a66
CH		 405 btrfs_set_root_level(&root_item, 0);
406 btrfs_set_root_refs(&root_item, 1);
86b9f2ec 407 btrfs_set_root_used(&root_item, leaf->len);
80ff3856 408 btrfs_set_root_last_snapshot(&root_item, 0);
f46b5a66
CH		 409
410 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
411 root_item.drop_level = 0;
412
925baedd 413 btrfs_tree_unlock(leaf);
f46b5a66
CH		 414 free_extent_buffer(leaf);
415 leaf = NULL;
416
417 btrfs_set_root_dirid(&root_item, new_dirid);
418
419 key.objectid = objectid;
5d4f98a2 420 key.offset = 0;
f46b5a66
CH		 421 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
422 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
423 &root_item);
424 if (ret)
425 goto fail;
426
76dda93c
YZ		 427 key.offset = (u64)-1;
428 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
79787eaa
JM		 429 if (IS_ERR(new_root)) {
430 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
431 ret = PTR_ERR(new_root);
432 goto fail;
433 }
76dda93c
YZ		 434
435 btrfs_record_root_in_trans(trans, new_root);
436
d82a6f1d 437 ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
ce598979
MF		 438 if (ret) {
439 /* We potentially lose an unused inode item here */
79787eaa 440 btrfs_abort_transaction(trans, root, ret);
ce598979
MF		 441 goto fail;
442 }
443
f46b5a66
CH		 444 /*
445 * insert the directory item
446 */
3de4586c 447 ret = btrfs_set_inode_index(dir, &index);
79787eaa
JM		 448 if (ret) {
449 btrfs_abort_transaction(trans, root, ret);
450 goto fail;
451 }
3de4586c
CM		 452
453 ret = btrfs_insert_dir_item(trans, root,
16cdcec7 454 name, namelen, dir, &key,
3de4586c 455 BTRFS_FT_DIR, index);
79787eaa
JM		 456 if (ret) {
457 btrfs_abort_transaction(trans, root, ret);
f46b5a66 458 goto fail;
79787eaa 459 }
0660b5af 460
52c26179
YZ		 461 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
462 ret = btrfs_update_inode(trans, root, dir);
463 BUG_ON(ret);
464
0660b5af 465 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
4df27c4d 466 objectid, root->root_key.objectid,
33345d01 467 btrfs_ino(dir), index, name, namelen);
0660b5af 468
76dda93c 469 BUG_ON(ret);
f46b5a66 470
76dda93c 471 d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
f46b5a66 472fail:
72fd032e
SW		 473 if (async_transid) {
474 *async_transid = trans->transid;
475 err = btrfs_commit_transaction_async(trans, root, 1);
476 } else {
477 err = btrfs_commit_transaction(trans, root);
478 }
f46b5a66
CH		 479 if (err && !ret)
480 ret = err;
f46b5a66
CH		 481 return ret;
482}
483
72fd032e 484static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
b83cc969
LZ		 485 char *name, int namelen, u64 *async_transid,
486 bool readonly)
f46b5a66 487{
2e4bfab9 488 struct inode *inode;
f46b5a66
CH		 489 struct btrfs_pending_snapshot *pending_snapshot;
490 struct btrfs_trans_handle *trans;
2e4bfab9 491 int ret;
f46b5a66
CH		 492
493 if (!root->ref_cows)
494 return -EINVAL;
495
3de4586c 496 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
a22285a6
YZ		 497 if (!pending_snapshot)
498 return -ENOMEM;
499
500 btrfs_init_block_rsv(&pending_snapshot->block_rsv);
3de4586c 501 pending_snapshot->dentry = dentry;
f46b5a66 502 pending_snapshot->root = root;
b83cc969 503 pending_snapshot->readonly = readonly;
a22285a6
YZ		 504
505 trans = btrfs_start_transaction(root->fs_info->extent_root, 5);
506 if (IS_ERR(trans)) {
507 ret = PTR_ERR(trans);
508 goto fail;
509 }
510
511 ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
512 BUG_ON(ret);
513
8351583e 514 spin_lock(&root->fs_info->trans_lock);
f46b5a66
CH		 515 list_add(&pending_snapshot->list,
516 &trans->transaction->pending_snapshots);
8351583e 517 spin_unlock(&root->fs_info->trans_lock);
72fd032e
SW		 518 if (async_transid) {
519 *async_transid = trans->transid;
520 ret = btrfs_commit_transaction_async(trans,
521 root->fs_info->extent_root, 1);
522 } else {
523 ret = btrfs_commit_transaction(trans,
524 root->fs_info->extent_root);
525 }
2e4bfab9 526 BUG_ON(ret);
a22285a6
YZ		 527
528 ret = pending_snapshot->error;
529 if (ret)
530 goto fail;
531
66b4ffd1
JB		 532 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
533 if (ret)
534 goto fail;
f46b5a66 535
2fbe8c8a 536 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
2e4bfab9
YZ		 537 if (IS_ERR(inode)) {
538 ret = PTR_ERR(inode);
539 goto fail;
540 }
541 BUG_ON(!inode);
542 d_instantiate(dentry, inode);
543 ret = 0;
544fail:
a22285a6 545 kfree(pending_snapshot);
f46b5a66
CH		 546 return ret;
547}
548
4260f7c7
SW		 549/* copy of check_sticky in fs/namei.c()
550* It's inline, so penalty for filesystems that don't use sticky bit is
551* minimal.
552*/
553static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
554{
555 uid_t fsuid = current_fsuid();
556
557 if (!(dir->i_mode & S_ISVTX))
558 return 0;
559 if (inode->i_uid == fsuid)
560 return 0;
561 if (dir->i_uid == fsuid)
562 return 0;
563 return !capable(CAP_FOWNER);
564}
565
566/* copy of may_delete in fs/namei.c()
567 * Check whether we can remove a link victim from directory dir, check
568 * whether the type of victim is right.
569 * 1. We can't do it if dir is read-only (done in permission())
570 * 2. We should have write and exec permissions on dir
571 * 3. We can't remove anything from append-only dir
572 * 4. We can't do anything with immutable dir (done in permission())
573 * 5. If the sticky bit on dir is set we should either
574 * a. be owner of dir, or
575 * b. be owner of victim, or
576 * c. have CAP_FOWNER capability
577 * 6. If the victim is append-only or immutable we can't do antyhing with
578 * links pointing to it.
579 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
580 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
581 * 9. We can't remove a root or mountpoint.
582 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
583 * nfs_async_unlink().
584 */
585
586static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
587{
588 int error;
589
590 if (!victim->d_inode)
591 return -ENOENT;
592
593 BUG_ON(victim->d_parent->d_inode != dir);
594 audit_inode_child(victim, dir);
595
596 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
597 if (error)
598 return error;
599 if (IS_APPEND(dir))
600 return -EPERM;
601 if (btrfs_check_sticky(dir, victim->d_inode)||
602 IS_APPEND(victim->d_inode)||
603 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
604 return -EPERM;
605 if (isdir) {
606 if (!S_ISDIR(victim->d_inode->i_mode))
607 return -ENOTDIR;
608 if (IS_ROOT(victim))
609 return -EBUSY;
610 } else if (S_ISDIR(victim->d_inode->i_mode))
611 return -EISDIR;
612 if (IS_DEADDIR(dir))
613 return -ENOENT;
614 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
615 return -EBUSY;
616 return 0;
617}
618
cb8e7090
CH		 619/* copy of may_create in fs/namei.c() */
620static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
621{
622 if (child->d_inode)
623 return -EEXIST;
624 if (IS_DEADDIR(dir))
625 return -ENOENT;
626 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
627}
628
629/*
630 * Create a new subvolume below @parent. This is largely modeled after
631 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
632 * inside this filesystem so it's quite a bit simpler.
633 */
76dda93c
YZ		 634static noinline int btrfs_mksubvol(struct path *parent,
635 char *name, int namelen,
72fd032e 636 struct btrfs_root *snap_src,
b83cc969 637 u64 *async_transid, bool readonly)
cb8e7090 638{
76dda93c 639 struct inode *dir = parent->dentry->d_inode;
cb8e7090
CH		 640 struct dentry *dentry;
641 int error;
642
76dda93c 643 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
cb8e7090
CH		 644
645 dentry = lookup_one_len(name, parent->dentry, namelen);
646 error = PTR_ERR(dentry);
647 if (IS_ERR(dentry))
648 goto out_unlock;
649
650 error = -EEXIST;
651 if (dentry->d_inode)
652 goto out_dput;
653
cb8e7090
CH		 654 error = mnt_want_write(parent->mnt);
655 if (error)
656 goto out_dput;
657
76dda93c 658 error = btrfs_may_create(dir, dentry);
cb8e7090
CH		 659 if (error)
660 goto out_drop_write;
661
76dda93c
YZ		 662 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
663
664 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
665 goto out_up_read;
666
3de4586c 667 if (snap_src) {
72fd032e 668 error = create_snapshot(snap_src, dentry,
b83cc969 669 name, namelen, async_transid, readonly);
3de4586c 670 } else {
76dda93c 671 error = create_subvol(BTRFS_I(dir)->root, dentry,
72fd032e 672 name, namelen, async_transid);
3de4586c 673 }
76dda93c
YZ		 674 if (!error)
675 fsnotify_mkdir(dir, dentry);
676out_up_read:
677 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
cb8e7090
CH		 678out_drop_write:
679 mnt_drop_write(parent->mnt);
680out_dput:
681 dput(dentry);
682out_unlock:
76dda93c 683 mutex_unlock(&dir->i_mutex);
cb8e7090
CH		 684 return error;
685}
686
4cb5300b
CM		 687/*
688 * When we're defragging a range, we don't want to kick it off again
689 * if it is really just waiting for delalloc to send it down.
690 * If we find a nice big extent or delalloc range for the bytes in the
691 * file you want to defrag, we return 0 to let you know to skip this
692 * part of the file
693 */
694static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
695{
696 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
697 struct extent_map *em = NULL;
698 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
699 u64 end;
700
701 read_lock(&em_tree->lock);
702 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
703 read_unlock(&em_tree->lock);
704
705 if (em) {
706 end = extent_map_end(em);
707 free_extent_map(em);
708 if (end - offset > thresh)
709 return 0;
710 }
711 /* if we already have a nice delalloc here, just stop */
712 thresh /= 2;
713 end = count_range_bits(io_tree, &offset, offset + thresh,
714 thresh, EXTENT_DELALLOC, 1);
715 if (end >= thresh)
716 return 0;
717 return 1;
718}
719
720/*
721 * helper function to walk through a file and find extents
722 * newer than a specific transid, and smaller than thresh.
723 *
724 * This is used by the defragging code to find new and small
725 * extents
726 */
727static int find_new_extents(struct btrfs_root *root,
728 struct inode *inode, u64 newer_than,
729 u64 *off, int thresh)
730{
731 struct btrfs_path *path;
732 struct btrfs_key min_key;
733 struct btrfs_key max_key;
734 struct extent_buffer *leaf;
735 struct btrfs_file_extent_item *extent;
736 int type;
737 int ret;
a4689d2b 738 u64 ino = btrfs_ino(inode);
4cb5300b
CM		 739
740 path = btrfs_alloc_path();
741 if (!path)
742 return -ENOMEM;
743
a4689d2b 744 min_key.objectid = ino;
4cb5300b
CM		 745 min_key.type = BTRFS_EXTENT_DATA_KEY;
746 min_key.offset = *off;
747
a4689d2b 748 max_key.objectid = ino;
4cb5300b
CM		 749 max_key.type = (u8)-1;
750 max_key.offset = (u64)-1;
751
752 path->keep_locks = 1;
753
754 while(1) {
755 ret = btrfs_search_forward(root, &min_key, &max_key,
756 path, 0, newer_than);
757 if (ret != 0)
758 goto none;
a4689d2b 759 if (min_key.objectid != ino)
4cb5300b
CM		 760 goto none;
761 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
762 goto none;
763
764 leaf = path->nodes[0];
765 extent = btrfs_item_ptr(leaf, path->slots[0],
766 struct btrfs_file_extent_item);
767
768 type = btrfs_file_extent_type(leaf, extent);
769 if (type == BTRFS_FILE_EXTENT_REG &&
770 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
771 check_defrag_in_cache(inode, min_key.offset, thresh)) {
772 *off = min_key.offset;
773 btrfs_free_path(path);
774 return 0;
775 }
776
777 if (min_key.offset == (u64)-1)
778 goto none;
779
780 min_key.offset++;
781 btrfs_release_path(path);
782 }
783none:
784 btrfs_free_path(path);
785 return -ENOENT;
786}
787
17ce6ef8
LB		 788/*
789 * Validaty check of prev em and next em:
790 * 1) no prev/next em
791 * 2) prev/next em is an hole/inline extent
792 */
793static int check_adjacent_extents(struct inode *inode, struct extent_map *em)
794{
795 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
796 struct extent_map *prev = NULL, *next = NULL;
797 int ret = 0;
798
799 read_lock(&em_tree->lock);
800 prev = lookup_extent_mapping(em_tree, em->start - 1, (u64)-1);
801 next = lookup_extent_mapping(em_tree, em->start + em->len, (u64)-1);
802 read_unlock(&em_tree->lock);
803
804 if ((!prev || prev->block_start >= EXTENT_MAP_LAST_BYTE) &&
805 (!next || next->block_start >= EXTENT_MAP_LAST_BYTE))
806 ret = 1;
807 free_extent_map(prev);
808 free_extent_map(next);
809
810 return ret;
811}
812
940100a4 813static int should_defrag_range(struct inode *inode, u64 start, u64 len,
1e701a32
CM		 814 int thresh, u64 *last_len, u64 *skip,
815 u64 *defrag_end)
940100a4
CM		 816{
817 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
818 struct extent_map *em = NULL;
819 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
820 int ret = 1;
821
822 /*
008873ea 823 * make sure that once we start defragging an extent, we keep on
940100a4
CM		 824 * defragging it
825 */
826 if (start < *defrag_end)
827 return 1;
828
829 *skip = 0;
830
831 /*
832 * hopefully we have this extent in the tree already, try without
833 * the full extent lock
834 */
835 read_lock(&em_tree->lock);
836 em = lookup_extent_mapping(em_tree, start, len);
837 read_unlock(&em_tree->lock);
838
839 if (!em) {
840 /* get the big lock and read metadata off disk */
d0082371 841 lock_extent(io_tree, start, start + len - 1);
940100a4 842 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
d0082371 843 unlock_extent(io_tree, start, start + len - 1);
940100a4 844
6cf8bfbf 845 if (IS_ERR(em))
940100a4
CM		 846 return 0;
847 }
848
849 /* this will cover holes, and inline extents */
17ce6ef8 850 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
940100a4 851 ret = 0;
17ce6ef8
LB		 852 goto out;
853 }
854
855 /* If we have nothing to merge with us, just skip. */
856 if (check_adjacent_extents(inode, em)) {
857 ret = 0;
858 goto out;
859 }
940100a4
CM		 860
861 /*
862 * we hit a real extent, if it is big don't bother defragging it again
863 */
1e701a32 864 if ((*last_len == 0 || *last_len >= thresh) && em->len >= thresh)
940100a4
CM		 865 ret = 0;
866
17ce6ef8 867out:
940100a4
CM		 868 /*
869 * last_len ends up being a counter of how many bytes we've defragged.
