projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
btrfs: ignore unfragmented file checks in defrag when compression enabled - rebased
[linux-2.6-block.git] / fs / btrfs / ioctl.c
CommitLineData
f46b5a66
CH		 1/*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
19#include <linux/kernel.h>
20#include <linux/bio.h>
21#include <linux/buffer_head.h>
22#include <linux/file.h>
23#include <linux/fs.h>
cb8e7090 24#include <linux/fsnotify.h>
f46b5a66
CH		 25#include <linux/pagemap.h>
26#include <linux/highmem.h>
27#include <linux/time.h>
28#include <linux/init.h>
29#include <linux/string.h>
f46b5a66 30#include <linux/backing-dev.h>
cb8e7090 31#include <linux/mount.h>
f46b5a66 32#include <linux/mpage.h>
cb8e7090 33#include <linux/namei.h>
f46b5a66
CH		 34#include <linux/swap.h>
35#include <linux/writeback.h>
36#include <linux/statfs.h>
37#include <linux/compat.h>
38#include <linux/bit_spinlock.h>
cb8e7090 39#include <linux/security.h>
f46b5a66 40#include <linux/xattr.h>
7ea394f1 41#include <linux/vmalloc.h>
5a0e3ad6 42#include <linux/slab.h>
f7039b1d 43#include <linux/blkdev.h>
4b4e25f2 44#include "compat.h"
f46b5a66
CH		 45#include "ctree.h"
46#include "disk-io.h"
47#include "transaction.h"
48#include "btrfs_inode.h"
49#include "ioctl.h"
50#include "print-tree.h"
51#include "volumes.h"
925baedd 52#include "locking.h"
581bb050 53#include "inode-map.h"
d7728c96 54#include "backref.h"
606686ee 55#include "rcu-string.h"
f46b5a66 56
6cbff00f
CH		 57/* Mask out flags that are inappropriate for the given type of inode. */
58static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
59{
60 if (S_ISDIR(mode))
61 return flags;
62 else if (S_ISREG(mode))
63 return flags & ~FS_DIRSYNC_FL;
64 else
65 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
66}
67
68/*
69 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
70 */
71static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
72{
73 unsigned int iflags = 0;
74
75 if (flags & BTRFS_INODE_SYNC)
76 iflags |= FS_SYNC_FL;
77 if (flags & BTRFS_INODE_IMMUTABLE)
78 iflags |= FS_IMMUTABLE_FL;
79 if (flags & BTRFS_INODE_APPEND)
80 iflags |= FS_APPEND_FL;
81 if (flags & BTRFS_INODE_NODUMP)
82 iflags |= FS_NODUMP_FL;
83 if (flags & BTRFS_INODE_NOATIME)
84 iflags |= FS_NOATIME_FL;
85 if (flags & BTRFS_INODE_DIRSYNC)
86 iflags |= FS_DIRSYNC_FL;
d0092bdd
LZ		 87 if (flags & BTRFS_INODE_NODATACOW)
88 iflags |= FS_NOCOW_FL;
89
90 if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
91 iflags |= FS_COMPR_FL;
92 else if (flags & BTRFS_INODE_NOCOMPRESS)
93 iflags |= FS_NOCOMP_FL;
6cbff00f
CH		 94
95 return iflags;
96}
97
98/*
99 * Update inode->i_flags based on the btrfs internal flags.
100 */
101void btrfs_update_iflags(struct inode *inode)
102{
103 struct btrfs_inode *ip = BTRFS_I(inode);
104
105 inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
106
107 if (ip->flags & BTRFS_INODE_SYNC)
108 inode->i_flags |= S_SYNC;
109 if (ip->flags & BTRFS_INODE_IMMUTABLE)
110 inode->i_flags |= S_IMMUTABLE;
111 if (ip->flags & BTRFS_INODE_APPEND)
112 inode->i_flags |= S_APPEND;
113 if (ip->flags & BTRFS_INODE_NOATIME)
114 inode->i_flags |= S_NOATIME;
115 if (ip->flags & BTRFS_INODE_DIRSYNC)
116 inode->i_flags |= S_DIRSYNC;
117}
118
119/*
120 * Inherit flags from the parent inode.
121 *
e27425d6 122 * Currently only the compression flags and the cow flags are inherited.
6cbff00f
CH		 123 */
124void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
125{
0b4dcea5
CM		 126 unsigned int flags;
127
128 if (!dir)
129 return;
130
131 flags = BTRFS_I(dir)->flags;
6cbff00f 132
e27425d6
JB		 133 if (flags & BTRFS_INODE_NOCOMPRESS) {
134 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
135 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
136 } else if (flags & BTRFS_INODE_COMPRESS) {
137 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
138 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
139 }
140
141 if (flags & BTRFS_INODE_NODATACOW)
142 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
6cbff00f 143
6cbff00f
CH		 144 btrfs_update_iflags(inode);
145}
146
147static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
148{
149 struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode);
150 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
151
152 if (copy_to_user(arg, &flags, sizeof(flags)))
153 return -EFAULT;
154 return 0;
155}
156
75e7cb7f
LB		 157static int check_flags(unsigned int flags)
158{
159 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
160 FS_NOATIME_FL | FS_NODUMP_FL | \
161 FS_SYNC_FL | FS_DIRSYNC_FL | \
e1e8fb6a
LZ		 162 FS_NOCOMP_FL | FS_COMPR_FL |
163 FS_NOCOW_FL))
75e7cb7f
LB		 164 return -EOPNOTSUPP;
165
166 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
167 return -EINVAL;
168
75e7cb7f
LB		 169 return 0;
170}
171
6cbff00f
CH		 172static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
173{
174 struct inode *inode = file->f_path.dentry->d_inode;
175 struct btrfs_inode *ip = BTRFS_I(inode);
176 struct btrfs_root *root = ip->root;
177 struct btrfs_trans_handle *trans;
178 unsigned int flags, oldflags;
179 int ret;
f062abf0
LZ		 180 u64 ip_oldflags;
181 unsigned int i_oldflags;
6cbff00f 182
b83cc969
LZ		 183 if (btrfs_root_readonly(root))
184 return -EROFS;
185
6cbff00f
CH		 186 if (copy_from_user(&flags, arg, sizeof(flags)))
187 return -EFAULT;
188
75e7cb7f
LB		 189 ret = check_flags(flags);
190 if (ret)
191 return ret;
f46b5a66 192
2e149670 193 if (!inode_owner_or_capable(inode))
6cbff00f
CH		 194 return -EACCES;
195
196 mutex_lock(&inode->i_mutex);
197
f062abf0
LZ		 198 ip_oldflags = ip->flags;
199 i_oldflags = inode->i_flags;
200
6cbff00f
CH		 201 flags = btrfs_mask_flags(inode->i_mode, flags);
202 oldflags = btrfs_flags_to_ioctl(ip->flags);
203 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
204 if (!capable(CAP_LINUX_IMMUTABLE)) {
205 ret = -EPERM;
206 goto out_unlock;
207 }
208 }
209
a561be71 210 ret = mnt_want_write_file(file);
6cbff00f
CH		 211 if (ret)
212 goto out_unlock;
213
214 if (flags & FS_SYNC_FL)
215 ip->flags |= BTRFS_INODE_SYNC;
216 else
217 ip->flags &= ~BTRFS_INODE_SYNC;
218 if (flags & FS_IMMUTABLE_FL)
219 ip->flags |= BTRFS_INODE_IMMUTABLE;
220 else
221 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
222 if (flags & FS_APPEND_FL)
223 ip->flags |= BTRFS_INODE_APPEND;
224 else
225 ip->flags &= ~BTRFS_INODE_APPEND;
226 if (flags & FS_NODUMP_FL)
227 ip->flags |= BTRFS_INODE_NODUMP;
228 else
229 ip->flags &= ~BTRFS_INODE_NODUMP;
230 if (flags & FS_NOATIME_FL)
231 ip->flags |= BTRFS_INODE_NOATIME;
232 else
233 ip->flags &= ~BTRFS_INODE_NOATIME;
234 if (flags & FS_DIRSYNC_FL)
235 ip->flags |= BTRFS_INODE_DIRSYNC;
236 else
237 ip->flags &= ~BTRFS_INODE_DIRSYNC;
e1e8fb6a
LZ		 238 if (flags & FS_NOCOW_FL)
239 ip->flags |= BTRFS_INODE_NODATACOW;
240 else
241 ip->flags &= ~BTRFS_INODE_NODATACOW;
6cbff00f 242
75e7cb7f
LB		 243 /*
244 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
245 * flag may be changed automatically if compression code won't make
246 * things smaller.
247 */
248 if (flags & FS_NOCOMP_FL) {
249 ip->flags &= ~BTRFS_INODE_COMPRESS;
250 ip->flags |= BTRFS_INODE_NOCOMPRESS;
251 } else if (flags & FS_COMPR_FL) {
252 ip->flags |= BTRFS_INODE_COMPRESS;
253 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
ebcb904d
LZ		 254 } else {
255 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
75e7cb7f 256 }
6cbff00f 257
4da6f1a3 258 trans = btrfs_start_transaction(root, 1);
f062abf0
LZ		 259 if (IS_ERR(trans)) {
260 ret = PTR_ERR(trans);
261 goto out_drop;
262 }
6cbff00f 263
306424cc 264 btrfs_update_iflags(inode);
0c4d2d95 265 inode_inc_iversion(inode);
306424cc 266 inode->i_ctime = CURRENT_TIME;
6cbff00f 267 ret = btrfs_update_inode(trans, root, inode);
6cbff00f 268
6cbff00f 269 btrfs_end_transaction(trans, root);
f062abf0
LZ		 270 out_drop:
271 if (ret) {
272 ip->flags = ip_oldflags;
273 inode->i_flags = i_oldflags;
274 }
6cbff00f 275
2a79f17e 276 mnt_drop_write_file(file);
6cbff00f
CH		 277 out_unlock:
278 mutex_unlock(&inode->i_mutex);
2d4e6f6a 279 return ret;
6cbff00f
CH		 280}
281
282static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
283{
284 struct inode *inode = file->f_path.dentry->d_inode;
285
286 return put_user(inode->i_generation, arg);
287}
f46b5a66 288
f7039b1d
LD		 289static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
290{
815745cf 291 struct btrfs_fs_info *fs_info = btrfs_sb(fdentry(file)->d_sb);
f7039b1d
LD		 292 struct btrfs_device *device;
293 struct request_queue *q;
294 struct fstrim_range range;
295 u64 minlen = ULLONG_MAX;
296 u64 num_devices = 0;
815745cf 297 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
f7039b1d
LD		 298 int ret;
299
300 if (!capable(CAP_SYS_ADMIN))
301 return -EPERM;
302
1f78160c
XG		 303 rcu_read_lock();
304 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
305 dev_list) {
f7039b1d
LD		 306 if (!device->bdev)
307 continue;
308 q = bdev_get_queue(device->bdev);
309 if (blk_queue_discard(q)) {
310 num_devices++;
311 minlen = min((u64)q->limits.discard_granularity,
312 minlen);
313 }
314 }
1f78160c 315 rcu_read_unlock();
f4c697e6 316
f7039b1d
LD		 317 if (!num_devices)
318 return -EOPNOTSUPP;
f7039b1d
LD		 319 if (copy_from_user(&range, arg, sizeof(range)))
320 return -EFAULT;
f4c697e6
LC		 321 if (range.start > total_bytes)
322 return -EINVAL;
f7039b1d 323
f4c697e6 324 range.len = min(range.len, total_bytes - range.start);
f7039b1d 325 range.minlen = max(range.minlen, minlen);
815745cf 326 ret = btrfs_trim_fs(fs_info->tree_root, &range);
f7039b1d
LD		 327 if (ret < 0)
328 return ret;
329
330 if (copy_to_user(arg, &range, sizeof(range)))
331 return -EFAULT;
332
333 return 0;
334}
335
cb8e7090
CH		 336static noinline int create_subvol(struct btrfs_root *root,
337 struct dentry *dentry,
72fd032e
SW		 338 char *name, int namelen,
339 u64 *async_transid)
f46b5a66
CH		 340{
341 struct btrfs_trans_handle *trans;
342 struct btrfs_key key;
343 struct btrfs_root_item root_item;
344 struct btrfs_inode_item *inode_item;
345 struct extent_buffer *leaf;
76dda93c 346 struct btrfs_root *new_root;
2fbe8c8a 347 struct dentry *parent = dentry->d_parent;
6a912213 348 struct inode *dir;
f46b5a66
CH		 349 int ret;
350 int err;
351 u64 objectid;
352 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
3de4586c 353 u64 index = 0;
f46b5a66 354
581bb050 355 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
2fbe8c8a 356 if (ret)
a22285a6 357 return ret;
6a912213
JB		 358
359 dir = parent->d_inode;
360
9ed74f2d
JB		 361 /*
362 * 1 - inode item
363 * 2 - refs
364 * 1 - root item
365 * 2 - dir items
366 */
a22285a6 367 trans = btrfs_start_transaction(root, 6);
2fbe8c8a 368 if (IS_ERR(trans))
a22285a6 369 return PTR_ERR(trans);
f46b5a66 370
5d4f98a2 371 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
5581a51a 372 0, objectid, NULL, 0, 0, 0);
8e8a1e31
JB		 373 if (IS_ERR(leaf)) {
374 ret = PTR_ERR(leaf);
375 goto fail;
376 }
f46b5a66 377
5d4f98a2 378 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
f46b5a66
CH		 379 btrfs_set_header_bytenr(leaf, leaf->start);
380 btrfs_set_header_generation(leaf, trans->transid);
5d4f98a2 381 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
f46b5a66
CH		 382 btrfs_set_header_owner(leaf, objectid);
383
384 write_extent_buffer(leaf, root->fs_info->fsid,
385 (unsigned long)btrfs_header_fsid(leaf),
386 BTRFS_FSID_SIZE);
5d4f98a2
YZ		 387 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
388 (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
389 BTRFS_UUID_SIZE);
f46b5a66
CH		 390 btrfs_mark_buffer_dirty(leaf);
391
392 inode_item = &root_item.inode;
393 memset(inode_item, 0, sizeof(*inode_item));
394 inode_item->generation = cpu_to_le64(1);
395 inode_item->size = cpu_to_le64(3);
396 inode_item->nlink = cpu_to_le32(1);
a76a3cd4 397 inode_item->nbytes = cpu_to_le64(root->leafsize);
f46b5a66
CH		 398 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
399
08fe4db1
LZ		 400 root_item.flags = 0;
401 root_item.byte_limit = 0;
402 inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
403
f46b5a66 404 btrfs_set_root_bytenr(&root_item, leaf->start);
84234f3a 405 btrfs_set_root_generation(&root_item, trans->transid);
f46b5a66
CH		 406 btrfs_set_root_level(&root_item, 0);
407 btrfs_set_root_refs(&root_item, 1);
86b9f2ec 408 btrfs_set_root_used(&root_item, leaf->len);
80ff3856 409 btrfs_set_root_last_snapshot(&root_item, 0);
f46b5a66
CH		 410
411 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
412 root_item.drop_level = 0;
413
925baedd 414 btrfs_tree_unlock(leaf);
f46b5a66
CH		 415 free_extent_buffer(leaf);
416 leaf = NULL;
417
418 btrfs_set_root_dirid(&root_item, new_dirid);
419
420 key.objectid = objectid;
5d4f98a2 421 key.offset = 0;
f46b5a66
CH		 422 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
423 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
424 &root_item);
425 if (ret)
426 goto fail;
427
76dda93c
YZ		 428 key.offset = (u64)-1;
429 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
79787eaa
JM		 430 if (IS_ERR(new_root)) {
431 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
432 ret = PTR_ERR(new_root);
433 goto fail;
434 }
76dda93c
YZ		 435
436 btrfs_record_root_in_trans(trans, new_root);
437
d82a6f1d 438 ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
ce598979
MF		 439 if (ret) {
440 /* We potentially lose an unused inode item here */
79787eaa 441 btrfs_abort_transaction(trans, root, ret);
ce598979
MF		 442 goto fail;
443 }
444
f46b5a66
CH		 445 /*
446 * insert the directory item
447 */
3de4586c 448 ret = btrfs_set_inode_index(dir, &index);
79787eaa
JM		 449 if (ret) {
450 btrfs_abort_transaction(trans, root, ret);
451 goto fail;
452 }
3de4586c
CM		 453
454 ret = btrfs_insert_dir_item(trans, root,
16cdcec7 455 name, namelen, dir, &key,
3de4586c 456 BTRFS_FT_DIR, index);
79787eaa
JM		 457 if (ret) {
458 btrfs_abort_transaction(trans, root, ret);
f46b5a66 459 goto fail;
79787eaa 460 }
0660b5af 461
52c26179
YZ		 462 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
463 ret = btrfs_update_inode(trans, root, dir);
464 BUG_ON(ret);
465
0660b5af 466 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
4df27c4d 467 objectid, root->root_key.objectid,
33345d01 468 btrfs_ino(dir), index, name, namelen);
0660b5af 469
76dda93c 470 BUG_ON(ret);
f46b5a66 471
76dda93c 472 d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
f46b5a66 473fail:
72fd032e
SW		 474 if (async_transid) {
475 *async_transid = trans->transid;
476 err = btrfs_commit_transaction_async(trans, root, 1);
477 } else {
478 err = btrfs_commit_transaction(trans, root);
479 }
f46b5a66
CH		 480 if (err && !ret)
481 ret = err;
f46b5a66
CH		 482 return ret;
483}
484
72fd032e 485static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
b83cc969
LZ		 486 char *name, int namelen, u64 *async_transid,
487 bool readonly)
f46b5a66 488{
2e4bfab9 489 struct inode *inode;
f46b5a66
CH		 490 struct btrfs_pending_snapshot *pending_snapshot;
491 struct btrfs_trans_handle *trans;
2e4bfab9 492 int ret;
f46b5a66
CH		 493
494 if (!root->ref_cows)
495 return -EINVAL;
496
3de4586c 497 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
a22285a6
YZ		 498 if (!pending_snapshot)
499 return -ENOMEM;
500
501 btrfs_init_block_rsv(&pending_snapshot->block_rsv);
3de4586c 502 pending_snapshot->dentry = dentry;
f46b5a66 503 pending_snapshot->root = root;
b83cc969 504 pending_snapshot->readonly = readonly;
a22285a6
YZ		 505
506 trans = btrfs_start_transaction(root->fs_info->extent_root, 5);
507 if (IS_ERR(trans)) {
508 ret = PTR_ERR(trans);
509 goto fail;
510 }
511
512 ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
513 BUG_ON(ret);
514
8351583e 515 spin_lock(&root->fs_info->trans_lock);
f46b5a66
CH		 516 list_add(&pending_snapshot->list,
517 &trans->transaction->pending_snapshots);
8351583e 518 spin_unlock(&root->fs_info->trans_lock);
72fd032e
SW		 519 if (async_transid) {
520 *async_transid = trans->transid;
521 ret = btrfs_commit_transaction_async(trans,
522 root->fs_info->extent_root, 1);
523 } else {
524 ret = btrfs_commit_transaction(trans,
525 root->fs_info->extent_root);
526 }
2e4bfab9 527 BUG_ON(ret);
a22285a6
YZ		 528
529 ret = pending_snapshot->error;
530 if (ret)
531 goto fail;
532
66b4ffd1
JB		 533 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
534 if (ret)
535 goto fail;
f46b5a66 536
2fbe8c8a 537 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
2e4bfab9
YZ		 538 if (IS_ERR(inode)) {
539 ret = PTR_ERR(inode);
540 goto fail;
541 }
542 BUG_ON(!inode);
543 d_instantiate(dentry, inode);
544 ret = 0;
545fail:
a22285a6 546 kfree(pending_snapshot);
f46b5a66
CH		 547 return ret;
548}
549
4260f7c7
SW		 550/* copy of check_sticky in fs/namei.c()
551* It's inline, so penalty for filesystems that don't use sticky bit is
552* minimal.
