cifs: Check server capability before attempting silly rename
[deliverable/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>
4b4e25f2 46#include "compat.h"
f46b5a66
CH
47#include "ctree.h"
48#include "disk-io.h"
49#include "transaction.h"
50#include "btrfs_inode.h"
f46b5a66
CH
51#include "print-tree.h"
52#include "volumes.h"
925baedd 53#include "locking.h"
581bb050 54#include "inode-map.h"
d7728c96 55#include "backref.h"
606686ee 56#include "rcu-string.h"
31db9f7c 57#include "send.h"
3f6bcfbd 58#include "dev-replace.h"
f46b5a66 59
6cbff00f
CH
60/* Mask out flags that are inappropriate for the given type of inode. */
61static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
62{
63 if (S_ISDIR(mode))
64 return flags;
65 else if (S_ISREG(mode))
66 return flags & ~FS_DIRSYNC_FL;
67 else
68 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
69}
70
71/*
72 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
73 */
74static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
75{
76 unsigned int iflags = 0;
77
78 if (flags & BTRFS_INODE_SYNC)
79 iflags |= FS_SYNC_FL;
80 if (flags & BTRFS_INODE_IMMUTABLE)
81 iflags |= FS_IMMUTABLE_FL;
82 if (flags & BTRFS_INODE_APPEND)
83 iflags |= FS_APPEND_FL;
84 if (flags & BTRFS_INODE_NODUMP)
85 iflags |= FS_NODUMP_FL;
86 if (flags & BTRFS_INODE_NOATIME)
87 iflags |= FS_NOATIME_FL;
88 if (flags & BTRFS_INODE_DIRSYNC)
89 iflags |= FS_DIRSYNC_FL;
d0092bdd
LZ
90 if (flags & BTRFS_INODE_NODATACOW)
91 iflags |= FS_NOCOW_FL;
92
93 if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
94 iflags |= FS_COMPR_FL;
95 else if (flags & BTRFS_INODE_NOCOMPRESS)
96 iflags |= FS_NOCOMP_FL;
6cbff00f
CH
97
98 return iflags;
99}
100
101/*
102 * Update inode->i_flags based on the btrfs internal flags.
103 */
104void btrfs_update_iflags(struct inode *inode)
105{
106 struct btrfs_inode *ip = BTRFS_I(inode);
107
108 inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
109
110 if (ip->flags & BTRFS_INODE_SYNC)
111 inode->i_flags |= S_SYNC;
112 if (ip->flags & BTRFS_INODE_IMMUTABLE)
113 inode->i_flags |= S_IMMUTABLE;
114 if (ip->flags & BTRFS_INODE_APPEND)
115 inode->i_flags |= S_APPEND;
116 if (ip->flags & BTRFS_INODE_NOATIME)
117 inode->i_flags |= S_NOATIME;
118 if (ip->flags & BTRFS_INODE_DIRSYNC)
119 inode->i_flags |= S_DIRSYNC;
120}
121
122/*
123 * Inherit flags from the parent inode.
124 *
e27425d6 125 * Currently only the compression flags and the cow flags are inherited.
6cbff00f
CH
126 */
127void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
128{
0b4dcea5
CM
129 unsigned int flags;
130
131 if (!dir)
132 return;
133
134 flags = BTRFS_I(dir)->flags;
6cbff00f 135
e27425d6
JB
136 if (flags & BTRFS_INODE_NOCOMPRESS) {
137 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
138 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
139 } else if (flags & BTRFS_INODE_COMPRESS) {
140 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
141 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
142 }
143
213490b3 144 if (flags & BTRFS_INODE_NODATACOW) {
e27425d6 145 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
213490b3
LB
146 if (S_ISREG(inode->i_mode))
147 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
148 }
6cbff00f 149
6cbff00f
CH
150 btrfs_update_iflags(inode);
151}
152
153static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
154{
496ad9aa 155 struct btrfs_inode *ip = BTRFS_I(file_inode(file));
6cbff00f
CH
156 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
157
158 if (copy_to_user(arg, &flags, sizeof(flags)))
159 return -EFAULT;
160 return 0;
161}
162
75e7cb7f
LB
163static int check_flags(unsigned int flags)
164{
165 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
166 FS_NOATIME_FL | FS_NODUMP_FL | \
167 FS_SYNC_FL | FS_DIRSYNC_FL | \
e1e8fb6a
LZ
168 FS_NOCOMP_FL | FS_COMPR_FL |
169 FS_NOCOW_FL))
75e7cb7f
LB
170 return -EOPNOTSUPP;
171
172 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
173 return -EINVAL;
174
75e7cb7f
LB
175 return 0;
176}
177
6cbff00f
CH
178static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
179{
496ad9aa 180 struct inode *inode = file_inode(file);
6cbff00f
CH
181 struct btrfs_inode *ip = BTRFS_I(inode);
182 struct btrfs_root *root = ip->root;
183 struct btrfs_trans_handle *trans;
184 unsigned int flags, oldflags;
185 int ret;
f062abf0
LZ
186 u64 ip_oldflags;
187 unsigned int i_oldflags;
7e97b8da 188 umode_t mode;
6cbff00f 189
b83cc969
LZ
190 if (btrfs_root_readonly(root))
191 return -EROFS;
192
6cbff00f
CH
193 if (copy_from_user(&flags, arg, sizeof(flags)))
194 return -EFAULT;
195
75e7cb7f
LB
196 ret = check_flags(flags);
197 if (ret)
198 return ret;
f46b5a66 199
2e149670 200 if (!inode_owner_or_capable(inode))
6cbff00f
CH
201 return -EACCES;
202
e7848683
JK
203 ret = mnt_want_write_file(file);
204 if (ret)
205 return ret;
206
6cbff00f
CH
207 mutex_lock(&inode->i_mutex);
208
f062abf0
LZ
209 ip_oldflags = ip->flags;
210 i_oldflags = inode->i_flags;
7e97b8da 211 mode = inode->i_mode;
f062abf0 212
6cbff00f
CH
213 flags = btrfs_mask_flags(inode->i_mode, flags);
214 oldflags = btrfs_flags_to_ioctl(ip->flags);
215 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
216 if (!capable(CAP_LINUX_IMMUTABLE)) {
217 ret = -EPERM;
218 goto out_unlock;
219 }
220 }
221
6cbff00f
CH
222 if (flags & FS_SYNC_FL)
223 ip->flags |= BTRFS_INODE_SYNC;
224 else
225 ip->flags &= ~BTRFS_INODE_SYNC;
226 if (flags & FS_IMMUTABLE_FL)
227 ip->flags |= BTRFS_INODE_IMMUTABLE;
228 else
229 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
230 if (flags & FS_APPEND_FL)
231 ip->flags |= BTRFS_INODE_APPEND;
232 else
233 ip->flags &= ~BTRFS_INODE_APPEND;
234 if (flags & FS_NODUMP_FL)
235 ip->flags |= BTRFS_INODE_NODUMP;
236 else
237 ip->flags &= ~BTRFS_INODE_NODUMP;
238 if (flags & FS_NOATIME_FL)
239 ip->flags |= BTRFS_INODE_NOATIME;
240 else
241 ip->flags &= ~BTRFS_INODE_NOATIME;
242 if (flags & FS_DIRSYNC_FL)
243 ip->flags |= BTRFS_INODE_DIRSYNC;
244 else
245 ip->flags &= ~BTRFS_INODE_DIRSYNC;
7e97b8da
DS
246 if (flags & FS_NOCOW_FL) {
247 if (S_ISREG(mode)) {
248 /*
249 * It's safe to turn csums off here, no extents exist.
250 * Otherwise we want the flag to reflect the real COW
251 * status of the file and will not set it.
252 */
253 if (inode->i_size == 0)
254 ip->flags |= BTRFS_INODE_NODATACOW
255 | BTRFS_INODE_NODATASUM;
256 } else {
257 ip->flags |= BTRFS_INODE_NODATACOW;
258 }
259 } else {
260 /*
261 * Revert back under same assuptions as above
262 */
263 if (S_ISREG(mode)) {
264 if (inode->i_size == 0)
265 ip->flags &= ~(BTRFS_INODE_NODATACOW
266 | BTRFS_INODE_NODATASUM);
267 } else {
268 ip->flags &= ~BTRFS_INODE_NODATACOW;
269 }
270 }
6cbff00f 271
75e7cb7f
LB
272 /*
273 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
274 * flag may be changed automatically if compression code won't make
275 * things smaller.
276 */
277 if (flags & FS_NOCOMP_FL) {
278 ip->flags &= ~BTRFS_INODE_COMPRESS;
279 ip->flags |= BTRFS_INODE_NOCOMPRESS;
280 } else if (flags & FS_COMPR_FL) {
281 ip->flags |= BTRFS_INODE_COMPRESS;
282 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
ebcb904d
LZ
283 } else {
284 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
75e7cb7f 285 }
6cbff00f 286
4da6f1a3 287 trans = btrfs_start_transaction(root, 1);
f062abf0
LZ
288 if (IS_ERR(trans)) {
289 ret = PTR_ERR(trans);
290 goto out_drop;
291 }
6cbff00f 292
306424cc 293 btrfs_update_iflags(inode);
0c4d2d95 294 inode_inc_iversion(inode);
306424cc 295 inode->i_ctime = CURRENT_TIME;
6cbff00f 296 ret = btrfs_update_inode(trans, root, inode);
6cbff00f 297
6cbff00f 298 btrfs_end_transaction(trans, root);
f062abf0
LZ
299 out_drop:
300 if (ret) {
301 ip->flags = ip_oldflags;
302 inode->i_flags = i_oldflags;
303 }
6cbff00f 304
6cbff00f
CH
305 out_unlock:
306 mutex_unlock(&inode->i_mutex);
e7848683 307 mnt_drop_write_file(file);
2d4e6f6a 308 return ret;
6cbff00f
CH
309}
310
311static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
312{
496ad9aa 313 struct inode *inode = file_inode(file);
6cbff00f
CH
314
315 return put_user(inode->i_generation, arg);
316}
f46b5a66 317
f7039b1d
LD
318static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
319{
815745cf 320 struct btrfs_fs_info *fs_info = btrfs_sb(fdentry(file)->d_sb);
f7039b1d
LD
321 struct btrfs_device *device;
322 struct request_queue *q;
323 struct fstrim_range range;
324 u64 minlen = ULLONG_MAX;
325 u64 num_devices = 0;
815745cf 326 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
f7039b1d
LD
327 int ret;
328
329 if (!capable(CAP_SYS_ADMIN))
330 return -EPERM;
331
1f78160c
XG
332 rcu_read_lock();
333 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
334 dev_list) {
f7039b1d
LD
335 if (!device->bdev)
336 continue;
337 q = bdev_get_queue(device->bdev);
338 if (blk_queue_discard(q)) {
339 num_devices++;
340 minlen = min((u64)q->limits.discard_granularity,
341 minlen);
342 }
343 }
1f78160c 344 rcu_read_unlock();
f4c697e6 345
f7039b1d
LD
346 if (!num_devices)
347 return -EOPNOTSUPP;
f7039b1d
LD
348 if (copy_from_user(&range, arg, sizeof(range)))
349 return -EFAULT;
e515c18b
LC
350 if (range.start > total_bytes ||
351 range.len < fs_info->sb->s_blocksize)
f4c697e6 352 return -EINVAL;
f7039b1d 353
f4c697e6 354 range.len = min(range.len, total_bytes - range.start);
f7039b1d 355 range.minlen = max(range.minlen, minlen);
815745cf 356 ret = btrfs_trim_fs(fs_info->tree_root, &range);
f7039b1d
LD
357 if (ret < 0)
358 return ret;
359
360 if (copy_to_user(arg, &range, sizeof(range)))
361 return -EFAULT;
362
363 return 0;
364}
365
d5c12070 366static noinline int create_subvol(struct inode *dir,
cb8e7090 367 struct dentry *dentry,
72fd032e 368 char *name, int namelen,
6f72c7e2 369 u64 *async_transid,
8696c533 370 struct btrfs_qgroup_inherit *inherit)
f46b5a66
CH
371{
372 struct btrfs_trans_handle *trans;
373 struct btrfs_key key;
374 struct btrfs_root_item root_item;
375 struct btrfs_inode_item *inode_item;
376 struct extent_buffer *leaf;
d5c12070 377 struct btrfs_root *root = BTRFS_I(dir)->root;
76dda93c 378 struct btrfs_root *new_root;
d5c12070 379 struct btrfs_block_rsv block_rsv;
8ea05e3a 380 struct timespec cur_time = CURRENT_TIME;
f46b5a66
CH
381 int ret;
382 int err;
383 u64 objectid;
384 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
3de4586c 385 u64 index = 0;
d5c12070 386 u64 qgroup_reserved;
8ea05e3a 387 uuid_le new_uuid;
f46b5a66 388
581bb050 389 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
2fbe8c8a 390 if (ret)
a22285a6 391 return ret;
6a912213 392
d5c12070 393 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
9ed74f2d 394 /*
d5c12070
MX
395 * The same as the snapshot creation, please see the comment
396 * of create_snapshot().
9ed74f2d 397 */
d5c12070
MX
398 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
399 7, &qgroup_reserved);
400 if (ret)
401 return ret;
402
403 trans = btrfs_start_transaction(root, 0);
404 if (IS_ERR(trans)) {
405 ret = PTR_ERR(trans);
406 goto out;
407 }
408 trans->block_rsv = &block_rsv;
409 trans->bytes_reserved = block_rsv.size;
f46b5a66 410
8696c533 411 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
6f72c7e2
AJ
412 if (ret)
413 goto fail;
414
5d4f98a2 415 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
5581a51a 416 0, objectid, NULL, 0, 0, 0);
8e8a1e31
JB
417 if (IS_ERR(leaf)) {
418 ret = PTR_ERR(leaf);
419 goto fail;
420 }
f46b5a66 421
5d4f98a2 422 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
f46b5a66
CH
423 btrfs_set_header_bytenr(leaf, leaf->start);
424 btrfs_set_header_generation(leaf, trans->transid);
5d4f98a2 425 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
f46b5a66
CH
426 btrfs_set_header_owner(leaf, objectid);
427
428 write_extent_buffer(leaf, root->fs_info->fsid,
429 (unsigned long)btrfs_header_fsid(leaf),
430 BTRFS_FSID_SIZE);
5d4f98a2
YZ
431 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
432 (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
433 BTRFS_UUID_SIZE);
f46b5a66
CH
434 btrfs_mark_buffer_dirty(leaf);
435
8ea05e3a
AB
436 memset(&root_item, 0, sizeof(root_item));
437
f46b5a66 438 inode_item = &root_item.inode;
f46b5a66
CH
439 inode_item->generation = cpu_to_le64(1);
440 inode_item->size = cpu_to_le64(3);
441 inode_item->nlink = cpu_to_le32(1);
a76a3cd4 442 inode_item->nbytes = cpu_to_le64(root->leafsize);
f46b5a66
CH
443 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
444
08fe4db1
LZ
445 root_item.flags = 0;
446 root_item.byte_limit = 0;
447 inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
448
f46b5a66 449 btrfs_set_root_bytenr(&root_item, leaf->start);
84234f3a 450 btrfs_set_root_generation(&root_item, trans->transid);
f46b5a66
CH
451 btrfs_set_root_level(&root_item, 0);
452 btrfs_set_root_refs(&root_item, 1);
86b9f2ec 453 btrfs_set_root_used(&root_item, leaf->len);
80ff3856 454 btrfs_set_root_last_snapshot(&root_item, 0);
f46b5a66 455
8ea05e3a
AB
456 btrfs_set_root_generation_v2(&root_item,
457 btrfs_root_generation(&root_item));
458 uuid_le_gen(&new_uuid);
459 memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
460 root_item.otime.sec = cpu_to_le64(cur_time.tv_sec);
dadd1105 461 root_item.otime.nsec = cpu_to_le32(cur_time.tv_nsec);
8ea05e3a
AB
462 root_item.ctime = root_item.otime;
463 btrfs_set_root_ctransid(&root_item, trans->transid);
464 btrfs_set_root_otransid(&root_item, trans->transid);
f46b5a66 465
925baedd 466 btrfs_tree_unlock(leaf);
f46b5a66
CH
467 free_extent_buffer(leaf);
468 leaf = NULL;
469
470 btrfs_set_root_dirid(&root_item, new_dirid);
471
472 key.objectid = objectid;
5d4f98a2 473 key.offset = 0;
f46b5a66
CH
474 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
475 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
476 &root_item);
477 if (ret)
478 goto fail;
479
76dda93c
YZ
480 key.offset = (u64)-1;
481 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
79787eaa
JM
482 if (IS_ERR(new_root)) {
483 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
484 ret = PTR_ERR(new_root);
485 goto fail;
486 }
76dda93c
YZ
487
488 btrfs_record_root_in_trans(trans, new_root);
489
d82a6f1d 490 ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
ce598979
MF
491 if (ret) {
492 /* We potentially lose an unused inode item here */
79787eaa 493 btrfs_abort_transaction(trans, root, ret);
ce598979
MF
494 goto fail;
495 }
496
f46b5a66
CH
497 /*
498 * insert the directory item
499 */
3de4586c 500 ret = btrfs_set_inode_index(dir, &index);
79787eaa
JM
501 if (ret) {
502 btrfs_abort_transaction(trans, root, ret);
503 goto fail;
504 }
3de4586c
CM
505
506 ret = btrfs_insert_dir_item(trans, root,
16cdcec7 507 name, namelen, dir, &key,
3de4586c 508 BTRFS_FT_DIR, index);
79787eaa
JM
509 if (ret) {
510 btrfs_abort_transaction(trans, root, ret);
f46b5a66 511 goto fail;
79787eaa 512 }
0660b5af 513
52c26179
YZ
514 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
515 ret = btrfs_update_inode(trans, root, dir);
516 BUG_ON(ret);
517
0660b5af 518 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
4df27c4d 519 objectid, root->root_key.objectid,
33345d01 520 btrfs_ino(dir), index, name, namelen);
0660b5af 521
76dda93c 522 BUG_ON(ret);
f46b5a66 523
f46b5a66 524fail:
d5c12070
MX
525 trans->block_rsv = NULL;
526 trans->bytes_reserved = 0;
72fd032e
SW
527 if (async_transid) {
528 *async_transid = trans->transid;
529 err = btrfs_commit_transaction_async(trans, root, 1);
530 } else {
531 err = btrfs_commit_transaction(trans, root);
532 }
f46b5a66
CH
533 if (err && !ret)
534 ret = err;
1a65e24b
CM
535
536 if (!ret)
537 d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
d5c12070
MX
538out:
539 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
f46b5a66
CH
540 return ret;
541}
542
e9662f70
MX
543static int create_snapshot(struct btrfs_root *root, struct inode *dir,
544 struct dentry *dentry, char *name, int namelen,
545 u64 *async_transid, bool readonly,
546 struct btrfs_qgroup_inherit *inherit)
f46b5a66 547{
2e4bfab9 548 struct inode *inode;
f46b5a66
CH
549 struct btrfs_pending_snapshot *pending_snapshot;
550 struct btrfs_trans_handle *trans;
2e4bfab9 551 int ret;
f46b5a66
CH
552
553 if (!root->ref_cows)
554 return -EINVAL;
555
3de4586c 556 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
a22285a6
YZ
557 if (!pending_snapshot)
558 return -ENOMEM;
559
66d8f3dd
MX
560 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
561 BTRFS_BLOCK_RSV_TEMP);
d5c12070
MX
562 /*
563 * 1 - parent dir inode
564 * 2 - dir entries
565 * 1 - root item
566 * 2 - root ref/backref
567 * 1 - root of snapshot
568 */
569 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
570 &pending_snapshot->block_rsv, 7,
571 &pending_snapshot->qgroup_reserved);
572 if (ret)
573 goto out;
574
3de4586c 575 pending_snapshot->dentry = dentry;
f46b5a66 576 pending_snapshot->root = root;
b83cc969 577 pending_snapshot->readonly = readonly;
e9662f70 578 pending_snapshot->dir = dir;
8696c533 579 pending_snapshot->inherit = inherit;
a22285a6 580
d5c12070 581 trans = btrfs_start_transaction(root, 0);
a22285a6
YZ
582 if (IS_ERR(trans)) {
583 ret = PTR_ERR(trans);
584 goto fail;
585 }
586
8351583e 587 spin_lock(&root->fs_info->trans_lock);
f46b5a66
CH
588 list_add(&pending_snapshot->list,
589 &trans->transaction->pending_snapshots);
8351583e 590 spin_unlock(&root->fs_info->trans_lock);
72fd032e
SW
591 if (async_transid) {
592 *async_transid = trans->transid;
593 ret = btrfs_commit_transaction_async(trans,
594 root->fs_info->extent_root, 1);
595 } else {
596 ret = btrfs_commit_transaction(trans,
597 root->fs_info->extent_root);
598 }
109f2365
LB
599 if (ret) {
600 /* cleanup_transaction has freed this for us */
601 if (trans->aborted)
602 pending_snapshot = NULL;
c37b2b62 603 goto fail;
109f2365 604 }
a22285a6
YZ
605
606 ret = pending_snapshot->error;
607 if (ret)
608 goto fail;
609
66b4ffd1
JB
610 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
611 if (ret)
612 goto fail;
f46b5a66 613
2fbe8c8a 614 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
2e4bfab9
YZ
615 if (IS_ERR(inode)) {
616 ret = PTR_ERR(inode);
617 goto fail;
618 }
619 BUG_ON(!inode);
620 d_instantiate(dentry, inode);
621 ret = 0;
622fail:
d5c12070
MX
623 btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
624 &pending_snapshot->block_rsv,
625 pending_snapshot->qgroup_reserved);
626out:
a22285a6 627 kfree(pending_snapshot);
f46b5a66
CH
628 return ret;
629}
630
4260f7c7
SW
631/* copy of check_sticky in fs/namei.c()
632* It's inline, so penalty for filesystems that don't use sticky bit is
633* minimal.
