projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Btrfs: fix resize a readonly device
[linux-2.6-block.git] / fs / btrfs / ioctl.c
CommitLineData
f46b5a66
CH		 1/*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
19#include <linux/kernel.h>
20#include <linux/bio.h>
21#include <linux/buffer_head.h>
22#include <linux/file.h>
23#include <linux/fs.h>
cb8e7090 24#include <linux/fsnotify.h>
f46b5a66
CH		 25#include <linux/pagemap.h>
26#include <linux/highmem.h>
27#include <linux/time.h>
28#include <linux/init.h>
29#include <linux/string.h>
f46b5a66 30#include <linux/backing-dev.h>
cb8e7090 31#include <linux/mount.h>
f46b5a66 32#include <linux/mpage.h>
cb8e7090 33#include <linux/namei.h>
f46b5a66
CH		 34#include <linux/swap.h>
35#include <linux/writeback.h>
36#include <linux/statfs.h>
37#include <linux/compat.h>
38#include <linux/bit_spinlock.h>
cb8e7090 39#include <linux/security.h>
f46b5a66 40#include <linux/xattr.h>
7ea394f1 41#include <linux/vmalloc.h>
5a0e3ad6 42#include <linux/slab.h>
f7039b1d 43#include <linux/blkdev.h>
8ea05e3a 44#include <linux/uuid.h>
4b4e25f2 45#include "compat.h"
f46b5a66
CH		 46#include "ctree.h"
47#include "disk-io.h"
48#include "transaction.h"
49#include "btrfs_inode.h"
50#include "ioctl.h"
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{
155 struct btrfs_inode *ip = BTRFS_I(file->f_path.dentry->d_inode);
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{
180 struct inode *inode = file->f_path.dentry->d_inode;
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{
313 struct inode *inode = file->f_path.dentry->d_inode;
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
cb8e7090
CH		 366static noinline int create_subvol(struct btrfs_root *root,
367 struct dentry *dentry,
72fd032e 368 char *name, int namelen,
6f72c7e2
AJ		 369 u64 *async_transid,
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;
76dda93c 377 struct btrfs_root *new_root;
2fbe8c8a 378 struct dentry *parent = dentry->d_parent;
6a912213 379 struct inode *dir;
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;
8ea05e3a 386 uuid_le new_uuid;
f46b5a66 387
581bb050 388 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
2fbe8c8a 389 if (ret)
a22285a6 390 return ret;
6a912213
JB		 391
392 dir = parent->d_inode;
393
9ed74f2d
JB		 394 /*
395 * 1 - inode item
396 * 2 - refs
397 * 1 - root item
398 * 2 - dir items
399 */
a22285a6 400 trans = btrfs_start_transaction(root, 6);
2fbe8c8a 401 if (IS_ERR(trans))
a22285a6 402 return PTR_ERR(trans);
f46b5a66 403
6f72c7e2
AJ		 404 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid,
405 inherit ? *inherit : NULL);
406 if (ret)
407 goto fail;
408
5d4f98a2 409 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
5581a51a 410 0, objectid, NULL, 0, 0, 0);
8e8a1e31
JB		 411 if (IS_ERR(leaf)) {
412 ret = PTR_ERR(leaf);
413 goto fail;
414 }
f46b5a66 415
5d4f98a2 416 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
f46b5a66
CH		 417 btrfs_set_header_bytenr(leaf, leaf->start);
418 btrfs_set_header_generation(leaf, trans->transid);
5d4f98a2 419 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
f46b5a66
CH		 420 btrfs_set_header_owner(leaf, objectid);
421
422 write_extent_buffer(leaf, root->fs_info->fsid,
423 (unsigned long)btrfs_header_fsid(leaf),
424 BTRFS_FSID_SIZE);
5d4f98a2
YZ		 425 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
426 (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
427 BTRFS_UUID_SIZE);
f46b5a66
CH		 428 btrfs_mark_buffer_dirty(leaf);
429
8ea05e3a
AB		 430 memset(&root_item, 0, sizeof(root_item));
431
f46b5a66 432 inode_item = &root_item.inode;
f46b5a66
CH		 433 inode_item->generation = cpu_to_le64(1);
434 inode_item->size = cpu_to_le64(3);
435 inode_item->nlink = cpu_to_le32(1);
a76a3cd4 436 inode_item->nbytes = cpu_to_le64(root->leafsize);
f46b5a66
CH		 437 inode_item->mode = cpu_to_le32(S_IFDIR | 0755);
438
08fe4db1
LZ		 439 root_item.flags = 0;
440 root_item.byte_limit = 0;
441 inode_item->flags = cpu_to_le64(BTRFS_INODE_ROOT_ITEM_INIT);
442
f46b5a66 443 btrfs_set_root_bytenr(&root_item, leaf->start);
84234f3a 444 btrfs_set_root_generation(&root_item, trans->transid);
f46b5a66
CH		 445 btrfs_set_root_level(&root_item, 0);
446 btrfs_set_root_refs(&root_item, 1);
86b9f2ec 447 btrfs_set_root_used(&root_item, leaf->len);
80ff3856 448 btrfs_set_root_last_snapshot(&root_item, 0);
f46b5a66 449
8ea05e3a
AB		 450 btrfs_set_root_generation_v2(&root_item,
451 btrfs_root_generation(&root_item));
452 uuid_le_gen(&new_uuid);
453 memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
454 root_item.otime.sec = cpu_to_le64(cur_time.tv_sec);
dadd1105 455 root_item.otime.nsec = cpu_to_le32(cur_time.tv_nsec);
8ea05e3a
AB		 456 root_item.ctime = root_item.otime;
457 btrfs_set_root_ctransid(&root_item, trans->transid);
458 btrfs_set_root_otransid(&root_item, trans->transid);
f46b5a66 459
925baedd 460 btrfs_tree_unlock(leaf);
f46b5a66
CH		 461 free_extent_buffer(leaf);
462 leaf = NULL;
463
464 btrfs_set_root_dirid(&root_item, new_dirid);
465
466 key.objectid = objectid;
5d4f98a2 467 key.offset = 0;
f46b5a66
CH		 468 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
469 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
470 &root_item);
471 if (ret)
472 goto fail;
473
76dda93c
YZ		 474 key.offset = (u64)-1;
475 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
79787eaa
JM		 476 if (IS_ERR(new_root)) {
477 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
478 ret = PTR_ERR(new_root);
479 goto fail;
480 }
76dda93c
YZ		 481
482 btrfs_record_root_in_trans(trans, new_root);
483
d82a6f1d 484 ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
ce598979
MF		 485 if (ret) {
486 /* We potentially lose an unused inode item here */
79787eaa 487 btrfs_abort_transaction(trans, root, ret);
ce598979
MF		 488 goto fail;
489 }
490
f46b5a66
CH		 491 /*
492 * insert the directory item
493 */
3de4586c 494 ret = btrfs_set_inode_index(dir, &index);
79787eaa
JM		 495 if (ret) {
496 btrfs_abort_transaction(trans, root, ret);
497 goto fail;
498 }
3de4586c
CM		 499
500 ret = btrfs_insert_dir_item(trans, root,
16cdcec7 501 name, namelen, dir, &key,
3de4586c 502 BTRFS_FT_DIR, index);
79787eaa
JM		 503 if (ret) {
504 btrfs_abort_transaction(trans, root, ret);
f46b5a66 505 goto fail;
79787eaa 506 }
0660b5af 507
52c26179
YZ		 508 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
509 ret = btrfs_update_inode(trans, root, dir);
510 BUG_ON(ret);
511
0660b5af 512 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
4df27c4d 513 objectid, root->root_key.objectid,
33345d01 514 btrfs_ino(dir), index, name, namelen);
0660b5af 515
76dda93c 516 BUG_ON(ret);
f46b5a66 517
76dda93c 518 d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
f46b5a66 519fail:
72fd032e
SW		 520 if (async_transid) {
521 *async_transid = trans->transid;
522 err = btrfs_commit_transaction_async(trans, root, 1);
523 } else {
524 err = btrfs_commit_transaction(trans, root);
525 }
f46b5a66
CH		 526 if (err && !ret)
527 ret = err;
f46b5a66
CH		 528 return ret;
529}
530
72fd032e 531static int create_snapshot(struct btrfs_root *root, struct dentry *dentry,
b83cc969 532 char *name, int namelen, u64 *async_transid,
6f72c7e2 533 bool readonly, struct btrfs_qgroup_inherit **inherit)
f46b5a66 534{
2e4bfab9 535 struct inode *inode;
f46b5a66
CH		 536 struct btrfs_pending_snapshot *pending_snapshot;
537 struct btrfs_trans_handle *trans;
2e4bfab9 538 int ret;
f46b5a66
CH		 539
540 if (!root->ref_cows)
541 return -EINVAL;
542
3de4586c 543 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
a22285a6
YZ		 544 if (!pending_snapshot)
545 return -ENOMEM;
546
66d8f3dd
MX		 547 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
548 BTRFS_BLOCK_RSV_TEMP);
3de4586c 549 pending_snapshot->dentry = dentry;
f46b5a66 550 pending_snapshot->root = root;
b83cc969 551 pending_snapshot->readonly = readonly;
6f72c7e2
AJ		 552 if (inherit) {
553 pending_snapshot->inherit = *inherit;
554 *inherit = NULL; /* take responsibility to free it */
555 }
a22285a6 556
48c03c4b 557 trans = btrfs_start_transaction(root->fs_info->extent_root, 6);
a22285a6
YZ		 558 if (IS_ERR(trans)) {
559 ret = PTR_ERR(trans);
560 goto fail;
561 }
562
563 ret = btrfs_snap_reserve_metadata(trans, pending_snapshot);
564 BUG_ON(ret);
565
8351583e 566 spin_lock(&root->fs_info->trans_lock);
f46b5a66
CH		 567 list_add(&pending_snapshot->list,
568 &trans->transaction->pending_snapshots);
8351583e 569 spin_unlock(&root->fs_info->trans_lock);
72fd032e
SW		 570 if (async_transid) {
571 *async_transid = trans->transid;
572 ret = btrfs_commit_transaction_async(trans,
573 root->fs_info->extent_root, 1);
574 } else {
575 ret = btrfs_commit_transaction(trans,
576 root->fs_info->extent_root);
577 }
109f2365
LB		 578 if (ret) {
579 /* cleanup_transaction has freed this for us */
580 if (trans->aborted)
581 pending_snapshot = NULL;
c37b2b62 582 goto fail;
109f2365 583 }
a22285a6
YZ		 584
585 ret = pending_snapshot->error;
586 if (ret)
587 goto fail;
588
66b4ffd1
JB		 589 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
590 if (ret)
591 goto fail;
f46b5a66 592
2fbe8c8a 593 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
2e4bfab9
YZ		 594 if (IS_ERR(inode)) {
595 ret = PTR_ERR(inode);
596 goto fail;
597 }
598 BUG_ON(!inode);
599 d_instantiate(dentry, inode);
600 ret = 0;
601fail:
a22285a6 602 kfree(pending_snapshot);
f46b5a66
CH		 603 return ret;
604}
605
4260f7c7
SW		 606/* copy of check_sticky in fs/namei.c()
607* It's inline, so penalty for filesystems that don't use sticky bit is
608* minimal.
609*/
610static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
611{
2f2f43d3 612 kuid_t fsuid = current_fsuid();
4260f7c7
SW		 613
614 if (!(dir->i_mode & S_ISVTX))
615 return 0;
2f2f43d3 616 if (uid_eq(inode->i_uid, fsuid))
4260f7c7 617 return 0;
2f2f43d3 618 if (uid_eq(dir->i_uid, fsuid))
4260f7c7
SW		 619 return 0;
620 return !capable(CAP_FOWNER);
621}
622
623/* copy of may_delete in fs/namei.c()
624 * Check whether we can remove a link victim from directory dir, check
625 * whether the type of victim is right.
626 * 1. We can't do it if dir is read-only (done in permission())
627 * 2. We should have write and exec permissions on dir
628 * 3. We can't remove anything from append-only dir
629 * 4. We can't do anything with immutable dir (done in permission())
630 * 5. If the sticky bit on dir is set we should either
631 * a. be owner of dir, or
632 * b. be owner of victim, or
633 * c. have CAP_FOWNER capability
634 * 6. If the victim is append-only or immutable we can't do antyhing with
635 * links pointing to it.
636 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
637 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
638 * 9. We can't remove a root or mountpoint.
639 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
640 * nfs_async_unlink().
641 */
642
643static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
644{
645 int error;
646
647 if (!victim->d_inode)
648 return -ENOENT;
649
650 BUG_ON(victim->d_parent->d_inode != dir);
4fa6b5ec 651 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
4260f7c7
SW		 652
653 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
654 if (error)
655 return error;
656 if (IS_APPEND(dir))
657 return -EPERM;
658 if (btrfs_check_sticky(dir, victim->d_inode)||
659 IS_APPEND(victim->d_inode)||
660 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
661 return -EPERM;
662 if (isdir) {
663 if (!S_ISDIR(victim->d_inode->i_mode))
664 return -ENOTDIR;
665 if (IS_ROOT(victim))
666 return -EBUSY;
667 } else if (S_ISDIR(victim->d_inode->i_mode))
668 return -EISDIR;
669 if (IS_DEADDIR(dir))
670 return -ENOENT;
671 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
672 return -EBUSY;
673 return 0;
674}
675
cb8e7090
CH		 676/* copy of may_create in fs/namei.c() */
677static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
678{
679 if (child->d_inode)
680 return -EEXIST;
681 if (IS_DEADDIR(dir))
682 return -ENOENT;
683 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
684}
685
686/*
687 * Create a new subvolume below @parent. This is largely modeled after
688 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
689 * inside this filesystem so it's quite a bit simpler.
690 */
76dda93c
YZ		 691static noinline int btrfs_mksubvol(struct path *parent,
692 char *name, int namelen,
72fd032e 693 struct btrfs_root *snap_src,
6f72c7e2
AJ		 694 u64 *async_transid, bool readonly,
695 struct btrfs_qgroup_inherit **inherit)
cb8e7090 696{
76dda93c 697 struct inode *dir = parent->dentry->d_inode;
cb8e7090
CH		 698 struct dentry *dentry;
699 int error;
700
76dda93c 701 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
cb8e7090
CH		 702
703 dentry = lookup_one_len(name, parent->dentry, namelen);
704 error = PTR_ERR(dentry);
705 if (IS_ERR(dentry))
706 goto out_unlock;
707
708 error = -EEXIST;
709 if (dentry->d_inode)
710 goto out_dput;
711
76dda93c 712 error = btrfs_may_create(dir, dentry);
cb8e7090 713 if (error)
a874a63e 714 goto out_dput;
cb8e7090 715
9c52057c
CM		 716 /*
717 * even if this name doesn't exist, we may get hash collisions.
718 * check for them now when we can safely fail
719 */
720 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
721 dir->i_ino, name,
722 namelen);
723 if (error)
724 goto out_dput;
725
76dda93c
YZ		 726 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
727
728 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
729 goto out_up_read;
730
3de4586c 731 if (snap_src) {
6f72c7e2
AJ		 732 error = create_snapshot(snap_src, dentry, name, namelen,
733 async_transid, readonly, inherit);
3de4586c 734 } else {
76dda93c 735 error = create_subvol(BTRFS_I(dir)->root, dentry,
6f72c7e2 736 name, namelen, async_transid, inherit);
3de4586c 737 }
76dda93c
YZ		 738 if (!error)
739 fsnotify_mkdir(dir, dentry);
740out_up_read:
741 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
cb8e7090
CH		 742out_dput:
743 dput(dentry);
744out_unlock:
76dda93c 745 mutex_unlock(&dir->i_mutex);
cb8e7090
CH		 746 return error;
747}
748
4cb5300b
CM		 749/*
750 * When we're defragging a range, we don't want to kick it off again
751 * if it is really just waiting for delalloc to send it down.
752 * If we find a nice big extent or delalloc range for the bytes in the
753 * file you want to defrag, we return 0 to let you know to skip this
754 * part of the file
755 */
756static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
757{
758 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
759 struct extent_map *em = NULL;
760 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
761 u64 end;
762
763 read_lock(&em_tree->lock);
764 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
765 read_unlock(&em_tree->lock);
766
767 if (em) {
768 end = extent_map_end(em);
769 free_extent_map(em);
770 if (end - offset > thresh)
771 return 0;
772 }
773 /* if we already have a nice delalloc here, just stop */
774 thresh /= 2;
775 end = count_range_bits(io_tree, &offset, offset + thresh,
776 thresh, EXTENT_DELALLOC, 1);
777 if (end >= thresh)
778 return 0;
779 return 1;
780}
781
782/*
783 * helper function to walk through a file and find extents
784 * newer than a specific transid, and smaller than thresh.
