]> Git Repo - linux.git/blame - fs/btrfs/ioctl.c
projects / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
btrfs: enhance transaction abort infrastructure
[linux.git] / fs / btrfs / ioctl.c
CommitLineData
f46b5a66
CH		 1/*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
19#include <linux/kernel.h>
20#include <linux/bio.h>
21#include <linux/buffer_head.h>
22#include <linux/file.h>
23#include <linux/fs.h>
cb8e7090 24#include <linux/fsnotify.h>
f46b5a66
CH		 25#include <linux/pagemap.h>
26#include <linux/highmem.h>
27#include <linux/time.h>
28#include <linux/init.h>
29#include <linux/string.h>
f46b5a66 30#include <linux/backing-dev.h>
cb8e7090 31#include <linux/mount.h>
f46b5a66 32#include <linux/mpage.h>
cb8e7090 33#include <linux/namei.h>
f46b5a66
CH		 34#include <linux/swap.h>
35#include <linux/writeback.h>
36#include <linux/statfs.h>
37#include <linux/compat.h>
38#include <linux/bit_spinlock.h>
cb8e7090 39#include <linux/security.h>
f46b5a66 40#include <linux/xattr.h>
7ea394f1 41#include <linux/vmalloc.h>
5a0e3ad6 42#include <linux/slab.h>
f7039b1d 43#include <linux/blkdev.h>
4b4e25f2 44#include "compat.h"
f46b5a66
CH		 45#include "ctree.h"
46#include "disk-io.h"
47#include "transaction.h"
48#include "btrfs_inode.h"
49#include "ioctl.h"
50#include "print-tree.h"
51#include "volumes.h"
925baedd 52#include "locking.h"
581bb050 53#include "inode-map.h"
d7728c96 54#include "backref.h"
f46b5a66 55
6cbff00f
CH		 56/* Mask out flags that are inappropriate for the given type of inode. */
57static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
58{
59 if (S_ISDIR(mode))
60 return flags;
61 else if (S_ISREG(mode))
62 return flags & ~FS_DIRSYNC_FL;
63 else
64 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
65}
66
67/*
68 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
69 */
70static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
71{
72 unsigned int iflags = 0;
73
74 if (flags & BTRFS_INODE_SYNC)
75 iflags |= FS_SYNC_FL;
76 if (flags & BTRFS_INODE_IMMUTABLE)
77 iflags |= FS_IMMUTABLE_FL;
78 if (flags & BTRFS_INODE_APPEND)
79 iflags |= FS_APPEND_FL;
80 if (flags & BTRFS_INODE_NODUMP)
81 iflags |= FS_NODUMP_FL;
82 if (flags & BTRFS_INODE_NOATIME)
83 iflags |= FS_NOATIME_FL;
84 if (flags & BTRFS_INODE_DIRSYNC)
85 iflags |= FS_DIRSYNC_FL;
d0092bdd
LZ		 86 if (flags & BTRFS_INODE_NODATACOW)
87 iflags |= FS_NOCOW_FL;
88
89 if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
90 iflags |= FS_COMPR_FL;
91 else if (flags & BTRFS_INODE_NOCOMPRESS)
92 iflags |= FS_NOCOMP_FL;
6cbff00f
CH		 93
94 return iflags;
95}
96
97/*
98 * Update inode->i_flags based on the btrfs internal flags.
99 */
100void btrfs_update_iflags(struct inode *inode)
101{
102 struct btrfs_inode *ip = BTRFS_I(inode);
103
104 inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
105
106 if (ip->flags & BTRFS_INODE_SYNC)
107 inode->i_flags |= S_SYNC;
108 if (ip->flags & BTRFS_INODE_IMMUTABLE)
109 inode->i_flags |= S_IMMUTABLE;
110 if (ip->flags & BTRFS_INODE_APPEND)
111 inode->i_flags |= S_APPEND;
112 if (ip->flags & BTRFS_INODE_NOATIME)
113 inode->i_flags |= S_NOATIME;
114 if (ip->flags & BTRFS_INODE_DIRSYNC)
115 inode->i_flags |= S_DIRSYNC;
116}
117
118/*
119 * Inherit flags from the parent inode.
120 *
e27425d6 121 * Currently only the compression flags and the cow flags are inherited.
6cbff00f
CH		 122 */
123void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
124{
0b4dcea5
CM		 125 unsigned int flags;
126
127 if (!dir)
128 return;
129
130 flags = BTRFS_I(dir)->flags;
6cbff00f 131
e27425d6
JB		 132 if (flags & BTRFS_INODE_NOCOMPRESS) {
133 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
134 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
135 } else if (flags & BTRFS_INODE_COMPRESS) {
136 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
137 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
138 }
139
140 if (flags & BTRFS_INODE_NODATACOW)
141 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
6cbff00f 142
6cbff00f
CH		 143 btrfs_update_iflags(inode);
144}
145
146static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
147{
148 struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode);
149 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
150
151 if (copy_to_user(arg, &flags, sizeof(flags)))
152 return -EFAULT;
153 return 0;
154}
155
75e7cb7f
LB		 156static int check_flags(unsigned int flags)
157{
158 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
159 FS_NOATIME_FL | FS_NODUMP_FL | \
160 FS_SYNC_FL | FS_DIRSYNC_FL | \
e1e8fb6a
LZ		 161 FS_NOCOMP_FL | FS_COMPR_FL |
162 FS_NOCOW_FL))
75e7cb7f
LB		 163 return -EOPNOTSUPP;
164
165 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
166 return -EINVAL;
167
75e7cb7f
LB		 168 return 0;
169}
170
6cbff00f
CH		 171static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
172{
173 struct inode *inode = file->f_path.dentry->d_inode;
174 struct btrfs_inode *ip = BTRFS_I(inode);
175 struct btrfs_root *root = ip->root;
176 struct btrfs_trans_handle *trans;
177 unsigned int flags, oldflags;
178 int ret;
f062abf0
LZ		 179 u64 ip_oldflags;
180 unsigned int i_oldflags;
6cbff00f 181
b83cc969
LZ		 182 if (btrfs_root_readonly(root))
183 return -EROFS;
184
6cbff00f
CH		 185 if (copy_from_user(&flags, arg, sizeof(flags)))
186 return -EFAULT;
187
75e7cb7f
LB		 188 ret = check_flags(flags);
189 if (ret)
190 return ret;
f46b5a66 191
2e149670 192 if (!inode_owner_or_capable(inode))
6cbff00f
CH		 193 return -EACCES;
194
195 mutex_lock(&inode->i_mutex);
196
f062abf0
LZ		 197 ip_oldflags = ip->flags;
198 i_oldflags = inode->i_flags;
199
6cbff00f
CH		 200 flags = btrfs_mask_flags(inode->i_mode, flags);
201 oldflags = btrfs_flags_to_ioctl(ip->flags);
202 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
203 if (!capable(CAP_LINUX_IMMUTABLE)) {
204 ret = -EPERM;
205 goto out_unlock;
206 }
207 }
208
a561be71 209 ret = mnt_want_write_file(file);
6cbff00f
CH		 210 if (ret)
211 goto out_unlock;
212
213 if (flags & FS_SYNC_FL)
214 ip->flags |= BTRFS_INODE_SYNC;
215 else
216 ip->flags &= ~BTRFS_INODE_SYNC;
217 if (flags & FS_IMMUTABLE_FL)
218 ip->flags |= BTRFS_INODE_IMMUTABLE;
219 else
220 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
221 if (flags & FS_APPEND_FL)
222 ip->flags |= BTRFS_INODE_APPEND;
223 else
224 ip->flags &= ~BTRFS_INODE_APPEND;
225 if (flags & FS_NODUMP_FL)
226 ip->flags |= BTRFS_INODE_NODUMP;
227 else
228 ip->flags &= ~BTRFS_INODE_NODUMP;
229 if (flags & FS_NOATIME_FL)
230 ip->flags |= BTRFS_INODE_NOATIME;
231 else
232 ip->flags &= ~BTRFS_INODE_NOATIME;
233 if (flags & FS_DIRSYNC_FL)
234 ip->flags |= BTRFS_INODE_DIRSYNC;
235 else
236 ip->flags &= ~BTRFS_INODE_DIRSYNC;
e1e8fb6a
LZ		 237 if (flags & FS_NOCOW_FL)
238 ip->flags |= BTRFS_INODE_NODATACOW;
239 else
240 ip->flags &= ~BTRFS_INODE_NODATACOW;
6cbff00f 241
75e7cb7f
LB		 242 /*
243 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
244 * flag may be changed automatically if compression code won't make
245 * things smaller.
246 */
247 if (flags & FS_NOCOMP_FL) {
248 ip->flags &= ~BTRFS_INODE_COMPRESS;
249 ip->flags |= BTRFS_INODE_NOCOMPRESS;
250 } else if (flags & FS_COMPR_FL) {
251 ip->flags |= BTRFS_INODE_COMPRESS;
252 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
ebcb904d
LZ		 253 } else {
254 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
75e7cb7f 255 }
6cbff00f 256
4da6f1a3 257 trans = btrfs_start_transaction(root, 1);
f062abf0
LZ		 258 if (IS_ERR(trans)) {
259 ret = PTR_ERR(trans);
260 goto out_drop;
261 }
6cbff00f 262
306424cc
LZ		 263 btrfs_update_iflags(inode);
264 inode->i_ctime = CURRENT_TIME;
6cbff00f 265 ret = btrfs_update_inode(trans, root, inode);
6cbff00f 266
6cbff00f 267 btrfs_end_transaction(trans, root);
f062abf0
LZ		 268 out_drop:
269 if (ret) {
270 ip->flags = ip_oldflags;
271 inode->i_flags = i_oldflags;
272 }
6cbff00f 273
2a79f17e 274 mnt_drop_write_file(file);
6cbff00f
CH		 275 out_unlock:
276 mutex_unlock(&inode->i_mutex);
2d4e6f6a 277 return ret;
6cbff00f
CH		 278}
279
280static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
281{
282 struct inode *inode = file->f_path.dentry->d_inode;
283
284 return put_user(inode->i_generation, arg);
285}
f46b5a66 286
f7039b1d
LD		 287static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
288{
815745cf 289 struct btrfs_fs_info *fs_info = btrfs_sb(fdentry(file)->d_sb);
f7039b1d
LD		 290 struct btrfs_device *device;
291 struct request_queue *q;
292 struct fstrim_range range;
293 u64 minlen = ULLONG_MAX;
294 u64 num_devices = 0;
815745cf 295 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
f7039b1d
LD		 296 int ret;
297
298 if (!capable(CAP_SYS_ADMIN))
299 return -EPERM;
300
1f78160c
XG		 301 rcu_read_lock();
302 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
303 dev_list) {
f7039b1d
LD		 304 if (!device->bdev)
305 continue;
306 q = bdev_get_queue(device->bdev);
307 if (blk_queue_discard(q)) {
308 num_devices++;
309 minlen = min((u64)q->limits.discard_granularity,
310 minlen);
311 }
312 }
1f78160c 313 rcu_read_unlock();
f4c697e6 314
f7039b1d
LD		 315 if (!num_devices)
316 return -EOPNOTSUPP;
f7039b1d
LD		 317 if (copy_from_user(&range, arg, sizeof(range)))
318 return -EFAULT;
f4c697e6
LC		 319 if (range.start > total_bytes)
320 return -EINVAL;
f7039b1d 321
f4c697e6 322 range.len = min(range.len, total_bytes - range.start);
f7039b1d 323 range.minlen = max(range.minlen, minlen);
815745cf 324 ret = btrfs_trim_fs(fs_info->tree_root, &range);
f7039b1d
LD		 325 if (ret < 0)
326 return ret;
327
328 if (copy_to_user(arg, &range, sizeof(range)))
329 return -EFAULT;
330
331 return 0;
332}
333
cb8e7090
CH		 334static noinline int create_subvol(struct btrfs_root *root,
335 struct dentry *dentry,
72fd032e
SW		 336 char *name, int namelen,
337 u64 *async_transid)
f46b5a66
CH		 338{
339 struct btrfs_trans_handle *trans;
340 struct btrfs_key key;
341 struct btrfs_root_item root_item;
342 struct btrfs_inode_item *inode_item;
343 struct extent_buffer *leaf;
76dda93c 344 struct btrfs_root *new_root;
2fbe8c8a 345 struct dentry *parent = dentry->d_parent;
6a912213 346 struct inode *dir;
f46b5a66
CH		 347 int ret;
348 int err;
349 u64 objectid;
350 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
3de4586c 351 u64 index = 0;
f46b5a66 352
581bb050 353 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
2fbe8c8a 354 if (ret)
a22285a6 355 return ret;
6a912213
JB		 356
357 dir = parent->d_inode;
358
9ed74f2d
JB		 359 /*
360 * 1 - inode item
361 * 2 - refs
362 * 1 - root item
363 * 2 - dir items
364 */
a22285a6 365 trans = btrfs_start_transaction(root, 6);
2fbe8c8a 366 if (IS_ERR(trans))
a22285a6 367 return PTR_ERR(trans);
f46b5a66 368
5d4f98a2 369 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
66d7e7f0 370 0, objectid, NULL, 0, 0, 0, 0);
8e8a1e31
JB		 371 if (IS_ERR(leaf)) {
372 ret = PTR_ERR(leaf);
373 goto fail;
374 }
f46b5a66 375
5d4f98a2 376 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
f46b5a66
CH		 377 btrfs_set_header_bytenr(leaf, leaf->start);
378 btrfs_set_header_generation(leaf, trans->transid);
5d4f98a2 379 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
f46b5a66
CH		 380 btrfs_set_header_owner(leaf, objectid);
381
382 write_extent_buffer(leaf, root->fs_info->fsid,
383 (unsigned long)btrfs_header_fsid(leaf),
384 BTRFS_FSID_SIZE);
5d4f98a2
YZ		 385 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
386 (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
387 BTRFS_UUID_SIZE);
f46b5a66
CH		 388 btrfs_mark_buffer_dirty(leaf);
389
390 inode_item = &root_item.inode;
391 memset(inode_item, 0, sizeof(*inode_item));
392 inode_item->generation = cpu_to_le64(1);
393 inode_item->size = cpu_to_le64(3);
394 inode_item->nlink = cpu_to_le32(1);
a76a3cd4 395 inode_item->nbytes = cpu_to_le64(root->leafsize);
f46b5a66
CH		 396 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
397
08fe4db1
LZ		 398 root_item.flags = 0;
399 root_item.byte_limit = 0;
400 inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
401
f46b5a66 402 btrfs_set_root_bytenr(&root_item, leaf->start);
84234f3a 403 btrfs_set_root_generation(&root_item, trans->transid);
f46b5a66
CH		 404 btrfs_set_root_level(&root_item, 0);
405 btrfs_set_root_refs(&root_item, 1);
86b9f2ec 406 btrfs_set_root_used(&root_item, leaf->len);
80ff3856 407 btrfs_set_root_last_snapshot(&root_item, 0);
f46b5a66
CH		 408
409 memset(&root_item.drop_progress, 0, sizeof(root_item.drop_progress));
410 root_item.drop_level = 0;
411
925baedd 412 btrfs_tree_unlock(leaf);
f46b5a66
CH		 413 free_extent_buffer(leaf);
414 leaf = NULL;
415
416 btrfs_set_root_dirid(&root_item, new_dirid);
417
418 key.objectid = objectid;
5d4f98a2 419 key.offset = 0;
f46b5a66
CH		 420 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
421 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
422 &root_item);
423 if (ret)
424 goto fail;
425
76dda93c
YZ		 426 key.offset = (u64)-1;
