]> git.ipfire.org Git - thirdparty/linux.git/blame - fs/btrfs/ioctl.c
projects / thirdparty / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Btrfs: add mount option to force UUID tree checking
[thirdparty/linux.git] / fs / btrfs / ioctl.c
CommitLineData
f46b5a66
CH		 1/*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
19#include <linux/kernel.h>
20#include <linux/bio.h>
21#include <linux/buffer_head.h>
22#include <linux/file.h>
23#include <linux/fs.h>
cb8e7090 24#include <linux/fsnotify.h>
f46b5a66
CH		 25#include <linux/pagemap.h>
26#include <linux/highmem.h>
27#include <linux/time.h>
28#include <linux/init.h>
29#include <linux/string.h>
f46b5a66 30#include <linux/backing-dev.h>
cb8e7090 31#include <linux/mount.h>
f46b5a66 32#include <linux/mpage.h>
cb8e7090 33#include <linux/namei.h>
f46b5a66
CH		 34#include <linux/swap.h>
35#include <linux/writeback.h>
36#include <linux/statfs.h>
37#include <linux/compat.h>
38#include <linux/bit_spinlock.h>
cb8e7090 39#include <linux/security.h>
f46b5a66 40#include <linux/xattr.h>
7ea394f1 41#include <linux/vmalloc.h>
5a0e3ad6 42#include <linux/slab.h>
f7039b1d 43#include <linux/blkdev.h>
8ea05e3a 44#include <linux/uuid.h>
55e301fd 45#include <linux/btrfs.h>
416161db 46#include <linux/uaccess.h>
4b4e25f2 47#include "compat.h"
f46b5a66
CH		 48#include "ctree.h"
49#include "disk-io.h"
50#include "transaction.h"
51#include "btrfs_inode.h"
f46b5a66
CH		 52#include "print-tree.h"
53#include "volumes.h"
925baedd 54#include "locking.h"
581bb050 55#include "inode-map.h"
d7728c96 56#include "backref.h"
606686ee 57#include "rcu-string.h"
31db9f7c 58#include "send.h"
3f6bcfbd 59#include "dev-replace.h"
f46b5a66 60
416161db
MF		 61static int btrfs_clone(struct inode *src, struct inode *inode,
62 u64 off, u64 olen, u64 olen_aligned, u64 destoff);
63
6cbff00f
CH		 64/* Mask out flags that are inappropriate for the given type of inode. */
65static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
66{
67 if (S_ISDIR(mode))
68 return flags;
69 else if (S_ISREG(mode))
70 return flags & ~FS_DIRSYNC_FL;
71 else
72 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
73}
74
75/*
76 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
77 */
78static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
79{
80 unsigned int iflags = 0;
81
82 if (flags & BTRFS_INODE_SYNC)
83 iflags |= FS_SYNC_FL;
84 if (flags & BTRFS_INODE_IMMUTABLE)
85 iflags |= FS_IMMUTABLE_FL;
86 if (flags & BTRFS_INODE_APPEND)
87 iflags |= FS_APPEND_FL;
88 if (flags & BTRFS_INODE_NODUMP)
89 iflags |= FS_NODUMP_FL;
90 if (flags & BTRFS_INODE_NOATIME)
91 iflags |= FS_NOATIME_FL;
92 if (flags & BTRFS_INODE_DIRSYNC)
93 iflags |= FS_DIRSYNC_FL;
d0092bdd
LZ		 94 if (flags & BTRFS_INODE_NODATACOW)
95 iflags |= FS_NOCOW_FL;
96
97 if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
98 iflags |= FS_COMPR_FL;
99 else if (flags & BTRFS_INODE_NOCOMPRESS)
100 iflags |= FS_NOCOMP_FL;
6cbff00f
CH		 101
102 return iflags;
103}
104
105/*
106 * Update inode->i_flags based on the btrfs internal flags.
107 */
108void btrfs_update_iflags(struct inode *inode)
109{
110 struct btrfs_inode *ip = BTRFS_I(inode);
111
112 inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
113
114 if (ip->flags & BTRFS_INODE_SYNC)
115 inode->i_flags |= S_SYNC;
116 if (ip->flags & BTRFS_INODE_IMMUTABLE)
117 inode->i_flags |= S_IMMUTABLE;
118 if (ip->flags & BTRFS_INODE_APPEND)
119 inode->i_flags |= S_APPEND;
120 if (ip->flags & BTRFS_INODE_NOATIME)
121 inode->i_flags |= S_NOATIME;
122 if (ip->flags & BTRFS_INODE_DIRSYNC)
123 inode->i_flags |= S_DIRSYNC;
124}
125
126/*
127 * Inherit flags from the parent inode.
128 *
e27425d6 129 * Currently only the compression flags and the cow flags are inherited.
6cbff00f
CH		 130 */
131void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
132{
0b4dcea5
CM		 133 unsigned int flags;
134
135 if (!dir)
136 return;
137
138 flags = BTRFS_I(dir)->flags;
6cbff00f 139
e27425d6
JB		 140 if (flags & BTRFS_INODE_NOCOMPRESS) {
141 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
142 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
143 } else if (flags & BTRFS_INODE_COMPRESS) {
144 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
145 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
146 }
147
213490b3 148 if (flags & BTRFS_INODE_NODATACOW) {
e27425d6 149 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
213490b3
LB		 150 if (S_ISREG(inode->i_mode))
151 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
152 }
6cbff00f 153
6cbff00f
CH		 154 btrfs_update_iflags(inode);
155}
156
157static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
158{
496ad9aa 159 struct btrfs_inode *ip = BTRFS_I(file_inode(file));
6cbff00f
CH		 160 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
161
162 if (copy_to_user(arg, &flags, sizeof(flags)))
163 return -EFAULT;
164 return 0;
165}
166
75e7cb7f
LB		 167static int check_flags(unsigned int flags)
168{
169 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
170 FS_NOATIME_FL | FS_NODUMP_FL | \
171 FS_SYNC_FL | FS_DIRSYNC_FL | \
e1e8fb6a
LZ		 172 FS_NOCOMP_FL | FS_COMPR_FL |
173 FS_NOCOW_FL))
75e7cb7f
LB		 174 return -EOPNOTSUPP;
175
176 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
177 return -EINVAL;
178
75e7cb7f
LB		 179 return 0;
180}
181
6cbff00f
CH		 182static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
183{
496ad9aa 184 struct inode *inode = file_inode(file);
6cbff00f
CH		 185 struct btrfs_inode *ip = BTRFS_I(inode);
186 struct btrfs_root *root = ip->root;
187 struct btrfs_trans_handle *trans;
188 unsigned int flags, oldflags;
189 int ret;
f062abf0
LZ		 190 u64 ip_oldflags;
191 unsigned int i_oldflags;
7e97b8da 192 umode_t mode;
6cbff00f 193
b83cc969
LZ		 194 if (btrfs_root_readonly(root))
195 return -EROFS;
196
6cbff00f
CH		 197 if (copy_from_user(&flags, arg, sizeof(flags)))
198 return -EFAULT;
199
75e7cb7f
LB		 200 ret = check_flags(flags);
201 if (ret)
202 return ret;
f46b5a66 203
2e149670 204 if (!inode_owner_or_capable(inode))
6cbff00f
CH		 205 return -EACCES;
206
e7848683
JK		 207 ret = mnt_want_write_file(file);
208 if (ret)
209 return ret;
210
6cbff00f
CH		 211 mutex_lock(&inode->i_mutex);
212
f062abf0
LZ		 213 ip_oldflags = ip->flags;
214 i_oldflags = inode->i_flags;
7e97b8da 215 mode = inode->i_mode;
f062abf0 216
6cbff00f
CH		 217 flags = btrfs_mask_flags(inode->i_mode, flags);
218 oldflags = btrfs_flags_to_ioctl(ip->flags);
219 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
220 if (!capable(CAP_LINUX_IMMUTABLE)) {
221 ret = -EPERM;
222 goto out_unlock;
223 }
224 }
225
6cbff00f
CH		 226 if (flags & FS_SYNC_FL)
227 ip->flags |= BTRFS_INODE_SYNC;
228 else
229 ip->flags &= ~BTRFS_INODE_SYNC;
230 if (flags & FS_IMMUTABLE_FL)
231 ip->flags |= BTRFS_INODE_IMMUTABLE;
232 else
233 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
234 if (flags & FS_APPEND_FL)
235 ip->flags |= BTRFS_INODE_APPEND;
236 else
237 ip->flags &= ~BTRFS_INODE_APPEND;
238 if (flags & FS_NODUMP_FL)
239 ip->flags |= BTRFS_INODE_NODUMP;
240 else
241 ip->flags &= ~BTRFS_INODE_NODUMP;
242 if (flags & FS_NOATIME_FL)
243 ip->flags |= BTRFS_INODE_NOATIME;
244 else
245 ip->flags &= ~BTRFS_INODE_NOATIME;
246 if (flags & FS_DIRSYNC_FL)
247 ip->flags |= BTRFS_INODE_DIRSYNC;
248 else
249 ip->flags &= ~BTRFS_INODE_DIRSYNC;
7e97b8da
DS		 250 if (flags & FS_NOCOW_FL) {
251 if (S_ISREG(mode)) {
252 /*
253 * It's safe to turn csums off here, no extents exist.
254 * Otherwise we want the flag to reflect the real COW
255 * status of the file and will not set it.
256 */
257 if (inode->i_size == 0)
258 ip->flags |= BTRFS_INODE_NODATACOW
259 | BTRFS_INODE_NODATASUM;
260 } else {
261 ip->flags |= BTRFS_INODE_NODATACOW;
262 }
263 } else {
264 /*
265 * Revert back under same assuptions as above
266 */
267 if (S_ISREG(mode)) {
268 if (inode->i_size == 0)
269 ip->flags &= ~(BTRFS_INODE_NODATACOW
270 | BTRFS_INODE_NODATASUM);
271 } else {
272 ip->flags &= ~BTRFS_INODE_NODATACOW;
273 }
274 }
6cbff00f 275
75e7cb7f
LB		 276 /*
277 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
278 * flag may be changed automatically if compression code won't make
279 * things smaller.
280 */
281 if (flags & FS_NOCOMP_FL) {
282 ip->flags &= ~BTRFS_INODE_COMPRESS;
283 ip->flags |= BTRFS_INODE_NOCOMPRESS;
284 } else if (flags & FS_COMPR_FL) {
285 ip->flags |= BTRFS_INODE_COMPRESS;
286 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
ebcb904d
LZ		 287 } else {
288 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
75e7cb7f 289 }
6cbff00f 290
4da6f1a3 291 trans = btrfs_start_transaction(root, 1);
f062abf0
LZ		 292 if (IS_ERR(trans)) {
293 ret = PTR_ERR(trans);
294 goto out_drop;
295 }
6cbff00f 296
306424cc 297 btrfs_update_iflags(inode);
0c4d2d95 298 inode_inc_iversion(inode);
306424cc 299 inode->i_ctime = CURRENT_TIME;
6cbff00f 300 ret = btrfs_update_inode(trans, root, inode);
6cbff00f 301
6cbff00f 302 btrfs_end_transaction(trans, root);
f062abf0
LZ		 303 out_drop:
304 if (ret) {
305 ip->flags = ip_oldflags;
306 inode->i_flags = i_oldflags;
307 }
6cbff00f 308
6cbff00f
CH		 309 out_unlock:
310 mutex_unlock(&inode->i_mutex);
e7848683 311 mnt_drop_write_file(file);
2d4e6f6a 312 return ret;
6cbff00f
CH		 313}
314
315static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
316{
496ad9aa 317 struct inode *inode = file_inode(file);
6cbff00f
CH		 318
319 return put_user(inode->i_generation, arg);
320}
f46b5a66 321
f7039b1d
LD		 322static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
323{
815745cf 324 struct btrfs_fs_info *fs_info = btrfs_sb(fdentry(file)->d_sb);
f7039b1d
LD		 325 struct btrfs_device *device;
326 struct request_queue *q;
327 struct fstrim_range range;
328 u64 minlen = ULLONG_MAX;
329 u64 num_devices = 0;
815745cf 330 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
f7039b1d
LD		 331 int ret;
332
333 if (!capable(CAP_SYS_ADMIN))
334 return -EPERM;
335
1f78160c
XG		 336 rcu_read_lock();
337 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
338 dev_list) {
f7039b1d
LD		 339 if (!device->bdev)
340 continue;
341 q = bdev_get_queue(device->bdev);
342 if (blk_queue_discard(q)) {
343 num_devices++;
344 minlen = min((u64)q->limits.discard_granularity,
345 minlen);
346 }
347 }
1f78160c 348 rcu_read_unlock();
f4c697e6 349
f7039b1d
LD		 350 if (!num_devices)
351 return -EOPNOTSUPP;
f7039b1d
LD		 352 if (copy_from_user(&range, arg, sizeof(range)))
353 return -EFAULT;
e515c18b
LC		 354 if (range.start > total_bytes ||
355 range.len < fs_info->sb->s_blocksize)
f4c697e6 356 return -EINVAL;
f7039b1d 357
f4c697e6 358 range.len = min(range.len, total_bytes - range.start);
f7039b1d 359 range.minlen = max(range.minlen, minlen);
815745cf 360 ret = btrfs_trim_fs(fs_info->tree_root, &range);
f7039b1d
LD		 361 if (ret < 0)
362 return ret;
363
364 if (copy_to_user(arg, &range, sizeof(range)))
365 return -EFAULT;
366
367 return 0;
368}
369
dd5f9615
SB		 370int btrfs_is_empty_uuid(u8 *uuid)
371{
372 static char empty_uuid[BTRFS_UUID_SIZE] = {0};
373
374 return !memcmp(uuid, empty_uuid, BTRFS_UUID_SIZE);
375}
376
d5c12070 377static noinline int create_subvol(struct inode *dir,
cb8e7090 378 struct dentry *dentry,
72fd032e 379 char *name, int namelen,
6f72c7e2 380 u64 *async_transid,
8696c533 381 struct btrfs_qgroup_inherit *inherit)
f46b5a66
CH		 382{
383 struct btrfs_trans_handle *trans;
384 struct btrfs_key key;
385 struct btrfs_root_item root_item;
386 struct btrfs_inode_item *inode_item;
387 struct extent_buffer *leaf;
d5c12070 388 struct btrfs_root *root = BTRFS_I(dir)->root;
76dda93c 389 struct btrfs_root *new_root;
d5c12070 390 struct btrfs_block_rsv block_rsv;
8ea05e3a 391 struct timespec cur_time = CURRENT_TIME;
f46b5a66
CH		 392 int ret;
393 int err;
394 u64 objectid;
395 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
3de4586c 396 u64 index = 0;
d5c12070 397 u64 qgroup_reserved;
8ea05e3a 398 uuid_le new_uuid;
f46b5a66 399
581bb050 400 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
2fbe8c8a 401 if (ret)
a22285a6 402 return ret;
6a912213 403
d5c12070 404 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
9ed74f2d 405 /*
d5c12070
MX		 406 * The same as the snapshot creation, please see the comment
407 * of create_snapshot().
9ed74f2d 408 */
d5c12070 409 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
dd5f9615 410 8, &qgroup_reserved, false);
d5c12070
MX		 411 if (ret)
412 return ret;
413
414 trans = btrfs_start_transaction(root, 0);
415 if (IS_ERR(trans)) {
416 ret = PTR_ERR(trans);
417 goto out;
418 }
419 trans->block_rsv = &block_rsv;
420 trans->bytes_reserved = block_rsv.size;
f46b5a66 421
8696c533 422 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
6f72c7e2
AJ		 423 if (ret)
424 goto fail;
425
5d4f98a2 426 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
5581a51a 427 0, objectid, NULL, 0, 0, 0);
8e8a1e31
JB		 428 if (IS_ERR(leaf)) {
429 ret = PTR_ERR(leaf);
430 goto fail;
431 }
f46b5a66 432
5d4f98a2 433 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
f46b5a66
CH		 434 btrfs_set_header_bytenr(leaf, leaf->start);
435 btrfs_set_header_generation(leaf, trans->transid);
5d4f98a2 436 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
f46b5a66
CH		 437 btrfs_set_header_owner(leaf, objectid);
438
439 write_extent_buffer(leaf, root->fs_info->fsid,
440 (unsigned long)btrfs_header_fsid(leaf),
441 BTRFS_FSID_SIZE);
5d4f98a2
YZ		 442 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
443 (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
444 BTRFS_UUID_SIZE);
f46b5a66
CH		 445 btrfs_mark_buffer_dirty(leaf);
446
8ea05e3a
AB		 447 memset(&root_item, 0, sizeof(root_item));
448
f46b5a66 449 inode_item = &root_item.inode;
3cae210f
QW		 450 btrfs_set_stack_inode_generation(inode_item, 1);
451 btrfs_set_stack_inode_size(inode_item, 3);
452 btrfs_set_stack_inode_nlink(inode_item, 1);
453 btrfs_set_stack_inode_nbytes(inode_item, root->leafsize);
454 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
f46b5a66 455
3cae210f
QW		 456 btrfs_set_root_flags(&root_item, 0);
457 btrfs_set_root_limit(&root_item, 0);
458 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
08fe4db1 459
f46b5a66 460 btrfs_set_root_bytenr(&root_item, leaf->start);
84234f3a 461 btrfs_set_root_generation(&root_item, trans->transid);
f46b5a66
CH		 462 btrfs_set_root_level(&root_item, 0);
463 btrfs_set_root_refs(&root_item, 1);
86b9f2ec 464 btrfs_set_root_used(&root_item, leaf->len);
80ff3856 465 btrfs_set_root_last_snapshot(&root_item, 0);
f46b5a66 466
8ea05e3a
AB		 467 btrfs_set_root_generation_v2(&root_item,
468 btrfs_root_generation(&root_item));
469 uuid_le_gen(&new_uuid);
470 memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
3cae210f
QW		 471 btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec);
472 btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec);
8ea05e3a
AB		 473 root_item.ctime = root_item.otime;
474 btrfs_set_root_ctransid(&root_item, trans->transid);
475 btrfs_set_root_otransid(&root_item, trans->transid);
f46b5a66 476
925baedd 477 btrfs_tree_unlock(leaf);
f46b5a66
CH		 478 free_extent_buffer(leaf);
479 leaf = NULL;
480
481 btrfs_set_root_dirid(&root_item, new_dirid);
482
483 key.objectid = objectid;
5d4f98a2 484 key.offset = 0;
f46b5a66
CH		 485 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
486 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
487 &root_item);
488 if (ret)
489 goto fail;
490
76dda93c
YZ		 491 key.offset = (u64)-1;
492 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
79787eaa
JM		 493 if (IS_ERR(new_root)) {
494 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
495 ret = PTR_ERR(new_root);
496 goto fail;
497 }
76dda93c
YZ		 498
499 btrfs_record_root_in_trans(trans, new_root);
500
d82a6f1d 501 ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
ce598979
MF		 502 if (ret) {
503 /* We potentially lose an unused inode item here */
79787eaa 504 btrfs_abort_transaction(trans, root, ret);
ce598979
MF		 505 goto fail;
506 }
507
f46b5a66
CH		 508 /*
509 * insert the directory item
510 */
3de4586c 511 ret = btrfs_set_inode_index(dir, &index);
79787eaa
JM		 512 if (ret) {
513 btrfs_abort_transaction(trans, root, ret);
514 goto fail;
515 }
3de4586c
CM		 516
517 ret = btrfs_insert_dir_item(trans, root,
16cdcec7 518 name, namelen, dir, &key,
3de4586c 519 BTRFS_FT_DIR, index);
79787eaa
JM		 520 if (ret) {
521 btrfs_abort_transaction(trans, root, ret);
f46b5a66 522 goto fail;
79787eaa 523 }
0660b5af 524
52c26179
YZ		 525 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
526 ret = btrfs_update_inode(trans, root, dir);
527 BUG_ON(ret);
528
0660b5af 529 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
4df27c4d 530 objectid, root->root_key.objectid,
33345d01 531 btrfs_ino(dir), index, name, namelen);
76dda93c 532 BUG_ON(ret);
f46b5a66 533
dd5f9615
SB		 534 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
535 root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
536 objectid);
537 if (ret)
538 btrfs_abort_transaction(trans, root, ret);
539
f46b5a66 540fail:
d5c12070
MX		 541 trans->block_rsv = NULL;
542 trans->bytes_reserved = 0;
72fd032e
SW		 543 if (async_transid) {
544 *async_transid = trans->transid;
545 err = btrfs_commit_transaction_async(trans, root, 1);
00d71c9c
MX		 546 if (err)
547 err = btrfs_commit_transaction(trans, root);
72fd032e
SW		 548 } else {
549 err = btrfs_commit_transaction(trans, root);
550 }
f46b5a66
CH		 551 if (err && !ret)
552 ret = err;
1a65e24b
CM		 553
554 if (!ret)
555 d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
d5c12070
MX		 556out:
557 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
f46b5a66
CH		 558 return ret;
559}
560
e9662f70
MX		 561static int create_snapshot(struct btrfs_root *root, struct inode *dir,
562 struct dentry *dentry, char *name, int namelen,
563 u64 *async_transid, bool readonly,
564 struct btrfs_qgroup_inherit *inherit)
f46b5a66 565{
2e4bfab9 566 struct inode *inode;
f46b5a66
CH		 567 struct btrfs_pending_snapshot *pending_snapshot;
568 struct btrfs_trans_handle *trans;
2e4bfab9 569 int ret;
f46b5a66
CH		 570
571 if (!root->ref_cows)
572 return -EINVAL;
573
6a03843d
MX		 574 ret = btrfs_start_delalloc_inodes(root, 0);
575 if (ret)
576 return ret;
577
578 btrfs_wait_ordered_extents(root, 0);
579
3de4586c 580 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
a22285a6
YZ		 581 if (!pending_snapshot)
582 return -ENOMEM;
583
66d8f3dd
MX		 584 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
585 BTRFS_BLOCK_RSV_TEMP);
d5c12070
MX		 586 /*
587 * 1 - parent dir inode
588 * 2 - dir entries
589 * 1 - root item
590 * 2 - root ref/backref
591 * 1 - root of snapshot
dd5f9615 592 * 1 - UUID item
d5c12070
MX		 593 */
594 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
dd5f9615 595 &pending_snapshot->block_rsv, 8,
ee3441b4
JM		 596 &pending_snapshot->qgroup_reserved,
597 false);
d5c12070
MX		 598 if (ret)
599 goto out;
600
3de4586c 601 pending_snapshot->dentry = dentry;
f46b5a66 602 pending_snapshot->root = root;
b83cc969 603 pending_snapshot->readonly = readonly;
e9662f70 604 pending_snapshot->dir = dir;
8696c533 605 pending_snapshot->inherit = inherit;
a22285a6 606
d5c12070 607 trans = btrfs_start_transaction(root, 0);
a22285a6
YZ		 608 if (IS_ERR(trans)) {
609 ret = PTR_ERR(trans);
610 goto fail;
611 }
612
8351583e 613 spin_lock(&root->fs_info->trans_lock);
f46b5a66
CH		 614 list_add(&pending_snapshot->list,
615 &trans->transaction->pending_snapshots);
8351583e 616 spin_unlock(&root->fs_info->trans_lock);
72fd032e
SW		 617 if (async_transid) {
618 *async_transid = trans->transid;
619 ret = btrfs_commit_transaction_async(trans,
620 root->fs_info->extent_root, 1);
00d71c9c
MX		 621 if (ret)
622 ret = btrfs_commit_transaction(trans, root);
72fd032e
SW		 623 } else {
624 ret = btrfs_commit_transaction(trans,
625 root->fs_info->extent_root);
626 }
aec8030a 627 if (ret)
c37b2b62 628 goto fail;
a22285a6
YZ		 629
630 ret = pending_snapshot->error;
631 if (ret)
632 goto fail;
633
66b4ffd1
JB		 634 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
635 if (ret)
636 goto fail;
f46b5a66 637
2fbe8c8a 638 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
2e4bfab9
YZ		 639 if (IS_ERR(inode)) {
640 ret = PTR_ERR(inode);
641 goto fail;
642 }
643 BUG_ON(!inode);
644 d_instantiate(dentry, inode);
645 ret = 0;
646fail:
d5c12070
MX		 647 btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
648 &pending_snapshot->block_rsv,
649 pending_snapshot->qgroup_reserved);
650out:
a22285a6 651 kfree(pending_snapshot);
f46b5a66
CH		 652 return ret;
653}
654
4260f7c7
SW		 655/* copy of check_sticky in fs/namei.c()
656* It's inline, so penalty for filesystems that don't use sticky bit is
657* minimal.