553*/
554static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
555{
556 uid_t fsuid = current_fsuid();
557
558 if (!(dir->i_mode & S_ISVTX))
559 return 0;
560 if (inode->i_uid == fsuid)
561 return 0;
562 if (dir->i_uid == fsuid)
563 return 0;
564 return !capable(CAP_FOWNER);
565}
566
567/* copy of may_delete in fs/namei.c()
568 * Check whether we can remove a link victim from directory dir, check
569 * whether the type of victim is right.
570 * 1. We can't do it if dir is read-only (done in permission())
571 * 2. We should have write and exec permissions on dir
572 * 3. We can't remove anything from append-only dir
573 * 4. We can't do anything with immutable dir (done in permission())
574 * 5. If the sticky bit on dir is set we should either
575 * a. be owner of dir, or
576 * b. be owner of victim, or
577 * c. have CAP_FOWNER capability
578 * 6. If the victim is append-only or immutable we can't do antyhing with
579 * links pointing to it.
580 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
581 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
582 * 9. We can't remove a root or mountpoint.
583 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
584 * nfs_async_unlink().
585 */
586
587static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
588{
589 int error;
590
591 if (!victim->d_inode)
592 return -ENOENT;
593
594 BUG_ON(victim->d_parent->d_inode != dir);
595 audit_inode_child(victim, dir);
596
597 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
598 if (error)
599 return error;
600 if (IS_APPEND(dir))
601 return -EPERM;
602 if (btrfs_check_sticky(dir, victim->d_inode)||
603 IS_APPEND(victim->d_inode)||
604 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
605 return -EPERM;
606 if (isdir) {
607 if (!S_ISDIR(victim->d_inode->i_mode))
608 return -ENOTDIR;
609 if (IS_ROOT(victim))
610 return -EBUSY;
611 } else if (S_ISDIR(victim->d_inode->i_mode))
612 return -EISDIR;
613 if (IS_DEADDIR(dir))
614 return -ENOENT;
615 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
616 return -EBUSY;
617 return 0;
618}
619
cb8e7090
CH		 620/* copy of may_create in fs/namei.c() */
621static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
622{
623 if (child->d_inode)
624 return -EEXIST;
625 if (IS_DEADDIR(dir))
626 return -ENOENT;
627 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
628}
629
630/*
631 * Create a new subvolume below @parent. This is largely modeled after
632 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
633 * inside this filesystem so it's quite a bit simpler.
634 */
76dda93c
YZ		 635static noinline int btrfs_mksubvol(struct path *parent,
636 char *name, int namelen,
72fd032e 637 struct btrfs_root *snap_src,
b83cc969 638 u64 *async_transid, bool readonly)
cb8e7090 639{
76dda93c 640 struct inode *dir = parent->dentry->d_inode;
cb8e7090
CH		 641 struct dentry *dentry;
642 int error;
643
76dda93c 644 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
cb8e7090
CH		 645
646 dentry = lookup_one_len(name, parent->dentry, namelen);
647 error = PTR_ERR(dentry);
648 if (IS_ERR(dentry))
649 goto out_unlock;
650
651 error = -EEXIST;
652 if (dentry->d_inode)
653 goto out_dput;
654
cb8e7090
CH		 655 error = mnt_want_write(parent->mnt);
656 if (error)
657 goto out_dput;
658
76dda93c 659 error = btrfs_may_create(dir, dentry);
cb8e7090
CH		 660 if (error)
661 goto out_drop_write;
662
76dda93c
YZ		 663 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
664
665 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
666 goto out_up_read;
667
3de4586c 668 if (snap_src) {
72fd032e 669 error = create_snapshot(snap_src, dentry,
b83cc969 670 name, namelen, async_transid, readonly);
3de4586c 671 } else {
76dda93c 672 error = create_subvol(BTRFS_I(dir)->root, dentry,
72fd032e 673 name, namelen, async_transid);
3de4586c 674 }
76dda93c
YZ		 675 if (!error)
676 fsnotify_mkdir(dir, dentry);
677out_up_read:
678 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
cb8e7090
CH		 679out_drop_write:
680 mnt_drop_write(parent->mnt);
681out_dput:
682 dput(dentry);
683out_unlock:
76dda93c 684 mutex_unlock(&dir->i_mutex);
cb8e7090
CH		 685 return error;
686}
687
4cb5300b
CM		 688/*
689 * When we're defragging a range, we don't want to kick it off again
690 * if it is really just waiting for delalloc to send it down.
691 * If we find a nice big extent or delalloc range for the bytes in the
692 * file you want to defrag, we return 0 to let you know to skip this
693 * part of the file
694 */
695static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
696{
697 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
698 struct extent_map *em = NULL;
699 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
700 u64 end;
701
702 read_lock(&em_tree->lock);
703 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
704 read_unlock(&em_tree->lock);
705
706 if (em) {
707 end = extent_map_end(em);
708 free_extent_map(em);
709 if (end - offset > thresh)
710 return 0;
711 }
712 /* if we already have a nice delalloc here, just stop */
713 thresh /= 2;
714 end = count_range_bits(io_tree, &offset, offset + thresh,
715 thresh, EXTENT_DELALLOC, 1);
716 if (end >= thresh)
717 return 0;
718 return 1;
719}
720
721/*
722 * helper function to walk through a file and find extents
723 * newer than a specific transid, and smaller than thresh.
724 *
725 * This is used by the defragging code to find new and small
726 * extents
727 */
728static int find_new_extents(struct btrfs_root *root,
729 struct inode *inode, u64 newer_than,
730 u64 *off, int thresh)
731{
732 struct btrfs_path *path;
733 struct btrfs_key min_key;
734 struct btrfs_key max_key;
735 struct extent_buffer *leaf;
736 struct btrfs_file_extent_item *extent;
737 int type;
738 int ret;
a4689d2b 739 u64 ino = btrfs_ino(inode);
4cb5300b
CM		 740
741 path = btrfs_alloc_path();
742 if (!path)
743 return -ENOMEM;
744
a4689d2b 745 min_key.objectid = ino;
4cb5300b
CM		 746 min_key.type = BTRFS_EXTENT_DATA_KEY;
747 min_key.offset = *off;
748
a4689d2b 749 max_key.objectid = ino;
4cb5300b
CM		 750 max_key.type = (u8)-1;
751 max_key.offset = (u64)-1;
752
753 path->keep_locks = 1;
754
755 while(1) {
756 ret = btrfs_search_forward(root, &min_key, &max_key,
757 path, 0, newer_than);
758 if (ret != 0)
759 goto none;
a4689d2b 760 if (min_key.objectid != ino)
4cb5300b
CM		 761 goto none;
762 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
763 goto none;
764
765 leaf = path->nodes[0];
766 extent = btrfs_item_ptr(leaf, path->slots[0],
767 struct btrfs_file_extent_item);
768
769 type = btrfs_file_extent_type(leaf, extent);
770 if (type == BTRFS_FILE_EXTENT_REG &&
771 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
772 check_defrag_in_cache(inode, min_key.offset, thresh)) {
773 *off = min_key.offset;
774 btrfs_free_path(path);
775 return 0;
776 }
777
778 if (min_key.offset == (u64)-1)
779 goto none;
780
781 min_key.offset++;
782 btrfs_release_path(path);
783 }
784none:
785 btrfs_free_path(path);
786 return -ENOENT;
787}
788
6c282eb4 789static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
17ce6ef8
LB		 790{
791 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
6c282eb4
LZ		 792 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
793 struct extent_map *em;
794 u64 len = PAGE_CACHE_SIZE;
17ce6ef8 795
6c282eb4
LZ		 796 /*
797 * hopefully we have this extent in the tree already, try without
798 * the full extent lock
799 */
17ce6ef8 800 read_lock(&em_tree->lock);
6c282eb4 801 em = lookup_extent_mapping(em_tree, start, len);
17ce6ef8
LB		 802 read_unlock(&em_tree->lock);
803
6c282eb4
LZ		 804 if (!em) {
805 /* get the big lock and read metadata off disk */
806 lock_extent(io_tree, start, start + len - 1);
807 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
808 unlock_extent(io_tree, start, start + len - 1);
809
810 if (IS_ERR(em))
811 return NULL;
812 }
813
814 return em;
815}
17ce6ef8 816
6c282eb4
LZ		 817static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
818{
819 struct extent_map *next;
820 bool ret = true;
821
822 /* this is the last extent */
823 if (em->start + em->len >= i_size_read(inode))
824 return false;
825
826 next = defrag_lookup_extent(inode, em->start + em->len);
827 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
828 ret = false;
829
830 free_extent_map(next);
17ce6ef8
LB		 831 return ret;
832}
833
6c282eb4 834static int should_defrag_range(struct inode *inode, u64 start, int thresh,
a43a2111
AM		 835 u64 *last_len, u64 *skip, u64 *defrag_end,
836 int compress)
940100a4 837{
6c282eb4 838 struct extent_map *em;
940100a4 839 int ret = 1;
6c282eb4 840 bool next_mergeable = true;
940100a4
CM		 841
842 /*
008873ea 843 * make sure that once we start defragging an extent, we keep on
940100a4
CM		 844 * defragging it
845 */
846 if (start < *defrag_end)
847 return 1;
848
849 *skip = 0;
850
6c282eb4
LZ		 851 em = defrag_lookup_extent(inode, start);
852 if (!em)
853 return 0;
940100a4
CM		 854
855 /* this will cover holes, and inline extents */
17ce6ef8 856 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
940100a4 857 ret = 0;
17ce6ef8
LB		 858 goto out;
859 }
860
6c282eb4 861 next_mergeable = defrag_check_next_extent(inode, em);
940100a4
CM		 862
863 /*
6c282eb4
LZ		 864 * we hit a real extent, if it is big or the next extent is not a
865 * real extent, don't bother defragging it
940100a4 866 */
a43a2111 867 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
6c282eb4 868 (em->len >= thresh || !next_mergeable))
940100a4 869 ret = 0;
17ce6ef8 870out:
940100a4
CM		 871 /*
872 * last_len ends up being a counter of how many bytes we've defragged.
873 * every time we choose not to defrag an extent, we reset *last_len
874 * so that the next tiny extent will force a defrag.
875 *
876 * The end result of this is that tiny extents before a single big
877 * extent will force at least part of that big extent to be defragged.
878 */
879 if (ret) {
940100a4
CM		 880 *defrag_end = extent_map_end(em);
881 } else {
882 *last_len = 0;
883 *skip = extent_map_end(em);
884 *defrag_end = 0;
885 }
886
887 free_extent_map(em);
888 return ret;
889}
890
4cb5300b
CM		 891/*
892 * it doesn't do much good to defrag one or two pages
893 * at a time. This pulls in a nice chunk of pages
894 * to COW and defrag.
895 *
896 * It also makes sure the delalloc code has enough
897 * dirty data to avoid making new small extents as part
898 * of the defrag
899 *
900 * It's a good idea to start RA on this range
901 * before calling this.