870 * every time we choose not to defrag an extent, we reset *last_len
871 * so that the next tiny extent will force a defrag.
872 *
873 * The end result of this is that tiny extents before a single big
874 * extent will force at least part of that big extent to be defragged.
875 */
876 if (ret) {
940100a4
CM		 877 *defrag_end = extent_map_end(em);
878 } else {
879 *last_len = 0;
880 *skip = extent_map_end(em);
881 *defrag_end = 0;
882 }
883
884 free_extent_map(em);
885 return ret;
886}
887
4cb5300b
CM		 888/*
889 * it doesn't do much good to defrag one or two pages
890 * at a time. This pulls in a nice chunk of pages
891 * to COW and defrag.
892 *
893 * It also makes sure the delalloc code has enough
894 * dirty data to avoid making new small extents as part
895 * of the defrag
896 *
897 * It's a good idea to start RA on this range
898 * before calling this.
899 */
900static int cluster_pages_for_defrag(struct inode *inode,
901 struct page **pages,
902 unsigned long start_index,
903 int num_pages)
f46b5a66 904{
4cb5300b
CM		 905 unsigned long file_end;
906 u64 isize = i_size_read(inode);
907 u64 page_start;
908 u64 page_end;
1f12bd06 909 u64 page_cnt;
4cb5300b
CM		 910 int ret;
911 int i;
912 int i_done;
3eaa2885 913 struct btrfs_ordered_extent *ordered;
4cb5300b 914 struct extent_state *cached_state = NULL;
600a45e1 915 struct extent_io_tree *tree;
3b16a4e3 916 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
4cb5300b 917
4cb5300b 918 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1f12bd06
LB		 919 if (!isize || start_index > file_end)
920 return 0;
921
922 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
4cb5300b
CM		 923
924 ret = btrfs_delalloc_reserve_space(inode,
1f12bd06 925 page_cnt << PAGE_CACHE_SHIFT);
4cb5300b
CM		 926 if (ret)
927 return ret;
4cb5300b 928 i_done = 0;
600a45e1 929 tree = &BTRFS_I(inode)->io_tree;
4cb5300b
CM		 930
931 /* step one, lock all the pages */
1f12bd06 932 for (i = 0; i < page_cnt; i++) {
4cb5300b 933 struct page *page;
600a45e1 934again:
a94733d0 935 page = find_or_create_page(inode->i_mapping,
600a45e1 936 start_index + i, mask);
4cb5300b
CM		 937 if (!page)
938 break;
939
600a45e1
MX		 940 page_start = page_offset(page);
941 page_end = page_start + PAGE_CACHE_SIZE - 1;
942 while (1) {
d0082371 943 lock_extent(tree, page_start, page_end);
600a45e1
MX		 944 ordered = btrfs_lookup_ordered_extent(inode,
945 page_start);
d0082371 946 unlock_extent(tree, page_start, page_end);
600a45e1
MX		 947 if (!ordered)
948 break;
949
950 unlock_page(page);
951 btrfs_start_ordered_extent(inode, ordered, 1);
952 btrfs_put_ordered_extent(ordered);
953 lock_page(page);
1f12bd06
LB		 954 /*
955 * we unlocked the page above, so we need check if
956 * it was released or not.
957 */
958 if (page->mapping != inode->i_mapping) {
959 unlock_page(page);
960 page_cache_release(page);
961 goto again;
962 }
600a45e1
MX		 963 }
964
4cb5300b
CM		 965 if (!PageUptodate(page)) {
966 btrfs_readpage(NULL, page);
967 lock_page(page);
968 if (!PageUptodate(page)) {
969 unlock_page(page);
970 page_cache_release(page);
971 ret = -EIO;
972 break;
973 }
974 }
600a45e1 975
600a45e1
MX		 976 if (page->mapping != inode->i_mapping) {
977 unlock_page(page);
978 page_cache_release(page);
979 goto again;
980 }
981
4cb5300b
CM		 982 pages[i] = page;
983 i_done++;
984 }
985 if (!i_done || ret)
986 goto out;
987
988 if (!(inode->i_sb->s_flags & MS_ACTIVE))
989 goto out;
990
991 /*
992 * so now we have a nice long stream of locked
993 * and up to date pages, lets wait on them
994 */
995 for (i = 0; i < i_done; i++)
996 wait_on_page_writeback(pages[i]);
997
998 page_start = page_offset(pages[0]);
999 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1000
1001 lock_extent_bits(&BTRFS_I(inode)->io_tree,
d0082371 1002 page_start, page_end - 1, 0, &cached_state);
4cb5300b
CM		 1003 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1004 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1005 EXTENT_DO_ACCOUNTING, 0, 0, &cached_state,
1006 GFP_NOFS);
1007
1f12bd06 1008 if (i_done != page_cnt) {
9e0baf60
JB		 1009 spin_lock(&BTRFS_I(inode)->lock);
1010 BTRFS_I(inode)->outstanding_extents++;
1011 spin_unlock(&BTRFS_I(inode)->lock);
4cb5300b 1012 btrfs_delalloc_release_space(inode,
1f12bd06 1013 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1014 }
1015
1016
1017 btrfs_set_extent_delalloc(inode, page_start, page_end - 1,
1018 &cached_state);
1019
1020 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1021 page_start, page_end - 1, &cached_state,
1022 GFP_NOFS);
1023
1024 for (i = 0; i < i_done; i++) {
1025 clear_page_dirty_for_io(pages[i]);
1026 ClearPageChecked(pages[i]);
1027 set_page_extent_mapped(pages[i]);
1028 set_page_dirty(pages[i]);
1029 unlock_page(pages[i]);
1030 page_cache_release(pages[i]);
1031 }
1032 return i_done;
1033out:
1034 for (i = 0; i < i_done; i++) {
1035 unlock_page(pages[i]);
1036 page_cache_release(pages[i]);
1037 }
1f12bd06 1038 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1039 return ret;
1040
1041}
1042
1043int btrfs_defrag_file(struct inode *inode, struct file *file,
1044 struct btrfs_ioctl_defrag_range_args *range,
1045 u64 newer_than, unsigned long max_to_defrag)
1046{
1047 struct btrfs_root *root = BTRFS_I(inode)->root;
1a419d85 1048 struct btrfs_super_block *disk_super;
4cb5300b 1049 struct file_ra_state *ra = NULL;
f46b5a66 1050 unsigned long last_index;
151a31b2 1051 u64 isize = i_size_read(inode);
1a419d85 1052 u64 features;
940100a4
CM		 1053 u64 last_len = 0;
1054 u64 skip = 0;
1055 u64 defrag_end = 0;
4cb5300b 1056 u64 newer_off = range->start;
f46b5a66 1057 unsigned long i;
008873ea 1058 unsigned long ra_index = 0;
f46b5a66 1059 int ret;
4cb5300b 1060 int defrag_count = 0;
1a419d85 1061 int compress_type = BTRFS_COMPRESS_ZLIB;
4cb5300b 1062 int extent_thresh = range->extent_thresh;
008873ea
LZ		 1063 int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1064 int cluster = max_cluster;
4cb5300b
CM		 1065 u64 new_align = ~((u64)128 * 1024 - 1);
1066 struct page **pages = NULL;
1067
1068 if (extent_thresh == 0)
1069 extent_thresh = 256 * 1024;
1a419d85
LZ		 1070
1071 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1072 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1073 return -EINVAL;
1074 if (range->compress_type)
1075 compress_type = range->compress_type;
1076 }
f46b5a66 1077
151a31b2 1078 if (isize == 0)
940100a4
CM		 1079 return 0;
1080
4cb5300b
CM		 1081 /*
1082 * if we were not given a file, allocate a readahead
1083 * context
1084 */
1085 if (!file) {
1086 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1087 if (!ra)
1088 return -ENOMEM;
1089 file_ra_state_init(ra, inode->i_mapping);
1090 } else {
1091 ra = &file->f_ra;
1092 }
1093
008873ea 1094 pages = kmalloc(sizeof(struct page *) * max_cluster,
4cb5300b
CM		 1095 GFP_NOFS);
1096 if (!pages) {
1097 ret = -ENOMEM;
1098 goto out_ra;
1099 }
1100
1101 /* find the last page to defrag */
1e701a32 1102 if (range->start + range->len > range->start) {
151a31b2 1103 last_index = min_t(u64, isize - 1,
1e701a32
CM		 1104 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1105 } else {
151a31b2 1106 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1e701a32
CM		 1107 }
1108
4cb5300b
CM		 1109 if (newer_than) {
1110 ret = find_new_extents(root, inode, newer_than,
1111 &newer_off, 64 * 1024);
1112 if (!ret) {
1113 range->start = newer_off;
1114 /*
1115 * we always align our defrag to help keep
1116 * the extents in the file evenly spaced
1117 */
1118 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
4cb5300b
CM		 1119 } else
1120 goto out_ra;
1121 } else {
1122 i = range->start >> PAGE_CACHE_SHIFT;
1123 }
1124 if (!max_to_defrag)
7ec31b54 1125 max_to_defrag = last_index + 1;
4cb5300b 1126
2a0f7f57
LZ		 1127 /*
1128 * make writeback starts from i, so the defrag range can be
1129 * written sequentially.