427 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
428 BUG_ON(IS_ERR(new_root));
429
430 btrfs_record_root_in_trans(trans, new_root);
431
d82a6f1d 432 ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
ce598979
MF		 433 if (ret) {
434 /* We potentially lose an unused inode item here */
435 goto fail;
436 }
437
f46b5a66
CH		 438 /*
439 * insert the directory item
440 */
3de4586c
CM		 441 ret = btrfs_set_inode_index(dir, &index);
442 BUG_ON(ret);
443
444 ret = btrfs_insert_dir_item(trans, root,
16cdcec7 445 name, namelen, dir, &key,
3de4586c 446 BTRFS_FT_DIR, index);
f46b5a66
CH		 447 if (ret)
448 goto fail;
0660b5af 449
52c26179
YZ		 450 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
451 ret = btrfs_update_inode(trans, root, dir);
452 BUG_ON(ret);
453
0660b5af 454 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
4df27c4d 455 objectid, root->root_key.objectid,
33345d01 456 btrfs_ino(dir), index, name, namelen);
0660b5af 457
76dda93c 458 BUG_ON(ret);
f46b5a66 459
76dda93c 460 d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
f46b5a66 461fail:
72fd032e
SW		 462 if (async_transid) {
463 *async_transid = trans->transid;
464 err = btrfs_commit_transaction_async(trans, root, 1);
465 } else {
466 err = btrfs_commit_transaction(trans, root);
467 }
f46b5a66
CH		 468 if (err && !ret)
469 ret = err;
f46b5a66
CH		 470 return ret;
471}
472
72fd032e 473static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
b83cc969
LZ		 474 char *name, int namelen, u64 *async_transid,
475 bool readonly)
f46b5a66 476{
2e4bfab9 477 struct inode *inode;
f46b5a66
CH		 478 struct btrfs_pending_snapshot *pending_snapshot;
479 struct btrfs_trans_handle *trans;
2e4bfab9 480 int ret;
f46b5a66
CH		 481
482 if (!root->ref_cows)
483 return -EINVAL;
484
3de4586c 485 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
a22285a6
YZ		 486 if (!pending_snapshot)
487 return -ENOMEM;
488
489 btrfs_init_block_rsv(&pending_snapshot->block_rsv);
3de4586c 490 pending_snapshot->dentry = dentry;
f46b5a66 491 pending_snapshot->root = root;
b83cc969 492 pending_snapshot->readonly = readonly;
a22285a6
YZ		 493
494 trans = btrfs_start_transaction(root->fs_info->extent_root, 5);
495 if (IS_ERR(trans)) {
496 ret = PTR_ERR(trans);
497 goto fail;
498 }
499
500 ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
501 BUG_ON(ret);
502
8351583e 503 spin_lock(&root->fs_info->trans_lock);
f46b5a66
CH		 504 list_add(&pending_snapshot->list,
505 &trans->transaction->pending_snapshots);
8351583e 506 spin_unlock(&root->fs_info->trans_lock);
72fd032e
SW		 507 if (async_transid) {
508 *async_transid = trans->transid;
509 ret = btrfs_commit_transaction_async(trans,
510 root->fs_info->extent_root, 1);
511 } else {
512 ret = btrfs_commit_transaction(trans,
513 root->fs_info->extent_root);
514 }
2e4bfab9 515 BUG_ON(ret);
a22285a6
YZ		 516
517 ret = pending_snapshot->error;
518 if (ret)
519 goto fail;
520
66b4ffd1
JB		 521 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
522 if (ret)
523 goto fail;
f46b5a66 524
2fbe8c8a 525 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
2e4bfab9
YZ		 526 if (IS_ERR(inode)) {
527 ret = PTR_ERR(inode);
528 goto fail;
529 }
530 BUG_ON(!inode);
531 d_instantiate(dentry, inode);
532 ret = 0;
533fail:
a22285a6 534 kfree(pending_snapshot);
f46b5a66
CH		 535 return ret;
536}
537
4260f7c7
SW		 538/* copy of check_sticky in fs/namei.c()
539* It's inline, so penalty for filesystems that don't use sticky bit is
540* minimal.
541*/
542static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
543{
544 uid_t fsuid = current_fsuid();
545
546 if (!(dir->i_mode & S_ISVTX))
547 return 0;
548 if (inode->i_uid == fsuid)
549 return 0;
550 if (dir->i_uid == fsuid)
551 return 0;
552 return !capable(CAP_FOWNER);
553}
554
555/* copy of may_delete in fs/namei.c()
556 * Check whether we can remove a link victim from directory dir, check
557 * whether the type of victim is right.
558 * 1. We can't do it if dir is read-only (done in permission())
559 * 2. We should have write and exec permissions on dir
560 * 3. We can't remove anything from append-only dir
561 * 4. We can't do anything with immutable dir (done in permission())
562 * 5. If the sticky bit on dir is set we should either
563 * a. be owner of dir, or
564 * b. be owner of victim, or
565 * c. have CAP_FOWNER capability
566 * 6. If the victim is append-only or immutable we can't do antyhing with
567 * links pointing to it.
568 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
569 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
570 * 9. We can't remove a root or mountpoint.
571 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
572 * nfs_async_unlink().
573 */
574
575static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
576{
577 int error;
578
579 if (!victim->d_inode)
580 return -ENOENT;
581
582 BUG_ON(victim->d_parent->d_inode != dir);
583 audit_inode_child(victim, dir);
584
585 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
586 if (error)
587 return error;
588 if (IS_APPEND(dir))
589 return -EPERM;
590 if (btrfs_check_sticky(dir, victim->d_inode)||
591 IS_APPEND(victim->d_inode)||
592 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
593 return -EPERM;
594 if (isdir) {
595 if (!S_ISDIR(victim->d_inode->i_mode))
596 return -ENOTDIR;
597 if (IS_ROOT(victim))
598 return -EBUSY;
599 } else if (S_ISDIR(victim->d_inode->i_mode))
600 return -EISDIR;
601 if (IS_DEADDIR(dir))
602 return -ENOENT;
603 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
604 return -EBUSY;
605 return 0;
606}
607
cb8e7090
CH		 608/* copy of may_create in fs/namei.c() */
609static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
610{
611 if (child->d_inode)
612 return -EEXIST;
613 if (IS_DEADDIR(dir))
614 return -ENOENT;
615 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
616}
617
618/*
619 * Create a new subvolume below @parent. This is largely modeled after
620 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
621 * inside this filesystem so it's quite a bit simpler.
622 */
76dda93c
YZ		 623static noinline int btrfs_mksubvol(struct path *parent,
624 char *name, int namelen,
72fd032e 625 struct btrfs_root *snap_src,
b83cc969 626 u64 *async_transid, bool readonly)
cb8e7090 627{
76dda93c 628 struct inode *dir = parent->dentry->d_inode;
cb8e7090
CH		 629 struct dentry *dentry;
630 int error;
631
76dda93c 632 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
cb8e7090
CH		 633
634 dentry = lookup_one_len(name, parent->dentry, namelen);
635 error = PTR_ERR(dentry);
636 if (IS_ERR(dentry))
637 goto out_unlock;
638
639 error = -EEXIST;
640 if (dentry->d_inode)
641 goto out_dput;
642
cb8e7090
CH		 643 error = mnt_want_write(parent->mnt);
644 if (error)
645 goto out_dput;
646
76dda93c 647 error = btrfs_may_create(dir, dentry);
cb8e7090
CH		 648 if (error)
649 goto out_drop_write;
650
76dda93c
YZ		 651 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
652
653 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
654 goto out_up_read;
655
3de4586c 656 if (snap_src) {
72fd032e 657 error = create_snapshot(snap_src, dentry,
b83cc969 658 name, namelen, async_transid, readonly);
3de4586c 659 } else {
76dda93c 660 error = create_subvol(BTRFS_I(dir)->root, dentry,
72fd032e 661 name, namelen, async_transid);
3de4586c 662 }
76dda93c
YZ		 663 if (!error)
664 fsnotify_mkdir(dir, dentry);
665out_up_read:
666 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
cb8e7090
CH		 667out_drop_write:
668 mnt_drop_write(parent->mnt);
669out_dput:
670 dput(dentry);
671out_unlock:
76dda93c 672 mutex_unlock(&dir->i_mutex);
cb8e7090
CH		 673 return error;
674}
675
4cb5300b
CM		 676/*
677 * When we're defragging a range, we don't want to kick it off again
678 * if it is really just waiting for delalloc to send it down.
679 * If we find a nice big extent or delalloc range for the bytes in the
680 * file you want to defrag, we return 0 to let you know to skip this
681 * part of the file
682 */
683static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
684{
685 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
686 struct extent_map *em = NULL;
687 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
688 u64 end;
689
690 read_lock(&em_tree->lock);
691 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
692 read_unlock(&em_tree->lock);
693
694 if (em) {
695 end = extent_map_end(em);
696 free_extent_map(em);
697 if (end - offset > thresh)
698 return 0;
699 }
700 /* if we already have a nice delalloc here, just stop */
701 thresh /= 2;
702 end = count_range_bits(io_tree, &offset, offset + thresh,
703 thresh, EXTENT_DELALLOC, 1);
704 if (end >= thresh)
705 return 0;
706 return 1;
707}
708
709/*
710 * helper function to walk through a file and find extents
711 * newer than a specific transid, and smaller than thresh.
712 *
713 * This is used by the defragging code to find new and small
714 * extents
715 */
716static int find_new_extents(struct btrfs_root *root,
717 struct inode *inode, u64 newer_than,
718 u64 *off, int thresh)
719{
720 struct btrfs_path *path;
721 struct btrfs_key min_key;
722 struct btrfs_key max_key;
723 struct extent_buffer *leaf;
724 struct btrfs_file_extent_item *extent;
725 int type;
726 int ret;
a4689d2b 727 u64 ino = btrfs_ino(inode);
4cb5300b
CM		 728
729 path = btrfs_alloc_path();
730 if (!path)
731 return -ENOMEM;
732
a4689d2b 733 min_key.objectid = ino;
4cb5300b
CM		 734 min_key.type = BTRFS_EXTENT_DATA_KEY;
735 min_key.offset = *off;
736
a4689d2b 737 max_key.objectid = ino;
4cb5300b
CM		 738 max_key.type = (u8)-1;
739 max_key.offset = (u64)-1;
740
741 path->keep_locks = 1;
742
743 while(1) {
744 ret = btrfs_search_forward(root, &min_key, &max_key,
745 path, 0, newer_than);
746 if (ret != 0)
747 goto none;
a4689d2b 748 if (min_key.objectid != ino)
4cb5300b
CM		 749 goto none;
750 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
751 goto none;
752
753 leaf = path->nodes[0];
754 extent = btrfs_item_ptr(leaf, path->slots[0],
755 struct btrfs_file_extent_item);
756
757 type = btrfs_file_extent_type(leaf, extent);
758 if (type == BTRFS_FILE_EXTENT_REG &&
759 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
760 check_defrag_in_cache(inode, min_key.offset, thresh)) {
761 *off = min_key.offset;
762 btrfs_free_path(path);
763 return 0;
764 }
765
766 if (min_key.offset == (u64)-1)
767 goto none;
768
769 min_key.offset++;
770 btrfs_release_path(path);
771 }
772none:
773 btrfs_free_path(path);
774 return -ENOENT;
775}
776
940100a4 777static int should_defrag_range(struct inode *inode, u64 start, u64 len,
1e701a32
CM		 778 int thresh, u64 *last_len, u64 *skip,
779 u64 *defrag_end)
940100a4
CM		 780{
781 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
782 struct extent_map *em = NULL;
783 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
784 int ret = 1;
785
786 /*
008873ea 787 * make sure that once we start defragging an extent, we keep on
940100a4
CM		 788 * defragging it
789 */
790 if (start < *defrag_end)
791 return 1;
792
793 *skip = 0;
794
795 /*
796 * hopefully we have this extent in the tree already, try without
797 * the full extent lock
798 */
799 read_lock(&em_tree->lock);
800 em = lookup_extent_mapping(em_tree, start, len);
801 read_unlock(&em_tree->lock);
802
803 if (!em) {
804 /* get the big lock and read metadata off disk */
d0082371 805 lock_extent(io_tree, start, start + len - 1);
940100a4 806 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
d0082371 807 unlock_extent(io_tree, start, start + len - 1);
940100a4 808
6cf8bfbf 809 if (IS_ERR(em))
940100a4
CM		 810 return 0;
811 }
812
813 /* this will cover holes, and inline extents */
814 if (em->block_start >= EXTENT_MAP_LAST_BYTE)
815 ret = 0;
816
817 /*
818 * we hit a real extent, if it is big don't bother defragging it again
819 */
1e701a32 820 if ((*last_len == 0 || *last_len >= thresh) && em->len >= thresh)
940100a4
CM		 821 ret = 0;
822
823 /*
824 * last_len ends up being a counter of how many bytes we've defragged.
825 * every time we choose not to defrag an extent, we reset *last_len
826 * so that the next tiny extent will force a defrag.
827 *
828 * The end result of this is that tiny extents before a single big
829 * extent will force at least part of that big extent to be defragged.
830 */
831 if (ret) {
940100a4
CM		 832 *defrag_end = extent_map_end(em);
833 } else {
834 *last_len = 0;
835 *skip = extent_map_end(em);
836 *defrag_end = 0;
837 }
838
839 free_extent_map(em);
840 return ret;
841}
842
4cb5300b
CM		 843/*
844 * it doesn't do much good to defrag one or two pages
845 * at a time. This pulls in a nice chunk of pages
846 * to COW and defrag.
847 *
848 * It also makes sure the delalloc code has enough
849 * dirty data to avoid making new small extents as part
850 * of the defrag
851 *
852 * It's a good idea to start RA on this range
853 * before calling this.