902 */
903static int cluster_pages_for_defrag(struct inode *inode,
904 struct page **pages,
905 unsigned long start_index,
906 int num_pages)
f46b5a66 907{
4cb5300b
CM		 908 unsigned long file_end;
909 u64 isize = i_size_read(inode);
910 u64 page_start;
911 u64 page_end;
1f12bd06 912 u64 page_cnt;
4cb5300b
CM		 913 int ret;
914 int i;
915 int i_done;
3eaa2885 916 struct btrfs_ordered_extent *ordered;
4cb5300b 917 struct extent_state *cached_state = NULL;
600a45e1 918 struct extent_io_tree *tree;
3b16a4e3 919 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
4cb5300b 920
4cb5300b 921 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1f12bd06
LB		 922 if (!isize || start_index > file_end)
923 return 0;
924
925 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
4cb5300b
CM		 926
927 ret = btrfs_delalloc_reserve_space(inode,
1f12bd06 928 page_cnt << PAGE_CACHE_SHIFT);
4cb5300b
CM		 929 if (ret)
930 return ret;
4cb5300b 931 i_done = 0;
600a45e1 932 tree = &BTRFS_I(inode)->io_tree;
4cb5300b
CM		 933
934 /* step one, lock all the pages */
1f12bd06 935 for (i = 0; i < page_cnt; i++) {
4cb5300b 936 struct page *page;
600a45e1 937again:
a94733d0 938 page = find_or_create_page(inode->i_mapping,
600a45e1 939 start_index + i, mask);
4cb5300b
CM		 940 if (!page)
941 break;
942
600a45e1
MX		 943 page_start = page_offset(page);
944 page_end = page_start + PAGE_CACHE_SIZE - 1;
945 while (1) {
d0082371 946 lock_extent(tree, page_start, page_end);
600a45e1
MX		 947 ordered = btrfs_lookup_ordered_extent(inode,
948 page_start);
d0082371 949 unlock_extent(tree, page_start, page_end);
600a45e1
MX		 950 if (!ordered)
951 break;
952
953 unlock_page(page);
954 btrfs_start_ordered_extent(inode, ordered, 1);
955 btrfs_put_ordered_extent(ordered);
956 lock_page(page);
1f12bd06
LB		 957 /*
958 * we unlocked the page above, so we need check if
959 * it was released or not.
960 */
961 if (page->mapping != inode->i_mapping) {
962 unlock_page(page);
963 page_cache_release(page);
964 goto again;
965 }
600a45e1
MX		 966 }
967
4cb5300b
CM		 968 if (!PageUptodate(page)) {
969 btrfs_readpage(NULL, page);
970 lock_page(page);
971 if (!PageUptodate(page)) {
972 unlock_page(page);
973 page_cache_release(page);
974 ret = -EIO;
975 break;
976 }
977 }
600a45e1 978
600a45e1
MX		 979 if (page->mapping != inode->i_mapping) {
980 unlock_page(page);
981 page_cache_release(page);
982 goto again;
983 }
984
4cb5300b
CM		 985 pages[i] = page;
986 i_done++;
987 }
988 if (!i_done || ret)
989 goto out;
990
991 if (!(inode->i_sb->s_flags & MS_ACTIVE))
992 goto out;
993
994 /*
995 * so now we have a nice long stream of locked
996 * and up to date pages, lets wait on them
997 */
998 for (i = 0; i < i_done; i++)
999 wait_on_page_writeback(pages[i]);
1000
1001 page_start = page_offset(pages[0]);
1002 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1003
1004 lock_extent_bits(&BTRFS_I(inode)->io_tree,
d0082371 1005 page_start, page_end - 1, 0, &cached_state);
4cb5300b
CM		 1006 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1007 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
1008 EXTENT_DO_ACCOUNTING, 0, 0, &cached_state,
1009 GFP_NOFS);
1010
1f12bd06 1011 if (i_done != page_cnt) {
9e0baf60
JB		 1012 spin_lock(&BTRFS_I(inode)->lock);
1013 BTRFS_I(inode)->outstanding_extents++;
1014 spin_unlock(&BTRFS_I(inode)->lock);
4cb5300b 1015 btrfs_delalloc_release_space(inode,
1f12bd06 1016 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1017 }
1018
1019
1020 btrfs_set_extent_delalloc(inode, page_start, page_end - 1,
1021 &cached_state);
1022
1023 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1024 page_start, page_end - 1, &cached_state,
1025 GFP_NOFS);
1026
1027 for (i = 0; i < i_done; i++) {
1028 clear_page_dirty_for_io(pages[i]);
1029 ClearPageChecked(pages[i]);
1030 set_page_extent_mapped(pages[i]);
1031 set_page_dirty(pages[i]);
1032 unlock_page(pages[i]);
1033 page_cache_release(pages[i]);
1034 }
1035 return i_done;
1036out:
1037 for (i = 0; i < i_done; i++) {
1038 unlock_page(pages[i]);
1039 page_cache_release(pages[i]);
1040 }
1f12bd06 1041 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1042 return ret;
1043
1044}
1045
1046int btrfs_defrag_file(struct inode *inode, struct file *file,
1047 struct btrfs_ioctl_defrag_range_args *range,
1048 u64 newer_than, unsigned long max_to_defrag)
1049{
1050 struct btrfs_root *root = BTRFS_I(inode)->root;
1a419d85 1051 struct btrfs_super_block *disk_super;
4cb5300b 1052 struct file_ra_state *ra = NULL;
f46b5a66 1053 unsigned long last_index;
151a31b2 1054 u64 isize = i_size_read(inode);
1a419d85 1055 u64 features;
940100a4
CM		 1056 u64 last_len = 0;
1057 u64 skip = 0;
1058 u64 defrag_end = 0;
4cb5300b 1059 u64 newer_off = range->start;
f46b5a66 1060 unsigned long i;
008873ea 1061 unsigned long ra_index = 0;
f46b5a66 1062 int ret;
4cb5300b 1063 int defrag_count = 0;
1a419d85 1064 int compress_type = BTRFS_COMPRESS_ZLIB;
4cb5300b 1065 int extent_thresh = range->extent_thresh;
008873ea
LZ		 1066 int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1067 int cluster = max_cluster;
4cb5300b
CM		 1068 u64 new_align = ~((u64)128 * 1024 - 1);
1069 struct page **pages = NULL;
1070
1071 if (extent_thresh == 0)
1072 extent_thresh = 256 * 1024;
1a419d85
LZ		 1073
1074 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1075 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1076 return -EINVAL;
1077 if (range->compress_type)
1078 compress_type = range->compress_type;
1079 }
f46b5a66 1080
151a31b2 1081 if (isize == 0)
940100a4
CM		 1082 return 0;
1083
4cb5300b
CM		 1084 /*
1085 * if we were not given a file, allocate a readahead
1086 * context
1087 */
1088 if (!file) {
1089 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1090 if (!ra)
1091 return -ENOMEM;
1092 file_ra_state_init(ra, inode->i_mapping);
1093 } else {
1094 ra = &file->f_ra;
1095 }
1096
008873ea 1097 pages = kmalloc(sizeof(struct page *) * max_cluster,
4cb5300b
CM		 1098 GFP_NOFS);
1099 if (!pages) {
1100 ret = -ENOMEM;
1101 goto out_ra;
1102 }
1103
1104 /* find the last page to defrag */
1e701a32 1105 if (range->start + range->len > range->start) {
151a31b2 1106 last_index = min_t(u64, isize - 1,
1e701a32
CM		 1107 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1108 } else {
151a31b2 1109 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1e701a32
CM		 1110 }
1111
4cb5300b
CM		 1112 if (newer_than) {
1113 ret = find_new_extents(root, inode, newer_than,
1114 &newer_off, 64 * 1024);
1115 if (!ret) {
1116 range->start = newer_off;
1117 /*
1118 * we always align our defrag to help keep
1119 * the extents in the file evenly spaced
1120 */
1121 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
4cb5300b
CM		 1122 } else
1123 goto out_ra;
1124 } else {
1125 i = range->start >> PAGE_CACHE_SHIFT;
1126 }
1127 if (!max_to_defrag)
7ec31b54 1128 max_to_defrag = last_index + 1;
4cb5300b 1129
2a0f7f57
LZ		 1130 /*
1131 * make writeback starts from i, so the defrag range can be
1132 * written sequentially.
1133 */
1134 if (i < inode->i_mapping->writeback_index)
1135 inode->i_mapping->writeback_index = i;
1136
f7f43cc8
CM		 1137 while (i <= last_index && defrag_count < max_to_defrag &&
1138 (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1139 PAGE_CACHE_SHIFT)) {
4cb5300b
CM		 1140 /*
1141 * make sure we stop running if someone unmounts
1142 * the FS
1143 */
1144 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1145 break;
1146
66c26892 1147 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
6c282eb4 1148 extent_thresh, &last_len, &skip,
a43a2111
AM		 1149 &defrag_end, range->flags &
1150 BTRFS_DEFRAG_RANGE_COMPRESS)) {
940100a4
CM		 1151 unsigned long next;
1152 /*
1153 * the should_defrag function tells us how much to skip
1154 * bump our counter by the suggested amount
1155 */
1156 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1157 i = max(i + 1, next);
1158 continue;
1159 }
008873ea
LZ		 1160
1161 if (!newer_than) {
1162 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1163 PAGE_CACHE_SHIFT) - i;
1164 cluster = min(cluster, max_cluster);
1165 } else {
1166 cluster = max_cluster;
1167 }
1168
1e701a32 1169 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1a419d85 1170 BTRFS_I(inode)->force_compress = compress_type;
940100a4 1171
008873ea
LZ		 1172 if (i + cluster > ra_index) {
1173 ra_index = max(i, ra_index);
1174 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1175 cluster);
1176 ra_index += max_cluster;
1177 }
940100a4 1178
ecb8bea8 1179 mutex_lock(&inode->i_mutex);
008873ea 1180 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
ecb8bea8
LB		 1181 if (ret < 0) {
1182 mutex_unlock(&inode->i_mutex);
4cb5300b 1183 goto out_ra;
ecb8bea8 1184 }
4cb5300b
CM		 1185
1186 defrag_count += ret;
1187 balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret);
ecb8bea8 1188 mutex_unlock(&inode->i_mutex);
4cb5300b
CM		 1189
1190 if (newer_than) {
1191 if (newer_off == (u64)-1)
1192 break;
1193
e1f041e1
LB		 1194 if (ret > 0)
1195 i += ret;
1196
4cb5300b
CM		 1197 newer_off = max(newer_off + 1,
1198 (u64)i << PAGE_CACHE_SHIFT);
1199
1200 ret = find_new_extents(root, inode,
1201 newer_than, &newer_off,
1202 64 * 1024);
1203 if (!ret) {
1204 range->start = newer_off;
1205 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
4cb5300b
CM		 1206 } else {
1207 break;
f46b5a66 1208 }
4cb5300b 1209 } else {
008873ea 1210 if (ret > 0) {
cbcc8326 1211 i += ret;
008873ea
LZ		 1212 last_len += ret << PAGE_CACHE_SHIFT;
1213 } else {
cbcc8326 1214 i++;
008873ea
LZ		 1215 last_len = 0;
1216 }
f46b5a66 1217 }
f46b5a66
CH		 1218 }
1219
1e701a32
CM		 1220 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1221 filemap_flush(inode->i_mapping);
1222
1223 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1224 /* the filemap_flush will queue IO into the worker threads, but
1225 * we have to make sure the IO is actually started and that
1226 * ordered extents get created before we return
1227 */
1228 atomic_inc(&root->fs_info->async_submit_draining);
1229 while (atomic_read(&root->fs_info->nr_async_submits) ||
1230 atomic_read(&root->fs_info->async_delalloc_pages)) {
1231 wait_event(root->fs_info->async_submit_wait,
1232 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1233 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1234 }
1235 atomic_dec(&root->fs_info->async_submit_draining);
1236
1237 mutex_lock(&inode->i_mutex);
261507a0 1238 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1e701a32
CM		 1239 mutex_unlock(&inode->i_mutex);
1240 }
1241
6c41761f 1242 disk_super = root->fs_info->super_copy;
1a419d85
LZ		 1243 features = btrfs_super_incompat_flags(disk_super);
1244 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1245 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
1246 btrfs_set_super_incompat_flags(disk_super, features);
1247 }
1248
60ccf82f 1249 ret = defrag_count;
940100a4 1250
4cb5300b
CM		 1251out_ra:
1252 if (!file)
1253 kfree(ra);
1254 kfree(pages);
940100a4 1255 return ret;
f46b5a66
CH		 1256}
1257
76dda93c
YZ		 1258static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
1259 void __user *arg)
f46b5a66
CH		 1260{
1261 u64 new_size;
1262 u64 old_size;
1263 u64 devid = 1;
1264 struct btrfs_ioctl_vol_args *vol_args;
1265 struct btrfs_trans_handle *trans;
1266 struct btrfs_device *device = NULL;
1267 char *sizestr;
1268 char *devstr = NULL;
1269 int ret = 0;
f46b5a66
CH		 1270 int mod = 0;
1271
c146afad
YZ		 1272 if (root->fs_info->sb->s_flags & MS_RDONLY)
1273 return -EROFS;
1274
e441d54d
CM		 1275 if (!capable(CAP_SYS_ADMIN))
1276 return -EPERM;
1277
c9e9f97b
ID		 1278 mutex_lock(&root->fs_info->volume_mutex);
1279 if (root->fs_info->balance_ctl) {
1280 printk(KERN_INFO "btrfs: balance in progress\n");
1281 ret = -EINVAL;
1282 goto out;
1283 }
1284
dae7b665 1285 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 1286 if (IS_ERR(vol_args)) {
1287 ret = PTR_ERR(vol_args);
1288 goto out;
1289 }
5516e595
MF		 1290
1291 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 1292
f46b5a66
CH		 1293 sizestr = vol_args->name;
1294 devstr = strchr(sizestr, ':');
1295 if (devstr) {
1296 char *end;
1297 sizestr = devstr + 1;
1298 *devstr = '\0';
1299 devstr = vol_args->name;
1300 devid = simple_strtoull(devstr, &end, 10);
5bb14682 1301 printk(KERN_INFO "btrfs: resizing devid %llu\n",
21380931 1302 (unsigned long long)devid);
f46b5a66 1303 }
2b82032c 1304 device = btrfs_find_device(root, devid, NULL, NULL);
f46b5a66 1305 if (!device) {
5bb14682 1306 printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
21380931 1307 (unsigned long long)devid);
f46b5a66 1308 ret = -EINVAL;
c9e9f97b 1309 goto out_free;
f46b5a66 1310 }
4e42ae1b
LB		 1311 if (device->fs_devices && device->fs_devices->seeding) {
1312 printk(KERN_INFO "btrfs: resizer unable to apply on "
a8c4a33b
CM		 1313 "seeding device %llu\n",
1314 (unsigned long long)devid);
4e42ae1b
LB		 1315 ret = -EINVAL;
1316 goto out_free;
1317 }
1318
f46b5a66
CH		 1319 if (!strcmp(sizestr, "max"))
1320 new_size = device->bdev->bd_inode->i_size;
1321 else {
1322 if (sizestr[0] == '-') {
1323 mod = -1;
1324 sizestr++;
1325 } else if (sizestr[0] == '+') {
1326 mod = 1;
1327 sizestr++;
1328 }
91748467 1329 new_size = memparse(sizestr, NULL);
f46b5a66
CH		 1330 if (new_size == 0) {
1331 ret = -EINVAL;
c9e9f97b 1332 goto out_free;
f46b5a66
CH		 1333 }
1334 }
1335