1130 */
1131 if (i < inode->i_mapping->writeback_index)
1132 inode->i_mapping->writeback_index = i;
1133
f7f43cc8
CM		 1134 while (i <= last_index && defrag_count < max_to_defrag &&
1135 (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1136 PAGE_CACHE_SHIFT)) {
4cb5300b
CM		 1137 /*
1138 * make sure we stop running if someone unmounts
1139 * the FS
1140 */
1141 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1142 break;
1143
66c26892
LB		 1144 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1145 PAGE_CACHE_SIZE, extent_thresh,
1146 &last_len, &skip, &defrag_end)) {
940100a4
CM		 1147 unsigned long next;
1148 /*
1149 * the should_defrag function tells us how much to skip
1150 * bump our counter by the suggested amount
1151 */
1152 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1153 i = max(i + 1, next);
1154 continue;
1155 }
008873ea
LZ		 1156
1157 if (!newer_than) {
1158 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1159 PAGE_CACHE_SHIFT) - i;
1160 cluster = min(cluster, max_cluster);
1161 } else {
1162 cluster = max_cluster;
1163 }
1164
1e701a32 1165 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1a419d85 1166 BTRFS_I(inode)->force_compress = compress_type;
940100a4 1167
008873ea
LZ		 1168 if (i + cluster > ra_index) {
1169 ra_index = max(i, ra_index);
1170 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1171 cluster);
1172 ra_index += max_cluster;
1173 }
940100a4 1174
ecb8bea8 1175 mutex_lock(&inode->i_mutex);
008873ea 1176 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
ecb8bea8
LB		 1177 if (ret < 0) {
1178 mutex_unlock(&inode->i_mutex);
4cb5300b 1179 goto out_ra;
ecb8bea8 1180 }
4cb5300b
CM		 1181
1182 defrag_count += ret;
1183 balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret);
ecb8bea8 1184 mutex_unlock(&inode->i_mutex);
4cb5300b
CM		 1185
1186 if (newer_than) {
1187 if (newer_off == (u64)-1)
1188 break;
1189
e1f041e1
LB		 1190 if (ret > 0)
1191 i += ret;
1192
4cb5300b
CM		 1193 newer_off = max(newer_off + 1,
1194 (u64)i << PAGE_CACHE_SHIFT);
1195
1196 ret = find_new_extents(root, inode,
1197 newer_than, &newer_off,
1198 64 * 1024);
1199 if (!ret) {
1200 range->start = newer_off;
1201 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
4cb5300b
CM		 1202 } else {
1203 break;
f46b5a66 1204 }
4cb5300b 1205 } else {
008873ea 1206 if (ret > 0) {
cbcc8326 1207 i += ret;
008873ea
LZ		 1208 last_len += ret << PAGE_CACHE_SHIFT;
1209 } else {
cbcc8326 1210 i++;
008873ea
LZ		 1211 last_len = 0;
1212 }
f46b5a66 1213 }
f46b5a66
CH		 1214 }
1215
1e701a32
CM		 1216 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1217 filemap_flush(inode->i_mapping);
1218
1219 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1220 /* the filemap_flush will queue IO into the worker threads, but
1221 * we have to make sure the IO is actually started and that
1222 * ordered extents get created before we return
1223 */
1224 atomic_inc(&root->fs_info->async_submit_draining);
1225 while (atomic_read(&root->fs_info->nr_async_submits) ||
1226 atomic_read(&root->fs_info->async_delalloc_pages)) {
1227 wait_event(root->fs_info->async_submit_wait,
1228 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1229 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1230 }
1231 atomic_dec(&root->fs_info->async_submit_draining);
1232
1233 mutex_lock(&inode->i_mutex);
261507a0 1234 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1e701a32
CM		 1235 mutex_unlock(&inode->i_mutex);
1236 }
1237
6c41761f 1238 disk_super = root->fs_info->super_copy;
1a419d85
LZ		 1239 features = btrfs_super_incompat_flags(disk_super);
1240 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1241 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
1242 btrfs_set_super_incompat_flags(disk_super, features);
1243 }
1244
60ccf82f 1245 ret = defrag_count;
940100a4 1246
4cb5300b
CM		 1247out_ra:
1248 if (!file)
1249 kfree(ra);
1250 kfree(pages);
940100a4 1251 return ret;
f46b5a66
CH		 1252}
1253
76dda93c
YZ		 1254static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
1255 void __user *arg)
f46b5a66
CH		 1256{
1257 u64 new_size;
1258 u64 old_size;
1259 u64 devid = 1;
1260 struct btrfs_ioctl_vol_args *vol_args;
1261 struct btrfs_trans_handle *trans;
1262 struct btrfs_device *device = NULL;
1263 char *sizestr;
1264 char *devstr = NULL;
1265 int ret = 0;
f46b5a66
CH		 1266 int mod = 0;
1267
c146afad
YZ		 1268 if (root->fs_info->sb->s_flags & MS_RDONLY)
1269 return -EROFS;
1270
e441d54d
CM		 1271 if (!capable(CAP_SYS_ADMIN))
1272 return -EPERM;
1273
c9e9f97b
ID		 1274 mutex_lock(&root->fs_info->volume_mutex);
1275 if (root->fs_info->balance_ctl) {
1276 printk(KERN_INFO "btrfs: balance in progress\n");
1277 ret = -EINVAL;
1278 goto out;
1279 }
1280
dae7b665 1281 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 1282 if (IS_ERR(vol_args)) {
1283 ret = PTR_ERR(vol_args);
1284 goto out;
1285 }
5516e595
MF		 1286
1287 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 1288
f46b5a66
CH		 1289 sizestr = vol_args->name;
1290 devstr = strchr(sizestr, ':');
1291 if (devstr) {
1292 char *end;
1293 sizestr = devstr + 1;
1294 *devstr = '\0';
1295 devstr = vol_args->name;
1296 devid = simple_strtoull(devstr, &end, 10);
5bb14682 1297 printk(KERN_INFO "btrfs: resizing devid %llu\n",
21380931 1298 (unsigned long long)devid);
f46b5a66 1299 }
2b82032c 1300 device = btrfs_find_device(root, devid, NULL, NULL);
f46b5a66 1301 if (!device) {
5bb14682 1302 printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
21380931 1303 (unsigned long long)devid);
f46b5a66 1304 ret = -EINVAL;
c9e9f97b 1305 goto out_free;
f46b5a66
CH		 1306 }
1307 if (!strcmp(sizestr, "max"))
1308 new_size = device->bdev->bd_inode->i_size;
1309 else {
1310 if (sizestr[0] == '-') {
1311 mod = -1;
1312 sizestr++;
1313 } else if (sizestr[0] == '+') {
1314 mod = 1;
1315 sizestr++;
1316 }
91748467 1317 new_size = memparse(sizestr, NULL);
f46b5a66
CH		 1318 if (new_size == 0) {
1319 ret = -EINVAL;
c9e9f97b 1320 goto out_free;
f46b5a66
CH		 1321 }
1322 }
1323
1324 old_size = device->total_bytes;
1325
1326 if (mod < 0) {
1327 if (new_size > old_size) {
1328 ret = -EINVAL;
c9e9f97b 1329 goto out_free;
f46b5a66
CH		 1330 }
1331 new_size = old_size - new_size;
1332 } else if (mod > 0) {
1333 new_size = old_size + new_size;
1334 }
1335
1336 if (new_size < 256 * 1024 * 1024) {
1337 ret = -EINVAL;
c9e9f97b 1338 goto out_free;
f46b5a66
CH		 1339 }
1340 if (new_size > device->bdev->bd_inode->i_size) {
1341 ret = -EFBIG;
c9e9f97b 1342 goto out_free;
f46b5a66
CH		 1343 }
1344
1345 do_div(new_size, root->sectorsize);
1346 new_size *= root->sectorsize;
1347
5bb14682 1348 printk(KERN_INFO "btrfs: new size for %s is %llu\n",
f46b5a66
CH		 1349 device->name, (unsigned long long)new_size);
1350
1351 if (new_size > old_size) {
a22285a6 1352 trans = btrfs_start_transaction(root, 0);
98d5dc13
TI		 1353 if (IS_ERR(trans)) {
1354 ret = PTR_ERR(trans);
c9e9f97b 1355 goto out_free;
98d5dc13 1356 }
f46b5a66
CH		 1357 ret = btrfs_grow_device(trans, device, new_size);
1358 btrfs_commit_transaction(trans, root);
ece7d20e 1359 } else if (new_size < old_size) {
f46b5a66
CH		 1360 ret = btrfs_shrink_device(device, new_size);
1361 }
1362
c9e9f97b 1363out_free:
f46b5a66 1364 kfree(vol_args);
c9e9f97b
ID		 1365out:
1366 mutex_unlock(&root->fs_info->volume_mutex);
f46b5a66
CH		 1367 return ret;
1368}
1369
72fd032e
SW		 1370static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1371 char *name,
1372 unsigned long fd,
1373 int subvol,
b83cc969
LZ		 1374 u64 *transid,
1375 bool readonly)
f46b5a66 1376{
cb8e7090 1377 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3de4586c 1378 struct file *src_file;
f46b5a66 1379 int namelen;
3de4586c 1380 int ret = 0;
f46b5a66 1381
c146afad
YZ		 1382 if (root->fs_info->sb->s_flags & MS_RDONLY)
1383 return -EROFS;
1384
72fd032e
SW		 1385 namelen = strlen(name);
1386 if (strchr(name, '/')) {
f46b5a66
CH		 1387 ret = -EINVAL;
1388 goto out;
1389 }
1390
16780cab
CM		 1391 if (name[0] == '.' &&
1392 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1393 ret = -EEXIST;
1394 goto out;
1395 }
1396
3de4586c 1397 if (subvol) {
72fd032e 1398 ret = btrfs_mksubvol(&file->f_path, name, namelen,
b83cc969 1399 NULL, transid, readonly);
cb8e7090 1400 } else {
3de4586c 1401 struct inode *src_inode;
72fd032e 1402 src_file = fget(fd);
3de4586c
CM		 1403 if (!src_file) {
1404 ret = -EINVAL;
1405 goto out;
1406 }
1407
1408 src_inode = src_file->f_path.dentry->d_inode;
1409 if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
d397712b
CM		 1410 printk(KERN_INFO "btrfs: Snapshot src from "
1411 "another FS\n");
3de4586c
CM		 1412 ret = -EINVAL;
1413 fput(src_file);
1414 goto out;
1415 }
72fd032e
SW		 1416 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1417 BTRFS_I(src_inode)->root,
b83cc969 1418 transid, readonly);
3de4586c 1419 fput(src_file);
cb8e7090 1420 }
f46b5a66 1421out:
72fd032e
SW		 1422 return ret;
1423}
1424
1425static noinline int btrfs_ioctl_snap_create(struct file *file,
fa0d2b9b 1426 void __user *arg, int subvol)
72fd032e 1427{
fa0d2b9b 1428 struct btrfs_ioctl_vol_args *vol_args;
72fd032e
SW		 1429 int ret;
1430
fa0d2b9b
LZ		 1431 vol_args = memdup_user(arg, sizeof(*vol_args));
1432 if (IS_ERR(vol_args))
1433 return PTR_ERR(vol_args);
1434 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
72fd032e 1435
fa0d2b9b 1436 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
b83cc969
LZ		 1437 vol_args->fd, subvol,
1438 NULL, false);
fdfb1e4f 1439
fa0d2b9b
LZ		 1440 kfree(vol_args);
1441 return ret;
1442}
fdfb1e4f 1443
fa0d2b9b
LZ		 1444static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1445 void __user *arg, int subvol)
1446{
1447 struct btrfs_ioctl_vol_args_v2 *vol_args;
1448 int ret;
1449 u64 transid = 0;
1450 u64 *ptr = NULL;
b83cc969 1451 bool readonly = false;
75eaa0e2 1452
fa0d2b9b
LZ		 1453 vol_args = memdup_user(arg, sizeof(*vol_args));
1454 if (IS_ERR(vol_args))
1455 return PTR_ERR(vol_args);
1456 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
75eaa0e2 1457
b83cc969
LZ		 1458 if (vol_args->flags &
1459 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY)) {
1460 ret = -EOPNOTSUPP;
fa0d2b9b 1461 goto out;
72fd032e 1462 }
fa0d2b9b
LZ		 1463
1464 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1465 ptr = &transid;
b83cc969
LZ		 1466 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1467 readonly = true;
fa0d2b9b
LZ		 1468
1469 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
b83cc969
LZ		 1470 vol_args->fd, subvol,
1471 ptr, readonly);
fa0d2b9b
LZ		 1472
1473 if (ret == 0 && ptr &&
1474 copy_to_user(arg +
1475 offsetof(struct btrfs_ioctl_vol_args_v2,
1476 transid), ptr, sizeof(*ptr)))
1477 ret = -EFAULT;
fdfb1e4f 1478out:
f46b5a66
CH		 1479 kfree(vol_args);
1480 return ret;
1481}
1482
0caa102d
LZ		 1483static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1484 void __user *arg)
1485{
1486 struct inode *inode = fdentry(file)->d_inode;
1487 struct btrfs_root *root = BTRFS_I(inode)->root;
1488 int ret = 0;
1489 u64 flags = 0;
1490
33345d01 1491 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
0caa102d
LZ		 1492 return -EINVAL;
1493
1494 down_read(&root->fs_info->subvol_sem);
1495 if (btrfs_root_readonly(root))
1496 flags |= BTRFS_SUBVOL_RDONLY;
1497 up_read(&root->fs_info->subvol_sem);
1498
1499 if (copy_to_user(arg, &flags, sizeof(flags)))
1500 ret = -EFAULT;
1501
1502 return ret;
1503}
1504
1505static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1506 void __user *arg)
1507{
1508 struct inode *inode = fdentry(file)->d_inode;
1509 struct btrfs_root *root = BTRFS_I(inode)->root;
1510 struct btrfs_trans_handle *trans;
1511 u64 root_flags;
1512 u64 flags;
1513 int ret = 0;
1514
1515 if (root->fs_info->sb->s_flags & MS_RDONLY)
1516 return -EROFS;
1517
33345d01 1518 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
0caa102d
LZ		 1519 return -EINVAL;
1520
1521 if (copy_from_user(&flags, arg, sizeof(flags)))
1522 return -EFAULT;
1523
b4dc2b8c 1524 if (flags & BTRFS_SUBVOL_CREATE_ASYNC)
0caa102d
LZ		 1525 return -EINVAL;
1526
1527 if (flags & ~BTRFS_SUBVOL_RDONLY)
1528 return -EOPNOTSUPP;
1529
2e149670 1530 if (!inode_owner_or_capable(inode))
b4dc2b8c
LZ		 1531 return -EACCES;
1532
0caa102d
LZ		 1533 down_write(&root->fs_info->subvol_sem);
1534
1535 /* nothing to do */