785 *
786 * This is used by the defragging code to find new and small
787 * extents
788 */
789static int find_new_extents(struct btrfs_root *root,
790 struct inode *inode, u64 newer_than,
791 u64 *off, int thresh)
792{
793 struct btrfs_path *path;
794 struct btrfs_key min_key;
795 struct btrfs_key max_key;
796 struct extent_buffer *leaf;
797 struct btrfs_file_extent_item *extent;
798 int type;
799 int ret;
a4689d2b 800 u64 ino = btrfs_ino(inode);
4cb5300b
CM		 801
802 path = btrfs_alloc_path();
803 if (!path)
804 return -ENOMEM;
805
a4689d2b 806 min_key.objectid = ino;
4cb5300b
CM		 807 min_key.type = BTRFS_EXTENT_DATA_KEY;
808 min_key.offset = *off;
809
a4689d2b 810 max_key.objectid = ino;
4cb5300b
CM		 811 max_key.type = (u8)-1;
812 max_key.offset = (u64)-1;
813
814 path->keep_locks = 1;
815
816 while(1) {
817 ret = btrfs_search_forward(root, &min_key, &max_key,
818 path, 0, newer_than);
819 if (ret != 0)
820 goto none;
a4689d2b 821 if (min_key.objectid != ino)
4cb5300b
CM		 822 goto none;
823 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
824 goto none;
825
826 leaf = path->nodes[0];
827 extent = btrfs_item_ptr(leaf, path->slots[0],
828 struct btrfs_file_extent_item);
829
830 type = btrfs_file_extent_type(leaf, extent);
831 if (type == BTRFS_FILE_EXTENT_REG &&
832 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
833 check_defrag_in_cache(inode, min_key.offset, thresh)) {
834 *off = min_key.offset;
835 btrfs_free_path(path);
836 return 0;
837 }
838
839 if (min_key.offset == (u64)-1)
840 goto none;
841
842 min_key.offset++;
843 btrfs_release_path(path);
844 }
845none:
846 btrfs_free_path(path);
847 return -ENOENT;
848}
849
6c282eb4 850static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
17ce6ef8
LB		 851{
852 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
6c282eb4
LZ		 853 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
854 struct extent_map *em;
855 u64 len = PAGE_CACHE_SIZE;
17ce6ef8 856
6c282eb4
LZ		 857 /*
858 * hopefully we have this extent in the tree already, try without
859 * the full extent lock
860 */
17ce6ef8 861 read_lock(&em_tree->lock);
6c282eb4 862 em = lookup_extent_mapping(em_tree, start, len);
17ce6ef8
LB		 863 read_unlock(&em_tree->lock);
864
6c282eb4
LZ		 865 if (!em) {
866 /* get the big lock and read metadata off disk */
867 lock_extent(io_tree, start, start + len - 1);
868 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
869 unlock_extent(io_tree, start, start + len - 1);
870
871 if (IS_ERR(em))
872 return NULL;
873 }
874
875 return em;
876}
17ce6ef8 877
6c282eb4
LZ		 878static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
879{
880 struct extent_map *next;
881 bool ret = true;
882
883 /* this is the last extent */
884 if (em->start + em->len >= i_size_read(inode))
885 return false;
886
887 next = defrag_lookup_extent(inode, em->start + em->len);
888 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
889 ret = false;
890
891 free_extent_map(next);
17ce6ef8
LB		 892 return ret;
893}
894
6c282eb4 895static int should_defrag_range(struct inode *inode, u64 start, int thresh,
a43a2111
AM		 896 u64 *last_len, u64 *skip, u64 *defrag_end,
897 int compress)
940100a4 898{
6c282eb4 899 struct extent_map *em;
940100a4 900 int ret = 1;
6c282eb4 901 bool next_mergeable = true;
940100a4
CM		 902
903 /*
008873ea 904 * make sure that once we start defragging an extent, we keep on
940100a4
CM		 905 * defragging it
906 */
907 if (start < *defrag_end)
908 return 1;
909
910 *skip = 0;
911
6c282eb4
LZ		 912 em = defrag_lookup_extent(inode, start);
913 if (!em)
914 return 0;
940100a4
CM		 915
916 /* this will cover holes, and inline extents */
17ce6ef8 917 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
940100a4 918 ret = 0;
17ce6ef8
LB		 919 goto out;
920 }
921
6c282eb4 922 next_mergeable = defrag_check_next_extent(inode, em);
940100a4
CM		 923
924 /*
6c282eb4
LZ		 925 * we hit a real extent, if it is big or the next extent is not a
926 * real extent, don't bother defragging it
940100a4 927 */
a43a2111 928 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
6c282eb4 929 (em->len >= thresh || !next_mergeable))
940100a4 930 ret = 0;
17ce6ef8 931out:
940100a4
CM		 932 /*
933 * last_len ends up being a counter of how many bytes we've defragged.
934 * every time we choose not to defrag an extent, we reset *last_len
935 * so that the next tiny extent will force a defrag.
936 *
937 * The end result of this is that tiny extents before a single big
938 * extent will force at least part of that big extent to be defragged.
939 */
940 if (ret) {
940100a4
CM		 941 *defrag_end = extent_map_end(em);
942 } else {
943 *last_len = 0;
944 *skip = extent_map_end(em);
945 *defrag_end = 0;
946 }
947
948 free_extent_map(em);
949 return ret;
950}
951
4cb5300b
CM		 952/*
953 * it doesn't do much good to defrag one or two pages
954 * at a time. This pulls in a nice chunk of pages
955 * to COW and defrag.
956 *
957 * It also makes sure the delalloc code has enough
958 * dirty data to avoid making new small extents as part
959 * of the defrag
960 *
961 * It's a good idea to start RA on this range
962 * before calling this.
963 */
964static int cluster_pages_for_defrag(struct inode *inode,
965 struct page **pages,
966 unsigned long start_index,
967 int num_pages)
f46b5a66 968{
4cb5300b
CM		 969 unsigned long file_end;
970 u64 isize = i_size_read(inode);
971 u64 page_start;
972 u64 page_end;
1f12bd06 973 u64 page_cnt;
4cb5300b
CM		 974 int ret;
975 int i;
976 int i_done;
3eaa2885 977 struct btrfs_ordered_extent *ordered;
4cb5300b 978 struct extent_state *cached_state = NULL;
600a45e1 979 struct extent_io_tree *tree;
3b16a4e3 980 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
4cb5300b 981
4cb5300b 982 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1f12bd06
LB		 983 if (!isize || start_index > file_end)
984 return 0;
985
986 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
4cb5300b
CM		 987
988 ret = btrfs_delalloc_reserve_space(inode,
1f12bd06 989 page_cnt << PAGE_CACHE_SHIFT);
4cb5300b
CM		 990 if (ret)
991 return ret;
4cb5300b 992 i_done = 0;
600a45e1 993 tree = &BTRFS_I(inode)->io_tree;
4cb5300b
CM		 994
995 /* step one, lock all the pages */
1f12bd06 996 for (i = 0; i < page_cnt; i++) {
4cb5300b 997 struct page *page;
600a45e1 998again:
a94733d0 999 page = find_or_create_page(inode->i_mapping,
600a45e1 1000 start_index + i, mask);
4cb5300b
CM		 1001 if (!page)
1002 break;
1003
600a45e1
MX		 1004 page_start = page_offset(page);
1005 page_end = page_start + PAGE_CACHE_SIZE - 1;
1006 while (1) {
d0082371 1007 lock_extent(tree, page_start, page_end);
600a45e1
MX		 1008 ordered = btrfs_lookup_ordered_extent(inode,
1009 page_start);
d0082371 1010 unlock_extent(tree, page_start, page_end);
600a45e1
MX		 1011 if (!ordered)
1012 break;
1013
1014 unlock_page(page);
1015 btrfs_start_ordered_extent(inode, ordered, 1);
1016 btrfs_put_ordered_extent(ordered);
1017 lock_page(page);
1f12bd06
LB		 1018 /*
1019 * we unlocked the page above, so we need check if
1020 * it was released or not.
1021 */
1022 if (page->mapping != inode->i_mapping) {
1023 unlock_page(page);
1024 page_cache_release(page);
1025 goto again;
1026 }
600a45e1
MX		 1027 }
1028
4cb5300b
CM		 1029 if (!PageUptodate(page)) {
1030 btrfs_readpage(NULL, page);
1031 lock_page(page);
1032 if (!PageUptodate(page)) {
1033 unlock_page(page);
1034 page_cache_release(page);
1035 ret = -EIO;
1036 break;
1037 }
1038 }
600a45e1 1039
600a45e1
MX		 1040 if (page->mapping != inode->i_mapping) {
1041 unlock_page(page);
1042 page_cache_release(page);
1043 goto again;
1044 }
1045
4cb5300b
CM		 1046 pages[i] = page;
1047 i_done++;
1048 }
1049 if (!i_done || ret)
1050 goto out;
1051
1052 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1053 goto out;
1054
1055 /*
1056 * so now we have a nice long stream of locked
1057 * and up to date pages, lets wait on them
1058 */
1059 for (i = 0; i < i_done; i++)
1060 wait_on_page_writeback(pages[i]);
1061
1062 page_start = page_offset(pages[0]);
1063 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1064
1065 lock_extent_bits(&BTRFS_I(inode)->io_tree,
d0082371 1066 page_start, page_end - 1, 0, &cached_state);
4cb5300b
CM		 1067 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1068 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
9e8a4a8b
LB		 1069 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1070 &cached_state, GFP_NOFS);
4cb5300b 1071
1f12bd06 1072 if (i_done != page_cnt) {
9e0baf60
JB		 1073 spin_lock(&BTRFS_I(inode)->lock);
1074 BTRFS_I(inode)->outstanding_extents++;
1075 spin_unlock(&BTRFS_I(inode)->lock);
4cb5300b 1076 btrfs_delalloc_release_space(inode,
1f12bd06 1077 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1078 }
1079
1080
9e8a4a8b
LB		 1081 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1082 &cached_state, GFP_NOFS);
4cb5300b
CM		 1083
1084 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1085 page_start, page_end - 1, &cached_state,
1086 GFP_NOFS);
1087
1088 for (i = 0; i < i_done; i++) {
1089 clear_page_dirty_for_io(pages[i]);
1090 ClearPageChecked(pages[i]);
1091 set_page_extent_mapped(pages[i]);
1092 set_page_dirty(pages[i]);
1093 unlock_page(pages[i]);
1094 page_cache_release(pages[i]);
1095 }
1096 return i_done;
1097out:
1098 for (i = 0; i < i_done; i++) {
1099 unlock_page(pages[i]);
1100 page_cache_release(pages[i]);
1101 }
1f12bd06 1102 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1103 return ret;
1104
1105}
1106
1107int btrfs_defrag_file(struct inode *inode, struct file *file,
1108 struct btrfs_ioctl_defrag_range_args *range,
1109 u64 newer_than, unsigned long max_to_defrag)
1110{
1111 struct btrfs_root *root = BTRFS_I(inode)->root;
4cb5300b 1112 struct file_ra_state *ra = NULL;
f46b5a66 1113 unsigned long last_index;
151a31b2 1114 u64 isize = i_size_read(inode);
940100a4
CM		 1115 u64 last_len = 0;
1116 u64 skip = 0;
1117 u64 defrag_end = 0;
4cb5300b 1118 u64 newer_off = range->start;
f46b5a66 1119 unsigned long i;
008873ea 1120 unsigned long ra_index = 0;
f46b5a66 1121 int ret;
4cb5300b 1122 int defrag_count = 0;
1a419d85 1123 int compress_type = BTRFS_COMPRESS_ZLIB;
4cb5300b 1124 int extent_thresh = range->extent_thresh;
008873ea
LZ		 1125 int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1126 int cluster = max_cluster;
4cb5300b
CM		 1127 u64 new_align = ~((u64)128 * 1024 - 1);
1128 struct page **pages = NULL;
1129
1130 if (extent_thresh == 0)
1131 extent_thresh = 256 * 1024;
1a419d85
LZ		 1132
1133 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1134 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1135 return -EINVAL;
1136 if (range->compress_type)
1137 compress_type = range->compress_type;
1138 }
f46b5a66 1139
151a31b2 1140 if (isize == 0)
940100a4
CM		 1141 return 0;
1142
4cb5300b
CM		 1143 /*
1144 * if we were not given a file, allocate a readahead
1145 * context
1146 */
1147 if (!file) {
1148 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1149 if (!ra)
1150 return -ENOMEM;
1151 file_ra_state_init(ra, inode->i_mapping);
1152 } else {
1153 ra = &file->f_ra;
1154 }
1155
008873ea 1156 pages = kmalloc(sizeof(struct page *) * max_cluster,
4cb5300b
CM		 1157 GFP_NOFS);
1158 if (!pages) {
1159 ret = -ENOMEM;
1160 goto out_ra;
1161 }
1162
1163 /* find the last page to defrag */
1e701a32 1164 if (range->start + range->len > range->start) {
151a31b2 1165 last_index = min_t(u64, isize - 1,
1e701a32
CM		 1166 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1167 } else {
151a31b2 1168 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1e701a32
CM		 1169 }
1170
4cb5300b
CM		 1171 if (newer_than) {
1172 ret = find_new_extents(root, inode, newer_than,
1173 &newer_off, 64 * 1024);
1174 if (!ret) {
1175 range->start = newer_off;
1176 /*
1177 * we always align our defrag to help keep
1178 * the extents in the file evenly spaced
1179 */
1180 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
4cb5300b
CM		 1181 } else
1182 goto out_ra;
1183 } else {
1184 i = range->start >> PAGE_CACHE_SHIFT;
1185 }
1186 if (!max_to_defrag)
7ec31b54 1187 max_to_defrag = last_index + 1;
4cb5300b 1188
2a0f7f57
LZ		 1189 /*
1190 * make writeback starts from i, so the defrag range can be
1191 * written sequentially.