1336 old_size = device->total_bytes;
1337
1338 if (mod < 0) {
1339 if (new_size > old_size) {
1340 ret = -EINVAL;
c9e9f97b 1341 goto out_free;
f46b5a66
CH		 1342 }
1343 new_size = old_size - new_size;
1344 } else if (mod > 0) {
1345 new_size = old_size + new_size;
1346 }
1347
1348 if (new_size < 256 * 1024 * 1024) {
1349 ret = -EINVAL;
c9e9f97b 1350 goto out_free;
f46b5a66
CH		 1351 }
1352 if (new_size > device->bdev->bd_inode->i_size) {
1353 ret = -EFBIG;
c9e9f97b 1354 goto out_free;
f46b5a66
CH		 1355 }
1356
1357 do_div(new_size, root->sectorsize);
1358 new_size *= root->sectorsize;
1359
606686ee
JB		 1360 printk_in_rcu(KERN_INFO "btrfs: new size for %s is %llu\n",
1361 rcu_str_deref(device->name),
1362 (unsigned long long)new_size);
f46b5a66
CH		 1363
1364 if (new_size > old_size) {
a22285a6 1365 trans = btrfs_start_transaction(root, 0);
98d5dc13
TI		 1366 if (IS_ERR(trans)) {
1367 ret = PTR_ERR(trans);
c9e9f97b 1368 goto out_free;
98d5dc13 1369 }
f46b5a66
CH		 1370 ret = btrfs_grow_device(trans, device, new_size);
1371 btrfs_commit_transaction(trans, root);
ece7d20e 1372 } else if (new_size < old_size) {
f46b5a66
CH		 1373 ret = btrfs_shrink_device(device, new_size);
1374 }
1375
c9e9f97b 1376out_free:
f46b5a66 1377 kfree(vol_args);
c9e9f97b
ID		 1378out:
1379 mutex_unlock(&root->fs_info->volume_mutex);
f46b5a66
CH		 1380 return ret;
1381}
1382
72fd032e
SW		 1383static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1384 char *name,
1385 unsigned long fd,
1386 int subvol,
b83cc969
LZ		 1387 u64 *transid,
1388 bool readonly)
f46b5a66 1389{
cb8e7090 1390 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3de4586c 1391 struct file *src_file;
f46b5a66 1392 int namelen;
3de4586c 1393 int ret = 0;
f46b5a66 1394
c146afad
YZ		 1395 if (root->fs_info->sb->s_flags & MS_RDONLY)
1396 return -EROFS;
1397
72fd032e
SW		 1398 namelen = strlen(name);
1399 if (strchr(name, '/')) {
f46b5a66
CH		 1400 ret = -EINVAL;
1401 goto out;
1402 }
1403
16780cab
CM		 1404 if (name[0] == '.' &&
1405 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1406 ret = -EEXIST;
1407 goto out;
1408 }
1409
3de4586c 1410 if (subvol) {
72fd032e 1411 ret = btrfs_mksubvol(&file->f_path, name, namelen,
b83cc969 1412 NULL, transid, readonly);
cb8e7090 1413 } else {
3de4586c 1414 struct inode *src_inode;
72fd032e 1415 src_file = fget(fd);
3de4586c
CM		 1416 if (!src_file) {
1417 ret = -EINVAL;
1418 goto out;
1419 }
1420
1421 src_inode = src_file->f_path.dentry->d_inode;
1422 if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
d397712b
CM		 1423 printk(KERN_INFO "btrfs: Snapshot src from "
1424 "another FS\n");
3de4586c
CM		 1425 ret = -EINVAL;
1426 fput(src_file);
1427 goto out;
1428 }
72fd032e
SW		 1429 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1430 BTRFS_I(src_inode)->root,
b83cc969 1431 transid, readonly);
3de4586c 1432 fput(src_file);
cb8e7090 1433 }
f46b5a66 1434out:
72fd032e
SW		 1435 return ret;
1436}
1437
1438static noinline int btrfs_ioctl_snap_create(struct file *file,
fa0d2b9b 1439 void __user *arg, int subvol)
72fd032e 1440{
fa0d2b9b 1441 struct btrfs_ioctl_vol_args *vol_args;
72fd032e
SW		 1442 int ret;
1443
fa0d2b9b
LZ		 1444 vol_args = memdup_user(arg, sizeof(*vol_args));
1445 if (IS_ERR(vol_args))
1446 return PTR_ERR(vol_args);
1447 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
72fd032e 1448
fa0d2b9b 1449 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
b83cc969
LZ		 1450 vol_args->fd, subvol,
1451 NULL, false);
fdfb1e4f 1452
fa0d2b9b
LZ		 1453 kfree(vol_args);
1454 return ret;
1455}
fdfb1e4f 1456
fa0d2b9b
LZ		 1457static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1458 void __user *arg, int subvol)
1459{
1460 struct btrfs_ioctl_vol_args_v2 *vol_args;
1461 int ret;
1462 u64 transid = 0;
1463 u64 *ptr = NULL;
b83cc969 1464 bool readonly = false;
75eaa0e2 1465
fa0d2b9b
LZ		 1466 vol_args = memdup_user(arg, sizeof(*vol_args));
1467 if (IS_ERR(vol_args))
1468 return PTR_ERR(vol_args);
1469 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
75eaa0e2 1470
b83cc969
LZ		 1471 if (vol_args->flags &
1472 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY)) {
1473 ret = -EOPNOTSUPP;
fa0d2b9b 1474 goto out;
72fd032e 1475 }
fa0d2b9b
LZ		 1476
1477 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1478 ptr = &transid;
b83cc969
LZ		 1479 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1480 readonly = true;
fa0d2b9b
LZ		 1481
1482 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
b83cc969
LZ		 1483 vol_args->fd, subvol,
1484 ptr, readonly);
fa0d2b9b
LZ		 1485
1486 if (ret == 0 && ptr &&
1487 copy_to_user(arg +
1488 offsetof(struct btrfs_ioctl_vol_args_v2,
1489 transid), ptr, sizeof(*ptr)))
1490 ret = -EFAULT;
fdfb1e4f 1491out:
f46b5a66
CH		 1492 kfree(vol_args);
1493 return ret;
1494}
1495
0caa102d
LZ		 1496static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1497 void __user *arg)
1498{
1499 struct inode *inode = fdentry(file)->d_inode;
1500 struct btrfs_root *root = BTRFS_I(inode)->root;
1501 int ret = 0;
1502 u64 flags = 0;
1503
33345d01 1504 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
0caa102d
LZ		 1505 return -EINVAL;
1506
1507 down_read(&root->fs_info->subvol_sem);
1508 if (btrfs_root_readonly(root))
1509 flags |= BTRFS_SUBVOL_RDONLY;
1510 up_read(&root->fs_info->subvol_sem);
1511
1512 if (copy_to_user(arg, &flags, sizeof(flags)))
1513 ret = -EFAULT;
1514
1515 return ret;
1516}
1517
1518static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1519 void __user *arg)
1520{
1521 struct inode *inode = fdentry(file)->d_inode;
1522 struct btrfs_root *root = BTRFS_I(inode)->root;
1523 struct btrfs_trans_handle *trans;
1524 u64 root_flags;
1525 u64 flags;
1526 int ret = 0;
1527
1528 if (root->fs_info->sb->s_flags & MS_RDONLY)
1529 return -EROFS;
1530
33345d01 1531 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
0caa102d
LZ		 1532 return -EINVAL;
1533
1534 if (copy_from_user(&flags, arg, sizeof(flags)))
1535 return -EFAULT;
1536
b4dc2b8c 1537 if (flags & BTRFS_SUBVOL_CREATE_ASYNC)
0caa102d
LZ		 1538 return -EINVAL;
1539
1540 if (flags & ~BTRFS_SUBVOL_RDONLY)
1541 return -EOPNOTSUPP;
1542
2e149670 1543 if (!inode_owner_or_capable(inode))
b4dc2b8c
LZ		 1544 return -EACCES;
1545
0caa102d
LZ		 1546 down_write(&root->fs_info->subvol_sem);
1547
1548 /* nothing to do */
1549 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1550 goto out;
1551
1552 root_flags = btrfs_root_flags(&root->root_item);
1553 if (flags & BTRFS_SUBVOL_RDONLY)
1554 btrfs_set_root_flags(&root->root_item,
1555 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1556 else
1557 btrfs_set_root_flags(&root->root_item,
1558 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1559
1560 trans = btrfs_start_transaction(root, 1);
1561 if (IS_ERR(trans)) {
1562 ret = PTR_ERR(trans);
1563 goto out_reset;
1564 }
1565
b4dc2b8c 1566 ret = btrfs_update_root(trans, root->fs_info->tree_root,
0caa102d
LZ		 1567 &root->root_key, &root->root_item);
1568
1569 btrfs_commit_transaction(trans, root);
1570out_reset:
1571 if (ret)
1572 btrfs_set_root_flags(&root->root_item, root_flags);
1573out:
1574 up_write(&root->fs_info->subvol_sem);
1575 return ret;
1576}
1577
76dda93c
YZ		 1578/*
1579 * helper to check if the subvolume references other subvolumes
1580 */
1581static noinline int may_destroy_subvol(struct btrfs_root *root)
1582{
1583 struct btrfs_path *path;
1584 struct btrfs_key key;
1585 int ret;
1586
1587 path = btrfs_alloc_path();
1588 if (!path)
1589 return -ENOMEM;
1590
1591 key.objectid = root->root_key.objectid;
1592 key.type = BTRFS_ROOT_REF_KEY;
1593 key.offset = (u64)-1;
1594
1595 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1596 &key, path, 0, 0);
1597 if (ret < 0)
1598 goto out;
1599 BUG_ON(ret == 0);
1600
1601 ret = 0;
1602 if (path->slots[0] > 0) {
1603 path->slots[0]--;
1604 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1605 if (key.objectid == root->root_key.objectid &&
1606 key.type == BTRFS_ROOT_REF_KEY)
1607 ret = -ENOTEMPTY;
1608 }
1609out:
1610 btrfs_free_path(path);
1611 return ret;
1612}
1613
ac8e9819
CM		 1614static noinline int key_in_sk(struct btrfs_key *key,
1615 struct btrfs_ioctl_search_key *sk)
1616{
abc6e134
CM		 1617 struct btrfs_key test;
1618 int ret;
1619
1620 test.objectid = sk->min_objectid;
1621 test.type = sk->min_type;
1622 test.offset = sk->min_offset;
1623
1624 ret = btrfs_comp_cpu_keys(key, &test);
1625 if (ret < 0)
ac8e9819 1626 return 0;
abc6e134
CM		 1627
1628 test.objectid = sk->max_objectid;
1629 test.type = sk->max_type;
1630 test.offset = sk->max_offset;
1631
1632 ret = btrfs_comp_cpu_keys(key, &test);
1633 if (ret > 0)
ac8e9819
CM		 1634 return 0;
1635 return 1;
1636}
1637
1638static noinline int copy_to_sk(struct btrfs_root *root,
1639 struct btrfs_path *path,
1640 struct btrfs_key *key,
1641 struct btrfs_ioctl_search_key *sk,
1642 char *buf,
1643 unsigned long *sk_offset,
1644 int *num_found)
1645{
1646 u64 found_transid;
1647 struct extent_buffer *leaf;
1648 struct btrfs_ioctl_search_header sh;
1649 unsigned long item_off;
1650 unsigned long item_len;
1651 int nritems;
1652 int i;
1653 int slot;
ac8e9819
CM		 1654 int ret = 0;
1655
1656 leaf = path->nodes[0];
1657 slot = path->slots[0];
1658 nritems = btrfs_header_nritems(leaf);
1659
1660 if (btrfs_header_generation(leaf) > sk->max_transid) {
1661 i = nritems;
1662 goto advance_key;
1663 }
1664 found_transid = btrfs_header_generation(leaf);
1665
1666 for (i = slot; i < nritems; i++) {
1667 item_off = btrfs_item_ptr_offset(leaf, i);
1668 item_len = btrfs_item_size_nr(leaf, i);
1669
1670 if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1671 item_len = 0;
1672
1673 if (sizeof(sh) + item_len + *sk_offset >
1674 BTRFS_SEARCH_ARGS_BUFSIZE) {
1675 ret = 1;
1676 goto overflow;
1677 }
1678
1679 btrfs_item_key_to_cpu(leaf, key, i);
1680 if (!key_in_sk(key, sk))
1681 continue;
1682
1683 sh.objectid = key->objectid;
1684 sh.offset = key->offset;
1685 sh.type = key->type;
1686 sh.len = item_len;
1687 sh.transid = found_transid;
1688
1689 /* copy search result header */
1690 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1691 *sk_offset += sizeof(sh);
1692
1693 if (item_len) {
1694 char *p = buf + *sk_offset;
1695 /* copy the item */
1696 read_extent_buffer(leaf, p,
1697 item_off, item_len);
1698 *sk_offset += item_len;
ac8e9819 1699 }
e2156867 1700 (*num_found)++;
ac8e9819
CM		 1701
1702 if (*num_found >= sk->nr_items)
1703 break;
1704 }
1705advance_key:
abc6e134
CM		 1706 ret = 0;
1707 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
ac8e9819 1708 key->offset++;
abc6e134
CM		 1709 else if (key->type < (u8)-1 && key->type < sk->max_type) {
1710 key->offset = 0;
ac8e9819 1711 key->type++;
abc6e134
CM		 1712 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1713 key->offset = 0;
1714 key->type = 0;
ac8e9819 1715 key->objectid++;
abc6e134
CM		 1716 } else
1717 ret = 1;
ac8e9819 1718overflow:
ac8e9819
CM		 1719 return ret;
1720}
1721
1722static noinline int search_ioctl(struct inode *inode,
1723 struct btrfs_ioctl_search_args *args)
1724{
1725 struct btrfs_root *root;
1726 struct btrfs_key key;
1727 struct btrfs_key max_key;
1728 struct btrfs_path *path;
1729 struct btrfs_ioctl_search_key *sk = &args->key;
1730 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1731 int ret;
1732 int num_found = 0;
1733 unsigned long sk_offset = 0;
1734
1735 path = btrfs_alloc_path();
1736 if (!path)
1737 return -ENOMEM;
1738
1739 if (sk->tree_id == 0) {
1740 /* search the root of the inode that was passed */
1741 root = BTRFS_I(inode)->root;
1742 } else {
1743 key.objectid = sk->tree_id;
1744 key.type = BTRFS_ROOT_ITEM_KEY;
1745 key.offset = (u64)-1;
1746 root = btrfs_read_fs_root_no_name(info, &key);
1747 if (IS_ERR(root)) {
1748 printk(KERN_ERR "could not find root %llu\n",
1749 sk->tree_id);
1750 btrfs_free_path(path);
1751 return -ENOENT;
1752 }
1753 }
1754
1755 key.objectid = sk->min_objectid;
1756 key.type = sk->min_type;
1757 key.offset = sk->min_offset;
1758
1759 max_key.objectid = sk->max_objectid;
1760 max_key.type = sk->max_type;
1761 max_key.offset = sk->max_offset;
1762
1763 path->keep_locks = 1;
1764
1765 while(1) {
1766 ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1767 sk->min_transid);
1768 if (ret != 0) {
1769 if (ret > 0)
1770 ret = 0;
1771 goto err;
1772 }
1773 ret = copy_to_sk(root, path, &key, sk, args->buf,
1774 &sk_offset, &num_found);
b3b4aa74 1775 btrfs_release_path(path);
ac8e9819
CM		 1776 if (ret || num_found >= sk->nr_items)
1777 break;
1778
1779 }
1780 ret = 0;
1781err:
1782 sk->nr_items = num_found;
1783 btrfs_free_path(path);
1784 return ret;
1785}
1786
1787static noinline int btrfs_ioctl_tree_search(struct file *file,
1788 void __user *argp)
1789{
1790 struct btrfs_ioctl_search_args *args;
1791 struct inode *inode;
1792 int ret;
1793
1794 if (!capable(CAP_SYS_ADMIN))
1795 return -EPERM;
1796
2354d08f
JL		 1797 args = memdup_user(argp, sizeof(*args));
1798 if (IS_ERR(args))
1799 return PTR_ERR(args);
ac8e9819 1800
ac8e9819
CM		 1801 inode = fdentry(file)->d_inode;
1802 ret = search_ioctl(inode, args);
1803 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1804 ret = -EFAULT;
1805 kfree(args);
1806 return ret;
1807}
1808
98d377a0 1809/*
ac8e9819
CM		 1810 * Search INODE_REFs to identify path name of 'dirid' directory
1811 * in a 'tree_id' tree. and sets path name to 'name'.