1536 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1537 goto out;
1538
1539 root_flags = btrfs_root_flags(&root->root_item);
1540 if (flags & BTRFS_SUBVOL_RDONLY)
1541 btrfs_set_root_flags(&root->root_item,
1542 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1543 else
1544 btrfs_set_root_flags(&root->root_item,
1545 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1546
1547 trans = btrfs_start_transaction(root, 1);
1548 if (IS_ERR(trans)) {
1549 ret = PTR_ERR(trans);
1550 goto out_reset;
1551 }
1552
b4dc2b8c 1553 ret = btrfs_update_root(trans, root->fs_info->tree_root,
0caa102d
LZ		 1554 &root->root_key, &root->root_item);
1555
1556 btrfs_commit_transaction(trans, root);
1557out_reset:
1558 if (ret)
1559 btrfs_set_root_flags(&root->root_item, root_flags);
1560out:
1561 up_write(&root->fs_info->subvol_sem);
1562 return ret;
1563}
1564
76dda93c
YZ		 1565/*
1566 * helper to check if the subvolume references other subvolumes
1567 */
1568static noinline int may_destroy_subvol(struct btrfs_root *root)
1569{
1570 struct btrfs_path *path;
1571 struct btrfs_key key;
1572 int ret;
1573
1574 path = btrfs_alloc_path();
1575 if (!path)
1576 return -ENOMEM;
1577
1578 key.objectid = root->root_key.objectid;
1579 key.type = BTRFS_ROOT_REF_KEY;
1580 key.offset = (u64)-1;
1581
1582 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1583 &key, path, 0, 0);
1584 if (ret < 0)
1585 goto out;
1586 BUG_ON(ret == 0);
1587
1588 ret = 0;
1589 if (path->slots[0] > 0) {
1590 path->slots[0]--;
1591 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1592 if (key.objectid == root->root_key.objectid &&
1593 key.type == BTRFS_ROOT_REF_KEY)
1594 ret = -ENOTEMPTY;
1595 }
1596out:
1597 btrfs_free_path(path);
1598 return ret;
1599}
1600
ac8e9819
CM		 1601static noinline int key_in_sk(struct btrfs_key *key,
1602 struct btrfs_ioctl_search_key *sk)
1603{
abc6e134
CM		 1604 struct btrfs_key test;
1605 int ret;
1606
1607 test.objectid = sk->min_objectid;
1608 test.type = sk->min_type;
1609 test.offset = sk->min_offset;
1610
1611 ret = btrfs_comp_cpu_keys(key, &test);
1612 if (ret < 0)
ac8e9819 1613 return 0;
abc6e134
CM		 1614
1615 test.objectid = sk->max_objectid;
1616 test.type = sk->max_type;
1617 test.offset = sk->max_offset;
1618
1619 ret = btrfs_comp_cpu_keys(key, &test);
1620 if (ret > 0)
ac8e9819
CM		 1621 return 0;
1622 return 1;
1623}
1624
1625static noinline int copy_to_sk(struct btrfs_root *root,
1626 struct btrfs_path *path,
1627 struct btrfs_key *key,
1628 struct btrfs_ioctl_search_key *sk,
1629 char *buf,
1630 unsigned long *sk_offset,
1631 int *num_found)
1632{
1633 u64 found_transid;
1634 struct extent_buffer *leaf;
1635 struct btrfs_ioctl_search_header sh;
1636 unsigned long item_off;
1637 unsigned long item_len;
1638 int nritems;
1639 int i;
1640 int slot;
ac8e9819
CM		 1641 int ret = 0;
1642
1643 leaf = path->nodes[0];
1644 slot = path->slots[0];
1645 nritems = btrfs_header_nritems(leaf);
1646
1647 if (btrfs_header_generation(leaf) > sk->max_transid) {
1648 i = nritems;
1649 goto advance_key;
1650 }
1651 found_transid = btrfs_header_generation(leaf);
1652
1653 for (i = slot; i < nritems; i++) {
1654 item_off = btrfs_item_ptr_offset(leaf, i);
1655 item_len = btrfs_item_size_nr(leaf, i);
1656
1657 if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1658 item_len = 0;
1659
1660 if (sizeof(sh) + item_len + *sk_offset >
1661 BTRFS_SEARCH_ARGS_BUFSIZE) {
1662 ret = 1;
1663 goto overflow;
1664 }
1665
1666 btrfs_item_key_to_cpu(leaf, key, i);
1667 if (!key_in_sk(key, sk))
1668 continue;
1669
1670 sh.objectid = key->objectid;
1671 sh.offset = key->offset;
1672 sh.type = key->type;
1673 sh.len = item_len;
1674 sh.transid = found_transid;
1675
1676 /* copy search result header */
1677 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1678 *sk_offset += sizeof(sh);
1679
1680 if (item_len) {
1681 char *p = buf + *sk_offset;
1682 /* copy the item */
1683 read_extent_buffer(leaf, p,
1684 item_off, item_len);
1685 *sk_offset += item_len;
ac8e9819 1686 }
e2156867 1687 (*num_found)++;
ac8e9819
CM		 1688
1689 if (*num_found >= sk->nr_items)
1690 break;
1691 }
1692advance_key:
abc6e134
CM		 1693 ret = 0;
1694 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
ac8e9819 1695 key->offset++;
abc6e134
CM		 1696 else if (key->type < (u8)-1 && key->type < sk->max_type) {
1697 key->offset = 0;
ac8e9819 1698 key->type++;
abc6e134
CM		 1699 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1700 key->offset = 0;
1701 key->type = 0;
ac8e9819 1702 key->objectid++;
abc6e134
CM		 1703 } else
1704 ret = 1;
ac8e9819 1705overflow:
ac8e9819
CM		 1706 return ret;
1707}
1708
1709static noinline int search_ioctl(struct inode *inode,
1710 struct btrfs_ioctl_search_args *args)
1711{
1712 struct btrfs_root *root;
1713 struct btrfs_key key;
1714 struct btrfs_key max_key;
1715 struct btrfs_path *path;
1716 struct btrfs_ioctl_search_key *sk = &args->key;
1717 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1718 int ret;
1719 int num_found = 0;
1720 unsigned long sk_offset = 0;
1721
1722 path = btrfs_alloc_path();
1723 if (!path)
1724 return -ENOMEM;
1725
1726 if (sk->tree_id == 0) {
1727 /* search the root of the inode that was passed */
1728 root = BTRFS_I(inode)->root;
1729 } else {
1730 key.objectid = sk->tree_id;
1731 key.type = BTRFS_ROOT_ITEM_KEY;
1732 key.offset = (u64)-1;
1733 root = btrfs_read_fs_root_no_name(info, &key);
1734 if (IS_ERR(root)) {
1735 printk(KERN_ERR "could not find root %llu\n",
1736 sk->tree_id);
1737 btrfs_free_path(path);
1738 return -ENOENT;
1739 }
1740 }
1741
1742 key.objectid = sk->min_objectid;
1743 key.type = sk->min_type;
1744 key.offset = sk->min_offset;
1745
1746 max_key.objectid = sk->max_objectid;
1747 max_key.type = sk->max_type;
1748 max_key.offset = sk->max_offset;
1749
1750 path->keep_locks = 1;
1751
1752 while(1) {
1753 ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1754 sk->min_transid);
1755 if (ret != 0) {
1756 if (ret > 0)
1757 ret = 0;
1758 goto err;
1759 }
1760 ret = copy_to_sk(root, path, &key, sk, args->buf,
1761 &sk_offset, &num_found);
b3b4aa74 1762 btrfs_release_path(path);
ac8e9819
CM		 1763 if (ret || num_found >= sk->nr_items)
1764 break;
1765
1766 }
1767 ret = 0;
1768err:
1769 sk->nr_items = num_found;
1770 btrfs_free_path(path);
1771 return ret;
1772}
1773
1774static noinline int btrfs_ioctl_tree_search(struct file *file,
1775 void __user *argp)
1776{
1777 struct btrfs_ioctl_search_args *args;
1778 struct inode *inode;
1779 int ret;
1780
1781 if (!capable(CAP_SYS_ADMIN))
1782 return -EPERM;
1783
2354d08f
JL		 1784 args = memdup_user(argp, sizeof(*args));
1785 if (IS_ERR(args))
1786 return PTR_ERR(args);
ac8e9819 1787
ac8e9819
CM		 1788 inode = fdentry(file)->d_inode;
1789 ret = search_ioctl(inode, args);
1790 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1791 ret = -EFAULT;
1792 kfree(args);
1793 return ret;
1794}
1795
98d377a0 1796/*
ac8e9819
CM		 1797 * Search INODE_REFs to identify path name of 'dirid' directory
1798 * in a 'tree_id' tree. and sets path name to 'name'.
1799 */
98d377a0
TH		 1800static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1801 u64 tree_id, u64 dirid, char *name)
1802{
1803 struct btrfs_root *root;
1804 struct btrfs_key key;
ac8e9819 1805 char *ptr;
98d377a0
TH		 1806 int ret = -1;
1807 int slot;
1808 int len;
1809 int total_len = 0;
1810 struct btrfs_inode_ref *iref;
1811 struct extent_buffer *l;
1812 struct btrfs_path *path;
1813
1814 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1815 name[0]='\0';
1816 return 0;
1817 }
1818
1819 path = btrfs_alloc_path();
1820 if (!path)
1821 return -ENOMEM;
1822
ac8e9819 1823 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
98d377a0
TH		 1824
1825 key.objectid = tree_id;
1826 key.type = BTRFS_ROOT_ITEM_KEY;
1827 key.offset = (u64)-1;
1828 root = btrfs_read_fs_root_no_name(info, &key);
1829 if (IS_ERR(root)) {
1830 printk(KERN_ERR "could not find root %llu\n", tree_id);
8ad6fcab
CM		 1831 ret = -ENOENT;
1832 goto out;
98d377a0
TH		 1833 }
1834
1835 key.objectid = dirid;
1836 key.type = BTRFS_INODE_REF_KEY;
8ad6fcab 1837 key.offset = (u64)-1;
98d377a0
TH		 1838
1839 while(1) {
1840 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1841 if (ret < 0)
1842 goto out;
1843
1844 l = path->nodes[0];
1845 slot = path->slots[0];
8ad6fcab
CM		 1846 if (ret > 0 && slot > 0)
1847 slot--;
98d377a0
TH		 1848 btrfs_item_key_to_cpu(l, &key, slot);
1849
1850 if (ret > 0 && (key.objectid != dirid ||
ac8e9819
CM		 1851 key.type != BTRFS_INODE_REF_KEY)) {
1852 ret = -ENOENT;
98d377a0 1853 goto out;
ac8e9819 1854 }
98d377a0
TH		 1855
1856 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1857 len = btrfs_inode_ref_name_len(l, iref);
1858 ptr -= len + 1;
1859 total_len += len + 1;
ac8e9819 1860 if (ptr < name)
98d377a0
TH		 1861 goto out;
1862
1863 *(ptr + len) = '/';
1864 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1865
1866 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1867 break;
1868
b3b4aa74 1869 btrfs_release_path(path);
98d377a0 1870 key.objectid = key.offset;
8ad6fcab 1871 key.offset = (u64)-1;
98d377a0 1872 dirid = key.objectid;
98d377a0 1873 }
ac8e9819 1874 if (ptr < name)
98d377a0 1875 goto out;
77906a50 1876 memmove(name, ptr, total_len);
98d377a0
TH		 1877 name[total_len]='\0';
1878 ret = 0;
1879out:
1880 btrfs_free_path(path);
ac8e9819
CM		 1881 return ret;
1882}
1883
1884static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1885 void __user *argp)
1886{
1887 struct btrfs_ioctl_ino_lookup_args *args;
1888 struct inode *inode;
1889 int ret;
1890
1891 if (!capable(CAP_SYS_ADMIN))
1892 return -EPERM;
1893
2354d08f
JL		 1894 args = memdup_user(argp, sizeof(*args));
1895 if (IS_ERR(args))
1896 return PTR_ERR(args);
c2b96929 1897
ac8e9819
CM		 1898 inode = fdentry(file)->d_inode;
1899
1b53ac4d
CM		 1900 if (args->treeid == 0)
1901 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1902
ac8e9819
CM		 1903 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1904 args->treeid, args->objectid,
1905 args->name);
1906
1907 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1908 ret = -EFAULT;
1909
1910 kfree(args);
98d377a0
TH		 1911 return ret;
1912}
1913
76dda93c
YZ		 1914static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1915 void __user *arg)
1916{
1917 struct dentry *parent = fdentry(file);
1918 struct dentry *dentry;
1919 struct inode *dir = parent->d_inode;
1920 struct inode *inode;
1921 struct btrfs_root *root = BTRFS_I(dir)->root;
1922 struct btrfs_root *dest = NULL;
1923 struct btrfs_ioctl_vol_args *vol_args;
1924 struct btrfs_trans_handle *trans;
1925 int namelen;
1926 int ret;
1927 int err = 0;
1928
76dda93c
YZ		 1929 vol_args = memdup_user(arg, sizeof(*vol_args));
1930 if (IS_ERR(vol_args))
1931 return PTR_ERR(vol_args);
1932
1933 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1934 namelen = strlen(vol_args->name);
1935 if (strchr(vol_args->name, '/') ||
1936 strncmp(vol_args->name, "..", namelen) == 0) {
1937 err = -EINVAL;
1938 goto out;
1939 }
1940
a561be71 1941 err = mnt_want_write_file(file);
76dda93c
YZ		 1942 if (err)
1943 goto out;
1944
1945 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
1946 dentry = lookup_one_len(vol_args->name, parent, namelen);
1947 if (IS_ERR(dentry)) {
1948 err = PTR_ERR(dentry);
1949 goto out_unlock_dir;
1950 }
1951
1952 if (!dentry->d_inode) {
1953 err = -ENOENT;
1954 goto out_dput;
1955 }
1956
1957 inode = dentry->d_inode;
4260f7c7
SW		 1958 dest = BTRFS_I(inode)->root;
1959 if (!capable(CAP_SYS_ADMIN)){
1960 /*
1961 * Regular user. Only allow this with a special mount
1962 * option, when the user has write+exec access to the
1963 * subvol root, and when rmdir(2) would have been
1964 * allowed.
1965 *
1966 * Note that this is _not_ check that the subvol is
1967 * empty or doesn't contain data that we wouldn't
1968 * otherwise be able to delete.
1969 *
1970 * Users who want to delete empty subvols should try
1971 * rmdir(2).
1972 */
1973 err = -EPERM;
1974 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1975 goto out_dput;
1976
1977 /*
1978 * Do not allow deletion if the parent dir is the same
1979 * as the dir to be deleted. That means the ioctl
1980 * must be called on the dentry referencing the root
1981 * of the subvol, not a random directory contained
1982 * within it.