1192 */
1193 if (i < inode->i_mapping->writeback_index)
1194 inode->i_mapping->writeback_index = i;
1195
f7f43cc8
CM		 1196 while (i <= last_index && defrag_count < max_to_defrag &&
1197 (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1198 PAGE_CACHE_SHIFT)) {
4cb5300b
CM		 1199 /*
1200 * make sure we stop running if someone unmounts
1201 * the FS
1202 */
1203 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1204 break;
1205
66c26892 1206 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
6c282eb4 1207 extent_thresh, &last_len, &skip,
a43a2111
AM		 1208 &defrag_end, range->flags &
1209 BTRFS_DEFRAG_RANGE_COMPRESS)) {
940100a4
CM		 1210 unsigned long next;
1211 /*
1212 * the should_defrag function tells us how much to skip
1213 * bump our counter by the suggested amount
1214 */
1215 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1216 i = max(i + 1, next);
1217 continue;
1218 }
008873ea
LZ		 1219
1220 if (!newer_than) {
1221 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1222 PAGE_CACHE_SHIFT) - i;
1223 cluster = min(cluster, max_cluster);
1224 } else {
1225 cluster = max_cluster;
1226 }
1227
1e701a32 1228 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1a419d85 1229 BTRFS_I(inode)->force_compress = compress_type;
940100a4 1230
008873ea
LZ		 1231 if (i + cluster > ra_index) {
1232 ra_index = max(i, ra_index);
1233 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1234 cluster);
1235 ra_index += max_cluster;
1236 }
940100a4 1237
ecb8bea8 1238 mutex_lock(&inode->i_mutex);
008873ea 1239 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
ecb8bea8
LB		 1240 if (ret < 0) {
1241 mutex_unlock(&inode->i_mutex);
4cb5300b 1242 goto out_ra;
ecb8bea8 1243 }
4cb5300b
CM		 1244
1245 defrag_count += ret;
1246 balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret);
ecb8bea8 1247 mutex_unlock(&inode->i_mutex);
4cb5300b
CM		 1248
1249 if (newer_than) {
1250 if (newer_off == (u64)-1)
1251 break;
1252
e1f041e1
LB		 1253 if (ret > 0)
1254 i += ret;
1255
4cb5300b
CM		 1256 newer_off = max(newer_off + 1,
1257 (u64)i << PAGE_CACHE_SHIFT);
1258
1259 ret = find_new_extents(root, inode,
1260 newer_than, &newer_off,
1261 64 * 1024);
1262 if (!ret) {
1263 range->start = newer_off;
1264 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
4cb5300b
CM		 1265 } else {
1266 break;
f46b5a66 1267 }
4cb5300b 1268 } else {
008873ea 1269 if (ret > 0) {
cbcc8326 1270 i += ret;
008873ea
LZ		 1271 last_len += ret << PAGE_CACHE_SHIFT;
1272 } else {
cbcc8326 1273 i++;
008873ea
LZ		 1274 last_len = 0;
1275 }
f46b5a66 1276 }
f46b5a66
CH		 1277 }
1278
1e701a32
CM		 1279 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1280 filemap_flush(inode->i_mapping);
1281
1282 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1283 /* the filemap_flush will queue IO into the worker threads, but
1284 * we have to make sure the IO is actually started and that
1285 * ordered extents get created before we return
1286 */
1287 atomic_inc(&root->fs_info->async_submit_draining);
1288 while (atomic_read(&root->fs_info->nr_async_submits) ||
1289 atomic_read(&root->fs_info->async_delalloc_pages)) {
1290 wait_event(root->fs_info->async_submit_wait,
1291 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1292 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1293 }
1294 atomic_dec(&root->fs_info->async_submit_draining);
1295
1296 mutex_lock(&inode->i_mutex);
261507a0 1297 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1e701a32
CM		 1298 mutex_unlock(&inode->i_mutex);
1299 }
1300
1a419d85 1301 if (range->compress_type == BTRFS_COMPRESS_LZO) {
2b0ce2c2 1302 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1a419d85
LZ		 1303 }
1304
60ccf82f 1305 ret = defrag_count;
940100a4 1306
4cb5300b
CM		 1307out_ra:
1308 if (!file)
1309 kfree(ra);
1310 kfree(pages);
940100a4 1311 return ret;
f46b5a66
CH		 1312}
1313
198605a8 1314static noinline int btrfs_ioctl_resize(struct file *file,
76dda93c 1315 void __user *arg)
f46b5a66
CH		 1316{
1317 u64 new_size;
1318 u64 old_size;
1319 u64 devid = 1;
198605a8 1320 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
f46b5a66
CH		 1321 struct btrfs_ioctl_vol_args *vol_args;
1322 struct btrfs_trans_handle *trans;
1323 struct btrfs_device *device = NULL;
1324 char *sizestr;
1325 char *devstr = NULL;
1326 int ret = 0;
f46b5a66
CH		 1327 int mod = 0;
1328
c146afad
YZ		 1329 if (root->fs_info->sb->s_flags & MS_RDONLY)
1330 return -EROFS;
1331
e441d54d
CM		 1332 if (!capable(CAP_SYS_ADMIN))
1333 return -EPERM;
1334
198605a8
MX		 1335 ret = mnt_want_write_file(file);
1336 if (ret)
1337 return ret;
1338
5ac00add
SB		 1339 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1340 1)) {
1341 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
1342 return -EINPROGRESS;
c9e9f97b
ID		 1343 }
1344
5ac00add 1345 mutex_lock(&root->fs_info->volume_mutex);
dae7b665 1346 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 1347 if (IS_ERR(vol_args)) {
1348 ret = PTR_ERR(vol_args);
1349 goto out;
1350 }
5516e595
MF		 1351
1352 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 1353
f46b5a66
CH		 1354 sizestr = vol_args->name;
1355 devstr = strchr(sizestr, ':');
1356 if (devstr) {
1357 char *end;
1358 sizestr = devstr + 1;
1359 *devstr = '\0';
1360 devstr = vol_args->name;
1361 devid = simple_strtoull(devstr, &end, 10);
5bb14682 1362 printk(KERN_INFO "btrfs: resizing devid %llu\n",
21380931 1363 (unsigned long long)devid);
f46b5a66 1364 }
dba60f3f 1365
aa1b8cd4 1366 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
f46b5a66 1367 if (!device) {
5bb14682 1368 printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
21380931 1369 (unsigned long long)devid);
f46b5a66 1370 ret = -EINVAL;
c9e9f97b 1371 goto out_free;
f46b5a66 1372 }
dba60f3f
MX		 1373
1374 if (!device->writeable) {
4e42ae1b 1375 printk(KERN_INFO "btrfs: resizer unable to apply on "
dba60f3f 1376 "readonly device %llu\n",
a8c4a33b 1377 (unsigned long long)devid);
4e42ae1b
LB		 1378 ret = -EINVAL;
1379 goto out_free;
1380 }
1381
f46b5a66
CH		 1382 if (!strcmp(sizestr, "max"))
1383 new_size = device->bdev->bd_inode->i_size;
1384 else {
1385 if (sizestr[0] == '-') {
1386 mod = -1;
1387 sizestr++;
1388 } else if (sizestr[0] == '+') {
1389 mod = 1;
1390 sizestr++;
1391 }
91748467 1392 new_size = memparse(sizestr, NULL);
f46b5a66
CH		 1393 if (new_size == 0) {
1394 ret = -EINVAL;
c9e9f97b 1395 goto out_free;
f46b5a66
CH		 1396 }
1397 }
1398
63a212ab
SB		 1399 if (device->is_tgtdev_for_dev_replace) {
1400 ret = -EINVAL;
1401 goto out_free;
1402 }
1403
f46b5a66
CH		 1404 old_size = device->total_bytes;
1405
1406 if (mod < 0) {
1407 if (new_size > old_size) {
1408 ret = -EINVAL;
c9e9f97b 1409 goto out_free;
f46b5a66
CH		 1410 }
1411 new_size = old_size - new_size;
1412 } else if (mod > 0) {
1413 new_size = old_size + new_size;
1414 }
1415
1416 if (new_size < 256 * 1024 * 1024) {
1417 ret = -EINVAL;
c9e9f97b 1418 goto out_free;
f46b5a66
CH		 1419 }
1420 if (new_size > device->bdev->bd_inode->i_size) {
1421 ret = -EFBIG;
c9e9f97b 1422 goto out_free;
f46b5a66
CH		 1423 }
1424
1425 do_div(new_size, root->sectorsize);
1426 new_size *= root->sectorsize;
1427
606686ee
JB		 1428 printk_in_rcu(KERN_INFO "btrfs: new size for %s is %llu\n",
1429 rcu_str_deref(device->name),
1430 (unsigned long long)new_size);
f46b5a66
CH		 1431
1432 if (new_size > old_size) {
a22285a6 1433 trans = btrfs_start_transaction(root, 0);
98d5dc13
TI		 1434 if (IS_ERR(trans)) {
1435 ret = PTR_ERR(trans);
c9e9f97b 1436 goto out_free;
98d5dc13 1437 }
f46b5a66
CH		 1438 ret = btrfs_grow_device(trans, device, new_size);
1439 btrfs_commit_transaction(trans, root);
ece7d20e 1440 } else if (new_size < old_size) {
f46b5a66 1441 ret = btrfs_shrink_device(device, new_size);
0253f40e 1442 } /* equal, nothing need to do */
f46b5a66 1443
c9e9f97b 1444out_free:
f46b5a66 1445 kfree(vol_args);
c9e9f97b
ID		 1446out:
1447 mutex_unlock(&root->fs_info->volume_mutex);
198605a8 1448 mnt_drop_write_file(file);
5ac00add 1449 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
f46b5a66
CH		 1450 return ret;
1451}
1452
72fd032e 1453static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
6f72c7e2
AJ		 1454 char *name, unsigned long fd, int subvol,
1455 u64 *transid, bool readonly,
1456 struct btrfs_qgroup_inherit **inherit)
f46b5a66 1457{
f46b5a66 1458 int namelen;
3de4586c 1459 int ret = 0;
f46b5a66 1460
a874a63e
LB		 1461 ret = mnt_want_write_file(file);
1462 if (ret)
1463 goto out;
1464
72fd032e
SW		 1465 namelen = strlen(name);
1466 if (strchr(name, '/')) {
f46b5a66 1467 ret = -EINVAL;
a874a63e 1468 goto out_drop_write;
f46b5a66
CH		 1469 }
1470
16780cab
CM		 1471 if (name[0] == '.' &&
1472 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1473 ret = -EEXIST;
a874a63e 1474 goto out_drop_write;
16780cab
CM		 1475 }
1476
3de4586c 1477 if (subvol) {
72fd032e 1478 ret = btrfs_mksubvol(&file->f_path, name, namelen,
6f72c7e2 1479 NULL, transid, readonly, inherit);
cb8e7090 1480 } else {
2903ff01 1481 struct fd src = fdget(fd);
3de4586c 1482 struct inode *src_inode;
2903ff01 1483 if (!src.file) {
3de4586c 1484 ret = -EINVAL;
a874a63e 1485 goto out_drop_write;
3de4586c
CM		 1486 }
1487
2903ff01 1488 src_inode = src.file->f_path.dentry->d_inode;
3de4586c 1489 if (src_inode->i_sb != file->f_path.dentry->d_inode->i_sb) {
d397712b
CM		 1490 printk(KERN_INFO "btrfs: Snapshot src from "
1491 "another FS\n");
3de4586c 1492 ret = -EINVAL;
ecd18815
AV		 1493 } else {
1494 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1495 BTRFS_I(src_inode)->root,
1496 transid, readonly, inherit);
3de4586c 1497 }
2903ff01 1498 fdput(src);
cb8e7090 1499 }
a874a63e
LB		 1500out_drop_write:
1501 mnt_drop_write_file(file);
f46b5a66 1502out:
72fd032e
SW		 1503 return ret;
1504}
1505
1506static noinline int btrfs_ioctl_snap_create(struct file *file,
fa0d2b9b 1507 void __user *arg, int subvol)
72fd032e 1508{
fa0d2b9b 1509 struct btrfs_ioctl_vol_args *vol_args;
72fd032e
SW		 1510 int ret;
1511
fa0d2b9b
LZ		 1512 vol_args = memdup_user(arg, sizeof(*vol_args));
1513 if (IS_ERR(vol_args))
1514 return PTR_ERR(vol_args);
1515 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
72fd032e 1516
fa0d2b9b 1517 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
b83cc969 1518 vol_args->fd, subvol,
6f72c7e2 1519 NULL, false, NULL);
fdfb1e4f 1520
fa0d2b9b
LZ		 1521 kfree(vol_args);
1522 return ret;
1523}
fdfb1e4f 1524
fa0d2b9b
LZ		 1525static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1526 void __user *arg, int subvol)
1527{
1528 struct btrfs_ioctl_vol_args_v2 *vol_args;
1529 int ret;
1530 u64 transid = 0;
1531 u64 *ptr = NULL;
b83cc969 1532 bool readonly = false;
6f72c7e2 1533 struct btrfs_qgroup_inherit *inherit = NULL;
75eaa0e2 1534
fa0d2b9b
LZ		 1535 vol_args = memdup_user(arg, sizeof(*vol_args));
1536 if (IS_ERR(vol_args))
1537 return PTR_ERR(vol_args);
1538 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
75eaa0e2 1539
b83cc969 1540 if (vol_args->flags &
6f72c7e2
AJ		 1541 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1542 BTRFS_SUBVOL_QGROUP_INHERIT)) {
b83cc969 1543 ret = -EOPNOTSUPP;
fa0d2b9b 1544 goto out;
72fd032e 1545 }
fa0d2b9b
LZ		 1546
1547 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1548 ptr = &transid;
b83cc969
LZ		 1549 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1550 readonly = true;
6f72c7e2
AJ		 1551 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1552 if (vol_args->size > PAGE_CACHE_SIZE) {
1553 ret = -EINVAL;
1554 goto out;
1555 }
1556 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1557 if (IS_ERR(inherit)) {
1558 ret = PTR_ERR(inherit);
1559 goto out;
1560 }
1561 }
fa0d2b9b
LZ		 1562
1563 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
6f72c7e2
AJ		 1564 vol_args->fd, subvol, ptr,
1565 readonly, &inherit);
fa0d2b9b
LZ		 1566
1567 if (ret == 0 && ptr &&
1568 copy_to_user(arg +
1569 offsetof(struct btrfs_ioctl_vol_args_v2,
1570 transid), ptr, sizeof(*ptr)))
1571 ret = -EFAULT;
fdfb1e4f 1572out:
f46b5a66 1573 kfree(vol_args);
6f72c7e2 1574 kfree(inherit);
f46b5a66
CH		 1575 return ret;
1576}
1577
0caa102d
LZ		 1578static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1579 void __user *arg)
1580{
1581 struct inode *inode = fdentry(file)->d_inode;
1582 struct btrfs_root *root = BTRFS_I(inode)->root;
1583 int ret = 0;
1584 u64 flags = 0;
1585
33345d01 1586 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
0caa102d
LZ		 1587 return -EINVAL;
1588
1589 down_read(&root->fs_info->subvol_sem);
1590 if (btrfs_root_readonly(root))
1591 flags |= BTRFS_SUBVOL_RDONLY;
1592 up_read(&root->fs_info->subvol_sem);
1593
1594 if (copy_to_user(arg, &flags, sizeof(flags)))
1595 ret = -EFAULT;
1596
1597 return ret;
1598}
1599
1600static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1601 void __user *arg)
1602{
1603 struct inode *inode = fdentry(file)->d_inode;
1604 struct btrfs_root *root = BTRFS_I(inode)->root;
1605 struct btrfs_trans_handle *trans;
1606 u64 root_flags;
1607 u64 flags;
1608 int ret = 0;
1609
b9ca0664
LB		 1610 ret = mnt_want_write_file(file);
1611 if (ret)
1612 goto out;
0caa102d 1613
b9ca0664
LB		 1614 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1615 ret = -EINVAL;