854 */
855static int cluster_pages_for_defrag(struct inode *inode,
856 struct page **pages,
857 unsigned long start_index,
858 int num_pages)
f46b5a66 859{
4cb5300b
CM		 860 unsigned long file_end;
861 u64 isize = i_size_read(inode);
862 u64 page_start;
863 u64 page_end;
864 int ret;
865 int i;
866 int i_done;
3eaa2885 867 struct btrfs_ordered_extent *ordered;
4cb5300b 868 struct extent_state *cached_state = NULL;
600a45e1 869 struct extent_io_tree *tree;
3b16a4e3 870 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
4cb5300b
CM		 871
872 if (isize == 0)
873 return 0;
874 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
875
876 ret = btrfs_delalloc_reserve_space(inode,
877 num_pages << PAGE_CACHE_SHIFT);
878 if (ret)
879 return ret;
4cb5300b 880 i_done = 0;
600a45e1 881 tree = &BTRFS_I(inode)->io_tree;
4cb5300b
CM		 882
883 /* step one, lock all the pages */
884 for (i = 0; i < num_pages; i++) {
885 struct page *page;
600a45e1 886again:
a94733d0 887 page = find_or_create_page(inode->i_mapping,
600a45e1 888 start_index + i, mask);
4cb5300b
CM		 889 if (!page)
890 break;
891
600a45e1
MX		 892 page_start = page_offset(page);
893 page_end = page_start + PAGE_CACHE_SIZE - 1;
894 while (1) {
d0082371 895 lock_extent(tree, page_start, page_end);
600a45e1
MX		 896 ordered = btrfs_lookup_ordered_extent(inode,
897 page_start);
d0082371 898 unlock_extent(tree, page_start, page_end);
600a45e1
MX		 899 if (!ordered)
900 break;
901
902 unlock_page(page);
903 btrfs_start_ordered_extent(inode, ordered, 1);
904 btrfs_put_ordered_extent(ordered);
905 lock_page(page);
906 }
907
4cb5300b
CM		 908 if (!PageUptodate(page)) {
909 btrfs_readpage(NULL, page);
910 lock_page(page);
911 if (!PageUptodate(page)) {
912 unlock_page(page);
913 page_cache_release(page);
914 ret = -EIO;
915 break;
916 }
917 }
600a45e1 918
4cb5300b
CM		 919 isize = i_size_read(inode);
920 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
600a45e1 921 if (!isize || page->index > file_end) {
4cb5300b
CM		 922 /* whoops, we blew past eof, skip this page */
923 unlock_page(page);
924 page_cache_release(page);
925 break;
926 }
600a45e1
MX		 927
928 if (page->mapping != inode->i_mapping) {
929 unlock_page(page);
930 page_cache_release(page);
931 goto again;
932 }
933
4cb5300b
CM		 934 pages[i] = page;
935 i_done++;
936 }
937 if (!i_done || ret)
938 goto out;
939
940 if (!(inode->i_sb->s_flags & MS_ACTIVE))
941 goto out;
942
943 /*
944 * so now we have a nice long stream of locked
945 * and up to date pages, lets wait on them
946 */
947 for (i = 0; i < i_done; i++)
948 wait_on_page_writeback(pages[i]);
949
950 page_start = page_offset(pages[0]);
951 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
952
953 lock_extent_bits(&BTRFS_I(inode)->io_tree,
d0082371 954 page_start, page_end - 1, 0, &cached_state);
4cb5300b
CM		 955 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
956 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
957 EXTENT_DO_ACCOUNTING, 0, 0, &cached_state,
958 GFP_NOFS);
959
960 if (i_done != num_pages) {
9e0baf60
JB		 961 spin_lock(&BTRFS_I(inode)->lock);
962 BTRFS_I(inode)->outstanding_extents++;
963 spin_unlock(&BTRFS_I(inode)->lock);
4cb5300b
CM		 964 btrfs_delalloc_release_space(inode,
965 (num_pages - i_done) << PAGE_CACHE_SHIFT);
966 }
967
968
969 btrfs_set_extent_delalloc(inode, page_start, page_end - 1,
970 &cached_state);
971
972 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
973 page_start, page_end - 1, &cached_state,
974 GFP_NOFS);
975
976 for (i = 0; i < i_done; i++) {
977 clear_page_dirty_for_io(pages[i]);
978 ClearPageChecked(pages[i]);
979 set_page_extent_mapped(pages[i]);
980 set_page_dirty(pages[i]);
981 unlock_page(pages[i]);
982 page_cache_release(pages[i]);
983 }
984 return i_done;
985out:
986 for (i = 0; i < i_done; i++) {
987 unlock_page(pages[i]);
988 page_cache_release(pages[i]);
989 }
990 btrfs_delalloc_release_space(inode, num_pages << PAGE_CACHE_SHIFT);
991 return ret;
992
993}
994
995int btrfs_defrag_file(struct inode *inode, struct file *file,
996 struct btrfs_ioctl_defrag_range_args *range,
997 u64 newer_than, unsigned long max_to_defrag)
998{
999 struct btrfs_root *root = BTRFS_I(inode)->root;
1a419d85 1000 struct btrfs_super_block *disk_super;
4cb5300b 1001 struct file_ra_state *ra = NULL;
f46b5a66 1002 unsigned long last_index;
151a31b2 1003 u64 isize = i_size_read(inode);
1a419d85 1004 u64 features;
940100a4
CM		 1005 u64 last_len = 0;
1006 u64 skip = 0;
1007 u64 defrag_end = 0;
4cb5300b 1008 u64 newer_off = range->start;
f46b5a66 1009 unsigned long i;
008873ea 1010 unsigned long ra_index = 0;
f46b5a66 1011 int ret;
4cb5300b 1012 int defrag_count = 0;
1a419d85 1013 int compress_type = BTRFS_COMPRESS_ZLIB;
4cb5300b 1014 int extent_thresh = range->extent_thresh;
008873ea
LZ		 1015 int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1016 int cluster = max_cluster;
4cb5300b
CM		 1017 u64 new_align = ~((u64)128 * 1024 - 1);
1018 struct page **pages = NULL;
1019
1020 if (extent_thresh == 0)
1021 extent_thresh = 256 * 1024;
1a419d85
LZ		 1022
1023 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1024 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1025 return -EINVAL;
1026 if (range->compress_type)
1027 compress_type = range->compress_type;
1028 }
f46b5a66 1029
151a31b2 1030 if (isize == 0)
940100a4
CM		 1031 return 0;
1032
4cb5300b
CM		 1033 /*
1034 * if we were not given a file, allocate a readahead
1035 * context
1036 */
1037 if (!file) {
1038 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1039 if (!ra)
1040 return -ENOMEM;
1041 file_ra_state_init(ra, inode->i_mapping);
1042 } else {
1043 ra = &file->f_ra;
1044 }
1045
008873ea 1046 pages = kmalloc(sizeof(struct page *) * max_cluster,
4cb5300b
CM		 1047 GFP_NOFS);
1048 if (!pages) {
1049 ret = -ENOMEM;
1050 goto out_ra;
1051 }
1052
1053 /* find the last page to defrag */
1e701a32 1054 if (range->start + range->len > range->start) {
151a31b2 1055 last_index = min_t(u64, isize - 1,
1e701a32
CM		 1056 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1057 } else {
151a31b2 1058 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1e701a32
CM		 1059 }
1060
4cb5300b
CM		 1061 if (newer_than) {
1062 ret = find_new_extents(root, inode, newer_than,
1063 &newer_off, 64 * 1024);
1064 if (!ret) {
1065 range->start = newer_off;
1066 /*
1067 * we always align our defrag to help keep
1068 * the extents in the file evenly spaced
1069 */
1070 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
4cb5300b
CM		 1071 } else
1072 goto out_ra;
1073 } else {
1074 i = range->start >> PAGE_CACHE_SHIFT;
1075 }
1076 if (!max_to_defrag)
7ec31b54 1077 max_to_defrag = last_index + 1;
4cb5300b 1078
2a0f7f57
LZ		 1079 /*
1080 * make writeback starts from i, so the defrag range can be
1081 * written sequentially.
1082 */
1083 if (i < inode->i_mapping->writeback_index)
1084 inode->i_mapping->writeback_index = i;
1085
f7f43cc8
CM		 1086 while (i <= last_index && defrag_count < max_to_defrag &&
1087 (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1088 PAGE_CACHE_SHIFT)) {
4cb5300b
CM		 1089 /*
1090 * make sure we stop running if someone unmounts
1091 * the FS
1092 */
1093 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1094 break;
1095
1096 if (!newer_than &&
1097 !should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
1e701a32 1098 PAGE_CACHE_SIZE,
4cb5300b 1099 extent_thresh,
1e701a32 1100 &last_len, &skip,
940100a4
CM		 1101 &defrag_end)) {
1102 unsigned long next;
1103 /*
1104 * the should_defrag function tells us how much to skip
1105 * bump our counter by the suggested amount
1106 */
1107 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1108 i = max(i + 1, next);
1109 continue;
1110 }
008873ea
LZ		 1111
1112 if (!newer_than) {
1113 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1114 PAGE_CACHE_SHIFT) - i;
1115 cluster = min(cluster, max_cluster);
1116 } else {
1117 cluster = max_cluster;
1118 }
1119
1e701a32 1120 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1a419d85 1121 BTRFS_I(inode)->force_compress = compress_type;
940100a4 1122
008873ea
LZ		 1123 if (i + cluster > ra_index) {
1124 ra_index = max(i, ra_index);
1125 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1126 cluster);
1127 ra_index += max_cluster;
1128 }
940100a4 1129
008873ea 1130 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
4cb5300b
CM		 1131 if (ret < 0)
1132 goto out_ra;
1133
1134 defrag_count += ret;
1135 balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret);
4cb5300b
CM		 1136
1137 if (newer_than) {
1138 if (newer_off == (u64)-1)
1139 break;
1140
1141 newer_off = max(newer_off + 1,
1142 (u64)i << PAGE_CACHE_SHIFT);
1143
1144 ret = find_new_extents(root, inode,
1145 newer_than, &newer_off,
1146 64 * 1024);
1147 if (!ret) {
1148 range->start = newer_off;
1149 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
4cb5300b
CM		 1150 } else {
1151 break;
f46b5a66 1152 }
4cb5300b 1153 } else {
008873ea 1154 if (ret > 0) {
cbcc8326 1155 i += ret;
008873ea
LZ		 1156 last_len += ret << PAGE_CACHE_SHIFT;
1157 } else {
cbcc8326 1158 i++;
008873ea
LZ		 1159 last_len = 0;
1160 }
f46b5a66 1161 }
f46b5a66
CH		 1162 }
1163
1e701a32
CM		 1164 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1165 filemap_flush(inode->i_mapping);
1166
1167 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1168 /* the filemap_flush will queue IO into the worker threads, but
1169 * we have to make sure the IO is actually started and that
1170 * ordered extents get created before we return
1171 */
1172 atomic_inc(&root->fs_info->async_submit_draining);
1173 while (atomic_read(&root->fs_info->nr_async_submits) ||
1174 atomic_read(&root->fs_info->async_delalloc_pages)) {
1175 wait_event(root->fs_info->async_submit_wait,
1176 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1177 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1178 }
1179 atomic_dec(&root->fs_info->async_submit_draining);
1180
1181 mutex_lock(&inode->i_mutex);
261507a0 1182 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1e701a32
CM		 1183 mutex_unlock(&inode->i_mutex);
1184 }
1185
6c41761f 1186 disk_super = root->fs_info->super_copy;
1a419d85
LZ		 1187 features = btrfs_super_incompat_flags(disk_super);
1188 if (range->compress_type == BTRFS_COMPRESS_LZO) {
1189 features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
1190 btrfs_set_super_incompat_flags(disk_super, features);
1191 }
1192
60ccf82f 1193 ret = defrag_count;
940100a4 1194
4cb5300b
CM		 1195out_ra:
1196 if (!file)
1197 kfree(ra);
1198 kfree(pages);
940100a4 1199 return ret;
f46b5a66
CH		 1200}
1201
76dda93c
YZ		 1202static noinline int btrfs_ioctl_resize(struct btrfs_root *root,
1203 void __user *arg)
f46b5a66
CH		 1204{
1205 u64 new_size;
1206 u64 old_size;
1207 u64 devid = 1;
1208 struct btrfs_ioctl_vol_args *vol_args;
1209 struct btrfs_trans_handle *trans;
1210 struct btrfs_device *device = NULL;
1211 char *sizestr;
1212 char *devstr = NULL;
1213 int ret = 0;
f46b5a66
CH		 1214 int mod = 0;
1215
c146afad
YZ		 1216 if (root->fs_info->sb->s_flags & MS_RDONLY)
1217 return -EROFS;
1218
e441d54d
CM		 1219 if (!capable(CAP_SYS_ADMIN))
1220 return -EPERM;
1221
c9e9f97b
ID		 1222 mutex_lock(&root->fs_info->volume_mutex);
1223 if (root->fs_info->balance_ctl) {
1224 printk(KERN_INFO "btrfs: balance in progress\n");
1225 ret = -EINVAL;
1226 goto out;
1227 }
1228
dae7b665 1229 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 1230 if (IS_ERR(vol_args)) {
1231 ret = PTR_ERR(vol_args);
1232 goto out;
1233 }
5516e595
MF		 1234
1235 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 1236
f46b5a66
CH		 1237 sizestr = vol_args->name;
1238 devstr = strchr(sizestr, ':');
1239 if (devstr) {
1240 char *end;
1241 sizestr = devstr + 1;
1242 *devstr = '\0';
1243 devstr = vol_args->name;
1244 devid = simple_strtoull(devstr, &end, 10);
5bb14682 1245 printk(KERN_INFO "btrfs: resizing devid %llu\n",
21380931 1246 (unsigned long long)devid);
f46b5a66 1247 }
2b82032c 1248 device = btrfs_find_device(root, devid, NULL, NULL);
f46b5a66 1249 if (!device) {
5bb14682 1250 printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
21380931 1251 (unsigned long long)devid);
f46b5a66 1252 ret = -EINVAL;
c9e9f97b 1253 goto out_free;
f46b5a66
CH		 1254 }
1255 if (!strcmp(sizestr, "max"))
1256 new_size = device->bdev->bd_inode->i_size;
1257 else {
1258 if (sizestr[0] == '-') {
1259 mod = -1;
1260 sizestr++;
1261 } else if (sizestr[0] == '+') {
1262 mod = 1;
1263 sizestr++;
1264 }
91748467 1265 new_size = memparse(sizestr, NULL);
f46b5a66
CH		 1266 if (new_size == 0) {
1267 ret = -EINVAL;
c9e9f97b 1268 goto out_free;
f46b5a66
CH		 1269 }
1270 }
1271
1272 old_size = device->total_bytes;
1273
1274 if (mod < 0) {
1275 if (new_size > old_size) {
1276 ret = -EINVAL;
c9e9f97b 1277 goto out_free;
f46b5a66
CH		 1278 }
1279 new_size = old_size - new_size;
1280 } else if (mod > 0) {
1281 new_size = old_size + new_size;
1282 }
1283
1284 if (new_size < 256 * 1024 * 1024) {
1285 ret = -EINVAL;
