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