1616 goto out_drop_write;
1617 }
0caa102d 1618
b9ca0664
LB		 1619 if (copy_from_user(&flags, arg, sizeof(flags))) {
1620 ret = -EFAULT;
1621 goto out_drop_write;
1622 }
0caa102d 1623
b9ca0664
LB		 1624 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1625 ret = -EINVAL;
1626 goto out_drop_write;
1627 }
0caa102d 1628
b9ca0664
LB		 1629 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1630 ret = -EOPNOTSUPP;
1631 goto out_drop_write;
1632 }
0caa102d 1633
b9ca0664
LB		 1634 if (!inode_owner_or_capable(inode)) {
1635 ret = -EACCES;
1636 goto out_drop_write;
1637 }
b4dc2b8c 1638
0caa102d
LZ		 1639 down_write(&root->fs_info->subvol_sem);
1640
1641 /* nothing to do */
1642 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
b9ca0664 1643 goto out_drop_sem;
0caa102d
LZ		 1644
1645 root_flags = btrfs_root_flags(&root->root_item);
1646 if (flags & BTRFS_SUBVOL_RDONLY)
1647 btrfs_set_root_flags(&root->root_item,
1648 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1649 else
1650 btrfs_set_root_flags(&root->root_item,
1651 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1652
1653 trans = btrfs_start_transaction(root, 1);
1654 if (IS_ERR(trans)) {
1655 ret = PTR_ERR(trans);
1656 goto out_reset;
1657 }
1658
b4dc2b8c 1659 ret = btrfs_update_root(trans, root->fs_info->tree_root,
0caa102d
LZ		 1660 &root->root_key, &root->root_item);
1661
1662 btrfs_commit_transaction(trans, root);
1663out_reset:
1664 if (ret)
1665 btrfs_set_root_flags(&root->root_item, root_flags);
b9ca0664 1666out_drop_sem:
0caa102d 1667 up_write(&root->fs_info->subvol_sem);
b9ca0664
LB		 1668out_drop_write:
1669 mnt_drop_write_file(file);
1670out:
0caa102d
LZ		 1671 return ret;
1672}
1673
76dda93c
YZ		 1674/*
1675 * helper to check if the subvolume references other subvolumes
1676 */
1677static noinline int may_destroy_subvol(struct btrfs_root *root)
1678{
1679 struct btrfs_path *path;
1680 struct btrfs_key key;
1681 int ret;
1682
1683 path = btrfs_alloc_path();
1684 if (!path)
1685 return -ENOMEM;
1686
1687 key.objectid = root->root_key.objectid;
1688 key.type = BTRFS_ROOT_REF_KEY;
1689 key.offset = (u64)-1;
1690
1691 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1692 &key, path, 0, 0);
1693 if (ret < 0)
1694 goto out;
1695 BUG_ON(ret == 0);
1696
1697 ret = 0;
1698 if (path->slots[0] > 0) {
1699 path->slots[0]--;
1700 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1701 if (key.objectid == root->root_key.objectid &&
1702 key.type == BTRFS_ROOT_REF_KEY)
1703 ret = -ENOTEMPTY;
1704 }
1705out:
1706 btrfs_free_path(path);
1707 return ret;
1708}
1709
ac8e9819
CM		 1710static noinline int key_in_sk(struct btrfs_key *key,
1711 struct btrfs_ioctl_search_key *sk)
1712{
abc6e134
CM		 1713 struct btrfs_key test;
1714 int ret;
1715
1716 test.objectid = sk->min_objectid;
1717 test.type = sk->min_type;
1718 test.offset = sk->min_offset;
1719
1720 ret = btrfs_comp_cpu_keys(key, &test);
1721 if (ret < 0)
ac8e9819 1722 return 0;
abc6e134
CM		 1723
1724 test.objectid = sk->max_objectid;
1725 test.type = sk->max_type;
1726 test.offset = sk->max_offset;
1727
1728 ret = btrfs_comp_cpu_keys(key, &test);
1729 if (ret > 0)
ac8e9819
CM		 1730 return 0;
1731 return 1;
1732}
1733
1734static noinline int copy_to_sk(struct btrfs_root *root,
1735 struct btrfs_path *path,
1736 struct btrfs_key *key,
1737 struct btrfs_ioctl_search_key *sk,
1738 char *buf,
1739 unsigned long *sk_offset,
1740 int *num_found)
1741{
1742 u64 found_transid;
1743 struct extent_buffer *leaf;
1744 struct btrfs_ioctl_search_header sh;
1745 unsigned long item_off;
1746 unsigned long item_len;
1747 int nritems;
1748 int i;
1749 int slot;
ac8e9819
CM		 1750 int ret = 0;
1751
1752 leaf = path->nodes[0];
1753 slot = path->slots[0];
1754 nritems = btrfs_header_nritems(leaf);
1755
1756 if (btrfs_header_generation(leaf) > sk->max_transid) {
1757 i = nritems;
1758 goto advance_key;
1759 }
1760 found_transid = btrfs_header_generation(leaf);
1761
1762 for (i = slot; i < nritems; i++) {
1763 item_off = btrfs_item_ptr_offset(leaf, i);
1764 item_len = btrfs_item_size_nr(leaf, i);
1765
1766 if (item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
1767 item_len = 0;
1768
1769 if (sizeof(sh) + item_len + *sk_offset >
1770 BTRFS_SEARCH_ARGS_BUFSIZE) {
1771 ret = 1;
1772 goto overflow;
1773 }
1774
1775 btrfs_item_key_to_cpu(leaf, key, i);
1776 if (!key_in_sk(key, sk))
1777 continue;
1778
1779 sh.objectid = key->objectid;
1780 sh.offset = key->offset;
1781 sh.type = key->type;
1782 sh.len = item_len;
1783 sh.transid = found_transid;
1784
1785 /* copy search result header */
1786 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1787 *sk_offset += sizeof(sh);
1788
1789 if (item_len) {
1790 char *p = buf + *sk_offset;
1791 /* copy the item */
1792 read_extent_buffer(leaf, p,
1793 item_off, item_len);
1794 *sk_offset += item_len;
ac8e9819 1795 }
e2156867 1796 (*num_found)++;
ac8e9819
CM		 1797
1798 if (*num_found >= sk->nr_items)
1799 break;
1800 }
1801advance_key:
abc6e134
CM		 1802 ret = 0;
1803 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
ac8e9819 1804 key->offset++;
abc6e134
CM		 1805 else if (key->type < (u8)-1 && key->type < sk->max_type) {
1806 key->offset = 0;
ac8e9819 1807 key->type++;
abc6e134
CM		 1808 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1809 key->offset = 0;
1810 key->type = 0;
ac8e9819 1811 key->objectid++;
abc6e134
CM		 1812 } else
1813 ret = 1;
ac8e9819 1814overflow:
ac8e9819
CM		 1815 return ret;
1816}
1817
1818static noinline int search_ioctl(struct inode *inode,
1819 struct btrfs_ioctl_search_args *args)
1820{
1821 struct btrfs_root *root;
1822 struct btrfs_key key;
1823 struct btrfs_key max_key;
1824 struct btrfs_path *path;
1825 struct btrfs_ioctl_search_key *sk = &args->key;
1826 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1827 int ret;
1828 int num_found = 0;
1829 unsigned long sk_offset = 0;
1830
1831 path = btrfs_alloc_path();
1832 if (!path)
1833 return -ENOMEM;
1834
1835 if (sk->tree_id == 0) {
1836 /* search the root of the inode that was passed */
1837 root = BTRFS_I(inode)->root;
1838 } else {
1839 key.objectid = sk->tree_id;
1840 key.type = BTRFS_ROOT_ITEM_KEY;
1841 key.offset = (u64)-1;
1842 root = btrfs_read_fs_root_no_name(info, &key);
1843 if (IS_ERR(root)) {
1844 printk(KERN_ERR "could not find root %llu\n",
1845 sk->tree_id);
1846 btrfs_free_path(path);
1847 return -ENOENT;
1848 }
1849 }
1850
1851 key.objectid = sk->min_objectid;
1852 key.type = sk->min_type;
1853 key.offset = sk->min_offset;
1854
1855 max_key.objectid = sk->max_objectid;
1856 max_key.type = sk->max_type;
1857 max_key.offset = sk->max_offset;
1858
1859 path->keep_locks = 1;
1860
1861 while(1) {
1862 ret = btrfs_search_forward(root, &key, &max_key, path, 0,
1863 sk->min_transid);
1864 if (ret != 0) {
1865 if (ret > 0)
1866 ret = 0;
1867 goto err;
1868 }
1869 ret = copy_to_sk(root, path, &key, sk, args->buf,
1870 &sk_offset, &num_found);
b3b4aa74 1871 btrfs_release_path(path);
ac8e9819
CM		 1872 if (ret || num_found >= sk->nr_items)
1873 break;
1874
1875 }
1876 ret = 0;
1877err:
1878 sk->nr_items = num_found;
1879 btrfs_free_path(path);
1880 return ret;
1881}
1882
1883static noinline int btrfs_ioctl_tree_search(struct file *file,
1884 void __user *argp)
1885{
1886 struct btrfs_ioctl_search_args *args;
1887 struct inode *inode;
1888 int ret;
1889
1890 if (!capable(CAP_SYS_ADMIN))
1891 return -EPERM;
1892
2354d08f
JL		 1893 args = memdup_user(argp, sizeof(*args));
1894 if (IS_ERR(args))
1895 return PTR_ERR(args);
ac8e9819 1896
ac8e9819
CM		 1897 inode = fdentry(file)->d_inode;
1898 ret = search_ioctl(inode, args);
1899 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1900 ret = -EFAULT;
1901 kfree(args);
1902 return ret;
1903}
1904
98d377a0 1905/*
ac8e9819
CM		 1906 * Search INODE_REFs to identify path name of 'dirid' directory
1907 * in a 'tree_id' tree. and sets path name to 'name'.
1908 */
98d377a0
TH		 1909static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1910 u64 tree_id, u64 dirid, char *name)
1911{
1912 struct btrfs_root *root;
1913 struct btrfs_key key;
ac8e9819 1914 char *ptr;
98d377a0
TH		 1915 int ret = -1;
1916 int slot;
1917 int len;
1918 int total_len = 0;
1919 struct btrfs_inode_ref *iref;
1920 struct extent_buffer *l;
1921 struct btrfs_path *path;
1922
1923 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1924 name[0]='\0';
1925 return 0;
1926 }
1927
1928 path = btrfs_alloc_path();
1929 if (!path)
1930 return -ENOMEM;
1931
ac8e9819 1932 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
98d377a0
TH		 1933
1934 key.objectid = tree_id;
1935 key.type = BTRFS_ROOT_ITEM_KEY;
1936 key.offset = (u64)-1;
1937 root = btrfs_read_fs_root_no_name(info, &key);
1938 if (IS_ERR(root)) {
1939 printk(KERN_ERR "could not find root %llu\n", tree_id);
8ad6fcab
CM		 1940 ret = -ENOENT;
1941 goto out;
98d377a0
TH		 1942 }
1943
1944 key.objectid = dirid;
1945 key.type = BTRFS_INODE_REF_KEY;
8ad6fcab 1946 key.offset = (u64)-1;
98d377a0
TH		 1947
1948 while(1) {
1949 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
1950 if (ret < 0)
1951 goto out;
1952
1953 l = path->nodes[0];
1954 slot = path->slots[0];
8ad6fcab
CM		 1955 if (ret > 0 && slot > 0)
1956 slot--;
98d377a0
TH		 1957 btrfs_item_key_to_cpu(l, &key, slot);
1958
1959 if (ret > 0 && (key.objectid != dirid ||
ac8e9819
CM		 1960 key.type != BTRFS_INODE_REF_KEY)) {
1961 ret = -ENOENT;
98d377a0 1962 goto out;
ac8e9819 1963 }
98d377a0
TH		 1964
1965 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
1966 len = btrfs_inode_ref_name_len(l, iref);
1967 ptr -= len + 1;
1968 total_len += len + 1;
ac8e9819 1969 if (ptr < name)
98d377a0
TH		 1970 goto out;
1971
1972 *(ptr + len) = '/';
1973 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
1974
1975 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
1976 break;
1977
b3b4aa74 1978 btrfs_release_path(path);
98d377a0 1979 key.objectid = key.offset;
8ad6fcab 1980 key.offset = (u64)-1;
98d377a0 1981 dirid = key.objectid;
98d377a0 1982 }
ac8e9819 1983 if (ptr < name)
98d377a0 1984 goto out;
77906a50 1985 memmove(name, ptr, total_len);
98d377a0
TH		 1986 name[total_len]='\0';
1987 ret = 0;
1988out:
1989 btrfs_free_path(path);
ac8e9819
CM		 1990 return ret;
1991}
1992
1993static noinline int btrfs_ioctl_ino_lookup(struct file *file,
1994 void __user *argp)
1995{
1996 struct btrfs_ioctl_ino_lookup_args *args;
1997 struct inode *inode;
1998 int ret;
1999
2000 if (!capable(CAP_SYS_ADMIN))
2001 return -EPERM;
2002
2354d08f
JL		 2003 args = memdup_user(argp, sizeof(*args));
2004 if (IS_ERR(args))
2005 return PTR_ERR(args);
c2b96929 2006
ac8e9819
CM		 2007 inode = fdentry(file)->d_inode;
2008
1b53ac4d
CM		 2009 if (args->treeid == 0)
2010 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2011
ac8e9819
CM		 2012 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2013 args->treeid, args->objectid,
2014 args->name);
2015
2016 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2017 ret = -EFAULT;
2018
2019 kfree(args);
98d377a0
TH		 2020 return ret;
2021}
2022
76dda93c
YZ		 2023static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2024 void __user *arg)
2025{
2026 struct dentry *parent = fdentry(file);
2027 struct dentry *dentry;
2028 struct inode *dir = parent->d_inode;
2029 struct inode *inode;
2030 struct btrfs_root *root = BTRFS_I(dir)->root;
2031 struct btrfs_root *dest = NULL;
2032 struct btrfs_ioctl_vol_args *vol_args;
2033 struct btrfs_trans_handle *trans;
2034 int namelen;
2035 int ret;
2036 int err = 0;
2037
76dda93c
YZ		 2038 vol_args = memdup_user(arg, sizeof(*vol_args));
2039 if (IS_ERR(vol_args))
2040 return PTR_ERR(vol_args);
2041
2042 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2043 namelen = strlen(vol_args->name);
2044 if (strchr(vol_args->name, '/') ||
2045 strncmp(vol_args->name, "..", namelen) == 0) {
2046 err = -EINVAL;
2047 goto out;
2048 }
2049
a561be71 2050 err = mnt_want_write_file(file);
76dda93c
YZ		 2051 if (err)
2052 goto out;
2053
2054 mutex_lock_nested(&dir->i_mutex, I_MUTEX_PARENT);
2055 dentry = lookup_one_len(vol_args->name, parent, namelen);
2056 if (IS_ERR(dentry)) {
2057 err = PTR_ERR(dentry);
2058 goto out_unlock_dir;
2059 }
2060
2061 if (!dentry->d_inode) {
2062 err = -ENOENT;
2063 goto out_dput;
2064 }
2065
2066 inode = dentry->d_inode;
4260f7c7
SW		 2067 dest = BTRFS_I(inode)->root;
2068 if (!capable(CAP_SYS_ADMIN)){
2069 /*
2070 * Regular user. Only allow this with a special mount
2071 * option, when the user has write+exec access to the
2072 * subvol root, and when rmdir(2) would have been
2073 * allowed.
2074 *
2075 * Note that this is _not_ check that the subvol is
2076 * empty or doesn't contain data that we wouldn't
2077 * otherwise be able to delete.
2078 *
2079 * Users who want to delete empty subvols should try
2080 * rmdir(2).
2081 */
2082 err = -EPERM;
2083 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2084 goto out_dput;
2085
2086 /*
2087 * Do not allow deletion if the parent dir is the same
2088 * as the dir to be deleted. That means the ioctl
2089 * must be called on the dentry referencing the root
2090 * of the subvol, not a random directory contained
2091 * within it.