1812 */
98d377a0
TH		 1813static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1814 u64 tree_id, u64 dirid, char *name)
1815{
1816 struct btrfs_root *root;
1817 struct btrfs_key key;
ac8e9819 1818 char *ptr;
98d377a0
TH		 1819 int ret = -1;
1820 int slot;
1821 int len;
1822 int total_len = 0;
1823 struct btrfs_inode_ref *iref;
1824 struct extent_buffer *l;
1825 struct btrfs_path *path;
1826
1827 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1828 name[0]='\0';
1829 return 0;
1830 }
1831
1832 path = btrfs_alloc_path();
1833 if (!path)
1834 return -ENOMEM;
1835
ac8e9819 1836 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
98d377a0
TH		 1837
1838 key.objectid = tree_id;
1839 key.type = BTRFS_ROOT_ITEM_KEY;
1840 key.offset = (u64)-1;
1841 root = btrfs_read_fs_root_no_name(info, &key);
1842 if (IS_ERR(root)) {
1843 printk(KERN_ERR "could not find root %llu\n", tree_id);
8ad6fcab
CM		 1844 ret = -ENOENT;
1845 goto out;
98d377a0
TH		 1846 }
1847
1848 key.objectid = dirid;
1849 key.type = BTRFS_INODE_REF_KEY;
8ad6fcab 1850 key.offset = (u64)-1;
98d377a0
TH		 1851
1852 while(1) {
1853 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1854 if (ret < 0)
1855 goto out;
1856
1857 l = path->nodes[0];
1858 slot = path->slots[0];
8ad6fcab
CM		 1859 if (ret > 0 && slot > 0)
1860 slot--;
98d377a0
TH		 1861 btrfs_item_key_to_cpu(l, &key, slot);
1862
1863 if (ret > 0 && (key.objectid != dirid ||
ac8e9819
CM		 1864 key.type != BTRFS_INODE_REF_KEY)) {
1865 ret = -ENOENT;
98d377a0 1866 goto out;
ac8e9819 1867 }
98d377a0
TH		 1868
1869 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1870 len = btrfs_inode_ref_name_len(l, iref);
1871 ptr -= len + 1;
1872 total_len += len + 1;
ac8e9819 1873 if (ptr < name)
98d377a0
TH		 1874 goto out;
1875
1876 *(ptr + len) = '/';
1877 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1878
1879 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1880 break;
1881
b3b4aa74 1882 btrfs_release_path(path);
98d377a0 1883 key.objectid = key.offset;
8ad6fcab 1884 key.offset = (u64)-1;
98d377a0 1885 dirid = key.objectid;
98d377a0 1886 }
ac8e9819 1887 if (ptr < name)
98d377a0 1888 goto out;
77906a50 1889 memmove(name, ptr, total_len);
98d377a0
TH		 1890 name[total_len]='\0';
1891 ret = 0;
1892out:
1893 btrfs_free_path(path);
ac8e9819
CM		 1894 return ret;
1895}
1896
1897static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1898 void __user *argp)
1899{
1900 struct btrfs_ioctl_ino_lookup_args *args;
1901 struct inode *inode;
1902 int ret;
1903
1904 if (!capable(CAP_SYS_ADMIN))
1905 return -EPERM;
1906
2354d08f
JL		 1907 args = memdup_user(argp, sizeof(*args));
1908 if (IS_ERR(args))
1909 return PTR_ERR(args);
c2b96929 1910
ac8e9819
CM		 1911 inode = fdentry(file)->d_inode;
1912
1b53ac4d
CM		 1913 if (args->treeid == 0)
1914 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1915
ac8e9819
CM		 1916 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1917 args->treeid, args->objectid,
1918 args->name);
1919
1920 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1921 ret = -EFAULT;
1922
1923 kfree(args);
98d377a0
TH		 1924 return ret;
1925}
1926
76dda93c
YZ		 1927static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1928 void __user *arg)
1929{
1930 struct dentry *parent = fdentry(file);
1931 struct dentry *dentry;
1932 struct inode *dir = parent->d_inode;
1933 struct inode *inode;
1934 struct btrfs_root *root = BTRFS_I(dir)->root;
1935 struct btrfs_root *dest = NULL;
1936 struct btrfs_ioctl_vol_args *vol_args;
1937 struct btrfs_trans_handle *trans;
1938 int namelen;
1939 int ret;
1940 int err = 0;
1941
76dda93c
YZ		 1942 vol_args = memdup_user(arg, sizeof(*vol_args));
1943 if (IS_ERR(vol_args))
1944 return PTR_ERR(vol_args);
1945
1946 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1947 namelen = strlen(vol_args->name);
1948 if (strchr(vol_args->name, '/') ||
1949 strncmp(vol_args->name, "..", namelen) == 0) {
1950 err = -EINVAL;
1951 goto out;
1952 }
1953
a561be71 1954 err = mnt_want_write_file(file);
76dda93c
YZ		 1955 if (err)
1956 goto out;
1957
1958 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
1959 dentry = lookup_one_len(vol_args->name, parent, namelen);
1960 if (IS_ERR(dentry)) {
1961 err = PTR_ERR(dentry);
1962 goto out_unlock_dir;
1963 }
1964
1965 if (!dentry->d_inode) {
1966 err = -ENOENT;
1967 goto out_dput;
1968 }
1969
1970 inode = dentry->d_inode;
4260f7c7
SW		 1971 dest = BTRFS_I(inode)->root;
1972 if (!capable(CAP_SYS_ADMIN)){
1973 /*
1974 * Regular user. Only allow this with a special mount
1975 * option, when the user has write+exec access to the
1976 * subvol root, and when rmdir(2) would have been
1977 * allowed.
1978 *
1979 * Note that this is _not_ check that the subvol is
1980 * empty or doesn't contain data that we wouldn't
1981 * otherwise be able to delete.
1982 *
1983 * Users who want to delete empty subvols should try
1984 * rmdir(2).
1985 */
1986 err = -EPERM;
1987 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1988 goto out_dput;
1989
1990 /*
1991 * Do not allow deletion if the parent dir is the same
1992 * as the dir to be deleted. That means the ioctl
1993 * must be called on the dentry referencing the root
1994 * of the subvol, not a random directory contained
1995 * within it.
1996 */
1997 err = -EINVAL;
1998 if (root == dest)
1999 goto out_dput;
2000
2001 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2002 if (err)
2003 goto out_dput;
2004
2005 /* check if subvolume may be deleted by a non-root user */
2006 err = btrfs_may_delete(dir, dentry, 1);
2007 if (err)
2008 goto out_dput;
2009 }
2010
33345d01 2011 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
76dda93c
YZ		 2012 err = -EINVAL;
2013 goto out_dput;
2014 }
2015
76dda93c
YZ		 2016 mutex_lock(&inode->i_mutex);
2017 err = d_invalidate(dentry);
2018 if (err)
2019 goto out_unlock;
2020
2021 down_write(&root->fs_info->subvol_sem);
2022
2023 err = may_destroy_subvol(dest);
2024 if (err)
2025 goto out_up_write;
2026
a22285a6
YZ		 2027 trans = btrfs_start_transaction(root, 0);
2028 if (IS_ERR(trans)) {
2029 err = PTR_ERR(trans);
d327099a 2030 goto out_up_write;
a22285a6
YZ		 2031 }
2032 trans->block_rsv = &root->fs_info->global_block_rsv;
2033
76dda93c
YZ		 2034 ret = btrfs_unlink_subvol(trans, root, dir,
2035 dest->root_key.objectid,
2036 dentry->d_name.name,
2037 dentry->d_name.len);
79787eaa
JM		 2038 if (ret) {
2039 err = ret;
2040 btrfs_abort_transaction(trans, root, ret);
2041 goto out_end_trans;
2042 }
76dda93c
YZ		 2043
2044 btrfs_record_root_in_trans(trans, dest);
2045
2046 memset(&dest->root_item.drop_progress, 0,
2047 sizeof(dest->root_item.drop_progress));
2048 dest->root_item.drop_level = 0;
2049 btrfs_set_root_refs(&dest->root_item, 0);
2050
d68fc57b
YZ		 2051 if (!xchg(&dest->orphan_item_inserted, 1)) {
2052 ret = btrfs_insert_orphan_item(trans,
2053 root->fs_info->tree_root,
2054 dest->root_key.objectid);
79787eaa
JM		 2055 if (ret) {
2056 btrfs_abort_transaction(trans, root, ret);
2057 err = ret;
2058 goto out_end_trans;
2059 }
d68fc57b 2060 }
79787eaa 2061out_end_trans:
531cb13f 2062 ret = btrfs_end_transaction(trans, root);
79787eaa
JM		 2063 if (ret && !err)
2064 err = ret;
76dda93c
YZ		 2065 inode->i_flags |= S_DEAD;
2066out_up_write:
2067 up_write(&root->fs_info->subvol_sem);
2068out_unlock:
2069 mutex_unlock(&inode->i_mutex);
2070 if (!err) {
efefb143 2071 shrink_dcache_sb(root->fs_info->sb);
76dda93c
YZ		 2072 btrfs_invalidate_inodes(dest);
2073 d_delete(dentry);
2074 }
2075out_dput:
2076 dput(dentry);
2077out_unlock_dir:
2078 mutex_unlock(&dir->i_mutex);
2a79f17e 2079 mnt_drop_write_file(file);
76dda93c
YZ		 2080out:
2081 kfree(vol_args);
2082 return err;
2083}
2084
1e701a32 2085static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
f46b5a66
CH		 2086{
2087 struct inode *inode = fdentry(file)->d_inode;
2088 struct btrfs_root *root = BTRFS_I(inode)->root;
1e701a32 2089 struct btrfs_ioctl_defrag_range_args *range;
c146afad
YZ		 2090 int ret;
2091
b83cc969
LZ		 2092 if (btrfs_root_readonly(root))
2093 return -EROFS;
2094
a561be71 2095 ret = mnt_want_write_file(file);
c146afad
YZ		 2096 if (ret)
2097 return ret;
f46b5a66
CH		 2098
2099 switch (inode->i_mode & S_IFMT) {
2100 case S_IFDIR:
e441d54d
CM		 2101 if (!capable(CAP_SYS_ADMIN)) {
2102 ret = -EPERM;
2103 goto out;
2104 }
8929ecfa
YZ		 2105 ret = btrfs_defrag_root(root, 0);
2106 if (ret)
2107 goto out;
2108 ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
f46b5a66
CH		 2109 break;
2110 case S_IFREG:
e441d54d
CM		 2111 if (!(file->f_mode & FMODE_WRITE)) {
2112 ret = -EINVAL;
2113 goto out;
2114 }
1e701a32
CM		 2115
2116 range = kzalloc(sizeof(*range), GFP_KERNEL);
2117 if (!range) {
2118 ret = -ENOMEM;
2119 goto out;
2120 }
2121
2122 if (argp) {
2123 if (copy_from_user(range, argp,
2124 sizeof(*range))) {
2125 ret = -EFAULT;
2126 kfree(range);
683be16e 2127 goto out;
1e701a32
CM		 2128 }
2129 /* compression requires us to start the IO */
2130 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2131 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2132 range->extent_thresh = (u32)-1;
2133 }
2134 } else {
2135 /* the rest are all set to zero by kzalloc */
2136 range->len = (u64)-1;
2137 }
4cb5300b
CM		 2138 ret = btrfs_defrag_file(fdentry(file)->d_inode, file,
2139 range, 0, 0);
2140 if (ret > 0)
2141 ret = 0;
1e701a32 2142 kfree(range);
f46b5a66 2143 break;
8929ecfa
YZ		 2144 default:
2145 ret = -EINVAL;
f46b5a66 2146 }
e441d54d 2147out:
2a79f17e 2148 mnt_drop_write_file(file);
e441d54d 2149 return ret;
f46b5a66
CH		 2150}
2151
b2950863 2152static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
f46b5a66
CH		 2153{
2154 struct btrfs_ioctl_vol_args *vol_args;
2155 int ret;
2156
e441d54d
CM		 2157 if (!capable(CAP_SYS_ADMIN))
2158 return -EPERM;
2159
c9e9f97b
ID		 2160 mutex_lock(&root->fs_info->volume_mutex);
2161 if (root->fs_info->balance_ctl) {
2162 printk(KERN_INFO "btrfs: balance in progress\n");
2163 ret = -EINVAL;
2164 goto out;
2165 }
2166
dae7b665 2167 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 2168 if (IS_ERR(vol_args)) {
2169 ret = PTR_ERR(vol_args);
2170 goto out;
2171 }
f46b5a66 2172
5516e595 2173 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66
CH		 2174 ret = btrfs_init_new_device(root, vol_args->name);
2175
f46b5a66 2176 kfree(vol_args);
c9e9f97b
ID		 2177out:
2178 mutex_unlock(&root->fs_info->volume_mutex);
f46b5a66
CH		 2179 return ret;
2180}
2181
b2950863 2182static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
f46b5a66
CH		 2183{
2184 struct btrfs_ioctl_vol_args *vol_args;
2185 int ret;
2186
e441d54d
CM		 2187 if (!capable(CAP_SYS_ADMIN))
2188 return -EPERM;
2189
c146afad
YZ		 2190 if (root->fs_info->sb->s_flags & MS_RDONLY)
2191 return -EROFS;
2192
c9e9f97b
ID		 2193 mutex_lock(&root->fs_info->volume_mutex);
2194 if (root->fs_info->balance_ctl) {
2195 printk(KERN_INFO "btrfs: balance in progress\n");
2196 ret = -EINVAL;
2197 goto out;
2198 }
2199
dae7b665 2200 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 2201 if (IS_ERR(vol_args)) {
2202 ret = PTR_ERR(vol_args);
2203 goto out;
2204 }
f46b5a66 2205
5516e595 2206 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66
CH		 2207 ret = btrfs_rm_device(root, vol_args->name);
2208
f46b5a66 2209 kfree(vol_args);
c9e9f97b
ID		 2210out:
2211 mutex_unlock(&root->fs_info->volume_mutex);
f46b5a66
CH		 2212 return ret;
2213}
2214
475f6387
JS		 2215static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2216{
027ed2f0 2217 struct btrfs_ioctl_fs_info_args *fi_args;
475f6387
JS		 2218 struct btrfs_device *device;
2219 struct btrfs_device *next;
2220 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
027ed2f0 2221 int ret = 0;
475f6387
JS		 2222
2223 if (!capable(CAP_SYS_ADMIN))
2224 return -EPERM;
2225
027ed2f0
LZ		 2226 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2227 if (!fi_args)
2228 return -ENOMEM;
2229
2230 fi_args->num_devices = fs_devices->num_devices;
2231 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
475f6387
JS		 2232
2233 mutex_lock(&fs_devices->device_list_mutex);
2234 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
027ed2f0
LZ		 2235 if (device->devid > fi_args->max_id)
2236 fi_args->max_id = device->devid;
475f6387
JS		 2237 }
2238 mutex_unlock(&fs_devices->device_list_mutex);
2239
027ed2f0
LZ		 2240 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2241 ret = -EFAULT;
475f6387 2242
027ed2f0
LZ		 2243 kfree(fi_args);
2244 return ret;
475f6387
JS		 2245}
2246
2247static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2248{
2249 struct btrfs_ioctl_dev_info_args *di_args;
2250 struct btrfs_device *dev;
2251 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2252 int ret = 0;
2253 char *s_uuid = NULL;
2254 char empty_uuid[BTRFS_UUID_SIZE] = {0};
2255
2256 if (!capable(CAP_SYS_ADMIN))
2257 return -EPERM;
2258
2259 di_args = memdup_user(arg, sizeof(*di_args));
2260 if (IS_ERR(di_args))
2261 return PTR_ERR(di_args);
2262
2263 if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2264 s_uuid = di_args->uuid;
2265
2266 mutex_lock(&fs_devices->device_list_mutex);
2267 dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL);
2268 mutex_unlock(&fs_devices->device_list_mutex);
2269
2270 if (!dev) {
2271 ret = -ENODEV;
2272 goto out;
2273 }
2274
2275 di_args->devid = dev->devid;
2276 di_args->bytes_used = dev->bytes_used;
2277 di_args->total_bytes = dev->total_bytes;
2278 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
a27202fb 2279 if (dev->name) {
606686ee
JB		 2280 struct rcu_string *name;
2281
2282 rcu_read_lock();
2283 name = rcu_dereference(dev->name);
2284 strncpy(di_args->path, name->str, sizeof(di_args->path));
2285 rcu_read_unlock();
a27202fb
JM		 2286 di_args->path[sizeof(di_args->path) - 1] = 0;
2287 } else {
99ba55ad 2288 di_args->path[0] = '\0';
a27202fb 2289 }
475f6387
JS		 2290
2291out:
2292 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2293 ret = -EFAULT;
2294
2295 kfree(di_args);
2296 return ret;
2297}
2298
76dda93c
YZ		 2299static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2300 u64 off, u64 olen, u64 destoff)
f46b5a66
CH		 2301{
2302 struct inode *inode = fdentry(file)->d_inode;
2303 struct btrfs_root *root = BTRFS_I(inode)->root;
2304 struct file *src_file;
2305 struct inode *src;
2306 struct btrfs_trans_handle *trans;
f46b5a66 2307 struct btrfs_path *path;
f46b5a66 2308 struct extent_buffer *leaf;
ae01a0ab
YZ		 2309 char *buf;
2310 struct btrfs_key key;
f46b5a66
CH		 2311 u32 nritems;
2312 int slot;
ae01a0ab 2313 int ret;
c5c9cd4d
SW		 2314 u64 len = olen;
2315 u64 bs = root->fs_info->sb->s_blocksize;
2316 u64 hint_byte;
d20f7043 2317
c5c9cd4d
SW		 2318 /*
2319 * TODO:
2320 * - split compressed inline extents. annoying: we need to
2321 * decompress into destination's address_space (the file offset
2322 * may change, so source mapping won't do), then recompress (or
2323 * otherwise reinsert) a subrange.