c9e9f97b 1286 goto out_free;
f46b5a66
CH		 1287 }
1288 if (new_size > device->bdev->bd_inode->i_size) {
1289 ret = -EFBIG;
c9e9f97b 1290 goto out_free;
f46b5a66
CH		 1291 }
1292
1293 do_div(new_size, root->sectorsize);
1294 new_size *= root->sectorsize;
1295
5bb14682 1296 printk(KERN_INFO "btrfs: new size for %s is %llu\n",
f46b5a66
CH		 1297 device->name, (unsigned long long)new_size);
1298
1299 if (new_size > old_size) {
a22285a6 1300 trans = btrfs_start_transaction(root, 0);
98d5dc13
TI		 1301 if (IS_ERR(trans)) {
1302 ret = PTR_ERR(trans);
c9e9f97b 1303 goto out_free;
98d5dc13 1304 }
f46b5a66
CH		 1305 ret = btrfs_grow_device(trans, device, new_size);
1306 btrfs_commit_transaction(trans, root);
ece7d20e 1307 } else if (new_size < old_size) {
f46b5a66
CH		 1308 ret = btrfs_shrink_device(device, new_size);
1309 }
1310
c9e9f97b 1311out_free:
f46b5a66 1312 kfree(vol_args);
c9e9f97b
ID		 1313out:
1314 mutex_unlock(&root->fs_info->volume_mutex);
f46b5a66
CH		 1315 return ret;
1316}
1317
72fd032e
SW		 1318static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
1319 char *name,
1320 unsigned long fd,
1321 int subvol,
b83cc969
LZ		 1322 u64 *transid,
1323 bool readonly)
f46b5a66 1324{
cb8e7090 1325 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3de4586c 1326 struct file *src_file;
f46b5a66 1327 int namelen;
3de4586c 1328 int ret = 0;
f46b5a66 1329
c146afad
YZ		 1330 if (root->fs_info->sb->s_flags & MS_RDONLY)
1331 return -EROFS;
1332
72fd032e
SW		 1333 namelen = strlen(name);
1334 if (strchr(name, '/')) {
f46b5a66
CH		 1335 ret = -EINVAL;
1336 goto out;
1337 }
1338
16780cab
CM		 1339 if (name[0] == '.' &&
1340 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1341 ret = -EEXIST;
1342 goto out;
1343 }
1344
3de4586c 1345 if (subvol) {
72fd032e 1346 ret = btrfs_mksubvol(&file->f_path, name, namelen,
b83cc969 1347 NULL, transid, readonly);
cb8e7090 1348 } else {
3de4586c 1349 struct inode *src_inode;
72fd032e 1350 src_file = fget(fd);
3de4586c
CM		 1351 if (!src_file) {
1352 ret = -EINVAL;
1353 goto out;
1354 }
1355
1356 src_inode = src_file->f_path.dentry->d_inode;
1357 if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
d397712b
CM		 1358 printk(KERN_INFO "btrfs: Snapshot src from "
1359 "another FS\n");
3de4586c
CM		 1360 ret = -EINVAL;
1361 fput(src_file);
1362 goto out;
1363 }
72fd032e
SW		 1364 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1365 BTRFS_I(src_inode)->root,
b83cc969 1366 transid, readonly);
3de4586c 1367 fput(src_file);
cb8e7090 1368 }
f46b5a66 1369out:
72fd032e
SW		 1370 return ret;
1371}
1372
1373static noinline int btrfs_ioctl_snap_create(struct file *file,
fa0d2b9b 1374 void __user *arg, int subvol)
72fd032e 1375{
fa0d2b9b 1376 struct btrfs_ioctl_vol_args *vol_args;
72fd032e
SW		 1377 int ret;
1378
fa0d2b9b
LZ		 1379 vol_args = memdup_user(arg, sizeof(*vol_args));
1380 if (IS_ERR(vol_args))
1381 return PTR_ERR(vol_args);
1382 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
72fd032e 1383
fa0d2b9b 1384 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
b83cc969
LZ		 1385 vol_args->fd, subvol,
1386 NULL, false);
fdfb1e4f 1387
fa0d2b9b
LZ		 1388 kfree(vol_args);
1389 return ret;
1390}
fdfb1e4f 1391
fa0d2b9b
LZ		 1392static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1393 void __user *arg, int subvol)
1394{
1395 struct btrfs_ioctl_vol_args_v2 *vol_args;
1396 int ret;
1397 u64 transid = 0;
1398 u64 *ptr = NULL;
b83cc969 1399 bool readonly = false;
75eaa0e2 1400
fa0d2b9b
LZ		 1401 vol_args = memdup_user(arg, sizeof(*vol_args));
1402 if (IS_ERR(vol_args))
1403 return PTR_ERR(vol_args);
1404 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
75eaa0e2 1405
b83cc969
LZ		 1406 if (vol_args->flags &
1407 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY)) {
1408 ret = -EOPNOTSUPP;
fa0d2b9b 1409 goto out;
72fd032e 1410 }
fa0d2b9b
LZ		 1411
1412 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1413 ptr = &transid;
b83cc969
LZ		 1414 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1415 readonly = true;
fa0d2b9b
LZ		 1416
1417 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
b83cc969
LZ		 1418 vol_args->fd, subvol,
1419 ptr, readonly);
fa0d2b9b
LZ		 1420
1421 if (ret == 0 && ptr &&
1422 copy_to_user(arg +
1423 offsetof(struct btrfs_ioctl_vol_args_v2,
1424 transid), ptr, sizeof(*ptr)))
1425 ret = -EFAULT;
fdfb1e4f 1426out:
f46b5a66
CH		 1427 kfree(vol_args);
1428 return ret;
1429}
1430
0caa102d
LZ		 1431static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1432 void __user *arg)
1433{
1434 struct inode *inode = fdentry(file)->d_inode;
1435 struct btrfs_root *root = BTRFS_I(inode)->root;
1436 int ret = 0;
1437 u64 flags = 0;
1438
33345d01 1439 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
0caa102d
LZ		 1440 return -EINVAL;
1441
1442 down_read(&root->fs_info->subvol_sem);
1443 if (btrfs_root_readonly(root))
1444 flags |= BTRFS_SUBVOL_RDONLY;
1445 up_read(&root->fs_info->subvol_sem);
1446
1447 if (copy_to_user(arg, &flags, sizeof(flags)))
1448 ret = -EFAULT;
1449
1450 return ret;
1451}
1452
1453static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1454 void __user *arg)
1455{
1456 struct inode *inode = fdentry(file)->d_inode;
1457 struct btrfs_root *root = BTRFS_I(inode)->root;
1458 struct btrfs_trans_handle *trans;
1459 u64 root_flags;
1460 u64 flags;
1461 int ret = 0;
1462
1463 if (root->fs_info->sb->s_flags & MS_RDONLY)
1464 return -EROFS;
1465
33345d01 1466 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
0caa102d
LZ		 1467 return -EINVAL;
1468
1469 if (copy_from_user(&flags, arg, sizeof(flags)))
1470 return -EFAULT;
1471
b4dc2b8c 1472 if (flags & BTRFS_SUBVOL_CREATE_ASYNC)
0caa102d
LZ		 1473 return -EINVAL;
1474
1475 if (flags & ~BTRFS_SUBVOL_RDONLY)
1476 return -EOPNOTSUPP;
1477
2e149670 1478 if (!inode_owner_or_capable(inode))
b4dc2b8c
LZ		 1479 return -EACCES;
1480
0caa102d
LZ		 1481 down_write(&root->fs_info->subvol_sem);
1482
1483 /* nothing to do */
1484 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
1485 goto out;
1486
1487 root_flags = btrfs_root_flags(&root->root_item);
1488 if (flags & BTRFS_SUBVOL_RDONLY)
1489 btrfs_set_root_flags(&root->root_item,
1490 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1491 else
1492 btrfs_set_root_flags(&root->root_item,
1493 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1494
1495 trans = btrfs_start_transaction(root, 1);
1496 if (IS_ERR(trans)) {
1497 ret = PTR_ERR(trans);
1498 goto out_reset;
1499 }
1500
b4dc2b8c 1501 ret = btrfs_update_root(trans, root->fs_info->tree_root,
0caa102d
LZ		 1502 &root->root_key, &root->root_item);
1503
1504 btrfs_commit_transaction(trans, root);
1505out_reset:
1506 if (ret)
1507 btrfs_set_root_flags(&root->root_item, root_flags);
1508out:
1509 up_write(&root->fs_info->subvol_sem);
1510 return ret;
1511}
1512
76dda93c
YZ		 1513/*
1514 * helper to check if the subvolume references other subvolumes
1515 */
1516static noinline int may_destroy_subvol(struct btrfs_root *root)
1517{
1518 struct btrfs_path *path;
1519 struct btrfs_key key;
1520 int ret;
1521
1522 path = btrfs_alloc_path();
1523 if (!path)
1524 return -ENOMEM;
1525
1526 key.objectid = root->root_key.objectid;
1527 key.type = BTRFS_ROOT_REF_KEY;
1528 key.offset = (u64)-1;
1529
1530 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1531 &key, path, 0, 0);
1532 if (ret < 0)
1533 goto out;
1534 BUG_ON(ret == 0);
1535
1536 ret = 0;
1537 if (path->slots[0] > 0) {
1538 path->slots[0]--;
1539 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1540 if (key.objectid == root->root_key.objectid &&
1541 key.type == BTRFS_ROOT_REF_KEY)
1542 ret = -ENOTEMPTY;
1543 }
1544out:
1545 btrfs_free_path(path);
1546 return ret;
1547}
1548
ac8e9819
CM		 1549static noinline int key_in_sk(struct btrfs_key *key,
1550 struct btrfs_ioctl_search_key *sk)
1551{
abc6e134
CM		 1552 struct btrfs_key test;
1553 int ret;
1554
1555 test.objectid = sk->min_objectid;
1556 test.type = sk->min_type;
1557 test.offset = sk->min_offset;
1558
1559 ret = btrfs_comp_cpu_keys(key, &test);
1560 if (ret < 0)
ac8e9819 1561 return 0;
abc6e134
CM		 1562
1563 test.objectid = sk->max_objectid;
1564 test.type = sk->max_type;
1565 test.offset = sk->max_offset;
1566
1567 ret = btrfs_comp_cpu_keys(key, &test);
1568 if (ret > 0)
ac8e9819
CM		 1569 return 0;
1570 return 1;
1571}
1572
1573static noinline int copy_to_sk(struct btrfs_root *root,
1574 struct btrfs_path *path,
1575 struct btrfs_key *key,
1576 struct btrfs_ioctl_search_key *sk,
1577 char *buf,
1578 unsigned long *sk_offset,
1579 int *num_found)
1580{
1581 u64 found_transid;
1582 struct extent_buffer *leaf;
1583 struct btrfs_ioctl_search_header sh;
1584 unsigned long item_off;
1585 unsigned long item_len;
1586 int nritems;
1587 int i;
1588 int slot;
ac8e9819
CM		 1589 int ret = 0;
1590
1591 leaf = path->nodes[0];
1592 slot = path->slots[0];
1593 nritems = btrfs_header_nritems(leaf);
1594
1595 if (btrfs_header_generation(leaf) > sk->max_transid) {
1596 i = nritems;
1597 goto advance_key;
1598 }
1599 found_transid = btrfs_header_generation(leaf);
1600
1601 for (i = slot; i < nritems; i++) {
1602 item_off = btrfs_item_ptr_offset(leaf, i);
1603 item_len = btrfs_item_size_nr(leaf, i);
1604
1605 if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1606 item_len = 0;
1607
1608 if (sizeof(sh) + item_len + *sk_offset >
1609 BTRFS_SEARCH_ARGS_BUFSIZE) {
1610 ret = 1;
1611 goto overflow;
1612 }
1613
1614 btrfs_item_key_to_cpu(leaf, key, i);
1615 if (!key_in_sk(key, sk))
1616 continue;
1617
1618 sh.objectid = key->objectid;
1619 sh.offset = key->offset;
1620 sh.type = key->type;
1621 sh.len = item_len;
1622 sh.transid = found_transid;
1623
1624 /* copy search result header */
1625 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1626 *sk_offset += sizeof(sh);
1627
1628 if (item_len) {
1629 char *p = buf + *sk_offset;
1630 /* copy the item */
1631 read_extent_buffer(leaf, p,
1632 item_off, item_len);
1633 *sk_offset += item_len;
ac8e9819 1634 }
e2156867 1635 (*num_found)++;
ac8e9819
CM		 1636
1637 if (*num_found >= sk->nr_items)
1638 break;
1639 }
1640advance_key:
abc6e134
CM		 1641 ret = 0;
1642 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
ac8e9819 1643 key->offset++;
abc6e134
CM		 1644 else if (key->type < (u8)-1 && key->type < sk->max_type) {
1645 key->offset = 0;
ac8e9819 1646 key->type++;
abc6e134
CM		 1647 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1648 key->offset = 0;
1649 key->type = 0;
ac8e9819 1650 key->objectid++;
abc6e134
CM		 1651 } else
1652 ret = 1;
ac8e9819 1653overflow:
ac8e9819
CM		 1654 return ret;
1655}
1656
1657static noinline int search_ioctl(struct inode *inode,
1658 struct btrfs_ioctl_search_args *args)
1659{
1660 struct btrfs_root *root;
1661 struct btrfs_key key;
1662 struct btrfs_key max_key;
1663 struct btrfs_path *path;
1664 struct btrfs_ioctl_search_key *sk = &args->key;
1665 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1666 int ret;
1667 int num_found = 0;
1668 unsigned long sk_offset = 0;
1669
1670 path = btrfs_alloc_path();
1671 if (!path)
1672 return -ENOMEM;
1673
1674 if (sk->tree_id == 0) {
1675 /* search the root of the inode that was passed */
1676 root = BTRFS_I(inode)->root;
1677 } else {
1678 key.objectid = sk->tree_id;
1679 key.type = BTRFS_ROOT_ITEM_KEY;
1680 key.offset = (u64)-1;
1681 root = btrfs_read_fs_root_no_name(info, &key);
1682 if (IS_ERR(root)) {
1683 printk(KERN_ERR "could not find root %llu\n",
1684 sk->tree_id);
1685 btrfs_free_path(path);
1686 return -ENOENT;
1687 }
1688 }
1689
1690 key.objectid = sk->min_objectid;
1691 key.type = sk->min_type;
1692 key.offset = sk->min_offset;
1693
1694 max_key.objectid = sk->max_objectid;
1695 max_key.type = sk->max_type;
1696 max_key.offset = sk->max_offset;
1697
1698 path->keep_locks = 1;
1699
1700 while(1) {
1701 ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1702 sk->min_transid);
1703 if (ret != 0) {
1704 if (ret > 0)
1705 ret = 0;
1706 goto err;
1707 }
1708 ret = copy_to_sk(root, path, &key, sk, args->buf,
1709 &sk_offset, &num_found);
b3b4aa74 1710 btrfs_release_path(path);
ac8e9819
CM		 1711 if (ret || num_found >= sk->nr_items)
1712 break;
1713
1714 }
1715 ret = 0;
1716err:
1717 sk->nr_items = num_found;
1718 btrfs_free_path(path);
1719 return ret;
1720}
1721
1722static noinline int btrfs_ioctl_tree_search(struct file *file,
1723 void __user *argp)
1724{
1725 struct btrfs_ioctl_search_args *args;
1726 struct inode *inode;
1727 int ret;
1728
1729 if (!capable(CAP_SYS_ADMIN))
1730 return -EPERM;
1731
2354d08f
JL		 1732 args = memdup_user(argp, sizeof(*args));
1733 if (IS_ERR(args))
1734 return PTR_ERR(args);
ac8e9819 1735
ac8e9819
CM		 1736 inode = fdentry(file)->d_inode;
1737 ret = search_ioctl(inode, args);
1738 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1739 ret = -EFAULT;
1740 kfree(args);
1741 return ret;
1742}
1743
98d377a0 1744/*
ac8e9819
CM		 1745 * Search INODE_REFs to identify path name of 'dirid' directory
1746 * in a 'tree_id' tree. and sets path name to 'name'.