2092 */
2093 err = -EINVAL;
2094 if (root == dest)
2095 goto out_dput;
2096
2097 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2098 if (err)
2099 goto out_dput;
4260f7c7
SW		 2100 }
2101
5c39da5b
MX		 2102 /* check if subvolume may be deleted by a user */
2103 err = btrfs_may_delete(dir, dentry, 1);
2104 if (err)
2105 goto out_dput;
2106
33345d01 2107 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
76dda93c
YZ		 2108 err = -EINVAL;
2109 goto out_dput;
2110 }
2111
76dda93c
YZ		 2112 mutex_lock(&inode->i_mutex);
2113 err = d_invalidate(dentry);
2114 if (err)
2115 goto out_unlock;
2116
2117 down_write(&root->fs_info->subvol_sem);
2118
2119 err = may_destroy_subvol(dest);
2120 if (err)
2121 goto out_up_write;
2122
a22285a6
YZ		 2123 trans = btrfs_start_transaction(root, 0);
2124 if (IS_ERR(trans)) {
2125 err = PTR_ERR(trans);
d327099a 2126 goto out_up_write;
a22285a6
YZ		 2127 }
2128 trans->block_rsv = &root->fs_info->global_block_rsv;
2129
76dda93c
YZ		 2130 ret = btrfs_unlink_subvol(trans, root, dir,
2131 dest->root_key.objectid,
2132 dentry->d_name.name,
2133 dentry->d_name.len);
79787eaa
JM		 2134 if (ret) {
2135 err = ret;
2136 btrfs_abort_transaction(trans, root, ret);
2137 goto out_end_trans;
2138 }
76dda93c
YZ		 2139
2140 btrfs_record_root_in_trans(trans, dest);
2141
2142 memset(&dest->root_item.drop_progress, 0,
2143 sizeof(dest->root_item.drop_progress));
2144 dest->root_item.drop_level = 0;
2145 btrfs_set_root_refs(&dest->root_item, 0);
2146
d68fc57b
YZ		 2147 if (!xchg(&dest->orphan_item_inserted, 1)) {
2148 ret = btrfs_insert_orphan_item(trans,
2149 root->fs_info->tree_root,
2150 dest->root_key.objectid);
79787eaa
JM		 2151 if (ret) {
2152 btrfs_abort_transaction(trans, root, ret);
2153 err = ret;
2154 goto out_end_trans;
2155 }
d68fc57b 2156 }
79787eaa 2157out_end_trans:
531cb13f 2158 ret = btrfs_end_transaction(trans, root);
79787eaa
JM		 2159 if (ret && !err)
2160 err = ret;
76dda93c
YZ		 2161 inode->i_flags |= S_DEAD;
2162out_up_write:
2163 up_write(&root->fs_info->subvol_sem);
2164out_unlock:
2165 mutex_unlock(&inode->i_mutex);
2166 if (!err) {
efefb143 2167 shrink_dcache_sb(root->fs_info->sb);
76dda93c
YZ		 2168 btrfs_invalidate_inodes(dest);
2169 d_delete(dentry);
2170 }
2171out_dput:
2172 dput(dentry);
2173out_unlock_dir:
2174 mutex_unlock(&dir->i_mutex);
2a79f17e 2175 mnt_drop_write_file(file);
76dda93c
YZ		 2176out:
2177 kfree(vol_args);
2178 return err;
2179}
2180
1e701a32 2181static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
f46b5a66
CH		 2182{
2183 struct inode *inode = fdentry(file)->d_inode;
2184 struct btrfs_root *root = BTRFS_I(inode)->root;
1e701a32 2185 struct btrfs_ioctl_defrag_range_args *range;
c146afad
YZ		 2186 int ret;
2187
b83cc969
LZ		 2188 if (btrfs_root_readonly(root))
2189 return -EROFS;
2190
5ac00add
SB		 2191 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2192 1)) {
2193 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2194 return -EINPROGRESS;
2195 }
a561be71 2196 ret = mnt_want_write_file(file);
5ac00add
SB		 2197 if (ret) {
2198 atomic_set(&root->fs_info->mutually_exclusive_operation_running,
2199 0);
c146afad 2200 return ret;
5ac00add 2201 }
f46b5a66
CH		 2202
2203 switch (inode->i_mode & S_IFMT) {
2204 case S_IFDIR:
e441d54d
CM		 2205 if (!capable(CAP_SYS_ADMIN)) {
2206 ret = -EPERM;
2207 goto out;
2208 }
8929ecfa
YZ		 2209 ret = btrfs_defrag_root(root, 0);
2210 if (ret)
2211 goto out;
2212 ret = btrfs_defrag_root(root->fs_info->extent_root, 0);
f46b5a66
CH		 2213 break;
2214 case S_IFREG:
e441d54d
CM		 2215 if (!(file->f_mode & FMODE_WRITE)) {
2216 ret = -EINVAL;
2217 goto out;
2218 }
1e701a32
CM		 2219
2220 range = kzalloc(sizeof(*range), GFP_KERNEL);
2221 if (!range) {
2222 ret = -ENOMEM;
2223 goto out;
2224 }
2225
2226 if (argp) {
2227 if (copy_from_user(range, argp,
2228 sizeof(*range))) {
2229 ret = -EFAULT;
2230 kfree(range);
683be16e 2231 goto out;
1e701a32
CM		 2232 }
2233 /* compression requires us to start the IO */
2234 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2235 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2236 range->extent_thresh = (u32)-1;
2237 }
2238 } else {
2239 /* the rest are all set to zero by kzalloc */
2240 range->len = (u64)-1;
2241 }
4cb5300b
CM		 2242 ret = btrfs_defrag_file(fdentry(file)->d_inode, file,
2243 range, 0, 0);
2244 if (ret > 0)
2245 ret = 0;
1e701a32 2246 kfree(range);
f46b5a66 2247 break;
8929ecfa
YZ		 2248 default:
2249 ret = -EINVAL;
f46b5a66 2250 }
e441d54d 2251out:
2a79f17e 2252 mnt_drop_write_file(file);
5ac00add 2253 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
e441d54d 2254 return ret;
f46b5a66
CH		 2255}
2256
b2950863 2257static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
f46b5a66
CH		 2258{
2259 struct btrfs_ioctl_vol_args *vol_args;
2260 int ret;
2261
e441d54d
CM		 2262 if (!capable(CAP_SYS_ADMIN))
2263 return -EPERM;
2264
5ac00add
SB		 2265 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2266 1)) {
2267 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2268 return -EINPROGRESS;
c9e9f97b
ID		 2269 }
2270
5ac00add 2271 mutex_lock(&root->fs_info->volume_mutex);
dae7b665 2272 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 2273 if (IS_ERR(vol_args)) {
2274 ret = PTR_ERR(vol_args);
2275 goto out;
2276 }
f46b5a66 2277
5516e595 2278 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66
CH		 2279 ret = btrfs_init_new_device(root, vol_args->name);
2280
f46b5a66 2281 kfree(vol_args);
c9e9f97b
ID		 2282out:
2283 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 2284 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
f46b5a66
CH		 2285 return ret;
2286}
2287
da24927b 2288static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
f46b5a66 2289{
da24927b 2290 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
f46b5a66
CH		 2291 struct btrfs_ioctl_vol_args *vol_args;
2292 int ret;
2293
e441d54d
CM		 2294 if (!capable(CAP_SYS_ADMIN))
2295 return -EPERM;
2296
da24927b
MX		 2297 ret = mnt_want_write_file(file);
2298 if (ret)
2299 return ret;
c146afad 2300
5ac00add
SB		 2301 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2302 1)) {
2303 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
da24927b 2304 mnt_drop_write_file(file);
5ac00add 2305 return -EINPROGRESS;
c9e9f97b
ID		 2306 }
2307
5ac00add 2308 mutex_lock(&root->fs_info->volume_mutex);
dae7b665 2309 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 2310 if (IS_ERR(vol_args)) {
2311 ret = PTR_ERR(vol_args);
2312 goto out;
2313 }
f46b5a66 2314
5516e595 2315 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66
CH		 2316 ret = btrfs_rm_device(root, vol_args->name);
2317
f46b5a66 2318 kfree(vol_args);
c9e9f97b
ID		 2319out:
2320 mutex_unlock(&root->fs_info->volume_mutex);
da24927b 2321 mnt_drop_write_file(file);
5ac00add 2322 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
f46b5a66
CH		 2323 return ret;
2324}
2325
475f6387
JS		 2326static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2327{
027ed2f0 2328 struct btrfs_ioctl_fs_info_args *fi_args;
475f6387
JS		 2329 struct btrfs_device *device;
2330 struct btrfs_device *next;
2331 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
027ed2f0 2332 int ret = 0;
475f6387
JS		 2333
2334 if (!capable(CAP_SYS_ADMIN))
2335 return -EPERM;
2336
027ed2f0
LZ		 2337 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2338 if (!fi_args)
2339 return -ENOMEM;
2340
2341 fi_args->num_devices = fs_devices->num_devices;
2342 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
475f6387
JS		 2343
2344 mutex_lock(&fs_devices->device_list_mutex);
2345 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
027ed2f0
LZ		 2346 if (device->devid > fi_args->max_id)
2347 fi_args->max_id = device->devid;
475f6387
JS		 2348 }
2349 mutex_unlock(&fs_devices->device_list_mutex);
2350
027ed2f0
LZ		 2351 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2352 ret = -EFAULT;
475f6387 2353
027ed2f0
LZ		 2354 kfree(fi_args);
2355 return ret;
475f6387
JS		 2356}
2357
2358static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2359{
2360 struct btrfs_ioctl_dev_info_args *di_args;
2361 struct btrfs_device *dev;
2362 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2363 int ret = 0;
2364 char *s_uuid = NULL;
2365 char empty_uuid[BTRFS_UUID_SIZE] = {0};
2366
2367 if (!capable(CAP_SYS_ADMIN))
2368 return -EPERM;
2369
2370 di_args = memdup_user(arg, sizeof(*di_args));
2371 if (IS_ERR(di_args))
2372 return PTR_ERR(di_args);
2373
2374 if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2375 s_uuid = di_args->uuid;
2376
2377 mutex_lock(&fs_devices->device_list_mutex);
aa1b8cd4 2378 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
475f6387
JS		 2379 mutex_unlock(&fs_devices->device_list_mutex);
2380
2381 if (!dev) {
2382 ret = -ENODEV;
2383 goto out;
2384 }
2385
2386 di_args->devid = dev->devid;
2387 di_args->bytes_used = dev->bytes_used;
2388 di_args->total_bytes = dev->total_bytes;
2389 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
a27202fb 2390 if (dev->name) {
606686ee
JB		 2391 struct rcu_string *name;
2392
2393 rcu_read_lock();
2394 name = rcu_dereference(dev->name);
2395 strncpy(di_args->path, name->str, sizeof(di_args->path));
2396 rcu_read_unlock();
a27202fb
JM		 2397 di_args->path[sizeof(di_args->path) - 1] = 0;
2398 } else {
99ba55ad 2399 di_args->path[0] = '\0';
a27202fb 2400 }
475f6387
JS		 2401
2402out:
2403 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2404 ret = -EFAULT;
2405
2406 kfree(di_args);
2407 return ret;
2408}
2409
76dda93c
YZ		 2410static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
2411 u64 off, u64 olen, u64 destoff)
f46b5a66
CH		 2412{
2413 struct inode *inode = fdentry(file)->d_inode;
2414 struct btrfs_root *root = BTRFS_I(inode)->root;
2903ff01 2415 struct fd src_file;
f46b5a66
CH		 2416 struct inode *src;
2417 struct btrfs_trans_handle *trans;
f46b5a66 2418 struct btrfs_path *path;
f46b5a66 2419 struct extent_buffer *leaf;
ae01a0ab
YZ		 2420 char *buf;
2421 struct btrfs_key key;
f46b5a66
CH		 2422 u32 nritems;
2423 int slot;
ae01a0ab 2424 int ret;
c5c9cd4d
SW		 2425 u64 len = olen;
2426 u64 bs = root->fs_info->sb->s_blocksize;
d20f7043 2427
c5c9cd4d
SW		 2428 /*
2429 * TODO:
2430 * - split compressed inline extents. annoying: we need to
2431 * decompress into destination's address_space (the file offset
2432 * may change, so source mapping won't do), then recompress (or
2433 * otherwise reinsert) a subrange.
2434 * - allow ranges within the same file to be cloned (provided
2435 * they don't overlap)?
2436 */
2437
e441d54d 2438 /* the destination must be opened for writing */
2ebc3464 2439 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
e441d54d
CM		 2440 return -EINVAL;
2441
b83cc969
LZ		 2442 if (btrfs_root_readonly(root))
2443 return -EROFS;
2444
a561be71 2445 ret = mnt_want_write_file(file);
c146afad
YZ		 2446 if (ret)
2447 return ret;
2448
2903ff01
AV		 2449 src_file = fdget(srcfd);
2450 if (!src_file.file) {
ab67b7c1
YZ		 2451 ret = -EBADF;
2452 goto out_drop_write;
2453 }
5dc64164 2454
362a20c5 2455 ret = -EXDEV;
2903ff01 2456 if (src_file.file->f_path.mnt != file->f_path.mnt)
362a20c5
DS		 2457 goto out_fput;
2458
2903ff01 2459 src = src_file.file->f_dentry->d_inode;
f46b5a66 2460
c5c9cd4d
SW		 2461 ret = -EINVAL;
2462 if (src == inode)
2463 goto out_fput;
2464
5dc64164 2465 /* the src must be open for reading */
2903ff01 2466 if (!(src_file.file->f_mode & FMODE_READ))
5dc64164
DR		 2467 goto out_fput;
2468
0e7b824c
LZ		 2469 /* don't make the dst file partly checksummed */
2470 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2471 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
2472 goto out_fput;
2473
ae01a0ab
YZ		 2474 ret = -EISDIR;
2475 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
2476 goto out_fput;
2477
f46b5a66 2478 ret = -EXDEV;
362a20c5 2479 if (src->i_sb != inode->i_sb)
ae01a0ab
YZ		 2480 goto out_fput;
2481
2482 ret = -ENOMEM;
2483 buf = vmalloc(btrfs_level_size(root, 0));
2484 if (!buf)
2485 goto out_fput;
2486
2487 path = btrfs_alloc_path();
2488 if (!path) {
2489 vfree(buf);
f46b5a66 2490 goto out_fput;
ae01a0ab
YZ		 2491 }
2492 path->reada = 2;
f46b5a66
CH		 2493
2494 if (inode < src) {
fccdae43
SW		 2495 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
2496 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
f46b5a66 2497 } else {
fccdae43
SW		 2498 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
2499 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
f46b5a66
CH		 2500 }
2501
c5c9cd4d
SW		 2502 /* determine range to clone */
2503 ret = -EINVAL;
2ebc3464 2504 if (off + len > src->i_size || off + len < off)
f46b5a66 2505 goto out_unlock;
c5c9cd4d
SW		 2506 if (len == 0)
2507 olen = len = src->i_size - off;
2508 /* if we extend to eof, continue to block boundary */
2509 if (off + len == src->i_size)
2a6b8dae 2510 len = ALIGN(src->i_size, bs) - off;
c5c9cd4d
SW		 2511
2512 /* verify the end result is block aligned */
2a6b8dae
LZ		 2513 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
2514 !IS_ALIGNED(destoff, bs))
c5c9cd4d
SW		 2515 goto out_unlock;
2516
d525e8ab
LZ		 2517 if (destoff > inode->i_size) {
2518 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
2519 if (ret)
2520 goto out_unlock;
2521 }
2522
71ef0786
LZ		 2523 /* truncate page cache pages from target inode range */
2524 truncate_inode_pages_range(&inode->i_data, destoff,
2525 PAGE_CACHE_ALIGN(destoff + len) - 1);
2526
f46b5a66
CH		 2527 /* do any pending delalloc/csum calc on src, one way or
2528 another, and lock file content */
2529 while (1) {
31840ae1 2530 struct btrfs_ordered_extent *ordered;
aa42ffd9
LB		 2531 lock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
2532 ordered = btrfs_lookup_first_ordered_extent(src, off + len - 1);
9a019196 2533 if (!ordered &&
aa42ffd9
LB		 2534 !test_range_bit(&BTRFS_I(src)->io_tree, off, off + len - 1,
2535 EXTENT_DELALLOC, 0, NULL))
f46b5a66 2536 break;
aa42ffd9 2537 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
ae01a0ab
YZ		 2538 if (ordered)
2539 btrfs_put_ordered_extent(ordered);
9a019196 2540 btrfs_wait_ordered_range(src, off, len);
f46b5a66
CH		 2541 }
2542
c5c9cd4d 2543 /* clone data */
33345d01 2544 key.objectid = btrfs_ino(src);
ae01a0ab
YZ		 2545 key.type = BTRFS_EXTENT_DATA_KEY;
2546 key.offset = 0;
f46b5a66
CH		 2547
2548 while (1) {
2549 /*
2550 * note the key will change type as we walk through the
2551 * tree.