634*/
635static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
636{
2f2f43d3 637 kuid_t fsuid = current_fsuid();
4260f7c7
SW
638
639 if (!(dir->i_mode & S_ISVTX))
640 return 0;
2f2f43d3 641 if (uid_eq(inode->i_uid, fsuid))
4260f7c7 642 return 0;
2f2f43d3 643 if (uid_eq(dir->i_uid, fsuid))
4260f7c7
SW
644 return 0;
645 return !capable(CAP_FOWNER);
646}
647
648/* copy of may_delete in fs/namei.c()
649 * Check whether we can remove a link victim from directory dir, check
650 * whether the type of victim is right.
651 * 1. We can't do it if dir is read-only (done in permission())
652 * 2. We should have write and exec permissions on dir
653 * 3. We can't remove anything from append-only dir
654 * 4. We can't do anything with immutable dir (done in permission())
655 * 5. If the sticky bit on dir is set we should either
656 * a. be owner of dir, or
657 * b. be owner of victim, or
658 * c. have CAP_FOWNER capability
659 * 6. If the victim is append-only or immutable we can't do antyhing with
660 * links pointing to it.
661 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
662 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
663 * 9. We can't remove a root or mountpoint.
664 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
665 * nfs_async_unlink().
666 */
667
668static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
669{
670 int error;
671
672 if (!victim->d_inode)
673 return -ENOENT;
674
675 BUG_ON(victim->d_parent->d_inode != dir);
4fa6b5ec 676 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
4260f7c7
SW
677
678 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
679 if (error)
680 return error;
681 if (IS_APPEND(dir))
682 return -EPERM;
683 if (btrfs_check_sticky(dir, victim->d_inode)||
684 IS_APPEND(victim->d_inode)||
685 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
686 return -EPERM;
687 if (isdir) {
688 if (!S_ISDIR(victim->d_inode->i_mode))
689 return -ENOTDIR;
690 if (IS_ROOT(victim))
691 return -EBUSY;
692 } else if (S_ISDIR(victim->d_inode->i_mode))
693 return -EISDIR;
694 if (IS_DEADDIR(dir))
695 return -ENOENT;
696 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
697 return -EBUSY;
698 return 0;
699}
700
cb8e7090
CH
701/* copy of may_create in fs/namei.c() */
702static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
703{
704 if (child->d_inode)
705 return -EEXIST;
706 if (IS_DEADDIR(dir))
707 return -ENOENT;
708 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
709}
710
711/*
712 * Create a new subvolume below @parent. This is largely modeled after
713 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
714 * inside this filesystem so it's quite a bit simpler.
715 */
76dda93c
YZ
716static noinline int btrfs_mksubvol(struct path *parent,
717 char *name, int namelen,
72fd032e 718 struct btrfs_root *snap_src,
6f72c7e2 719 u64 *async_transid, bool readonly,
8696c533 720 struct btrfs_qgroup_inherit *inherit)
cb8e7090 721{
76dda93c 722 struct inode *dir = parent->dentry->d_inode;
cb8e7090
CH
723 struct dentry *dentry;
724 int error;
725
76dda93c 726 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
cb8e7090
CH
727
728 dentry = lookup_one_len(name, parent->dentry, namelen);
729 error = PTR_ERR(dentry);
730 if (IS_ERR(dentry))
731 goto out_unlock;
732
733 error = -EEXIST;
734 if (dentry->d_inode)
735 goto out_dput;
736
76dda93c 737 error = btrfs_may_create(dir, dentry);
cb8e7090 738 if (error)
a874a63e 739 goto out_dput;
cb8e7090 740
9c52057c
CM
741 /*
742 * even if this name doesn't exist, we may get hash collisions.
743 * check for them now when we can safely fail
744 */
745 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
746 dir->i_ino, name,
747 namelen);
748 if (error)
749 goto out_dput;
750
76dda93c
YZ
751 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
752
753 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
754 goto out_up_read;
755
3de4586c 756 if (snap_src) {
e9662f70 757 error = create_snapshot(snap_src, dir, dentry, name, namelen,
6f72c7e2 758 async_transid, readonly, inherit);
3de4586c 759 } else {
d5c12070
MX
760 error = create_subvol(dir, dentry, name, namelen,
761 async_transid, inherit);
3de4586c 762 }
76dda93c
YZ
763 if (!error)
764 fsnotify_mkdir(dir, dentry);
765out_up_read:
766 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
cb8e7090
CH
767out_dput:
768 dput(dentry);
769out_unlock:
76dda93c 770 mutex_unlock(&dir->i_mutex);
cb8e7090
CH
771 return error;
772}
773
4cb5300b
CM
774/*
775 * When we're defragging a range, we don't want to kick it off again
776 * if it is really just waiting for delalloc to send it down.
777 * If we find a nice big extent or delalloc range for the bytes in the
778 * file you want to defrag, we return 0 to let you know to skip this
779 * part of the file
780 */
781static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
782{
783 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
784 struct extent_map *em = NULL;
785 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
786 u64 end;
787
788 read_lock(&em_tree->lock);
789 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
790 read_unlock(&em_tree->lock);
791
792 if (em) {
793 end = extent_map_end(em);
794 free_extent_map(em);
795 if (end - offset > thresh)
796 return 0;
797 }
798 /* if we already have a nice delalloc here, just stop */
799 thresh /= 2;
800 end = count_range_bits(io_tree, &offset, offset + thresh,
801 thresh, EXTENT_DELALLOC, 1);
802 if (end >= thresh)
803 return 0;
804 return 1;
805}
806
807/*
808 * helper function to walk through a file and find extents
809 * newer than a specific transid, and smaller than thresh.
810 *
811 * This is used by the defragging code to find new and small
812 * extents
813 */
814static int find_new_extents(struct btrfs_root *root,
815 struct inode *inode, u64 newer_than,
816 u64 *off, int thresh)
817{
818 struct btrfs_path *path;
819 struct btrfs_key min_key;
820 struct btrfs_key max_key;
821 struct extent_buffer *leaf;
822 struct btrfs_file_extent_item *extent;
823 int type;
824 int ret;
a4689d2b 825 u64 ino = btrfs_ino(inode);
4cb5300b
CM
826
827 path = btrfs_alloc_path();
828 if (!path)
829 return -ENOMEM;
830
a4689d2b 831 min_key.objectid = ino;
4cb5300b
CM
832 min_key.type = BTRFS_EXTENT_DATA_KEY;
833 min_key.offset = *off;
834
a4689d2b 835 max_key.objectid = ino;
4cb5300b
CM
836 max_key.type = (u8)-1;
837 max_key.offset = (u64)-1;
838
839 path->keep_locks = 1;
840
841 while(1) {
842 ret = btrfs_search_forward(root, &min_key, &max_key,
de78b51a 843 path, newer_than);
4cb5300b
CM
844 if (ret != 0)
845 goto none;
a4689d2b 846 if (min_key.objectid != ino)
4cb5300b
CM
847 goto none;
848 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
849 goto none;
850
851 leaf = path->nodes[0];
852 extent = btrfs_item_ptr(leaf, path->slots[0],
853 struct btrfs_file_extent_item);
854
855 type = btrfs_file_extent_type(leaf, extent);
856 if (type == BTRFS_FILE_EXTENT_REG &&
857 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
858 check_defrag_in_cache(inode, min_key.offset, thresh)) {
859 *off = min_key.offset;
860 btrfs_free_path(path);
861 return 0;
862 }
863
864 if (min_key.offset == (u64)-1)
865 goto none;
866
867 min_key.offset++;
868 btrfs_release_path(path);
869 }
870none:
871 btrfs_free_path(path);
872 return -ENOENT;
873}
874
6c282eb4 875static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
17ce6ef8
LB
876{
877 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
6c282eb4
LZ
878 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
879 struct extent_map *em;
880 u64 len = PAGE_CACHE_SIZE;
17ce6ef8 881
6c282eb4
LZ
882 /*
883 * hopefully we have this extent in the tree already, try without
884 * the full extent lock
885 */
17ce6ef8 886 read_lock(&em_tree->lock);
6c282eb4 887 em = lookup_extent_mapping(em_tree, start, len);
17ce6ef8
LB
888 read_unlock(&em_tree->lock);
889
6c282eb4
LZ
890 if (!em) {
891 /* get the big lock and read metadata off disk */
892 lock_extent(io_tree, start, start + len - 1);
893 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
894 unlock_extent(io_tree, start, start + len - 1);
895
896 if (IS_ERR(em))
897 return NULL;
898 }
899
900 return em;
901}
17ce6ef8 902
6c282eb4
LZ
903static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
904{
905 struct extent_map *next;
906 bool ret = true;
907
908 /* this is the last extent */
909 if (em->start + em->len >= i_size_read(inode))
910 return false;
911
912 next = defrag_lookup_extent(inode, em->start + em->len);
913 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
914 ret = false;
915
916 free_extent_map(next);
17ce6ef8
LB
917 return ret;
918}
919
6c282eb4 920static int should_defrag_range(struct inode *inode, u64 start, int thresh,
a43a2111
AM
921 u64 *last_len, u64 *skip, u64 *defrag_end,
922 int compress)
940100a4 923{
6c282eb4 924 struct extent_map *em;
940100a4 925 int ret = 1;
6c282eb4 926 bool next_mergeable = true;
940100a4
CM
927
928 /*
008873ea 929 * make sure that once we start defragging an extent, we keep on
940100a4
CM
930 * defragging it
931 */
932 if (start < *defrag_end)
933 return 1;
934
935 *skip = 0;
936
6c282eb4
LZ
937 em = defrag_lookup_extent(inode, start);
938 if (!em)
939 return 0;
940100a4
CM
940
941 /* this will cover holes, and inline extents */
17ce6ef8 942 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
940100a4 943 ret = 0;
17ce6ef8
LB
944 goto out;
945 }
946
6c282eb4 947 next_mergeable = defrag_check_next_extent(inode, em);
940100a4
CM
948
949 /*
6c282eb4
LZ
950 * we hit a real extent, if it is big or the next extent is not a
951 * real extent, don't bother defragging it
940100a4 952 */
a43a2111 953 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
6c282eb4 954 (em->len >= thresh || !next_mergeable))
940100a4 955 ret = 0;
17ce6ef8 956out:
940100a4
CM
957 /*
958 * last_len ends up being a counter of how many bytes we've defragged.
959 * every time we choose not to defrag an extent, we reset *last_len
960 * so that the next tiny extent will force a defrag.
961 *
962 * The end result of this is that tiny extents before a single big
963 * extent will force at least part of that big extent to be defragged.
964 */
965 if (ret) {
940100a4
CM
966 *defrag_end = extent_map_end(em);
967 } else {
968 *last_len = 0;
969 *skip = extent_map_end(em);
970 *defrag_end = 0;
971 }
972
973 free_extent_map(em);
974 return ret;
975}
976
4cb5300b
CM
977/*
978 * it doesn't do much good to defrag one or two pages
979 * at a time. This pulls in a nice chunk of pages
980 * to COW and defrag.
981 *
982 * It also makes sure the delalloc code has enough
983 * dirty data to avoid making new small extents as part
984 * of the defrag
985 *
986 * It's a good idea to start RA on this range
987 * before calling this.
988 */
989static int cluster_pages_for_defrag(struct inode *inode,
990 struct page **pages,
991 unsigned long start_index,
992 int num_pages)
f46b5a66 993{
4cb5300b
CM
994 unsigned long file_end;
995 u64 isize = i_size_read(inode);
996 u64 page_start;
997 u64 page_end;
1f12bd06 998 u64 page_cnt;
4cb5300b
CM
999 int ret;
1000 int i;
1001 int i_done;
3eaa2885 1002 struct btrfs_ordered_extent *ordered;
4cb5300b 1003 struct extent_state *cached_state = NULL;
600a45e1 1004 struct extent_io_tree *tree;
3b16a4e3 1005 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
4cb5300b 1006
4cb5300b 1007 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1f12bd06
LB
1008 if (!isize || start_index > file_end)
1009 return 0;
1010
1011 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
4cb5300b
CM
1012
1013 ret = btrfs_delalloc_reserve_space(inode,
1f12bd06 1014 page_cnt << PAGE_CACHE_SHIFT);
4cb5300b
CM
1015 if (ret)
1016 return ret;
4cb5300b 1017 i_done = 0;
600a45e1 1018 tree = &BTRFS_I(inode)->io_tree;
4cb5300b
CM
1019
1020 /* step one, lock all the pages */
1f12bd06 1021 for (i = 0; i < page_cnt; i++) {
4cb5300b 1022 struct page *page;
600a45e1 1023again:
a94733d0 1024 page = find_or_create_page(inode->i_mapping,
600a45e1 1025 start_index + i, mask);
4cb5300b
CM
1026 if (!page)
1027 break;
1028
600a45e1
MX
1029 page_start = page_offset(page);
1030 page_end = page_start + PAGE_CACHE_SIZE - 1;
1031 while (1) {
d0082371 1032 lock_extent(tree, page_start, page_end);
600a45e1
MX
1033 ordered = btrfs_lookup_ordered_extent(inode,
1034 page_start);
d0082371 1035 unlock_extent(tree, page_start, page_end);
600a45e1
MX
1036 if (!ordered)
1037 break;
1038
1039 unlock_page(page);
1040 btrfs_start_ordered_extent(inode, ordered, 1);
1041 btrfs_put_ordered_extent(ordered);
1042 lock_page(page);
1f12bd06
LB
1043 /*
1044 * we unlocked the page above, so we need check if
1045 * it was released or not.