1983 */
1984 err = -EINVAL;
1985 if (root == dest)
1986 goto out_dput;
1987
1988 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
1989 if (err)
1990 goto out_dput;
1991
1992 /* check if subvolume may be deleted by a non-root user */
1993 err = btrfs_may_delete(dir, dentry, 1);
1994 if (err)
1995 goto out_dput;
1996 }
1997
33345d01 1998 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
76dda93c
YZ		 1999 err = -EINVAL;
2000 goto out_dput;
2001 }
2002
76dda93c
YZ		 2003 mutex_lock(&inode->i_mutex);
2004 err = d_invalidate(dentry);
2005 if (err)
2006 goto out_unlock;
2007
2008 down_write(&root->fs_info->subvol_sem);
2009
2010 err = may_destroy_subvol(dest);
2011 if (err)
2012 goto out_up_write;
2013
a22285a6
YZ		 2014 trans = btrfs_start_transaction(root, 0);
2015 if (IS_ERR(trans)) {
2016 err = PTR_ERR(trans);
d327099a 2017 goto out_up_write;
a22285a6
YZ		 2018 }
2019 trans->block_rsv = &root->fs_info->global_block_rsv;
2020
76dda93c
YZ		 2021 ret = btrfs_unlink_subvol(trans, root, dir,
2022 dest->root_key.objectid,
2023 dentry->d_name.name,
2024 dentry->d_name.len);
79787eaa
JM		 2025 if (ret) {
2026 err = ret;
2027 btrfs_abort_transaction(trans, root, ret);
2028 goto out_end_trans;
2029 }
76dda93c
YZ		 2030
2031 btrfs_record_root_in_trans(trans, dest);
2032
2033 memset(&dest->root_item.drop_progress, 0,
2034 sizeof(dest->root_item.drop_progress));
2035 dest->root_item.drop_level = 0;
2036 btrfs_set_root_refs(&dest->root_item, 0);
2037
d68fc57b
YZ		 2038 if (!xchg(&dest->orphan_item_inserted, 1)) {
2039 ret = btrfs_insert_orphan_item(trans,
2040 root->fs_info->tree_root,
2041 dest->root_key.objectid);
79787eaa
JM		 2042 if (ret) {
2043 btrfs_abort_transaction(trans, root, ret);
2044 err = ret;
2045 goto out_end_trans;
2046 }
d68fc57b 2047 }
79787eaa 2048out_end_trans:
531cb13f 2049 ret = btrfs_end_transaction(trans, root);
79787eaa
JM		 2050 if (ret && !err)
2051 err = ret;
76dda93c
YZ		 2052 inode->i_flags |= S_DEAD;
2053out_up_write:
2054 up_write(&root->fs_info->subvol_sem);
2055out_unlock:
2056 mutex_unlock(&inode->i_mutex);
2057 if (!err) {
efefb143 2058 shrink_dcache_sb(root->fs_info->sb);
76dda93c
YZ		 2059 btrfs_invalidate_inodes(dest);
2060 d_delete(dentry);
2061 }
2062out_dput:
2063 dput(dentry);
2064out_unlock_dir:
2065 mutex_unlock(&dir->i_mutex);
2a79f17e 2066 mnt_drop_write_file(file);
76dda93c
YZ		 2067out:
2068 kfree(vol_args);
2069 return err;
2070}
2071
1e701a32 2072static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
f46b5a66
CH		 2073{
2074 struct inode *inode = fdentry(file)->d_inode;
2075 struct btrfs_root *root = BTRFS_I(inode)->root;
1e701a32 2076 struct btrfs_ioctl_defrag_range_args *range;
c146afad
YZ		 2077 int ret;
2078
b83cc969
LZ		 2079 if (btrfs_root_readonly(root))
2080 return -EROFS;
2081
a561be71 2082 ret = mnt_want_write_file(file);
c146afad
YZ		 2083 if (ret)
2084 return ret;
f46b5a66
CH		 2085
2086 switch (inode->i_mode & S_IFMT) {
2087 case S_IFDIR:
e441d54d
CM		 2088 if (!capable(CAP_SYS_ADMIN)) {
2089 ret = -EPERM;
2090 goto out;
2091 }
8929ecfa
YZ		 2092 ret = btrfs_defrag_root(root, 0);
2093 if (ret)
2094 goto out;
2095 ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
f46b5a66
CH		 2096 break;
2097 case S_IFREG:
e441d54d
CM		 2098 if (!(file->f_mode & FMODE_WRITE)) {
2099 ret = -EINVAL;
2100 goto out;
2101 }
1e701a32
CM		 2102
2103 range = kzalloc(sizeof(*range), GFP_KERNEL);
2104 if (!range) {
2105 ret = -ENOMEM;
2106 goto out;
2107 }
2108
2109 if (argp) {
2110 if (copy_from_user(range, argp,
2111 sizeof(*range))) {
2112 ret = -EFAULT;
2113 kfree(range);
683be16e 2114 goto out;
1e701a32
CM		 2115 }
2116 /* compression requires us to start the IO */
2117 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2118 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2119 range->extent_thresh = (u32)-1;
2120 }
2121 } else {
2122 /* the rest are all set to zero by kzalloc */
2123 range->len = (u64)-1;
2124 }
4cb5300b
CM		 2125 ret = btrfs_defrag_file(fdentry(file)->d_inode, file,
2126 range, 0, 0);
2127 if (ret > 0)
2128 ret = 0;
1e701a32 2129 kfree(range);
f46b5a66 2130 break;
8929ecfa
YZ		 2131 default:
2132 ret = -EINVAL;
f46b5a66 2133 }
e441d54d 2134out:
2a79f17e 2135 mnt_drop_write_file(file);
e441d54d 2136 return ret;
f46b5a66
CH		 2137}
2138
b2950863 2139static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
f46b5a66
CH		 2140{
2141 struct btrfs_ioctl_vol_args *vol_args;
2142 int ret;
2143
e441d54d
CM		 2144 if (!capable(CAP_SYS_ADMIN))
2145 return -EPERM;
2146
c9e9f97b
ID		 2147 mutex_lock(&root->fs_info->volume_mutex);
2148 if (root->fs_info->balance_ctl) {
2149 printk(KERN_INFO "btrfs: balance in progress\n");
2150 ret = -EINVAL;
2151 goto out;
2152 }
2153
dae7b665 2154 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 2155 if (IS_ERR(vol_args)) {
2156 ret = PTR_ERR(vol_args);
2157 goto out;
2158 }
f46b5a66 2159
5516e595 2160 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66
CH		 2161 ret = btrfs_init_new_device(root, vol_args->name);
2162
f46b5a66 2163 kfree(vol_args);
c9e9f97b
ID		 2164out:
2165 mutex_unlock(&root->fs_info->volume_mutex);
f46b5a66
CH		 2166 return ret;
2167}
2168
b2950863 2169static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
f46b5a66
CH		 2170{
2171 struct btrfs_ioctl_vol_args *vol_args;
2172 int ret;
2173
e441d54d
CM		 2174 if (!capable(CAP_SYS_ADMIN))
2175 return -EPERM;
2176
c146afad
YZ		 2177 if (root->fs_info->sb->s_flags & MS_RDONLY)
2178 return -EROFS;
2179
c9e9f97b
ID		 2180 mutex_lock(&root->fs_info->volume_mutex);
2181 if (root->fs_info->balance_ctl) {
2182 printk(KERN_INFO "btrfs: balance in progress\n");
2183 ret = -EINVAL;
2184 goto out;
2185 }
2186
dae7b665 2187 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 2188 if (IS_ERR(vol_args)) {
2189 ret = PTR_ERR(vol_args);
2190 goto out;
2191 }
f46b5a66 2192
5516e595 2193 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66
CH		 2194 ret = btrfs_rm_device(root, vol_args->name);
2195
f46b5a66 2196 kfree(vol_args);
c9e9f97b
ID		 2197out:
2198 mutex_unlock(&root->fs_info->volume_mutex);
f46b5a66
CH		 2199 return ret;
2200}
2201
475f6387
JS		 2202static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2203{
027ed2f0 2204 struct btrfs_ioctl_fs_info_args *fi_args;
475f6387
JS		 2205 struct btrfs_device *device;
2206 struct btrfs_device *next;
2207 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
027ed2f0 2208 int ret = 0;
475f6387
JS		 2209
2210 if (!capable(CAP_SYS_ADMIN))
2211 return -EPERM;
2212
027ed2f0
LZ		 2213 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2214 if (!fi_args)
2215 return -ENOMEM;
2216
2217 fi_args->num_devices = fs_devices->num_devices;
2218 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
475f6387
JS		 2219
2220 mutex_lock(&fs_devices->device_list_mutex);
2221 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
027ed2f0
LZ		 2222 if (device->devid > fi_args->max_id)
2223 fi_args->max_id = device->devid;
475f6387
JS		 2224 }
2225 mutex_unlock(&fs_devices->device_list_mutex);
2226
027ed2f0
LZ		 2227 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2228 ret = -EFAULT;
475f6387 2229
027ed2f0
LZ		 2230 kfree(fi_args);
2231 return ret;
475f6387
JS		 2232}
2233
2234static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2235{
2236 struct btrfs_ioctl_dev_info_args *di_args;
2237 struct btrfs_device *dev;
2238 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2239 int ret = 0;
2240 char *s_uuid = NULL;
2241 char empty_uuid[BTRFS_UUID_SIZE] = {0};
2242
2243 if (!capable(CAP_SYS_ADMIN))
2244 return -EPERM;
2245
2246 di_args = memdup_user(arg, sizeof(*di_args));
2247 if (IS_ERR(di_args))
2248 return PTR_ERR(di_args);
2249
2250 if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2251 s_uuid = di_args->uuid;
2252
2253 mutex_lock(&fs_devices->device_list_mutex);
2254 dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL);
2255 mutex_unlock(&fs_devices->device_list_mutex);
2256
2257 if (!dev) {
2258 ret = -ENODEV;
2259 goto out;
2260 }
2261
2262 di_args->devid = dev->devid;
2263 di_args->bytes_used = dev->bytes_used;
2264 di_args->total_bytes = dev->total_bytes;
2265 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
99ba55ad
SB		 2266 if (dev->name)
2267 strncpy(di_args->path, dev->name, sizeof(di_args->path));
2268 else
2269 di_args->path[0] = '\0';
475f6387
JS		 2270
2271out:
2272 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2273 ret = -EFAULT;
2274
2275 kfree(di_args);
2276 return ret;
2277}
2278
76dda93c
YZ		 2279static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2280 u64 off, u64 olen, u64 destoff)
f46b5a66
CH		 2281{
2282 struct inode *inode = fdentry(file)->d_inode;
2283 struct btrfs_root *root = BTRFS_I(inode)->root;
2284 struct file *src_file;
2285 struct inode *src;
2286 struct btrfs_trans_handle *trans;
f46b5a66 2287 struct btrfs_path *path;
f46b5a66 2288 struct extent_buffer *leaf;
ae01a0ab
YZ		 2289 char *buf;
2290 struct btrfs_key key;
f46b5a66
CH		 2291 u32 nritems;
2292 int slot;
ae01a0ab 2293 int ret;
c5c9cd4d
SW		 2294 u64 len = olen;
2295 u64 bs = root->fs_info->sb->s_blocksize;
2296 u64 hint_byte;
d20f7043 2297
c5c9cd4d
SW		 2298 /*
2299 * TODO:
2300 * - split compressed inline extents. annoying: we need to
2301 * decompress into destination's address_space (the file offset
2302 * may change, so source mapping won't do), then recompress (or
2303 * otherwise reinsert) a subrange.
2304 * - allow ranges within the same file to be cloned (provided
2305 * they don't overlap)?
2306 */
2307
e441d54d 2308 /* the destination must be opened for writing */
2ebc3464 2309 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
e441d54d
CM		 2310 return -EINVAL;
2311
b83cc969
LZ		 2312 if (btrfs_root_readonly(root))
2313 return -EROFS;
2314
a561be71 2315 ret = mnt_want_write_file(file);
c146afad
YZ		 2316 if (ret)
2317 return ret;
2318
c5c9cd4d 2319 src_file = fget(srcfd);
ab67b7c1
YZ		 2320 if (!src_file) {
2321 ret = -EBADF;
2322 goto out_drop_write;
2323 }
5dc64164 2324
f46b5a66
CH		 2325 src = src_file->f_dentry->d_inode;
2326
c5c9cd4d
SW		 2327 ret = -EINVAL;
2328 if (src == inode)
2329 goto out_fput;
2330
5dc64164
DR		 2331 /* the src must be open for reading */
2332 if (!(src_file->f_mode & FMODE_READ))
2333 goto out_fput;
2334
0e7b824c
LZ		 2335 /* don't make the dst file partly checksummed */
2336 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2337 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2338 goto out_fput;
2339
ae01a0ab
YZ		 2340 ret = -EISDIR;
2341 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2342 goto out_fput;
2343
f46b5a66 2344 ret = -EXDEV;
ae01a0ab
YZ		 2345 if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
2346 goto out_fput;
2347
2348 ret = -ENOMEM;
2349 buf = vmalloc(btrfs_level_size(root, 0));
2350 if (!buf)
2351 goto out_fput;
2352
2353 path = btrfs_alloc_path();
2354 if (!path) {
2355 vfree(buf);
f46b5a66 2356 goto out_fput;
ae01a0ab
YZ		 2357 }
2358 path->reada = 2;
f46b5a66
CH		 2359
2360 if (inode < src) {
fccdae43
SW		 2361 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2362 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
f46b5a66 2363 } else {
fccdae43
SW		 2364 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2365 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
f46b5a66
CH		 2366 }
2367
c5c9cd4d
SW		 2368 /* determine range to clone */
2369 ret = -EINVAL;
2ebc3464 2370 if (off + len > src->i_size || off + len < off)
f46b5a66 2371 goto out_unlock;
c5c9cd4d
SW		 2372 if (len == 0)
2373 olen = len = src->i_size - off;
2374 /* if we extend to eof, continue to block boundary */
2375 if (off + len == src->i_size)
2a6b8dae 2376 len = ALIGN(src->i_size, bs) - off;
c5c9cd4d
SW		 2377
2378 /* verify the end result is block aligned */
2a6b8dae
LZ		 2379 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2380 !IS_ALIGNED(destoff, bs))
c5c9cd4d
SW		 2381 goto out_unlock;
2382
d525e8ab
LZ		 2383 if (destoff > inode->i_size) {
2384 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2385 if (ret)
2386 goto out_unlock;
2387 }
2388
71ef0786
LZ		 2389 /* truncate page cache pages from target inode range */
2390 truncate_inode_pages_range(&inode->i_data, destoff,
2391 PAGE_CACHE_ALIGN(destoff + len) - 1);
2392
f46b5a66
CH		 2393 /* do any pending delalloc/csum calc on src, one way or
2394 another, and lock file content */
2395 while (1) {
31840ae1 2396 struct btrfs_ordered_extent *ordered;
d0082371 2397 lock_extent(&BTRFS_I(src)->io_tree, off, off+len);
9a019196
SW		 2398 ordered = btrfs_lookup_first_ordered_extent(src, off+len);
2399 if (!ordered &&
2400 !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len,
2401 EXTENT_DELALLOC, 0, NULL))
f46b5a66 2402 break;
d0082371 2403 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len);
ae01a0ab
YZ		 2404 if (ordered)
2405 btrfs_put_ordered_extent(ordered);
9a019196 2406 btrfs_wait_ordered_range(src, off, len);
f46b5a66
CH		 2407 }
2408
c5c9cd4d 2409 /* clone data */
33345d01 2410 key.objectid = btrfs_ino(src);
ae01a0ab
YZ		 2411 key.type = BTRFS_EXTENT_DATA_KEY;
2412 key.offset = 0;
f46b5a66
CH		 2413
2414 while (1) {
2415 /*
2416 * note the key will change type as we walk through the
2417 * tree.