2552 */
362a20c5
DS		 2553 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
2554 0, 0);
f46b5a66
CH		 2555 if (ret < 0)
2556 goto out;
2557
ae01a0ab
YZ		 2558 nritems = btrfs_header_nritems(path->nodes[0]);
2559 if (path->slots[0] >= nritems) {
362a20c5 2560 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
f46b5a66
CH		 2561 if (ret < 0)
2562 goto out;
2563 if (ret > 0)
2564 break;
ae01a0ab 2565 nritems = btrfs_header_nritems(path->nodes[0]);
f46b5a66
CH		 2566 }
2567 leaf = path->nodes[0];
2568 slot = path->slots[0];
f46b5a66 2569
ae01a0ab 2570 btrfs_item_key_to_cpu(leaf, &key, slot);
d20f7043 2571 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
33345d01 2572 key.objectid != btrfs_ino(src))
f46b5a66
CH		 2573 break;
2574
c5c9cd4d
SW		 2575 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2576 struct btrfs_file_extent_item *extent;
2577 int type;
31840ae1
ZY		 2578 u32 size;
2579 struct btrfs_key new_key;
c5c9cd4d
SW		 2580 u64 disko = 0, diskl = 0;
2581 u64 datao = 0, datal = 0;
2582 u8 comp;
b5384d48 2583 u64 endoff;
31840ae1
ZY		 2584
2585 size = btrfs_item_size_nr(leaf, slot);
2586 read_extent_buffer(leaf, buf,
2587 btrfs_item_ptr_offset(leaf, slot),
2588 size);
c5c9cd4d
SW		 2589
2590 extent = btrfs_item_ptr(leaf, slot,
2591 struct btrfs_file_extent_item);
2592 comp = btrfs_file_extent_compression(leaf, extent);
2593 type = btrfs_file_extent_type(leaf, extent);
c8a894d7
CM		 2594 if (type == BTRFS_FILE_EXTENT_REG ||
2595 type == BTRFS_FILE_EXTENT_PREALLOC) {
d397712b
CM		 2596 disko = btrfs_file_extent_disk_bytenr(leaf,
2597 extent);
2598 diskl = btrfs_file_extent_disk_num_bytes(leaf,
2599 extent);
c5c9cd4d 2600 datao = btrfs_file_extent_offset(leaf, extent);
d397712b
CM		 2601 datal = btrfs_file_extent_num_bytes(leaf,
2602 extent);
c5c9cd4d
SW		 2603 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2604 /* take upper bound, may be compressed */
2605 datal = btrfs_file_extent_ram_bytes(leaf,
2606 extent);
2607 }
b3b4aa74 2608 btrfs_release_path(path);
31840ae1 2609
050006a7 2610 if (key.offset + datal <= off ||
aa42ffd9 2611 key.offset >= off + len - 1)
c5c9cd4d
SW		 2612 goto next;
2613
31840ae1 2614 memcpy(&new_key, &key, sizeof(new_key));
33345d01 2615 new_key.objectid = btrfs_ino(inode);
4d728ec7
LZ		 2616 if (off <= key.offset)
2617 new_key.offset = key.offset + destoff - off;
2618 else
2619 new_key.offset = destoff;
31840ae1 2620
b6f3409b
SW		 2621 /*
2622 * 1 - adjusting old extent (we may have to split it)
2623 * 1 - add new extent
2624 * 1 - inode update
2625 */
2626 trans = btrfs_start_transaction(root, 3);
a22285a6
YZ		 2627 if (IS_ERR(trans)) {
2628 ret = PTR_ERR(trans);
2629 goto out;
2630 }
2631
c8a894d7
CM		 2632 if (type == BTRFS_FILE_EXTENT_REG ||
2633 type == BTRFS_FILE_EXTENT_PREALLOC) {
d72c0842
LZ		 2634 /*
2635 * a | --- range to clone ---| b
2636 * | ------------- extent ------------- |
2637 */
2638
2639 /* substract range b */
2640 if (key.offset + datal > off + len)
2641 datal = off + len - key.offset;
2642
2643 /* substract range a */
a22285a6
YZ		 2644 if (off > key.offset) {
2645 datao += off - key.offset;
2646 datal -= off - key.offset;
2647 }
2648
5dc562c5 2649 ret = btrfs_drop_extents(trans, root, inode,
a22285a6
YZ		 2650 new_key.offset,
2651 new_key.offset + datal,
2671485d 2652 1);
79787eaa
JM		 2653 if (ret) {
2654 btrfs_abort_transaction(trans, root,
2655 ret);
2656 btrfs_end_transaction(trans, root);
2657 goto out;
2658 }
a22285a6 2659
c5c9cd4d
SW		 2660 ret = btrfs_insert_empty_item(trans, root, path,
2661 &new_key, size);
79787eaa
JM		 2662 if (ret) {
2663 btrfs_abort_transaction(trans, root,
2664 ret);
2665 btrfs_end_transaction(trans, root);
2666 goto out;
2667 }
c5c9cd4d
SW		 2668
2669 leaf = path->nodes[0];
2670 slot = path->slots[0];
2671 write_extent_buffer(leaf, buf,
31840ae1
ZY		 2672 btrfs_item_ptr_offset(leaf, slot),
2673 size);
ae01a0ab 2674
c5c9cd4d 2675 extent = btrfs_item_ptr(leaf, slot,
f46b5a66 2676 struct btrfs_file_extent_item);
c5c9cd4d 2677
c5c9cd4d
SW		 2678 /* disko == 0 means it's a hole */
2679 if (!disko)
2680 datao = 0;
c5c9cd4d
SW		 2681
2682 btrfs_set_file_extent_offset(leaf, extent,
2683 datao);
2684 btrfs_set_file_extent_num_bytes(leaf, extent,
2685 datal);
2686 if (disko) {
2687 inode_add_bytes(inode, datal);
ae01a0ab 2688 ret = btrfs_inc_extent_ref(trans, root,
5d4f98a2
YZ		 2689 disko, diskl, 0,
2690 root->root_key.objectid,
33345d01 2691 btrfs_ino(inode),
66d7e7f0
AJ		 2692 new_key.offset - datao,
2693 0);
79787eaa
JM		 2694 if (ret) {
2695 btrfs_abort_transaction(trans,
2696 root,
2697 ret);
2698 btrfs_end_transaction(trans,
2699 root);
2700 goto out;
2701
2702 }
f46b5a66 2703 }
c5c9cd4d
SW		 2704 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2705 u64 skip = 0;
2706 u64 trim = 0;
2707 if (off > key.offset) {
2708 skip = off - key.offset;
2709 new_key.offset += skip;
2710 }
d397712b 2711
aa42ffd9
LB		 2712 if (key.offset + datal > off + len)
2713 trim = key.offset + datal - (off + len);
d397712b 2714
c5c9cd4d 2715 if (comp && (skip || trim)) {
c5c9cd4d 2716 ret = -EINVAL;
a22285a6 2717 btrfs_end_transaction(trans, root);
c5c9cd4d
SW		 2718 goto out;
2719 }
2720 size -= skip + trim;
2721 datal -= skip + trim;
a22285a6 2722
5dc562c5 2723 ret = btrfs_drop_extents(trans, root, inode,
a22285a6
YZ		 2724 new_key.offset,
2725 new_key.offset + datal,
2671485d 2726 1);
79787eaa
JM		 2727 if (ret) {
2728 btrfs_abort_transaction(trans, root,
2729 ret);
2730 btrfs_end_transaction(trans, root);
2731 goto out;
2732 }
a22285a6 2733
c5c9cd4d
SW		 2734 ret = btrfs_insert_empty_item(trans, root, path,
2735 &new_key, size);
79787eaa
JM		 2736 if (ret) {
2737 btrfs_abort_transaction(trans, root,
2738 ret);
2739 btrfs_end_transaction(trans, root);
2740 goto out;
2741 }
c5c9cd4d
SW		 2742
2743 if (skip) {
d397712b
CM		 2744 u32 start =
2745 btrfs_file_extent_calc_inline_size(0);
c5c9cd4d
SW		 2746 memmove(buf+start, buf+start+skip,
2747 datal);
2748 }
2749
2750 leaf = path->nodes[0];
2751 slot = path->slots[0];
2752 write_extent_buffer(leaf, buf,
2753 btrfs_item_ptr_offset(leaf, slot),
2754 size);
2755 inode_add_bytes(inode, datal);
f46b5a66 2756 }
c5c9cd4d
SW		 2757
2758 btrfs_mark_buffer_dirty(leaf);
b3b4aa74 2759 btrfs_release_path(path);
c5c9cd4d 2760
0c4d2d95 2761 inode_inc_iversion(inode);
a22285a6 2762 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
b5384d48
SW		 2763
2764 /*
2765 * we round up to the block size at eof when
2766 * determining which extents to clone above,
2767 * but shouldn't round up the file size
2768 */
2769 endoff = new_key.offset + datal;
5f3888ff
LZ		 2770 if (endoff > destoff+olen)
2771 endoff = destoff+olen;
b5384d48
SW		 2772 if (endoff > inode->i_size)
2773 btrfs_i_size_write(inode, endoff);
2774
a22285a6 2775 ret = btrfs_update_inode(trans, root, inode);
79787eaa
JM		 2776 if (ret) {
2777 btrfs_abort_transaction(trans, root, ret);
2778 btrfs_end_transaction(trans, root);
2779 goto out;
2780 }
2781 ret = btrfs_end_transaction(trans, root);
a22285a6 2782 }
d397712b 2783next:
b3b4aa74 2784 btrfs_release_path(path);
f46b5a66 2785 key.offset++;
f46b5a66 2786 }
f46b5a66
CH		 2787 ret = 0;
2788out:
b3b4aa74 2789 btrfs_release_path(path);
aa42ffd9 2790 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
f46b5a66
CH		 2791out_unlock:
2792 mutex_unlock(&src->i_mutex);
2793 mutex_unlock(&inode->i_mutex);
ae01a0ab
YZ		 2794 vfree(buf);
2795 btrfs_free_path(path);
f46b5a66 2796out_fput:
2903ff01 2797 fdput(src_file);
ab67b7c1 2798out_drop_write:
2a79f17e 2799 mnt_drop_write_file(file);
f46b5a66
CH		 2800 return ret;
2801}
2802
7a865e8a 2803static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
c5c9cd4d
SW		 2804{
2805 struct btrfs_ioctl_clone_range_args args;
2806
7a865e8a 2807 if (copy_from_user(&args, argp, sizeof(args)))
c5c9cd4d
SW		 2808 return -EFAULT;
2809 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
2810 args.src_length, args.dest_offset);
2811}
2812
f46b5a66
CH		 2813/*
2814 * there are many ways the trans_start and trans_end ioctls can lead
2815 * to deadlocks. They should only be used by applications that
2816 * basically own the machine, and have a very in depth understanding
2817 * of all the possible deadlocks and enospc problems.
2818 */
b2950863 2819static long btrfs_ioctl_trans_start(struct file *file)
f46b5a66
CH		 2820{
2821 struct inode *inode = fdentry(file)->d_inode;
2822 struct btrfs_root *root = BTRFS_I(inode)->root;
2823 struct btrfs_trans_handle *trans;
1ab86aed 2824 int ret;
f46b5a66 2825
1ab86aed 2826 ret = -EPERM;
df5b5520 2827 if (!capable(CAP_SYS_ADMIN))
1ab86aed 2828 goto out;
df5b5520 2829
1ab86aed
SW		 2830 ret = -EINPROGRESS;
2831 if (file->private_data)
f46b5a66 2832 goto out;
9ca9ee09 2833
b83cc969
LZ		 2834 ret = -EROFS;
2835 if (btrfs_root_readonly(root))
2836 goto out;
2837
a561be71 2838 ret = mnt_want_write_file(file);
c146afad
YZ		 2839 if (ret)
2840 goto out;
2841
a4abeea4 2842 atomic_inc(&root->fs_info->open_ioctl_trans);
9ca9ee09 2843
1ab86aed 2844 ret = -ENOMEM;
7a7eaa40 2845 trans = btrfs_start_ioctl_transaction(root);
abd30bb0 2846 if (IS_ERR(trans))
1ab86aed
SW		 2847 goto out_drop;
2848
2849 file->private_data = trans;
2850 return 0;
2851
2852out_drop:
a4abeea4 2853 atomic_dec(&root->fs_info->open_ioctl_trans);
2a79f17e 2854 mnt_drop_write_file(file);
f46b5a66 2855out:
f46b5a66
CH		 2856 return ret;
2857}
2858
6ef5ed0d
JB		 2859static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
2860{
2861 struct inode *inode = fdentry(file)->d_inode;
2862 struct btrfs_root *root = BTRFS_I(inode)->root;
2863 struct btrfs_root *new_root;
2864 struct btrfs_dir_item *di;
2865 struct btrfs_trans_handle *trans;
2866 struct btrfs_path *path;
2867 struct btrfs_key location;
2868 struct btrfs_disk_key disk_key;
6ef5ed0d
JB		 2869 u64 objectid = 0;
2870 u64 dir_id;
3c04ce01 2871 int ret;
6ef5ed0d
JB		 2872
2873 if (!capable(CAP_SYS_ADMIN))
2874 return -EPERM;
2875
3c04ce01
MX		 2876 ret = mnt_want_write_file(file);
2877 if (ret)
2878 return ret;
2879
2880 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
2881 ret = -EFAULT;
2882 goto out;
2883 }
6ef5ed0d
JB		 2884
2885 if (!objectid)
2886 objectid = root->root_key.objectid;
2887
2888 location.objectid = objectid;
2889 location.type = BTRFS_ROOT_ITEM_KEY;
2890 location.offset = (u64)-1;
2891
2892 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
3c04ce01
MX		 2893 if (IS_ERR(new_root)) {
2894 ret = PTR_ERR(new_root);
2895 goto out;
2896 }
6ef5ed0d 2897
3c04ce01
MX		 2898 if (btrfs_root_refs(&new_root->root_item) == 0) {
2899 ret = -ENOENT;
2900 goto out;
2901 }
6ef5ed0d
JB		 2902
2903 path = btrfs_alloc_path();
3c04ce01
MX		 2904 if (!path) {
2905 ret = -ENOMEM;
2906 goto out;
2907 }
6ef5ed0d
JB		 2908 path->leave_spinning = 1;
2909
2910 trans = btrfs_start_transaction(root, 1);
98d5dc13 2911 if (IS_ERR(trans)) {
6ef5ed0d 2912 btrfs_free_path(path);
3c04ce01
MX		 2913 ret = PTR_ERR(trans);
2914 goto out;
6ef5ed0d
JB		 2915 }
2916
6c41761f 2917 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
6ef5ed0d
JB		 2918 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
2919 dir_id, "default", 7, 1);
cf1e99a4 2920 if (IS_ERR_OR_NULL(di)) {
6ef5ed0d
JB		 2921 btrfs_free_path(path);
2922 btrfs_end_transaction(trans, root);
2923 printk(KERN_ERR "Umm, you don't have the default dir item, "
2924 "this isn't going to work\n");
3c04ce01
MX		 2925 ret = -ENOENT;
2926 goto out;
6ef5ed0d
JB		 2927 }
2928
2929 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
2930 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
2931 btrfs_mark_buffer_dirty(path->nodes[0]);
2932 btrfs_free_path(path);
2933
2b0ce2c2 2934 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
6ef5ed0d 2935 btrfs_end_transaction(trans, root);
3c04ce01
MX		 2936out:
2937 mnt_drop_write_file(file);
2938 return ret;
6ef5ed0d
JB		 2939}
2940
5af3e8cc
SB		 2941void btrfs_get_block_group_info(struct list_head *groups_list,
2942 struct btrfs_ioctl_space_info *space)
bf5fc093
JB		 2943{
2944 struct btrfs_block_group_cache *block_group;
2945
2946 space->total_bytes = 0;
2947 space->used_bytes = 0;
2948 space->flags = 0;
2949 list_for_each_entry(block_group, groups_list, list) {
2950 space->flags = block_group->flags;
2951 space->total_bytes += block_group->key.offset;
2952 space->used_bytes +=
2953 btrfs_block_group_used(&block_group->item);
2954 }
2955}
2956
1406e432
JB		 2957long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
2958{
2959 struct btrfs_ioctl_space_args space_args;
2960 struct btrfs_ioctl_space_info space;
2961 struct btrfs_ioctl_space_info *dest;
7fde62bf 2962 struct btrfs_ioctl_space_info *dest_orig;
13f2696f 2963 struct btrfs_ioctl_space_info __user *user_dest;
1406e432 2964 struct btrfs_space_info *info;
bf5fc093
JB		 2965 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
2966 BTRFS_BLOCK_GROUP_SYSTEM,
2967 BTRFS_BLOCK_GROUP_METADATA,
2968 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
2969 int num_types = 4;
7fde62bf 2970 int alloc_size;
1406e432 2971 int ret = 0;
51788b1b 2972 u64 slot_count = 0;
bf5fc093 2973 int i, c;
1406e432
JB		 2974
2975 if (copy_from_user(&space_args,
2976 (struct btrfs_ioctl_space_args __user *)arg,
2977 sizeof(space_args)))
2978 return -EFAULT;
2979
bf5fc093
JB		 2980 for (i = 0; i < num_types; i++) {
2981 struct btrfs_space_info *tmp;
2982
2983 info = NULL;
2984 rcu_read_lock();
2985 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
2986 list) {
2987 if (tmp->flags == types[i]) {
2988 info = tmp;
2989 break;
2990 }
2991 }
2992 rcu_read_unlock();
2993
2994 if (!info)
2995 continue;
2996
2997 down_read(&info->groups_sem);
2998 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
2999 if (!list_empty(&info->block_groups[c]))
3000 slot_count++;
3001 }
3002 up_read(&info->groups_sem);
3003 }
7fde62bf
CM		 3004
3005 /* space_slots == 0 means they are asking for a count */
3006 if (space_args.space_slots == 0) {
3007 space_args.total_spaces = slot_count;
3008 goto out;
3009 }
bf5fc093 3010
51788b1b 3011 slot_count = min_t(u64, space_args.space_slots, slot_count);
bf5fc093 3012
7fde62bf 3013 alloc_size = sizeof(*dest) * slot_count;
bf5fc093 3014
7fde62bf
CM		 3015 /* we generally have at most 6 or so space infos, one for each raid
3016 * level. So, a whole page should be more than enough for everyone
3017 */
3018 if (alloc_size > PAGE_CACHE_SIZE)
3019 return -ENOMEM;
3020
1406e432 3021 space_args.total_spaces = 0;
7fde62bf
CM		 3022 dest = kmalloc(alloc_size, GFP_NOFS);
3023 if (!dest)
3024 return -ENOMEM;
3025 dest_orig = dest;
1406e432 3026
7fde62bf 3027 /* now we have a buffer to copy into */
bf5fc093
JB		 3028 for (i = 0; i < num_types; i++) {
3029 struct btrfs_space_info *tmp;
3030
51788b1b
DR		 3031 if (!slot_count)
3032 break;
3033
bf5fc093
JB		 3034 info = NULL;
3035 rcu_read_lock();
3036 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3037 list) {
3038 if (tmp->flags == types[i]) {
3039 info = tmp;
3040 break;
3041 }
3042 }
3043 rcu_read_unlock();
7fde62bf 3044
bf5fc093
JB		 3045 if (!info)
3046 continue;
3047 down_read(&info->groups_sem);
3048 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3049 if (!list_empty(&info->block_groups[c])) {
5af3e8cc
SB		 3050 btrfs_get_block_group_info(
3051 &info->block_groups[c], &space);
bf5fc093
JB		 3052 memcpy(dest, &space, sizeof(space));
3053 dest++;
3054 space_args.total_spaces++;
51788b1b 3055 slot_count--;
bf5fc093 3056 }
51788b1b
DR		 3057 if (!slot_count)
3058 break;
bf5fc093
JB		 3059 }
3060 up_read(&info->groups_sem);
1406e432 3061 }
1406e432 3062
2eec6c81 3063 user_dest = (struct btrfs_ioctl_space_info __user *)
7fde62bf
CM		 3064 (arg + sizeof(struct btrfs_ioctl_space_args));
3065
3066 if (copy_to_user(user_dest, dest_orig, alloc_size))
3067 ret = -EFAULT;
3068
3069 kfree(dest_orig);
3070out:
3071 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
1406e432
JB		 3072 ret = -EFAULT;
3073
3074 return ret;
3075}
3076
f46b5a66
CH		 3077/*
3078 * there are many ways the trans_start and trans_end ioctls can lead
3079 * to deadlocks. They should only be used by applications that
3080 * basically own the machine, and have a very in depth understanding
3081 * of all the possible deadlocks and enospc problems.