1046 */
1047 if (page->mapping != inode->i_mapping) {
1048 unlock_page(page);
1049 page_cache_release(page);
1050 goto again;
1051 }
600a45e1
MX
1052 }
1053
4cb5300b
CM
1054 if (!PageUptodate(page)) {
1055 btrfs_readpage(NULL, page);
1056 lock_page(page);
1057 if (!PageUptodate(page)) {
1058 unlock_page(page);
1059 page_cache_release(page);
1060 ret = -EIO;
1061 break;
1062 }
1063 }
600a45e1 1064
600a45e1
MX
1065 if (page->mapping != inode->i_mapping) {
1066 unlock_page(page);
1067 page_cache_release(page);
1068 goto again;
1069 }
1070
4cb5300b
CM
1071 pages[i] = page;
1072 i_done++;
1073 }
1074 if (!i_done || ret)
1075 goto out;
1076
1077 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1078 goto out;
1079
1080 /*
1081 * so now we have a nice long stream of locked
1082 * and up to date pages, lets wait on them
1083 */
1084 for (i = 0; i < i_done; i++)
1085 wait_on_page_writeback(pages[i]);
1086
1087 page_start = page_offset(pages[0]);
1088 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1089
1090 lock_extent_bits(&BTRFS_I(inode)->io_tree,
d0082371 1091 page_start, page_end - 1, 0, &cached_state);
4cb5300b
CM
1092 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1093 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
9e8a4a8b
LB
1094 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1095 &cached_state, GFP_NOFS);
4cb5300b 1096
1f12bd06 1097 if (i_done != page_cnt) {
9e0baf60
JB
1098 spin_lock(&BTRFS_I(inode)->lock);
1099 BTRFS_I(inode)->outstanding_extents++;
1100 spin_unlock(&BTRFS_I(inode)->lock);
4cb5300b 1101 btrfs_delalloc_release_space(inode,
1f12bd06 1102 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
4cb5300b
CM
1103 }
1104
1105
9e8a4a8b
LB
1106 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1107 &cached_state, GFP_NOFS);
4cb5300b
CM
1108
1109 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1110 page_start, page_end - 1, &cached_state,
1111 GFP_NOFS);
1112
1113 for (i = 0; i < i_done; i++) {
1114 clear_page_dirty_for_io(pages[i]);
1115 ClearPageChecked(pages[i]);
1116 set_page_extent_mapped(pages[i]);
1117 set_page_dirty(pages[i]);
1118 unlock_page(pages[i]);
1119 page_cache_release(pages[i]);
1120 }
1121 return i_done;
1122out:
1123 for (i = 0; i < i_done; i++) {
1124 unlock_page(pages[i]);
1125 page_cache_release(pages[i]);
1126 }
1f12bd06 1127 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
4cb5300b
CM
1128 return ret;
1129
1130}
1131
1132int btrfs_defrag_file(struct inode *inode, struct file *file,
1133 struct btrfs_ioctl_defrag_range_args *range,
1134 u64 newer_than, unsigned long max_to_defrag)
1135{
1136 struct btrfs_root *root = BTRFS_I(inode)->root;
4cb5300b 1137 struct file_ra_state *ra = NULL;
f46b5a66 1138 unsigned long last_index;
151a31b2 1139 u64 isize = i_size_read(inode);
940100a4
CM
1140 u64 last_len = 0;
1141 u64 skip = 0;
1142 u64 defrag_end = 0;
4cb5300b 1143 u64 newer_off = range->start;
f46b5a66 1144 unsigned long i;
008873ea 1145 unsigned long ra_index = 0;
f46b5a66 1146 int ret;
4cb5300b 1147 int defrag_count = 0;
1a419d85 1148 int compress_type = BTRFS_COMPRESS_ZLIB;
4cb5300b 1149 int extent_thresh = range->extent_thresh;
008873ea
LZ
1150 int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1151 int cluster = max_cluster;
4cb5300b
CM
1152 u64 new_align = ~((u64)128 * 1024 - 1);
1153 struct page **pages = NULL;
1154
1155 if (extent_thresh == 0)
1156 extent_thresh = 256 * 1024;
1a419d85
LZ
1157
1158 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1159 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1160 return -EINVAL;
1161 if (range->compress_type)
1162 compress_type = range->compress_type;
1163 }
f46b5a66 1164
151a31b2 1165 if (isize == 0)
940100a4
CM
1166 return 0;
1167
4cb5300b
CM
1168 /*
1169 * if we were not given a file, allocate a readahead
1170 * context
1171 */
1172 if (!file) {
1173 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1174 if (!ra)
1175 return -ENOMEM;
1176 file_ra_state_init(ra, inode->i_mapping);
1177 } else {
1178 ra = &file->f_ra;
1179 }
1180
008873ea 1181 pages = kmalloc(sizeof(struct page *) * max_cluster,
4cb5300b
CM
1182 GFP_NOFS);
1183 if (!pages) {
1184 ret = -ENOMEM;
1185 goto out_ra;
1186 }
1187
1188 /* find the last page to defrag */
1e701a32 1189 if (range->start + range->len > range->start) {
151a31b2 1190 last_index = min_t(u64, isize - 1,
1e701a32
CM
1191 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1192 } else {
151a31b2 1193 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1e701a32
CM
1194 }
1195
4cb5300b
CM
1196 if (newer_than) {
1197 ret = find_new_extents(root, inode, newer_than,
1198 &newer_off, 64 * 1024);
1199 if (!ret) {
1200 range->start = newer_off;
1201 /*
1202 * we always align our defrag to help keep
1203 * the extents in the file evenly spaced
1204 */
1205 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
4cb5300b
CM
1206 } else
1207 goto out_ra;
1208 } else {
1209 i = range->start >> PAGE_CACHE_SHIFT;
1210 }
1211 if (!max_to_defrag)
7ec31b54 1212 max_to_defrag = last_index + 1;
4cb5300b 1213
2a0f7f57
LZ
1214 /*
1215 * make writeback starts from i, so the defrag range can be
1216 * written sequentially.
1217 */
1218 if (i < inode->i_mapping->writeback_index)
1219 inode->i_mapping->writeback_index = i;
1220
f7f43cc8
CM
1221 while (i <= last_index && defrag_count < max_to_defrag &&
1222 (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1223 PAGE_CACHE_SHIFT)) {
4cb5300b
CM
1224 /*
1225 * make sure we stop running if someone unmounts
1226 * the FS
1227 */
1228 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1229 break;
1230
210549eb
DS
1231 if (btrfs_defrag_cancelled(root->fs_info)) {
1232 printk(KERN_DEBUG "btrfs: defrag_file cancelled\n");
1233 ret = -EAGAIN;
1234 break;
1235 }
1236
66c26892 1237 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
6c282eb4 1238 extent_thresh, &last_len, &skip,
a43a2111
AM
1239 &defrag_end, range->flags &
1240 BTRFS_DEFRAG_RANGE_COMPRESS)) {
940100a4
CM
1241 unsigned long next;
1242 /*
1243 * the should_defrag function tells us how much to skip
1244 * bump our counter by the suggested amount
1245 */
1246 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1247 i = max(i + 1, next);
1248 continue;
1249 }
008873ea
LZ
1250
1251 if (!newer_than) {
1252 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1253 PAGE_CACHE_SHIFT) - i;
1254 cluster = min(cluster, max_cluster);
1255 } else {
1256 cluster = max_cluster;
1257 }
1258
1e701a32 1259 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1a419d85 1260 BTRFS_I(inode)->force_compress = compress_type;
940100a4 1261
008873ea
LZ
1262 if (i + cluster > ra_index) {
1263 ra_index = max(i, ra_index);
1264 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1265 cluster);
1266 ra_index += max_cluster;
1267 }
940100a4 1268
ecb8bea8 1269 mutex_lock(&inode->i_mutex);
008873ea 1270 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
ecb8bea8
LB
1271 if (ret < 0) {
1272 mutex_unlock(&inode->i_mutex);
4cb5300b 1273 goto out_ra;
ecb8bea8 1274 }
4cb5300b
CM
1275
1276 defrag_count += ret;
d0e1d66b 1277 balance_dirty_pages_ratelimited(inode->i_mapping);
ecb8bea8 1278 mutex_unlock(&inode->i_mutex);
4cb5300b
CM
1279
1280 if (newer_than) {
1281 if (newer_off == (u64)-1)
1282 break;
1283
e1f041e1
LB
1284 if (ret > 0)
1285 i += ret;
1286
4cb5300b
CM
1287 newer_off = max(newer_off + 1,
1288 (u64)i << PAGE_CACHE_SHIFT);
1289
1290 ret = find_new_extents(root, inode,
1291 newer_than, &newer_off,
1292 64 * 1024);
1293 if (!ret) {
1294 range->start = newer_off;
1295 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
4cb5300b
CM
1296 } else {
1297 break;
f46b5a66 1298 }
4cb5300b 1299 } else {
008873ea 1300 if (ret > 0) {
cbcc8326 1301 i += ret;
008873ea
LZ
1302 last_len += ret << PAGE_CACHE_SHIFT;
1303 } else {
cbcc8326 1304 i++;
008873ea
LZ
1305 last_len = 0;
1306 }
f46b5a66 1307 }
f46b5a66
CH
1308 }
1309
1e701a32
CM
1310 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1311 filemap_flush(inode->i_mapping);
1312
1313 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1314 /* the filemap_flush will queue IO into the worker threads, but
1315 * we have to make sure the IO is actually started and that
1316 * ordered extents get created before we return
1317 */
1318 atomic_inc(&root->fs_info->async_submit_draining);
1319 while (atomic_read(&root->fs_info->nr_async_submits) ||
1320 atomic_read(&root->fs_info->async_delalloc_pages)) {
1321 wait_event(root->fs_info->async_submit_wait,
1322 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1323 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1324 }
1325 atomic_dec(&root->fs_info->async_submit_draining);
1326
1327 mutex_lock(&inode->i_mutex);
261507a0 1328 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1e701a32
CM
1329 mutex_unlock(&inode->i_mutex);
1330 }
1331
1a419d85 1332 if (range->compress_type == BTRFS_COMPRESS_LZO) {
2b0ce2c2 1333 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1a419d85
LZ
1334 }
1335
60ccf82f 1336 ret = defrag_count;
940100a4 1337
4cb5300b
CM
1338out_ra:
1339 if (!file)
1340 kfree(ra);
1341 kfree(pages);
940100a4 1342 return ret;
f46b5a66
CH
1343}
1344
198605a8 1345static noinline int btrfs_ioctl_resize(struct file *file,
76dda93c 1346 void __user *arg)
f46b5a66
CH
1347{
1348 u64 new_size;
1349 u64 old_size;
1350 u64 devid = 1;
496ad9aa 1351 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
f46b5a66
CH
1352 struct btrfs_ioctl_vol_args *vol_args;
1353 struct btrfs_trans_handle *trans;
1354 struct btrfs_device *device = NULL;
1355 char *sizestr;
1356 char *devstr = NULL;
1357 int ret = 0;
f46b5a66
CH
1358 int mod = 0;
1359
e441d54d
CM
1360 if (!capable(CAP_SYS_ADMIN))
1361 return -EPERM;
1362
198605a8
MX
1363 ret = mnt_want_write_file(file);
1364 if (ret)
1365 return ret;
1366
5ac00add
SB
1367 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1368 1)) {
1369 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
97547676 1370 mnt_drop_write_file(file);
2c0c9da0 1371 return -EINVAL;
c9e9f97b
ID
1372 }
1373
5ac00add 1374 mutex_lock(&root->fs_info->volume_mutex);
dae7b665 1375 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID
1376 if (IS_ERR(vol_args)) {
1377 ret = PTR_ERR(vol_args);
1378 goto out;
1379 }
5516e595
MF
1380
1381 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 1382
f46b5a66
CH
1383 sizestr = vol_args->name;
1384 devstr = strchr(sizestr, ':');
1385 if (devstr) {
1386 char *end;
1387 sizestr = devstr + 1;
1388 *devstr = '\0';
1389 devstr = vol_args->name;
1390 devid = simple_strtoull(devstr, &end, 10);
dfd79829
MX
1391 if (!devid) {
1392 ret = -EINVAL;
1393 goto out_free;
1394 }
5bb14682 1395 printk(KERN_INFO "btrfs: resizing devid %llu\n",
21380931 1396 (unsigned long long)devid);
f46b5a66 1397 }
dba60f3f 1398
aa1b8cd4 1399 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
f46b5a66 1400 if (!device) {
5bb14682 1401 printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
21380931 1402 (unsigned long long)devid);
dfd79829 1403 ret = -ENODEV;
c9e9f97b 1404 goto out_free;
f46b5a66 1405 }
dba60f3f
MX
1406
1407 if (!device->writeable) {
4e42ae1b 1408 printk(KERN_INFO "btrfs: resizer unable to apply on "
dba60f3f 1409 "readonly device %llu\n",
a8c4a33b 1410 (unsigned long long)devid);
dfd79829 1411 ret = -EPERM;
4e42ae1b
LB
1412 goto out_free;
1413 }
1414
f46b5a66
CH
1415 if (!strcmp(sizestr, "max"))
1416 new_size = device->bdev->bd_inode->i_size;
1417 else {
1418 if (sizestr[0] == '-') {
1419 mod = -1;
1420 sizestr++;
1421 } else if (sizestr[0] == '+') {
1422 mod = 1;
1423 sizestr++;
1424 }
91748467 1425 new_size = memparse(sizestr, NULL);
f46b5a66
CH
1426 if (new_size == 0) {
1427 ret = -EINVAL;
c9e9f97b 1428 goto out_free;
f46b5a66
CH
1429 }
1430 }
1431
63a212ab 1432 if (device->is_tgtdev_for_dev_replace) {
dfd79829 1433 ret = -EPERM;
63a212ab
SB
1434 goto out_free;
1435 }
1436
f46b5a66
CH
1437 old_size = device->total_bytes;
1438
1439 if (mod < 0) {
1440 if (new_size > old_size) {
1441 ret = -EINVAL;
c9e9f97b 1442 goto out_free;
f46b5a66
CH
1443 }
1444 new_size = old_size - new_size;
1445 } else if (mod > 0) {
1446 new_size = old_size + new_size;
1447 }
1448
1449 if (new_size < 256 * 1024 * 1024) {
1450 ret = -EINVAL;
c9e9f97b 1451 goto out_free;
f46b5a66
CH
1452 }
1453 if (new_size > device->bdev->bd_inode->i_size) {
1454 ret = -EFBIG;
c9e9f97b 1455 goto out_free;
f46b5a66
CH
1456 }
1457
1458 do_div(new_size, root->sectorsize);
1459 new_size *= root->sectorsize;
1460
606686ee
JB
1461 printk_in_rcu(KERN_INFO "btrfs: new size for %s is %llu\n",
1462 rcu_str_deref(device->name),
1463 (unsigned long long)new_size);
f46b5a66
CH
1464
1465 if (new_size > old_size) {
a22285a6 1466 trans = btrfs_start_transaction(root, 0);
98d5dc13
TI
1467 if (IS_ERR(trans)) {
1468 ret = PTR_ERR(trans);
c9e9f97b 1469 goto out_free;
98d5dc13 1470 }
f46b5a66
CH
1471 ret = btrfs_grow_device(trans, device, new_size);
1472 btrfs_commit_transaction(trans, root);
ece7d20e 1473 } else if (new_size < old_size) {
f46b5a66 1474 ret = btrfs_shrink_device(device, new_size);
0253f40e 1475 } /* equal, nothing need to do */
f46b5a66 1476
c9e9f97b 1477out_free:
f46b5a66 1478 kfree(vol_args);
c9e9f97b
ID
1479out:
1480 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 1481 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
18f39c41 1482 mnt_drop_write_file(file);
f46b5a66
CH
1483 return ret;
1484}
1485
72fd032e 1486static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
6f72c7e2
AJ
1487 char *name, unsigned long fd, int subvol,
1488 u64 *transid, bool readonly,
8696c533 1489 struct btrfs_qgroup_inherit *inherit)
f46b5a66 1490{
f46b5a66 1491 int namelen;
3de4586c 1492 int ret = 0;
f46b5a66 1493
a874a63e
LB
1494 ret = mnt_want_write_file(file);
1495 if (ret)
1496 goto out;
1497
72fd032e
SW
1498 namelen = strlen(name);
1499 if (strchr(name, '/')) {
f46b5a66 1500 ret = -EINVAL;
a874a63e 1501 goto out_drop_write;
f46b5a66
CH
1502 }
1503
16780cab
CM
1504 if (name[0] == '.' &&
1505 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1506 ret = -EEXIST;
a874a63e 1507 goto out_drop_write;
16780cab
CM
1508 }
1509
3de4586c 1510 if (subvol) {
72fd032e 1511 ret = btrfs_mksubvol(&file->f_path, name, namelen,
6f72c7e2 1512 NULL, transid, readonly, inherit);
cb8e7090 1513 } else {
2903ff01 1514 struct fd src = fdget(fd);
3de4586c 1515 struct inode *src_inode;
2903ff01 1516 if (!src.file) {
3de4586c 1517 ret = -EINVAL;
a874a63e 1518 goto out_drop_write;
3de4586c
CM
1519 }
1520
496ad9aa
AV
1521 src_inode = file_inode(src.file);
1522 if (src_inode->i_sb != file_inode(file)->i_sb) {
d397712b
CM
1523 printk(KERN_INFO "btrfs: Snapshot src from "
1524 "another FS\n");
3de4586c 1525 ret = -EINVAL;
ecd18815
AV
1526 } else {
1527 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1528 BTRFS_I(src_inode)->root,
1529 transid, readonly, inherit);
3de4586c 1530 }
2903ff01 1531 fdput(src);
cb8e7090 1532 }
a874a63e
LB
1533out_drop_write:
1534 mnt_drop_write_file(file);
f46b5a66 1535out:
72fd032e
SW
1536 return ret;
1537}
1538
1539static noinline int btrfs_ioctl_snap_create(struct file *file,
fa0d2b9b 1540 void __user *arg, int subvol)
72fd032e 1541{
fa0d2b9b 1542 struct btrfs_ioctl_vol_args *vol_args;
72fd032e
SW
1543 int ret;
1544
fa0d2b9b
LZ
1545 vol_args = memdup_user(arg, sizeof(*vol_args));
1546 if (IS_ERR(vol_args))
1547 return PTR_ERR(vol_args);
1548 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
72fd032e 1549
fa0d2b9b 1550 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
b83cc969 1551 vol_args->fd, subvol,
6f72c7e2 1552 NULL, false, NULL);
fdfb1e4f 1553
fa0d2b9b
LZ
1554 kfree(vol_args);
1555 return ret;
1556}
fdfb1e4f 1557
fa0d2b9b
LZ
1558static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1559 void __user *arg, int subvol)
1560{
1561 struct btrfs_ioctl_vol_args_v2 *vol_args;