658*/
659static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
660{
2f2f43d3 661 kuid_t fsuid = current_fsuid();
4260f7c7
SW		 662
663 if (!(dir->i_mode & S_ISVTX))
664 return 0;
2f2f43d3 665 if (uid_eq(inode->i_uid, fsuid))
4260f7c7 666 return 0;
2f2f43d3 667 if (uid_eq(dir->i_uid, fsuid))
4260f7c7
SW		 668 return 0;
669 return !capable(CAP_FOWNER);
670}
671
672/* copy of may_delete in fs/namei.c()
673 * Check whether we can remove a link victim from directory dir, check
674 * whether the type of victim is right.
675 * 1. We can't do it if dir is read-only (done in permission())
676 * 2. We should have write and exec permissions on dir
677 * 3. We can't remove anything from append-only dir
678 * 4. We can't do anything with immutable dir (done in permission())
679 * 5. If the sticky bit on dir is set we should either
680 * a. be owner of dir, or
681 * b. be owner of victim, or
682 * c. have CAP_FOWNER capability
683 * 6. If the victim is append-only or immutable we can't do antyhing with
684 * links pointing to it.
685 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
686 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
687 * 9. We can't remove a root or mountpoint.
688 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
689 * nfs_async_unlink().
690 */
691
692static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
693{
694 int error;
695
696 if (!victim->d_inode)
697 return -ENOENT;
698
699 BUG_ON(victim->d_parent->d_inode != dir);
4fa6b5ec 700 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
4260f7c7
SW		 701
702 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
703 if (error)
704 return error;
705 if (IS_APPEND(dir))
706 return -EPERM;
707 if (btrfs_check_sticky(dir, victim->d_inode)||
708 IS_APPEND(victim->d_inode)||
709 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
710 return -EPERM;
711 if (isdir) {
712 if (!S_ISDIR(victim->d_inode->i_mode))
713 return -ENOTDIR;
714 if (IS_ROOT(victim))
715 return -EBUSY;
716 } else if (S_ISDIR(victim->d_inode->i_mode))
717 return -EISDIR;
718 if (IS_DEADDIR(dir))
719 return -ENOENT;
720 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
721 return -EBUSY;
722 return 0;
723}
724
cb8e7090
CH		 725/* copy of may_create in fs/namei.c() */
726static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
727{
728 if (child->d_inode)
729 return -EEXIST;
730 if (IS_DEADDIR(dir))
731 return -ENOENT;
732 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
733}
734
735/*
736 * Create a new subvolume below @parent. This is largely modeled after
737 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
738 * inside this filesystem so it's quite a bit simpler.
739 */
76dda93c
YZ		 740static noinline int btrfs_mksubvol(struct path *parent,
741 char *name, int namelen,
72fd032e 742 struct btrfs_root *snap_src,
6f72c7e2 743 u64 *async_transid, bool readonly,
8696c533 744 struct btrfs_qgroup_inherit *inherit)
cb8e7090 745{
76dda93c 746 struct inode *dir = parent->dentry->d_inode;
cb8e7090
CH		 747 struct dentry *dentry;
748 int error;
749
5c50c9b8
DS		 750 error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
751 if (error == -EINTR)
752 return error;
cb8e7090
CH		 753
754 dentry = lookup_one_len(name, parent->dentry, namelen);
755 error = PTR_ERR(dentry);
756 if (IS_ERR(dentry))
757 goto out_unlock;
758
759 error = -EEXIST;
760 if (dentry->d_inode)
761 goto out_dput;
762
76dda93c 763 error = btrfs_may_create(dir, dentry);
cb8e7090 764 if (error)
a874a63e 765 goto out_dput;
cb8e7090 766
9c52057c
CM		 767 /*
768 * even if this name doesn't exist, we may get hash collisions.
769 * check for them now when we can safely fail
770 */
771 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
772 dir->i_ino, name,
773 namelen);
774 if (error)
775 goto out_dput;
776
76dda93c
YZ		 777 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
778
779 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
780 goto out_up_read;
781
3de4586c 782 if (snap_src) {
e9662f70 783 error = create_snapshot(snap_src, dir, dentry, name, namelen,
6f72c7e2 784 async_transid, readonly, inherit);
3de4586c 785 } else {
d5c12070
MX		 786 error = create_subvol(dir, dentry, name, namelen,
787 async_transid, inherit);
3de4586c 788 }
76dda93c
YZ		 789 if (!error)
790 fsnotify_mkdir(dir, dentry);
791out_up_read:
792 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
cb8e7090
CH		 793out_dput:
794 dput(dentry);
795out_unlock:
76dda93c 796 mutex_unlock(&dir->i_mutex);
cb8e7090
CH		 797 return error;
798}
799
4cb5300b
CM		 800/*
801 * When we're defragging a range, we don't want to kick it off again
802 * if it is really just waiting for delalloc to send it down.
803 * If we find a nice big extent or delalloc range for the bytes in the
804 * file you want to defrag, we return 0 to let you know to skip this
805 * part of the file
806 */
807static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
808{
809 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
810 struct extent_map *em = NULL;
811 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
812 u64 end;
813
814 read_lock(&em_tree->lock);
815 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
816 read_unlock(&em_tree->lock);
817
818 if (em) {
819 end = extent_map_end(em);
820 free_extent_map(em);
821 if (end - offset > thresh)
822 return 0;
823 }
824 /* if we already have a nice delalloc here, just stop */
825 thresh /= 2;
826 end = count_range_bits(io_tree, &offset, offset + thresh,
827 thresh, EXTENT_DELALLOC, 1);
828 if (end >= thresh)
829 return 0;
830 return 1;
831}
832
833/*
834 * helper function to walk through a file and find extents
835 * newer than a specific transid, and smaller than thresh.
836 *
837 * This is used by the defragging code to find new and small
838 * extents
839 */
840static int find_new_extents(struct btrfs_root *root,
841 struct inode *inode, u64 newer_than,
842 u64 *off, int thresh)
843{
844 struct btrfs_path *path;
845 struct btrfs_key min_key;
846 struct btrfs_key max_key;
847 struct extent_buffer *leaf;
848 struct btrfs_file_extent_item *extent;
849 int type;
850 int ret;
a4689d2b 851 u64 ino = btrfs_ino(inode);
4cb5300b
CM		 852
853 path = btrfs_alloc_path();
854 if (!path)
855 return -ENOMEM;
856
a4689d2b 857 min_key.objectid = ino;
4cb5300b
CM		 858 min_key.type = BTRFS_EXTENT_DATA_KEY;
859 min_key.offset = *off;
860
a4689d2b 861 max_key.objectid = ino;
4cb5300b
CM		 862 max_key.type = (u8)-1;
863 max_key.offset = (u64)-1;
864
865 path->keep_locks = 1;
866
867 while(1) {
868 ret = btrfs_search_forward(root, &min_key, &max_key,
de78b51a 869 path, newer_than);
4cb5300b
CM		 870 if (ret != 0)
871 goto none;
a4689d2b 872 if (min_key.objectid != ino)
4cb5300b
CM		 873 goto none;
874 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
875 goto none;
876
877 leaf = path->nodes[0];
878 extent = btrfs_item_ptr(leaf, path->slots[0],
879 struct btrfs_file_extent_item);
880
881 type = btrfs_file_extent_type(leaf, extent);
882 if (type == BTRFS_FILE_EXTENT_REG &&
883 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
884 check_defrag_in_cache(inode, min_key.offset, thresh)) {
885 *off = min_key.offset;
886 btrfs_free_path(path);
887 return 0;
888 }
889
890 if (min_key.offset == (u64)-1)
891 goto none;
892
893 min_key.offset++;
894 btrfs_release_path(path);
895 }
896none:
897 btrfs_free_path(path);
898 return -ENOENT;
899}
900
6c282eb4 901static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
17ce6ef8
LB		 902{
903 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
6c282eb4
LZ		 904 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
905 struct extent_map *em;
906 u64 len = PAGE_CACHE_SIZE;
17ce6ef8 907
6c282eb4
LZ		 908 /*
909 * hopefully we have this extent in the tree already, try without
910 * the full extent lock
911 */
17ce6ef8 912 read_lock(&em_tree->lock);
6c282eb4 913 em = lookup_extent_mapping(em_tree, start, len);
17ce6ef8
LB		 914 read_unlock(&em_tree->lock);
915
6c282eb4
LZ		 916 if (!em) {
917 /* get the big lock and read metadata off disk */
918 lock_extent(io_tree, start, start + len - 1);
919 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
920 unlock_extent(io_tree, start, start + len - 1);
921
922 if (IS_ERR(em))
923 return NULL;
924 }
925
926 return em;
927}
17ce6ef8 928
6c282eb4
LZ		 929static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
930{
931 struct extent_map *next;
932 bool ret = true;
933
934 /* this is the last extent */
935 if (em->start + em->len >= i_size_read(inode))
936 return false;
937
938 next = defrag_lookup_extent(inode, em->start + em->len);
939 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
940 ret = false;
941
942 free_extent_map(next);
17ce6ef8
LB		 943 return ret;
944}
945
6c282eb4 946static int should_defrag_range(struct inode *inode, u64 start, int thresh,
a43a2111
AM		 947 u64 *last_len, u64 *skip, u64 *defrag_end,
948 int compress)
940100a4 949{
6c282eb4 950 struct extent_map *em;
940100a4 951 int ret = 1;
6c282eb4 952 bool next_mergeable = true;
940100a4
CM		 953
954 /*
008873ea 955 * make sure that once we start defragging an extent, we keep on
940100a4
CM		 956 * defragging it
957 */
958 if (start < *defrag_end)
959 return 1;
960
961 *skip = 0;
962
6c282eb4
LZ		 963 em = defrag_lookup_extent(inode, start);
964 if (!em)
965 return 0;
940100a4
CM		 966
967 /* this will cover holes, and inline extents */
17ce6ef8 968 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
940100a4 969 ret = 0;
17ce6ef8
LB		 970 goto out;
971 }
972
6c282eb4 973 next_mergeable = defrag_check_next_extent(inode, em);
940100a4
CM		 974
975 /*
6c282eb4
LZ		 976 * we hit a real extent, if it is big or the next extent is not a
977 * real extent, don't bother defragging it
940100a4 978 */
a43a2111 979 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
6c282eb4 980 (em->len >= thresh || !next_mergeable))
940100a4 981 ret = 0;
17ce6ef8 982out:
940100a4
CM		 983 /*
984 * last_len ends up being a counter of how many bytes we've defragged.
985 * every time we choose not to defrag an extent, we reset *last_len
986 * so that the next tiny extent will force a defrag.
987 *
988 * The end result of this is that tiny extents before a single big
989 * extent will force at least part of that big extent to be defragged.
990 */
991 if (ret) {
940100a4
CM		 992 *defrag_end = extent_map_end(em);
993 } else {
994 *last_len = 0;
995 *skip = extent_map_end(em);
996 *defrag_end = 0;
997 }
998
999 free_extent_map(em);
1000 return ret;
1001}
1002
4cb5300b
CM		 1003/*
1004 * it doesn't do much good to defrag one or two pages
1005 * at a time. This pulls in a nice chunk of pages
1006 * to COW and defrag.
1007 *
1008 * It also makes sure the delalloc code has enough
1009 * dirty data to avoid making new small extents as part
1010 * of the defrag
1011 *
1012 * It's a good idea to start RA on this range
1013 * before calling this.
1014 */
1015static int cluster_pages_for_defrag(struct inode *inode,
1016 struct page **pages,
1017 unsigned long start_index,
1018 int num_pages)
f46b5a66 1019{
4cb5300b
CM		 1020 unsigned long file_end;
1021 u64 isize = i_size_read(inode);
1022 u64 page_start;
1023 u64 page_end;
1f12bd06 1024 u64 page_cnt;
4cb5300b
CM		 1025 int ret;
1026 int i;
1027 int i_done;
3eaa2885 1028 struct btrfs_ordered_extent *ordered;
4cb5300b 1029 struct extent_state *cached_state = NULL;
600a45e1 1030 struct extent_io_tree *tree;
3b16a4e3 1031 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
4cb5300b 1032
4cb5300b 1033 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1f12bd06
LB		 1034 if (!isize || start_index > file_end)
1035 return 0;
1036
1037 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
4cb5300b
CM		 1038
1039 ret = btrfs_delalloc_reserve_space(inode,
1f12bd06 1040 page_cnt << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1041 if (ret)
1042 return ret;
4cb5300b 1043 i_done = 0;
600a45e1 1044 tree = &BTRFS_I(inode)->io_tree;
4cb5300b
CM		 1045
1046 /* step one, lock all the pages */
1f12bd06 1047 for (i = 0; i < page_cnt; i++) {
4cb5300b 1048 struct page *page;
600a45e1 1049again:
a94733d0 1050 page = find_or_create_page(inode->i_mapping,
600a45e1 1051 start_index + i, mask);
4cb5300b
CM		 1052 if (!page)
1053 break;
1054
600a45e1
MX		 1055 page_start = page_offset(page);
1056 page_end = page_start + PAGE_CACHE_SIZE - 1;
1057 while (1) {
d0082371 1058 lock_extent(tree, page_start, page_end);
600a45e1
MX		 1059 ordered = btrfs_lookup_ordered_extent(inode,
1060 page_start);
d0082371 1061 unlock_extent(tree, page_start, page_end);
600a45e1
MX		 1062 if (!ordered)
1063 break;
1064
1065 unlock_page(page);
1066 btrfs_start_ordered_extent(inode, ordered, 1);
1067 btrfs_put_ordered_extent(ordered);
1068 lock_page(page);
1f12bd06
LB		 1069 /*
1070 * we unlocked the page above, so we need check if
1071 * it was released or not.