2324 * - allow ranges within the same file to be cloned (provided
2325 * they don't overlap)?
2326 */
2327
e441d54d 2328 /* the destination must be opened for writing */
2ebc3464 2329 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
e441d54d
CM		 2330 return -EINVAL;
2331
b83cc969
LZ		 2332 if (btrfs_root_readonly(root))
2333 return -EROFS;
2334
a561be71 2335 ret = mnt_want_write_file(file);
c146afad
YZ		 2336 if (ret)
2337 return ret;
2338
c5c9cd4d 2339 src_file = fget(srcfd);
ab67b7c1
YZ		 2340 if (!src_file) {
2341 ret = -EBADF;
2342 goto out_drop_write;
2343 }
5dc64164 2344
f46b5a66
CH		 2345 src = src_file->f_dentry->d_inode;
2346
c5c9cd4d
SW		 2347 ret = -EINVAL;
2348 if (src == inode)
2349 goto out_fput;
2350
5dc64164
DR		 2351 /* the src must be open for reading */
2352 if (!(src_file->f_mode & FMODE_READ))
2353 goto out_fput;
2354
0e7b824c
LZ		 2355 /* don't make the dst file partly checksummed */
2356 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2357 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2358 goto out_fput;
2359
ae01a0ab
YZ		 2360 ret = -EISDIR;
2361 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2362 goto out_fput;
2363
f46b5a66 2364 ret = -EXDEV;
ae01a0ab
YZ		 2365 if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
2366 goto out_fput;
2367
2368 ret = -ENOMEM;
2369 buf = vmalloc(btrfs_level_size(root, 0));
2370 if (!buf)
2371 goto out_fput;
2372
2373 path = btrfs_alloc_path();
2374 if (!path) {
2375 vfree(buf);
f46b5a66 2376 goto out_fput;
ae01a0ab
YZ		 2377 }
2378 path->reada = 2;
f46b5a66
CH		 2379
2380 if (inode < src) {
fccdae43
SW		 2381 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2382 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
f46b5a66 2383 } else {
fccdae43
SW		 2384 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2385 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
f46b5a66
CH		 2386 }
2387
c5c9cd4d
SW		 2388 /* determine range to clone */
2389 ret = -EINVAL;
2ebc3464 2390 if (off + len > src->i_size || off + len < off)
f46b5a66 2391 goto out_unlock;
c5c9cd4d
SW		 2392 if (len == 0)
2393 olen = len = src->i_size - off;
2394 /* if we extend to eof, continue to block boundary */
2395 if (off + len == src->i_size)
2a6b8dae 2396 len = ALIGN(src->i_size, bs) - off;
c5c9cd4d
SW		 2397
2398 /* verify the end result is block aligned */
2a6b8dae
LZ		 2399 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2400 !IS_ALIGNED(destoff, bs))
c5c9cd4d
SW		 2401 goto out_unlock;
2402
d525e8ab
LZ		 2403 if (destoff > inode->i_size) {
2404 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2405 if (ret)
2406 goto out_unlock;
2407 }
2408
71ef0786
LZ		 2409 /* truncate page cache pages from target inode range */
2410 truncate_inode_pages_range(&inode->i_data, destoff,
2411 PAGE_CACHE_ALIGN(destoff + len) - 1);
2412
f46b5a66
CH		 2413 /* do any pending delalloc/csum calc on src, one way or
2414 another, and lock file content */
2415 while (1) {
31840ae1 2416 struct btrfs_ordered_extent *ordered;
d0082371 2417 lock_extent(&BTRFS_I(src)->io_tree, off, off+len);
9a019196
SW		 2418 ordered = btrfs_lookup_first_ordered_extent(src, off+len);
2419 if (!ordered &&
2420 !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len,
2421 EXTENT_DELALLOC, 0, NULL))
f46b5a66 2422 break;
d0082371 2423 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len);
ae01a0ab
YZ		 2424 if (ordered)
2425 btrfs_put_ordered_extent(ordered);
9a019196 2426 btrfs_wait_ordered_range(src, off, len);
f46b5a66
CH		 2427 }
2428
c5c9cd4d 2429 /* clone data */
33345d01 2430 key.objectid = btrfs_ino(src);
ae01a0ab
YZ		 2431 key.type = BTRFS_EXTENT_DATA_KEY;
2432 key.offset = 0;
f46b5a66
CH		 2433
2434 while (1) {
2435 /*
2436 * note the key will change type as we walk through the
2437 * tree.
2438 */
a22285a6 2439 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
f46b5a66
CH		 2440 if (ret < 0)
2441 goto out;
2442
ae01a0ab
YZ		 2443 nritems = btrfs_header_nritems(path->nodes[0]);
2444 if (path->slots[0] >= nritems) {
f46b5a66
CH		 2445 ret = btrfs_next_leaf(root, path);
2446 if (ret < 0)
2447 goto out;
2448 if (ret > 0)
2449 break;
ae01a0ab 2450 nritems = btrfs_header_nritems(path->nodes[0]);
f46b5a66
CH		 2451 }
2452 leaf = path->nodes[0];
2453 slot = path->slots[0];
f46b5a66 2454
ae01a0ab 2455 btrfs_item_key_to_cpu(leaf, &key, slot);
d20f7043 2456 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
33345d01 2457 key.objectid != btrfs_ino(src))
f46b5a66
CH		 2458 break;
2459
c5c9cd4d
SW		 2460 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2461 struct btrfs_file_extent_item *extent;
2462 int type;
31840ae1
ZY		 2463 u32 size;
2464 struct btrfs_key new_key;
c5c9cd4d
SW		 2465 u64 disko = 0, diskl = 0;
2466 u64 datao = 0, datal = 0;
2467 u8 comp;
b5384d48 2468 u64 endoff;
31840ae1
ZY		 2469
2470 size = btrfs_item_size_nr(leaf, slot);
2471 read_extent_buffer(leaf, buf,
2472 btrfs_item_ptr_offset(leaf, slot),
2473 size);
c5c9cd4d
SW		 2474
2475 extent = btrfs_item_ptr(leaf, slot,
2476 struct btrfs_file_extent_item);
2477 comp = btrfs_file_extent_compression(leaf, extent);
2478 type = btrfs_file_extent_type(leaf, extent);
c8a894d7
CM		 2479 if (type == BTRFS_FILE_EXTENT_REG ||
2480 type == BTRFS_FILE_EXTENT_PREALLOC) {
d397712b
CM		 2481 disko = btrfs_file_extent_disk_bytenr(leaf,
2482 extent);
2483 diskl = btrfs_file_extent_disk_num_bytes(leaf,
2484 extent);
c5c9cd4d 2485 datao = btrfs_file_extent_offset(leaf, extent);
d397712b
CM		 2486 datal = btrfs_file_extent_num_bytes(leaf,
2487 extent);
c5c9cd4d
SW		 2488 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2489 /* take upper bound, may be compressed */
2490 datal = btrfs_file_extent_ram_bytes(leaf,
2491 extent);
2492 }
b3b4aa74 2493 btrfs_release_path(path);
31840ae1 2494
050006a7 2495 if (key.offset + datal <= off ||
c5c9cd4d
SW		 2496 key.offset >= off+len)
2497 goto next;
2498
31840ae1 2499 memcpy(&new_key, &key, sizeof(new_key));
33345d01 2500 new_key.objectid = btrfs_ino(inode);
4d728ec7
LZ		 2501 if (off <= key.offset)
2502 new_key.offset = key.offset + destoff - off;
2503 else
2504 new_key.offset = destoff;
31840ae1 2505
b6f3409b
SW		 2506 /*
2507 * 1 - adjusting old extent (we may have to split it)
2508 * 1 - add new extent
2509 * 1 - inode update
2510 */
2511 trans = btrfs_start_transaction(root, 3);
a22285a6
YZ		 2512 if (IS_ERR(trans)) {
2513 ret = PTR_ERR(trans);
2514 goto out;
2515 }
2516
c8a894d7
CM		 2517 if (type == BTRFS_FILE_EXTENT_REG ||
2518 type == BTRFS_FILE_EXTENT_PREALLOC) {
d72c0842
LZ		 2519 /*
2520 * a | --- range to clone ---| b
2521 * | ------------- extent ------------- |
2522 */
2523
2524 /* substract range b */
2525 if (key.offset + datal > off + len)
2526 datal = off + len - key.offset;
2527
2528 /* substract range a */
a22285a6
YZ		 2529 if (off > key.offset) {
2530 datao += off - key.offset;
2531 datal -= off - key.offset;
2532 }
2533
a22285a6
YZ		 2534 ret = btrfs_drop_extents(trans, inode,
2535 new_key.offset,
2536 new_key.offset + datal,
2537 &hint_byte, 1);
79787eaa
JM		 2538 if (ret) {
2539 btrfs_abort_transaction(trans, root,
2540 ret);
2541 btrfs_end_transaction(trans, root);
2542 goto out;
2543 }
a22285a6 2544
c5c9cd4d
SW		 2545 ret = btrfs_insert_empty_item(trans, root, path,
2546 &new_key, size);
79787eaa
JM		 2547 if (ret) {
2548 btrfs_abort_transaction(trans, root,
2549 ret);
2550 btrfs_end_transaction(trans, root);
2551 goto out;
2552 }
c5c9cd4d
SW		 2553
2554 leaf = path->nodes[0];
2555 slot = path->slots[0];
2556 write_extent_buffer(leaf, buf,
31840ae1
ZY		 2557 btrfs_item_ptr_offset(leaf, slot),
2558 size);
ae01a0ab 2559
c5c9cd4d 2560 extent = btrfs_item_ptr(leaf, slot,
f46b5a66 2561 struct btrfs_file_extent_item);
c5c9cd4d 2562
c5c9cd4d
SW		 2563 /* disko == 0 means it's a hole */
2564 if (!disko)
2565 datao = 0;
c5c9cd4d
SW		 2566
2567 btrfs_set_file_extent_offset(leaf, extent,
2568 datao);
2569 btrfs_set_file_extent_num_bytes(leaf, extent,
2570 datal);
2571 if (disko) {
2572 inode_add_bytes(inode, datal);
ae01a0ab 2573 ret = btrfs_inc_extent_ref(trans, root,
5d4f98a2
YZ		 2574 disko, diskl, 0,
2575 root->root_key.objectid,
33345d01 2576 btrfs_ino(inode),
66d7e7f0
AJ		 2577 new_key.offset - datao,
2578 0);
79787eaa
JM		 2579 if (ret) {
2580 btrfs_abort_transaction(trans,
2581 root,
2582 ret);
2583 btrfs_end_transaction(trans,
2584 root);
2585 goto out;
2586
2587 }
f46b5a66 2588 }
c5c9cd4d
SW		 2589 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2590 u64 skip = 0;
2591 u64 trim = 0;
2592 if (off > key.offset) {
2593 skip = off - key.offset;
2594 new_key.offset += skip;
2595 }
d397712b 2596
c5c9cd4d
SW		 2597 if (key.offset + datal > off+len)
2598 trim = key.offset + datal - (off+len);
d397712b 2599
c5c9cd4d 2600 if (comp && (skip || trim)) {
c5c9cd4d 2601 ret = -EINVAL;
a22285a6 2602 btrfs_end_transaction(trans, root);
c5c9cd4d
SW		 2603 goto out;
2604 }
2605 size -= skip + trim;
2606 datal -= skip + trim;
a22285a6
YZ		 2607
2608 ret = btrfs_drop_extents(trans, inode,
2609 new_key.offset,
2610 new_key.offset + datal,
2611 &hint_byte, 1);
79787eaa
JM		 2612 if (ret) {
2613 btrfs_abort_transaction(trans, root,
2614 ret);
2615 btrfs_end_transaction(trans, root);
2616 goto out;
2617 }
a22285a6 2618
c5c9cd4d
SW		 2619 ret = btrfs_insert_empty_item(trans, root, path,
2620 &new_key, size);
79787eaa
JM		 2621 if (ret) {
2622 btrfs_abort_transaction(trans, root,
2623 ret);
2624 btrfs_end_transaction(trans, root);
2625 goto out;
2626 }
c5c9cd4d
SW		 2627
2628 if (skip) {
d397712b
CM		 2629 u32 start =
2630 btrfs_file_extent_calc_inline_size(0);
c5c9cd4d
SW		 2631 memmove(buf+start, buf+start+skip,
2632 datal);
2633 }
2634
2635 leaf = path->nodes[0];
2636 slot = path->slots[0];
2637 write_extent_buffer(leaf, buf,
2638 btrfs_item_ptr_offset(leaf, slot),
2639 size);
2640 inode_add_bytes(inode, datal);
f46b5a66 2641 }
c5c9cd4d
SW		 2642
2643 btrfs_mark_buffer_dirty(leaf);
b3b4aa74 2644 btrfs_release_path(path);
c5c9cd4d 2645
0c4d2d95 2646 inode_inc_iversion(inode);
a22285a6 2647 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
b5384d48
SW		 2648
2649 /*
2650 * we round up to the block size at eof when
2651 * determining which extents to clone above,
2652 * but shouldn't round up the file size
2653 */
2654 endoff = new_key.offset + datal;
5f3888ff
LZ		 2655 if (endoff > destoff+olen)
2656 endoff = destoff+olen;
b5384d48
SW		 2657 if (endoff > inode->i_size)
2658 btrfs_i_size_write(inode, endoff);
2659
a22285a6 2660 ret = btrfs_update_inode(trans, root, inode);
79787eaa
JM		 2661 if (ret) {
2662 btrfs_abort_transaction(trans, root, ret);
2663 btrfs_end_transaction(trans, root);
2664 goto out;
2665 }
2666 ret = btrfs_end_transaction(trans, root);
a22285a6 2667 }
d397712b 2668next:
b3b4aa74 2669 btrfs_release_path(path);
f46b5a66 2670 key.offset++;
f46b5a66 2671 }
f46b5a66
CH		 2672 ret = 0;
2673out:
b3b4aa74 2674 btrfs_release_path(path);
d0082371 2675 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len);
f46b5a66
CH		 2676out_unlock:
2677 mutex_unlock(&src->i_mutex);
2678 mutex_unlock(&inode->i_mutex);
ae01a0ab
YZ		 2679 vfree(buf);
2680 btrfs_free_path(path);
f46b5a66
CH		 2681out_fput:
2682 fput(src_file);
ab67b7c1 2683out_drop_write:
2a79f17e 2684 mnt_drop_write_file(file);
f46b5a66
CH		 2685 return ret;
2686}
2687
7a865e8a 2688static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
c5c9cd4d
SW		 2689{
2690 struct btrfs_ioctl_clone_range_args args;
2691
7a865e8a 2692 if (copy_from_user(&args, argp, sizeof(args)))
c5c9cd4d
SW		 2693 return -EFAULT;
2694 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2695 args.src_length, args.dest_offset);
2696}
2697
f46b5a66
CH		 2698/*
2699 * there are many ways the trans_start and trans_end ioctls can lead
2700 * to deadlocks. They should only be used by applications that
2701 * basically own the machine, and have a very in depth understanding
2702 * of all the possible deadlocks and enospc problems.