1562 int ret;
1563 u64 transid = 0;
1564 u64 *ptr = NULL;
b83cc969 1565 bool readonly = false;
6f72c7e2 1566 struct btrfs_qgroup_inherit *inherit = NULL;
75eaa0e2 1567
fa0d2b9b
LZ
1568 vol_args = memdup_user(arg, sizeof(*vol_args));
1569 if (IS_ERR(vol_args))
1570 return PTR_ERR(vol_args);
1571 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
75eaa0e2 1572
b83cc969 1573 if (vol_args->flags &
6f72c7e2
AJ
1574 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1575 BTRFS_SUBVOL_QGROUP_INHERIT)) {
b83cc969 1576 ret = -EOPNOTSUPP;
fa0d2b9b 1577 goto out;
72fd032e 1578 }
fa0d2b9b
LZ
1579
1580 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1581 ptr = &transid;
b83cc969
LZ
1582 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1583 readonly = true;
6f72c7e2
AJ
1584 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1585 if (vol_args->size > PAGE_CACHE_SIZE) {
1586 ret = -EINVAL;
1587 goto out;
1588 }
1589 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1590 if (IS_ERR(inherit)) {
1591 ret = PTR_ERR(inherit);
1592 goto out;
1593 }
1594 }
fa0d2b9b
LZ
1595
1596 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
6f72c7e2 1597 vol_args->fd, subvol, ptr,
8696c533 1598 readonly, inherit);
fa0d2b9b
LZ
1599
1600 if (ret == 0 && ptr &&
1601 copy_to_user(arg +
1602 offsetof(struct btrfs_ioctl_vol_args_v2,
1603 transid), ptr, sizeof(*ptr)))
1604 ret = -EFAULT;
fdfb1e4f 1605out:
f46b5a66 1606 kfree(vol_args);
6f72c7e2 1607 kfree(inherit);
f46b5a66
CH
1608 return ret;
1609}
1610
0caa102d
LZ
1611static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1612 void __user *arg)
1613{
496ad9aa 1614 struct inode *inode = file_inode(file);
0caa102d
LZ
1615 struct btrfs_root *root = BTRFS_I(inode)->root;
1616 int ret = 0;
1617 u64 flags = 0;
1618
33345d01 1619 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
0caa102d
LZ
1620 return -EINVAL;
1621
1622 down_read(&root->fs_info->subvol_sem);
1623 if (btrfs_root_readonly(root))
1624 flags |= BTRFS_SUBVOL_RDONLY;
1625 up_read(&root->fs_info->subvol_sem);
1626
1627 if (copy_to_user(arg, &flags, sizeof(flags)))
1628 ret = -EFAULT;
1629
1630 return ret;
1631}
1632
1633static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1634 void __user *arg)
1635{
496ad9aa 1636 struct inode *inode = file_inode(file);
0caa102d
LZ
1637 struct btrfs_root *root = BTRFS_I(inode)->root;
1638 struct btrfs_trans_handle *trans;
1639 u64 root_flags;
1640 u64 flags;
1641 int ret = 0;
1642
b9ca0664
LB
1643 ret = mnt_want_write_file(file);
1644 if (ret)
1645 goto out;
0caa102d 1646
b9ca0664
LB
1647 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1648 ret = -EINVAL;
1649 goto out_drop_write;
1650 }
0caa102d 1651
b9ca0664
LB
1652 if (copy_from_user(&flags, arg, sizeof(flags))) {
1653 ret = -EFAULT;
1654 goto out_drop_write;
1655 }
0caa102d 1656
b9ca0664
LB
1657 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1658 ret = -EINVAL;
1659 goto out_drop_write;
1660 }
0caa102d 1661
b9ca0664
LB
1662 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1663 ret = -EOPNOTSUPP;
1664 goto out_drop_write;
1665 }
0caa102d 1666
b9ca0664
LB
1667 if (!inode_owner_or_capable(inode)) {
1668 ret = -EACCES;
1669 goto out_drop_write;
1670 }
b4dc2b8c 1671
0caa102d
LZ
1672 down_write(&root->fs_info->subvol_sem);
1673
1674 /* nothing to do */
1675 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
b9ca0664 1676 goto out_drop_sem;
0caa102d
LZ
1677
1678 root_flags = btrfs_root_flags(&root->root_item);
1679 if (flags & BTRFS_SUBVOL_RDONLY)
1680 btrfs_set_root_flags(&root->root_item,
1681 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1682 else
1683 btrfs_set_root_flags(&root->root_item,
1684 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1685
1686 trans = btrfs_start_transaction(root, 1);
1687 if (IS_ERR(trans)) {
1688 ret = PTR_ERR(trans);
1689 goto out_reset;
1690 }
1691
b4dc2b8c 1692 ret = btrfs_update_root(trans, root->fs_info->tree_root,
0caa102d
LZ
1693 &root->root_key, &root->root_item);
1694
1695 btrfs_commit_transaction(trans, root);
1696out_reset:
1697 if (ret)
1698 btrfs_set_root_flags(&root->root_item, root_flags);
b9ca0664 1699out_drop_sem:
0caa102d 1700 up_write(&root->fs_info->subvol_sem);
b9ca0664
LB
1701out_drop_write:
1702 mnt_drop_write_file(file);
1703out:
0caa102d
LZ
1704 return ret;
1705}
1706
76dda93c
YZ
1707/*
1708 * helper to check if the subvolume references other subvolumes
1709 */
1710static noinline int may_destroy_subvol(struct btrfs_root *root)
1711{
1712 struct btrfs_path *path;
1713 struct btrfs_key key;
1714 int ret;
1715
1716 path = btrfs_alloc_path();
1717 if (!path)
1718 return -ENOMEM;
1719
1720 key.objectid = root->root_key.objectid;
1721 key.type = BTRFS_ROOT_REF_KEY;
1722 key.offset = (u64)-1;
1723
1724 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1725 &key, path, 0, 0);
1726 if (ret < 0)
1727 goto out;
1728 BUG_ON(ret == 0);
1729
1730 ret = 0;
1731 if (path->slots[0] > 0) {
1732 path->slots[0]--;
1733 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1734 if (key.objectid == root->root_key.objectid &&
1735 key.type == BTRFS_ROOT_REF_KEY)
1736 ret = -ENOTEMPTY;
1737 }
1738out:
1739 btrfs_free_path(path);
1740 return ret;
1741}
1742
ac8e9819
CM
1743static noinline int key_in_sk(struct btrfs_key *key,
1744 struct btrfs_ioctl_search_key *sk)
1745{
abc6e134
CM
1746 struct btrfs_key test;
1747 int ret;
1748
1749 test.objectid = sk->min_objectid;
1750 test.type = sk->min_type;
1751 test.offset = sk->min_offset;
1752
1753 ret = btrfs_comp_cpu_keys(key, &test);
1754 if (ret < 0)
ac8e9819 1755 return 0;
abc6e134
CM
1756
1757 test.objectid = sk->max_objectid;
1758 test.type = sk->max_type;
1759 test.offset = sk->max_offset;
1760
1761 ret = btrfs_comp_cpu_keys(key, &test);
1762 if (ret > 0)
ac8e9819
CM
1763 return 0;
1764 return 1;
1765}
1766
1767static noinline int copy_to_sk(struct btrfs_root *root,
1768 struct btrfs_path *path,
1769 struct btrfs_key *key,
1770 struct btrfs_ioctl_search_key *sk,
1771 char *buf,
1772 unsigned long *sk_offset,
1773 int *num_found)
1774{
1775 u64 found_transid;
1776 struct extent_buffer *leaf;
1777 struct btrfs_ioctl_search_header sh;
1778 unsigned long item_off;
1779 unsigned long item_len;
1780 int nritems;
1781 int i;
1782 int slot;
ac8e9819
CM
1783 int ret = 0;
1784
1785 leaf = path->nodes[0];
1786 slot = path->slots[0];
1787 nritems = btrfs_header_nritems(leaf);
1788
1789 if (btrfs_header_generation(leaf) > sk->max_transid) {
1790 i = nritems;
1791 goto advance_key;
1792 }
1793 found_transid = btrfs_header_generation(leaf);
1794
1795 for (i = slot; i < nritems; i++) {
1796 item_off = btrfs_item_ptr_offset(leaf, i);
1797 item_len = btrfs_item_size_nr(leaf, i);
1798
1799 if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1800 item_len = 0;
1801
1802 if (sizeof(sh) + item_len + *sk_offset >
1803 BTRFS_SEARCH_ARGS_BUFSIZE) {
1804 ret = 1;
1805 goto overflow;
1806 }
1807
1808 btrfs_item_key_to_cpu(leaf, key, i);
1809 if (!key_in_sk(key, sk))
1810 continue;
1811
1812 sh.objectid = key->objectid;
1813 sh.offset = key->offset;
1814 sh.type = key->type;
1815 sh.len = item_len;
1816 sh.transid = found_transid;
1817
1818 /* copy search result header */
1819 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1820 *sk_offset += sizeof(sh);
1821
1822 if (item_len) {
1823 char *p = buf + *sk_offset;
1824 /* copy the item */
1825 read_extent_buffer(leaf, p,
1826 item_off, item_len);
1827 *sk_offset += item_len;
ac8e9819 1828 }
e2156867 1829 (*num_found)++;
ac8e9819
CM
1830
1831 if (*num_found >= sk->nr_items)
1832 break;
1833 }
1834advance_key:
abc6e134
CM
1835 ret = 0;
1836 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
ac8e9819 1837 key->offset++;
abc6e134
CM
1838 else if (key->type < (u8)-1 && key->type < sk->max_type) {
1839 key->offset = 0;
ac8e9819 1840 key->type++;
abc6e134
CM
1841 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1842 key->offset = 0;
1843 key->type = 0;
ac8e9819 1844 key->objectid++;
abc6e134
CM
1845 } else
1846 ret = 1;
ac8e9819 1847overflow:
ac8e9819
CM
1848 return ret;
1849}
1850
1851static noinline int search_ioctl(struct inode *inode,
1852 struct btrfs_ioctl_search_args *args)
1853{
1854 struct btrfs_root *root;
1855 struct btrfs_key key;
1856 struct btrfs_key max_key;
1857 struct btrfs_path *path;
1858 struct btrfs_ioctl_search_key *sk = &args->key;
1859 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1860 int ret;
1861 int num_found = 0;
1862 unsigned long sk_offset = 0;
1863
1864 path = btrfs_alloc_path();
1865 if (!path)
1866 return -ENOMEM;
1867
1868 if (sk->tree_id == 0) {
1869 /* search the root of the inode that was passed */
1870 root = BTRFS_I(inode)->root;
1871 } else {
1872 key.objectid = sk->tree_id;
1873 key.type = BTRFS_ROOT_ITEM_KEY;
1874 key.offset = (u64)-1;
1875 root = btrfs_read_fs_root_no_name(info, &key);
1876 if (IS_ERR(root)) {
1877 printk(KERN_ERR "could not find root %llu\n",
1878 sk->tree_id);
1879 btrfs_free_path(path);
1880 return -ENOENT;
1881 }
1882 }
1883
1884 key.objectid = sk->min_objectid;
1885 key.type = sk->min_type;
1886 key.offset = sk->min_offset;
1887
1888 max_key.objectid = sk->max_objectid;
1889 max_key.type = sk->max_type;
1890 max_key.offset = sk->max_offset;
1891
1892 path->keep_locks = 1;
1893
1894 while(1) {
de78b51a 1895 ret = btrfs_search_forward(root, &key, &max_key, path,
ac8e9819
CM
1896 sk->min_transid);
1897 if (ret != 0) {
1898 if (ret > 0)
1899 ret = 0;
1900 goto err;
1901 }
1902 ret = copy_to_sk(root, path, &key, sk, args->buf,
1903 &sk_offset, &num_found);
b3b4aa74 1904 btrfs_release_path(path);
ac8e9819
CM
1905 if (ret || num_found >= sk->nr_items)
1906 break;
1907
1908 }
1909 ret = 0;
1910err:
1911 sk->nr_items = num_found;
1912 btrfs_free_path(path);
1913 return ret;
1914}
1915
1916static noinline int btrfs_ioctl_tree_search(struct file *file,
1917 void __user *argp)
1918{
1919 struct btrfs_ioctl_search_args *args;
1920 struct inode *inode;
1921 int ret;
1922
1923 if (!capable(CAP_SYS_ADMIN))
1924 return -EPERM;
1925
2354d08f
JL
1926 args = memdup_user(argp, sizeof(*args));
1927 if (IS_ERR(args))
1928 return PTR_ERR(args);
ac8e9819 1929
496ad9aa 1930 inode = file_inode(file);
ac8e9819
CM
1931 ret = search_ioctl(inode, args);
1932 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1933 ret = -EFAULT;
1934 kfree(args);
1935 return ret;
1936}
1937
98d377a0 1938/*
ac8e9819
CM
1939 * Search INODE_REFs to identify path name of 'dirid' directory
1940 * in a 'tree_id' tree. and sets path name to 'name'.
1941 */
98d377a0
TH
1942static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1943 u64 tree_id, u64 dirid, char *name)
1944{
1945 struct btrfs_root *root;
1946 struct btrfs_key key;
ac8e9819 1947 char *ptr;
98d377a0
TH
1948 int ret = -1;
1949 int slot;
1950 int len;
1951 int total_len = 0;
1952 struct btrfs_inode_ref *iref;
1953 struct extent_buffer *l;
1954 struct btrfs_path *path;
1955
1956 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1957 name[0]='\0';
1958 return 0;
1959 }
1960
1961 path = btrfs_alloc_path();
1962 if (!path)
1963 return -ENOMEM;
1964
ac8e9819 1965 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
98d377a0
TH
1966
1967 key.objectid = tree_id;
1968 key.type = BTRFS_ROOT_ITEM_KEY;
1969 key.offset = (u64)-1;
1970 root = btrfs_read_fs_root_no_name(info, &key);
1971 if (IS_ERR(root)) {
1972 printk(KERN_ERR "could not find root %llu\n", tree_id);
8ad6fcab
CM
1973 ret = -ENOENT;
1974 goto out;
98d377a0
TH
1975 }
1976
1977 key.objectid = dirid;
1978 key.type = BTRFS_INODE_REF_KEY;
8ad6fcab 1979 key.offset = (u64)-1;
98d377a0
TH
1980
1981 while(1) {
1982 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1983 if (ret < 0)
1984 goto out;
1985
1986 l = path->nodes[0];
1987 slot = path->slots[0];
8ad6fcab
CM
1988 if (ret > 0 && slot > 0)
1989 slot--;
98d377a0
TH
1990 btrfs_item_key_to_cpu(l, &key, slot);
1991
1992 if (ret > 0 && (key.objectid != dirid ||
ac8e9819
CM
1993 key.type != BTRFS_INODE_REF_KEY)) {
1994 ret = -ENOENT;
98d377a0 1995 goto out;
ac8e9819 1996 }
98d377a0
TH
1997
1998 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1999 len = btrfs_inode_ref_name_len(l, iref);
2000 ptr -= len + 1;
2001 total_len += len + 1;
ac8e9819 2002 if (ptr < name)
98d377a0
TH
2003 goto out;
2004
2005 *(ptr + len) = '/';
2006 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
2007
2008 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2009 break;
2010
b3b4aa74 2011 btrfs_release_path(path);
98d377a0 2012 key.objectid = key.offset;
8ad6fcab 2013 key.offset = (u64)-1;
98d377a0 2014 dirid = key.objectid;
98d377a0 2015 }
ac8e9819 2016 if (ptr < name)
98d377a0 2017 goto out;
77906a50 2018 memmove(name, ptr, total_len);
98d377a0
TH
2019 name[total_len]='\0';
2020 ret = 0;
2021out:
2022 btrfs_free_path(path);
ac8e9819
CM
2023 return ret;
2024}
2025
2026static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2027 void __user *argp)
2028{
2029 struct btrfs_ioctl_ino_lookup_args *args;
2030 struct inode *inode;
2031 int ret;
2032
2033 if (!capable(CAP_SYS_ADMIN))
2034 return -EPERM;
2035
2354d08f
JL
2036 args = memdup_user(argp, sizeof(*args));
2037 if (IS_ERR(args))
2038 return PTR_ERR(args);
c2b96929 2039
496ad9aa 2040 inode = file_inode(file);
ac8e9819 2041
1b53ac4d
CM
2042 if (args->treeid == 0)
2043 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2044
ac8e9819
CM
2045 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2046 args->treeid, args->objectid,
2047 args->name);
2048
2049 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2050 ret = -EFAULT;
2051
2052 kfree(args);
98d377a0
TH
2053 return ret;
2054}
2055
76dda93c
YZ
2056static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2057 void __user *arg)
2058{
2059 struct dentry *parent = fdentry(file);
2060 struct dentry *dentry;
2061 struct inode *dir = parent->d_inode;
2062 struct inode *inode;
2063 struct btrfs_root *root = BTRFS_I(dir)->root;
2064 struct btrfs_root *dest = NULL;
2065 struct btrfs_ioctl_vol_args *vol_args;
2066 struct btrfs_trans_handle *trans;
c58aaad2
MX
2067 struct btrfs_block_rsv block_rsv;
2068 u64 qgroup_reserved;
76dda93c
YZ
2069 int namelen;
2070 int ret;
2071 int err = 0;
2072
76dda93c
YZ
2073 vol_args = memdup_user(arg, sizeof(*vol_args));
2074 if (IS_ERR(vol_args))
2075 return PTR_ERR(vol_args);
2076
2077 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2078 namelen = strlen(vol_args->name);
2079 if (strchr(vol_args->name, '/') ||
2080 strncmp(vol_args->name, "..", namelen) == 0) {
2081 err = -EINVAL;
2082 goto out;
2083 }
2084
a561be71 2085 err = mnt_want_write_file(file);
76dda93c
YZ
2086 if (err)
2087 goto out;
2088
2089 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
2090 dentry = lookup_one_len(vol_args->name, parent, namelen);
2091 if (IS_ERR(dentry)) {
2092 err = PTR_ERR(dentry);
2093 goto out_unlock_dir;
2094 }
2095
2096 if (!dentry->d_inode) {
2097 err = -ENOENT;
2098 goto out_dput;
2099 }
2100
2101 inode = dentry->d_inode;
4260f7c7
SW
2102 dest = BTRFS_I(inode)->root;
2103 if (!capable(CAP_SYS_ADMIN)){
2104 /*
2105 * Regular user. Only allow this with a special mount
2106 * option, when the user has write+exec access to the
2107 * subvol root, and when rmdir(2) would have been
2108 * allowed.
2109 *
2110 * Note that this is _not_ check that the subvol is
2111 * empty or doesn't contain data that we wouldn't
2112 * otherwise be able to delete.
2113 *
2114 * Users who want to delete empty subvols should try
2115 * rmdir(2).
2116 */
2117 err = -EPERM;
2118 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2119 goto out_dput;
2120
2121 /*
2122 * Do not allow deletion if the parent dir is the same
2123 * as the dir to be deleted. That means the ioctl
2124 * must be called on the dentry referencing the root
2125 * of the subvol, not a random directory contained
2126 * within it.