3082 */
3083long btrfs_ioctl_trans_end(struct file *file)
3084{
3085 struct inode *inode = fdentry(file)->d_inode;
3086 struct btrfs_root *root = BTRFS_I(inode)->root;
3087 struct btrfs_trans_handle *trans;
f46b5a66 3088
f46b5a66 3089 trans = file->private_data;
1ab86aed
SW		 3090 if (!trans)
3091 return -EINVAL;
b214107e 3092 file->private_data = NULL;
9ca9ee09 3093
1ab86aed
SW		 3094 btrfs_end_transaction(trans, root);
3095
a4abeea4 3096 atomic_dec(&root->fs_info->open_ioctl_trans);
9ca9ee09 3097
2a79f17e 3098 mnt_drop_write_file(file);
1ab86aed 3099 return 0;
f46b5a66
CH		 3100}
3101
9a8c28be
MX		 3102static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3103 void __user *argp)
46204592 3104{
46204592
SW		 3105 struct btrfs_trans_handle *trans;
3106 u64 transid;
db5b493a 3107 int ret;
46204592 3108
ff7c1d33
MX		 3109 trans = btrfs_attach_transaction(root);
3110 if (IS_ERR(trans)) {
3111 if (PTR_ERR(trans) != -ENOENT)
3112 return PTR_ERR(trans);
3113
3114 /* No running transaction, don't bother */
3115 transid = root->fs_info->last_trans_committed;
3116 goto out;
3117 }
46204592 3118 transid = trans->transid;
db5b493a 3119 ret = btrfs_commit_transaction_async(trans, root, 0);
8b2b2d3c
TI		 3120 if (ret) {
3121 btrfs_end_transaction(trans, root);
db5b493a 3122 return ret;
8b2b2d3c 3123 }
ff7c1d33 3124out:
46204592
SW		 3125 if (argp)
3126 if (copy_to_user(argp, &transid, sizeof(transid)))
3127 return -EFAULT;
3128 return 0;
3129}
3130
9a8c28be
MX		 3131static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
3132 void __user *argp)
46204592 3133{
46204592
SW		 3134 u64 transid;
3135
3136 if (argp) {
3137 if (copy_from_user(&transid, argp, sizeof(transid)))
3138 return -EFAULT;
3139 } else {
3140 transid = 0; /* current trans */
3141 }
3142 return btrfs_wait_for_commit(root, transid);
3143}
3144
b8e95489 3145static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
475f6387 3146{
b8e95489 3147 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
475f6387 3148 struct btrfs_ioctl_scrub_args *sa;
b8e95489 3149 int ret;
475f6387
JS		 3150
3151 if (!capable(CAP_SYS_ADMIN))
3152 return -EPERM;
3153
3154 sa = memdup_user(arg, sizeof(*sa));
3155 if (IS_ERR(sa))
3156 return PTR_ERR(sa);
3157
b8e95489
MX		 3158 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3159 ret = mnt_want_write_file(file);
3160 if (ret)
3161 goto out;
3162 }
3163
aa1b8cd4 3164 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
63a212ab
SB		 3165 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3166 0);
475f6387
JS		 3167
3168 if (copy_to_user(arg, sa, sizeof(*sa)))
3169 ret = -EFAULT;
3170
b8e95489
MX		 3171 if (!(sa->flags & BTRFS_SCRUB_READONLY))
3172 mnt_drop_write_file(file);
3173out:
475f6387
JS		 3174 kfree(sa);
3175 return ret;
3176}
3177
3178static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3179{
3180 if (!capable(CAP_SYS_ADMIN))
3181 return -EPERM;
3182
aa1b8cd4 3183 return btrfs_scrub_cancel(root->fs_info);
475f6387
JS		 3184}
3185
3186static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3187 void __user *arg)
3188{
3189 struct btrfs_ioctl_scrub_args *sa;
3190 int ret;
3191
3192 if (!capable(CAP_SYS_ADMIN))
3193 return -EPERM;
3194
3195 sa = memdup_user(arg, sizeof(*sa));
3196 if (IS_ERR(sa))
3197 return PTR_ERR(sa);
3198
3199 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3200
3201 if (copy_to_user(arg, sa, sizeof(*sa)))
3202 ret = -EFAULT;
3203
3204 kfree(sa);
3205 return ret;
3206}
3207
c11d2c23 3208static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
b27f7c0c 3209 void __user *arg)
c11d2c23
SB		 3210{
3211 struct btrfs_ioctl_get_dev_stats *sa;
3212 int ret;
3213
c11d2c23
SB		 3214 sa = memdup_user(arg, sizeof(*sa));
3215 if (IS_ERR(sa))
3216 return PTR_ERR(sa);
3217
b27f7c0c
DS		 3218 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3219 kfree(sa);
3220 return -EPERM;
3221 }
3222
3223 ret = btrfs_get_dev_stats(root, sa);
c11d2c23
SB		 3224
3225 if (copy_to_user(arg, sa, sizeof(*sa)))
3226 ret = -EFAULT;
3227
3228 kfree(sa);
3229 return ret;
3230}
3231
3f6bcfbd
SB		 3232static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
3233{
3234 struct btrfs_ioctl_dev_replace_args *p;
3235 int ret;
3236
3237 if (!capable(CAP_SYS_ADMIN))
3238 return -EPERM;
3239
3240 p = memdup_user(arg, sizeof(*p));
3241 if (IS_ERR(p))
3242 return PTR_ERR(p);
3243
3244 switch (p->cmd) {
3245 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
3246 if (atomic_xchg(
3247 &root->fs_info->mutually_exclusive_operation_running,
3248 1)) {
3249 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3250 ret = -EINPROGRESS;
3251 } else {
3252 ret = btrfs_dev_replace_start(root, p);
3253 atomic_set(
3254 &root->fs_info->mutually_exclusive_operation_running,
3255 0);
3256 }
3257 break;
3258 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3259 btrfs_dev_replace_status(root->fs_info, p);
3260 ret = 0;
3261 break;
3262 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3263 ret = btrfs_dev_replace_cancel(root->fs_info, p);
3264 break;
3265 default:
3266 ret = -EINVAL;
3267 break;
3268 }
3269
3270 if (copy_to_user(arg, p, sizeof(*p)))
3271 ret = -EFAULT;
3272
3273 kfree(p);
3274 return ret;
3275}
3276
d7728c96
JS		 3277static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3278{
3279 int ret = 0;
3280 int i;
740c3d22 3281 u64 rel_ptr;
d7728c96 3282 int size;
806468f8 3283 struct btrfs_ioctl_ino_path_args *ipa = NULL;
d7728c96
JS		 3284 struct inode_fs_paths *ipath = NULL;
3285 struct btrfs_path *path;
3286
3287 if (!capable(CAP_SYS_ADMIN))
3288 return -EPERM;
3289
3290 path = btrfs_alloc_path();
3291 if (!path) {
3292 ret = -ENOMEM;
3293 goto out;
3294 }
3295
3296 ipa = memdup_user(arg, sizeof(*ipa));
3297 if (IS_ERR(ipa)) {
3298 ret = PTR_ERR(ipa);
3299 ipa = NULL;
3300 goto out;
3301 }
3302
3303 size = min_t(u32, ipa->size, 4096);
3304 ipath = init_ipath(size, root, path);
3305 if (IS_ERR(ipath)) {
3306 ret = PTR_ERR(ipath);
3307 ipath = NULL;
3308 goto out;
3309 }
3310
3311 ret = paths_from_inode(ipa->inum, ipath);
3312 if (ret < 0)
3313 goto out;
3314
3315 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
745c4d8e
JM		 3316 rel_ptr = ipath->fspath->val[i] -
3317 (u64)(unsigned long)ipath->fspath->val;
740c3d22 3318 ipath->fspath->val[i] = rel_ptr;
d7728c96
JS		 3319 }
3320
745c4d8e
JM		 3321 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3322 (void *)(unsigned long)ipath->fspath, size);
d7728c96
JS		 3323 if (ret) {
3324 ret = -EFAULT;
3325 goto out;
3326 }
3327
3328out:
3329 btrfs_free_path(path);
3330 free_ipath(ipath);
3331 kfree(ipa);
3332
3333 return ret;
3334}
3335
3336static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3337{
3338 struct btrfs_data_container *inodes = ctx;
3339 const size_t c = 3 * sizeof(u64);
3340
3341 if (inodes->bytes_left >= c) {
3342 inodes->bytes_left -= c;
3343 inodes->val[inodes->elem_cnt] = inum;
3344 inodes->val[inodes->elem_cnt + 1] = offset;
3345 inodes->val[inodes->elem_cnt + 2] = root;
3346 inodes->elem_cnt += 3;
3347 } else {
3348 inodes->bytes_missing += c - inodes->bytes_left;
3349 inodes->bytes_left = 0;
3350 inodes->elem_missed += 3;
3351 }
3352
3353 return 0;
3354}
3355
3356static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3357 void __user *arg)
3358{
3359 int ret = 0;
3360 int size;
d7728c96
JS		 3361 struct btrfs_ioctl_logical_ino_args *loi;
3362 struct btrfs_data_container *inodes = NULL;
3363 struct btrfs_path *path = NULL;
d7728c96
JS		 3364
3365 if (!capable(CAP_SYS_ADMIN))
3366 return -EPERM;
3367
3368 loi = memdup_user(arg, sizeof(*loi));
3369 if (IS_ERR(loi)) {
3370 ret = PTR_ERR(loi);
3371 loi = NULL;
3372 goto out;
3373 }
3374
3375 path = btrfs_alloc_path();
3376 if (!path) {
3377 ret = -ENOMEM;
3378 goto out;
3379 }
3380
425d17a2 3381 size = min_t(u32, loi->size, 64 * 1024);
d7728c96
JS		 3382 inodes = init_data_container(size);
3383 if (IS_ERR(inodes)) {
3384 ret = PTR_ERR(inodes);
3385 inodes = NULL;
3386 goto out;
3387 }
3388
df031f07
LB		 3389 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
3390 build_ino_list, inodes);
3391 if (ret == -EINVAL)
d7728c96
JS		 3392 ret = -ENOENT;
3393 if (ret < 0)
3394 goto out;
3395
745c4d8e
JM		 3396 ret = copy_to_user((void *)(unsigned long)loi->inodes,
3397 (void *)(unsigned long)inodes, size);
d7728c96
JS		 3398 if (ret)
3399 ret = -EFAULT;
3400
3401out:
3402 btrfs_free_path(path);
425d17a2 3403 vfree(inodes);
d7728c96
JS		 3404 kfree(loi);
3405
3406 return ret;
3407}
3408
19a39dce 3409void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
c9e9f97b
ID		 3410 struct btrfs_ioctl_balance_args *bargs)
3411{
3412 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3413
3414 bargs->flags = bctl->flags;
3415
837d5b6e
ID		 3416 if (atomic_read(&fs_info->balance_running))
3417 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3418 if (atomic_read(&fs_info->balance_pause_req))
3419 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
a7e99c69
ID		 3420 if (atomic_read(&fs_info->balance_cancel_req))
3421 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
837d5b6e 3422
c9e9f97b
ID		 3423 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3424 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3425 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
19a39dce
ID		 3426
3427 if (lock) {
3428 spin_lock(&fs_info->balance_lock);
3429 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3430 spin_unlock(&fs_info->balance_lock);
3431 } else {
3432 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3433 }
c9e9f97b
ID		 3434}
3435
9ba1f6e4 3436static long btrfs_ioctl_balance(struct file *file, void __user *arg)
c9e9f97b 3437{
9ba1f6e4 3438 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
c9e9f97b
ID		 3439 struct btrfs_fs_info *fs_info = root->fs_info;
3440 struct btrfs_ioctl_balance_args *bargs;
3441 struct btrfs_balance_control *bctl;
3442 int ret;
5ac00add 3443 int need_to_clear_lock = 0;
c9e9f97b
ID		 3444
3445 if (!capable(CAP_SYS_ADMIN))
3446 return -EPERM;
3447
e54bfa31 3448 ret = mnt_want_write_file(file);
9ba1f6e4
LB		 3449 if (ret)
3450 return ret;
3451
c9e9f97b
ID		 3452 mutex_lock(&fs_info->volume_mutex);
3453 mutex_lock(&fs_info->balance_mutex);
3454
3455 if (arg) {
3456 bargs = memdup_user(arg, sizeof(*bargs));
3457 if (IS_ERR(bargs)) {
3458 ret = PTR_ERR(bargs);
3459 goto out;
3460 }
de322263
ID		 3461
3462 if (bargs->flags & BTRFS_BALANCE_RESUME) {
3463 if (!fs_info->balance_ctl) {
3464 ret = -ENOTCONN;
3465 goto out_bargs;
3466 }
3467
3468 bctl = fs_info->balance_ctl;
3469 spin_lock(&fs_info->balance_lock);
3470 bctl->flags |= BTRFS_BALANCE_RESUME;
3471 spin_unlock(&fs_info->balance_lock);
3472
3473 goto do_balance;
3474 }
c9e9f97b
ID		 3475 } else {
3476 bargs = NULL;
3477 }
3478
5ac00add
SB		 3479 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
3480 1)) {
3481 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
837d5b6e
ID		 3482 ret = -EINPROGRESS;
3483 goto out_bargs;
3484 }
5ac00add 3485 need_to_clear_lock = 1;
837d5b6e 3486
c9e9f97b
ID		 3487 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3488 if (!bctl) {
3489 ret = -ENOMEM;
3490 goto out_bargs;
3491 }
3492
3493 bctl->fs_info = fs_info;
3494 if (arg) {
3495 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3496 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3497 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3498
3499 bctl->flags = bargs->flags;
f43ffb60
ID		 3500 } else {
3501 /* balance everything - no filters */
3502 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
c9e9f97b
ID		 3503 }
3504
de322263 3505do_balance:
c9e9f97b
ID		 3506 ret = btrfs_balance(bctl, bargs);
3507 /*
837d5b6e
ID		 3508 * bctl is freed in __cancel_balance or in free_fs_info if
3509 * restriper was paused all the way until unmount
c9e9f97b
ID		 3510 */
3511 if (arg) {
3512 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3513 ret = -EFAULT;
3514 }
3515
3516out_bargs:
3517 kfree(bargs);
3518out:
5ac00add
SB		 3519 if (need_to_clear_lock)
3520 atomic_set(&root->fs_info->mutually_exclusive_operation_running,
3521 0);
c9e9f97b
ID		 3522 mutex_unlock(&fs_info->balance_mutex);
3523 mutex_unlock(&fs_info->volume_mutex);
e54bfa31 3524 mnt_drop_write_file(file);
c9e9f97b
ID		 3525 return ret;
3526}
3527
837d5b6e
ID		 3528static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
3529{
3530 if (!capable(CAP_SYS_ADMIN))
3531 return -EPERM;
3532
3533 switch (cmd) {
3534 case BTRFS_BALANCE_CTL_PAUSE:
3535 return btrfs_pause_balance(root->fs_info);
a7e99c69
ID		 3536 case BTRFS_BALANCE_CTL_CANCEL:
3537 return btrfs_cancel_balance(root->fs_info);
837d5b6e
ID		 3538 }
3539
3540 return -EINVAL;
3541}
3542
19a39dce
ID		 3543static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