1747 */
98d377a0
TH		 1748static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1749 u64 tree_id, u64 dirid, char *name)
1750{
1751 struct btrfs_root *root;
1752 struct btrfs_key key;
ac8e9819 1753 char *ptr;
98d377a0
TH		 1754 int ret = -1;
1755 int slot;
1756 int len;
1757 int total_len = 0;
1758 struct btrfs_inode_ref *iref;
1759 struct extent_buffer *l;
1760 struct btrfs_path *path;
1761
1762 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1763 name[0]='\0';
1764 return 0;
1765 }
1766
1767 path = btrfs_alloc_path();
1768 if (!path)
1769 return -ENOMEM;
1770
ac8e9819 1771 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
98d377a0
TH		 1772
1773 key.objectid = tree_id;
1774 key.type = BTRFS_ROOT_ITEM_KEY;
1775 key.offset = (u64)-1;
1776 root = btrfs_read_fs_root_no_name(info, &key);
1777 if (IS_ERR(root)) {
1778 printk(KERN_ERR "could not find root %llu\n", tree_id);
8ad6fcab
CM		 1779 ret = -ENOENT;
1780 goto out;
98d377a0
TH		 1781 }
1782
1783 key.objectid = dirid;
1784 key.type = BTRFS_INODE_REF_KEY;
8ad6fcab 1785 key.offset = (u64)-1;
98d377a0
TH		 1786
1787 while(1) {
1788 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1789 if (ret < 0)
1790 goto out;
1791
1792 l = path->nodes[0];
1793 slot = path->slots[0];
8ad6fcab
CM		 1794 if (ret > 0 && slot > 0)
1795 slot--;
98d377a0
TH		 1796 btrfs_item_key_to_cpu(l, &key, slot);
1797
1798 if (ret > 0 && (key.objectid != dirid ||
ac8e9819
CM		 1799 key.type != BTRFS_INODE_REF_KEY)) {
1800 ret = -ENOENT;
98d377a0 1801 goto out;
ac8e9819 1802 }
98d377a0
TH		 1803
1804 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1805 len = btrfs_inode_ref_name_len(l, iref);
1806 ptr -= len + 1;
1807 total_len += len + 1;
ac8e9819 1808 if (ptr < name)
98d377a0
TH		 1809 goto out;
1810
1811 *(ptr + len) = '/';
1812 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1813
1814 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1815 break;
1816
b3b4aa74 1817 btrfs_release_path(path);
98d377a0 1818 key.objectid = key.offset;
8ad6fcab 1819 key.offset = (u64)-1;
98d377a0 1820 dirid = key.objectid;
98d377a0 1821 }
ac8e9819 1822 if (ptr < name)
98d377a0 1823 goto out;
77906a50 1824 memmove(name, ptr, total_len);
98d377a0
TH		 1825 name[total_len]='\0';
1826 ret = 0;
1827out:
1828 btrfs_free_path(path);
ac8e9819
CM		 1829 return ret;
1830}
1831
1832static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1833 void __user *argp)
1834{
1835 struct btrfs_ioctl_ino_lookup_args *args;
1836 struct inode *inode;
1837 int ret;
1838
1839 if (!capable(CAP_SYS_ADMIN))
1840 return -EPERM;
1841
2354d08f
JL		 1842 args = memdup_user(argp, sizeof(*args));
1843 if (IS_ERR(args))
1844 return PTR_ERR(args);
c2b96929 1845
ac8e9819
CM		 1846 inode = fdentry(file)->d_inode;
1847
1b53ac4d
CM		 1848 if (args->treeid == 0)
1849 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
1850
ac8e9819
CM		 1851 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
1852 args->treeid, args->objectid,
1853 args->name);
1854
1855 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1856 ret = -EFAULT;
1857
1858 kfree(args);
98d377a0
TH		 1859 return ret;
1860}
1861
76dda93c
YZ		 1862static noinline int btrfs_ioctl_snap_destroy(struct file *file,
1863 void __user *arg)
1864{
1865 struct dentry *parent = fdentry(file);
1866 struct dentry *dentry;
1867 struct inode *dir = parent->d_inode;
1868 struct inode *inode;
1869 struct btrfs_root *root = BTRFS_I(dir)->root;
1870 struct btrfs_root *dest = NULL;
1871 struct btrfs_ioctl_vol_args *vol_args;
1872 struct btrfs_trans_handle *trans;
1873 int namelen;
1874 int ret;
1875 int err = 0;
1876
76dda93c
YZ		 1877 vol_args = memdup_user(arg, sizeof(*vol_args));
1878 if (IS_ERR(vol_args))
1879 return PTR_ERR(vol_args);
1880
1881 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
1882 namelen = strlen(vol_args->name);
1883 if (strchr(vol_args->name, '/') ||
1884 strncmp(vol_args->name, "..", namelen) == 0) {
1885 err = -EINVAL;
1886 goto out;
1887 }
1888
a561be71 1889 err = mnt_want_write_file(file);
76dda93c
YZ		 1890 if (err)
1891 goto out;
1892
1893 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
1894 dentry = lookup_one_len(vol_args->name, parent, namelen);
1895 if (IS_ERR(dentry)) {
1896 err = PTR_ERR(dentry);
1897 goto out_unlock_dir;
1898 }
1899
1900 if (!dentry->d_inode) {
1901 err = -ENOENT;
1902 goto out_dput;
1903 }
1904
1905 inode = dentry->d_inode;
4260f7c7
SW		 1906 dest = BTRFS_I(inode)->root;
1907 if (!capable(CAP_SYS_ADMIN)){
1908 /*
1909 * Regular user. Only allow this with a special mount
1910 * option, when the user has write+exec access to the
1911 * subvol root, and when rmdir(2) would have been
1912 * allowed.
1913 *
1914 * Note that this is _not_ check that the subvol is
1915 * empty or doesn't contain data that we wouldn't
1916 * otherwise be able to delete.
1917 *
1918 * Users who want to delete empty subvols should try
1919 * rmdir(2).
1920 */
1921 err = -EPERM;
1922 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
1923 goto out_dput;
1924
1925 /*
1926 * Do not allow deletion if the parent dir is the same
1927 * as the dir to be deleted. That means the ioctl
1928 * must be called on the dentry referencing the root
1929 * of the subvol, not a random directory contained
1930 * within it.
1931 */
1932 err = -EINVAL;
1933 if (root == dest)
1934 goto out_dput;
1935
1936 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
1937 if (err)
1938 goto out_dput;
1939
1940 /* check if subvolume may be deleted by a non-root user */
1941 err = btrfs_may_delete(dir, dentry, 1);
1942 if (err)
1943 goto out_dput;
1944 }
1945
33345d01 1946 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
76dda93c
YZ		 1947 err = -EINVAL;
1948 goto out_dput;
1949 }
1950
76dda93c
YZ		 1951 mutex_lock(&inode->i_mutex);
1952 err = d_invalidate(dentry);
1953 if (err)
1954 goto out_unlock;
1955
1956 down_write(&root->fs_info->subvol_sem);
1957
1958 err = may_destroy_subvol(dest);
1959 if (err)
1960 goto out_up_write;
1961
a22285a6
YZ		 1962 trans = btrfs_start_transaction(root, 0);
1963 if (IS_ERR(trans)) {
1964 err = PTR_ERR(trans);
d327099a 1965 goto out_up_write;
a22285a6
YZ		 1966 }
1967 trans->block_rsv = &root->fs_info->global_block_rsv;
1968
76dda93c
YZ		 1969 ret = btrfs_unlink_subvol(trans, root, dir,
1970 dest->root_key.objectid,
1971 dentry->d_name.name,
1972 dentry->d_name.len);
1973 BUG_ON(ret);
1974
1975 btrfs_record_root_in_trans(trans, dest);
1976
1977 memset(&dest->root_item.drop_progress, 0,
1978 sizeof(dest->root_item.drop_progress));
1979 dest->root_item.drop_level = 0;
1980 btrfs_set_root_refs(&dest->root_item, 0);
1981
d68fc57b
YZ		 1982 if (!xchg(&dest->orphan_item_inserted, 1)) {
1983 ret = btrfs_insert_orphan_item(trans,
1984 root->fs_info->tree_root,
1985 dest->root_key.objectid);
1986 BUG_ON(ret);
1987 }
76dda93c 1988
531cb13f 1989 ret = btrfs_end_transaction(trans, root);
76dda93c
YZ		 1990 BUG_ON(ret);
1991 inode->i_flags |= S_DEAD;
1992out_up_write:
1993 up_write(&root->fs_info->subvol_sem);
1994out_unlock:
1995 mutex_unlock(&inode->i_mutex);
1996 if (!err) {
efefb143 1997 shrink_dcache_sb(root->fs_info->sb);
76dda93c
YZ		 1998 btrfs_invalidate_inodes(dest);
1999 d_delete(dentry);
2000 }
2001out_dput:
2002 dput(dentry);
2003out_unlock_dir:
2004 mutex_unlock(&dir->i_mutex);
2a79f17e 2005 mnt_drop_write_file(file);
76dda93c
YZ		 2006out:
2007 kfree(vol_args);
2008 return err;
2009}
2010
1e701a32 2011static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
f46b5a66
CH		 2012{
2013 struct inode *inode = fdentry(file)->d_inode;
2014 struct btrfs_root *root = BTRFS_I(inode)->root;
1e701a32 2015 struct btrfs_ioctl_defrag_range_args *range;
c146afad
YZ		 2016 int ret;
2017
b83cc969
LZ		 2018 if (btrfs_root_readonly(root))
2019 return -EROFS;
2020
a561be71 2021 ret = mnt_want_write_file(file);
c146afad
YZ		 2022 if (ret)
2023 return ret;
f46b5a66
CH		 2024
2025 switch (inode->i_mode & S_IFMT) {
2026 case S_IFDIR:
e441d54d
CM		 2027 if (!capable(CAP_SYS_ADMIN)) {
2028 ret = -EPERM;
2029 goto out;
2030 }
8929ecfa
YZ		 2031 ret = btrfs_defrag_root(root, 0);
2032 if (ret)
2033 goto out;
2034 ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
f46b5a66
CH		 2035 break;
2036 case S_IFREG:
e441d54d
CM		 2037 if (!(file->f_mode & FMODE_WRITE)) {
2038 ret = -EINVAL;
2039 goto out;
2040 }
1e701a32
CM		 2041
2042 range = kzalloc(sizeof(*range), GFP_KERNEL);
2043 if (!range) {
2044 ret = -ENOMEM;
2045 goto out;
2046 }
2047
2048 if (argp) {
2049 if (copy_from_user(range, argp,
2050 sizeof(*range))) {
2051 ret = -EFAULT;
2052 kfree(range);
683be16e 2053 goto out;
1e701a32
CM		 2054 }
2055 /* compression requires us to start the IO */
2056 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2057 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2058 range->extent_thresh = (u32)-1;
2059 }
2060 } else {
2061 /* the rest are all set to zero by kzalloc */
2062 range->len = (u64)-1;
2063 }
4cb5300b
CM		 2064 ret = btrfs_defrag_file(fdentry(file)->d_inode, file,
2065 range, 0, 0);
2066 if (ret > 0)
2067 ret = 0;
1e701a32 2068 kfree(range);
f46b5a66 2069 break;
8929ecfa
YZ		 2070 default:
2071 ret = -EINVAL;
f46b5a66 2072 }
e441d54d 2073out:
2a79f17e 2074 mnt_drop_write_file(file);
e441d54d 2075 return ret;
f46b5a66
CH		 2076}
2077
b2950863 2078static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
f46b5a66
CH		 2079{
2080 struct btrfs_ioctl_vol_args *vol_args;
2081 int ret;
2082
e441d54d
CM		 2083 if (!capable(CAP_SYS_ADMIN))
2084 return -EPERM;
2085
c9e9f97b
ID		 2086 mutex_lock(&root->fs_info->volume_mutex);
2087 if (root->fs_info->balance_ctl) {
2088 printk(KERN_INFO "btrfs: balance in progress\n");
2089 ret = -EINVAL;
2090 goto out;
2091 }
2092
dae7b665 2093 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 2094 if (IS_ERR(vol_args)) {
2095 ret = PTR_ERR(vol_args);
2096 goto out;
2097 }
f46b5a66 2098
5516e595 2099 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66
CH		 2100 ret = btrfs_init_new_device(root, vol_args->name);
2101
f46b5a66 2102 kfree(vol_args);
c9e9f97b
ID		 2103out:
2104 mutex_unlock(&root->fs_info->volume_mutex);
f46b5a66
CH		 2105 return ret;
2106}
2107
b2950863 2108static long btrfs_ioctl_rm_dev(struct btrfs_root *root, void __user *arg)
f46b5a66
CH		 2109{
2110 struct btrfs_ioctl_vol_args *vol_args;
2111 int ret;
2112
e441d54d
CM		 2113 if (!capable(CAP_SYS_ADMIN))
2114 return -EPERM;
2115
c146afad
YZ		 2116 if (root->fs_info->sb->s_flags & MS_RDONLY)
2117 return -EROFS;
2118
c9e9f97b
ID		 2119 mutex_lock(&root->fs_info->volume_mutex);
2120 if (root->fs_info->balance_ctl) {
2121 printk(KERN_INFO "btrfs: balance in progress\n");
2122 ret = -EINVAL;
2123 goto out;
2124 }
2125
dae7b665 2126 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 2127 if (IS_ERR(vol_args)) {
2128 ret = PTR_ERR(vol_args);
2129 goto out;
2130 }
f46b5a66 2131
5516e595 2132 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66
CH		 2133 ret = btrfs_rm_device(root, vol_args->name);
2134
f46b5a66 2135 kfree(vol_args);
c9e9f97b
ID		 2136out:
2137 mutex_unlock(&root->fs_info->volume_mutex);
f46b5a66
CH		 2138 return ret;
2139}
2140
475f6387
JS		 2141static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2142{
027ed2f0 2143 struct btrfs_ioctl_fs_info_args *fi_args;
475f6387
JS		 2144 struct btrfs_device *device;
2145 struct btrfs_device *next;
2146 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
027ed2f0 2147 int ret = 0;
475f6387
JS		 2148
2149 if (!capable(CAP_SYS_ADMIN))
2150 return -EPERM;
2151
027ed2f0
LZ		 2152 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2153 if (!fi_args)
2154 return -ENOMEM;
2155
2156 fi_args->num_devices = fs_devices->num_devices;
2157 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
475f6387
JS		 2158
2159 mutex_lock(&fs_devices->device_list_mutex);
2160 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
027ed2f0
LZ		 2161 if (device->devid > fi_args->max_id)
2162 fi_args->max_id = device->devid;
475f6387
JS		 2163 }
2164 mutex_unlock(&fs_devices->device_list_mutex);
2165
027ed2f0
LZ		 2166 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2167 ret = -EFAULT;
475f6387 2168
027ed2f0
LZ		 2169 kfree(fi_args);
2170 return ret;
475f6387
JS		 2171}
2172
2173static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2174{
2175 struct btrfs_ioctl_dev_info_args *di_args;
2176 struct btrfs_device *dev;
2177 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2178 int ret = 0;
2179 char *s_uuid = NULL;
2180 char empty_uuid[BTRFS_UUID_SIZE] = {0};
2181
2182 if (!capable(CAP_SYS_ADMIN))
2183 return -EPERM;
2184
2185 di_args = memdup_user(arg, sizeof(*di_args));
2186 if (IS_ERR(di_args))
2187 return PTR_ERR(di_args);
2188
2189 if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2190 s_uuid = di_args->uuid;
2191
2192 mutex_lock(&fs_devices->device_list_mutex);
2193 dev = btrfs_find_device(root, di_args->devid, s_uuid, NULL);
2194 mutex_unlock(&fs_devices->device_list_mutex);
2195
2196 if (!dev) {
2197 ret = -ENODEV;
2198 goto out;
2199 }
2200
2201 di_args->devid = dev->devid;
2202 di_args->bytes_used = dev->bytes_used;
2203 di_args->total_bytes = dev->total_bytes;
2204 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
2205 strncpy(di_args->path, dev->name, sizeof(di_args->path));
2206
2207out:
2208 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2209 ret = -EFAULT;
2210
2211 kfree(di_args);
2212 return ret;
2213}
2214
76dda93c
YZ		 2215static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2216 u64 off, u64 olen, u64 destoff)
f46b5a66
CH		 2217{
2218 struct inode *inode = fdentry(file)->d_inode;
2219 struct btrfs_root *root = BTRFS_I(inode)->root;
2220 struct file *src_file;
2221 struct inode *src;
2222 struct btrfs_trans_handle *trans;
f46b5a66 2223 struct btrfs_path *path;
f46b5a66 2224 struct extent_buffer *leaf;
ae01a0ab
YZ		 2225 char *buf;
2226 struct btrfs_key key;
f46b5a66
CH		 2227 u32 nritems;
2228 int slot;
ae01a0ab 2229 int ret;
c5c9cd4d
SW		 2230 u64 len = olen;
2231 u64 bs = root->fs_info->sb->s_blocksize;
2232 u64 hint_byte;
d20f7043 2233
c5c9cd4d
SW		 2234 /*
2235 * TODO:
2236 * - split compressed inline extents. annoying: we need to
2237 * decompress into destination's address_space (the file offset
2238 * may change, so source mapping won't do), then recompress (or
2239 * otherwise reinsert) a subrange.