2127 */
2128 err = -EINVAL;
2129 if (root == dest)
2130 goto out_dput;
2131
2132 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2133 if (err)
2134 goto out_dput;
4260f7c7
SW
2135 }
2136
5c39da5b
MX
2137 /* check if subvolume may be deleted by a user */
2138 err = btrfs_may_delete(dir, dentry, 1);
2139 if (err)
2140 goto out_dput;
2141
33345d01 2142 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
76dda93c
YZ
2143 err = -EINVAL;
2144 goto out_dput;
2145 }
2146
76dda93c
YZ
2147 mutex_lock(&inode->i_mutex);
2148 err = d_invalidate(dentry);
2149 if (err)
2150 goto out_unlock;
2151
2152 down_write(&root->fs_info->subvol_sem);
2153
2154 err = may_destroy_subvol(dest);
2155 if (err)
2156 goto out_up_write;
2157
c58aaad2
MX
2158 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2159 /*
2160 * One for dir inode, two for dir entries, two for root
2161 * ref/backref.
2162 */
2163 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
2164 5, &qgroup_reserved);
2165 if (err)
2166 goto out_up_write;
2167
a22285a6
YZ
2168 trans = btrfs_start_transaction(root, 0);
2169 if (IS_ERR(trans)) {
2170 err = PTR_ERR(trans);
c58aaad2 2171 goto out_release;
a22285a6 2172 }
c58aaad2
MX
2173 trans->block_rsv = &block_rsv;
2174 trans->bytes_reserved = block_rsv.size;
a22285a6 2175
76dda93c
YZ
2176 ret = btrfs_unlink_subvol(trans, root, dir,
2177 dest->root_key.objectid,
2178 dentry->d_name.name,
2179 dentry->d_name.len);
79787eaa
JM
2180 if (ret) {
2181 err = ret;
2182 btrfs_abort_transaction(trans, root, ret);
2183 goto out_end_trans;
2184 }
76dda93c
YZ
2185
2186 btrfs_record_root_in_trans(trans, dest);
2187
2188 memset(&dest->root_item.drop_progress, 0,
2189 sizeof(dest->root_item.drop_progress));
2190 dest->root_item.drop_level = 0;
2191 btrfs_set_root_refs(&dest->root_item, 0);
2192
d68fc57b
YZ
2193 if (!xchg(&dest->orphan_item_inserted, 1)) {
2194 ret = btrfs_insert_orphan_item(trans,
2195 root->fs_info->tree_root,
2196 dest->root_key.objectid);
79787eaa
JM
2197 if (ret) {
2198 btrfs_abort_transaction(trans, root, ret);
2199 err = ret;
2200 goto out_end_trans;
2201 }
d68fc57b 2202 }
79787eaa 2203out_end_trans:
c58aaad2
MX
2204 trans->block_rsv = NULL;
2205 trans->bytes_reserved = 0;
531cb13f 2206 ret = btrfs_end_transaction(trans, root);
79787eaa
JM
2207 if (ret && !err)
2208 err = ret;
76dda93c 2209 inode->i_flags |= S_DEAD;
c58aaad2
MX
2210out_release:
2211 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
76dda93c
YZ
2212out_up_write:
2213 up_write(&root->fs_info->subvol_sem);
2214out_unlock:
2215 mutex_unlock(&inode->i_mutex);
2216 if (!err) {
efefb143 2217 shrink_dcache_sb(root->fs_info->sb);
76dda93c
YZ
2218 btrfs_invalidate_inodes(dest);
2219 d_delete(dentry);
fa6ac876
LB
2220
2221 /* the last ref */
2222 if (dest->cache_inode) {
2223 iput(dest->cache_inode);
2224 dest->cache_inode = NULL;
2225 }
76dda93c
YZ
2226 }
2227out_dput:
2228 dput(dentry);
2229out_unlock_dir:
2230 mutex_unlock(&dir->i_mutex);
2a79f17e 2231 mnt_drop_write_file(file);
76dda93c
YZ
2232out:
2233 kfree(vol_args);
2234 return err;
2235}
2236
1e701a32 2237static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
f46b5a66 2238{
496ad9aa 2239 struct inode *inode = file_inode(file);
f46b5a66 2240 struct btrfs_root *root = BTRFS_I(inode)->root;
1e701a32 2241 struct btrfs_ioctl_defrag_range_args *range;
c146afad
YZ
2242 int ret;
2243
25122d15
ID
2244 ret = mnt_want_write_file(file);
2245 if (ret)
2246 return ret;
b83cc969 2247
5ac00add
SB
2248 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2249 1)) {
2250 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
25122d15 2251 mnt_drop_write_file(file);
2c0c9da0 2252 return -EINVAL;
5ac00add 2253 }
25122d15
ID
2254
2255 if (btrfs_root_readonly(root)) {
2256 ret = -EROFS;
2257 goto out;
5ac00add 2258 }
f46b5a66
CH
2259
2260 switch (inode->i_mode & S_IFMT) {
2261 case S_IFDIR:
e441d54d
CM
2262 if (!capable(CAP_SYS_ADMIN)) {
2263 ret = -EPERM;
2264 goto out;
2265 }
de78b51a 2266 ret = btrfs_defrag_root(root);
8929ecfa
YZ
2267 if (ret)
2268 goto out;
de78b51a 2269 ret = btrfs_defrag_root(root->fs_info->extent_root);
f46b5a66
CH
2270 break;
2271 case S_IFREG:
e441d54d
CM
2272 if (!(file->f_mode & FMODE_WRITE)) {
2273 ret = -EINVAL;
2274 goto out;
2275 }
1e701a32
CM
2276
2277 range = kzalloc(sizeof(*range), GFP_KERNEL);
2278 if (!range) {
2279 ret = -ENOMEM;
2280 goto out;
2281 }
2282
2283 if (argp) {
2284 if (copy_from_user(range, argp,
2285 sizeof(*range))) {
2286 ret = -EFAULT;
2287 kfree(range);
683be16e 2288 goto out;
1e701a32
CM
2289 }
2290 /* compression requires us to start the IO */
2291 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2292 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2293 range->extent_thresh = (u32)-1;
2294 }
2295 } else {
2296 /* the rest are all set to zero by kzalloc */
2297 range->len = (u64)-1;
2298 }
496ad9aa 2299 ret = btrfs_defrag_file(file_inode(file), file,
4cb5300b
CM
2300 range, 0, 0);
2301 if (ret > 0)
2302 ret = 0;
1e701a32 2303 kfree(range);
f46b5a66 2304 break;
8929ecfa
YZ
2305 default:
2306 ret = -EINVAL;
f46b5a66 2307 }
e441d54d 2308out:
5ac00add 2309 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
25122d15 2310 mnt_drop_write_file(file);
e441d54d 2311 return ret;
f46b5a66
CH
2312}
2313
b2950863 2314static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
f46b5a66
CH
2315{
2316 struct btrfs_ioctl_vol_args *vol_args;
2317 int ret;
2318
e441d54d
CM
2319 if (!capable(CAP_SYS_ADMIN))
2320 return -EPERM;
2321
5ac00add
SB
2322 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2323 1)) {
2324 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2c0c9da0 2325 return -EINVAL;
c9e9f97b
ID
2326 }
2327
5ac00add 2328 mutex_lock(&root->fs_info->volume_mutex);
dae7b665 2329 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID
2330 if (IS_ERR(vol_args)) {
2331 ret = PTR_ERR(vol_args);
2332 goto out;
2333 }
f46b5a66 2334
5516e595 2335 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66
CH
2336 ret = btrfs_init_new_device(root, vol_args->name);
2337
f46b5a66 2338 kfree(vol_args);
c9e9f97b
ID
2339out:
2340 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 2341 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
f46b5a66
CH
2342 return ret;
2343}
2344
da24927b 2345static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
f46b5a66 2346{
496ad9aa 2347 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
f46b5a66
CH
2348 struct btrfs_ioctl_vol_args *vol_args;
2349 int ret;
2350
e441d54d
CM
2351 if (!capable(CAP_SYS_ADMIN))
2352 return -EPERM;
2353
da24927b
MX
2354 ret = mnt_want_write_file(file);
2355 if (ret)
2356 return ret;
c146afad 2357
5ac00add
SB
2358 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2359 1)) {
2360 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
da24927b 2361 mnt_drop_write_file(file);
2c0c9da0 2362 return -EINVAL;
c9e9f97b
ID
2363 }
2364
5ac00add 2365 mutex_lock(&root->fs_info->volume_mutex);
dae7b665 2366 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID
2367 if (IS_ERR(vol_args)) {
2368 ret = PTR_ERR(vol_args);
2369 goto out;
2370 }
f46b5a66 2371
5516e595 2372 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66
CH
2373 ret = btrfs_rm_device(root, vol_args->name);
2374
f46b5a66 2375 kfree(vol_args);
c9e9f97b
ID
2376out:
2377 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 2378 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
4ac20c70 2379 mnt_drop_write_file(file);
f46b5a66
CH
2380 return ret;
2381}
2382
475f6387
JS
2383static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2384{
027ed2f0 2385 struct btrfs_ioctl_fs_info_args *fi_args;
475f6387
JS
2386 struct btrfs_device *device;
2387 struct btrfs_device *next;
2388 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
027ed2f0 2389 int ret = 0;
475f6387
JS
2390
2391 if (!capable(CAP_SYS_ADMIN))
2392 return -EPERM;
2393
027ed2f0
LZ
2394 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2395 if (!fi_args)
2396 return -ENOMEM;
2397
2398 fi_args->num_devices = fs_devices->num_devices;
2399 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
475f6387
JS
2400
2401 mutex_lock(&fs_devices->device_list_mutex);
2402 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
027ed2f0
LZ
2403 if (device->devid > fi_args->max_id)
2404 fi_args->max_id = device->devid;
475f6387
JS
2405 }
2406 mutex_unlock(&fs_devices->device_list_mutex);
2407
027ed2f0
LZ
2408 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2409 ret = -EFAULT;
475f6387 2410
027ed2f0
LZ
2411 kfree(fi_args);
2412 return ret;
475f6387
JS
2413}
2414
2415static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2416{
2417 struct btrfs_ioctl_dev_info_args *di_args;
2418 struct btrfs_device *dev;
2419 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2420 int ret = 0;
2421 char *s_uuid = NULL;
2422 char empty_uuid[BTRFS_UUID_SIZE] = {0};
2423
2424 if (!capable(CAP_SYS_ADMIN))
2425 return -EPERM;
2426
2427 di_args = memdup_user(arg, sizeof(*di_args));
2428 if (IS_ERR(di_args))
2429 return PTR_ERR(di_args);
2430
2431 if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2432 s_uuid = di_args->uuid;
2433
2434 mutex_lock(&fs_devices->device_list_mutex);
aa1b8cd4 2435 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
475f6387
JS
2436 mutex_unlock(&fs_devices->device_list_mutex);
2437
2438 if (!dev) {
2439 ret = -ENODEV;
2440 goto out;
2441 }
2442
2443 di_args->devid = dev->devid;
2444 di_args->bytes_used = dev->bytes_used;
2445 di_args->total_bytes = dev->total_bytes;
2446 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
a27202fb 2447 if (dev->name) {
606686ee
JB
2448 struct rcu_string *name;
2449
2450 rcu_read_lock();
2451 name = rcu_dereference(dev->name);
2452 strncpy(di_args->path, name->str, sizeof(di_args->path));
2453 rcu_read_unlock();
a27202fb
JM
2454 di_args->path[sizeof(di_args->path) - 1] = 0;
2455 } else {
99ba55ad 2456 di_args->path[0] = '\0';
a27202fb 2457 }
475f6387
JS
2458
2459out:
2460 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2461 ret = -EFAULT;
2462
2463 kfree(di_args);
2464 return ret;
2465}
2466
76dda93c
YZ
2467static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2468 u64 off, u64 olen, u64 destoff)
f46b5a66 2469{
496ad9aa 2470 struct inode *inode = file_inode(file);
f46b5a66 2471 struct btrfs_root *root = BTRFS_I(inode)->root;
2903ff01 2472 struct fd src_file;
f46b5a66
CH
2473 struct inode *src;
2474 struct btrfs_trans_handle *trans;
f46b5a66 2475 struct btrfs_path *path;
f46b5a66 2476 struct extent_buffer *leaf;
ae01a0ab
YZ
2477 char *buf;
2478 struct btrfs_key key;
f46b5a66
CH
2479 u32 nritems;
2480 int slot;
ae01a0ab 2481 int ret;
c5c9cd4d
SW
2482 u64 len = olen;
2483 u64 bs = root->fs_info->sb->s_blocksize;
d20f7043 2484
c5c9cd4d
SW
2485 /*
2486 * TODO:
2487 * - split compressed inline extents. annoying: we need to
2488 * decompress into destination's address_space (the file offset
2489 * may change, so source mapping won't do), then recompress (or
2490 * otherwise reinsert) a subrange.
2491 * - allow ranges within the same file to be cloned (provided
2492 * they don't overlap)?
2493 */
2494
e441d54d 2495 /* the destination must be opened for writing */
2ebc3464 2496 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
e441d54d
CM
2497 return -EINVAL;
2498
b83cc969
LZ
2499 if (btrfs_root_readonly(root))
2500 return -EROFS;
2501
a561be71 2502 ret = mnt_want_write_file(file);
c146afad
YZ
2503 if (ret)
2504 return ret;
2505
2903ff01
AV
2506 src_file = fdget(srcfd);
2507 if (!src_file.file) {
ab67b7c1
YZ
2508 ret = -EBADF;
2509 goto out_drop_write;
2510 }
5dc64164 2511
362a20c5 2512 ret = -EXDEV;
2903ff01 2513 if (src_file.file->f_path.mnt != file->f_path.mnt)
362a20c5
DS
2514 goto out_fput;
2515
496ad9aa 2516 src = file_inode(src_file.file);
f46b5a66 2517
c5c9cd4d
SW
2518 ret = -EINVAL;
2519 if (src == inode)
2520 goto out_fput;
2521
5dc64164 2522 /* the src must be open for reading */
2903ff01 2523 if (!(src_file.file->f_mode & FMODE_READ))
5dc64164
DR
2524 goto out_fput;
2525
0e7b824c
LZ
2526 /* don't make the dst file partly checksummed */
2527 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2528 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2529 goto out_fput;
2530
ae01a0ab
YZ
2531 ret = -EISDIR;
2532 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2533 goto out_fput;
2534
f46b5a66 2535 ret = -EXDEV;
362a20c5 2536 if (src->i_sb != inode->i_sb)
ae01a0ab
YZ
2537 goto out_fput;
2538
2539 ret = -ENOMEM;
2540 buf = vmalloc(btrfs_level_size(root, 0));
2541 if (!buf)
2542 goto out_fput;
2543
2544 path = btrfs_alloc_path();
2545 if (!path) {
2546 vfree(buf);
f46b5a66 2547 goto out_fput;
ae01a0ab
YZ
2548 }
2549 path->reada = 2;
f46b5a66
CH
2550
2551 if (inode < src) {
fccdae43
SW
2552 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2553 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
f46b5a66 2554 } else {
fccdae43
SW
2555 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2556 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
f46b5a66
CH
2557 }
2558
c5c9cd4d
SW
2559 /* determine range to clone */
2560 ret = -EINVAL;
2ebc3464 2561 if (off + len > src->i_size || off + len < off)
f46b5a66 2562 goto out_unlock;
c5c9cd4d
SW
2563 if (len == 0)
2564 olen = len = src->i_size - off;
2565 /* if we extend to eof, continue to block boundary */
2566 if (off + len == src->i_size)
2a6b8dae 2567 len = ALIGN(src->i_size, bs) - off;
c5c9cd4d
SW
2568
2569 /* verify the end result is block aligned */
2a6b8dae
LZ
2570 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2571 !IS_ALIGNED(destoff, bs))
c5c9cd4d
SW
2572 goto out_unlock;
2573
d525e8ab
LZ
2574 if (destoff > inode->i_size) {
2575 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2576 if (ret)
2577 goto out_unlock;
2578 }
2579
71ef0786
LZ
2580 /* truncate page cache pages from target inode range */
2581 truncate_inode_pages_range(&inode->i_data, destoff,
2582 PAGE_CACHE_ALIGN(destoff + len) - 1);
2583
f46b5a66
CH
2584 /* do any pending delalloc/csum calc on src, one way or
2585 another, and lock file content */
2586 while (1) {
31840ae1 2587 struct btrfs_ordered_extent *ordered;
aa42ffd9
LB
2588 lock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
2589 ordered = btrfs_lookup_first_ordered_extent(src, off + len - 1);
9a019196 2590 if (!ordered &&
aa42ffd9
LB
2591 !test_range_bit(&BTRFS_I(src)->io_tree, off, off + len - 1,
2592 EXTENT_DELALLOC, 0, NULL))
f46b5a66 2593 break;
aa42ffd9 2594 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
ae01a0ab
YZ
2595 if (ordered)
2596 btrfs_put_ordered_extent(ordered);
9a019196 2597 btrfs_wait_ordered_range(src, off, len);
f46b5a66
CH
2598 }
2599
c5c9cd4d 2600 /* clone data */
33345d01 2601 key.objectid = btrfs_ino(src);
ae01a0ab
YZ
2602 key.type = BTRFS_EXTENT_DATA_KEY;
2603 key.offset = 0;
f46b5a66
CH
2604
2605 while (1) {
2606 /*
2607 * note the key will change type as we walk through the
2608 * tree.