3544 void __user *arg)
3545{
3546 struct btrfs_fs_info *fs_info = root->fs_info;
3547 struct btrfs_ioctl_balance_args *bargs;
3548 int ret = 0;
3549
3550 if (!capable(CAP_SYS_ADMIN))
3551 return -EPERM;
3552
3553 mutex_lock(&fs_info->balance_mutex);
3554 if (!fs_info->balance_ctl) {
3555 ret = -ENOTCONN;
3556 goto out;
3557 }
3558
3559 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
3560 if (!bargs) {
3561 ret = -ENOMEM;
3562 goto out;
3563 }
3564
3565 update_ioctl_balance_args(fs_info, 1, bargs);
3566
3567 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3568 ret = -EFAULT;
3569
3570 kfree(bargs);
3571out:
3572 mutex_unlock(&fs_info->balance_mutex);
3573 return ret;
3574}
3575
905b0dda 3576static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
5d13a37b 3577{
905b0dda 3578 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
5d13a37b
AJ		 3579 struct btrfs_ioctl_quota_ctl_args *sa;
3580 struct btrfs_trans_handle *trans = NULL;
3581 int ret;
3582 int err;
3583
3584 if (!capable(CAP_SYS_ADMIN))
3585 return -EPERM;
3586
905b0dda
MX		 3587 ret = mnt_want_write_file(file);
3588 if (ret)
3589 return ret;
5d13a37b
AJ		 3590
3591 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 3592 if (IS_ERR(sa)) {
3593 ret = PTR_ERR(sa);
3594 goto drop_write;
3595 }
5d13a37b
AJ		 3596
3597 if (sa->cmd != BTRFS_QUOTA_CTL_RESCAN) {
3598 trans = btrfs_start_transaction(root, 2);
3599 if (IS_ERR(trans)) {
3600 ret = PTR_ERR(trans);
3601 goto out;
3602 }
3603 }
3604
3605 switch (sa->cmd) {
3606 case BTRFS_QUOTA_CTL_ENABLE:
3607 ret = btrfs_quota_enable(trans, root->fs_info);
3608 break;
3609 case BTRFS_QUOTA_CTL_DISABLE:
3610 ret = btrfs_quota_disable(trans, root->fs_info);
3611 break;
3612 case BTRFS_QUOTA_CTL_RESCAN:
3613 ret = btrfs_quota_rescan(root->fs_info);
3614 break;
3615 default:
3616 ret = -EINVAL;
3617 break;
3618 }
3619
3620 if (copy_to_user(arg, sa, sizeof(*sa)))
3621 ret = -EFAULT;
3622
3623 if (trans) {
3624 err = btrfs_commit_transaction(trans, root);
3625 if (err && !ret)
3626 ret = err;
3627 }
5d13a37b
AJ		 3628out:
3629 kfree(sa);
905b0dda
MX		 3630drop_write:
3631 mnt_drop_write_file(file);
5d13a37b
AJ		 3632 return ret;
3633}
3634
905b0dda 3635static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
5d13a37b 3636{
905b0dda 3637 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
5d13a37b
AJ		 3638 struct btrfs_ioctl_qgroup_assign_args *sa;
3639 struct btrfs_trans_handle *trans;
3640 int ret;
3641 int err;
3642
3643 if (!capable(CAP_SYS_ADMIN))
3644 return -EPERM;
3645
905b0dda
MX		 3646 ret = mnt_want_write_file(file);
3647 if (ret)
3648 return ret;
5d13a37b
AJ		 3649
3650 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 3651 if (IS_ERR(sa)) {
3652 ret = PTR_ERR(sa);
3653 goto drop_write;
3654 }
5d13a37b
AJ		 3655
3656 trans = btrfs_join_transaction(root);
3657 if (IS_ERR(trans)) {
3658 ret = PTR_ERR(trans);
3659 goto out;
3660 }
3661
3662 /* FIXME: check if the IDs really exist */
3663 if (sa->assign) {
3664 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
3665 sa->src, sa->dst);
3666 } else {
3667 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
3668 sa->src, sa->dst);
3669 }
3670
3671 err = btrfs_end_transaction(trans, root);
3672 if (err && !ret)
3673 ret = err;
3674
3675out:
3676 kfree(sa);
905b0dda
MX		 3677drop_write:
3678 mnt_drop_write_file(file);
5d13a37b
AJ		 3679 return ret;
3680}
3681
905b0dda 3682static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
5d13a37b 3683{
905b0dda 3684 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
5d13a37b
AJ		 3685 struct btrfs_ioctl_qgroup_create_args *sa;
3686 struct btrfs_trans_handle *trans;
3687 int ret;
3688 int err;
3689
3690 if (!capable(CAP_SYS_ADMIN))
3691 return -EPERM;
3692
905b0dda
MX		 3693 ret = mnt_want_write_file(file);
3694 if (ret)
3695 return ret;
5d13a37b
AJ		 3696
3697 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 3698 if (IS_ERR(sa)) {
3699 ret = PTR_ERR(sa);
3700 goto drop_write;
3701 }
5d13a37b 3702
d86e56cf
MX		 3703 if (!sa->qgroupid) {
3704 ret = -EINVAL;
3705 goto out;
3706 }
3707
5d13a37b
AJ		 3708 trans = btrfs_join_transaction(root);
3709 if (IS_ERR(trans)) {
3710 ret = PTR_ERR(trans);
3711 goto out;
3712 }
3713
3714 /* FIXME: check if the IDs really exist */
3715 if (sa->create) {
3716 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
3717 NULL);
3718 } else {
3719 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
3720 }
3721
3722 err = btrfs_end_transaction(trans, root);
3723 if (err && !ret)
3724 ret = err;
3725
3726out:
3727 kfree(sa);
905b0dda
MX		 3728drop_write:
3729 mnt_drop_write_file(file);
5d13a37b
AJ		 3730 return ret;
3731}
3732
905b0dda 3733static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
5d13a37b 3734{
905b0dda 3735 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
5d13a37b
AJ		 3736 struct btrfs_ioctl_qgroup_limit_args *sa;
3737 struct btrfs_trans_handle *trans;
3738 int ret;
3739 int err;
3740 u64 qgroupid;
3741
3742 if (!capable(CAP_SYS_ADMIN))
3743 return -EPERM;
3744
905b0dda
MX		 3745 ret = mnt_want_write_file(file);
3746 if (ret)
3747 return ret;
5d13a37b
AJ		 3748
3749 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 3750 if (IS_ERR(sa)) {
3751 ret = PTR_ERR(sa);
3752 goto drop_write;
3753 }
5d13a37b
AJ		 3754
3755 trans = btrfs_join_transaction(root);
3756 if (IS_ERR(trans)) {
3757 ret = PTR_ERR(trans);
3758 goto out;
3759 }
3760
3761 qgroupid = sa->qgroupid;
3762 if (!qgroupid) {
3763 /* take the current subvol as qgroup */
3764 qgroupid = root->root_key.objectid;
3765 }
3766
3767 /* FIXME: check if the IDs really exist */
3768 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
3769
3770 err = btrfs_end_transaction(trans, root);
3771 if (err && !ret)
3772 ret = err;
3773
3774out:
3775 kfree(sa);
905b0dda
MX		 3776drop_write:
3777 mnt_drop_write_file(file);
5d13a37b
AJ		 3778 return ret;
3779}
3780
8ea05e3a
AB		 3781static long btrfs_ioctl_set_received_subvol(struct file *file,
3782 void __user *arg)
3783{
3784 struct btrfs_ioctl_received_subvol_args *sa = NULL;
3785 struct inode *inode = fdentry(file)->d_inode;
3786 struct btrfs_root *root = BTRFS_I(inode)->root;
3787 struct btrfs_root_item *root_item = &root->root_item;
3788 struct btrfs_trans_handle *trans;
3789 struct timespec ct = CURRENT_TIME;
3790 int ret = 0;
3791
3792 ret = mnt_want_write_file(file);
3793 if (ret < 0)
3794 return ret;
3795
3796 down_write(&root->fs_info->subvol_sem);
3797
3798 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
3799 ret = -EINVAL;
3800 goto out;
3801 }
3802
3803 if (btrfs_root_readonly(root)) {
3804 ret = -EROFS;
3805 goto out;
3806 }
3807
3808 if (!inode_owner_or_capable(inode)) {
3809 ret = -EACCES;
3810 goto out;
3811 }
3812
3813 sa = memdup_user(arg, sizeof(*sa));
3814 if (IS_ERR(sa)) {
3815 ret = PTR_ERR(sa);
3816 sa = NULL;
3817 goto out;
3818 }
3819
3820 trans = btrfs_start_transaction(root, 1);
3821 if (IS_ERR(trans)) {
3822 ret = PTR_ERR(trans);
3823 trans = NULL;
3824 goto out;
3825 }
3826
3827 sa->rtransid = trans->transid;
3828 sa->rtime.sec = ct.tv_sec;
3829 sa->rtime.nsec = ct.tv_nsec;
3830
3831 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
3832 btrfs_set_root_stransid(root_item, sa->stransid);
3833 btrfs_set_root_rtransid(root_item, sa->rtransid);
3834 root_item->stime.sec = cpu_to_le64(sa->stime.sec);
3835 root_item->stime.nsec = cpu_to_le32(sa->stime.nsec);
3836 root_item->rtime.sec = cpu_to_le64(sa->rtime.sec);
3837 root_item->rtime.nsec = cpu_to_le32(sa->rtime.nsec);
3838
3839 ret = btrfs_update_root(trans, root->fs_info->tree_root,
3840 &root->root_key, &root->root_item);
3841 if (ret < 0) {
3842 btrfs_end_transaction(trans, root);
3843 trans = NULL;
3844 goto out;
3845 } else {
3846 ret = btrfs_commit_transaction(trans, root);
3847 if (ret < 0)
3848 goto out;
3849 }
3850
3851 ret = copy_to_user(arg, sa, sizeof(*sa));
3852 if (ret)
3853 ret = -EFAULT;
3854
3855out:
3856 kfree(sa);
3857 up_write(&root->fs_info->subvol_sem);
3858 mnt_drop_write_file(file);
3859 return ret;
3860}
3861
f46b5a66
CH		 3862long btrfs_ioctl(struct file *file, unsigned int
3863 cmd, unsigned long arg)
3864{
3865 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
4bcabaa3 3866 void __user *argp = (void __user *)arg;
f46b5a66
CH		 3867
3868 switch (cmd) {
6cbff00f
CH		 3869 case FS_IOC_GETFLAGS:
3870 return btrfs_ioctl_getflags(file, argp);
3871 case FS_IOC_SETFLAGS:
3872 return btrfs_ioctl_setflags(file, argp);
3873 case FS_IOC_GETVERSION:
3874 return btrfs_ioctl_getversion(file, argp);
f7039b1d
LD		 3875 case FITRIM:
3876 return btrfs_ioctl_fitrim(file, argp);
f46b5a66 3877 case BTRFS_IOC_SNAP_CREATE:
fa0d2b9b 3878 return btrfs_ioctl_snap_create(file, argp, 0);
fdfb1e4f 3879 case BTRFS_IOC_SNAP_CREATE_V2:
fa0d2b9b 3880 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3de4586c 3881 case BTRFS_IOC_SUBVOL_CREATE:
fa0d2b9b 3882 return btrfs_ioctl_snap_create(file, argp, 1);
6f72c7e2
AJ		 3883 case BTRFS_IOC_SUBVOL_CREATE_V2:
3884 return btrfs_ioctl_snap_create_v2(file, argp, 1);
76dda93c
YZ		 3885 case BTRFS_IOC_SNAP_DESTROY:
3886 return btrfs_ioctl_snap_destroy(file, argp);
0caa102d
LZ		 3887 case BTRFS_IOC_SUBVOL_GETFLAGS:
3888 return btrfs_ioctl_subvol_getflags(file, argp);
3889 case BTRFS_IOC_SUBVOL_SETFLAGS:
3890 return btrfs_ioctl_subvol_setflags(file, argp);
6ef5ed0d
JB		 3891 case BTRFS_IOC_DEFAULT_SUBVOL:
3892 return btrfs_ioctl_default_subvol(file, argp);
f46b5a66 3893 case BTRFS_IOC_DEFRAG:
1e701a32
CM		 3894 return btrfs_ioctl_defrag(file, NULL);
3895 case BTRFS_IOC_DEFRAG_RANGE:
3896 return btrfs_ioctl_defrag(file, argp);
f46b5a66 3897 case BTRFS_IOC_RESIZE:
198605a8 3898 return btrfs_ioctl_resize(file, argp);
f46b5a66 3899 case BTRFS_IOC_ADD_DEV:
4bcabaa3 3900 return btrfs_ioctl_add_dev(root, argp);
f46b5a66 3901 case BTRFS_IOC_RM_DEV:
da24927b 3902 return btrfs_ioctl_rm_dev(file, argp);
475f6387
JS		 3903 case BTRFS_IOC_FS_INFO:
3904 return btrfs_ioctl_fs_info(root, argp);
3905 case BTRFS_IOC_DEV_INFO:
3906 return btrfs_ioctl_dev_info(root, argp);
f46b5a66 3907 case BTRFS_IOC_BALANCE:
9ba1f6e4 3908 return btrfs_ioctl_balance(file, NULL);
f46b5a66 3909 case BTRFS_IOC_CLONE:
c5c9cd4d
SW		 3910 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
3911 case BTRFS_IOC_CLONE_RANGE:
7a865e8a 3912 return btrfs_ioctl_clone_range(file, argp);
f46b5a66
CH		 3913 case BTRFS_IOC_TRANS_START:
3914 return btrfs_ioctl_trans_start(file);
3915 case BTRFS_IOC_TRANS_END:
3916 return btrfs_ioctl_trans_end(file);
ac8e9819
CM		 3917 case BTRFS_IOC_TREE_SEARCH:
3918 return btrfs_ioctl_tree_search(file, argp);
3919 case BTRFS_IOC_INO_LOOKUP:
3920 return btrfs_ioctl_ino_lookup(file, argp);
d7728c96
JS		 3921 case BTRFS_IOC_INO_PATHS:
3922 return btrfs_ioctl_ino_to_path(root, argp);
3923 case BTRFS_IOC_LOGICAL_INO:
3924 return btrfs_ioctl_logical_to_ino(root, argp);
1406e432
JB		 3925 case BTRFS_IOC_SPACE_INFO:
3926 return btrfs_ioctl_space_info(root, argp);
f46b5a66
CH		 3927 case BTRFS_IOC_SYNC:
3928 btrfs_sync_fs(file->f_dentry->d_sb, 1);
3929 return 0;
46204592 3930 case BTRFS_IOC_START_SYNC:
9a8c28be 3931 return btrfs_ioctl_start_sync(root, argp);
46204592 3932 case BTRFS_IOC_WAIT_SYNC:
9a8c28be 3933 return btrfs_ioctl_wait_sync(root, argp);
475f6387 3934 case BTRFS_IOC_SCRUB:
b8e95489 3935 return btrfs_ioctl_scrub(file, argp);
475f6387
JS		 3936 case BTRFS_IOC_SCRUB_CANCEL:
3937 return btrfs_ioctl_scrub_cancel(root, argp);
3938 case BTRFS_IOC_SCRUB_PROGRESS:
3939 return btrfs_ioctl_scrub_progress(root, argp);
c9e9f97b 3940 case BTRFS_IOC_BALANCE_V2:
9ba1f6e4 3941 return btrfs_ioctl_balance(file, argp);
837d5b6e
ID		 3942 case BTRFS_IOC_BALANCE_CTL:
3943 return btrfs_ioctl_balance_ctl(root, arg);
19a39dce
ID		 3944 case BTRFS_IOC_BALANCE_PROGRESS:
3945 return btrfs_ioctl_balance_progress(root, argp);
8ea05e3a
AB		 3946 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
3947 return btrfs_ioctl_set_received_subvol(file, argp);
31db9f7c
AB		 3948 case BTRFS_IOC_SEND:
3949 return btrfs_ioctl_send(file, argp);
c11d2c23 3950 case BTRFS_IOC_GET_DEV_STATS:
b27f7c0c 3951 return btrfs_ioctl_get_dev_stats(root, argp);
5d13a37b 3952 case BTRFS_IOC_QUOTA_CTL:
905b0dda 3953 return btrfs_ioctl_quota_ctl(file, argp);
5d13a37b 3954 case BTRFS_IOC_QGROUP_ASSIGN:
905b0dda 3955 return btrfs_ioctl_qgroup_assign(file, argp);
5d13a37b 3956 case BTRFS_IOC_QGROUP_CREATE:
905b0dda 3957 return btrfs_ioctl_qgroup_create(file, argp);
5d13a37b 3958 case BTRFS_IOC_QGROUP_LIMIT:
905b0dda 3959 return btrfs_ioctl_qgroup_limit(file, argp);
3f6bcfbd
SB		 3960 case BTRFS_IOC_DEV_REPLACE:
3961 return btrfs_ioctl_dev_replace(root, argp);
f46b5a66
CH		 3962 }
3963
3964 return -ENOTTY;
3965}
Linux 6.x block layer and io_uring tree(s)