2240 * - allow ranges within the same file to be cloned (provided
2241 * they don't overlap)?
2242 */
2243
e441d54d 2244 /* the destination must be opened for writing */
2ebc3464 2245 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
e441d54d
CM		 2246 return -EINVAL;
2247
b83cc969
LZ		 2248 if (btrfs_root_readonly(root))
2249 return -EROFS;
2250
a561be71 2251 ret = mnt_want_write_file(file);
c146afad
YZ		 2252 if (ret)
2253 return ret;
2254
c5c9cd4d 2255 src_file = fget(srcfd);
ab67b7c1
YZ		 2256 if (!src_file) {
2257 ret = -EBADF;
2258 goto out_drop_write;
2259 }
5dc64164 2260
f46b5a66
CH		 2261 src = src_file->f_dentry->d_inode;
2262
c5c9cd4d
SW		 2263 ret = -EINVAL;
2264 if (src == inode)
2265 goto out_fput;
2266
5dc64164
DR		 2267 /* the src must be open for reading */
2268 if (!(src_file->f_mode & FMODE_READ))
2269 goto out_fput;
2270
0e7b824c
LZ		 2271 /* don't make the dst file partly checksummed */
2272 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2273 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2274 goto out_fput;
2275
ae01a0ab
YZ		 2276 ret = -EISDIR;
2277 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2278 goto out_fput;
2279
f46b5a66 2280 ret = -EXDEV;
ae01a0ab
YZ		 2281 if (src->i_sb != inode->i_sb || BTRFS_I(src)->root != root)
2282 goto out_fput;
2283
2284 ret = -ENOMEM;
2285 buf = vmalloc(btrfs_level_size(root, 0));
2286 if (!buf)
2287 goto out_fput;
2288
2289 path = btrfs_alloc_path();
2290 if (!path) {
2291 vfree(buf);
f46b5a66 2292 goto out_fput;
ae01a0ab
YZ		 2293 }
2294 path->reada = 2;
f46b5a66
CH		 2295
2296 if (inode < src) {
fccdae43
SW		 2297 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2298 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
f46b5a66 2299 } else {
fccdae43
SW		 2300 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2301 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
f46b5a66
CH		 2302 }
2303
c5c9cd4d
SW		 2304 /* determine range to clone */
2305 ret = -EINVAL;
2ebc3464 2306 if (off + len > src->i_size || off + len < off)
f46b5a66 2307 goto out_unlock;
c5c9cd4d
SW		 2308 if (len == 0)
2309 olen = len = src->i_size - off;
2310 /* if we extend to eof, continue to block boundary */
2311 if (off + len == src->i_size)
2a6b8dae 2312 len = ALIGN(src->i_size, bs) - off;
c5c9cd4d
SW		 2313
2314 /* verify the end result is block aligned */
2a6b8dae
LZ		 2315 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2316 !IS_ALIGNED(destoff, bs))
c5c9cd4d
SW		 2317 goto out_unlock;
2318
d525e8ab
LZ		 2319 if (destoff > inode->i_size) {
2320 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2321 if (ret)
2322 goto out_unlock;
2323 }
2324
71ef0786
LZ		 2325 /* truncate page cache pages from target inode range */
2326 truncate_inode_pages_range(&inode->i_data, destoff,
2327 PAGE_CACHE_ALIGN(destoff + len) - 1);
2328
f46b5a66
CH		 2329 /* do any pending delalloc/csum calc on src, one way or
2330 another, and lock file content */
2331 while (1) {
31840ae1 2332 struct btrfs_ordered_extent *ordered;
d0082371 2333 lock_extent(&BTRFS_I(src)->io_tree, off, off+len);
9a019196
SW		 2334 ordered = btrfs_lookup_first_ordered_extent(src, off+len);
2335 if (!ordered &&
2336 !test_range_bit(&BTRFS_I(src)->io_tree, off, off+len,
2337 EXTENT_DELALLOC, 0, NULL))
f46b5a66 2338 break;
d0082371 2339 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len);
ae01a0ab
YZ		 2340 if (ordered)
2341 btrfs_put_ordered_extent(ordered);
9a019196 2342 btrfs_wait_ordered_range(src, off, len);
f46b5a66
CH		 2343 }
2344
c5c9cd4d 2345 /* clone data */
33345d01 2346 key.objectid = btrfs_ino(src);
ae01a0ab
YZ		 2347 key.type = BTRFS_EXTENT_DATA_KEY;
2348 key.offset = 0;
f46b5a66
CH		 2349
2350 while (1) {
2351 /*
2352 * note the key will change type as we walk through the
2353 * tree.
2354 */
a22285a6 2355 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
f46b5a66
CH		 2356 if (ret < 0)
2357 goto out;
2358
ae01a0ab
YZ		 2359 nritems = btrfs_header_nritems(path->nodes[0]);
2360 if (path->slots[0] >= nritems) {
f46b5a66
CH		 2361 ret = btrfs_next_leaf(root, path);
2362 if (ret < 0)
2363 goto out;
2364 if (ret > 0)
2365 break;
ae01a0ab 2366 nritems = btrfs_header_nritems(path->nodes[0]);
f46b5a66
CH		 2367 }
2368 leaf = path->nodes[0];
2369 slot = path->slots[0];
f46b5a66 2370
ae01a0ab 2371 btrfs_item_key_to_cpu(leaf, &key, slot);
d20f7043 2372 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
33345d01 2373 key.objectid != btrfs_ino(src))
f46b5a66
CH		 2374 break;
2375
c5c9cd4d
SW		 2376 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2377 struct btrfs_file_extent_item *extent;
2378 int type;
31840ae1
ZY		 2379 u32 size;
2380 struct btrfs_key new_key;
c5c9cd4d
SW		 2381 u64 disko = 0, diskl = 0;
2382 u64 datao = 0, datal = 0;
2383 u8 comp;
b5384d48 2384 u64 endoff;
31840ae1
ZY		 2385
2386 size = btrfs_item_size_nr(leaf, slot);
2387 read_extent_buffer(leaf, buf,
2388 btrfs_item_ptr_offset(leaf, slot),
2389 size);
c5c9cd4d
SW		 2390
2391 extent = btrfs_item_ptr(leaf, slot,
2392 struct btrfs_file_extent_item);
2393 comp = btrfs_file_extent_compression(leaf, extent);
2394 type = btrfs_file_extent_type(leaf, extent);
c8a894d7
CM		 2395 if (type == BTRFS_FILE_EXTENT_REG ||
2396 type == BTRFS_FILE_EXTENT_PREALLOC) {
d397712b
CM		 2397 disko = btrfs_file_extent_disk_bytenr(leaf,
2398 extent);
2399 diskl = btrfs_file_extent_disk_num_bytes(leaf,
2400 extent);
c5c9cd4d 2401 datao = btrfs_file_extent_offset(leaf, extent);
d397712b
CM		 2402 datal = btrfs_file_extent_num_bytes(leaf,
2403 extent);
c5c9cd4d
SW		 2404 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2405 /* take upper bound, may be compressed */
2406 datal = btrfs_file_extent_ram_bytes(leaf,
2407 extent);
2408 }
b3b4aa74 2409 btrfs_release_path(path);
31840ae1 2410
050006a7 2411 if (key.offset + datal <= off ||
c5c9cd4d
SW		 2412 key.offset >= off+len)
2413 goto next;
2414
31840ae1 2415 memcpy(&new_key, &key, sizeof(new_key));
33345d01 2416 new_key.objectid = btrfs_ino(inode);
4d728ec7
LZ		 2417 if (off <= key.offset)
2418 new_key.offset = key.offset + destoff - off;
2419 else
2420 new_key.offset = destoff;
31840ae1 2421
b6f3409b
SW		 2422 /*
2423 * 1 - adjusting old extent (we may have to split it)
2424 * 1 - add new extent
2425 * 1 - inode update
2426 */
2427 trans = btrfs_start_transaction(root, 3);
a22285a6
YZ		 2428 if (IS_ERR(trans)) {
2429 ret = PTR_ERR(trans);
2430 goto out;
2431 }
2432
c8a894d7
CM		 2433 if (type == BTRFS_FILE_EXTENT_REG ||
2434 type == BTRFS_FILE_EXTENT_PREALLOC) {
d72c0842
LZ		 2435 /*
2436 * a | --- range to clone ---| b
2437 * | ------------- extent ------------- |
2438 */
2439
2440 /* substract range b */
2441 if (key.offset + datal > off + len)
2442 datal = off + len - key.offset;
2443
2444 /* substract range a */
a22285a6
YZ		 2445 if (off > key.offset) {
2446 datao += off - key.offset;
2447 datal -= off - key.offset;
2448 }
2449
a22285a6
YZ		 2450 ret = btrfs_drop_extents(trans, inode,
2451 new_key.offset,
2452 new_key.offset + datal,
2453 &hint_byte, 1);
2454 BUG_ON(ret);
2455
c5c9cd4d
SW		 2456 ret = btrfs_insert_empty_item(trans, root, path,
2457 &new_key, size);
a22285a6 2458 BUG_ON(ret);
c5c9cd4d
SW		 2459
2460 leaf = path->nodes[0];
2461 slot = path->slots[0];
2462 write_extent_buffer(leaf, buf,
31840ae1
ZY		 2463 btrfs_item_ptr_offset(leaf, slot),
2464 size);
ae01a0ab 2465
c5c9cd4d 2466 extent = btrfs_item_ptr(leaf, slot,
f46b5a66 2467 struct btrfs_file_extent_item);
c5c9cd4d 2468
c5c9cd4d
SW		 2469 /* disko == 0 means it's a hole */
2470 if (!disko)
2471 datao = 0;
c5c9cd4d
SW		 2472
2473 btrfs_set_file_extent_offset(leaf, extent,
2474 datao);
2475 btrfs_set_file_extent_num_bytes(leaf, extent,
2476 datal);
2477 if (disko) {
2478 inode_add_bytes(inode, datal);
ae01a0ab 2479 ret = btrfs_inc_extent_ref(trans, root,
5d4f98a2
YZ		 2480 disko, diskl, 0,
2481 root->root_key.objectid,
33345d01 2482 btrfs_ino(inode),
66d7e7f0
AJ		 2483 new_key.offset - datao,
2484 0);
31840ae1 2485 BUG_ON(ret);
f46b5a66 2486 }
c5c9cd4d
SW		 2487 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2488 u64 skip = 0;
2489 u64 trim = 0;
2490 if (off > key.offset) {
2491 skip = off - key.offset;
2492 new_key.offset += skip;
2493 }
d397712b 2494
c5c9cd4d
SW		 2495 if (key.offset + datal > off+len)
2496 trim = key.offset + datal - (off+len);
d397712b 2497
c5c9cd4d 2498 if (comp && (skip || trim)) {
c5c9cd4d 2499 ret = -EINVAL;
a22285a6 2500 btrfs_end_transaction(trans, root);
c5c9cd4d
SW		 2501 goto out;
2502 }
2503 size -= skip + trim;
2504 datal -= skip + trim;
a22285a6
YZ		 2505
2506 ret = btrfs_drop_extents(trans, inode,
2507 new_key.offset,
2508 new_key.offset + datal,
2509 &hint_byte, 1);
2510 BUG_ON(ret);
2511
c5c9cd4d
SW		 2512 ret = btrfs_insert_empty_item(trans, root, path,
2513 &new_key, size);
a22285a6 2514 BUG_ON(ret);
c5c9cd4d
SW		 2515
2516 if (skip) {
d397712b
CM		 2517 u32 start =
2518 btrfs_file_extent_calc_inline_size(0);
c5c9cd4d
SW		 2519 memmove(buf+start, buf+start+skip,
2520 datal);
2521 }
2522
2523 leaf = path->nodes[0];
2524 slot = path->slots[0];
2525 write_extent_buffer(leaf, buf,
2526 btrfs_item_ptr_offset(leaf, slot),
2527 size);
2528 inode_add_bytes(inode, datal);
f46b5a66 2529 }
c5c9cd4d
SW		 2530
2531 btrfs_mark_buffer_dirty(leaf);
b3b4aa74 2532 btrfs_release_path(path);
c5c9cd4d 2533
a22285a6 2534 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
b5384d48
SW		 2535
2536 /*
2537 * we round up to the block size at eof when
2538 * determining which extents to clone above,
2539 * but shouldn't round up the file size
2540 */
2541 endoff = new_key.offset + datal;
5f3888ff
LZ		 2542 if (endoff > destoff+olen)
2543 endoff = destoff+olen;
b5384d48
SW		 2544 if (endoff > inode->i_size)
2545 btrfs_i_size_write(inode, endoff);
2546
a22285a6
YZ		 2547 ret = btrfs_update_inode(trans, root, inode);
2548 BUG_ON(ret);
2549 btrfs_end_transaction(trans, root);
2550 }
d397712b 2551next:
b3b4aa74 2552 btrfs_release_path(path);
f46b5a66 2553 key.offset++;
f46b5a66 2554 }
f46b5a66
CH		 2555 ret = 0;
2556out:
b3b4aa74 2557 btrfs_release_path(path);
d0082371 2558 unlock_extent(&BTRFS_I(src)->io_tree, off, off+len);
f46b5a66
CH		 2559out_unlock:
2560 mutex_unlock(&src->i_mutex);
2561 mutex_unlock(&inode->i_mutex);
ae01a0ab
YZ		 2562 vfree(buf);
2563 btrfs_free_path(path);
f46b5a66
CH		 2564out_fput:
2565 fput(src_file);
ab67b7c1 2566out_drop_write:
2a79f17e 2567 mnt_drop_write_file(file);
f46b5a66
CH		 2568 return ret;
2569}
2570
7a865e8a 2571static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
c5c9cd4d
SW		 2572{
2573 struct btrfs_ioctl_clone_range_args args;
2574
7a865e8a 2575 if (copy_from_user(&args, argp, sizeof(args)))
c5c9cd4d
SW		 2576 return -EFAULT;
2577 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2578 args.src_length, args.dest_offset);
2579}
2580
f46b5a66
CH		 2581/*
2582 * there are many ways the trans_start and trans_end ioctls can lead
2583 * to deadlocks. They should only be used by applications that
2584 * basically own the machine, and have a very in depth understanding
2585 * of all the possible deadlocks and enospc problems.