1072 */
1073 if (page->mapping != inode->i_mapping) {
1074 unlock_page(page);
1075 page_cache_release(page);
1076 goto again;
1077 }
600a45e1
MX		 1078 }
1079
4cb5300b
CM		 1080 if (!PageUptodate(page)) {
1081 btrfs_readpage(NULL, page);
1082 lock_page(page);
1083 if (!PageUptodate(page)) {
1084 unlock_page(page);
1085 page_cache_release(page);
1086 ret = -EIO;
1087 break;
1088 }
1089 }
600a45e1 1090
600a45e1
MX		 1091 if (page->mapping != inode->i_mapping) {
1092 unlock_page(page);
1093 page_cache_release(page);
1094 goto again;
1095 }
1096
4cb5300b
CM		 1097 pages[i] = page;
1098 i_done++;
1099 }
1100 if (!i_done || ret)
1101 goto out;
1102
1103 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1104 goto out;
1105
1106 /*
1107 * so now we have a nice long stream of locked
1108 * and up to date pages, lets wait on them
1109 */
1110 for (i = 0; i < i_done; i++)
1111 wait_on_page_writeback(pages[i]);
1112
1113 page_start = page_offset(pages[0]);
1114 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1115
1116 lock_extent_bits(&BTRFS_I(inode)->io_tree,
d0082371 1117 page_start, page_end - 1, 0, &cached_state);
4cb5300b
CM		 1118 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1119 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
9e8a4a8b
LB		 1120 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1121 &cached_state, GFP_NOFS);
4cb5300b 1122
1f12bd06 1123 if (i_done != page_cnt) {
9e0baf60
JB		 1124 spin_lock(&BTRFS_I(inode)->lock);
1125 BTRFS_I(inode)->outstanding_extents++;
1126 spin_unlock(&BTRFS_I(inode)->lock);
4cb5300b 1127 btrfs_delalloc_release_space(inode,
1f12bd06 1128 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1129 }
1130
1131
9e8a4a8b
LB		 1132 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1133 &cached_state, GFP_NOFS);
4cb5300b
CM		 1134
1135 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1136 page_start, page_end - 1, &cached_state,
1137 GFP_NOFS);
1138
1139 for (i = 0; i < i_done; i++) {
1140 clear_page_dirty_for_io(pages[i]);
1141 ClearPageChecked(pages[i]);
1142 set_page_extent_mapped(pages[i]);
1143 set_page_dirty(pages[i]);
1144 unlock_page(pages[i]);
1145 page_cache_release(pages[i]);
1146 }
1147 return i_done;
1148out:
1149 for (i = 0; i < i_done; i++) {
1150 unlock_page(pages[i]);
1151 page_cache_release(pages[i]);
1152 }
1f12bd06 1153 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1154 return ret;
1155
1156}
1157
1158int btrfs_defrag_file(struct inode *inode, struct file *file,
1159 struct btrfs_ioctl_defrag_range_args *range,
1160 u64 newer_than, unsigned long max_to_defrag)
1161{
1162 struct btrfs_root *root = BTRFS_I(inode)->root;
4cb5300b 1163 struct file_ra_state *ra = NULL;
f46b5a66 1164 unsigned long last_index;
151a31b2 1165 u64 isize = i_size_read(inode);
940100a4
CM		 1166 u64 last_len = 0;
1167 u64 skip = 0;
1168 u64 defrag_end = 0;
4cb5300b 1169 u64 newer_off = range->start;
f46b5a66 1170 unsigned long i;
008873ea 1171 unsigned long ra_index = 0;
f46b5a66 1172 int ret;
4cb5300b 1173 int defrag_count = 0;
1a419d85 1174 int compress_type = BTRFS_COMPRESS_ZLIB;
4cb5300b 1175 int extent_thresh = range->extent_thresh;
008873ea
LZ		 1176 int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1177 int cluster = max_cluster;
4cb5300b
CM		 1178 u64 new_align = ~((u64)128 * 1024 - 1);
1179 struct page **pages = NULL;
1180
0abd5b17
LB		 1181 if (isize == 0)
1182 return 0;
1183
1184 if (range->start >= isize)
1185 return -EINVAL;
1a419d85
LZ		 1186
1187 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1188 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1189 return -EINVAL;
1190 if (range->compress_type)
1191 compress_type = range->compress_type;
1192 }
f46b5a66 1193
0abd5b17
LB		 1194 if (extent_thresh == 0)
1195 extent_thresh = 256 * 1024;
940100a4 1196
4cb5300b
CM		 1197 /*
1198 * if we were not given a file, allocate a readahead
1199 * context
1200 */
1201 if (!file) {
1202 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1203 if (!ra)
1204 return -ENOMEM;
1205 file_ra_state_init(ra, inode->i_mapping);
1206 } else {
1207 ra = &file->f_ra;
1208 }
1209
008873ea 1210 pages = kmalloc(sizeof(struct page *) * max_cluster,
4cb5300b
CM		 1211 GFP_NOFS);
1212 if (!pages) {
1213 ret = -ENOMEM;
1214 goto out_ra;
1215 }
1216
1217 /* find the last page to defrag */
1e701a32 1218 if (range->start + range->len > range->start) {
151a31b2 1219 last_index = min_t(u64, isize - 1,
1e701a32
CM		 1220 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1221 } else {
151a31b2 1222 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1e701a32
CM		 1223 }
1224
4cb5300b
CM		 1225 if (newer_than) {
1226 ret = find_new_extents(root, inode, newer_than,
1227 &newer_off, 64 * 1024);
1228 if (!ret) {
1229 range->start = newer_off;
1230 /*
1231 * we always align our defrag to help keep
1232 * the extents in the file evenly spaced
1233 */
1234 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
4cb5300b
CM		 1235 } else
1236 goto out_ra;
1237 } else {
1238 i = range->start >> PAGE_CACHE_SHIFT;
1239 }
1240 if (!max_to_defrag)
7ec31b54 1241 max_to_defrag = last_index + 1;
4cb5300b 1242
2a0f7f57
LZ		 1243 /*
1244 * make writeback starts from i, so the defrag range can be
1245 * written sequentially.
1246 */
1247 if (i < inode->i_mapping->writeback_index)
1248 inode->i_mapping->writeback_index = i;
1249
f7f43cc8
CM		 1250 while (i <= last_index && defrag_count < max_to_defrag &&
1251 (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1252 PAGE_CACHE_SHIFT)) {
4cb5300b
CM		 1253 /*
1254 * make sure we stop running if someone unmounts
1255 * the FS
1256 */
1257 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1258 break;
1259
210549eb
DS		 1260 if (btrfs_defrag_cancelled(root->fs_info)) {
1261 printk(KERN_DEBUG "btrfs: defrag_file cancelled\n");
1262 ret = -EAGAIN;
1263 break;
1264 }
1265
66c26892 1266 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
6c282eb4 1267 extent_thresh, &last_len, &skip,
a43a2111
AM		 1268 &defrag_end, range->flags &
1269 BTRFS_DEFRAG_RANGE_COMPRESS)) {
940100a4
CM		 1270 unsigned long next;
1271 /*
1272 * the should_defrag function tells us how much to skip
1273 * bump our counter by the suggested amount
1274 */
1275 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1276 i = max(i + 1, next);
1277 continue;
1278 }
008873ea
LZ		 1279
1280 if (!newer_than) {
1281 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1282 PAGE_CACHE_SHIFT) - i;
1283 cluster = min(cluster, max_cluster);
1284 } else {
1285 cluster = max_cluster;
1286 }
1287
1e701a32 1288 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1a419d85 1289 BTRFS_I(inode)->force_compress = compress_type;
940100a4 1290
008873ea
LZ		 1291 if (i + cluster > ra_index) {
1292 ra_index = max(i, ra_index);
1293 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1294 cluster);
1295 ra_index += max_cluster;
1296 }
940100a4 1297
ecb8bea8 1298 mutex_lock(&inode->i_mutex);
008873ea 1299 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
ecb8bea8
LB		 1300 if (ret < 0) {
1301 mutex_unlock(&inode->i_mutex);
4cb5300b 1302 goto out_ra;
ecb8bea8 1303 }
4cb5300b
CM		 1304
1305 defrag_count += ret;
d0e1d66b 1306 balance_dirty_pages_ratelimited(inode->i_mapping);
ecb8bea8 1307 mutex_unlock(&inode->i_mutex);
4cb5300b
CM		 1308
1309 if (newer_than) {
1310 if (newer_off == (u64)-1)
1311 break;
1312
e1f041e1
LB		 1313 if (ret > 0)
1314 i += ret;
1315
4cb5300b
CM		 1316 newer_off = max(newer_off + 1,
1317 (u64)i << PAGE_CACHE_SHIFT);
1318
1319 ret = find_new_extents(root, inode,
1320 newer_than, &newer_off,
1321 64 * 1024);
1322 if (!ret) {
1323 range->start = newer_off;
1324 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
4cb5300b
CM		 1325 } else {
1326 break;
f46b5a66 1327 }
4cb5300b 1328 } else {
008873ea 1329 if (ret > 0) {
cbcc8326 1330 i += ret;
008873ea
LZ		 1331 last_len += ret << PAGE_CACHE_SHIFT;
1332 } else {
cbcc8326 1333 i++;
008873ea
LZ		 1334 last_len = 0;
1335 }
f46b5a66 1336 }
f46b5a66
CH		 1337 }
1338
1e701a32
CM		 1339 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1340 filemap_flush(inode->i_mapping);
1341
1342 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1343 /* the filemap_flush will queue IO into the worker threads, but
1344 * we have to make sure the IO is actually started and that
1345 * ordered extents get created before we return
1346 */
1347 atomic_inc(&root->fs_info->async_submit_draining);
1348 while (atomic_read(&root->fs_info->nr_async_submits) ||
1349 atomic_read(&root->fs_info->async_delalloc_pages)) {
1350 wait_event(root->fs_info->async_submit_wait,
1351 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1352 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1353 }
1354 atomic_dec(&root->fs_info->async_submit_draining);
1355
1356 mutex_lock(&inode->i_mutex);
261507a0 1357 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1e701a32
CM		 1358 mutex_unlock(&inode->i_mutex);
1359 }
1360
1a419d85 1361 if (range->compress_type == BTRFS_COMPRESS_LZO) {
2b0ce2c2 1362 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1a419d85
LZ		 1363 }
1364
60ccf82f 1365 ret = defrag_count;
940100a4 1366
4cb5300b
CM		 1367out_ra:
1368 if (!file)
1369 kfree(ra);
1370 kfree(pages);
940100a4 1371 return ret;
f46b5a66
CH		 1372}
1373
198605a8 1374static noinline int btrfs_ioctl_resize(struct file *file,
76dda93c 1375 void __user *arg)
f46b5a66
CH		 1376{
1377 u64 new_size;
1378 u64 old_size;
1379 u64 devid = 1;
496ad9aa 1380 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
f46b5a66
CH		 1381 struct btrfs_ioctl_vol_args *vol_args;
1382 struct btrfs_trans_handle *trans;
1383 struct btrfs_device *device = NULL;
1384 char *sizestr;
1385 char *devstr = NULL;
1386 int ret = 0;
f46b5a66
CH		 1387 int mod = 0;
1388
e441d54d
CM		 1389 if (!capable(CAP_SYS_ADMIN))
1390 return -EPERM;
1391
198605a8
MX		 1392 ret = mnt_want_write_file(file);
1393 if (ret)
1394 return ret;
1395
5ac00add
SB		 1396 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1397 1)) {
1398 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
97547676 1399 mnt_drop_write_file(file);
2c0c9da0 1400 return -EINVAL;
c9e9f97b
ID		 1401 }
1402
5ac00add 1403 mutex_lock(&root->fs_info->volume_mutex);
dae7b665 1404 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 1405 if (IS_ERR(vol_args)) {
1406 ret = PTR_ERR(vol_args);
1407 goto out;
1408 }
5516e595
MF		 1409
1410 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 1411
f46b5a66
CH		 1412 sizestr = vol_args->name;
1413 devstr = strchr(sizestr, ':');
1414 if (devstr) {
1415 char *end;
1416 sizestr = devstr + 1;
1417 *devstr = '\0';
1418 devstr = vol_args->name;
1419 devid = simple_strtoull(devstr, &end, 10);
dfd79829
MX		 1420 if (!devid) {
1421 ret = -EINVAL;
1422 goto out_free;
1423 }
5bb14682 1424 printk(KERN_INFO "btrfs: resizing devid %llu\n",
21380931 1425 (unsigned long long)devid);
f46b5a66 1426 }
dba60f3f 1427
aa1b8cd4 1428 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
f46b5a66 1429 if (!device) {
5bb14682 1430 printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
21380931 1431 (unsigned long long)devid);
dfd79829 1432 ret = -ENODEV;
c9e9f97b 1433 goto out_free;
f46b5a66 1434 }
dba60f3f
MX		 1435
1436 if (!device->writeable) {
4e42ae1b 1437 printk(KERN_INFO "btrfs: resizer unable to apply on "
dba60f3f 1438 "readonly device %llu\n",
a8c4a33b 1439 (unsigned long long)devid);
dfd79829 1440 ret = -EPERM;
4e42ae1b
LB		 1441 goto out_free;
1442 }
1443
f46b5a66
CH		 1444 if (!strcmp(sizestr, "max"))
1445 new_size = device->bdev->bd_inode->i_size;
1446 else {
1447 if (sizestr[0] == '-') {
1448 mod = -1;
1449 sizestr++;
1450 } else if (sizestr[0] == '+') {
1451 mod = 1;
1452 sizestr++;
1453 }
91748467 1454 new_size = memparse(sizestr, NULL);
f46b5a66
CH		 1455 if (new_size == 0) {
1456 ret = -EINVAL;
c9e9f97b 1457 goto out_free;
f46b5a66
CH		 1458 }
1459 }
1460
63a212ab 1461 if (device->is_tgtdev_for_dev_replace) {
dfd79829 1462 ret = -EPERM;
63a212ab
SB		 1463 goto out_free;
1464 }
1465
f46b5a66
CH		 1466 old_size = device->total_bytes;
1467
1468 if (mod < 0) {
1469 if (new_size > old_size) {
1470 ret = -EINVAL;
c9e9f97b 1471 goto out_free;
f46b5a66
CH		 1472 }
1473 new_size = old_size - new_size;
1474 } else if (mod > 0) {
1475 new_size = old_size + new_size;
1476 }
1477
1478 if (new_size < 256 * 1024 * 1024) {
1479 ret = -EINVAL;
c9e9f97b 1480 goto out_free;
f46b5a66
CH		 1481 }
1482 if (new_size > device->bdev->bd_inode->i_size) {
1483 ret = -EFBIG;
c9e9f97b 1484 goto out_free;
f46b5a66
CH		 1485 }
1486
1487 do_div(new_size, root->sectorsize);
1488 new_size *= root->sectorsize;
1489
606686ee
JB		 1490 printk_in_rcu(KERN_INFO "btrfs: new size for %s is %llu\n",
1491 rcu_str_deref(device->name),
1492 (unsigned long long)new_size);
f46b5a66
CH		 1493
1494 if (new_size > old_size) {
a22285a6 1495 trans = btrfs_start_transaction(root, 0);
98d5dc13
TI		 1496 if (IS_ERR(trans)) {
1497 ret = PTR_ERR(trans);
c9e9f97b 1498 goto out_free;
98d5dc13 1499 }
f46b5a66
CH		 1500 ret = btrfs_grow_device(trans, device, new_size);
1501 btrfs_commit_transaction(trans, root);
ece7d20e 1502 } else if (new_size < old_size) {
f46b5a66 1503 ret = btrfs_shrink_device(device, new_size);
0253f40e 1504 } /* equal, nothing need to do */
f46b5a66 1505
c9e9f97b 1506out_free:
f46b5a66 1507 kfree(vol_args);
c9e9f97b
ID		 1508out:
1509 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 1510 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
18f39c41 1511 mnt_drop_write_file(file);
f46b5a66
CH		 1512 return ret;
1513}
1514
72fd032e 1515static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
6f72c7e2
AJ		 1516 char *name, unsigned long fd, int subvol,
1517 u64 *transid, bool readonly,
8696c533 1518 struct btrfs_qgroup_inherit *inherit)
f46b5a66 1519{
f46b5a66 1520 int namelen;
3de4586c 1521 int ret = 0;
f46b5a66 1522
a874a63e
LB		 1523 ret = mnt_want_write_file(file);
1524 if (ret)
1525 goto out;
1526
72fd032e
SW		 1527 namelen = strlen(name);
1528 if (strchr(name, '/')) {
f46b5a66 1529 ret = -EINVAL;
a874a63e 1530 goto out_drop_write;
f46b5a66
CH		 1531 }
1532
16780cab
CM		 1533 if (name[0] == '.' &&
1534 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1535 ret = -EEXIST;
a874a63e 1536 goto out_drop_write;
16780cab
CM		 1537 }
1538
3de4586c 1539 if (subvol) {
72fd032e 1540 ret = btrfs_mksubvol(&file->f_path, name, namelen,
6f72c7e2 1541 NULL, transid, readonly, inherit);
cb8e7090 1542 } else {
2903ff01 1543 struct fd src = fdget(fd);
3de4586c 1544 struct inode *src_inode;
2903ff01 1545 if (!src.file) {
3de4586c 1546 ret = -EINVAL;
a874a63e 1547 goto out_drop_write;
3de4586c
CM		 1548 }
1549
496ad9aa
AV		 1550 src_inode = file_inode(src.file);
1551 if (src_inode->i_sb != file_inode(file)->i_sb) {
d397712b
CM		 1552 printk(KERN_INFO "btrfs: Snapshot src from "
1553 "another FS\n");
3de4586c 1554 ret = -EINVAL;
ecd18815
AV		 1555 } else {
1556 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1557 BTRFS_I(src_inode)->root,
1558 transid, readonly, inherit);
3de4586c 1559 }
2903ff01 1560 fdput(src);
cb8e7090 1561 }
a874a63e
LB		 1562out_drop_write:
1563 mnt_drop_write_file(file);
f46b5a66 1564out:
72fd032e
SW		 1565 return ret;
1566}
1567
1568static noinline int btrfs_ioctl_snap_create(struct file *file,
fa0d2b9b 1569 void __user *arg, int subvol)
72fd032e 1570{
fa0d2b9b 1571 struct btrfs_ioctl_vol_args *vol_args;
72fd032e
SW		 1572 int ret;
1573
fa0d2b9b
LZ		 1574 vol_args = memdup_user(arg, sizeof(*vol_args));
1575 if (IS_ERR(vol_args))
1576 return PTR_ERR(vol_args);
1577 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
72fd032e 1578
fa0d2b9b 1579 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
b83cc969 1580 vol_args->fd, subvol,
6f72c7e2 1581 NULL, false, NULL);
fdfb1e4f 1582
fa0d2b9b
LZ		 1583 kfree(vol_args);
1584 return ret;
1585}
fdfb1e4f 1586
fa0d2b9b
LZ		 1587static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1588 void __user *arg, int subvol)
1589{
1590 struct btrfs_ioctl_vol_args_v2 *vol_args;
1591 int ret;
1592 u64 transid = 0;
1593 u64 *ptr = NULL;
b83cc969 1594 bool readonly = false;
6f72c7e2 1595 struct btrfs_qgroup_inherit *inherit = NULL;
75eaa0e2 1596
fa0d2b9b
LZ		 1597 vol_args = memdup_user(arg, sizeof(*vol_args));
1598 if (IS_ERR(vol_args))
1599 return PTR_ERR(vol_args);
1600 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
75eaa0e2 1601
b83cc969 1602 if (vol_args->flags &
6f72c7e2
AJ		 1603 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1604 BTRFS_SUBVOL_QGROUP_INHERIT)) {
b83cc969 1605 ret = -EOPNOTSUPP;
fa0d2b9b 1606 goto out;
72fd032e 1607 }
fa0d2b9b
LZ		 1608
1609 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1610 ptr = &transid;
b83cc969
LZ		 1611 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1612 readonly = true;