2703 */
b2950863 2704static long btrfs_ioctl_trans_start(struct file *file)
f46b5a66
CH		 2705{
2706 struct inode *inode = fdentry(file)->d_inode;
2707 struct btrfs_root *root = BTRFS_I(inode)->root;
2708 struct btrfs_trans_handle *trans;
1ab86aed 2709 int ret;
f46b5a66 2710
1ab86aed 2711 ret = -EPERM;
df5b5520 2712 if (!capable(CAP_SYS_ADMIN))
1ab86aed 2713 goto out;
df5b5520 2714
1ab86aed
SW		 2715 ret = -EINPROGRESS;
2716 if (file->private_data)
f46b5a66 2717 goto out;
9ca9ee09 2718
b83cc969
LZ		 2719 ret = -EROFS;
2720 if (btrfs_root_readonly(root))
2721 goto out;
2722
a561be71 2723 ret = mnt_want_write_file(file);
c146afad
YZ		 2724 if (ret)
2725 goto out;
2726
a4abeea4 2727 atomic_inc(&root->fs_info->open_ioctl_trans);
9ca9ee09 2728
1ab86aed 2729 ret = -ENOMEM;
7a7eaa40 2730 trans = btrfs_start_ioctl_transaction(root);
abd30bb0 2731 if (IS_ERR(trans))
1ab86aed
SW		 2732 goto out_drop;
2733
2734 file->private_data = trans;
2735 return 0;
2736
2737out_drop:
a4abeea4 2738 atomic_dec(&root->fs_info->open_ioctl_trans);
2a79f17e 2739 mnt_drop_write_file(file);
f46b5a66 2740out:
f46b5a66
CH		 2741 return ret;
2742}
2743
6ef5ed0d
JB		 2744static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2745{
2746 struct inode *inode = fdentry(file)->d_inode;
2747 struct btrfs_root *root = BTRFS_I(inode)->root;
2748 struct btrfs_root *new_root;
2749 struct btrfs_dir_item *di;
2750 struct btrfs_trans_handle *trans;
2751 struct btrfs_path *path;
2752 struct btrfs_key location;
2753 struct btrfs_disk_key disk_key;
2754 struct btrfs_super_block *disk_super;
2755 u64 features;
2756 u64 objectid = 0;
2757 u64 dir_id;
2758
2759 if (!capable(CAP_SYS_ADMIN))
2760 return -EPERM;
2761
2762 if (copy_from_user(&objectid, argp, sizeof(objectid)))
2763 return -EFAULT;
2764
2765 if (!objectid)
2766 objectid = root->root_key.objectid;
2767
2768 location.objectid = objectid;
2769 location.type = BTRFS_ROOT_ITEM_KEY;
2770 location.offset = (u64)-1;
2771
2772 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2773 if (IS_ERR(new_root))
2774 return PTR_ERR(new_root);
2775
2776 if (btrfs_root_refs(&new_root->root_item) == 0)
2777 return -ENOENT;
2778
2779 path = btrfs_alloc_path();
2780 if (!path)
2781 return -ENOMEM;
2782 path->leave_spinning = 1;
2783
2784 trans = btrfs_start_transaction(root, 1);
98d5dc13 2785 if (IS_ERR(trans)) {
6ef5ed0d 2786 btrfs_free_path(path);
98d5dc13 2787 return PTR_ERR(trans);
6ef5ed0d
JB		 2788 }
2789
6c41761f 2790 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
6ef5ed0d
JB		 2791 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2792 dir_id, "default", 7, 1);
cf1e99a4 2793 if (IS_ERR_OR_NULL(di)) {
6ef5ed0d
JB		 2794 btrfs_free_path(path);
2795 btrfs_end_transaction(trans, root);
2796 printk(KERN_ERR "Umm, you don't have the default dir item, "
2797 "this isn't going to work\n");
2798 return -ENOENT;
2799 }
2800
2801 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2802 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2803 btrfs_mark_buffer_dirty(path->nodes[0]);
2804 btrfs_free_path(path);
2805
6c41761f 2806 disk_super = root->fs_info->super_copy;
6ef5ed0d
JB		 2807 features = btrfs_super_incompat_flags(disk_super);
2808 if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) {
2809 features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL;
2810 btrfs_set_super_incompat_flags(disk_super, features);
2811 }
2812 btrfs_end_transaction(trans, root);
2813
2814 return 0;
2815}
2816
bf5fc093
JB		 2817static void get_block_group_info(struct list_head *groups_list,
2818 struct btrfs_ioctl_space_info *space)
2819{
2820 struct btrfs_block_group_cache *block_group;
2821
2822 space->total_bytes = 0;
2823 space->used_bytes = 0;
2824 space->flags = 0;
2825 list_for_each_entry(block_group, groups_list, list) {
2826 space->flags = block_group->flags;
2827 space->total_bytes += block_group->key.offset;
2828 space->used_bytes +=
2829 btrfs_block_group_used(&block_group->item);
2830 }
2831}
2832
1406e432
JB		 2833long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2834{
2835 struct btrfs_ioctl_space_args space_args;
2836 struct btrfs_ioctl_space_info space;
2837 struct btrfs_ioctl_space_info *dest;
7fde62bf 2838 struct btrfs_ioctl_space_info *dest_orig;
13f2696f 2839 struct btrfs_ioctl_space_info __user *user_dest;
1406e432 2840 struct btrfs_space_info *info;
bf5fc093
JB		 2841 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2842 BTRFS_BLOCK_GROUP_SYSTEM,
2843 BTRFS_BLOCK_GROUP_METADATA,
2844 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2845 int num_types = 4;
7fde62bf 2846 int alloc_size;
1406e432 2847 int ret = 0;
51788b1b 2848 u64 slot_count = 0;
bf5fc093 2849 int i, c;
1406e432
JB		 2850
2851 if (copy_from_user(&space_args,
2852 (struct btrfs_ioctl_space_args __user *)arg,
2853 sizeof(space_args)))
2854 return -EFAULT;
2855
bf5fc093
JB		 2856 for (i = 0; i < num_types; i++) {
2857 struct btrfs_space_info *tmp;
2858
2859 info = NULL;
2860 rcu_read_lock();
2861 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2862 list) {
2863 if (tmp->flags == types[i]) {
2864 info = tmp;
2865 break;
2866 }
2867 }
2868 rcu_read_unlock();
2869
2870 if (!info)
2871 continue;
2872
2873 down_read(&info->groups_sem);
2874 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2875 if (!list_empty(&info->block_groups[c]))
2876 slot_count++;
2877 }
2878 up_read(&info->groups_sem);
2879 }
7fde62bf
CM		 2880
2881 /* space_slots == 0 means they are asking for a count */
2882 if (space_args.space_slots == 0) {
2883 space_args.total_spaces = slot_count;
2884 goto out;
2885 }
bf5fc093 2886
51788b1b 2887 slot_count = min_t(u64, space_args.space_slots, slot_count);
bf5fc093 2888
7fde62bf 2889 alloc_size = sizeof(*dest) * slot_count;
bf5fc093 2890
7fde62bf
CM		 2891 /* we generally have at most 6 or so space infos, one for each raid
2892 * level. So, a whole page should be more than enough for everyone
2893 */
2894 if (alloc_size > PAGE_CACHE_SIZE)
2895 return -ENOMEM;
2896
1406e432 2897 space_args.total_spaces = 0;
7fde62bf
CM		 2898 dest = kmalloc(alloc_size, GFP_NOFS);
2899 if (!dest)
2900 return -ENOMEM;
2901 dest_orig = dest;
1406e432 2902
7fde62bf 2903 /* now we have a buffer to copy into */
bf5fc093
JB		 2904 for (i = 0; i < num_types; i++) {
2905 struct btrfs_space_info *tmp;
2906
51788b1b
DR		 2907 if (!slot_count)
2908 break;
2909
bf5fc093
JB		 2910 info = NULL;
2911 rcu_read_lock();
2912 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2913 list) {
2914 if (tmp->flags == types[i]) {
2915 info = tmp;
2916 break;
2917 }
2918 }
2919 rcu_read_unlock();
7fde62bf 2920
bf5fc093
JB		 2921 if (!info)
2922 continue;
2923 down_read(&info->groups_sem);
2924 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2925 if (!list_empty(&info->block_groups[c])) {
2926 get_block_group_info(&info->block_groups[c],
2927 &space);
2928 memcpy(dest, &space, sizeof(space));
2929 dest++;
2930 space_args.total_spaces++;
51788b1b 2931 slot_count--;
bf5fc093 2932 }
51788b1b
DR		 2933 if (!slot_count)
2934 break;
bf5fc093
JB		 2935 }
2936 up_read(&info->groups_sem);
1406e432 2937 }
1406e432 2938
2eec6c81 2939 user_dest = (struct btrfs_ioctl_space_info __user *)
7fde62bf
CM		 2940 (arg + sizeof(struct btrfs_ioctl_space_args));
2941
2942 if (copy_to_user(user_dest, dest_orig, alloc_size))
2943 ret = -EFAULT;
2944
2945 kfree(dest_orig);
2946out:
2947 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
1406e432
JB		 2948 ret = -EFAULT;
2949
2950 return ret;
2951}
2952
f46b5a66
CH		 2953/*
2954 * there are many ways the trans_start and trans_end ioctls can lead
2955 * to deadlocks. They should only be used by applications that
2956 * basically own the machine, and have a very in depth understanding
2957 * of all the possible deadlocks and enospc problems.