2418 */
a22285a6 2419 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
f46b5a66
CH		 2420 if (ret < 0)
2421 goto out;
2422
ae01a0ab
YZ		 2423 nritems = btrfs_header_nritems(path->nodes[0]);
2424 if (path->slots[0] >= nritems) {
f46b5a66
CH		 2425 ret = btrfs_next_leaf(root, path);
2426 if (ret < 0)
2427 goto out;
2428 if (ret > 0)
2429 break;
ae01a0ab 2430 nritems = btrfs_header_nritems(path->nodes[0]);
f46b5a66
CH		 2431 }
2432 leaf = path->nodes[0];
2433 slot = path->slots[0];
f46b5a66 2434
ae01a0ab 2435 btrfs_item_key_to_cpu(leaf, &key, slot);
d20f7043 2436 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
33345d01 2437 key.objectid != btrfs_ino(src))
f46b5a66
CH		 2438 break;
2439
c5c9cd4d
SW		 2440 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2441 struct btrfs_file_extent_item *extent;
2442 int type;
31840ae1
ZY		 2443 u32 size;
2444 struct btrfs_key new_key;
c5c9cd4d
SW		 2445 u64 disko = 0, diskl = 0;
2446 u64 datao = 0, datal = 0;
2447 u8 comp;
b5384d48 2448 u64 endoff;
31840ae1
ZY		 2449
2450 size = btrfs_item_size_nr(leaf, slot);
2451 read_extent_buffer(leaf, buf,
2452 btrfs_item_ptr_offset(leaf, slot),
2453 size);
c5c9cd4d
SW		 2454
2455 extent = btrfs_item_ptr(leaf, slot,
2456 struct btrfs_file_extent_item);
2457 comp = btrfs_file_extent_compression(leaf, extent);
2458 type = btrfs_file_extent_type(leaf, extent);
c8a894d7
CM		 2459 if (type == BTRFS_FILE_EXTENT_REG ||
2460 type == BTRFS_FILE_EXTENT_PREALLOC) {
d397712b
CM		 2461 disko = btrfs_file_extent_disk_bytenr(leaf,
2462 extent);
2463 diskl = btrfs_file_extent_disk_num_bytes(leaf,
2464 extent);
c5c9cd4d 2465 datao = btrfs_file_extent_offset(leaf, extent);
d397712b
CM		 2466 datal = btrfs_file_extent_num_bytes(leaf,
2467 extent);
c5c9cd4d
SW		 2468 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2469 /* take upper bound, may be compressed */
2470 datal = btrfs_file_extent_ram_bytes(leaf,
2471 extent);
2472 }
b3b4aa74 2473 btrfs_release_path(path);
31840ae1 2474
050006a7 2475 if (key.offset + datal <= off ||
c5c9cd4d
SW		 2476 key.offset >= off+len)
2477 goto next;
2478
31840ae1 2479 memcpy(&new_key, &key, sizeof(new_key));
33345d01 2480 new_key.objectid = btrfs_ino(inode);
4d728ec7
LZ		 2481 if (off <= key.offset)
2482 new_key.offset = key.offset + destoff - off;
2483 else
2484 new_key.offset = destoff;
31840ae1 2485
b6f3409b
SW		 2486 /*
2487 * 1 - adjusting old extent (we may have to split it)
2488 * 1 - add new extent
2489 * 1 - inode update
2490 */
2491 trans = btrfs_start_transaction(root, 3);
a22285a6
YZ		 2492 if (IS_ERR(trans)) {
2493 ret = PTR_ERR(trans);
2494 goto out;
2495 }
2496
c8a894d7
CM		 2497 if (type == BTRFS_FILE_EXTENT_REG ||
2498 type == BTRFS_FILE_EXTENT_PREALLOC) {
d72c0842
LZ		 2499 /*
2500 * a | --- range to clone ---| b
2501 * | ------------- extent ------------- |
2502 */
2503
2504 /* substract range b */
2505 if (key.offset + datal > off + len)
2506 datal = off + len - key.offset;
2507
2508 /* substract range a */
a22285a6
YZ		 2509 if (off > key.offset) {
2510 datao += off - key.offset;
2511 datal -= off - key.offset;
2512 }
2513
a22285a6
YZ		 2514 ret = btrfs_drop_extents(trans, inode,
2515 new_key.offset,
2516 new_key.offset + datal,
2517 &hint_byte, 1);
79787eaa
JM		 2518 if (ret) {
2519 btrfs_abort_transaction(trans, root,
2520 ret);
2521 btrfs_end_transaction(trans, root);
2522 goto out;
2523 }
a22285a6 2524
c5c9cd4d
SW		 2525 ret = btrfs_insert_empty_item(trans, root, path,
2526 &new_key, size);
79787eaa
JM		 2527 if (ret) {
2528 btrfs_abort_transaction(trans, root,
2529 ret);
2530 btrfs_end_transaction(trans, root);
2531 goto out;
2532 }
c5c9cd4d
SW		 2533
2534 leaf = path->nodes[0];
2535 slot = path->slots[0];
2536 write_extent_buffer(leaf, buf,
31840ae1
ZY		 2537 btrfs_item_ptr_offset(leaf, slot),
2538 size);
ae01a0ab 2539
c5c9cd4d 2540 extent = btrfs_item_ptr(leaf, slot,
f46b5a66 2541 struct btrfs_file_extent_item);
c5c9cd4d 2542
c5c9cd4d
SW		 2543 /* disko == 0 means it's a hole */
2544 if (!disko)
2545 datao = 0;
c5c9cd4d
SW		 2546
2547 btrfs_set_file_extent_offset(leaf, extent,
2548 datao);
2549 btrfs_set_file_extent_num_bytes(leaf, extent,
2550 datal);
2551 if (disko) {
2552 inode_add_bytes(inode, datal);
ae01a0ab 2553 ret = btrfs_inc_extent_ref(trans, root,
5d4f98a2
YZ		 2554 disko, diskl, 0,
2555 root->root_key.objectid,
33345d01 2556 btrfs_ino(inode),
66d7e7f0
AJ		 2557 new_key.offset - datao,
2558 0);
79787eaa
JM		 2559 if (ret) {
2560 btrfs_abort_transaction(trans,
2561 root,
2562 ret);
2563 btrfs_end_transaction(trans,
2564 root);
2565 goto out;
2566
2567 }
f46b5a66 2568 }
c5c9cd4d
SW		 2569 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2570 u64 skip = 0;
2571 u64 trim = 0;
2572 if (off > key.offset) {
2573 skip = off - key.offset;
2574 new_key.offset += skip;
2575 }
d397712b 2576
c5c9cd4d
SW		 2577 if (key.offset + datal > off+len)
2578 trim = key.offset + datal - (off+len);
d397712b 2579
c5c9cd4d 2580 if (comp && (skip || trim)) {
c5c9cd4d 2581 ret = -EINVAL;
a22285a6 2582 btrfs_end_transaction(trans, root);
c5c9cd4d
SW		 2583 goto out;
2584 }
2585 size -= skip + trim;
2586 datal -= skip + trim;
a22285a6
YZ		 2587
2588 ret = btrfs_drop_extents(trans, inode,
2589 new_key.offset,
2590 new_key.offset + datal,
2591 &hint_byte, 1);
79787eaa
JM		 2592 if (ret) {
2593 btrfs_abort_transaction(trans, root,
2594 ret);
2595 btrfs_end_transaction(trans, root);
2596 goto out;
2597 }
a22285a6 2598
c5c9cd4d
SW		 2599 ret = btrfs_insert_empty_item(trans, root, path,
2600 &new_key, size);
79787eaa
JM		 2601 if (ret) {
2602 btrfs_abort_transaction(trans, root,
2603 ret);
2604 btrfs_end_transaction(trans, root);
2605 goto out;
2606 }
c5c9cd4d
SW		 2607
2608 if (skip) {
d397712b
CM		 2609 u32 start =
2610 btrfs_file_extent_calc_inline_size(0);
c5c9cd4d
SW		 2611 memmove(buf+start, buf+start+skip,
2612 datal);
2613 }
2614
2615 leaf = path->nodes[0];
2616 slot = path->slots[0];
2617 write_extent_buffer(leaf, buf,
2618 btrfs_item_ptr_offset(leaf, slot),
2619 size);
2620 inode_add_bytes(inode, datal);
f46b5a66 2621 }
c5c9cd4d
SW		 2622
2623 btrfs_mark_buffer_dirty(leaf);
b3b4aa74 2624 btrfs_release_path(path);
c5c9cd4d 2625
0c4d2d95 2626 inode_inc_iversion(inode);
a22285a6 2627 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
b5384d48
SW		 2628
2629 /*
2630 * we round up to the block size at eof when
2631 * determining which extents to clone above,
2632 * but shouldn't round up the file size
2633 */
2634 endoff = new_key.offset + datal;
5f3888ff
LZ		 2635 if (endoff > destoff+olen)
2636 endoff = destoff+olen;
b5384d48
SW		 2637 if (endoff > inode->i_size)
2638 btrfs_i_size_write(inode, endoff);
2639
a22285a6 2640 ret = btrfs_update_inode(trans, root, inode);
79787eaa
JM		 2641 if (ret) {
2642 btrfs_abort_transaction(trans, root, ret);
2643 btrfs_end_transaction(trans, root);
2644 goto out;
2645 }
2646 ret = btrfs_end_transaction(trans, root);
a22285a6 2647 }
d397712b 2648next:
b3b4aa74 2649 btrfs_release_path(path);
f46b5a66 2650 key.offset++;
f46b5a66 2651 }
f46b5a66
CH		 2652 ret = 0;
2653out:
b3b4aa74 2654 btrfs_release_path(path);
d0082371 2655 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len);
f46b5a66
CH		 2656out_unlock:
2657 mutex_unlock(&src->i_mutex);
2658 mutex_unlock(&inode->i_mutex);
ae01a0ab
YZ		 2659 vfree(buf);
2660 btrfs_free_path(path);
f46b5a66
CH		 2661out_fput:
2662 fput(src_file);
ab67b7c1 2663out_drop_write:
2a79f17e 2664 mnt_drop_write_file(file);
f46b5a66
CH		 2665 return ret;
2666}
2667
7a865e8a 2668static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
c5c9cd4d
SW		 2669{
2670 struct btrfs_ioctl_clone_range_args args;
2671
7a865e8a 2672 if (copy_from_user(&args, argp, sizeof(args)))
c5c9cd4d
SW		 2673 return -EFAULT;
2674 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2675 args.src_length, args.dest_offset);
2676}
2677
f46b5a66
CH		 2678/*
2679 * there are many ways the trans_start and trans_end ioctls can lead
2680 * to deadlocks. They should only be used by applications that
2681 * basically own the machine, and have a very in depth understanding
2682 * of all the possible deadlocks and enospc problems.