6f72c7e2
AJ		 1613 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1614 if (vol_args->size > PAGE_CACHE_SIZE) {
1615 ret = -EINVAL;
1616 goto out;
1617 }
1618 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1619 if (IS_ERR(inherit)) {
1620 ret = PTR_ERR(inherit);
1621 goto out;
1622 }
1623 }
fa0d2b9b
LZ		 1624
1625 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
6f72c7e2 1626 vol_args->fd, subvol, ptr,
8696c533 1627 readonly, inherit);
fa0d2b9b
LZ		 1628
1629 if (ret == 0 && ptr &&
1630 copy_to_user(arg +
1631 offsetof(struct btrfs_ioctl_vol_args_v2,
1632 transid), ptr, sizeof(*ptr)))
1633 ret = -EFAULT;
fdfb1e4f 1634out:
f46b5a66 1635 kfree(vol_args);
6f72c7e2 1636 kfree(inherit);
f46b5a66
CH		 1637 return ret;
1638}
1639
0caa102d
LZ		 1640static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1641 void __user *arg)
1642{
496ad9aa 1643 struct inode *inode = file_inode(file);
0caa102d
LZ		 1644 struct btrfs_root *root = BTRFS_I(inode)->root;
1645 int ret = 0;
1646 u64 flags = 0;
1647
33345d01 1648 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
0caa102d
LZ		 1649 return -EINVAL;
1650
1651 down_read(&root->fs_info->subvol_sem);
1652 if (btrfs_root_readonly(root))
1653 flags |= BTRFS_SUBVOL_RDONLY;
1654 up_read(&root->fs_info->subvol_sem);
1655
1656 if (copy_to_user(arg, &flags, sizeof(flags)))
1657 ret = -EFAULT;
1658
1659 return ret;
1660}
1661
1662static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1663 void __user *arg)
1664{
496ad9aa 1665 struct inode *inode = file_inode(file);
0caa102d
LZ		 1666 struct btrfs_root *root = BTRFS_I(inode)->root;
1667 struct btrfs_trans_handle *trans;
1668 u64 root_flags;
1669 u64 flags;
1670 int ret = 0;
1671
b9ca0664
LB		 1672 ret = mnt_want_write_file(file);
1673 if (ret)
1674 goto out;
0caa102d 1675
b9ca0664
LB		 1676 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1677 ret = -EINVAL;
1678 goto out_drop_write;
1679 }
0caa102d 1680
b9ca0664
LB		 1681 if (copy_from_user(&flags, arg, sizeof(flags))) {
1682 ret = -EFAULT;
1683 goto out_drop_write;
1684 }
0caa102d 1685
b9ca0664
LB		 1686 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1687 ret = -EINVAL;
1688 goto out_drop_write;
1689 }
0caa102d 1690
b9ca0664
LB		 1691 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1692 ret = -EOPNOTSUPP;
1693 goto out_drop_write;
1694 }
0caa102d 1695
b9ca0664
LB		 1696 if (!inode_owner_or_capable(inode)) {
1697 ret = -EACCES;
1698 goto out_drop_write;
1699 }
b4dc2b8c 1700
0caa102d
LZ		 1701 down_write(&root->fs_info->subvol_sem);
1702
1703 /* nothing to do */
1704 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
b9ca0664 1705 goto out_drop_sem;
0caa102d
LZ		 1706
1707 root_flags = btrfs_root_flags(&root->root_item);
1708 if (flags & BTRFS_SUBVOL_RDONLY)
1709 btrfs_set_root_flags(&root->root_item,
1710 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1711 else
1712 btrfs_set_root_flags(&root->root_item,
1713 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1714
1715 trans = btrfs_start_transaction(root, 1);
1716 if (IS_ERR(trans)) {
1717 ret = PTR_ERR(trans);
1718 goto out_reset;
1719 }
1720
b4dc2b8c 1721 ret = btrfs_update_root(trans, root->fs_info->tree_root,
0caa102d
LZ		 1722 &root->root_key, &root->root_item);
1723
1724 btrfs_commit_transaction(trans, root);
1725out_reset:
1726 if (ret)
1727 btrfs_set_root_flags(&root->root_item, root_flags);
b9ca0664 1728out_drop_sem:
0caa102d 1729 up_write(&root->fs_info->subvol_sem);
b9ca0664
LB		 1730out_drop_write:
1731 mnt_drop_write_file(file);
1732out:
0caa102d
LZ		 1733 return ret;
1734}
1735
76dda93c
YZ		 1736/*
1737 * helper to check if the subvolume references other subvolumes
1738 */
1739static noinline int may_destroy_subvol(struct btrfs_root *root)
1740{
1741 struct btrfs_path *path;
175a2b87 1742 struct btrfs_dir_item *di;
76dda93c 1743 struct btrfs_key key;
175a2b87 1744 u64 dir_id;
76dda93c
YZ		 1745 int ret;
1746
1747 path = btrfs_alloc_path();
1748 if (!path)
1749 return -ENOMEM;
1750
175a2b87
JB		 1751 /* Make sure this root isn't set as the default subvol */
1752 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1753 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1754 dir_id, "default", 7, 0);
1755 if (di && !IS_ERR(di)) {
1756 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1757 if (key.objectid == root->root_key.objectid) {
1758 ret = -ENOTEMPTY;
1759 goto out;
1760 }
1761 btrfs_release_path(path);
1762 }
1763
76dda93c
YZ		 1764 key.objectid = root->root_key.objectid;
1765 key.type = BTRFS_ROOT_REF_KEY;
1766 key.offset = (u64)-1;
1767
1768 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1769 &key, path, 0, 0);
1770 if (ret < 0)
1771 goto out;
1772 BUG_ON(ret == 0);
1773
1774 ret = 0;
1775 if (path->slots[0] > 0) {
1776 path->slots[0]--;
1777 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1778 if (key.objectid == root->root_key.objectid &&
1779 key.type == BTRFS_ROOT_REF_KEY)
1780 ret = -ENOTEMPTY;
1781 }
1782out:
1783 btrfs_free_path(path);
1784 return ret;
1785}
1786
ac8e9819
CM		 1787static noinline int key_in_sk(struct btrfs_key *key,
1788 struct btrfs_ioctl_search_key *sk)
1789{
abc6e134
CM		 1790 struct btrfs_key test;
1791 int ret;
1792
1793 test.objectid = sk->min_objectid;
1794 test.type = sk->min_type;
1795 test.offset = sk->min_offset;
1796
1797 ret = btrfs_comp_cpu_keys(key, &test);
1798 if (ret < 0)
ac8e9819 1799 return 0;
abc6e134
CM		 1800
1801 test.objectid = sk->max_objectid;
1802 test.type = sk->max_type;
1803 test.offset = sk->max_offset;
1804
1805 ret = btrfs_comp_cpu_keys(key, &test);
1806 if (ret > 0)
ac8e9819
CM		 1807 return 0;
1808 return 1;
1809}
1810
1811static noinline int copy_to_sk(struct btrfs_root *root,
1812 struct btrfs_path *path,
1813 struct btrfs_key *key,
1814 struct btrfs_ioctl_search_key *sk,
1815 char *buf,
1816 unsigned long *sk_offset,
1817 int *num_found)
1818{
1819 u64 found_transid;
1820 struct extent_buffer *leaf;
1821 struct btrfs_ioctl_search_header sh;
1822 unsigned long item_off;
1823 unsigned long item_len;
1824 int nritems;
1825 int i;
1826 int slot;
ac8e9819
CM		 1827 int ret = 0;
1828
1829 leaf = path->nodes[0];
1830 slot = path->slots[0];
1831 nritems = btrfs_header_nritems(leaf);
1832
1833 if (btrfs_header_generation(leaf) > sk->max_transid) {
1834 i = nritems;
1835 goto advance_key;
1836 }
1837 found_transid = btrfs_header_generation(leaf);
1838
1839 for (i = slot; i < nritems; i++) {
1840 item_off = btrfs_item_ptr_offset(leaf, i);
1841 item_len = btrfs_item_size_nr(leaf, i);
1842
03b71c6c
GP		 1843 btrfs_item_key_to_cpu(leaf, key, i);
1844 if (!key_in_sk(key, sk))
1845 continue;
1846
1847 if (sizeof(sh) + item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
ac8e9819
CM		 1848 item_len = 0;
1849
1850 if (sizeof(sh) + item_len + *sk_offset >
1851 BTRFS_SEARCH_ARGS_BUFSIZE) {
1852 ret = 1;
1853 goto overflow;
1854 }
1855
ac8e9819
CM		 1856 sh.objectid = key->objectid;
1857 sh.offset = key->offset;
1858 sh.type = key->type;
1859 sh.len = item_len;
1860 sh.transid = found_transid;
1861
1862 /* copy search result header */
1863 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1864 *sk_offset += sizeof(sh);
1865
1866 if (item_len) {
1867 char *p = buf + *sk_offset;
1868 /* copy the item */
1869 read_extent_buffer(leaf, p,
1870 item_off, item_len);
1871 *sk_offset += item_len;
ac8e9819 1872 }
e2156867 1873 (*num_found)++;
ac8e9819
CM		 1874
1875 if (*num_found >= sk->nr_items)
1876 break;
1877 }
1878advance_key:
abc6e134
CM		 1879 ret = 0;
1880 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
ac8e9819 1881 key->offset++;
abc6e134
CM		 1882 else if (key->type < (u8)-1 && key->type < sk->max_type) {
1883 key->offset = 0;
ac8e9819 1884 key->type++;
abc6e134
CM		 1885 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1886 key->offset = 0;
1887 key->type = 0;
ac8e9819 1888 key->objectid++;
abc6e134
CM		 1889 } else
1890 ret = 1;
ac8e9819 1891overflow:
ac8e9819
CM		 1892 return ret;
1893}
1894
1895static noinline int search_ioctl(struct inode *inode,
1896 struct btrfs_ioctl_search_args *args)
1897{
1898 struct btrfs_root *root;
1899 struct btrfs_key key;
1900 struct btrfs_key max_key;
1901 struct btrfs_path *path;
1902 struct btrfs_ioctl_search_key *sk = &args->key;
1903 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1904 int ret;
1905 int num_found = 0;
1906 unsigned long sk_offset = 0;
1907
1908 path = btrfs_alloc_path();
1909 if (!path)
1910 return -ENOMEM;
1911
1912 if (sk->tree_id == 0) {
1913 /* search the root of the inode that was passed */
1914 root = BTRFS_I(inode)->root;
1915 } else {
1916 key.objectid = sk->tree_id;
1917 key.type = BTRFS_ROOT_ITEM_KEY;
1918 key.offset = (u64)-1;
1919 root = btrfs_read_fs_root_no_name(info, &key);
1920 if (IS_ERR(root)) {
1921 printk(KERN_ERR "could not find root %llu\n",
1922 sk->tree_id);
1923 btrfs_free_path(path);
1924 return -ENOENT;
1925 }
1926 }
1927
1928 key.objectid = sk->min_objectid;
1929 key.type = sk->min_type;
1930 key.offset = sk->min_offset;
1931
1932 max_key.objectid = sk->max_objectid;
1933 max_key.type = sk->max_type;
1934 max_key.offset = sk->max_offset;
1935
1936 path->keep_locks = 1;
1937
1938 while(1) {
de78b51a 1939 ret = btrfs_search_forward(root, &key, &max_key, path,
ac8e9819
CM		 1940 sk->min_transid);
1941 if (ret != 0) {
1942 if (ret > 0)
1943 ret = 0;
1944 goto err;
1945 }
1946 ret = copy_to_sk(root, path, &key, sk, args->buf,
1947 &sk_offset, &num_found);
b3b4aa74 1948 btrfs_release_path(path);
ac8e9819
CM		 1949 if (ret || num_found >= sk->nr_items)
1950 break;
1951
1952 }
1953 ret = 0;
1954err:
1955 sk->nr_items = num_found;
1956 btrfs_free_path(path);
1957 return ret;
1958}
1959
1960static noinline int btrfs_ioctl_tree_search(struct file *file,
1961 void __user *argp)
1962{
1963 struct btrfs_ioctl_search_args *args;
1964 struct inode *inode;
1965 int ret;
1966
1967 if (!capable(CAP_SYS_ADMIN))
1968 return -EPERM;
1969
2354d08f
JL		 1970 args = memdup_user(argp, sizeof(*args));
1971 if (IS_ERR(args))
1972 return PTR_ERR(args);
ac8e9819 1973
496ad9aa 1974 inode = file_inode(file);
ac8e9819
CM		 1975 ret = search_ioctl(inode, args);
1976 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1977 ret = -EFAULT;
1978 kfree(args);
1979 return ret;
1980}
1981
98d377a0 1982/*
ac8e9819
CM		 1983 * Search INODE_REFs to identify path name of 'dirid' directory
1984 * in a 'tree_id' tree. and sets path name to 'name'.
1985 */
98d377a0
TH		 1986static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1987 u64 tree_id, u64 dirid, char *name)
1988{
1989 struct btrfs_root *root;
1990 struct btrfs_key key;
ac8e9819 1991 char *ptr;
98d377a0
TH		 1992 int ret = -1;
1993 int slot;
1994 int len;
1995 int total_len = 0;
1996 struct btrfs_inode_ref *iref;
1997 struct extent_buffer *l;
1998 struct btrfs_path *path;
1999
2000 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
2001 name[0]='\0';
2002 return 0;
2003 }
2004
2005 path = btrfs_alloc_path();
2006 if (!path)
2007 return -ENOMEM;
2008
ac8e9819 2009 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
98d377a0
TH		 2010
2011 key.objectid = tree_id;
2012 key.type = BTRFS_ROOT_ITEM_KEY;
2013 key.offset = (u64)-1;
2014 root = btrfs_read_fs_root_no_name(info, &key);
2015 if (IS_ERR(root)) {
2016 printk(KERN_ERR "could not find root %llu\n", tree_id);
8ad6fcab
CM		 2017 ret = -ENOENT;
2018 goto out;
98d377a0
TH		 2019 }
2020
2021 key.objectid = dirid;
2022 key.type = BTRFS_INODE_REF_KEY;
8ad6fcab 2023 key.offset = (u64)-1;
98d377a0
TH		 2024
2025 while(1) {
2026 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2027 if (ret < 0)
2028 goto out;
18674c6c
FDBM		 2029 else if (ret > 0) {
2030 ret = btrfs_previous_item(root, path, dirid,
2031 BTRFS_INODE_REF_KEY);
2032 if (ret < 0)
2033 goto out;
2034 else if (ret > 0) {
2035 ret = -ENOENT;
2036 goto out;
2037 }
2038 }
98d377a0
TH		 2039
2040 l = path->nodes[0];
2041 slot = path->slots[0];
2042 btrfs_item_key_to_cpu(l, &key, slot);
2043
98d377a0
TH		 2044 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2045 len = btrfs_inode_ref_name_len(l, iref);
2046 ptr -= len + 1;
2047 total_len += len + 1;
a696cf35
FDBM		 2048 if (ptr < name) {
2049 ret = -ENAMETOOLONG;
98d377a0 2050 goto out;
a696cf35 2051 }
98d377a0
TH		 2052
2053 *(ptr + len) = '/';
2054 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
2055
2056 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2057 break;
2058
b3b4aa74 2059 btrfs_release_path(path);
98d377a0 2060 key.objectid = key.offset;
8ad6fcab 2061 key.offset = (u64)-1;
98d377a0 2062 dirid = key.objectid;
98d377a0 2063 }
77906a50 2064 memmove(name, ptr, total_len);
98d377a0
TH		 2065 name[total_len]='\0';
2066 ret = 0;
2067out:
2068 btrfs_free_path(path);
ac8e9819
CM		 2069 return ret;
2070}
2071
2072static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2073 void __user *argp)
2074{
2075 struct btrfs_ioctl_ino_lookup_args *args;
2076 struct inode *inode;
2077 int ret;
2078
2079 if (!capable(CAP_SYS_ADMIN))
2080 return -EPERM;
2081
2354d08f
JL		 2082 args = memdup_user(argp, sizeof(*args));
2083 if (IS_ERR(args))
2084 return PTR_ERR(args);
c2b96929 2085
496ad9aa 2086 inode = file_inode(file);
ac8e9819 2087
1b53ac4d
CM		 2088 if (args->treeid == 0)
2089 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2090
ac8e9819
CM		 2091 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2092 args->treeid, args->objectid,
2093 args->name);
2094
2095 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2096 ret = -EFAULT;
2097
2098 kfree(args);
98d377a0
TH		 2099 return ret;
2100}
2101
76dda93c
YZ		 2102static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2103 void __user *arg)
2104{
2105 struct dentry *parent = fdentry(file);
2106 struct dentry *dentry;
2107 struct inode *dir = parent->d_inode;
2108 struct inode *inode;
2109 struct btrfs_root *root = BTRFS_I(dir)->root;
2110 struct btrfs_root *dest = NULL;
2111 struct btrfs_ioctl_vol_args *vol_args;
2112 struct btrfs_trans_handle *trans;
c58aaad2
MX		 2113 struct btrfs_block_rsv block_rsv;
2114 u64 qgroup_reserved;
76dda93c
YZ		 2115 int namelen;
2116 int ret;
2117 int err = 0;
2118
76dda93c
YZ		 2119 vol_args = memdup_user(arg, sizeof(*vol_args));
2120 if (IS_ERR(vol_args))
2121 return PTR_ERR(vol_args);
2122
2123 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2124 namelen = strlen(vol_args->name);
2125 if (strchr(vol_args->name, '/') ||
2126 strncmp(vol_args->name, "..", namelen) == 0) {
2127 err = -EINVAL;
2128 goto out;
2129 }
2130
a561be71 2131 err = mnt_want_write_file(file);
76dda93c
YZ		 2132 if (err)
2133 goto out;
2134
5c50c9b8
DS		 2135 err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2136 if (err == -EINTR)
2137 goto out;
76dda93c
YZ		 2138 dentry = lookup_one_len(vol_args->name, parent, namelen);
2139 if (IS_ERR(dentry)) {
2140 err = PTR_ERR(dentry);
2141 goto out_unlock_dir;
2142 }
2143
2144 if (!dentry->d_inode) {
2145 err = -ENOENT;
2146 goto out_dput;
2147 }
2148
2149 inode = dentry->d_inode;
4260f7c7
SW		 2150 dest = BTRFS_I(inode)->root;
2151 if (!capable(CAP_SYS_ADMIN)){
2152 /*
2153 * Regular user. Only allow this with a special mount
2154 * option, when the user has write+exec access to the
2155 * subvol root, and when rmdir(2) would have been
2156 * allowed.
2157 *
2158 * Note that this is _not_ check that the subvol is
2159 * empty or doesn't contain data that we wouldn't
2160 * otherwise be able to delete.
2161 *
2162 * Users who want to delete empty subvols should try
2163 * rmdir(2).
2164 */
2165 err = -EPERM;
2166 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2167 goto out_dput;
2168
2169 /*
2170 * Do not allow deletion if the parent dir is the same
2171 * as the dir to be deleted. That means the ioctl
2172 * must be called on the dentry referencing the root
2173 * of the subvol, not a random directory contained
2174 * within it.
2175 */
2176 err = -EINVAL;
2177 if (root == dest)
2178 goto out_dput;
2179
2180 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2181 if (err)
2182 goto out_dput;
4260f7c7
SW		 2183 }
2184
5c39da5b
MX		 2185 /* check if subvolume may be deleted by a user */
2186 err = btrfs_may_delete(dir, dentry, 1);
2187 if (err)
2188 goto out_dput;
2189
33345d01 2190 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
76dda93c
YZ		 2191 err = -EINVAL;
2192 goto out_dput;
2193 }
2194
76dda93c
YZ		 2195 mutex_lock(&inode->i_mutex);
2196 err = d_invalidate(dentry);
2197 if (err)
2198 goto out_unlock;
2199
2200 down_write(&root->fs_info->subvol_sem);
2201
2202 err = may_destroy_subvol(dest);
2203 if (err)
2204 goto out_up_write;
2205
c58aaad2
MX		 2206 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2207 /*
2208 * One for dir inode, two for dir entries, two for root
2209 * ref/backref.