2958 */
2959long btrfs_ioctl_trans_end(struct file *file)
2960{
2961 struct inode *inode = fdentry(file)->d_inode;
2962 struct btrfs_root *root = BTRFS_I(inode)->root;
2963 struct btrfs_trans_handle *trans;
f46b5a66 2964
f46b5a66 2965 trans = file->private_data;
1ab86aed
SW		 2966 if (!trans)
2967 return -EINVAL;
b214107e 2968 file->private_data = NULL;
9ca9ee09 2969
1ab86aed
SW		 2970 btrfs_end_transaction(trans, root);
2971
a4abeea4 2972 atomic_dec(&root->fs_info->open_ioctl_trans);
9ca9ee09 2973
2a79f17e 2974 mnt_drop_write_file(file);
1ab86aed 2975 return 0;
f46b5a66
CH		 2976}
2977
46204592
SW		 2978static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
2979{
2980 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2981 struct btrfs_trans_handle *trans;
2982 u64 transid;
db5b493a 2983 int ret;
46204592
SW		 2984
2985 trans = btrfs_start_transaction(root, 0);
98d5dc13
TI		 2986 if (IS_ERR(trans))
2987 return PTR_ERR(trans);
46204592 2988 transid = trans->transid;
db5b493a 2989 ret = btrfs_commit_transaction_async(trans, root, 0);
8b2b2d3c
TI		 2990 if (ret) {
2991 btrfs_end_transaction(trans, root);
db5b493a 2992 return ret;
8b2b2d3c 2993 }
46204592
SW		 2994
2995 if (argp)
2996 if (copy_to_user(argp, &transid, sizeof(transid)))
2997 return -EFAULT;
2998 return 0;
2999}
3000
3001static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
3002{
3003 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
3004 u64 transid;
3005
3006 if (argp) {
3007 if (copy_from_user(&transid, argp, sizeof(transid)))
3008 return -EFAULT;
3009 } else {
3010 transid = 0; /* current trans */
3011 }
3012 return btrfs_wait_for_commit(root, transid);
3013}
3014
475f6387
JS		 3015static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg)
3016{
3017 int ret;
3018 struct btrfs_ioctl_scrub_args *sa;
3019
3020 if (!capable(CAP_SYS_ADMIN))
3021 return -EPERM;
3022
3023 sa = memdup_user(arg, sizeof(*sa));
3024 if (IS_ERR(sa))
3025 return PTR_ERR(sa);
3026
3027 ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end,
8628764e 3028 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY);
475f6387
JS		 3029
3030 if (copy_to_user(arg, sa, sizeof(*sa)))
3031 ret = -EFAULT;
3032
3033 kfree(sa);
3034 return ret;
3035}
3036
3037static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3038{
3039 if (!capable(CAP_SYS_ADMIN))
3040 return -EPERM;
3041
3042 return btrfs_scrub_cancel(root);
3043}
3044
3045static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3046 void __user *arg)
3047{
3048 struct btrfs_ioctl_scrub_args *sa;
3049 int ret;
3050
3051 if (!capable(CAP_SYS_ADMIN))
3052 return -EPERM;
3053
3054 sa = memdup_user(arg, sizeof(*sa));
3055 if (IS_ERR(sa))
3056 return PTR_ERR(sa);
3057
3058 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3059
3060 if (copy_to_user(arg, sa, sizeof(*sa)))
3061 ret = -EFAULT;
3062
3063 kfree(sa);
3064 return ret;
3065}
3066
c11d2c23
SB		 3067static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
3068 void __user *arg, int reset_after_read)
3069{
3070 struct btrfs_ioctl_get_dev_stats *sa;
3071 int ret;
3072
3073 if (reset_after_read && !capable(CAP_SYS_ADMIN))
3074 return -EPERM;
3075
3076 sa = memdup_user(arg, sizeof(*sa));
3077 if (IS_ERR(sa))
3078 return PTR_ERR(sa);
3079
3080 ret = btrfs_get_dev_stats(root, sa, reset_after_read);
3081
3082 if (copy_to_user(arg, sa, sizeof(*sa)))
3083 ret = -EFAULT;
3084
3085 kfree(sa);
3086 return ret;
3087}
3088
d7728c96
JS		 3089static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3090{
3091 int ret = 0;
3092 int i;
740c3d22 3093 u64 rel_ptr;
d7728c96 3094 int size;
806468f8 3095 struct btrfs_ioctl_ino_path_args *ipa = NULL;
d7728c96
JS		 3096 struct inode_fs_paths *ipath = NULL;
3097 struct btrfs_path *path;
3098
3099 if (!capable(CAP_SYS_ADMIN))
3100 return -EPERM;
3101
3102 path = btrfs_alloc_path();
3103 if (!path) {
3104 ret = -ENOMEM;
3105 goto out;
3106 }
3107
3108 ipa = memdup_user(arg, sizeof(*ipa));
3109 if (IS_ERR(ipa)) {
3110 ret = PTR_ERR(ipa);
3111 ipa = NULL;
3112 goto out;
3113 }
3114
3115 size = min_t(u32, ipa->size, 4096);
3116 ipath = init_ipath(size, root, path);
3117 if (IS_ERR(ipath)) {
3118 ret = PTR_ERR(ipath);
3119 ipath = NULL;
3120 goto out;
3121 }
3122
3123 ret = paths_from_inode(ipa->inum, ipath);
3124 if (ret < 0)
3125 goto out;
3126
3127 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
745c4d8e
JM		 3128 rel_ptr = ipath->fspath->val[i] -
3129 (u64)(unsigned long)ipath->fspath->val;
740c3d22 3130 ipath->fspath->val[i] = rel_ptr;
d7728c96
JS		 3131 }
3132
745c4d8e
JM		 3133 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3134 (void *)(unsigned long)ipath->fspath, size);
d7728c96
JS		 3135 if (ret) {
3136 ret = -EFAULT;
3137 goto out;
3138 }
3139
3140out:
3141 btrfs_free_path(path);
3142 free_ipath(ipath);
3143 kfree(ipa);
3144
3145 return ret;
3146}
3147
3148static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3149{
3150 struct btrfs_data_container *inodes = ctx;
3151 const size_t c = 3 * sizeof(u64);
3152
3153 if (inodes->bytes_left >= c) {
3154 inodes->bytes_left -= c;
3155 inodes->val[inodes->elem_cnt] = inum;
3156 inodes->val[inodes->elem_cnt + 1] = offset;
3157 inodes->val[inodes->elem_cnt + 2] = root;
3158 inodes->elem_cnt += 3;
3159 } else {
3160 inodes->bytes_missing += c - inodes->bytes_left;
3161 inodes->bytes_left = 0;
3162 inodes->elem_missed += 3;
3163 }
3164
3165 return 0;
3166}
3167
3168static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3169 void __user *arg)
3170{
3171 int ret = 0;
3172 int size;
4692cf58 3173 u64 extent_item_pos;
d7728c96
JS		 3174 struct btrfs_ioctl_logical_ino_args *loi;
3175 struct btrfs_data_container *inodes = NULL;
3176 struct btrfs_path *path = NULL;
3177 struct btrfs_key key;
3178
3179 if (!capable(CAP_SYS_ADMIN))
3180 return -EPERM;
3181
3182 loi = memdup_user(arg, sizeof(*loi));
3183 if (IS_ERR(loi)) {
3184 ret = PTR_ERR(loi);
3185 loi = NULL;
3186 goto out;
3187 }
3188
3189 path = btrfs_alloc_path();
3190 if (!path) {
3191 ret = -ENOMEM;
3192 goto out;
3193 }
3194
3195 size = min_t(u32, loi->size, 4096);
3196 inodes = init_data_container(size);
3197 if (IS_ERR(inodes)) {
3198 ret = PTR_ERR(inodes);
3199 inodes = NULL;
3200 goto out;
3201 }
3202
3203 ret = extent_from_logical(root->fs_info, loi->logical, path, &key);
4692cf58 3204 btrfs_release_path(path);
d7728c96
JS		 3205
3206 if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK)
3207 ret = -ENOENT;
3208 if (ret < 0)
3209 goto out;
3210
4692cf58 3211 extent_item_pos = loi->logical - key.objectid;
7a3ae2f8
JS		 3212 ret = iterate_extent_inodes(root->fs_info, key.objectid,
3213 extent_item_pos, 0, build_ino_list,
4692cf58 3214 inodes);
d7728c96
JS		 3215
3216 if (ret < 0)
3217 goto out;
3218
745c4d8e
JM		 3219 ret = copy_to_user((void *)(unsigned long)loi->inodes,
3220 (void *)(unsigned long)inodes, size);
d7728c96
JS		 3221 if (ret)
3222 ret = -EFAULT;
3223
3224out:
3225 btrfs_free_path(path);
3226 kfree(inodes);
3227 kfree(loi);
3228
3229 return ret;
3230}
3231
19a39dce 3232void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
c9e9f97b
ID		 3233 struct btrfs_ioctl_balance_args *bargs)
3234{
3235 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3236
3237 bargs->flags = bctl->flags;
3238
837d5b6e
ID		 3239 if (atomic_read(&fs_info->balance_running))
3240 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3241 if (atomic_read(&fs_info->balance_pause_req))
3242 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
a7e99c69
ID		 3243 if (atomic_read(&fs_info->balance_cancel_req))
3244 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
837d5b6e 3245
c9e9f97b
ID		 3246 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3247 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3248 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
19a39dce
ID		 3249
3250 if (lock) {
3251 spin_lock(&fs_info->balance_lock);
3252 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3253 spin_unlock(&fs_info->balance_lock);
3254 } else {
3255 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3256 }
c9e9f97b
ID		 3257}
3258
9ba1f6e4 3259static long btrfs_ioctl_balance(struct file *file, void __user *arg)
c9e9f97b 3260{
9ba1f6e4 3261 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
c9e9f97b
ID		 3262 struct btrfs_fs_info *fs_info = root->fs_info;
3263 struct btrfs_ioctl_balance_args *bargs;
3264 struct btrfs_balance_control *bctl;
3265 int ret;
3266
3267 if (!capable(CAP_SYS_ADMIN))
3268 return -EPERM;
3269
3270 if (fs_info->sb->s_flags & MS_RDONLY)
3271 return -EROFS;
3272
9ba1f6e4
LB		 3273 ret = mnt_want_write(file->f_path.mnt);
3274 if (ret)
3275 return ret;
3276
c9e9f97b
ID		 3277 mutex_lock(&fs_info->volume_mutex);
3278 mutex_lock(&fs_info->balance_mutex);
3279
3280 if (arg) {
3281 bargs = memdup_user(arg, sizeof(*bargs));
3282 if (IS_ERR(bargs)) {
3283 ret = PTR_ERR(bargs);
3284 goto out;
3285 }
de322263
ID		 3286
3287 if (bargs->flags & BTRFS_BALANCE_RESUME) {
3288 if (!fs_info->balance_ctl) {
3289 ret = -ENOTCONN;
3290 goto out_bargs;
3291 }
3292
3293 bctl = fs_info->balance_ctl;
3294 spin_lock(&fs_info->balance_lock);
3295 bctl->flags |= BTRFS_BALANCE_RESUME;
3296 spin_unlock(&fs_info->balance_lock);
3297
3298 goto do_balance;
3299 }
c9e9f97b
ID		 3300 } else {
3301 bargs = NULL;
3302 }
3303
837d5b6e
ID		 3304 if (fs_info->balance_ctl) {
3305 ret = -EINPROGRESS;
3306 goto out_bargs;
3307 }
3308
c9e9f97b
ID		 3309 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3310 if (!bctl) {
3311 ret = -ENOMEM;
3312 goto out_bargs;
3313 }
3314
3315 bctl->fs_info = fs_info;
3316 if (arg) {
3317 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3318 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3319 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3320
3321 bctl->flags = bargs->flags;
f43ffb60
ID		 3322 } else {
3323 /* balance everything - no filters */
3324 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
c9e9f97b
ID		 3325 }
3326
de322263 3327do_balance:
c9e9f97b
ID		 3328 ret = btrfs_balance(bctl, bargs);
3329 /*
837d5b6e
ID		 3330 * bctl is freed in __cancel_balance or in free_fs_info if
3331 * restriper was paused all the way until unmount
c9e9f97b
ID		 3332 */
3333 if (arg) {
3334 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3335 ret = -EFAULT;
3336 }
3337
3338out_bargs:
3339 kfree(bargs);
3340out:
3341 mutex_unlock(&fs_info->balance_mutex);
3342 mutex_unlock(&fs_info->volume_mutex);
9ba1f6e4 3343 mnt_drop_write(file->f_path.mnt);
c9e9f97b
ID		 3344 return ret;
3345}
3346
837d5b6e
ID		 3347static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
3348{
3349 if (!capable(CAP_SYS_ADMIN))
3350 return -EPERM;
3351
3352 switch (cmd) {
3353 case BTRFS_BALANCE_CTL_PAUSE:
3354 return btrfs_pause_balance(root->fs_info);
a7e99c69
ID		 3355 case BTRFS_BALANCE_CTL_CANCEL:
3356 return btrfs_cancel_balance(root->fs_info);
837d5b6e
ID		 3357 }
3358
3359 return -EINVAL;
3360}
3361
19a39dce
ID		 3362static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
3363 void __user *arg)
3364{
3365 struct btrfs_fs_info *fs_info = root->fs_info;
3366 struct btrfs_ioctl_balance_args *bargs;
3367 int ret = 0;
3368
3369 if (!capable(CAP_SYS_ADMIN))
3370 return -EPERM;
3371
3372 mutex_lock(&fs_info->balance_mutex);
3373 if (!fs_info->balance_ctl) {
3374 ret = -ENOTCONN;
3375 goto out;
3376 }
3377
3378 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
3379 if (!bargs) {
3380 ret = -ENOMEM;
3381 goto out;
3382 }
3383
3384 update_ioctl_balance_args(fs_info, 1, bargs);
3385
3386 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3387 ret = -EFAULT;
3388
3389 kfree(bargs);
3390out:
3391 mutex_unlock(&fs_info->balance_mutex);
3392 return ret;
3393}
3394
f46b5a66
CH		 3395long btrfs_ioctl(struct file *file, unsigned int
3396 cmd, unsigned long arg)
3397{
3398 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
4bcabaa3 3399 void __user *argp = (void __user *)arg;
f46b5a66
CH		 3400
3401 switch (cmd) {
6cbff00f
CH		 3402 case FS_IOC_GETFLAGS:
3403 return btrfs_ioctl_getflags(file, argp);
3404 case FS_IOC_SETFLAGS:
3405 return btrfs_ioctl_setflags(file, argp);
3406 case FS_IOC_GETVERSION:
3407 return btrfs_ioctl_getversion(file, argp);
f7039b1d
LD		 3408 case FITRIM:
3409 return btrfs_ioctl_fitrim(file, argp);
f46b5a66 3410 case BTRFS_IOC_SNAP_CREATE:
fa0d2b9b 3411 return btrfs_ioctl_snap_create(file, argp, 0);
fdfb1e4f 3412 case BTRFS_IOC_SNAP_CREATE_V2:
fa0d2b9b 3413 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3de4586c 3414 case BTRFS_IOC_SUBVOL_CREATE:
fa0d2b9b 3415 return btrfs_ioctl_snap_create(file, argp, 1);
76dda93c
YZ		 3416 case BTRFS_IOC_SNAP_DESTROY:
3417 return btrfs_ioctl_snap_destroy(file, argp);
0caa102d
LZ		 3418 case BTRFS_IOC_SUBVOL_GETFLAGS:
3419 return btrfs_ioctl_subvol_getflags(file, argp);
3420 case BTRFS_IOC_SUBVOL_SETFLAGS:
3421 return btrfs_ioctl_subvol_setflags(file, argp);
6ef5ed0d
JB		 3422 case BTRFS_IOC_DEFAULT_SUBVOL:
3423 return btrfs_ioctl_default_subvol(file, argp);
f46b5a66 3424 case BTRFS_IOC_DEFRAG:
1e701a32
CM		 3425 return btrfs_ioctl_defrag(file, NULL);
3426 case BTRFS_IOC_DEFRAG_RANGE:
3427 return btrfs_ioctl_defrag(file, argp);
f46b5a66 3428 case BTRFS_IOC_RESIZE:
4bcabaa3 3429 return btrfs_ioctl_resize(root, argp);
f46b5a66 3430 case BTRFS_IOC_ADD_DEV:
4bcabaa3 3431 return btrfs_ioctl_add_dev(root, argp);
f46b5a66 3432 case BTRFS_IOC_RM_DEV:
4bcabaa3 3433 return btrfs_ioctl_rm_dev(root, argp);
475f6387
JS		 3434 case BTRFS_IOC_FS_INFO:
3435 return btrfs_ioctl_fs_info(root, argp);
3436 case BTRFS_IOC_DEV_INFO:
3437 return btrfs_ioctl_dev_info(root, argp);
f46b5a66 3438 case BTRFS_IOC_BALANCE:
9ba1f6e4 3439 return btrfs_ioctl_balance(file, NULL);
f46b5a66 3440 case BTRFS_IOC_CLONE:
c5c9cd4d
SW		 3441 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
3442 case BTRFS_IOC_CLONE_RANGE:
7a865e8a 3443 return btrfs_ioctl_clone_range(file, argp);
f46b5a66
CH		 3444 case BTRFS_IOC_TRANS_START:
3445 return btrfs_ioctl_trans_start(file);
3446 case BTRFS_IOC_TRANS_END:
3447 return btrfs_ioctl_trans_end(file);
ac8e9819
CM		 3448 case BTRFS_IOC_TREE_SEARCH:
3449 return btrfs_ioctl_tree_search(file, argp);
3450 case BTRFS_IOC_INO_LOOKUP:
3451 return btrfs_ioctl_ino_lookup(file, argp);
d7728c96
JS		 3452 case BTRFS_IOC_INO_PATHS:
3453 return btrfs_ioctl_ino_to_path(root, argp);
3454 case BTRFS_IOC_LOGICAL_INO:
3455 return btrfs_ioctl_logical_to_ino(root, argp);
1406e432
JB		 3456 case BTRFS_IOC_SPACE_INFO:
3457 return btrfs_ioctl_space_info(root, argp);
f46b5a66
CH		 3458 case BTRFS_IOC_SYNC:
3459 btrfs_sync_fs(file->f_dentry->d_sb, 1);
3460 return 0;
46204592
SW		 3461 case BTRFS_IOC_START_SYNC:
3462 return btrfs_ioctl_start_sync(file, argp);
3463 case BTRFS_IOC_WAIT_SYNC:
3464 return btrfs_ioctl_wait_sync(file, argp);
475f6387
JS		 3465 case BTRFS_IOC_SCRUB:
3466 return btrfs_ioctl_scrub(root, argp);
3467 case BTRFS_IOC_SCRUB_CANCEL:
3468 return btrfs_ioctl_scrub_cancel(root, argp);
3469 case BTRFS_IOC_SCRUB_PROGRESS:
3470 return btrfs_ioctl_scrub_progress(root, argp);
c9e9f97b 3471 case BTRFS_IOC_BALANCE_V2:
9ba1f6e4 3472 return btrfs_ioctl_balance(file, argp);
837d5b6e
ID		 3473 case BTRFS_IOC_BALANCE_CTL:
3474 return btrfs_ioctl_balance_ctl(root, arg);
19a39dce
ID		 3475 case BTRFS_IOC_BALANCE_PROGRESS:
3476 return btrfs_ioctl_balance_progress(root, argp);
c11d2c23
SB		 3477 case BTRFS_IOC_GET_DEV_STATS:
3478 return btrfs_ioctl_get_dev_stats(root, argp, 0);
3479 case BTRFS_IOC_GET_AND_RESET_DEV_STATS:
3480 return btrfs_ioctl_get_dev_stats(root, argp, 1);
f46b5a66
CH		 3481 }
3482
3483 return -ENOTTY;
3484}
Linux 6.x block layer and io_uring tree(s)