2683 */
b2950863 2684static long btrfs_ioctl_trans_start(struct file *file)
f46b5a66
CH		 2685{
2686 struct inode *inode = fdentry(file)->d_inode;
2687 struct btrfs_root *root = BTRFS_I(inode)->root;
2688 struct btrfs_trans_handle *trans;
1ab86aed 2689 int ret;
f46b5a66 2690
1ab86aed 2691 ret = -EPERM;
df5b5520 2692 if (!capable(CAP_SYS_ADMIN))
1ab86aed 2693 goto out;
df5b5520 2694
1ab86aed
SW		 2695 ret = -EINPROGRESS;
2696 if (file->private_data)
f46b5a66 2697 goto out;
9ca9ee09 2698
b83cc969
LZ		 2699 ret = -EROFS;
2700 if (btrfs_root_readonly(root))
2701 goto out;
2702
a561be71 2703 ret = mnt_want_write_file(file);
c146afad
YZ		 2704 if (ret)
2705 goto out;
2706
a4abeea4 2707 atomic_inc(&root->fs_info->open_ioctl_trans);
9ca9ee09 2708
1ab86aed 2709 ret = -ENOMEM;
7a7eaa40 2710 trans = btrfs_start_ioctl_transaction(root);
abd30bb0 2711 if (IS_ERR(trans))
1ab86aed
SW		 2712 goto out_drop;
2713
2714 file->private_data = trans;
2715 return 0;
2716
2717out_drop:
a4abeea4 2718 atomic_dec(&root->fs_info->open_ioctl_trans);
2a79f17e 2719 mnt_drop_write_file(file);
f46b5a66 2720out:
f46b5a66
CH		 2721 return ret;
2722}
2723
6ef5ed0d
JB		 2724static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2725{
2726 struct inode *inode = fdentry(file)->d_inode;
2727 struct btrfs_root *root = BTRFS_I(inode)->root;
2728 struct btrfs_root *new_root;
2729 struct btrfs_dir_item *di;
2730 struct btrfs_trans_handle *trans;
2731 struct btrfs_path *path;
2732 struct btrfs_key location;
2733 struct btrfs_disk_key disk_key;
2734 struct btrfs_super_block *disk_super;
2735 u64 features;
2736 u64 objectid = 0;
2737 u64 dir_id;
2738
2739 if (!capable(CAP_SYS_ADMIN))
2740 return -EPERM;
2741
2742 if (copy_from_user(&objectid, argp, sizeof(objectid)))
2743 return -EFAULT;
2744
2745 if (!objectid)
2746 objectid = root->root_key.objectid;
2747
2748 location.objectid = objectid;
2749 location.type = BTRFS_ROOT_ITEM_KEY;
2750 location.offset = (u64)-1;
2751
2752 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2753 if (IS_ERR(new_root))
2754 return PTR_ERR(new_root);
2755
2756 if (btrfs_root_refs(&new_root->root_item) == 0)
2757 return -ENOENT;
2758
2759 path = btrfs_alloc_path();
2760 if (!path)
2761 return -ENOMEM;
2762 path->leave_spinning = 1;
2763
2764 trans = btrfs_start_transaction(root, 1);
98d5dc13 2765 if (IS_ERR(trans)) {
6ef5ed0d 2766 btrfs_free_path(path);
98d5dc13 2767 return PTR_ERR(trans);
6ef5ed0d
JB		 2768 }
2769
6c41761f 2770 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
6ef5ed0d
JB		 2771 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2772 dir_id, "default", 7, 1);
cf1e99a4 2773 if (IS_ERR_OR_NULL(di)) {
6ef5ed0d
JB		 2774 btrfs_free_path(path);
2775 btrfs_end_transaction(trans, root);
2776 printk(KERN_ERR "Umm, you don't have the default dir item, "
2777 "this isn't going to work\n");
2778 return -ENOENT;
2779 }
2780
2781 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2782 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2783 btrfs_mark_buffer_dirty(path->nodes[0]);
2784 btrfs_free_path(path);
2785
6c41761f 2786 disk_super = root->fs_info->super_copy;
6ef5ed0d
JB		 2787 features = btrfs_super_incompat_flags(disk_super);
2788 if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) {
2789 features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL;
2790 btrfs_set_super_incompat_flags(disk_super, features);
2791 }
2792 btrfs_end_transaction(trans, root);
2793
2794 return 0;
2795}
2796
bf5fc093
JB		 2797static void get_block_group_info(struct list_head *groups_list,
2798 struct btrfs_ioctl_space_info *space)
2799{
2800 struct btrfs_block_group_cache *block_group;
2801
2802 space->total_bytes = 0;
2803 space->used_bytes = 0;
2804 space->flags = 0;
2805 list_for_each_entry(block_group, groups_list, list) {
2806 space->flags = block_group->flags;
2807 space->total_bytes += block_group->key.offset;
2808 space->used_bytes +=
2809 btrfs_block_group_used(&block_group->item);
2810 }
2811}
2812
1406e432
JB		 2813long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2814{
2815 struct btrfs_ioctl_space_args space_args;
2816 struct btrfs_ioctl_space_info space;
2817 struct btrfs_ioctl_space_info *dest;
7fde62bf 2818 struct btrfs_ioctl_space_info *dest_orig;
13f2696f 2819 struct btrfs_ioctl_space_info __user *user_dest;
1406e432 2820 struct btrfs_space_info *info;
bf5fc093
JB		 2821 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2822 BTRFS_BLOCK_GROUP_SYSTEM,
2823 BTRFS_BLOCK_GROUP_METADATA,
2824 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2825 int num_types = 4;
7fde62bf 2826 int alloc_size;
1406e432 2827 int ret = 0;
51788b1b 2828 u64 slot_count = 0;
bf5fc093 2829 int i, c;
1406e432
JB		 2830
2831 if (copy_from_user(&space_args,
2832 (struct btrfs_ioctl_space_args __user *)arg,
2833 sizeof(space_args)))
2834 return -EFAULT;
2835
bf5fc093
JB		 2836 for (i = 0; i < num_types; i++) {
2837 struct btrfs_space_info *tmp;
2838
2839 info = NULL;
2840 rcu_read_lock();
2841 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2842 list) {
2843 if (tmp->flags == types[i]) {
2844 info = tmp;
2845 break;
2846 }
2847 }
2848 rcu_read_unlock();
2849
2850 if (!info)
2851 continue;
2852
2853 down_read(&info->groups_sem);
2854 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2855 if (!list_empty(&info->block_groups[c]))
2856 slot_count++;
2857 }
2858 up_read(&info->groups_sem);
2859 }
7fde62bf
CM		 2860
2861 /* space_slots == 0 means they are asking for a count */
2862 if (space_args.space_slots == 0) {
2863 space_args.total_spaces = slot_count;
2864 goto out;
2865 }
bf5fc093 2866
51788b1b 2867 slot_count = min_t(u64, space_args.space_slots, slot_count);
bf5fc093 2868
7fde62bf 2869 alloc_size = sizeof(*dest) * slot_count;
bf5fc093 2870
7fde62bf
CM		 2871 /* we generally have at most 6 or so space infos, one for each raid
2872 * level. So, a whole page should be more than enough for everyone
2873 */
2874 if (alloc_size > PAGE_CACHE_SIZE)
2875 return -ENOMEM;
2876
1406e432 2877 space_args.total_spaces = 0;
7fde62bf
CM		 2878 dest = kmalloc(alloc_size, GFP_NOFS);
2879 if (!dest)
2880 return -ENOMEM;
2881 dest_orig = dest;
1406e432 2882
7fde62bf 2883 /* now we have a buffer to copy into */
bf5fc093
JB		 2884 for (i = 0; i < num_types; i++) {
2885 struct btrfs_space_info *tmp;
2886
51788b1b
DR		 2887 if (!slot_count)
2888 break;
2889
bf5fc093
JB		 2890 info = NULL;
2891 rcu_read_lock();
2892 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2893 list) {
2894 if (tmp->flags == types[i]) {
2895 info = tmp;
2896 break;
2897 }
2898 }
2899 rcu_read_unlock();
7fde62bf 2900
bf5fc093
JB		 2901 if (!info)
2902 continue;
2903 down_read(&info->groups_sem);
2904 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2905 if (!list_empty(&info->block_groups[c])) {
2906 get_block_group_info(&info->block_groups[c],
2907 &space);
2908 memcpy(dest, &space, sizeof(space));
2909 dest++;
2910 space_args.total_spaces++;
51788b1b 2911 slot_count--;
bf5fc093 2912 }
51788b1b
DR		 2913 if (!slot_count)
2914 break;
bf5fc093
JB		 2915 }
2916 up_read(&info->groups_sem);
1406e432 2917 }
1406e432 2918
2eec6c81 2919 user_dest = (struct btrfs_ioctl_space_info __user *)
7fde62bf
CM		 2920 (arg + sizeof(struct btrfs_ioctl_space_args));
2921
2922 if (copy_to_user(user_dest, dest_orig, alloc_size))
2923 ret = -EFAULT;
2924
2925 kfree(dest_orig);
2926out:
2927 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
1406e432
JB		 2928 ret = -EFAULT;
2929
2930 return ret;
2931}
2932
f46b5a66
CH		 2933/*
2934 * there are many ways the trans_start and trans_end ioctls can lead
2935 * to deadlocks. They should only be used by applications that
2936 * basically own the machine, and have a very in depth understanding
2937 * of all the possible deadlocks and enospc problems.
2938 */
2939long btrfs_ioctl_trans_end(struct file *file)
2940{
2941 struct inode *inode = fdentry(file)->d_inode;
2942 struct btrfs_root *root = BTRFS_I(inode)->root;
2943 struct btrfs_trans_handle *trans;
f46b5a66 2944
f46b5a66 2945 trans = file->private_data;
1ab86aed
SW		 2946 if (!trans)
2947 return -EINVAL;
b214107e 2948 file->private_data = NULL;
9ca9ee09 2949
1ab86aed
SW		 2950 btrfs_end_transaction(trans, root);
2951
a4abeea4 2952 atomic_dec(&root->fs_info->open_ioctl_trans);
9ca9ee09 2953
2a79f17e 2954 mnt_drop_write_file(file);
1ab86aed 2955 return 0;
f46b5a66
CH		 2956}
2957
46204592
SW		 2958static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
2959{
2960 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2961 struct btrfs_trans_handle *trans;
2962 u64 transid;
db5b493a 2963 int ret;
46204592
SW		 2964
2965 trans = btrfs_start_transaction(root, 0);
98d5dc13
TI		 2966 if (IS_ERR(trans))
2967 return PTR_ERR(trans);
46204592 2968 transid = trans->transid;
db5b493a 2969 ret = btrfs_commit_transaction_async(trans, root, 0);
8b2b2d3c
TI		 2970 if (ret) {
2971 btrfs_end_transaction(trans, root);
db5b493a 2972 return ret;
8b2b2d3c 2973 }
46204592
SW		 2974
2975 if (argp)
2976 if (copy_to_user(argp, &transid, sizeof(transid)))
2977 return -EFAULT;
2978 return 0;
2979}
2980
2981static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
2982{
2983 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2984 u64 transid;
2985
2986 if (argp) {
2987 if (copy_from_user(&transid, argp, sizeof(transid)))
2988 return -EFAULT;
2989 } else {
2990 transid = 0; /* current trans */
2991 }
2992 return btrfs_wait_for_commit(root, transid);
2993}
2994
475f6387
JS		 2995static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg)
2996{
2997 int ret;
2998 struct btrfs_ioctl_scrub_args *sa;
2999
3000 if (!capable(CAP_SYS_ADMIN))
3001 return -EPERM;
3002
3003 sa = memdup_user(arg, sizeof(*sa));
3004 if (IS_ERR(sa))
3005 return PTR_ERR(sa);
3006
3007 ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end,
8628764e 3008 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY);
475f6387
JS		 3009
3010 if (copy_to_user(arg, sa, sizeof(*sa)))
3011 ret = -EFAULT;
3012
3013 kfree(sa);
3014 return ret;
3015}
3016
3017static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3018{
3019 if (!capable(CAP_SYS_ADMIN))
3020 return -EPERM;
3021
3022 return btrfs_scrub_cancel(root);
3023}
3024
3025static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3026 void __user *arg)
3027{
3028 struct btrfs_ioctl_scrub_args *sa;
3029 int ret;
3030
3031 if (!capable(CAP_SYS_ADMIN))
3032 return -EPERM;
3033
3034 sa = memdup_user(arg, sizeof(*sa));
3035 if (IS_ERR(sa))
3036 return PTR_ERR(sa);
3037
3038 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3039
3040 if (copy_to_user(arg, sa, sizeof(*sa)))
3041 ret = -EFAULT;
3042
3043 kfree(sa);
3044 return ret;
3045}
3046
d7728c96
JS		 3047static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3048{
3049 int ret = 0;
3050 int i;
740c3d22 3051 u64 rel_ptr;
d7728c96 3052 int size;
806468f8 3053 struct btrfs_ioctl_ino_path_args *ipa = NULL;
d7728c96
JS		 3054 struct inode_fs_paths *ipath = NULL;
3055 struct btrfs_path *path;
3056
3057 if (!capable(CAP_SYS_ADMIN))
3058 return -EPERM;
3059
3060 path = btrfs_alloc_path();
3061 if (!path) {
3062 ret = -ENOMEM;
3063 goto out;
3064 }
3065
3066 ipa = memdup_user(arg, sizeof(*ipa));