2210 */
2211 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
ee3441b4 2212 5, &qgroup_reserved, true);
c58aaad2
MX		 2213 if (err)
2214 goto out_up_write;
2215
a22285a6
YZ		 2216 trans = btrfs_start_transaction(root, 0);
2217 if (IS_ERR(trans)) {
2218 err = PTR_ERR(trans);
c58aaad2 2219 goto out_release;
a22285a6 2220 }
c58aaad2
MX		 2221 trans->block_rsv = &block_rsv;
2222 trans->bytes_reserved = block_rsv.size;
a22285a6 2223
76dda93c
YZ		 2224 ret = btrfs_unlink_subvol(trans, root, dir,
2225 dest->root_key.objectid,
2226 dentry->d_name.name,
2227 dentry->d_name.len);
79787eaa
JM		 2228 if (ret) {
2229 err = ret;
2230 btrfs_abort_transaction(trans, root, ret);
2231 goto out_end_trans;
2232 }
76dda93c
YZ		 2233
2234 btrfs_record_root_in_trans(trans, dest);
2235
2236 memset(&dest->root_item.drop_progress, 0,
2237 sizeof(dest->root_item.drop_progress));
2238 dest->root_item.drop_level = 0;
2239 btrfs_set_root_refs(&dest->root_item, 0);
2240
d68fc57b
YZ		 2241 if (!xchg(&dest->orphan_item_inserted, 1)) {
2242 ret = btrfs_insert_orphan_item(trans,
2243 root->fs_info->tree_root,
2244 dest->root_key.objectid);
79787eaa
JM		 2245 if (ret) {
2246 btrfs_abort_transaction(trans, root, ret);
2247 err = ret;
2248 goto out_end_trans;
2249 }
d68fc57b 2250 }
dd5f9615
SB		 2251
2252 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2253 dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
2254 dest->root_key.objectid);
2255 if (ret && ret != -ENOENT) {
2256 btrfs_abort_transaction(trans, root, ret);
2257 err = ret;
2258 goto out_end_trans;
2259 }
2260 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2261 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2262 dest->root_item.received_uuid,
2263 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2264 dest->root_key.objectid);
2265 if (ret && ret != -ENOENT) {
2266 btrfs_abort_transaction(trans, root, ret);
2267 err = ret;
2268 goto out_end_trans;
2269 }
2270 }
2271
79787eaa 2272out_end_trans:
c58aaad2
MX		 2273 trans->block_rsv = NULL;
2274 trans->bytes_reserved = 0;
531cb13f 2275 ret = btrfs_end_transaction(trans, root);
79787eaa
JM		 2276 if (ret && !err)
2277 err = ret;
76dda93c 2278 inode->i_flags |= S_DEAD;
c58aaad2
MX		 2279out_release:
2280 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
76dda93c
YZ		 2281out_up_write:
2282 up_write(&root->fs_info->subvol_sem);
2283out_unlock:
2284 mutex_unlock(&inode->i_mutex);
2285 if (!err) {
efefb143 2286 shrink_dcache_sb(root->fs_info->sb);
76dda93c
YZ		 2287 btrfs_invalidate_inodes(dest);
2288 d_delete(dentry);
fa6ac876
LB		 2289
2290 /* the last ref */
2291 if (dest->cache_inode) {
2292 iput(dest->cache_inode);
2293 dest->cache_inode = NULL;
2294 }
76dda93c
YZ		 2295 }
2296out_dput:
2297 dput(dentry);
2298out_unlock_dir:
2299 mutex_unlock(&dir->i_mutex);
2a79f17e 2300 mnt_drop_write_file(file);
76dda93c
YZ		 2301out:
2302 kfree(vol_args);
2303 return err;
2304}
2305
1e701a32 2306static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
f46b5a66 2307{
496ad9aa 2308 struct inode *inode = file_inode(file);
f46b5a66 2309 struct btrfs_root *root = BTRFS_I(inode)->root;
1e701a32 2310 struct btrfs_ioctl_defrag_range_args *range;
c146afad
YZ		 2311 int ret;
2312
25122d15
ID		 2313 ret = mnt_want_write_file(file);
2314 if (ret)
2315 return ret;
b83cc969 2316
25122d15
ID		 2317 if (btrfs_root_readonly(root)) {
2318 ret = -EROFS;
2319 goto out;
5ac00add 2320 }
f46b5a66
CH		 2321
2322 switch (inode->i_mode & S_IFMT) {
2323 case S_IFDIR:
e441d54d
CM		 2324 if (!capable(CAP_SYS_ADMIN)) {
2325 ret = -EPERM;
2326 goto out;
2327 }
de78b51a 2328 ret = btrfs_defrag_root(root);
8929ecfa
YZ		 2329 if (ret)
2330 goto out;
de78b51a 2331 ret = btrfs_defrag_root(root->fs_info->extent_root);
f46b5a66
CH		 2332 break;
2333 case S_IFREG:
e441d54d
CM		 2334 if (!(file->f_mode & FMODE_WRITE)) {
2335 ret = -EINVAL;
2336 goto out;
2337 }
1e701a32
CM		 2338
2339 range = kzalloc(sizeof(*range), GFP_KERNEL);
2340 if (!range) {
2341 ret = -ENOMEM;
2342 goto out;
2343 }
2344
2345 if (argp) {
2346 if (copy_from_user(range, argp,
2347 sizeof(*range))) {
2348 ret = -EFAULT;
2349 kfree(range);
683be16e 2350 goto out;
1e701a32
CM		 2351 }
2352 /* compression requires us to start the IO */
2353 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2354 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2355 range->extent_thresh = (u32)-1;
2356 }
2357 } else {
2358 /* the rest are all set to zero by kzalloc */
2359 range->len = (u64)-1;
2360 }
496ad9aa 2361 ret = btrfs_defrag_file(file_inode(file), file,
4cb5300b
CM		 2362 range, 0, 0);
2363 if (ret > 0)
2364 ret = 0;
1e701a32 2365 kfree(range);
f46b5a66 2366 break;
8929ecfa
YZ		 2367 default:
2368 ret = -EINVAL;
f46b5a66 2369 }
e441d54d 2370out:
25122d15 2371 mnt_drop_write_file(file);
e441d54d 2372 return ret;
f46b5a66
CH		 2373}
2374
b2950863 2375static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
f46b5a66
CH		 2376{
2377 struct btrfs_ioctl_vol_args *vol_args;
2378 int ret;
2379
e441d54d
CM		 2380 if (!capable(CAP_SYS_ADMIN))
2381 return -EPERM;
2382
5ac00add
SB		 2383 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2384 1)) {
2385 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2c0c9da0 2386 return -EINVAL;
c9e9f97b
ID		 2387 }
2388
5ac00add 2389 mutex_lock(&root->fs_info->volume_mutex);
dae7b665 2390 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 2391 if (IS_ERR(vol_args)) {
2392 ret = PTR_ERR(vol_args);
2393 goto out;
2394 }
f46b5a66 2395
5516e595 2396 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66
CH		 2397 ret = btrfs_init_new_device(root, vol_args->name);
2398
f46b5a66 2399 kfree(vol_args);
c9e9f97b
ID		 2400out:
2401 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 2402 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
f46b5a66
CH		 2403 return ret;
2404}
2405
da24927b 2406static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
f46b5a66 2407{
496ad9aa 2408 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
f46b5a66
CH		 2409 struct btrfs_ioctl_vol_args *vol_args;
2410 int ret;
2411
e441d54d
CM		 2412 if (!capable(CAP_SYS_ADMIN))
2413 return -EPERM;
2414
da24927b
MX		 2415 ret = mnt_want_write_file(file);
2416 if (ret)
2417 return ret;
c146afad 2418
dae7b665 2419 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 2420 if (IS_ERR(vol_args)) {
2421 ret = PTR_ERR(vol_args);
2422 goto out;
2423 }
f46b5a66 2424
5516e595 2425 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 2426
183860f6
AJ		 2427 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2428 1)) {
2429 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2430 goto out;
2431 }
2432
2433 mutex_lock(&root->fs_info->volume_mutex);
2434 ret = btrfs_rm_device(root, vol_args->name);
c9e9f97b 2435 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 2436 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
183860f6
AJ		 2437
2438out:
2439 kfree(vol_args);
4ac20c70 2440 mnt_drop_write_file(file);
f46b5a66
CH		 2441 return ret;
2442}
2443
475f6387
JS		 2444static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2445{
027ed2f0 2446 struct btrfs_ioctl_fs_info_args *fi_args;
475f6387
JS		 2447 struct btrfs_device *device;
2448 struct btrfs_device *next;
2449 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
027ed2f0 2450 int ret = 0;
475f6387
JS		 2451
2452 if (!capable(CAP_SYS_ADMIN))
2453 return -EPERM;
2454
027ed2f0
LZ		 2455 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2456 if (!fi_args)
2457 return -ENOMEM;
2458
2459 fi_args->num_devices = fs_devices->num_devices;
2460 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
475f6387
JS		 2461
2462 mutex_lock(&fs_devices->device_list_mutex);
2463 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
027ed2f0
LZ		 2464 if (device->devid > fi_args->max_id)
2465 fi_args->max_id = device->devid;
475f6387
JS		 2466 }
2467 mutex_unlock(&fs_devices->device_list_mutex);
2468
027ed2f0
LZ		 2469 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2470 ret = -EFAULT;
475f6387 2471
027ed2f0
LZ		 2472 kfree(fi_args);
2473 return ret;
475f6387
JS		 2474}
2475
2476static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2477{
2478 struct btrfs_ioctl_dev_info_args *di_args;
2479 struct btrfs_device *dev;
2480 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2481 int ret = 0;
2482 char *s_uuid = NULL;
475f6387
JS		 2483
2484 if (!capable(CAP_SYS_ADMIN))
2485 return -EPERM;
2486
2487 di_args = memdup_user(arg, sizeof(*di_args));
2488 if (IS_ERR(di_args))
2489 return PTR_ERR(di_args);
2490
dd5f9615 2491 if (!btrfs_is_empty_uuid(di_args->uuid))
475f6387
JS		 2492 s_uuid = di_args->uuid;
2493
2494 mutex_lock(&fs_devices->device_list_mutex);
aa1b8cd4 2495 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
475f6387
JS		 2496
2497 if (!dev) {
2498 ret = -ENODEV;
2499 goto out;
2500 }
2501
2502 di_args->devid = dev->devid;
2503 di_args->bytes_used = dev->bytes_used;
2504 di_args->total_bytes = dev->total_bytes;
2505 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
a27202fb 2506 if (dev->name) {
606686ee
JB		 2507 struct rcu_string *name;
2508
2509 rcu_read_lock();
2510 name = rcu_dereference(dev->name);
2511 strncpy(di_args->path, name->str, sizeof(di_args->path));
2512 rcu_read_unlock();
a27202fb
JM		 2513 di_args->path[sizeof(di_args->path) - 1] = 0;
2514 } else {
99ba55ad 2515 di_args->path[0] = '\0';
a27202fb 2516 }
475f6387
JS		 2517
2518out:
55793c0d 2519 mutex_unlock(&fs_devices->device_list_mutex);
475f6387
JS		 2520 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2521 ret = -EFAULT;
2522
2523 kfree(di_args);
2524 return ret;
2525}
2526
416161db
MF		 2527static struct page *extent_same_get_page(struct inode *inode, u64 off)
2528{
2529 struct page *page;
2530 pgoff_t index;
2531 struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
2532
2533 index = off >> PAGE_CACHE_SHIFT;
2534
2535 page = grab_cache_page(inode->i_mapping, index);
2536 if (!page)
2537 return NULL;
2538
2539 if (!PageUptodate(page)) {
2540 if (extent_read_full_page_nolock(tree, page, btrfs_get_extent,
2541 0))
2542 return NULL;
2543 lock_page(page);
2544 if (!PageUptodate(page)) {
2545 unlock_page(page);
2546 page_cache_release(page);
2547 return NULL;
2548 }
2549 }
2550 unlock_page(page);
2551
2552 return page;
2553}
2554
77fe20dc
MF		 2555static inline void lock_extent_range(struct inode *inode, u64 off, u64 len)
2556{
2557 /* do any pending delalloc/csum calc on src, one way or
2558 another, and lock file content */
2559 while (1) {
2560 struct btrfs_ordered_extent *ordered;
2561 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2562 ordered = btrfs_lookup_first_ordered_extent(inode,
2563 off + len - 1);
2564 if (!ordered &&
2565 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2566 off + len - 1, EXTENT_DELALLOC, 0, NULL))
2567 break;
2568 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2569 if (ordered)
2570 btrfs_put_ordered_extent(ordered);
2571 btrfs_wait_ordered_range(inode, off, len);
2572 }
2573}
2574
416161db
MF		 2575static void btrfs_double_unlock(struct inode *inode1, u64 loff1,
2576 struct inode *inode2, u64 loff2, u64 len)
2577{
2578 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2579 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2580
2581 mutex_unlock(&inode1->i_mutex);
2582 mutex_unlock(&inode2->i_mutex);
2583}
2584
2585static void btrfs_double_lock(struct inode *inode1, u64 loff1,
2586 struct inode *inode2, u64 loff2, u64 len)
2587{
2588 if (inode1 < inode2) {
2589 swap(inode1, inode2);
2590 swap(loff1, loff2);
2591 }
2592
2593 mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT);
2594 lock_extent_range(inode1, loff1, len);
2595 if (inode1 != inode2) {
2596 mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD);
2597 lock_extent_range(inode2, loff2, len);
2598 }
2599}
2600
2601static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
2602 u64 dst_loff, u64 len)
2603{
2604 int ret = 0;
2605 struct page *src_page, *dst_page;
2606 unsigned int cmp_len = PAGE_CACHE_SIZE;
2607 void *addr, *dst_addr;
2608
2609 while (len) {
2610 if (len < PAGE_CACHE_SIZE)
2611 cmp_len = len;
2612
2613 src_page = extent_same_get_page(src, loff);
2614 if (!src_page)
2615 return -EINVAL;
2616 dst_page = extent_same_get_page(dst, dst_loff);
2617 if (!dst_page) {
2618 page_cache_release(src_page);
2619 return -EINVAL;
2620 }
2621 addr = kmap_atomic(src_page);
2622 dst_addr = kmap_atomic(dst_page);
2623
2624 flush_dcache_page(src_page);
2625 flush_dcache_page(dst_page);
2626
2627 if (memcmp(addr, dst_addr, cmp_len))
2628 ret = BTRFS_SAME_DATA_DIFFERS;
2629
2630 kunmap_atomic(addr);
2631 kunmap_atomic(dst_addr);
2632 page_cache_release(src_page);
2633 page_cache_release(dst_page);
2634
2635 if (ret)
2636 break;
2637
2638 loff += cmp_len;
2639 dst_loff += cmp_len;
2640 len -= cmp_len;
2641 }
2642
2643 return ret;
2644}
2645
2646static int extent_same_check_offsets(struct inode *inode, u64 off, u64 len)
2647{
2648 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
2649
2650 if (off + len > inode->i_size || off + len < off)
2651 return -EINVAL;
2652 /* Check that we are block aligned - btrfs_clone() requires this */
2653 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
2654 return -EINVAL;
2655
2656 return 0;
2657}
2658
2659static int btrfs_extent_same(struct inode *src, u64 loff, u64 len,
2660 struct inode *dst, u64 dst_loff)
2661{
2662 int ret;
2663
2664 /*
2665 * btrfs_clone() can't handle extents in the same file
2666 * yet. Once that works, we can drop this check and replace it
2667 * with a check for the same inode, but overlapping extents.
2668 */
2669 if (src == dst)
2670 return -EINVAL;
2671
2672 btrfs_double_lock(src, loff, dst, dst_loff, len);
2673
2674 ret = extent_same_check_offsets(src, loff, len);
2675 if (ret)
2676 goto out_unlock;
2677
2678 ret = extent_same_check_offsets(dst, dst_loff, len);
2679 if (ret)
2680 goto out_unlock;
2681
2682 /* don't make the dst file partly checksummed */
2683 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2684 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
2685 ret = -EINVAL;
2686 goto out_unlock;
2687 }
2688
2689 ret = btrfs_cmp_data(src, loff, dst, dst_loff, len);
2690 if (ret == 0)
2691 ret = btrfs_clone(src, dst, loff, len, len, dst_loff);
2692
2693out_unlock:
2694 btrfs_double_unlock(src, loff, dst, dst_loff, len);
2695
2696 return ret;
2697}
2698
2699#define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024)
2700
2701static long btrfs_ioctl_file_extent_same(struct file *file,
2702 void __user *argp)
2703{
2704 struct btrfs_ioctl_same_args *args = argp;
2705 struct btrfs_ioctl_same_args same;
2706 struct btrfs_ioctl_same_extent_info info;
2707 struct inode *src = file->f_dentry->d_inode;
2708 struct file *dst_file = NULL;
2709 struct inode *dst;
2710 u64 off;
2711 u64 len;
2712 int i;
2713 int ret;
2714 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
2715 bool is_admin = capable(CAP_SYS_ADMIN);
2716
2717 if (!(file->f_mode & FMODE_READ))
2718 return -EINVAL;
2719
2720 ret = mnt_want_write_file(file);
2721 if (ret)
2722 return ret;
2723
2724 if (copy_from_user(&same,
2725 (struct btrfs_ioctl_same_args __user *)argp,
2726 sizeof(same))) {
2727 ret = -EFAULT;
2728 goto out;
2729 }
2730
2731 off = same.logical_offset;
2732 len = same.length;
2733
2734 /*
2735 * Limit the total length we will dedupe for each operation.
2736 * This is intended to bound the total time spent in this
2737 * ioctl to something sane.
2738 */
2739 if (len > BTRFS_MAX_DEDUPE_LEN)
2740 len = BTRFS_MAX_DEDUPE_LEN;
2741
2742 if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) {
2743 /*
2744 * Btrfs does not support blocksize < page_size. As a
2745 * result, btrfs_cmp_data() won't correctly handle
2746 * this situation without an update.
2747 */
2748 ret = -EINVAL;
2749 goto out;
2750 }
2751
2752 ret = -EISDIR;
2753 if (S_ISDIR(src->i_mode))
2754 goto out;
2755
2756 ret = -EACCES;
2757 if (!S_ISREG(src->i_mode))
2758 goto out;
2759
2760 ret = 0;
2761 for (i = 0; i < same.dest_count; i++) {
2762 if (copy_from_user(&info, &args->info[i], sizeof(info))) {
2763 ret = -EFAULT;
2764 goto out;
2765 }
2766
2767 info.bytes_deduped = 0;
2768
2769 dst_file = fget(info.fd);
2770 if (!dst_file) {
2771 info.status = -EBADF;
2772 goto next;
2773 }
2774
2775 if (!(is_admin || (dst_file->f_mode & FMODE_WRITE))) {
2776 info.status = -EINVAL;
2777 goto next;
2778 }
2779
2780 info.status = -EXDEV;
2781 if (file->f_path.mnt != dst_file->f_path.mnt)
2782 goto next;
2783
2784 dst = dst_file->f_dentry->d_inode;
2785 if (src->i_sb != dst->i_sb)
2786 goto next;
2787
2788 if (S_ISDIR(dst->i_mode)) {
2789 info.status = -EISDIR;
2790 goto next;
2791 }
2792
2793 if (!S_ISREG(dst->i_mode)) {
2794 info.status = -EACCES;
2795 goto next;
2796 }
2797
2798 info.status = btrfs_extent_same(src, off, len, dst,
2799 info.logical_offset);
2800 if (info.status == 0)
2801 info.bytes_deduped += len;
2802
2803next:
2804 if (dst_file)
2805 fput(dst_file);
2806
2807 if (__put_user_unaligned(info.status, &args->info[i].status) ||
2808 __put_user_unaligned(info.bytes_deduped,
2809 &args->info[i].bytes_deduped)) {
2810 ret = -EFAULT;
2811 goto out;
2812 }
2813 }
2814
2815out:
2816 mnt_drop_write_file(file);
2817 return ret;
2818}
2819
32b7c687
MF		 2820/**
2821 * btrfs_clone() - clone a range from inode file to another
2822 *
2823 * @src: Inode to clone from
2824 * @inode: Inode to clone to
2825 * @off: Offset within source to start clone from
2826 * @olen: Original length, passed by user, of range to clone
2827 * @olen_aligned: Block-aligned value of olen, extent_same uses
2828 * identical values here
2829 * @destoff: Offset within @inode to start clone
2830 */
2831static int btrfs_clone(struct inode *src, struct inode *inode,
2832 u64 off, u64 olen, u64 olen_aligned, u64 destoff)
f46b5a66 2833{
f46b5a66 2834 struct btrfs_root *root = BTRFS_I(inode)->root;
32b7c687 2835 struct btrfs_path *path = NULL;
f46b5a66 2836 struct extent_buffer *leaf;
32b7c687
MF		 2837 struct btrfs_trans_handle *trans;
2838 char *buf = NULL;
ae01a0ab 2839 struct btrfs_key key;
f46b5a66
CH		 2840 u32 nritems;
2841 int slot;
ae01a0ab 2842 int ret;
32b7c687 2843 u64 len = olen_aligned;
ae01a0ab
YZ		 2844
2845 ret = -ENOMEM;
2846 buf = vmalloc(btrfs_level_size(root, 0));
2847 if (!buf)
32b7c687 2848 return ret;
ae01a0ab
YZ		 2849
2850 path = btrfs_alloc_path();
2851 if (!path) {
2852 vfree(buf);
32b7c687 2853 return ret;
d525e8ab
LZ		 2854 }
2855
32b7c687 2856 path->reada = 2;
c5c9cd4d 2857 /* clone data */
33345d01 2858 key.objectid = btrfs_ino(src);
ae01a0ab
YZ		 2859 key.type = BTRFS_EXTENT_DATA_KEY;
2860 key.offset = 0;
f46b5a66
CH		 2861
2862 while (1) {
2863 /*
2864 * note the key will change type as we walk through the
2865 * tree.