2586 */
b2950863 2587static long btrfs_ioctl_trans_start(struct file *file)
f46b5a66
CH		 2588{
2589 struct inode *inode = fdentry(file)->d_inode;
2590 struct btrfs_root *root = BTRFS_I(inode)->root;
2591 struct btrfs_trans_handle *trans;
1ab86aed 2592 int ret;
f46b5a66 2593
1ab86aed 2594 ret = -EPERM;
df5b5520 2595 if (!capable(CAP_SYS_ADMIN))
1ab86aed 2596 goto out;
df5b5520 2597
1ab86aed
SW		 2598 ret = -EINPROGRESS;
2599 if (file->private_data)
f46b5a66 2600 goto out;
9ca9ee09 2601
b83cc969
LZ		 2602 ret = -EROFS;
2603 if (btrfs_root_readonly(root))
2604 goto out;
2605
a561be71 2606 ret = mnt_want_write_file(file);
c146afad
YZ		 2607 if (ret)
2608 goto out;
2609
a4abeea4 2610 atomic_inc(&root->fs_info->open_ioctl_trans);
9ca9ee09 2611
1ab86aed 2612 ret = -ENOMEM;
7a7eaa40 2613 trans = btrfs_start_ioctl_transaction(root);
abd30bb0 2614 if (IS_ERR(trans))
1ab86aed
SW		 2615 goto out_drop;
2616
2617 file->private_data = trans;
2618 return 0;
2619
2620out_drop:
a4abeea4 2621 atomic_dec(&root->fs_info->open_ioctl_trans);
2a79f17e 2622 mnt_drop_write_file(file);
f46b5a66 2623out:
f46b5a66
CH		 2624 return ret;
2625}
2626
6ef5ed0d
JB		 2627static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2628{
2629 struct inode *inode = fdentry(file)->d_inode;
2630 struct btrfs_root *root = BTRFS_I(inode)->root;
2631 struct btrfs_root *new_root;
2632 struct btrfs_dir_item *di;
2633 struct btrfs_trans_handle *trans;
2634 struct btrfs_path *path;
2635 struct btrfs_key location;
2636 struct btrfs_disk_key disk_key;
2637 struct btrfs_super_block *disk_super;
2638 u64 features;
2639 u64 objectid = 0;
2640 u64 dir_id;
2641
2642 if (!capable(CAP_SYS_ADMIN))
2643 return -EPERM;
2644
2645 if (copy_from_user(&objectid, argp, sizeof(objectid)))
2646 return -EFAULT;
2647
2648 if (!objectid)
2649 objectid = root->root_key.objectid;
2650
2651 location.objectid = objectid;
2652 location.type = BTRFS_ROOT_ITEM_KEY;
2653 location.offset = (u64)-1;
2654
2655 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
2656 if (IS_ERR(new_root))
2657 return PTR_ERR(new_root);
2658
2659 if (btrfs_root_refs(&new_root->root_item) == 0)
2660 return -ENOENT;
2661
2662 path = btrfs_alloc_path();
2663 if (!path)
2664 return -ENOMEM;
2665 path->leave_spinning = 1;
2666
2667 trans = btrfs_start_transaction(root, 1);
98d5dc13 2668 if (IS_ERR(trans)) {
6ef5ed0d 2669 btrfs_free_path(path);
98d5dc13 2670 return PTR_ERR(trans);
6ef5ed0d
JB		 2671 }
2672
6c41761f 2673 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
6ef5ed0d
JB		 2674 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2675 dir_id, "default", 7, 1);
cf1e99a4 2676 if (IS_ERR_OR_NULL(di)) {
6ef5ed0d
JB		 2677 btrfs_free_path(path);
2678 btrfs_end_transaction(trans, root);
2679 printk(KERN_ERR "Umm, you don't have the default dir item, "
2680 "this isn't going to work\n");
2681 return -ENOENT;
2682 }
2683
2684 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2685 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2686 btrfs_mark_buffer_dirty(path->nodes[0]);
2687 btrfs_free_path(path);
2688
6c41761f 2689 disk_super = root->fs_info->super_copy;
6ef5ed0d
JB		 2690 features = btrfs_super_incompat_flags(disk_super);
2691 if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) {
2692 features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL;
2693 btrfs_set_super_incompat_flags(disk_super, features);
2694 }
2695 btrfs_end_transaction(trans, root);
2696
2697 return 0;
2698}
2699
bf5fc093
JB		 2700static void get_block_group_info(struct list_head *groups_list,
2701 struct btrfs_ioctl_space_info *space)
2702{
2703 struct btrfs_block_group_cache *block_group;
2704
2705 space->total_bytes = 0;
2706 space->used_bytes = 0;
2707 space->flags = 0;
2708 list_for_each_entry(block_group, groups_list, list) {
2709 space->flags = block_group->flags;
2710 space->total_bytes += block_group->key.offset;
2711 space->used_bytes +=
2712 btrfs_block_group_used(&block_group->item);
2713 }
2714}
2715
1406e432
JB		 2716long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2717{
2718 struct btrfs_ioctl_space_args space_args;
2719 struct btrfs_ioctl_space_info space;
2720 struct btrfs_ioctl_space_info *dest;
7fde62bf 2721 struct btrfs_ioctl_space_info *dest_orig;
13f2696f 2722 struct btrfs_ioctl_space_info __user *user_dest;
1406e432 2723 struct btrfs_space_info *info;
bf5fc093
JB		 2724 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2725 BTRFS_BLOCK_GROUP_SYSTEM,
2726 BTRFS_BLOCK_GROUP_METADATA,
2727 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2728 int num_types = 4;
7fde62bf 2729 int alloc_size;
1406e432 2730 int ret = 0;
51788b1b 2731 u64 slot_count = 0;
bf5fc093 2732 int i, c;
1406e432
JB		 2733
2734 if (copy_from_user(&space_args,
2735 (struct btrfs_ioctl_space_args __user *)arg,
2736 sizeof(space_args)))
2737 return -EFAULT;
2738
bf5fc093
JB		 2739 for (i = 0; i < num_types; i++) {
2740 struct btrfs_space_info *tmp;
2741
2742 info = NULL;
2743 rcu_read_lock();
2744 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2745 list) {
2746 if (tmp->flags == types[i]) {
2747 info = tmp;
2748 break;
2749 }
2750 }
2751 rcu_read_unlock();
2752
2753 if (!info)
2754 continue;
2755
2756 down_read(&info->groups_sem);
2757 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2758 if (!list_empty(&info->block_groups[c]))
2759 slot_count++;
2760 }
2761 up_read(&info->groups_sem);
2762 }
7fde62bf
CM		 2763
2764 /* space_slots == 0 means they are asking for a count */
2765 if (space_args.space_slots == 0) {
2766 space_args.total_spaces = slot_count;
2767 goto out;
2768 }
bf5fc093 2769
51788b1b 2770 slot_count = min_t(u64, space_args.space_slots, slot_count);
bf5fc093 2771
7fde62bf 2772 alloc_size = sizeof(*dest) * slot_count;
bf5fc093 2773
7fde62bf
CM		 2774 /* we generally have at most 6 or so space infos, one for each raid
2775 * level. So, a whole page should be more than enough for everyone
2776 */
2777 if (alloc_size > PAGE_CACHE_SIZE)
2778 return -ENOMEM;
2779
1406e432 2780 space_args.total_spaces = 0;
7fde62bf
CM		 2781 dest = kmalloc(alloc_size, GFP_NOFS);
2782 if (!dest)
2783 return -ENOMEM;
2784 dest_orig = dest;
1406e432 2785
7fde62bf 2786 /* now we have a buffer to copy into */
bf5fc093
JB		 2787 for (i = 0; i < num_types; i++) {
2788 struct btrfs_space_info *tmp;
2789
51788b1b
DR		 2790 if (!slot_count)
2791 break;
2792
bf5fc093
JB		 2793 info = NULL;
2794 rcu_read_lock();
2795 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2796 list) {
2797 if (tmp->flags == types[i]) {
2798 info = tmp;
2799 break;
2800 }
2801 }
2802 rcu_read_unlock();
7fde62bf 2803
bf5fc093
JB		 2804 if (!info)
2805 continue;
2806 down_read(&info->groups_sem);
2807 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2808 if (!list_empty(&info->block_groups[c])) {
2809 get_block_group_info(&info->block_groups[c],
2810 &space);
2811 memcpy(dest, &space, sizeof(space));
2812 dest++;
2813 space_args.total_spaces++;
51788b1b 2814 slot_count--;
bf5fc093 2815 }
51788b1b
DR		 2816 if (!slot_count)
2817 break;
bf5fc093
JB		 2818 }
2819 up_read(&info->groups_sem);
1406e432 2820 }
1406e432 2821
7fde62bf
CM		 2822 user_dest = (struct btrfs_ioctl_space_info *)
2823 (arg + sizeof(struct btrfs_ioctl_space_args));
2824
2825 if (copy_to_user(user_dest, dest_orig, alloc_size))
2826 ret = -EFAULT;
2827
2828 kfree(dest_orig);
2829out:
2830 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
1406e432
JB		 2831 ret = -EFAULT;
2832
2833 return ret;
2834}
2835
f46b5a66
CH		 2836/*
2837 * there are many ways the trans_start and trans_end ioctls can lead
2838 * to deadlocks. They should only be used by applications that
2839 * basically own the machine, and have a very in depth understanding
2840 * of all the possible deadlocks and enospc problems.