2609 */
362a20c5
DS
2610 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
2611 0, 0);
f46b5a66
CH
2612 if (ret < 0)
2613 goto out;
2614
ae01a0ab
YZ
2615 nritems = btrfs_header_nritems(path->nodes[0]);
2616 if (path->slots[0] >= nritems) {
362a20c5 2617 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
f46b5a66
CH
2618 if (ret < 0)
2619 goto out;
2620 if (ret > 0)
2621 break;
ae01a0ab 2622 nritems = btrfs_header_nritems(path->nodes[0]);
f46b5a66
CH
2623 }
2624 leaf = path->nodes[0];
2625 slot = path->slots[0];
f46b5a66 2626
ae01a0ab 2627 btrfs_item_key_to_cpu(leaf, &key, slot);
d20f7043 2628 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
33345d01 2629 key.objectid != btrfs_ino(src))
f46b5a66
CH
2630 break;
2631
c5c9cd4d
SW
2632 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2633 struct btrfs_file_extent_item *extent;
2634 int type;
31840ae1
ZY
2635 u32 size;
2636 struct btrfs_key new_key;
c5c9cd4d
SW
2637 u64 disko = 0, diskl = 0;
2638 u64 datao = 0, datal = 0;
2639 u8 comp;
b5384d48 2640 u64 endoff;
31840ae1
ZY
2641
2642 size = btrfs_item_size_nr(leaf, slot);
2643 read_extent_buffer(leaf, buf,
2644 btrfs_item_ptr_offset(leaf, slot),
2645 size);
c5c9cd4d
SW
2646
2647 extent = btrfs_item_ptr(leaf, slot,
2648 struct btrfs_file_extent_item);
2649 comp = btrfs_file_extent_compression(leaf, extent);
2650 type = btrfs_file_extent_type(leaf, extent);
c8a894d7
CM
2651 if (type == BTRFS_FILE_EXTENT_REG ||
2652 type == BTRFS_FILE_EXTENT_PREALLOC) {
d397712b
CM
2653 disko = btrfs_file_extent_disk_bytenr(leaf,
2654 extent);
2655 diskl = btrfs_file_extent_disk_num_bytes(leaf,
2656 extent);
c5c9cd4d 2657 datao = btrfs_file_extent_offset(leaf, extent);
d397712b
CM
2658 datal = btrfs_file_extent_num_bytes(leaf,
2659 extent);
c5c9cd4d
SW
2660 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2661 /* take upper bound, may be compressed */
2662 datal = btrfs_file_extent_ram_bytes(leaf,
2663 extent);
2664 }
b3b4aa74 2665 btrfs_release_path(path);
31840ae1 2666
050006a7 2667 if (key.offset + datal <= off ||
aa42ffd9 2668 key.offset >= off + len - 1)
c5c9cd4d
SW
2669 goto next;
2670
31840ae1 2671 memcpy(&new_key, &key, sizeof(new_key));
33345d01 2672 new_key.objectid = btrfs_ino(inode);
4d728ec7
LZ
2673 if (off <= key.offset)
2674 new_key.offset = key.offset + destoff - off;
2675 else
2676 new_key.offset = destoff;
31840ae1 2677
b6f3409b
SW
2678 /*
2679 * 1 - adjusting old extent (we may have to split it)
2680 * 1 - add new extent
2681 * 1 - inode update
2682 */
2683 trans = btrfs_start_transaction(root, 3);
a22285a6
YZ
2684 if (IS_ERR(trans)) {
2685 ret = PTR_ERR(trans);
2686 goto out;
2687 }
2688
c8a894d7
CM
2689 if (type == BTRFS_FILE_EXTENT_REG ||
2690 type == BTRFS_FILE_EXTENT_PREALLOC) {
d72c0842
LZ
2691 /*
2692 * a | --- range to clone ---| b
2693 * | ------------- extent ------------- |
2694 */
2695
2696 /* substract range b */
2697 if (key.offset + datal > off + len)
2698 datal = off + len - key.offset;
2699
2700 /* substract range a */
a22285a6
YZ
2701 if (off > key.offset) {
2702 datao += off - key.offset;
2703 datal -= off - key.offset;
2704 }
2705
5dc562c5 2706 ret = btrfs_drop_extents(trans, root, inode,
a22285a6
YZ
2707 new_key.offset,
2708 new_key.offset + datal,
2671485d 2709 1);
79787eaa
JM
2710 if (ret) {
2711 btrfs_abort_transaction(trans, root,
2712 ret);
2713 btrfs_end_transaction(trans, root);
2714 goto out;
2715 }
a22285a6 2716
c5c9cd4d
SW
2717 ret = btrfs_insert_empty_item(trans, root, path,
2718 &new_key, size);
79787eaa
JM
2719 if (ret) {
2720 btrfs_abort_transaction(trans, root,
2721 ret);
2722 btrfs_end_transaction(trans, root);
2723 goto out;
2724 }
c5c9cd4d
SW
2725
2726 leaf = path->nodes[0];
2727 slot = path->slots[0];
2728 write_extent_buffer(leaf, buf,
31840ae1
ZY
2729 btrfs_item_ptr_offset(leaf, slot),
2730 size);
ae01a0ab 2731
c5c9cd4d 2732 extent = btrfs_item_ptr(leaf, slot,
f46b5a66 2733 struct btrfs_file_extent_item);
c5c9cd4d 2734
c5c9cd4d
SW
2735 /* disko == 0 means it's a hole */
2736 if (!disko)
2737 datao = 0;
c5c9cd4d
SW
2738
2739 btrfs_set_file_extent_offset(leaf, extent,
2740 datao);
2741 btrfs_set_file_extent_num_bytes(leaf, extent,
2742 datal);
2743 if (disko) {
2744 inode_add_bytes(inode, datal);
ae01a0ab 2745 ret = btrfs_inc_extent_ref(trans, root,
5d4f98a2
YZ
2746 disko, diskl, 0,
2747 root->root_key.objectid,
33345d01 2748 btrfs_ino(inode),
66d7e7f0
AJ
2749 new_key.offset - datao,
2750 0);
79787eaa
JM
2751 if (ret) {
2752 btrfs_abort_transaction(trans,
2753 root,
2754 ret);
2755 btrfs_end_transaction(trans,
2756 root);
2757 goto out;
2758
2759 }
f46b5a66 2760 }
c5c9cd4d
SW
2761 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2762 u64 skip = 0;
2763 u64 trim = 0;
2764 if (off > key.offset) {
2765 skip = off - key.offset;
2766 new_key.offset += skip;
2767 }
d397712b 2768
aa42ffd9
LB
2769 if (key.offset + datal > off + len)
2770 trim = key.offset + datal - (off + len);
d397712b 2771
c5c9cd4d 2772 if (comp && (skip || trim)) {
c5c9cd4d 2773 ret = -EINVAL;
a22285a6 2774 btrfs_end_transaction(trans, root);
c5c9cd4d
SW
2775 goto out;
2776 }
2777 size -= skip + trim;
2778 datal -= skip + trim;
a22285a6 2779
5dc562c5 2780 ret = btrfs_drop_extents(trans, root, inode,
a22285a6
YZ
2781 new_key.offset,
2782 new_key.offset + datal,
2671485d 2783 1);
79787eaa
JM
2784 if (ret) {
2785 btrfs_abort_transaction(trans, root,
2786 ret);
2787 btrfs_end_transaction(trans, root);
2788 goto out;
2789 }
a22285a6 2790
c5c9cd4d
SW
2791 ret = btrfs_insert_empty_item(trans, root, path,
2792 &new_key, size);
79787eaa
JM
2793 if (ret) {
2794 btrfs_abort_transaction(trans, root,
2795 ret);
2796 btrfs_end_transaction(trans, root);
2797 goto out;
2798 }
c5c9cd4d
SW
2799
2800 if (skip) {
d397712b
CM
2801 u32 start =
2802 btrfs_file_extent_calc_inline_size(0);
c5c9cd4d
SW
2803 memmove(buf+start, buf+start+skip,
2804 datal);
2805 }
2806
2807 leaf = path->nodes[0];
2808 slot = path->slots[0];
2809 write_extent_buffer(leaf, buf,
2810 btrfs_item_ptr_offset(leaf, slot),
2811 size);
2812 inode_add_bytes(inode, datal);
f46b5a66 2813 }
c5c9cd4d
SW
2814
2815 btrfs_mark_buffer_dirty(leaf);
b3b4aa74 2816 btrfs_release_path(path);
c5c9cd4d 2817
0c4d2d95 2818 inode_inc_iversion(inode);
a22285a6 2819 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
b5384d48
SW
2820
2821 /*
2822 * we round up to the block size at eof when
2823 * determining which extents to clone above,
2824 * but shouldn't round up the file size
2825 */
2826 endoff = new_key.offset + datal;
5f3888ff
LZ
2827 if (endoff > destoff+olen)
2828 endoff = destoff+olen;
b5384d48
SW
2829 if (endoff > inode->i_size)
2830 btrfs_i_size_write(inode, endoff);
2831
a22285a6 2832 ret = btrfs_update_inode(trans, root, inode);
79787eaa
JM
2833 if (ret) {
2834 btrfs_abort_transaction(trans, root, ret);
2835 btrfs_end_transaction(trans, root);
2836 goto out;
2837 }
2838 ret = btrfs_end_transaction(trans, root);
a22285a6 2839 }
d397712b 2840next:
b3b4aa74 2841 btrfs_release_path(path);
f46b5a66 2842 key.offset++;
f46b5a66 2843 }
f46b5a66
CH
2844 ret = 0;
2845out:
b3b4aa74 2846 btrfs_release_path(path);
aa42ffd9 2847 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
f46b5a66
CH
2848out_unlock:
2849 mutex_unlock(&src->i_mutex);
2850 mutex_unlock(&inode->i_mutex);
ae01a0ab
YZ
2851 vfree(buf);
2852 btrfs_free_path(path);
f46b5a66 2853out_fput:
2903ff01 2854 fdput(src_file);
ab67b7c1 2855out_drop_write:
2a79f17e 2856 mnt_drop_write_file(file);
f46b5a66
CH
2857 return ret;
2858}
2859
7a865e8a 2860static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
c5c9cd4d
SW
2861{
2862 struct btrfs_ioctl_clone_range_args args;
2863
7a865e8a 2864 if (copy_from_user(&args, argp, sizeof(args)))
c5c9cd4d
SW
2865 return -EFAULT;
2866 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2867 args.src_length, args.dest_offset);
2868}
2869
f46b5a66
CH
2870/*
2871 * there are many ways the trans_start and trans_end ioctls can lead
2872 * to deadlocks. They should only be used by applications that
2873 * basically own the machine, and have a very in depth understanding
2874 * of all the possible deadlocks and enospc problems.
2875 */
b2950863 2876static long btrfs_ioctl_trans_start(struct file *file)
f46b5a66 2877{
496ad9aa 2878 struct inode *inode = file_inode(file);
f46b5a66
CH
2879 struct btrfs_root *root = BTRFS_I(inode)->root;
2880 struct btrfs_trans_handle *trans;
1ab86aed 2881 int ret;
f46b5a66 2882
1ab86aed 2883 ret = -EPERM;
df5b5520 2884 if (!capable(CAP_SYS_ADMIN))
1ab86aed 2885 goto out;
df5b5520 2886
1ab86aed
SW
2887 ret = -EINPROGRESS;
2888 if (file->private_data)
f46b5a66 2889 goto out;
9ca9ee09 2890
b83cc969
LZ
2891 ret = -EROFS;
2892 if (btrfs_root_readonly(root))
2893 goto out;
2894
a561be71 2895 ret = mnt_want_write_file(file);
c146afad
YZ
2896 if (ret)
2897 goto out;
2898
a4abeea4 2899 atomic_inc(&root->fs_info->open_ioctl_trans);
9ca9ee09 2900
1ab86aed 2901 ret = -ENOMEM;
7a7eaa40 2902 trans = btrfs_start_ioctl_transaction(root);
abd30bb0 2903 if (IS_ERR(trans))
1ab86aed
SW
2904 goto out_drop;
2905
2906 file->private_data = trans;
2907 return 0;
2908
2909out_drop:
a4abeea4 2910 atomic_dec(&root->fs_info->open_ioctl_trans);
2a79f17e 2911 mnt_drop_write_file(file);
f46b5a66 2912out:
f46b5a66
CH
2913 return ret;
2914}
2915
6ef5ed0d
JB
2916static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2917{
496ad9aa 2918 struct inode *inode = file_inode(file);
6ef5ed0d
JB
2919 struct btrfs_root *root = BTRFS_I(inode)->root;
2920 struct btrfs_root *new_root;
2921 struct btrfs_dir_item *di;
2922 struct btrfs_trans_handle *trans;
2923 struct btrfs_path *path;
2924 struct btrfs_key location;
2925 struct btrfs_disk_key disk_key;
6ef5ed0d
JB
2926 u64 objectid = 0;
2927 u64 dir_id;
3c04ce01 2928 int ret;
6ef5ed0d
JB
2929
2930 if (!capable(CAP_SYS_ADMIN))
2931 return -EPERM;
2932
3c04ce01
MX
2933 ret = mnt_want_write_file(file);
2934 if (ret)
2935 return ret;
2936
2937 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
2938 ret = -EFAULT;
2939 goto out;
2940 }
6ef5ed0d
JB
2941
2942 if (!objectid)
2943 objectid = root->root_key.objectid;
2944
2945 location.objectid = objectid;
2946 location.type = BTRFS_ROOT_ITEM_KEY;
2947 location.offset = (u64)-1;
2948
2949 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
3c04ce01
MX
2950 if (IS_ERR(new_root)) {
2951 ret = PTR_ERR(new_root);
2952 goto out;
2953 }
6ef5ed0d 2954
3c04ce01
MX
2955 if (btrfs_root_refs(&new_root->root_item) == 0) {
2956 ret = -ENOENT;
2957 goto out;
2958 }
6ef5ed0d
JB
2959
2960 path = btrfs_alloc_path();
3c04ce01
MX
2961 if (!path) {
2962 ret = -ENOMEM;
2963 goto out;
2964 }
6ef5ed0d
JB
2965 path->leave_spinning = 1;
2966
2967 trans = btrfs_start_transaction(root, 1);
98d5dc13 2968 if (IS_ERR(trans)) {
6ef5ed0d 2969 btrfs_free_path(path);
3c04ce01
MX
2970 ret = PTR_ERR(trans);
2971 goto out;
6ef5ed0d
JB
2972 }
2973
6c41761f 2974 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
6ef5ed0d
JB
2975 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2976 dir_id, "default", 7, 1);
cf1e99a4 2977 if (IS_ERR_OR_NULL(di)) {
6ef5ed0d
JB
2978 btrfs_free_path(path);
2979 btrfs_end_transaction(trans, root);
2980 printk(KERN_ERR "Umm, you don't have the default dir item, "
2981 "this isn't going to work\n");
3c04ce01
MX
2982 ret = -ENOENT;
2983 goto out;
6ef5ed0d
JB
2984 }
2985
2986 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2987 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2988 btrfs_mark_buffer_dirty(path->nodes[0]);
2989 btrfs_free_path(path);
2990
2b0ce2c2 2991 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
6ef5ed0d 2992 btrfs_end_transaction(trans, root);
3c04ce01
MX
2993out:
2994 mnt_drop_write_file(file);
2995 return ret;
6ef5ed0d
JB
2996}
2997
5af3e8cc
SB
2998void btrfs_get_block_group_info(struct list_head *groups_list,
2999 struct btrfs_ioctl_space_info *space)
bf5fc093
JB
3000{
3001 struct btrfs_block_group_cache *block_group;
3002
3003 space->total_bytes = 0;
3004 space->used_bytes = 0;
3005 space->flags = 0;
3006 list_for_each_entry(block_group, groups_list, list) {
3007 space->flags = block_group->flags;
3008 space->total_bytes += block_group->key.offset;
3009 space->used_bytes +=
3010 btrfs_block_group_used(&block_group->item);
3011 }
3012}
3013
1406e432
JB
3014long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
3015{
3016 struct btrfs_ioctl_space_args space_args;
3017 struct btrfs_ioctl_space_info space;
3018 struct btrfs_ioctl_space_info *dest;
7fde62bf 3019 struct btrfs_ioctl_space_info *dest_orig;
13f2696f 3020 struct btrfs_ioctl_space_info __user *user_dest;
1406e432 3021 struct btrfs_space_info *info;
bf5fc093
JB
3022 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3023 BTRFS_BLOCK_GROUP_SYSTEM,
3024 BTRFS_BLOCK_GROUP_METADATA,
3025 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3026 int num_types = 4;
7fde62bf 3027 int alloc_size;
1406e432 3028 int ret = 0;
51788b1b 3029 u64 slot_count = 0;
bf5fc093 3030 int i, c;
1406e432
JB
3031
3032 if (copy_from_user(&space_args,
3033 (struct btrfs_ioctl_space_args __user *)arg,
3034 sizeof(space_args)))
3035 return -EFAULT;
3036
bf5fc093
JB
3037 for (i = 0; i < num_types; i++) {
3038 struct btrfs_space_info *tmp;
3039
3040 info = NULL;
3041 rcu_read_lock();
3042 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3043 list) {
3044 if (tmp->flags == types[i]) {
3045 info = tmp;
3046 break;
3047 }
3048 }
3049 rcu_read_unlock();
3050
3051 if (!info)
3052 continue;
3053
3054 down_read(&info->groups_sem);
3055 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3056 if (!list_empty(&info->block_groups[c]))
3057 slot_count++;
3058 }
3059 up_read(&info->groups_sem);
3060 }
7fde62bf
CM
3061
3062 /* space_slots == 0 means they are asking for a count */
3063 if (space_args.space_slots == 0) {
3064 space_args.total_spaces = slot_count;
3065 goto out;
3066 }
bf5fc093 3067
51788b1b 3068 slot_count = min_t(u64, space_args.space_slots, slot_count);
bf5fc093 3069
7fde62bf 3070 alloc_size = sizeof(*dest) * slot_count;
bf5fc093 3071
7fde62bf
CM
3072 /* we generally have at most 6 or so space infos, one for each raid
3073 * level. So, a whole page should be more than enough for everyone
3074 */
3075 if (alloc_size > PAGE_CACHE_SIZE)
3076 return -ENOMEM;
3077
1406e432 3078 space_args.total_spaces = 0;
7fde62bf
CM
3079 dest = kmalloc(alloc_size, GFP_NOFS);
3080 if (!dest)
3081 return -ENOMEM;
3082 dest_orig = dest;
1406e432 3083
7fde62bf 3084 /* now we have a buffer to copy into */
bf5fc093
JB
3085 for (i = 0; i < num_types; i++) {
3086 struct btrfs_space_info *tmp;
3087
51788b1b
DR
3088 if (!slot_count)
3089 break;
3090
bf5fc093
JB
3091 info = NULL;
3092 rcu_read_lock();
3093 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3094 list) {
3095 if (tmp->flags == types[i]) {
3096 info = tmp;
3097 break;
3098 }
3099 }
3100 rcu_read_unlock();
7fde62bf 3101
bf5fc093
JB
3102 if (!info)
3103 continue;
3104 down_read(&info->groups_sem);
3105 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3106 if (!list_empty(&info->block_groups[c])) {
5af3e8cc
SB
3107 btrfs_get_block_group_info(
3108 &info->block_groups[c], &space);
bf5fc093
JB
3109 memcpy(dest, &space, sizeof(space));
3110 dest++;
3111 space_args.total_spaces++;
51788b1b 3112 slot_count--;
bf5fc093 3113 }
51788b1b
DR
3114 if (!slot_count)
3115 break;
bf5fc093
JB
3116 }
3117 up_read(&info->groups_sem);
1406e432 3118 }
1406e432 3119
2eec6c81 3120 user_dest = (struct btrfs_ioctl_space_info __user *)
7fde62bf
CM
3121 (arg + sizeof(struct btrfs_ioctl_space_args));
3122
3123 if (copy_to_user(user_dest, dest_orig, alloc_size))
3124 ret = -EFAULT;
3125
3126 kfree(dest_orig);
3127out:
3128 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
1406e432
JB
3129 ret = -EFAULT;
3130
3131 return ret;
3132}
3133
f46b5a66
CH
3134/*
3135 * there are many ways the trans_start and trans_end ioctls can lead
3136 * to deadlocks. They should only be used by applications that
3137 * basically own the machine, and have a very in depth understanding
3138 * of all the possible deadlocks and enospc problems.