2866 */
362a20c5
DS		 2867 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
2868 0, 0);
f46b5a66
CH		 2869 if (ret < 0)
2870 goto out;
2871
ae01a0ab
YZ		 2872 nritems = btrfs_header_nritems(path->nodes[0]);
2873 if (path->slots[0] >= nritems) {
362a20c5 2874 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
f46b5a66
CH		 2875 if (ret < 0)
2876 goto out;
2877 if (ret > 0)
2878 break;
ae01a0ab 2879 nritems = btrfs_header_nritems(path->nodes[0]);
f46b5a66
CH		 2880 }
2881 leaf = path->nodes[0];
2882 slot = path->slots[0];
f46b5a66 2883
ae01a0ab 2884 btrfs_item_key_to_cpu(leaf, &key, slot);
d20f7043 2885 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
33345d01 2886 key.objectid != btrfs_ino(src))
f46b5a66
CH		 2887 break;
2888
c5c9cd4d
SW		 2889 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2890 struct btrfs_file_extent_item *extent;
2891 int type;
31840ae1
ZY		 2892 u32 size;
2893 struct btrfs_key new_key;
c5c9cd4d
SW		 2894 u64 disko = 0, diskl = 0;
2895 u64 datao = 0, datal = 0;
2896 u8 comp;
b5384d48 2897 u64 endoff;
31840ae1
ZY		 2898
2899 size = btrfs_item_size_nr(leaf, slot);
2900 read_extent_buffer(leaf, buf,
2901 btrfs_item_ptr_offset(leaf, slot),
2902 size);
c5c9cd4d
SW		 2903
2904 extent = btrfs_item_ptr(leaf, slot,
2905 struct btrfs_file_extent_item);
2906 comp = btrfs_file_extent_compression(leaf, extent);
2907 type = btrfs_file_extent_type(leaf, extent);
c8a894d7
CM		 2908 if (type == BTRFS_FILE_EXTENT_REG ||
2909 type == BTRFS_FILE_EXTENT_PREALLOC) {
d397712b
CM		 2910 disko = btrfs_file_extent_disk_bytenr(leaf,
2911 extent);
2912 diskl = btrfs_file_extent_disk_num_bytes(leaf,
2913 extent);
c5c9cd4d 2914 datao = btrfs_file_extent_offset(leaf, extent);
d397712b
CM		 2915 datal = btrfs_file_extent_num_bytes(leaf,
2916 extent);
c5c9cd4d
SW		 2917 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2918 /* take upper bound, may be compressed */
2919 datal = btrfs_file_extent_ram_bytes(leaf,
2920 extent);
2921 }
b3b4aa74 2922 btrfs_release_path(path);
31840ae1 2923
050006a7 2924 if (key.offset + datal <= off ||
aa42ffd9 2925 key.offset >= off + len - 1)
c5c9cd4d
SW		 2926 goto next;
2927
31840ae1 2928 memcpy(&new_key, &key, sizeof(new_key));
33345d01 2929 new_key.objectid = btrfs_ino(inode);
4d728ec7
LZ		 2930 if (off <= key.offset)
2931 new_key.offset = key.offset + destoff - off;
2932 else
2933 new_key.offset = destoff;
31840ae1 2934
b6f3409b
SW		 2935 /*
2936 * 1 - adjusting old extent (we may have to split it)
2937 * 1 - add new extent
2938 * 1 - inode update
2939 */
2940 trans = btrfs_start_transaction(root, 3);
a22285a6
YZ		 2941 if (IS_ERR(trans)) {
2942 ret = PTR_ERR(trans);
2943 goto out;
2944 }
2945
c8a894d7
CM		 2946 if (type == BTRFS_FILE_EXTENT_REG ||
2947 type == BTRFS_FILE_EXTENT_PREALLOC) {
d72c0842
LZ		 2948 /*
2949 * a | --- range to clone ---| b
2950 * | ------------- extent ------------- |
2951 */
2952
2953 /* substract range b */
2954 if (key.offset + datal > off + len)
2955 datal = off + len - key.offset;
2956
2957 /* substract range a */
a22285a6
YZ		 2958 if (off > key.offset) {
2959 datao += off - key.offset;
2960 datal -= off - key.offset;
2961 }
2962
5dc562c5 2963 ret = btrfs_drop_extents(trans, root, inode,
a22285a6
YZ		 2964 new_key.offset,
2965 new_key.offset + datal,
2671485d 2966 1);
79787eaa
JM		 2967 if (ret) {
2968 btrfs_abort_transaction(trans, root,
2969 ret);
2970 btrfs_end_transaction(trans, root);
2971 goto out;
2972 }
a22285a6 2973
c5c9cd4d
SW		 2974 ret = btrfs_insert_empty_item(trans, root, path,
2975 &new_key, size);
79787eaa
JM		 2976 if (ret) {
2977 btrfs_abort_transaction(trans, root,
2978 ret);
2979 btrfs_end_transaction(trans, root);
2980 goto out;
2981 }
c5c9cd4d
SW		 2982
2983 leaf = path->nodes[0];
2984 slot = path->slots[0];
2985 write_extent_buffer(leaf, buf,
31840ae1
ZY		 2986 btrfs_item_ptr_offset(leaf, slot),
2987 size);
ae01a0ab 2988
c5c9cd4d 2989 extent = btrfs_item_ptr(leaf, slot,
f46b5a66 2990 struct btrfs_file_extent_item);
c5c9cd4d 2991
c5c9cd4d
SW		 2992 /* disko == 0 means it's a hole */
2993 if (!disko)
2994 datao = 0;
c5c9cd4d
SW		 2995
2996 btrfs_set_file_extent_offset(leaf, extent,
2997 datao);
2998 btrfs_set_file_extent_num_bytes(leaf, extent,
2999 datal);
3000 if (disko) {
3001 inode_add_bytes(inode, datal);
ae01a0ab 3002 ret = btrfs_inc_extent_ref(trans, root,
5d4f98a2
YZ		 3003 disko, diskl, 0,
3004 root->root_key.objectid,
33345d01 3005 btrfs_ino(inode),
66d7e7f0
AJ		 3006 new_key.offset - datao,
3007 0);
79787eaa
JM		 3008 if (ret) {
3009 btrfs_abort_transaction(trans,
3010 root,
3011 ret);
3012 btrfs_end_transaction(trans,
3013 root);
3014 goto out;
3015
3016 }
f46b5a66 3017 }
c5c9cd4d
SW		 3018 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3019 u64 skip = 0;
3020 u64 trim = 0;
3021 if (off > key.offset) {
3022 skip = off - key.offset;
3023 new_key.offset += skip;
3024 }
d397712b 3025
aa42ffd9
LB		 3026 if (key.offset + datal > off + len)
3027 trim = key.offset + datal - (off + len);
d397712b 3028
c5c9cd4d 3029 if (comp && (skip || trim)) {
c5c9cd4d 3030 ret = -EINVAL;
a22285a6 3031 btrfs_end_transaction(trans, root);
c5c9cd4d
SW		 3032 goto out;
3033 }
3034 size -= skip + trim;
3035 datal -= skip + trim;
a22285a6 3036
5dc562c5 3037 ret = btrfs_drop_extents(trans, root, inode,
a22285a6
YZ		 3038 new_key.offset,
3039 new_key.offset + datal,
2671485d 3040 1);
79787eaa
JM		 3041 if (ret) {
3042 btrfs_abort_transaction(trans, root,
3043 ret);
3044 btrfs_end_transaction(trans, root);
3045 goto out;
3046 }
a22285a6 3047
c5c9cd4d
SW		 3048 ret = btrfs_insert_empty_item(trans, root, path,
3049 &new_key, size);
79787eaa
JM		 3050 if (ret) {
3051 btrfs_abort_transaction(trans, root,
3052 ret);
3053 btrfs_end_transaction(trans, root);
3054 goto out;
3055 }
c5c9cd4d
SW		 3056
3057 if (skip) {
d397712b
CM		 3058 u32 start =
3059 btrfs_file_extent_calc_inline_size(0);
c5c9cd4d
SW		 3060 memmove(buf+start, buf+start+skip,
3061 datal);
3062 }
3063
3064 leaf = path->nodes[0];
3065 slot = path->slots[0];
3066 write_extent_buffer(leaf, buf,
3067 btrfs_item_ptr_offset(leaf, slot),
3068 size);
3069 inode_add_bytes(inode, datal);
f46b5a66 3070 }
c5c9cd4d
SW		 3071
3072 btrfs_mark_buffer_dirty(leaf);
b3b4aa74 3073 btrfs_release_path(path);
c5c9cd4d 3074
0c4d2d95 3075 inode_inc_iversion(inode);
a22285a6 3076 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
b5384d48
SW		 3077
3078 /*
3079 * we round up to the block size at eof when
3080 * determining which extents to clone above,
3081 * but shouldn't round up the file size
3082 */
3083 endoff = new_key.offset + datal;
5f3888ff
LZ		 3084 if (endoff > destoff+olen)
3085 endoff = destoff+olen;
b5384d48
SW		 3086 if (endoff > inode->i_size)
3087 btrfs_i_size_write(inode, endoff);
3088
a22285a6 3089 ret = btrfs_update_inode(trans, root, inode);
79787eaa
JM		 3090 if (ret) {
3091 btrfs_abort_transaction(trans, root, ret);
3092 btrfs_end_transaction(trans, root);
3093 goto out;
3094 }
3095 ret = btrfs_end_transaction(trans, root);
a22285a6 3096 }
d397712b 3097next:
b3b4aa74 3098 btrfs_release_path(path);
f46b5a66 3099 key.offset++;
f46b5a66 3100 }
f46b5a66 3101 ret = 0;
32b7c687 3102
f46b5a66 3103out:
b3b4aa74 3104 btrfs_release_path(path);
32b7c687
MF		 3105 btrfs_free_path(path);
3106 vfree(buf);
3107 return ret;
3108}
3109
3110static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
3111 u64 off, u64 olen, u64 destoff)
3112{
3113 struct inode *inode = fdentry(file)->d_inode;
3114 struct btrfs_root *root = BTRFS_I(inode)->root;
3115 struct fd src_file;
3116 struct inode *src;
3117 int ret;
3118 u64 len = olen;
3119 u64 bs = root->fs_info->sb->s_blocksize;
3120 int same_inode = 0;
3121
3122 /*
3123 * TODO:
3124 * - split compressed inline extents. annoying: we need to
3125 * decompress into destination's address_space (the file offset
3126 * may change, so source mapping won't do), then recompress (or
3127 * otherwise reinsert) a subrange.
3128 * - allow ranges within the same file to be cloned (provided
3129 * they don't overlap)?
3130 */
3131
3132 /* the destination must be opened for writing */
3133 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
3134 return -EINVAL;
3135
3136 if (btrfs_root_readonly(root))
3137 return -EROFS;
3138
3139 ret = mnt_want_write_file(file);
3140 if (ret)
3141 return ret;
3142
3143 src_file = fdget(srcfd);
3144 if (!src_file.file) {
3145 ret = -EBADF;
3146 goto out_drop_write;
3147 }
3148
3149 ret = -EXDEV;
3150 if (src_file.file->f_path.mnt != file->f_path.mnt)
3151 goto out_fput;
3152
3153 src = file_inode(src_file.file);
3154
3155 ret = -EINVAL;
3156 if (src == inode)
3157 same_inode = 1;
3158
3159 /* the src must be open for reading */
3160 if (!(src_file.file->f_mode & FMODE_READ))
3161 goto out_fput;
3162
3163 /* don't make the dst file partly checksummed */
3164 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3165 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3166 goto out_fput;
3167
3168 ret = -EISDIR;
3169 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3170 goto out_fput;
3171
3172 ret = -EXDEV;
3173 if (src->i_sb != inode->i_sb)
3174 goto out_fput;
3175
3176 if (!same_inode) {
3177 if (inode < src) {
3178 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
3179 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
3180 } else {
3181 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
3182 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
3183 }
3184 } else {
3185 mutex_lock(&src->i_mutex);
3186 }
3187
3188 /* determine range to clone */
3189 ret = -EINVAL;
3190 if (off + len > src->i_size || off + len < off)
3191 goto out_unlock;
3192 if (len == 0)
3193 olen = len = src->i_size - off;
3194 /* if we extend to eof, continue to block boundary */
3195 if (off + len == src->i_size)
3196 len = ALIGN(src->i_size, bs) - off;
3197
3198 /* verify the end result is block aligned */
3199 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3200 !IS_ALIGNED(destoff, bs))
3201 goto out_unlock;
3202
3203 /* verify if ranges are overlapped within the same file */
3204 if (same_inode) {
3205 if (destoff + len > off && destoff < off + len)
3206 goto out_unlock;
3207 }
3208
3209 if (destoff > inode->i_size) {
3210 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3211 if (ret)
3212 goto out_unlock;
3213 }
3214
3215 /* truncate page cache pages from target inode range */
3216 truncate_inode_pages_range(&inode->i_data, destoff,
3217 PAGE_CACHE_ALIGN(destoff + len) - 1);
3218
3219 lock_extent_range(src, off, len);
3220
3221 ret = btrfs_clone(src, inode, off, olen, len, destoff);
3222
aa42ffd9 3223 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
f46b5a66
CH		 3224out_unlock:
3225 mutex_unlock(&src->i_mutex);
a96fbc72
LB		 3226 if (!same_inode)
3227 mutex_unlock(&inode->i_mutex);
f46b5a66 3228out_fput:
2903ff01 3229 fdput(src_file);
ab67b7c1 3230out_drop_write:
2a79f17e 3231 mnt_drop_write_file(file);
f46b5a66
CH		 3232 return ret;
3233}
3234
7a865e8a 3235static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
c5c9cd4d
SW		 3236{
3237 struct btrfs_ioctl_clone_range_args args;
3238
7a865e8a 3239 if (copy_from_user(&args, argp, sizeof(args)))
c5c9cd4d
SW		 3240 return -EFAULT;
3241 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
3242 args.src_length, args.dest_offset);
3243}
3244
f46b5a66
CH		 3245/*
3246 * there are many ways the trans_start and trans_end ioctls can lead
3247 * to deadlocks. They should only be used by applications that
3248 * basically own the machine, and have a very in depth understanding
3249 * of all the possible deadlocks and enospc problems.
3250 */
b2950863 3251static long btrfs_ioctl_trans_start(struct file *file)
f46b5a66 3252{
496ad9aa 3253 struct inode *inode = file_inode(file);
f46b5a66
CH		 3254 struct btrfs_root *root = BTRFS_I(inode)->root;
3255 struct btrfs_trans_handle *trans;
1ab86aed 3256 int ret;
f46b5a66 3257
1ab86aed 3258 ret = -EPERM;
df5b5520 3259 if (!capable(CAP_SYS_ADMIN))
1ab86aed 3260 goto out;
df5b5520 3261
1ab86aed
SW		 3262 ret = -EINPROGRESS;
3263 if (file->private_data)
f46b5a66 3264 goto out;
9ca9ee09 3265
b83cc969
LZ		 3266 ret = -EROFS;
3267 if (btrfs_root_readonly(root))
3268 goto out;
3269
a561be71 3270 ret = mnt_want_write_file(file);
c146afad
YZ		 3271 if (ret)
3272 goto out;
3273
a4abeea4 3274 atomic_inc(&root->fs_info->open_ioctl_trans);
9ca9ee09 3275
1ab86aed 3276 ret = -ENOMEM;
7a7eaa40 3277 trans = btrfs_start_ioctl_transaction(root);
abd30bb0 3278 if (IS_ERR(trans))
1ab86aed
SW		 3279 goto out_drop;
3280
3281 file->private_data = trans;
3282 return 0;
3283
3284out_drop:
a4abeea4 3285 atomic_dec(&root->fs_info->open_ioctl_trans);
2a79f17e 3286 mnt_drop_write_file(file);
f46b5a66 3287out:
f46b5a66
CH		 3288 return ret;
3289}
3290
6ef5ed0d
JB		 3291static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3292{
496ad9aa 3293 struct inode *inode = file_inode(file);
6ef5ed0d
JB		 3294 struct btrfs_root *root = BTRFS_I(inode)->root;
3295 struct btrfs_root *new_root;
3296 struct btrfs_dir_item *di;
3297 struct btrfs_trans_handle *trans;
3298 struct btrfs_path *path;
3299 struct btrfs_key location;
3300 struct btrfs_disk_key disk_key;
6ef5ed0d
JB		 3301 u64 objectid = 0;
3302 u64 dir_id;
3c04ce01 3303 int ret;
6ef5ed0d
JB		 3304
3305 if (!capable(CAP_SYS_ADMIN))
3306 return -EPERM;
3307
3c04ce01
MX		 3308 ret = mnt_want_write_file(file);
3309 if (ret)
3310 return ret;
3311
3312 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3313 ret = -EFAULT;
3314 goto out;
3315 }
6ef5ed0d
JB		 3316
3317 if (!objectid)
3318 objectid = root->root_key.objectid;
3319
3320 location.objectid = objectid;
3321 location.type = BTRFS_ROOT_ITEM_KEY;
3322 location.offset = (u64)-1;
3323
3324 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
3c04ce01
MX		 3325 if (IS_ERR(new_root)) {
3326 ret = PTR_ERR(new_root);
3327 goto out;
3328 }
6ef5ed0d 3329
6ef5ed0d 3330 path = btrfs_alloc_path();
3c04ce01
MX		 3331 if (!path) {
3332 ret = -ENOMEM;
3333 goto out;
3334 }
6ef5ed0d
JB		 3335 path->leave_spinning = 1;
3336
3337 trans = btrfs_start_transaction(root, 1);
98d5dc13 3338 if (IS_ERR(trans)) {
6ef5ed0d 3339 btrfs_free_path(path);
3c04ce01
MX		 3340 ret = PTR_ERR(trans);
3341 goto out;
6ef5ed0d
JB		 3342 }
3343
6c41761f 3344 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
6ef5ed0d
JB		 3345 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
3346 dir_id, "default", 7, 1);
cf1e99a4 3347 if (IS_ERR_OR_NULL(di)) {
6ef5ed0d
JB		 3348 btrfs_free_path(path);
3349 btrfs_end_transaction(trans, root);
3350 printk(KERN_ERR "Umm, you don't have the default dir item, "
3351 "this isn't going to work\n");
3c04ce01
MX		 3352 ret = -ENOENT;
3353 goto out;
6ef5ed0d
JB		 3354 }
3355
3356 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3357 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3358 btrfs_mark_buffer_dirty(path->nodes[0]);
3359 btrfs_free_path(path);
3360
2b0ce2c2 3361 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
6ef5ed0d 3362 btrfs_end_transaction(trans, root);
3c04ce01
MX		 3363out:
3364 mnt_drop_write_file(file);
3365 return ret;
6ef5ed0d
JB		 3366}
3367
5af3e8cc
SB		 3368void btrfs_get_block_group_info(struct list_head *groups_list,
3369 struct btrfs_ioctl_space_info *space)
bf5fc093
JB		 3370{
3371 struct btrfs_block_group_cache *block_group;
3372
3373 space->total_bytes = 0;
3374 space->used_bytes = 0;
3375 space->flags = 0;
3376 list_for_each_entry(block_group, groups_list, list) {
3377 space->flags = block_group->flags;
3378 space->total_bytes += block_group->key.offset;
3379 space->used_bytes +=
3380 btrfs_block_group_used(&block_group->item);
3381 }
3382}
3383
48a3b636 3384static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
1406e432
JB		 3385{
3386 struct btrfs_ioctl_space_args space_args;
3387 struct btrfs_ioctl_space_info space;
3388 struct btrfs_ioctl_space_info *dest;
7fde62bf 3389 struct btrfs_ioctl_space_info *dest_orig;
13f2696f 3390 struct btrfs_ioctl_space_info __user *user_dest;
1406e432 3391 struct btrfs_space_info *info;
bf5fc093
JB		 3392 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3393 BTRFS_BLOCK_GROUP_SYSTEM,
3394 BTRFS_BLOCK_GROUP_METADATA,
3395 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3396 int num_types = 4;
7fde62bf 3397 int alloc_size;
1406e432 3398 int ret = 0;
51788b1b 3399 u64 slot_count = 0;
bf5fc093 3400 int i, c;
1406e432
JB		 3401
3402 if (copy_from_user(&space_args,
3403 (struct btrfs_ioctl_space_args __user *)arg,
3404 sizeof(space_args)))
3405 return -EFAULT;
3406
bf5fc093
JB		 3407 for (i = 0; i < num_types; i++) {
3408 struct btrfs_space_info *tmp;
3409
3410 info = NULL;
3411 rcu_read_lock();
3412 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3413 list) {
3414 if (tmp->flags == types[i]) {
3415 info = tmp;
3416 break;
3417 }
3418 }
3419 rcu_read_unlock();
3420
3421 if (!info)
3422 continue;
3423
3424 down_read(&info->groups_sem);
3425 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3426 if (!list_empty(&info->block_groups[c]))
3427 slot_count++;
3428 }
3429 up_read(&info->groups_sem);
3430 }
7fde62bf
CM		 3431
3432 /* space_slots == 0 means they are asking for a count */
3433 if (space_args.space_slots == 0) {
3434 space_args.total_spaces = slot_count;
3435 goto out;
3436 }
bf5fc093 3437
51788b1b 3438 slot_count = min_t(u64, space_args.space_slots, slot_count);
bf5fc093 3439
7fde62bf 3440 alloc_size = sizeof(*dest) * slot_count;
bf5fc093 3441
7fde62bf
CM		 3442 /* we generally have at most 6 or so space infos, one for each raid
3443 * level. So, a whole page should be more than enough for everyone
3444 */
3445 if (alloc_size > PAGE_CACHE_SIZE)
3446 return -ENOMEM;
3447
1406e432 3448 space_args.total_spaces = 0;
7fde62bf
CM		 3449 dest = kmalloc(alloc_size, GFP_NOFS);
3450 if (!dest)
3451 return -ENOMEM;
3452 dest_orig = dest;
1406e432 3453
7fde62bf 3454 /* now we have a buffer to copy into */
bf5fc093
JB		 3455 for (i = 0; i < num_types; i++) {
3456 struct btrfs_space_info *tmp;
3457
51788b1b
DR		 3458 if (!slot_count)
3459 break;
3460
bf5fc093
JB		 3461 info = NULL;
3462 rcu_read_lock();
3463 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3464 list) {
3465 if (tmp->flags == types[i]) {
3466 info = tmp;
3467 break;
3468 }
3469 }
3470 rcu_read_unlock();
7fde62bf 3471
bf5fc093
JB		 3472 if (!info)
3473 continue;
3474 down_read(&info->groups_sem);
3475 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3476 if (!list_empty(&info->block_groups[c])) {
5af3e8cc
SB		 3477 btrfs_get_block_group_info(
3478 &info->block_groups[c], &space);
bf5fc093
JB		 3479 memcpy(dest, &space, sizeof(space));
3480 dest++;
3481 space_args.total_spaces++;
51788b1b 3482 slot_count--;
bf5fc093 3483 }
51788b1b
DR		 3484 if (!slot_count)
3485 break;
bf5fc093
JB		 3486 }
3487 up_read(&info->groups_sem);
1406e432 3488 }
1406e432 3489
2eec6c81 3490 user_dest = (struct btrfs_ioctl_space_info __user *)
7fde62bf
CM		 3491 (arg + sizeof(struct btrfs_ioctl_space_args));
3492
3493 if (copy_to_user(user_dest, dest_orig, alloc_size))
3494 ret = -EFAULT;
3495
3496 kfree(dest_orig);
3497out:
3498 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
1406e432
JB		 3499 ret = -EFAULT;
3500
3501 return ret;
3502}
3503
f46b5a66
CH		 3504/*
3505 * there are many ways the trans_start and trans_end ioctls can lead
3506 * to deadlocks. They should only be used by applications that
3507 * basically own the machine, and have a very in depth understanding
3508 * of all the possible deadlocks and enospc problems.