2841 */
2842long btrfs_ioctl_trans_end(struct file *file)
2843{
2844 struct inode *inode = fdentry(file)->d_inode;
2845 struct btrfs_root *root = BTRFS_I(inode)->root;
2846 struct btrfs_trans_handle *trans;
f46b5a66 2847
f46b5a66 2848 trans = file->private_data;
1ab86aed
SW		 2849 if (!trans)
2850 return -EINVAL;
b214107e 2851 file->private_data = NULL;
9ca9ee09 2852
1ab86aed
SW		 2853 btrfs_end_transaction(trans, root);
2854
a4abeea4 2855 atomic_dec(&root->fs_info->open_ioctl_trans);
9ca9ee09 2856
2a79f17e 2857 mnt_drop_write_file(file);
1ab86aed 2858 return 0;
f46b5a66
CH		 2859}
2860
46204592
SW		 2861static noinline long btrfs_ioctl_start_sync(struct file *file, void __user *argp)
2862{
2863 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2864 struct btrfs_trans_handle *trans;
2865 u64 transid;
db5b493a 2866 int ret;
46204592
SW		 2867
2868 trans = btrfs_start_transaction(root, 0);
98d5dc13
TI		 2869 if (IS_ERR(trans))
2870 return PTR_ERR(trans);
46204592 2871 transid = trans->transid;
db5b493a 2872 ret = btrfs_commit_transaction_async(trans, root, 0);
8b2b2d3c
TI		 2873 if (ret) {
2874 btrfs_end_transaction(trans, root);
db5b493a 2875 return ret;
8b2b2d3c 2876 }
46204592
SW		 2877
2878 if (argp)
2879 if (copy_to_user(argp, &transid, sizeof(transid)))
2880 return -EFAULT;
2881 return 0;
2882}
2883
2884static noinline long btrfs_ioctl_wait_sync(struct file *file, void __user *argp)
2885{
2886 struct btrfs_root *root = BTRFS_I(file->f_dentry->d_inode)->root;
2887 u64 transid;
2888
2889 if (argp) {
2890 if (copy_from_user(&transid, argp, sizeof(transid)))
2891 return -EFAULT;
2892 } else {
2893 transid = 0; /* current trans */
2894 }
2895 return btrfs_wait_for_commit(root, transid);
2896}
2897
475f6387
JS		 2898static long btrfs_ioctl_scrub(struct btrfs_root *root, void __user *arg)
2899{
2900 int ret;
2901 struct btrfs_ioctl_scrub_args *sa;
2902
2903 if (!capable(CAP_SYS_ADMIN))
2904 return -EPERM;
2905
2906 sa = memdup_user(arg, sizeof(*sa));
2907 if (IS_ERR(sa))
2908 return PTR_ERR(sa);
2909
2910 ret = btrfs_scrub_dev(root, sa->devid, sa->start, sa->end,
8628764e 2911 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY);
475f6387
JS		 2912
2913 if (copy_to_user(arg, sa, sizeof(*sa)))
2914 ret = -EFAULT;
2915
2916 kfree(sa);
2917 return ret;
2918}
2919
2920static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
2921{
2922 if (!capable(CAP_SYS_ADMIN))
2923 return -EPERM;
2924
2925 return btrfs_scrub_cancel(root);
2926}
2927
2928static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
2929 void __user *arg)
2930{
2931 struct btrfs_ioctl_scrub_args *sa;
2932 int ret;
2933
2934 if (!capable(CAP_SYS_ADMIN))
2935 return -EPERM;
2936
2937 sa = memdup_user(arg, sizeof(*sa));
2938 if (IS_ERR(sa))
2939 return PTR_ERR(sa);
2940
2941 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
2942
2943 if (copy_to_user(arg, sa, sizeof(*sa)))
2944 ret = -EFAULT;
2945
2946 kfree(sa);
2947 return ret;
2948}
2949
d7728c96
JS		 2950static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
2951{
2952 int ret = 0;
2953 int i;
740c3d22 2954 u64 rel_ptr;
d7728c96 2955 int size;
806468f8 2956 struct btrfs_ioctl_ino_path_args *ipa = NULL;
d7728c96
JS		 2957 struct inode_fs_paths *ipath = NULL;
2958 struct btrfs_path *path;
2959
2960 if (!capable(CAP_SYS_ADMIN))
2961 return -EPERM;
2962
2963 path = btrfs_alloc_path();
2964 if (!path) {
2965 ret = -ENOMEM;
2966 goto out;
2967 }
2968
2969 ipa = memdup_user(arg, sizeof(*ipa));
2970 if (IS_ERR(ipa)) {
2971 ret = PTR_ERR(ipa);
2972 ipa = NULL;
2973 goto out;
2974 }
2975
2976 size = min_t(u32, ipa->size, 4096);
2977 ipath = init_ipath(size, root, path);
2978 if (IS_ERR(ipath)) {
2979 ret = PTR_ERR(ipath);
2980 ipath = NULL;
2981 goto out;
2982 }
2983
2984 ret = paths_from_inode(ipa->inum, ipath);
2985 if (ret < 0)
2986 goto out;
2987
2988 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
745c4d8e
JM		 2989 rel_ptr = ipath->fspath->val[i] -
2990 (u64)(unsigned long)ipath->fspath->val;
740c3d22 2991 ipath->fspath->val[i] = rel_ptr;
d7728c96
JS		 2992 }
2993
745c4d8e
JM		 2994 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
2995 (void *)(unsigned long)ipath->fspath, size);
d7728c96
JS		 2996 if (ret) {
2997 ret = -EFAULT;
2998 goto out;
2999 }
3000
3001out:
3002 btrfs_free_path(path);
3003 free_ipath(ipath);
3004 kfree(ipa);
3005
3006 return ret;
3007}
3008
3009static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3010{
3011 struct btrfs_data_container *inodes = ctx;
3012 const size_t c = 3 * sizeof(u64);
3013
3014 if (inodes->bytes_left >= c) {
3015 inodes->bytes_left -= c;
3016 inodes->val[inodes->elem_cnt] = inum;
3017 inodes->val[inodes->elem_cnt + 1] = offset;
3018 inodes->val[inodes->elem_cnt + 2] = root;
3019 inodes->elem_cnt += 3;
3020 } else {
3021 inodes->bytes_missing += c - inodes->bytes_left;
3022 inodes->bytes_left = 0;
3023 inodes->elem_missed += 3;
3024 }
3025
3026 return 0;
3027}
3028
3029static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3030 void __user *arg)
3031{
3032 int ret = 0;
3033 int size;
4692cf58 3034 u64 extent_item_pos;
d7728c96
JS		 3035 struct btrfs_ioctl_logical_ino_args *loi;
3036 struct btrfs_data_container *inodes = NULL;
3037 struct btrfs_path *path = NULL;
3038 struct btrfs_key key;
3039
3040 if (!capable(CAP_SYS_ADMIN))
3041 return -EPERM;
3042
3043 loi = memdup_user(arg, sizeof(*loi));
3044 if (IS_ERR(loi)) {
3045 ret = PTR_ERR(loi);
3046 loi = NULL;
3047 goto out;
3048 }
3049
3050 path = btrfs_alloc_path();
3051 if (!path) {
3052 ret = -ENOMEM;
3053 goto out;
3054 }
3055
3056 size = min_t(u32, loi->size, 4096);
3057 inodes = init_data_container(size);
3058 if (IS_ERR(inodes)) {
3059 ret = PTR_ERR(inodes);
3060 inodes = NULL;
3061 goto out;
3062 }
3063
3064 ret = extent_from_logical(root->fs_info, loi->logical, path, &key);
4692cf58 3065 btrfs_release_path(path);
d7728c96
JS		 3066
3067 if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK)
3068 ret = -ENOENT;
3069 if (ret < 0)
3070 goto out;
3071
4692cf58 3072 extent_item_pos = loi->logical - key.objectid;
d7728c96 3073 ret = iterate_extent_inodes(root->fs_info, path, key.objectid,
4692cf58
JS		 3074 extent_item_pos, build_ino_list,
3075 inodes);
d7728c96
JS		 3076
3077 if (ret < 0)
3078 goto out;
3079
745c4d8e
JM		 3080 ret = copy_to_user((void *)(unsigned long)loi->inodes,
3081 (void *)(unsigned long)inodes, size);
d7728c96
JS		 3082 if (ret)
3083 ret = -EFAULT;
3084
3085out:
3086 btrfs_free_path(path);
3087 kfree(inodes);
3088 kfree(loi);
3089
3090 return ret;
3091}
3092
19a39dce 3093void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
c9e9f97b
ID		 3094 struct btrfs_ioctl_balance_args *bargs)
3095{
3096 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3097
3098 bargs->flags = bctl->flags;
3099
837d5b6e
ID		 3100 if (atomic_read(&fs_info->balance_running))
3101 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3102 if (atomic_read(&fs_info->balance_pause_req))
3103 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
a7e99c69
ID		 3104 if (atomic_read(&fs_info->balance_cancel_req))
3105 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
837d5b6e 3106
c9e9f97b
ID		 3107 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3108 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3109 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
19a39dce
ID		 3110
3111 if (lock) {
3112 spin_lock(&fs_info->balance_lock);
3113 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3114 spin_unlock(&fs_info->balance_lock);
3115 } else {
3116 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3117 }
c9e9f97b
ID		 3118}
3119
3120static long btrfs_ioctl_balance(struct btrfs_root *root, void __user *arg)
3121{
3122 struct btrfs_fs_info *fs_info = root->fs_info;
3123 struct btrfs_ioctl_balance_args *bargs;
3124 struct btrfs_balance_control *bctl;
3125 int ret;
3126
3127 if (!capable(CAP_SYS_ADMIN))
3128 return -EPERM;
3129
3130 if (fs_info->sb->s_flags & MS_RDONLY)
3131 return -EROFS;
3132
3133 mutex_lock(&fs_info->volume_mutex);
3134 mutex_lock(&fs_info->balance_mutex);
3135
3136 if (arg) {
3137 bargs = memdup_user(arg, sizeof(*bargs));
3138 if (IS_ERR(bargs)) {
3139 ret = PTR_ERR(bargs);
3140 goto out;
3141 }
de322263
ID		 3142
3143 if (bargs->flags & BTRFS_BALANCE_RESUME) {
3144 if (!fs_info->balance_ctl) {
3145 ret = -ENOTCONN;
3146 goto out_bargs;
3147 }
3148
3149 bctl = fs_info->balance_ctl;
3150 spin_lock(&fs_info->balance_lock);
3151 bctl->flags |= BTRFS_BALANCE_RESUME;
3152 spin_unlock(&fs_info->balance_lock);
3153
3154 goto do_balance;
3155 }
c9e9f97b
ID		 3156 } else {
3157 bargs = NULL;
3158 }
3159
837d5b6e
ID		 3160 if (fs_info->balance_ctl) {
3161 ret = -EINPROGRESS;
3162 goto out_bargs;
3163 }
3164
c9e9f97b
ID		 3165 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3166 if (!bctl) {
3167 ret = -ENOMEM;
3168 goto out_bargs;
3169 }
3170
3171 bctl->fs_info = fs_info;
3172 if (arg) {
3173 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3174 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3175 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3176
3177 bctl->flags = bargs->flags;
f43ffb60
ID		 3178 } else {
3179 /* balance everything - no filters */
3180 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
c9e9f97b
ID		 3181 }
3182
de322263 3183do_balance:
c9e9f97b
ID		 3184 ret = btrfs_balance(bctl, bargs);
3185 /*
837d5b6e
ID		 3186 * bctl is freed in __cancel_balance or in free_fs_info if
3187 * restriper was paused all the way until unmount
c9e9f97b
ID		 3188 */
3189 if (arg) {
3190 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3191 ret = -EFAULT;
3192 }
3193
3194out_bargs:
3195 kfree(bargs);
3196out:
3197 mutex_unlock(&fs_info->balance_mutex);
3198 mutex_unlock(&fs_info->volume_mutex);
3199 return ret;
3200}
3201
837d5b6e
ID		 3202static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
3203{
3204 if (!capable(CAP_SYS_ADMIN))
3205 return -EPERM;
3206
3207 switch (cmd) {
3208 case BTRFS_BALANCE_CTL_PAUSE:
3209 return btrfs_pause_balance(root->fs_info);
a7e99c69
ID		 3210 case BTRFS_BALANCE_CTL_CANCEL:
3211 return btrfs_cancel_balance(root->fs_info);
837d5b6e
ID		 3212 }
3213
3214 return -EINVAL;
3215}
3216
19a39dce
ID		 3217static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
3218 void __user *arg)
3219{
3220 struct btrfs_fs_info *fs_info = root->fs_info;
3221 struct btrfs_ioctl_balance_args *bargs;
3222 int ret = 0;
3223
3224 if (!capable(CAP_SYS_ADMIN))
3225 return -EPERM;
3226
3227 mutex_lock(&fs_info->balance_mutex);
3228 if (!fs_info->balance_ctl) {
3229 ret = -ENOTCONN;
3230 goto out;
3231 }
3232
3233 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
3234 if (!bargs) {
3235 ret = -ENOMEM;
3236 goto out;
3237 }
3238
3239 update_ioctl_balance_args(fs_info, 1, bargs);
3240
3241 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3242 ret = -EFAULT;
3243
3244 kfree(bargs);
3245out:
3246 mutex_unlock(&fs_info->balance_mutex);
3247 return ret;
3248}
3249
f46b5a66
CH		 3250long btrfs_ioctl(struct file *file, unsigned int
3251 cmd, unsigned long arg)
3252{
3253 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
4bcabaa3 3254 void __user *argp = (void __user *)arg;
f46b5a66
CH		 3255
3256 switch (cmd) {
6cbff00f
CH		 3257 case FS_IOC_GETFLAGS:
3258 return btrfs_ioctl_getflags(file, argp);
3259 case FS_IOC_SETFLAGS:
3260 return btrfs_ioctl_setflags(file, argp);
3261 case FS_IOC_GETVERSION:
3262 return btrfs_ioctl_getversion(file, argp);
f7039b1d
LD		 3263 case FITRIM:
3264 return btrfs_ioctl_fitrim(file, argp);
f46b5a66 3265 case BTRFS_IOC_SNAP_CREATE:
fa0d2b9b 3266 return btrfs_ioctl_snap_create(file, argp, 0);
fdfb1e4f 3267 case BTRFS_IOC_SNAP_CREATE_V2:
fa0d2b9b 3268 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3de4586c 3269 case BTRFS_IOC_SUBVOL_CREATE:
fa0d2b9b 3270 return btrfs_ioctl_snap_create(file, argp, 1);
76dda93c
YZ		 3271 case BTRFS_IOC_SNAP_DESTROY:
3272 return btrfs_ioctl_snap_destroy(file, argp);
0caa102d
LZ		 3273 case BTRFS_IOC_SUBVOL_GETFLAGS:
3274 return btrfs_ioctl_subvol_getflags(file, argp);
3275 case BTRFS_IOC_SUBVOL_SETFLAGS:
3276 return btrfs_ioctl_subvol_setflags(file, argp);
6ef5ed0d
JB		 3277 case BTRFS_IOC_DEFAULT_SUBVOL:
3278 return btrfs_ioctl_default_subvol(file, argp);
f46b5a66 3279 case BTRFS_IOC_DEFRAG:
1e701a32
CM		 3280 return btrfs_ioctl_defrag(file, NULL);
3281 case BTRFS_IOC_DEFRAG_RANGE:
3282 return btrfs_ioctl_defrag(file, argp);
f46b5a66 3283 case BTRFS_IOC_RESIZE:
4bcabaa3 3284 return btrfs_ioctl_resize(root, argp);
f46b5a66 3285 case BTRFS_IOC_ADD_DEV:
4bcabaa3 3286 return btrfs_ioctl_add_dev(root, argp);
f46b5a66 3287 case BTRFS_IOC_RM_DEV:
4bcabaa3 3288 return btrfs_ioctl_rm_dev(root, argp);
475f6387
JS		 3289 case BTRFS_IOC_FS_INFO:
3290 return btrfs_ioctl_fs_info(root, argp);
3291 case BTRFS_IOC_DEV_INFO:
3292 return btrfs_ioctl_dev_info(root, argp);
f46b5a66 3293 case BTRFS_IOC_BALANCE:
c9e9f97b 3294 return btrfs_ioctl_balance(root, NULL);
f46b5a66 3295 case BTRFS_IOC_CLONE:
c5c9cd4d
SW		 3296 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
3297 case BTRFS_IOC_CLONE_RANGE:
7a865e8a 3298 return btrfs_ioctl_clone_range(file, argp);
f46b5a66
CH		 3299 case BTRFS_IOC_TRANS_START:
3300 return btrfs_ioctl_trans_start(file);
3301 case BTRFS_IOC_TRANS_END:
3302 return btrfs_ioctl_trans_end(file);
ac8e9819
CM		 3303 case BTRFS_IOC_TREE_SEARCH:
3304 return btrfs_ioctl_tree_search(file, argp);
3305 case BTRFS_IOC_INO_LOOKUP:
3306 return btrfs_ioctl_ino_lookup(file, argp);
d7728c96
JS		 3307 case BTRFS_IOC_INO_PATHS:
3308 return btrfs_ioctl_ino_to_path(root, argp);
3309 case BTRFS_IOC_LOGICAL_INO:
3310 return btrfs_ioctl_logical_to_ino(root, argp);
1406e432
JB		 3311 case BTRFS_IOC_SPACE_INFO:
3312 return btrfs_ioctl_space_info(root, argp);
f46b5a66
CH		 3313 case BTRFS_IOC_SYNC:
3314 btrfs_sync_fs(file->f_dentry->d_sb, 1);
3315 return 0;
46204592
SW		 3316 case BTRFS_IOC_START_SYNC:
3317 return btrfs_ioctl_start_sync(file, argp);
3318 case BTRFS_IOC_WAIT_SYNC:
3319 return btrfs_ioctl_wait_sync(file, argp);
475f6387
JS		 3320 case BTRFS_IOC_SCRUB:
3321 return btrfs_ioctl_scrub(root, argp);
3322 case BTRFS_IOC_SCRUB_CANCEL:
3323 return btrfs_ioctl_scrub_cancel(root, argp);
3324 case BTRFS_IOC_SCRUB_PROGRESS:
3325 return btrfs_ioctl_scrub_progress(root, argp);
c9e9f97b
ID		 3326 case BTRFS_IOC_BALANCE_V2:
3327 return btrfs_ioctl_balance(root, argp);
837d5b6e
ID		 3328 case BTRFS_IOC_BALANCE_CTL:
3329 return btrfs_ioctl_balance_ctl(root, arg);
19a39dce
ID		 3330 case BTRFS_IOC_BALANCE_PROGRESS:
3331 return btrfs_ioctl_balance_progress(root, argp);
f46b5a66
CH		 3332 }
3333
3334 return -ENOTTY;
3335}
Empty description
This page took 0.664948 seconds and 4 git commands to generate.