3139 */
3140long btrfs_ioctl_trans_end(struct file *file)
3141{
496ad9aa 3142 struct inode *inode = file_inode(file);
f46b5a66
CH
3143 struct btrfs_root *root = BTRFS_I(inode)->root;
3144 struct btrfs_trans_handle *trans;
f46b5a66 3145
f46b5a66 3146 trans = file->private_data;
1ab86aed
SW
3147 if (!trans)
3148 return -EINVAL;
b214107e 3149 file->private_data = NULL;
9ca9ee09 3150
1ab86aed
SW
3151 btrfs_end_transaction(trans, root);
3152
a4abeea4 3153 atomic_dec(&root->fs_info->open_ioctl_trans);
9ca9ee09 3154
2a79f17e 3155 mnt_drop_write_file(file);
1ab86aed 3156 return 0;
f46b5a66
CH
3157}
3158
9a8c28be
MX
3159static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3160 void __user *argp)
46204592 3161{
46204592
SW
3162 struct btrfs_trans_handle *trans;
3163 u64 transid;
db5b493a 3164 int ret;
46204592 3165
d4edf39b 3166 trans = btrfs_attach_transaction_barrier(root);
ff7c1d33
MX
3167 if (IS_ERR(trans)) {
3168 if (PTR_ERR(trans) != -ENOENT)
3169 return PTR_ERR(trans);
3170
3171 /* No running transaction, don't bother */
3172 transid = root->fs_info->last_trans_committed;
3173 goto out;
3174 }
46204592 3175 transid = trans->transid;
db5b493a 3176 ret = btrfs_commit_transaction_async(trans, root, 0);
8b2b2d3c
TI
3177 if (ret) {
3178 btrfs_end_transaction(trans, root);
db5b493a 3179 return ret;
8b2b2d3c 3180 }
ff7c1d33 3181out:
46204592
SW
3182 if (argp)
3183 if (copy_to_user(argp, &transid, sizeof(transid)))
3184 return -EFAULT;
3185 return 0;
3186}
3187
9a8c28be
MX
3188static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
3189 void __user *argp)
46204592 3190{
46204592
SW
3191 u64 transid;
3192
3193 if (argp) {
3194 if (copy_from_user(&transid, argp, sizeof(transid)))
3195 return -EFAULT;
3196 } else {
3197 transid = 0; /* current trans */
3198 }
3199 return btrfs_wait_for_commit(root, transid);
3200}
3201
b8e95489 3202static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
475f6387 3203{
496ad9aa 3204 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
475f6387 3205 struct btrfs_ioctl_scrub_args *sa;
b8e95489 3206 int ret;
475f6387
JS
3207
3208 if (!capable(CAP_SYS_ADMIN))
3209 return -EPERM;
3210
3211 sa = memdup_user(arg, sizeof(*sa));
3212 if (IS_ERR(sa))
3213 return PTR_ERR(sa);
3214
b8e95489
MX
3215 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3216 ret = mnt_want_write_file(file);
3217 if (ret)
3218 goto out;
3219 }
3220
aa1b8cd4 3221 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
63a212ab
SB
3222 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3223 0);
475f6387
JS
3224
3225 if (copy_to_user(arg, sa, sizeof(*sa)))
3226 ret = -EFAULT;
3227
b8e95489
MX
3228 if (!(sa->flags & BTRFS_SCRUB_READONLY))
3229 mnt_drop_write_file(file);
3230out:
475f6387
JS
3231 kfree(sa);
3232 return ret;
3233}
3234
3235static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3236{
3237 if (!capable(CAP_SYS_ADMIN))
3238 return -EPERM;
3239
aa1b8cd4 3240 return btrfs_scrub_cancel(root->fs_info);
475f6387
JS
3241}
3242
3243static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3244 void __user *arg)
3245{
3246 struct btrfs_ioctl_scrub_args *sa;
3247 int ret;
3248
3249 if (!capable(CAP_SYS_ADMIN))
3250 return -EPERM;
3251
3252 sa = memdup_user(arg, sizeof(*sa));
3253 if (IS_ERR(sa))
3254 return PTR_ERR(sa);
3255
3256 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3257
3258 if (copy_to_user(arg, sa, sizeof(*sa)))
3259 ret = -EFAULT;
3260
3261 kfree(sa);
3262 return ret;
3263}
3264
c11d2c23 3265static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
b27f7c0c 3266 void __user *arg)
c11d2c23
SB
3267{
3268 struct btrfs_ioctl_get_dev_stats *sa;
3269 int ret;
3270
c11d2c23
SB
3271 sa = memdup_user(arg, sizeof(*sa));
3272 if (IS_ERR(sa))
3273 return PTR_ERR(sa);
3274
b27f7c0c
DS
3275 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3276 kfree(sa);
3277 return -EPERM;
3278 }
3279
3280 ret = btrfs_get_dev_stats(root, sa);
c11d2c23
SB
3281
3282 if (copy_to_user(arg, sa, sizeof(*sa)))
3283 ret = -EFAULT;
3284
3285 kfree(sa);
3286 return ret;
3287}
3288
3f6bcfbd
SB
3289static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
3290{
3291 struct btrfs_ioctl_dev_replace_args *p;
3292 int ret;
3293
3294 if (!capable(CAP_SYS_ADMIN))
3295 return -EPERM;
3296
3297 p = memdup_user(arg, sizeof(*p));
3298 if (IS_ERR(p))
3299 return PTR_ERR(p);
3300
3301 switch (p->cmd) {
3302 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
3303 if (atomic_xchg(
3304 &root->fs_info->mutually_exclusive_operation_running,
3305 1)) {
3306 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3307 ret = -EINPROGRESS;
3308 } else {
3309 ret = btrfs_dev_replace_start(root, p);
3310 atomic_set(
3311 &root->fs_info->mutually_exclusive_operation_running,
3312 0);
3313 }
3314 break;
3315 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3316 btrfs_dev_replace_status(root->fs_info, p);
3317 ret = 0;
3318 break;
3319 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3320 ret = btrfs_dev_replace_cancel(root->fs_info, p);
3321 break;
3322 default:
3323 ret = -EINVAL;
3324 break;
3325 }
3326
3327 if (copy_to_user(arg, p, sizeof(*p)))
3328 ret = -EFAULT;
3329
3330 kfree(p);
3331 return ret;
3332}
3333
d7728c96
JS
3334static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3335{
3336 int ret = 0;
3337 int i;
740c3d22 3338 u64 rel_ptr;
d7728c96 3339 int size;
806468f8 3340 struct btrfs_ioctl_ino_path_args *ipa = NULL;
d7728c96
JS
3341 struct inode_fs_paths *ipath = NULL;
3342 struct btrfs_path *path;
3343
82b22ac8 3344 if (!capable(CAP_DAC_READ_SEARCH))
d7728c96
JS
3345 return -EPERM;
3346
3347 path = btrfs_alloc_path();
3348 if (!path) {
3349 ret = -ENOMEM;
3350 goto out;
3351 }
3352
3353 ipa = memdup_user(arg, sizeof(*ipa));
3354 if (IS_ERR(ipa)) {
3355 ret = PTR_ERR(ipa);
3356 ipa = NULL;
3357 goto out;
3358 }
3359
3360 size = min_t(u32, ipa->size, 4096);
3361 ipath = init_ipath(size, root, path);
3362 if (IS_ERR(ipath)) {
3363 ret = PTR_ERR(ipath);
3364 ipath = NULL;
3365 goto out;
3366 }
3367
3368 ret = paths_from_inode(ipa->inum, ipath);
3369 if (ret < 0)
3370 goto out;
3371
3372 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
745c4d8e
JM
3373 rel_ptr = ipath->fspath->val[i] -
3374 (u64)(unsigned long)ipath->fspath->val;
740c3d22 3375 ipath->fspath->val[i] = rel_ptr;
d7728c96
JS
3376 }
3377
745c4d8e
JM
3378 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3379 (void *)(unsigned long)ipath->fspath, size);
d7728c96
JS
3380 if (ret) {
3381 ret = -EFAULT;
3382 goto out;
3383 }
3384
3385out:
3386 btrfs_free_path(path);
3387 free_ipath(ipath);
3388 kfree(ipa);
3389
3390 return ret;
3391}
3392
3393static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3394{
3395 struct btrfs_data_container *inodes = ctx;
3396 const size_t c = 3 * sizeof(u64);
3397
3398 if (inodes->bytes_left >= c) {
3399 inodes->bytes_left -= c;
3400 inodes->val[inodes->elem_cnt] = inum;
3401 inodes->val[inodes->elem_cnt + 1] = offset;
3402 inodes->val[inodes->elem_cnt + 2] = root;
3403 inodes->elem_cnt += 3;
3404 } else {
3405 inodes->bytes_missing += c - inodes->bytes_left;
3406 inodes->bytes_left = 0;
3407 inodes->elem_missed += 3;
3408 }
3409
3410 return 0;
3411}
3412
3413static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3414 void __user *arg)
3415{
3416 int ret = 0;
3417 int size;
d7728c96
JS
3418 struct btrfs_ioctl_logical_ino_args *loi;
3419 struct btrfs_data_container *inodes = NULL;
3420 struct btrfs_path *path = NULL;
d7728c96
JS
3421
3422 if (!capable(CAP_SYS_ADMIN))
3423 return -EPERM;
3424
3425 loi = memdup_user(arg, sizeof(*loi));
3426 if (IS_ERR(loi)) {
3427 ret = PTR_ERR(loi);
3428 loi = NULL;
3429 goto out;
3430 }
3431
3432 path = btrfs_alloc_path();
3433 if (!path) {
3434 ret = -ENOMEM;
3435 goto out;
3436 }
3437
425d17a2 3438 size = min_t(u32, loi->size, 64 * 1024);
d7728c96
JS
3439 inodes = init_data_container(size);
3440 if (IS_ERR(inodes)) {
3441 ret = PTR_ERR(inodes);
3442 inodes = NULL;
3443 goto out;
3444 }
3445
df031f07
LB
3446 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
3447 build_ino_list, inodes);
3448 if (ret == -EINVAL)
d7728c96
JS
3449 ret = -ENOENT;
3450 if (ret < 0)
3451 goto out;
3452
745c4d8e
JM
3453 ret = copy_to_user((void *)(unsigned long)loi->inodes,
3454 (void *)(unsigned long)inodes, size);
d7728c96
JS
3455 if (ret)
3456 ret = -EFAULT;
3457
3458out:
3459 btrfs_free_path(path);
425d17a2 3460 vfree(inodes);
d7728c96
JS
3461 kfree(loi);
3462
3463 return ret;
3464}
3465
19a39dce 3466void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
c9e9f97b
ID
3467 struct btrfs_ioctl_balance_args *bargs)
3468{
3469 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3470
3471 bargs->flags = bctl->flags;
3472
837d5b6e
ID
3473 if (atomic_read(&fs_info->balance_running))
3474 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3475 if (atomic_read(&fs_info->balance_pause_req))
3476 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
a7e99c69
ID
3477 if (atomic_read(&fs_info->balance_cancel_req))
3478 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
837d5b6e 3479
c9e9f97b
ID
3480 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3481 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3482 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
19a39dce
ID
3483
3484 if (lock) {
3485 spin_lock(&fs_info->balance_lock);
3486 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3487 spin_unlock(&fs_info->balance_lock);
3488 } else {
3489 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3490 }
c9e9f97b
ID
3491}
3492
9ba1f6e4 3493static long btrfs_ioctl_balance(struct file *file, void __user *arg)
c9e9f97b 3494{
496ad9aa 3495 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
c9e9f97b
ID
3496 struct btrfs_fs_info *fs_info = root->fs_info;
3497 struct btrfs_ioctl_balance_args *bargs;
3498 struct btrfs_balance_control *bctl;
ed0fb78f 3499 bool need_unlock; /* for mut. excl. ops lock */
c9e9f97b
ID
3500 int ret;
3501
3502 if (!capable(CAP_SYS_ADMIN))
3503 return -EPERM;
3504
e54bfa31 3505 ret = mnt_want_write_file(file);
9ba1f6e4
LB
3506 if (ret)
3507 return ret;
3508
ed0fb78f
ID
3509again:
3510 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
3511 mutex_lock(&fs_info->volume_mutex);
3512 mutex_lock(&fs_info->balance_mutex);
3513 need_unlock = true;
3514 goto locked;
3515 }
3516
3517 /*
3518 * mut. excl. ops lock is locked. Three possibilites:
3519 * (1) some other op is running
3520 * (2) balance is running
3521 * (3) balance is paused -- special case (think resume)
3522 */
c9e9f97b 3523 mutex_lock(&fs_info->balance_mutex);
ed0fb78f
ID
3524 if (fs_info->balance_ctl) {
3525 /* this is either (2) or (3) */
3526 if (!atomic_read(&fs_info->balance_running)) {
3527 mutex_unlock(&fs_info->balance_mutex);
3528 if (!mutex_trylock(&fs_info->volume_mutex))
3529 goto again;
3530 mutex_lock(&fs_info->balance_mutex);
3531
3532 if (fs_info->balance_ctl &&
3533 !atomic_read(&fs_info->balance_running)) {
3534 /* this is (3) */
3535 need_unlock = false;
3536 goto locked;
3537 }
3538
3539 mutex_unlock(&fs_info->balance_mutex);
3540 mutex_unlock(&fs_info->volume_mutex);
3541 goto again;
3542 } else {
3543 /* this is (2) */
3544 mutex_unlock(&fs_info->balance_mutex);
3545 ret = -EINPROGRESS;
3546 goto out;
3547 }
3548 } else {
3549 /* this is (1) */
3550 mutex_unlock(&fs_info->balance_mutex);
3551 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3552 ret = -EINVAL;
3553 goto out;
3554 }
3555
3556locked:
3557 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
c9e9f97b
ID
3558
3559 if (arg) {
3560 bargs = memdup_user(arg, sizeof(*bargs));
3561 if (IS_ERR(bargs)) {
3562 ret = PTR_ERR(bargs);
ed0fb78f 3563 goto out_unlock;
c9e9f97b 3564 }
de322263
ID
3565
3566 if (bargs->flags & BTRFS_BALANCE_RESUME) {
3567 if (!fs_info->balance_ctl) {
3568 ret = -ENOTCONN;
3569 goto out_bargs;
3570 }
3571
3572 bctl = fs_info->balance_ctl;
3573 spin_lock(&fs_info->balance_lock);
3574 bctl->flags |= BTRFS_BALANCE_RESUME;
3575 spin_unlock(&fs_info->balance_lock);
3576
3577 goto do_balance;
3578 }
c9e9f97b
ID
3579 } else {
3580 bargs = NULL;
3581 }
3582
ed0fb78f 3583 if (fs_info->balance_ctl) {
837d5b6e
ID
3584 ret = -EINPROGRESS;
3585 goto out_bargs;
3586 }
3587
c9e9f97b
ID
3588 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3589 if (!bctl) {
3590 ret = -ENOMEM;
3591 goto out_bargs;
3592 }
3593
3594 bctl->fs_info = fs_info;
3595 if (arg) {
3596 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3597 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3598 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3599
3600 bctl->flags = bargs->flags;
f43ffb60
ID
3601 } else {
3602 /* balance everything - no filters */
3603 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
c9e9f97b
ID
3604 }
3605
de322263 3606do_balance:
c9e9f97b 3607 /*
ed0fb78f
ID
3608 * Ownership of bctl and mutually_exclusive_operation_running
3609 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
3610 * or, if restriper was paused all the way until unmount, in
3611 * free_fs_info. mutually_exclusive_operation_running is
3612 * cleared in __cancel_balance.