3509 */
3510long btrfs_ioctl_trans_end(struct file *file)
3511{
496ad9aa 3512 struct inode *inode = file_inode(file);
f46b5a66
CH		 3513 struct btrfs_root *root = BTRFS_I(inode)->root;
3514 struct btrfs_trans_handle *trans;
f46b5a66 3515
f46b5a66 3516 trans = file->private_data;
1ab86aed
SW		 3517 if (!trans)
3518 return -EINVAL;
b214107e 3519 file->private_data = NULL;
9ca9ee09 3520
1ab86aed
SW		 3521 btrfs_end_transaction(trans, root);
3522
a4abeea4 3523 atomic_dec(&root->fs_info->open_ioctl_trans);
9ca9ee09 3524
2a79f17e 3525 mnt_drop_write_file(file);
1ab86aed 3526 return 0;
f46b5a66
CH		 3527}
3528
9a8c28be
MX		 3529static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3530 void __user *argp)
46204592 3531{
46204592
SW		 3532 struct btrfs_trans_handle *trans;
3533 u64 transid;
db5b493a 3534 int ret;
46204592 3535
d4edf39b 3536 trans = btrfs_attach_transaction_barrier(root);
ff7c1d33
MX		 3537 if (IS_ERR(trans)) {
3538 if (PTR_ERR(trans) != -ENOENT)
3539 return PTR_ERR(trans);
3540
3541 /* No running transaction, don't bother */
3542 transid = root->fs_info->last_trans_committed;
3543 goto out;
3544 }
46204592 3545 transid = trans->transid;
db5b493a 3546 ret = btrfs_commit_transaction_async(trans, root, 0);
8b2b2d3c
TI		 3547 if (ret) {
3548 btrfs_end_transaction(trans, root);
db5b493a 3549 return ret;
8b2b2d3c 3550 }
ff7c1d33 3551out:
46204592
SW		 3552 if (argp)
3553 if (copy_to_user(argp, &transid, sizeof(transid)))
3554 return -EFAULT;
3555 return 0;
3556}
3557
9a8c28be
MX		 3558static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
3559 void __user *argp)
46204592 3560{
46204592
SW		 3561 u64 transid;
3562
3563 if (argp) {
3564 if (copy_from_user(&transid, argp, sizeof(transid)))
3565 return -EFAULT;
3566 } else {
3567 transid = 0; /* current trans */
3568 }
3569 return btrfs_wait_for_commit(root, transid);
3570}
3571
b8e95489 3572static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
475f6387 3573{
496ad9aa 3574 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
475f6387 3575 struct btrfs_ioctl_scrub_args *sa;
b8e95489 3576 int ret;
475f6387
JS		 3577
3578 if (!capable(CAP_SYS_ADMIN))
3579 return -EPERM;
3580
3581 sa = memdup_user(arg, sizeof(*sa));
3582 if (IS_ERR(sa))
3583 return PTR_ERR(sa);
3584
b8e95489
MX		 3585 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3586 ret = mnt_want_write_file(file);
3587 if (ret)
3588 goto out;
3589 }
3590
aa1b8cd4 3591 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
63a212ab
SB		 3592 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3593 0);
475f6387
JS		 3594
3595 if (copy_to_user(arg, sa, sizeof(*sa)))
3596 ret = -EFAULT;
3597
b8e95489
MX		 3598 if (!(sa->flags & BTRFS_SCRUB_READONLY))
3599 mnt_drop_write_file(file);
3600out:
475f6387
JS		 3601 kfree(sa);
3602 return ret;
3603}
3604
3605static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3606{
3607 if (!capable(CAP_SYS_ADMIN))
3608 return -EPERM;
3609
aa1b8cd4 3610 return btrfs_scrub_cancel(root->fs_info);
475f6387
JS		 3611}
3612
3613static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3614 void __user *arg)
3615{
3616 struct btrfs_ioctl_scrub_args *sa;
3617 int ret;
3618
3619 if (!capable(CAP_SYS_ADMIN))
3620 return -EPERM;
3621
3622 sa = memdup_user(arg, sizeof(*sa));
3623 if (IS_ERR(sa))
3624 return PTR_ERR(sa);
3625
3626 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3627
3628 if (copy_to_user(arg, sa, sizeof(*sa)))
3629 ret = -EFAULT;
3630
3631 kfree(sa);
3632 return ret;
3633}
3634
c11d2c23 3635static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
b27f7c0c 3636 void __user *arg)
c11d2c23
SB		 3637{
3638 struct btrfs_ioctl_get_dev_stats *sa;
3639 int ret;
3640
c11d2c23
SB		 3641 sa = memdup_user(arg, sizeof(*sa));
3642 if (IS_ERR(sa))
3643 return PTR_ERR(sa);
3644
b27f7c0c
DS		 3645 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3646 kfree(sa);
3647 return -EPERM;
3648 }
3649
3650 ret = btrfs_get_dev_stats(root, sa);
c11d2c23
SB		 3651
3652 if (copy_to_user(arg, sa, sizeof(*sa)))
3653 ret = -EFAULT;
3654
3655 kfree(sa);
3656 return ret;
3657}
3658
3f6bcfbd
SB		 3659static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
3660{
3661 struct btrfs_ioctl_dev_replace_args *p;
3662 int ret;
3663
3664 if (!capable(CAP_SYS_ADMIN))
3665 return -EPERM;
3666
3667 p = memdup_user(arg, sizeof(*p));
3668 if (IS_ERR(p))
3669 return PTR_ERR(p);
3670
3671 switch (p->cmd) {
3672 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
3673 if (atomic_xchg(
3674 &root->fs_info->mutually_exclusive_operation_running,
3675 1)) {
3676 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3677 ret = -EINPROGRESS;
3678 } else {
3679 ret = btrfs_dev_replace_start(root, p);
3680 atomic_set(
3681 &root->fs_info->mutually_exclusive_operation_running,
3682 0);
3683 }
3684 break;
3685 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3686 btrfs_dev_replace_status(root->fs_info, p);
3687 ret = 0;
3688 break;
3689 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3690 ret = btrfs_dev_replace_cancel(root->fs_info, p);
3691 break;
3692 default:
3693 ret = -EINVAL;
3694 break;
3695 }
3696
3697 if (copy_to_user(arg, p, sizeof(*p)))
3698 ret = -EFAULT;
3699
3700 kfree(p);
3701 return ret;
3702}
3703
d7728c96
JS		 3704static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3705{
3706 int ret = 0;
3707 int i;
740c3d22 3708 u64 rel_ptr;
d7728c96 3709 int size;
806468f8 3710 struct btrfs_ioctl_ino_path_args *ipa = NULL;
d7728c96
JS		 3711 struct inode_fs_paths *ipath = NULL;
3712 struct btrfs_path *path;
3713
82b22ac8 3714 if (!capable(CAP_DAC_READ_SEARCH))
d7728c96
JS		 3715 return -EPERM;
3716
3717 path = btrfs_alloc_path();
3718 if (!path) {
3719 ret = -ENOMEM;
3720 goto out;
3721 }
3722
3723 ipa = memdup_user(arg, sizeof(*ipa));
3724 if (IS_ERR(ipa)) {
3725 ret = PTR_ERR(ipa);
3726 ipa = NULL;
3727 goto out;
3728 }
3729
3730 size = min_t(u32, ipa->size, 4096);
3731 ipath = init_ipath(size, root, path);
3732 if (IS_ERR(ipath)) {
3733 ret = PTR_ERR(ipath);
3734 ipath = NULL;
3735 goto out;
3736 }
3737
3738 ret = paths_from_inode(ipa->inum, ipath);
3739 if (ret < 0)
3740 goto out;
3741
3742 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
745c4d8e
JM		 3743 rel_ptr = ipath->fspath->val[i] -
3744 (u64)(unsigned long)ipath->fspath->val;
740c3d22 3745 ipath->fspath->val[i] = rel_ptr;
d7728c96
JS		 3746 }
3747
745c4d8e
JM		 3748 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3749 (void *)(unsigned long)ipath->fspath, size);
d7728c96
JS		 3750 if (ret) {
3751 ret = -EFAULT;
3752 goto out;
3753 }
3754
3755out:
3756 btrfs_free_path(path);
3757 free_ipath(ipath);
3758 kfree(ipa);
3759
3760 return ret;
3761}
3762
3763static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3764{
3765 struct btrfs_data_container *inodes = ctx;
3766 const size_t c = 3 * sizeof(u64);
3767
3768 if (inodes->bytes_left >= c) {
3769 inodes->bytes_left -= c;
3770 inodes->val[inodes->elem_cnt] = inum;
3771 inodes->val[inodes->elem_cnt + 1] = offset;
3772 inodes->val[inodes->elem_cnt + 2] = root;
3773 inodes->elem_cnt += 3;
3774 } else {
3775 inodes->bytes_missing += c - inodes->bytes_left;
3776 inodes->bytes_left = 0;
3777 inodes->elem_missed += 3;
3778 }
3779
3780 return 0;
3781}
3782
3783static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3784 void __user *arg)
3785{
3786 int ret = 0;
3787 int size;
d7728c96
JS		 3788 struct btrfs_ioctl_logical_ino_args *loi;
3789 struct btrfs_data_container *inodes = NULL;
3790 struct btrfs_path *path = NULL;
d7728c96
JS		 3791
3792 if (!capable(CAP_SYS_ADMIN))
3793 return -EPERM;
3794
3795 loi = memdup_user(arg, sizeof(*loi));
3796 if (IS_ERR(loi)) {
3797 ret = PTR_ERR(loi);
3798 loi = NULL;
3799 goto out;
3800 }
3801
3802 path = btrfs_alloc_path();
3803 if (!path) {
3804 ret = -ENOMEM;
3805 goto out;
3806 }
3807
425d17a2 3808 size = min_t(u32, loi->size, 64 * 1024);
d7728c96
JS		 3809 inodes = init_data_container(size);
3810 if (IS_ERR(inodes)) {
3811 ret = PTR_ERR(inodes);
3812 inodes = NULL;
3813 goto out;
3814 }
3815
df031f07
LB		 3816 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
3817 build_ino_list, inodes);
3818 if (ret == -EINVAL)
d7728c96
JS		 3819 ret = -ENOENT;
3820 if (ret < 0)
3821 goto out;
3822
745c4d8e
JM		 3823 ret = copy_to_user((void *)(unsigned long)loi->inodes,
3824 (void *)(unsigned long)inodes, size);
d7728c96
JS		 3825 if (ret)
3826 ret = -EFAULT;
3827
3828out:
3829 btrfs_free_path(path);
425d17a2 3830 vfree(inodes);
d7728c96
JS		 3831 kfree(loi);
3832
3833 return ret;
3834}
3835
19a39dce 3836void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
c9e9f97b
ID		 3837 struct btrfs_ioctl_balance_args *bargs)
3838{
3839 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3840
3841 bargs->flags = bctl->flags;
3842
837d5b6e
ID		 3843 if (atomic_read(&fs_info->balance_running))
3844 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3845 if (atomic_read(&fs_info->balance_pause_req))
3846 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
a7e99c69
ID		 3847 if (atomic_read(&fs_info->balance_cancel_req))
3848 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
837d5b6e 3849
c9e9f97b
ID		 3850 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3851 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3852 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
19a39dce
ID		 3853
3854 if (lock) {
3855 spin_lock(&fs_info->balance_lock);
3856 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3857 spin_unlock(&fs_info->balance_lock);
3858 } else {
3859 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3860 }
c9e9f97b
ID		 3861}
3862
9ba1f6e4 3863static long btrfs_ioctl_balance(struct file *file, void __user *arg)
c9e9f97b 3864{
496ad9aa 3865 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
c9e9f97b
ID		 3866 struct btrfs_fs_info *fs_info = root->fs_info;
3867 struct btrfs_ioctl_balance_args *bargs;
3868 struct btrfs_balance_control *bctl;
ed0fb78f 3869 bool need_unlock; /* for mut. excl. ops lock */
c9e9f97b
ID		 3870 int ret;
3871
3872 if (!capable(CAP_SYS_ADMIN))
3873 return -EPERM;
3874
e54bfa31 3875 ret = mnt_want_write_file(file);
9ba1f6e4
LB		 3876 if (ret)
3877 return ret;
3878
ed0fb78f
ID		 3879again:
3880 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
3881 mutex_lock(&fs_info->volume_mutex);
3882 mutex_lock(&fs_info->balance_mutex);
3883 need_unlock = true;
3884 goto locked;
3885 }
3886
3887 /*
3888 * mut. excl. ops lock is locked. Three possibilites:
3889 * (1) some other op is running
3890 * (2) balance is running
3891 * (3) balance is paused -- special case (think resume)
3892 */
c9e9f97b 3893 mutex_lock(&fs_info->balance_mutex);
ed0fb78f
ID		 3894 if (fs_info->balance_ctl) {
3895 /* this is either (2) or (3) */
3896 if (!atomic_read(&fs_info->balance_running)) {
3897 mutex_unlock(&fs_info->balance_mutex);
3898 if (!mutex_trylock(&fs_info->volume_mutex))
3899 goto again;
3900 mutex_lock(&fs_info->balance_mutex);
3901
3902 if (fs_info->balance_ctl &&
3903 !atomic_read(&fs_info->balance_running)) {
3904 /* this is (3) */
3905 need_unlock = false;
3906 goto locked;
3907 }
3908
3909 mutex_unlock(&fs_info->balance_mutex);
3910 mutex_unlock(&fs_info->volume_mutex);
3911 goto again;
3912 } else {
3913 /* this is (2) */
3914 mutex_unlock(&fs_info->balance_mutex);
3915 ret = -EINPROGRESS;
3916 goto out;
3917 }
3918 } else {
3919 /* this is (1) */
3920 mutex_unlock(&fs_info->balance_mutex);
3921 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3922 ret = -EINVAL;
3923 goto out;
3924 }
3925
3926locked:
3927 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
c9e9f97b
ID		 3928
3929 if (arg) {
3930 bargs = memdup_user(arg, sizeof(*bargs));
3931 if (IS_ERR(bargs)) {
3932 ret = PTR_ERR(bargs);
ed0fb78f 3933 goto out_unlock;
c9e9f97b 3934 }
de322263
ID		 3935
3936 if (bargs->flags & BTRFS_BALANCE_RESUME) {
3937 if (!fs_info->balance_ctl) {
3938 ret = -ENOTCONN;
3939 goto out_bargs;
3940 }
3941
3942 bctl = fs_info->balance_ctl;
3943 spin_lock(&fs_info->balance_lock);
3944 bctl->flags |= BTRFS_BALANCE_RESUME;
3945 spin_unlock(&fs_info->balance_lock);
3946
3947 goto do_balance;
3948 }
c9e9f97b
ID		 3949 } else {
3950 bargs = NULL;
3951 }
3952
ed0fb78f 3953 if (fs_info->balance_ctl) {
837d5b6e
ID		 3954 ret = -EINPROGRESS;
3955 goto out_bargs;
3956 }
3957
c9e9f97b
ID		 3958 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3959 if (!bctl) {
3960 ret = -ENOMEM;
3961 goto out_bargs;
3962 }
3963
3964 bctl->fs_info = fs_info;
3965 if (arg) {
3966 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3967 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3968 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3969
3970 bctl->flags = bargs->flags;
f43ffb60
ID		 3971 } else {
3972 /* balance everything - no filters */
3973 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
c9e9f97b
ID		 3974 }
3975
de322263 3976do_balance:
c9e9f97b 3977 /*
ed0fb78f
ID		 3978 * Ownership of bctl and mutually_exclusive_operation_running
3979 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
3980 * or, if restriper was paused all the way until unmount, in
3981 * free_fs_info. mutually_exclusive_operation_running is
3982 * cleared in __cancel_balance.