3067 if (IS_ERR(ipa)) {
3068 ret = PTR_ERR(ipa);
3069 ipa = NULL;
3070 goto out;
3071 }
3072
3073 size = min_t(u32, ipa->size, 4096);
3074 ipath = init_ipath(size, root, path);
3075 if (IS_ERR(ipath)) {
3076 ret = PTR_ERR(ipath);
3077 ipath = NULL;
3078 goto out;
3079 }
3080
3081 ret = paths_from_inode(ipa->inum, ipath);
3082 if (ret < 0)
3083 goto out;
3084
3085 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
745c4d8e
JM		 3086 rel_ptr = ipath->fspath->val[i] -
3087 (u64)(unsigned long)ipath->fspath->val;
740c3d22 3088 ipath->fspath->val[i] = rel_ptr;
d7728c96
JS		 3089 }
3090
745c4d8e
JM		 3091 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3092 (void *)(unsigned long)ipath->fspath, size);
d7728c96
JS		 3093 if (ret) {
3094 ret = -EFAULT;
3095 goto out;
3096 }
3097
3098out:
3099 btrfs_free_path(path);
3100 free_ipath(ipath);
3101 kfree(ipa);
3102
3103 return ret;
3104}
3105
3106static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3107{
3108 struct btrfs_data_container *inodes = ctx;
3109 const size_t c = 3 * sizeof(u64);
3110
3111 if (inodes->bytes_left >= c) {
3112 inodes->bytes_left -= c;
3113 inodes->val[inodes->elem_cnt] = inum;
3114 inodes->val[inodes->elem_cnt + 1] = offset;
3115 inodes->val[inodes->elem_cnt + 2] = root;
3116 inodes->elem_cnt += 3;
3117 } else {
3118 inodes->bytes_missing += c - inodes->bytes_left;
3119 inodes->bytes_left = 0;
3120 inodes->elem_missed += 3;
3121 }
3122
3123 return 0;
3124}
3125
3126static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3127 void __user *arg)
3128{
3129 int ret = 0;
3130 int size;
4692cf58 3131 u64 extent_item_pos;
d7728c96
JS		 3132 struct btrfs_ioctl_logical_ino_args *loi;
3133 struct btrfs_data_container *inodes = NULL;
3134 struct btrfs_path *path = NULL;
3135 struct btrfs_key key;
3136
3137 if (!capable(CAP_SYS_ADMIN))
3138 return -EPERM;
3139
3140 loi = memdup_user(arg, sizeof(*loi));
3141 if (IS_ERR(loi)) {
3142 ret = PTR_ERR(loi);
3143 loi = NULL;
3144 goto out;
3145 }
3146
3147 path = btrfs_alloc_path();
3148 if (!path) {
3149 ret = -ENOMEM;
3150 goto out;
3151 }
3152
3153 size = min_t(u32, loi->size, 4096);
3154 inodes = init_data_container(size);
3155 if (IS_ERR(inodes)) {
3156 ret = PTR_ERR(inodes);
3157 inodes = NULL;
3158 goto out;
3159 }
3160
3161 ret = extent_from_logical(root->fs_info, loi->logical, path, &key);
4692cf58 3162 btrfs_release_path(path);
d7728c96
JS		 3163
3164 if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK)
3165 ret = -ENOENT;
3166 if (ret < 0)
3167 goto out;
3168
4692cf58 3169 extent_item_pos = loi->logical - key.objectid;
7a3ae2f8
JS		 3170 ret = iterate_extent_inodes(root->fs_info, key.objectid,
3171 extent_item_pos, 0, build_ino_list,
4692cf58 3172 inodes);
d7728c96
JS		 3173
3174 if (ret < 0)
3175 goto out;
3176
745c4d8e
JM		 3177 ret = copy_to_user((void *)(unsigned long)loi->inodes,
3178 (void *)(unsigned long)inodes, size);
d7728c96
JS		 3179 if (ret)
3180 ret = -EFAULT;
3181
3182out:
3183 btrfs_free_path(path);
3184 kfree(inodes);
3185 kfree(loi);
3186
3187 return ret;
3188}
3189
19a39dce 3190void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
c9e9f97b
ID		 3191 struct btrfs_ioctl_balance_args *bargs)
3192{
3193 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3194
3195 bargs->flags = bctl->flags;
3196
837d5b6e
ID		 3197 if (atomic_read(&fs_info->balance_running))
3198 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3199 if (atomic_read(&fs_info->balance_pause_req))
3200 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
a7e99c69
ID		 3201 if (atomic_read(&fs_info->balance_cancel_req))
3202 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
837d5b6e 3203
c9e9f97b
ID		 3204 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3205 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3206 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
19a39dce
ID		 3207
3208 if (lock) {
3209 spin_lock(&fs_info->balance_lock);
3210 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3211 spin_unlock(&fs_info->balance_lock);
3212 } else {
3213 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3214 }
c9e9f97b
ID		 3215}
3216
3217static long btrfs_ioctl_balance(struct btrfs_root *root, void __user *arg)
3218{
3219 struct btrfs_fs_info *fs_info = root->fs_info;
3220 struct btrfs_ioctl_balance_args *bargs;
3221 struct btrfs_balance_control *bctl;
3222 int ret;
3223
3224 if (!capable(CAP_SYS_ADMIN))
3225 return -EPERM;
3226
3227 if (fs_info->sb->s_flags & MS_RDONLY)
3228 return -EROFS;
3229
3230 mutex_lock(&fs_info->volume_mutex);
3231 mutex_lock(&fs_info->balance_mutex);
3232
3233 if (arg) {
3234 bargs = memdup_user(arg, sizeof(*bargs));
3235 if (IS_ERR(bargs)) {
3236 ret = PTR_ERR(bargs);
3237 goto out;
3238 }
de322263
ID		 3239
3240 if (bargs->flags & BTRFS_BALANCE_RESUME) {
3241 if (!fs_info->balance_ctl) {
3242 ret = -ENOTCONN;
3243 goto out_bargs;
3244 }
3245
3246 bctl = fs_info->balance_ctl;
3247 spin_lock(&fs_info->balance_lock);
3248 bctl->flags |= BTRFS_BALANCE_RESUME;
3249 spin_unlock(&fs_info->balance_lock);
3250
3251 goto do_balance;
3252 }
c9e9f97b
ID		 3253 } else {
3254 bargs = NULL;
3255 }
3256
837d5b6e
ID		 3257 if (fs_info->balance_ctl) {
3258 ret = -EINPROGRESS;
3259 goto out_bargs;
3260 }
3261
c9e9f97b
ID		 3262 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3263 if (!bctl) {
3264 ret = -ENOMEM;
3265 goto out_bargs;
3266 }
3267
3268 bctl->fs_info = fs_info;
3269 if (arg) {
3270 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3271 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3272 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3273
3274 bctl->flags = bargs->flags;
f43ffb60
ID		 3275 } else {
3276 /* balance everything - no filters */
3277 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
c9e9f97b
ID		 3278 }
3279
de322263 3280do_balance:
c9e9f97b
ID		 3281 ret = btrfs_balance(bctl, bargs);
3282 /*
837d5b6e
ID		 3283 * bctl is freed in __cancel_balance or in free_fs_info if
3284 * restriper was paused all the way until unmount
c9e9f97b
ID		 3285 */
3286 if (arg) {
3287 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3288 ret = -EFAULT;
3289 }
3290
3291out_bargs:
3292 kfree(bargs);
3293out:
3294 mutex_unlock(&fs_info->balance_mutex);
3295 mutex_unlock(&fs_info->volume_mutex);
3296 return ret;
3297}
3298
837d5b6e
ID		 3299static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
3300{
3301 if (!capable(CAP_SYS_ADMIN))
3302 return -EPERM;
3303
3304 switch (cmd) {
3305 case BTRFS_BALANCE_CTL_PAUSE:
3306 return btrfs_pause_balance(root->fs_info);
a7e99c69
ID		 3307 case BTRFS_BALANCE_CTL_CANCEL:
3308 return btrfs_cancel_balance(root->fs_info);
837d5b6e
ID		 3309 }
3310
3311 return -EINVAL;
3312}
3313
19a39dce
ID		 3314static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
3315 void __user *arg)
3316{
3317 struct btrfs_fs_info *fs_info = root->fs_info;
3318 struct btrfs_ioctl_balance_args *bargs;
3319 int ret = 0;
3320
3321 if (!capable(CAP_SYS_ADMIN))
3322 return -EPERM;
3323
3324 mutex_lock(&fs_info->balance_mutex);
3325 if (!fs_info->balance_ctl) {
3326 ret = -ENOTCONN;
3327 goto out;
3328 }
3329
3330 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
3331 if (!bargs) {
3332 ret = -ENOMEM;
3333 goto out;
3334 }
3335
3336 update_ioctl_balance_args(fs_info, 1, bargs);
3337
3338 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3339 ret = -EFAULT;
3340
3341 kfree(bargs);
3342out:
3343 mutex_unlock(&fs_info->balance_mutex);
3344 return ret;
3345}
3346
f46b5a66
CH		 3347long btrfs_ioctl(struct file *file, unsigned int
3348 cmd, unsigned long arg)
3349{
3350 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
4bcabaa3 3351 void __user *argp = (void __user *)arg;
f46b5a66
CH		 3352
3353 switch (cmd) {
6cbff00f
CH		 3354 case FS_IOC_GETFLAGS:
3355 return btrfs_ioctl_getflags(file, argp);
3356 case FS_IOC_SETFLAGS:
3357 return btrfs_ioctl_setflags(file, argp);
3358 case FS_IOC_GETVERSION:
3359 return btrfs_ioctl_getversion(file, argp);
f7039b1d
LD		 3360 case FITRIM:
3361 return btrfs_ioctl_fitrim(file, argp);
f46b5a66 3362 case BTRFS_IOC_SNAP_CREATE:
fa0d2b9b 3363 return btrfs_ioctl_snap_create(file, argp, 0);
fdfb1e4f 3364 case BTRFS_IOC_SNAP_CREATE_V2:
fa0d2b9b 3365 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3de4586c 3366 case BTRFS_IOC_SUBVOL_CREATE:
fa0d2b9b 3367 return btrfs_ioctl_snap_create(file, argp, 1);
76dda93c
YZ		 3368 case BTRFS_IOC_SNAP_DESTROY:
3369 return btrfs_ioctl_snap_destroy(file, argp);
0caa102d
LZ		 3370 case BTRFS_IOC_SUBVOL_GETFLAGS:
3371 return btrfs_ioctl_subvol_getflags(file, argp);
3372 case BTRFS_IOC_SUBVOL_SETFLAGS:
3373 return btrfs_ioctl_subvol_setflags(file, argp);
6ef5ed0d
JB		 3374 case BTRFS_IOC_DEFAULT_SUBVOL:
3375 return btrfs_ioctl_default_subvol(file, argp);
f46b5a66 3376 case BTRFS_IOC_DEFRAG:
1e701a32
CM		 3377 return btrfs_ioctl_defrag(file, NULL);
3378 case BTRFS_IOC_DEFRAG_RANGE:
3379 return btrfs_ioctl_defrag(file, argp);
f46b5a66 3380 case BTRFS_IOC_RESIZE:
4bcabaa3 3381 return btrfs_ioctl_resize(root, argp);
f46b5a66 3382 case BTRFS_IOC_ADD_DEV:
4bcabaa3 3383 return btrfs_ioctl_add_dev(root, argp);
f46b5a66 3384 case BTRFS_IOC_RM_DEV:
4bcabaa3 3385 return btrfs_ioctl_rm_dev(root, argp);
475f6387
JS		 3386 case BTRFS_IOC_FS_INFO:
3387 return btrfs_ioctl_fs_info(root, argp);
3388 case BTRFS_IOC_DEV_INFO:
3389 return btrfs_ioctl_dev_info(root, argp);
f46b5a66 3390 case BTRFS_IOC_BALANCE:
c9e9f97b 3391 return btrfs_ioctl_balance(root, NULL);
f46b5a66 3392 case BTRFS_IOC_CLONE:
c5c9cd4d
SW		 3393 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
3394 case BTRFS_IOC_CLONE_RANGE:
7a865e8a 3395 return btrfs_ioctl_clone_range(file, argp);
f46b5a66
CH		 3396 case BTRFS_IOC_TRANS_START:
3397 return btrfs_ioctl_trans_start(file);
3398 case BTRFS_IOC_TRANS_END:
3399 return btrfs_ioctl_trans_end(file);
ac8e9819
CM		 3400 case BTRFS_IOC_TREE_SEARCH:
3401 return btrfs_ioctl_tree_search(file, argp);
3402 case BTRFS_IOC_INO_LOOKUP:
3403 return btrfs_ioctl_ino_lookup(file, argp);
d7728c96
JS		 3404 case BTRFS_IOC_INO_PATHS:
3405 return btrfs_ioctl_ino_to_path(root, argp);
3406 case BTRFS_IOC_LOGICAL_INO:
3407 return btrfs_ioctl_logical_to_ino(root, argp);
1406e432
JB		 3408 case BTRFS_IOC_SPACE_INFO:
3409 return btrfs_ioctl_space_info(root, argp);
f46b5a66
CH		 3410 case BTRFS_IOC_SYNC:
3411 btrfs_sync_fs(file->f_dentry->d_sb, 1);
3412 return 0;
46204592
SW		 3413 case BTRFS_IOC_START_SYNC:
3414 return btrfs_ioctl_start_sync(file, argp);
3415 case BTRFS_IOC_WAIT_SYNC:
3416 return btrfs_ioctl_wait_sync(file, argp);
475f6387
JS		 3417 case BTRFS_IOC_SCRUB:
3418 return btrfs_ioctl_scrub(root, argp);
3419 case BTRFS_IOC_SCRUB_CANCEL:
3420 return btrfs_ioctl_scrub_cancel(root, argp);
3421 case BTRFS_IOC_SCRUB_PROGRESS:
3422 return btrfs_ioctl_scrub_progress(root, argp);
c9e9f97b
ID		 3423 case BTRFS_IOC_BALANCE_V2:
3424 return btrfs_ioctl_balance(root, argp);
837d5b6e
ID		 3425 case BTRFS_IOC_BALANCE_CTL:
3426 return btrfs_ioctl_balance_ctl(root, arg);
19a39dce
ID		 3427 case BTRFS_IOC_BALANCE_PROGRESS:
3428 return btrfs_ioctl_balance_progress(root, argp);
f46b5a66
CH		 3429 }
3430
3431 return -ENOTTY;
3432}
Arne's tree of linux kernel.