c9e9f97b 3613 */
ed0fb78f
ID
3614 need_unlock = false;
3615
3616 ret = btrfs_balance(bctl, bargs);
3617
c9e9f97b
ID
3618 if (arg) {
3619 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3620 ret = -EFAULT;
3621 }
3622
3623out_bargs:
3624 kfree(bargs);
ed0fb78f 3625out_unlock:
c9e9f97b
ID
3626 mutex_unlock(&fs_info->balance_mutex);
3627 mutex_unlock(&fs_info->volume_mutex);
ed0fb78f
ID
3628 if (need_unlock)
3629 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
3630out:
e54bfa31 3631 mnt_drop_write_file(file);
c9e9f97b
ID
3632 return ret;
3633}
3634
837d5b6e
ID
3635static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
3636{
3637 if (!capable(CAP_SYS_ADMIN))
3638 return -EPERM;
3639
3640 switch (cmd) {
3641 case BTRFS_BALANCE_CTL_PAUSE:
3642 return btrfs_pause_balance(root->fs_info);
a7e99c69
ID
3643 case BTRFS_BALANCE_CTL_CANCEL:
3644 return btrfs_cancel_balance(root->fs_info);
837d5b6e
ID
3645 }
3646
3647 return -EINVAL;
3648}
3649
19a39dce
ID
3650static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
3651 void __user *arg)
3652{
3653 struct btrfs_fs_info *fs_info = root->fs_info;
3654 struct btrfs_ioctl_balance_args *bargs;
3655 int ret = 0;
3656
3657 if (!capable(CAP_SYS_ADMIN))
3658 return -EPERM;
3659
3660 mutex_lock(&fs_info->balance_mutex);
3661 if (!fs_info->balance_ctl) {
3662 ret = -ENOTCONN;
3663 goto out;
3664 }
3665
3666 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
3667 if (!bargs) {
3668 ret = -ENOMEM;
3669 goto out;
3670 }
3671
3672 update_ioctl_balance_args(fs_info, 1, bargs);
3673
3674 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3675 ret = -EFAULT;
3676
3677 kfree(bargs);
3678out:
3679 mutex_unlock(&fs_info->balance_mutex);
3680 return ret;
3681}
3682
905b0dda 3683static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
5d13a37b 3684{
496ad9aa 3685 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ
3686 struct btrfs_ioctl_quota_ctl_args *sa;
3687 struct btrfs_trans_handle *trans = NULL;
3688 int ret;
3689 int err;
3690
3691 if (!capable(CAP_SYS_ADMIN))
3692 return -EPERM;
3693
905b0dda
MX
3694 ret = mnt_want_write_file(file);
3695 if (ret)
3696 return ret;
5d13a37b
AJ
3697
3698 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX
3699 if (IS_ERR(sa)) {
3700 ret = PTR_ERR(sa);
3701 goto drop_write;
3702 }
5d13a37b
AJ
3703
3704 if (sa->cmd != BTRFS_QUOTA_CTL_RESCAN) {
3705 trans = btrfs_start_transaction(root, 2);
3706 if (IS_ERR(trans)) {
3707 ret = PTR_ERR(trans);
3708 goto out;
3709 }
3710 }
3711
3712 switch (sa->cmd) {
3713 case BTRFS_QUOTA_CTL_ENABLE:
3714 ret = btrfs_quota_enable(trans, root->fs_info);
3715 break;
3716 case BTRFS_QUOTA_CTL_DISABLE:
3717 ret = btrfs_quota_disable(trans, root->fs_info);
3718 break;
3719 case BTRFS_QUOTA_CTL_RESCAN:
3720 ret = btrfs_quota_rescan(root->fs_info);
3721 break;
3722 default:
3723 ret = -EINVAL;
3724 break;
3725 }
3726
3727 if (copy_to_user(arg, sa, sizeof(*sa)))
3728 ret = -EFAULT;
3729
3730 if (trans) {
3731 err = btrfs_commit_transaction(trans, root);
3732 if (err && !ret)
3733 ret = err;
3734 }
5d13a37b
AJ
3735out:
3736 kfree(sa);
905b0dda
MX
3737drop_write:
3738 mnt_drop_write_file(file);
5d13a37b
AJ
3739 return ret;
3740}
3741
905b0dda 3742static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
5d13a37b 3743{
496ad9aa 3744 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ
3745 struct btrfs_ioctl_qgroup_assign_args *sa;
3746 struct btrfs_trans_handle *trans;
3747 int ret;
3748 int err;
3749
3750 if (!capable(CAP_SYS_ADMIN))
3751 return -EPERM;
3752
905b0dda
MX
3753 ret = mnt_want_write_file(file);
3754 if (ret)
3755 return ret;
5d13a37b
AJ
3756
3757 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX
3758 if (IS_ERR(sa)) {
3759 ret = PTR_ERR(sa);
3760 goto drop_write;
3761 }
5d13a37b
AJ
3762
3763 trans = btrfs_join_transaction(root);
3764 if (IS_ERR(trans)) {
3765 ret = PTR_ERR(trans);
3766 goto out;
3767 }
3768
3769 /* FIXME: check if the IDs really exist */
3770 if (sa->assign) {
3771 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
3772 sa->src, sa->dst);
3773 } else {
3774 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
3775 sa->src, sa->dst);
3776 }
3777
3778 err = btrfs_end_transaction(trans, root);
3779 if (err && !ret)
3780 ret = err;
3781
3782out:
3783 kfree(sa);
905b0dda
MX
3784drop_write:
3785 mnt_drop_write_file(file);
5d13a37b
AJ
3786 return ret;
3787}
3788
905b0dda 3789static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
5d13a37b 3790{
496ad9aa 3791 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ
3792 struct btrfs_ioctl_qgroup_create_args *sa;
3793 struct btrfs_trans_handle *trans;
3794 int ret;
3795 int err;
3796
3797 if (!capable(CAP_SYS_ADMIN))
3798 return -EPERM;
3799
905b0dda
MX
3800 ret = mnt_want_write_file(file);
3801 if (ret)
3802 return ret;
5d13a37b
AJ
3803
3804 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX
3805 if (IS_ERR(sa)) {
3806 ret = PTR_ERR(sa);
3807 goto drop_write;
3808 }
5d13a37b 3809
d86e56cf
MX
3810 if (!sa->qgroupid) {
3811 ret = -EINVAL;
3812 goto out;
3813 }
3814
5d13a37b
AJ
3815 trans = btrfs_join_transaction(root);
3816 if (IS_ERR(trans)) {
3817 ret = PTR_ERR(trans);
3818 goto out;
3819 }
3820
3821 /* FIXME: check if the IDs really exist */
3822 if (sa->create) {
3823 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
3824 NULL);
3825 } else {
3826 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
3827 }
3828
3829 err = btrfs_end_transaction(trans, root);
3830 if (err && !ret)
3831 ret = err;
3832
3833out:
3834 kfree(sa);
905b0dda
MX
3835drop_write:
3836 mnt_drop_write_file(file);
5d13a37b
AJ
3837 return ret;
3838}
3839
905b0dda 3840static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
5d13a37b 3841{
496ad9aa 3842 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ
3843 struct btrfs_ioctl_qgroup_limit_args *sa;
3844 struct btrfs_trans_handle *trans;
3845 int ret;
3846 int err;
3847 u64 qgroupid;
3848
3849 if (!capable(CAP_SYS_ADMIN))
3850 return -EPERM;
3851
905b0dda
MX
3852 ret = mnt_want_write_file(file);
3853 if (ret)
3854 return ret;
5d13a37b
AJ
3855
3856 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX
3857 if (IS_ERR(sa)) {
3858 ret = PTR_ERR(sa);
3859 goto drop_write;
3860 }
5d13a37b
AJ
3861
3862 trans = btrfs_join_transaction(root);
3863 if (IS_ERR(trans)) {
3864 ret = PTR_ERR(trans);
3865 goto out;
3866 }
3867
3868 qgroupid = sa->qgroupid;
3869 if (!qgroupid) {
3870 /* take the current subvol as qgroup */
3871 qgroupid = root->root_key.objectid;
3872 }
3873
3874 /* FIXME: check if the IDs really exist */
3875 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
3876
3877 err = btrfs_end_transaction(trans, root);
3878 if (err && !ret)
3879 ret = err;
3880
3881out:
3882 kfree(sa);
905b0dda
MX
3883drop_write:
3884 mnt_drop_write_file(file);
5d13a37b
AJ
3885 return ret;
3886}
3887
8ea05e3a
AB
3888static long btrfs_ioctl_set_received_subvol(struct file *file,
3889 void __user *arg)
3890{
3891 struct btrfs_ioctl_received_subvol_args *sa = NULL;
496ad9aa 3892 struct inode *inode = file_inode(file);
8ea05e3a
AB
3893 struct btrfs_root *root = BTRFS_I(inode)->root;
3894 struct btrfs_root_item *root_item = &root->root_item;
3895 struct btrfs_trans_handle *trans;
3896 struct timespec ct = CURRENT_TIME;
3897 int ret = 0;
3898
3899 ret = mnt_want_write_file(file);
3900 if (ret < 0)
3901 return ret;
3902
3903 down_write(&root->fs_info->subvol_sem);
3904
3905 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
3906 ret = -EINVAL;
3907 goto out;
3908 }
3909
3910 if (btrfs_root_readonly(root)) {
3911 ret = -EROFS;
3912 goto out;
3913 }
3914
3915 if (!inode_owner_or_capable(inode)) {
3916 ret = -EACCES;
3917 goto out;
3918 }
3919
3920 sa = memdup_user(arg, sizeof(*sa));
3921 if (IS_ERR(sa)) {
3922 ret = PTR_ERR(sa);
3923 sa = NULL;
3924 goto out;
3925 }
3926
3927 trans = btrfs_start_transaction(root, 1);
3928 if (IS_ERR(trans)) {
3929 ret = PTR_ERR(trans);
3930 trans = NULL;
3931 goto out;
3932 }
3933
3934 sa->rtransid = trans->transid;
3935 sa->rtime.sec = ct.tv_sec;
3936 sa->rtime.nsec = ct.tv_nsec;
3937
3938 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
3939 btrfs_set_root_stransid(root_item, sa->stransid);
3940 btrfs_set_root_rtransid(root_item, sa->rtransid);
3941 root_item->stime.sec = cpu_to_le64(sa->stime.sec);
3942 root_item->stime.nsec = cpu_to_le32(sa->stime.nsec);
3943 root_item->rtime.sec = cpu_to_le64(sa->rtime.sec);
3944 root_item->rtime.nsec = cpu_to_le32(sa->rtime.nsec);
3945
3946 ret = btrfs_update_root(trans, root->fs_info->tree_root,
3947 &root->root_key, &root->root_item);
3948 if (ret < 0) {
3949 btrfs_end_transaction(trans, root);
3950 trans = NULL;
3951 goto out;
3952 } else {
3953 ret = btrfs_commit_transaction(trans, root);
3954 if (ret < 0)
3955 goto out;
3956 }
3957
3958 ret = copy_to_user(arg, sa, sizeof(*sa));
3959 if (ret)
3960 ret = -EFAULT;
3961
3962out:
3963 kfree(sa);
3964 up_write(&root->fs_info->subvol_sem);
3965 mnt_drop_write_file(file);
3966 return ret;
3967}
3968
867ab667 3969static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
3970{
3971 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3972 const char *label = root->fs_info->super_copy->label;
3973 size_t len = strnlen(label, BTRFS_LABEL_SIZE);
3974 int ret;
3975
3976 if (len == BTRFS_LABEL_SIZE) {
3977 pr_warn("btrfs: label is too long, return the first %zu bytes\n",
3978 --len);
3979 }
3980
3981 mutex_lock(&root->fs_info->volume_mutex);
3982 ret = copy_to_user(arg, label, len);
3983 mutex_unlock(&root->fs_info->volume_mutex);
3984
3985 return ret ? -EFAULT : 0;
3986}
3987
a8bfd4ab 3988static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
3989{
3990 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
3991 struct btrfs_super_block *super_block = root->fs_info->super_copy;
3992 struct btrfs_trans_handle *trans;
3993 char label[BTRFS_LABEL_SIZE];
3994 int ret;
3995
3996 if (!capable(CAP_SYS_ADMIN))
3997 return -EPERM;
3998
3999 if (copy_from_user(label, arg, sizeof(label)))
4000 return -EFAULT;
4001
4002 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
4003 pr_err("btrfs: unable to set label with more than %d bytes\n",
4004 BTRFS_LABEL_SIZE - 1);
4005 return -EINVAL;
4006 }
4007
4008 ret = mnt_want_write_file(file);
4009 if (ret)
4010 return ret;
4011
4012 mutex_lock(&root->fs_info->volume_mutex);
4013 trans = btrfs_start_transaction(root, 0);
4014 if (IS_ERR(trans)) {
4015 ret = PTR_ERR(trans);
4016 goto out_unlock;
4017 }
4018
4019 strcpy(super_block->label, label);
4020 ret = btrfs_end_transaction(trans, root);
4021
4022out_unlock:
4023 mutex_unlock(&root->fs_info->volume_mutex);
4024 mnt_drop_write_file(file);
4025 return ret;
4026}
4027
f46b5a66
CH
4028long btrfs_ioctl(struct file *file, unsigned int
4029 cmd, unsigned long arg)
4030{
496ad9aa 4031 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4bcabaa3 4032 void __user *argp = (void __user *)arg;
f46b5a66
CH
4033
4034 switch (cmd) {
6cbff00f
CH
4035 case FS_IOC_GETFLAGS:
4036 return btrfs_ioctl_getflags(file, argp);
4037 case FS_IOC_SETFLAGS:
4038 return btrfs_ioctl_setflags(file, argp);
4039 case FS_IOC_GETVERSION:
4040 return btrfs_ioctl_getversion(file, argp);
f7039b1d
LD
4041 case FITRIM:
4042 return btrfs_ioctl_fitrim(file, argp);
f46b5a66 4043 case BTRFS_IOC_SNAP_CREATE:
fa0d2b9b 4044 return btrfs_ioctl_snap_create(file, argp, 0);
fdfb1e4f 4045 case BTRFS_IOC_SNAP_CREATE_V2:
fa0d2b9b 4046 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3de4586c 4047 case BTRFS_IOC_SUBVOL_CREATE:
fa0d2b9b 4048 return btrfs_ioctl_snap_create(file, argp, 1);
6f72c7e2
AJ
4049 case BTRFS_IOC_SUBVOL_CREATE_V2:
4050 return btrfs_ioctl_snap_create_v2(file, argp, 1);
76dda93c
YZ
4051 case BTRFS_IOC_SNAP_DESTROY:
4052 return btrfs_ioctl_snap_destroy(file, argp);
0caa102d
LZ
4053 case BTRFS_IOC_SUBVOL_GETFLAGS:
4054 return btrfs_ioctl_subvol_getflags(file, argp);
4055 case BTRFS_IOC_SUBVOL_SETFLAGS:
4056 return btrfs_ioctl_subvol_setflags(file, argp);
6ef5ed0d
JB
4057 case BTRFS_IOC_DEFAULT_SUBVOL:
4058 return btrfs_ioctl_default_subvol(file, argp);
f46b5a66 4059 case BTRFS_IOC_DEFRAG:
1e701a32
CM
4060 return btrfs_ioctl_defrag(file, NULL);
4061 case BTRFS_IOC_DEFRAG_RANGE:
4062 return btrfs_ioctl_defrag(file, argp);
f46b5a66 4063 case BTRFS_IOC_RESIZE:
198605a8 4064 return btrfs_ioctl_resize(file, argp);
f46b5a66 4065 case BTRFS_IOC_ADD_DEV:
4bcabaa3 4066 return btrfs_ioctl_add_dev(root, argp);
f46b5a66 4067 case BTRFS_IOC_RM_DEV:
da24927b 4068 return btrfs_ioctl_rm_dev(file, argp);
475f6387
JS
4069 case BTRFS_IOC_FS_INFO:
4070 return btrfs_ioctl_fs_info(root, argp);
4071 case BTRFS_IOC_DEV_INFO:
4072 return btrfs_ioctl_dev_info(root, argp);
f46b5a66 4073 case BTRFS_IOC_BALANCE:
9ba1f6e4 4074 return btrfs_ioctl_balance(file, NULL);
f46b5a66 4075 case BTRFS_IOC_CLONE:
c5c9cd4d
SW
4076 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
4077 case BTRFS_IOC_CLONE_RANGE:
7a865e8a 4078 return btrfs_ioctl_clone_range(file, argp);
f46b5a66
CH
4079 case BTRFS_IOC_TRANS_START:
4080 return btrfs_ioctl_trans_start(file);
4081 case BTRFS_IOC_TRANS_END:
4082 return btrfs_ioctl_trans_end(file);
ac8e9819
CM
4083 case BTRFS_IOC_TREE_SEARCH:
4084 return btrfs_ioctl_tree_search(file, argp);
4085 case BTRFS_IOC_INO_LOOKUP:
4086 return btrfs_ioctl_ino_lookup(file, argp);
d7728c96
JS
4087 case BTRFS_IOC_INO_PATHS:
4088 return btrfs_ioctl_ino_to_path(root, argp);
4089 case BTRFS_IOC_LOGICAL_INO:
4090 return btrfs_ioctl_logical_to_ino(root, argp);
1406e432
JB
4091 case BTRFS_IOC_SPACE_INFO:
4092 return btrfs_ioctl_space_info(root, argp);
f46b5a66
CH
4093 case BTRFS_IOC_SYNC:
4094 btrfs_sync_fs(file->f_dentry->d_sb, 1);
4095 return 0;
46204592 4096 case BTRFS_IOC_START_SYNC:
9a8c28be 4097 return btrfs_ioctl_start_sync(root, argp);
46204592 4098 case BTRFS_IOC_WAIT_SYNC:
9a8c28be 4099 return btrfs_ioctl_wait_sync(root, argp);
475f6387 4100 case BTRFS_IOC_SCRUB:
b8e95489 4101 return btrfs_ioctl_scrub(file, argp);
475f6387
JS
4102 case BTRFS_IOC_SCRUB_CANCEL:
4103 return btrfs_ioctl_scrub_cancel(root, argp);
4104 case BTRFS_IOC_SCRUB_PROGRESS:
4105 return btrfs_ioctl_scrub_progress(root, argp);
c9e9f97b 4106 case BTRFS_IOC_BALANCE_V2:
9ba1f6e4 4107 return btrfs_ioctl_balance(file, argp);
837d5b6e
ID
4108 case BTRFS_IOC_BALANCE_CTL:
4109 return btrfs_ioctl_balance_ctl(root, arg);
19a39dce
ID
4110 case BTRFS_IOC_BALANCE_PROGRESS:
4111 return btrfs_ioctl_balance_progress(root, argp);
8ea05e3a
AB
4112 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
4113 return btrfs_ioctl_set_received_subvol(file, argp);
31db9f7c
AB
4114 case BTRFS_IOC_SEND:
4115 return btrfs_ioctl_send(file, argp);
c11d2c23 4116 case BTRFS_IOC_GET_DEV_STATS:
b27f7c0c 4117 return btrfs_ioctl_get_dev_stats(root, argp);
5d13a37b 4118 case BTRFS_IOC_QUOTA_CTL:
905b0dda 4119 return btrfs_ioctl_quota_ctl(file, argp);
5d13a37b 4120 case BTRFS_IOC_QGROUP_ASSIGN:
905b0dda 4121 return btrfs_ioctl_qgroup_assign(file, argp);
5d13a37b 4122 case BTRFS_IOC_QGROUP_CREATE:
905b0dda 4123 return btrfs_ioctl_qgroup_create(file, argp);
5d13a37b 4124 case BTRFS_IOC_QGROUP_LIMIT:
905b0dda 4125 return btrfs_ioctl_qgroup_limit(file, argp);
3f6bcfbd
SB
4126 case BTRFS_IOC_DEV_REPLACE:
4127 return btrfs_ioctl_dev_replace(root, argp);
867ab667 4128 case BTRFS_IOC_GET_FSLABEL:
4129 return btrfs_ioctl_get_fslabel(file, argp);
a8bfd4ab 4130 case BTRFS_IOC_SET_FSLABEL:
4131 return btrfs_ioctl_set_fslabel(file, argp);
f46b5a66
CH
4132 }
4133
4134 return -ENOTTY;
4135}
This page took 0.479261 seconds and 5 git commands to generate.