c9e9f97b 3983 */
ed0fb78f
ID		 3984 need_unlock = false;
3985
3986 ret = btrfs_balance(bctl, bargs);
3987
c9e9f97b
ID		 3988 if (arg) {
3989 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3990 ret = -EFAULT;
3991 }
3992
3993out_bargs:
3994 kfree(bargs);
ed0fb78f 3995out_unlock:
c9e9f97b
ID		 3996 mutex_unlock(&fs_info->balance_mutex);
3997 mutex_unlock(&fs_info->volume_mutex);
ed0fb78f
ID		 3998 if (need_unlock)
3999 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4000out:
e54bfa31 4001 mnt_drop_write_file(file);
c9e9f97b
ID		 4002 return ret;
4003}
4004
837d5b6e
ID		 4005static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4006{
4007 if (!capable(CAP_SYS_ADMIN))
4008 return -EPERM;
4009
4010 switch (cmd) {
4011 case BTRFS_BALANCE_CTL_PAUSE:
4012 return btrfs_pause_balance(root->fs_info);
a7e99c69
ID		 4013 case BTRFS_BALANCE_CTL_CANCEL:
4014 return btrfs_cancel_balance(root->fs_info);
837d5b6e
ID		 4015 }
4016
4017 return -EINVAL;
4018}
4019
19a39dce
ID		 4020static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4021 void __user *arg)
4022{
4023 struct btrfs_fs_info *fs_info = root->fs_info;
4024 struct btrfs_ioctl_balance_args *bargs;
4025 int ret = 0;
4026
4027 if (!capable(CAP_SYS_ADMIN))
4028 return -EPERM;
4029
4030 mutex_lock(&fs_info->balance_mutex);
4031 if (!fs_info->balance_ctl) {
4032 ret = -ENOTCONN;
4033 goto out;
4034 }
4035
4036 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
4037 if (!bargs) {
4038 ret = -ENOMEM;
4039 goto out;
4040 }
4041
4042 update_ioctl_balance_args(fs_info, 1, bargs);
4043
4044 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4045 ret = -EFAULT;
4046
4047 kfree(bargs);
4048out:
4049 mutex_unlock(&fs_info->balance_mutex);
4050 return ret;
4051}
4052
905b0dda 4053static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
5d13a37b 4054{
496ad9aa 4055 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4056 struct btrfs_ioctl_quota_ctl_args *sa;
4057 struct btrfs_trans_handle *trans = NULL;
4058 int ret;
4059 int err;
4060
4061 if (!capable(CAP_SYS_ADMIN))
4062 return -EPERM;
4063
905b0dda
MX		 4064 ret = mnt_want_write_file(file);
4065 if (ret)
4066 return ret;
5d13a37b
AJ		 4067
4068 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4069 if (IS_ERR(sa)) {
4070 ret = PTR_ERR(sa);
4071 goto drop_write;
4072 }
5d13a37b 4073
7708f029 4074 down_write(&root->fs_info->subvol_sem);
2f232036
JS		 4075 trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4076 if (IS_ERR(trans)) {
4077 ret = PTR_ERR(trans);
4078 goto out;
5d13a37b
AJ		 4079 }
4080
4081 switch (sa->cmd) {
4082 case BTRFS_QUOTA_CTL_ENABLE:
4083 ret = btrfs_quota_enable(trans, root->fs_info);
4084 break;
4085 case BTRFS_QUOTA_CTL_DISABLE:
4086 ret = btrfs_quota_disable(trans, root->fs_info);
4087 break;
5d13a37b
AJ		 4088 default:
4089 ret = -EINVAL;
4090 break;
4091 }
4092
2f232036
JS		 4093 err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4094 if (err && !ret)
4095 ret = err;
5d13a37b
AJ		 4096out:
4097 kfree(sa);
7708f029 4098 up_write(&root->fs_info->subvol_sem);
905b0dda
MX		 4099drop_write:
4100 mnt_drop_write_file(file);
5d13a37b
AJ		 4101 return ret;
4102}
4103
905b0dda 4104static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
5d13a37b 4105{
496ad9aa 4106 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4107 struct btrfs_ioctl_qgroup_assign_args *sa;
4108 struct btrfs_trans_handle *trans;
4109 int ret;
4110 int err;
4111
4112 if (!capable(CAP_SYS_ADMIN))
4113 return -EPERM;
4114
905b0dda
MX		 4115 ret = mnt_want_write_file(file);
4116 if (ret)
4117 return ret;
5d13a37b
AJ		 4118
4119 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4120 if (IS_ERR(sa)) {
4121 ret = PTR_ERR(sa);
4122 goto drop_write;
4123 }
5d13a37b
AJ		 4124
4125 trans = btrfs_join_transaction(root);
4126 if (IS_ERR(trans)) {
4127 ret = PTR_ERR(trans);
4128 goto out;
4129 }
4130
4131 /* FIXME: check if the IDs really exist */
4132 if (sa->assign) {
4133 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4134 sa->src, sa->dst);
4135 } else {
4136 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4137 sa->src, sa->dst);
4138 }
4139
4140 err = btrfs_end_transaction(trans, root);
4141 if (err && !ret)
4142 ret = err;
4143
4144out:
4145 kfree(sa);
905b0dda
MX		 4146drop_write:
4147 mnt_drop_write_file(file);
5d13a37b
AJ		 4148 return ret;
4149}
4150
905b0dda 4151static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
5d13a37b 4152{
496ad9aa 4153 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4154 struct btrfs_ioctl_qgroup_create_args *sa;
4155 struct btrfs_trans_handle *trans;
4156 int ret;
4157 int err;
4158
4159 if (!capable(CAP_SYS_ADMIN))
4160 return -EPERM;
4161
905b0dda
MX		 4162 ret = mnt_want_write_file(file);
4163 if (ret)
4164 return ret;
5d13a37b
AJ		 4165
4166 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4167 if (IS_ERR(sa)) {
4168 ret = PTR_ERR(sa);
4169 goto drop_write;
4170 }
5d13a37b 4171
d86e56cf
MX		 4172 if (!sa->qgroupid) {
4173 ret = -EINVAL;
4174 goto out;
4175 }
4176
5d13a37b
AJ		 4177 trans = btrfs_join_transaction(root);
4178 if (IS_ERR(trans)) {
4179 ret = PTR_ERR(trans);
4180 goto out;
4181 }
4182
4183 /* FIXME: check if the IDs really exist */
4184 if (sa->create) {
4185 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
4186 NULL);
4187 } else {
4188 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4189 }
4190
4191 err = btrfs_end_transaction(trans, root);
4192 if (err && !ret)
4193 ret = err;
4194
4195out:
4196 kfree(sa);
905b0dda
MX		 4197drop_write:
4198 mnt_drop_write_file(file);
5d13a37b
AJ		 4199 return ret;
4200}
4201
905b0dda 4202static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
5d13a37b 4203{
496ad9aa 4204 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4205 struct btrfs_ioctl_qgroup_limit_args *sa;
4206 struct btrfs_trans_handle *trans;
4207 int ret;
4208 int err;
4209 u64 qgroupid;
4210
4211 if (!capable(CAP_SYS_ADMIN))
4212 return -EPERM;
4213
905b0dda
MX		 4214 ret = mnt_want_write_file(file);
4215 if (ret)
4216 return ret;
5d13a37b
AJ		 4217
4218 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4219 if (IS_ERR(sa)) {
4220 ret = PTR_ERR(sa);
4221 goto drop_write;
4222 }
5d13a37b
AJ		 4223
4224 trans = btrfs_join_transaction(root);
4225 if (IS_ERR(trans)) {
4226 ret = PTR_ERR(trans);
4227 goto out;
4228 }
4229
4230 qgroupid = sa->qgroupid;
4231 if (!qgroupid) {
4232 /* take the current subvol as qgroup */
4233 qgroupid = root->root_key.objectid;
4234 }
4235
4236 /* FIXME: check if the IDs really exist */
4237 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
4238
4239 err = btrfs_end_transaction(trans, root);
4240 if (err && !ret)
4241 ret = err;
4242
4243out:
4244 kfree(sa);
905b0dda
MX		 4245drop_write:
4246 mnt_drop_write_file(file);
5d13a37b
AJ		 4247 return ret;
4248}
4249
2f232036
JS		 4250static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4251{
6d0379ec 4252 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2f232036
JS		 4253 struct btrfs_ioctl_quota_rescan_args *qsa;
4254 int ret;
4255
4256 if (!capable(CAP_SYS_ADMIN))
4257 return -EPERM;
4258
4259 ret = mnt_want_write_file(file);
4260 if (ret)
4261 return ret;
4262
4263 qsa = memdup_user(arg, sizeof(*qsa));
4264 if (IS_ERR(qsa)) {
4265 ret = PTR_ERR(qsa);
4266 goto drop_write;
4267 }
4268
4269 if (qsa->flags) {
4270 ret = -EINVAL;
4271 goto out;
4272 }
4273
4274 ret = btrfs_qgroup_rescan(root->fs_info);
4275
4276out:
4277 kfree(qsa);
4278drop_write:
4279 mnt_drop_write_file(file);
4280 return ret;
4281}
4282
4283static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
4284{
6d0379ec 4285 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2f232036
JS		 4286 struct btrfs_ioctl_quota_rescan_args *qsa;
4287 int ret = 0;
4288
4289 if (!capable(CAP_SYS_ADMIN))
4290 return -EPERM;
4291
4292 qsa = kzalloc(sizeof(*qsa), GFP_NOFS);
4293 if (!qsa)
4294 return -ENOMEM;
4295
4296 if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4297 qsa->flags = 1;
4298 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
4299 }
4300
4301 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4302 ret = -EFAULT;
4303
4304 kfree(qsa);
4305 return ret;
4306}
4307
57254b6e
JS		 4308static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
4309{
4310 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
4311
4312 if (!capable(CAP_SYS_ADMIN))
4313 return -EPERM;
4314
4315 return btrfs_qgroup_wait_for_completion(root->fs_info);
4316}
4317
8ea05e3a
AB		 4318static long btrfs_ioctl_set_received_subvol(struct file *file,
4319 void __user *arg)
4320{
4321 struct btrfs_ioctl_received_subvol_args *sa = NULL;
496ad9aa 4322 struct inode *inode = file_inode(file);
8ea05e3a
AB		 4323 struct btrfs_root *root = BTRFS_I(inode)->root;
4324 struct btrfs_root_item *root_item = &root->root_item;
4325 struct btrfs_trans_handle *trans;
4326 struct timespec ct = CURRENT_TIME;
4327 int ret = 0;
dd5f9615 4328 int received_uuid_changed;
8ea05e3a
AB		 4329
4330 ret = mnt_want_write_file(file);
4331 if (ret < 0)
4332 return ret;
4333
4334 down_write(&root->fs_info->subvol_sem);
4335
4336 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
4337 ret = -EINVAL;
4338 goto out;
4339 }
4340
4341 if (btrfs_root_readonly(root)) {
4342 ret = -EROFS;
4343 goto out;
4344 }
4345
4346 if (!inode_owner_or_capable(inode)) {
4347 ret = -EACCES;
4348 goto out;
4349 }
4350
4351 sa = memdup_user(arg, sizeof(*sa));
4352 if (IS_ERR(sa)) {
4353 ret = PTR_ERR(sa);
4354 sa = NULL;
4355 goto out;
4356 }
4357
dd5f9615
SB		 4358 /*
4359 * 1 - root item
4360 * 2 - uuid items (received uuid + subvol uuid)
4361 */
4362 trans = btrfs_start_transaction(root, 3);
8ea05e3a
AB		 4363 if (IS_ERR(trans)) {
4364 ret = PTR_ERR(trans);
4365 trans = NULL;
4366 goto out;
4367 }
4368
4369 sa->rtransid = trans->transid;
4370 sa->rtime.sec = ct.tv_sec;
4371 sa->rtime.nsec = ct.tv_nsec;
4372
dd5f9615
SB		 4373 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4374 BTRFS_UUID_SIZE);
4375 if (received_uuid_changed &&
4376 !btrfs_is_empty_uuid(root_item->received_uuid))
4377 btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
4378 root_item->received_uuid,
4379 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4380 root->root_key.objectid);
8ea05e3a
AB		 4381 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4382 btrfs_set_root_stransid(root_item, sa->stransid);
4383 btrfs_set_root_rtransid(root_item, sa->rtransid);
3cae210f
QW		 4384 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4385 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4386 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4387 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
8ea05e3a
AB		 4388
4389 ret = btrfs_update_root(trans, root->fs_info->tree_root,
4390 &root->root_key, &root->root_item);
4391 if (ret < 0) {
4392 btrfs_end_transaction(trans, root);
8ea05e3a 4393 goto out;
dd5f9615
SB		 4394 }
4395 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4396 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
4397 sa->uuid,
4398 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4399 root->root_key.objectid);
4400 if (ret < 0 && ret != -EEXIST) {
4401 btrfs_abort_transaction(trans, root, ret);
8ea05e3a 4402 goto out;
dd5f9615
SB		 4403 }
4404 }
4405 ret = btrfs_commit_transaction(trans, root);
4406 if (ret < 0) {
4407 btrfs_abort_transaction(trans, root, ret);
4408 goto out;
8ea05e3a
AB		 4409 }
4410
4411 ret = copy_to_user(arg, sa, sizeof(*sa));
4412 if (ret)
4413 ret = -EFAULT;
4414
4415out:
4416 kfree(sa);
4417 up_write(&root->fs_info->subvol_sem);
4418 mnt_drop_write_file(file);
4419 return ret;
4420}
4421
867ab667 4422static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
4423{
6d0379ec 4424 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
a1b83ac5 4425 size_t len;
867ab667 4426 int ret;
a1b83ac5
AJ		 4427 char label[BTRFS_LABEL_SIZE];
4428
4429 spin_lock(&root->fs_info->super_lock);
4430 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4431 spin_unlock(&root->fs_info->super_lock);
4432
4433 len = strnlen(label, BTRFS_LABEL_SIZE);
867ab667 4434
4435 if (len == BTRFS_LABEL_SIZE) {
4436 pr_warn("btrfs: label is too long, return the first %zu bytes\n",
4437 --len);
4438 }
4439
867ab667 4440 ret = copy_to_user(arg, label, len);
867ab667 4441
4442 return ret ? -EFAULT : 0;
4443}
4444
a8bfd4ab 4445static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4446{
6d0379ec 4447 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
a8bfd4ab 4448 struct btrfs_super_block *super_block = root->fs_info->super_copy;
4449 struct btrfs_trans_handle *trans;
4450 char label[BTRFS_LABEL_SIZE];
4451 int ret;
4452
4453 if (!capable(CAP_SYS_ADMIN))
4454 return -EPERM;
4455
4456 if (copy_from_user(label, arg, sizeof(label)))
4457 return -EFAULT;
4458
4459 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
4460 pr_err("btrfs: unable to set label with more than %d bytes\n",
4461 BTRFS_LABEL_SIZE - 1);
4462 return -EINVAL;
4463 }
4464
4465 ret = mnt_want_write_file(file);
4466 if (ret)
4467 return ret;
4468
a8bfd4ab 4469 trans = btrfs_start_transaction(root, 0);
4470 if (IS_ERR(trans)) {
4471 ret = PTR_ERR(trans);
4472 goto out_unlock;
4473 }
4474
a1b83ac5 4475 spin_lock(&root->fs_info->super_lock);
a8bfd4ab 4476 strcpy(super_block->label, label);
a1b83ac5 4477 spin_unlock(&root->fs_info->super_lock);
a8bfd4ab 4478 ret = btrfs_end_transaction(trans, root);
4479
4480out_unlock:
a8bfd4ab 4481 mnt_drop_write_file(file);
4482 return ret;
4483}
4484
f46b5a66
CH		 4485long btrfs_ioctl(struct file *file, unsigned int
4486 cmd, unsigned long arg)
4487{
496ad9aa 4488 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4bcabaa3 4489 void __user *argp = (void __user *)arg;
f46b5a66
CH		 4490
4491 switch (cmd) {
6cbff00f
CH		 4492 case FS_IOC_GETFLAGS:
4493 return btrfs_ioctl_getflags(file, argp);
4494 case FS_IOC_SETFLAGS:
4495 return btrfs_ioctl_setflags(file, argp);
4496 case FS_IOC_GETVERSION:
4497 return btrfs_ioctl_getversion(file, argp);
f7039b1d
LD		 4498 case FITRIM:
4499 return btrfs_ioctl_fitrim(file, argp);
f46b5a66 4500 case BTRFS_IOC_SNAP_CREATE:
fa0d2b9b 4501 return btrfs_ioctl_snap_create(file, argp, 0);
fdfb1e4f 4502 case BTRFS_IOC_SNAP_CREATE_V2:
fa0d2b9b 4503 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3de4586c 4504 case BTRFS_IOC_SUBVOL_CREATE:
fa0d2b9b 4505 return btrfs_ioctl_snap_create(file, argp, 1);
6f72c7e2
AJ		 4506 case BTRFS_IOC_SUBVOL_CREATE_V2:
4507 return btrfs_ioctl_snap_create_v2(file, argp, 1);
76dda93c
YZ		 4508 case BTRFS_IOC_SNAP_DESTROY:
4509 return btrfs_ioctl_snap_destroy(file, argp);
0caa102d
LZ		 4510 case BTRFS_IOC_SUBVOL_GETFLAGS:
4511 return btrfs_ioctl_subvol_getflags(file, argp);
4512 case BTRFS_IOC_SUBVOL_SETFLAGS:
4513 return btrfs_ioctl_subvol_setflags(file, argp);
6ef5ed0d
JB		 4514 case BTRFS_IOC_DEFAULT_SUBVOL:
4515 return btrfs_ioctl_default_subvol(file, argp);
f46b5a66 4516 case BTRFS_IOC_DEFRAG:
1e701a32
CM		 4517 return btrfs_ioctl_defrag(file, NULL);
4518 case BTRFS_IOC_DEFRAG_RANGE:
4519 return btrfs_ioctl_defrag(file, argp);
f46b5a66 4520 case BTRFS_IOC_RESIZE:
198605a8 4521 return btrfs_ioctl_resize(file, argp);
f46b5a66 4522 case BTRFS_IOC_ADD_DEV:
4bcabaa3 4523 return btrfs_ioctl_add_dev(root, argp);
f46b5a66 4524 case BTRFS_IOC_RM_DEV:
da24927b 4525 return btrfs_ioctl_rm_dev(file, argp);
475f6387
JS		 4526 case BTRFS_IOC_FS_INFO:
4527 return btrfs_ioctl_fs_info(root, argp);
4528 case BTRFS_IOC_DEV_INFO:
4529 return btrfs_ioctl_dev_info(root, argp);
f46b5a66 4530 case BTRFS_IOC_BALANCE:
9ba1f6e4 4531 return btrfs_ioctl_balance(file, NULL);
f46b5a66 4532 case BTRFS_IOC_CLONE:
c5c9cd4d
SW		 4533 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
4534 case BTRFS_IOC_CLONE_RANGE:
7a865e8a 4535 return btrfs_ioctl_clone_range(file, argp);
f46b5a66
CH		 4536 case BTRFS_IOC_TRANS_START:
4537 return btrfs_ioctl_trans_start(file);
4538 case BTRFS_IOC_TRANS_END:
4539 return btrfs_ioctl_trans_end(file);
ac8e9819
CM		 4540 case BTRFS_IOC_TREE_SEARCH:
4541 return btrfs_ioctl_tree_search(file, argp);
4542 case BTRFS_IOC_INO_LOOKUP:
4543 return btrfs_ioctl_ino_lookup(file, argp);
d7728c96
JS		 4544 case BTRFS_IOC_INO_PATHS:
4545 return btrfs_ioctl_ino_to_path(root, argp);
4546 case BTRFS_IOC_LOGICAL_INO:
4547 return btrfs_ioctl_logical_to_ino(root, argp);
1406e432
JB		 4548 case BTRFS_IOC_SPACE_INFO:
4549 return btrfs_ioctl_space_info(root, argp);
f46b5a66
CH		 4550 case BTRFS_IOC_SYNC:
4551 btrfs_sync_fs(file->f_dentry->d_sb, 1);
4552 return 0;
46204592 4553 case BTRFS_IOC_START_SYNC:
9a8c28be 4554 return btrfs_ioctl_start_sync(root, argp);
46204592 4555 case BTRFS_IOC_WAIT_SYNC:
9a8c28be 4556 return btrfs_ioctl_wait_sync(root, argp);
475f6387 4557 case BTRFS_IOC_SCRUB:
b8e95489 4558 return btrfs_ioctl_scrub(file, argp);
475f6387
JS		 4559 case BTRFS_IOC_SCRUB_CANCEL:
4560 return btrfs_ioctl_scrub_cancel(root, argp);
4561 case BTRFS_IOC_SCRUB_PROGRESS:
4562 return btrfs_ioctl_scrub_progress(root, argp);
c9e9f97b 4563 case BTRFS_IOC_BALANCE_V2:
9ba1f6e4 4564 return btrfs_ioctl_balance(file, argp);
837d5b6e
ID		 4565 case BTRFS_IOC_BALANCE_CTL:
4566 return btrfs_ioctl_balance_ctl(root, arg);
19a39dce
ID		 4567 case BTRFS_IOC_BALANCE_PROGRESS:
4568 return btrfs_ioctl_balance_progress(root, argp);
8ea05e3a
AB		 4569 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
4570 return btrfs_ioctl_set_received_subvol(file, argp);
31db9f7c
AB		 4571 case BTRFS_IOC_SEND:
4572 return btrfs_ioctl_send(file, argp);
c11d2c23 4573 case BTRFS_IOC_GET_DEV_STATS:
b27f7c0c 4574 return btrfs_ioctl_get_dev_stats(root, argp);
5d13a37b 4575 case BTRFS_IOC_QUOTA_CTL:
905b0dda 4576 return btrfs_ioctl_quota_ctl(file, argp);
5d13a37b 4577 case BTRFS_IOC_QGROUP_ASSIGN:
905b0dda 4578 return btrfs_ioctl_qgroup_assign(file, argp);
5d13a37b 4579 case BTRFS_IOC_QGROUP_CREATE:
905b0dda 4580 return btrfs_ioctl_qgroup_create(file, argp);
5d13a37b 4581 case BTRFS_IOC_QGROUP_LIMIT:
905b0dda 4582 return btrfs_ioctl_qgroup_limit(file, argp);
2f232036
JS		 4583 case BTRFS_IOC_QUOTA_RESCAN:
4584 return btrfs_ioctl_quota_rescan(file, argp);
4585 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
4586 return btrfs_ioctl_quota_rescan_status(file, argp);
57254b6e
JS		 4587 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
4588 return btrfs_ioctl_quota_rescan_wait(file, argp);
3f6bcfbd
SB		 4589 case BTRFS_IOC_DEV_REPLACE:
4590 return btrfs_ioctl_dev_replace(root, argp);
867ab667 4591 case BTRFS_IOC_GET_FSLABEL:
4592 return btrfs_ioctl_get_fslabel(file, argp);
a8bfd4ab 4593 case BTRFS_IOC_SET_FSLABEL:
4594 return btrfs_ioctl_set_fslabel(file, argp);
416161db
MF		 4595 case BTRFS_IOC_FILE_EXTENT_SAME:
4596 return btrfs_ioctl_file_extent_same(file, argp);
f46b5a66
CH		 4597 }
4598
4599 return -ENOTTY;
4600}
A mirror of Linus' kernel repository