]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/blame - fs/btrfs/ioctl.c
projects / mirror_ubuntu-hirsute-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Btrfs: fix race between removing a dev and writing sbs
[mirror_ubuntu-hirsute-kernel.git] / fs / btrfs / ioctl.c
CommitLineData
f46b5a66
CH		 1/*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
19#include <linux/kernel.h>
20#include <linux/bio.h>
21#include <linux/buffer_head.h>
22#include <linux/file.h>
23#include <linux/fs.h>
cb8e7090 24#include <linux/fsnotify.h>
f46b5a66
CH		 25#include <linux/pagemap.h>
26#include <linux/highmem.h>
27#include <linux/time.h>
28#include <linux/init.h>
29#include <linux/string.h>
f46b5a66 30#include <linux/backing-dev.h>
cb8e7090 31#include <linux/mount.h>
f46b5a66 32#include <linux/mpage.h>
cb8e7090 33#include <linux/namei.h>
f46b5a66
CH		 34#include <linux/swap.h>
35#include <linux/writeback.h>
36#include <linux/statfs.h>
37#include <linux/compat.h>
38#include <linux/bit_spinlock.h>
cb8e7090 39#include <linux/security.h>
f46b5a66 40#include <linux/xattr.h>
7ea394f1 41#include <linux/vmalloc.h>
5a0e3ad6 42#include <linux/slab.h>
f7039b1d 43#include <linux/blkdev.h>
8ea05e3a 44#include <linux/uuid.h>
55e301fd 45#include <linux/btrfs.h>
416161db 46#include <linux/uaccess.h>
4b4e25f2 47#include "compat.h"
f46b5a66
CH		 48#include "ctree.h"
49#include "disk-io.h"
50#include "transaction.h"
51#include "btrfs_inode.h"
f46b5a66
CH		 52#include "print-tree.h"
53#include "volumes.h"
925baedd 54#include "locking.h"
581bb050 55#include "inode-map.h"
d7728c96 56#include "backref.h"
606686ee 57#include "rcu-string.h"
31db9f7c 58#include "send.h"
3f6bcfbd 59#include "dev-replace.h"
f46b5a66 60
416161db
MF		 61static int btrfs_clone(struct inode *src, struct inode *inode,
62 u64 off, u64 olen, u64 olen_aligned, u64 destoff);
63
6cbff00f
CH		 64/* Mask out flags that are inappropriate for the given type of inode. */
65static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
66{
67 if (S_ISDIR(mode))
68 return flags;
69 else if (S_ISREG(mode))
70 return flags & ~FS_DIRSYNC_FL;
71 else
72 return flags & (FS_NODUMP_FL | FS_NOATIME_FL);
73}
74
75/*
76 * Export inode flags to the format expected by the FS_IOC_GETFLAGS ioctl.
77 */
78static unsigned int btrfs_flags_to_ioctl(unsigned int flags)
79{
80 unsigned int iflags = 0;
81
82 if (flags & BTRFS_INODE_SYNC)
83 iflags |= FS_SYNC_FL;
84 if (flags & BTRFS_INODE_IMMUTABLE)
85 iflags |= FS_IMMUTABLE_FL;
86 if (flags & BTRFS_INODE_APPEND)
87 iflags |= FS_APPEND_FL;
88 if (flags & BTRFS_INODE_NODUMP)
89 iflags |= FS_NODUMP_FL;
90 if (flags & BTRFS_INODE_NOATIME)
91 iflags |= FS_NOATIME_FL;
92 if (flags & BTRFS_INODE_DIRSYNC)
93 iflags |= FS_DIRSYNC_FL;
d0092bdd
LZ		 94 if (flags & BTRFS_INODE_NODATACOW)
95 iflags |= FS_NOCOW_FL;
96
97 if ((flags & BTRFS_INODE_COMPRESS) && !(flags & BTRFS_INODE_NOCOMPRESS))
98 iflags |= FS_COMPR_FL;
99 else if (flags & BTRFS_INODE_NOCOMPRESS)
100 iflags |= FS_NOCOMP_FL;
6cbff00f
CH		 101
102 return iflags;
103}
104
105/*
106 * Update inode->i_flags based on the btrfs internal flags.
107 */
108void btrfs_update_iflags(struct inode *inode)
109{
110 struct btrfs_inode *ip = BTRFS_I(inode);
111
112 inode->i_flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
113
114 if (ip->flags & BTRFS_INODE_SYNC)
115 inode->i_flags |= S_SYNC;
116 if (ip->flags & BTRFS_INODE_IMMUTABLE)
117 inode->i_flags |= S_IMMUTABLE;
118 if (ip->flags & BTRFS_INODE_APPEND)
119 inode->i_flags |= S_APPEND;
120 if (ip->flags & BTRFS_INODE_NOATIME)
121 inode->i_flags |= S_NOATIME;
122 if (ip->flags & BTRFS_INODE_DIRSYNC)
123 inode->i_flags |= S_DIRSYNC;
124}
125
126/*
127 * Inherit flags from the parent inode.
128 *
e27425d6 129 * Currently only the compression flags and the cow flags are inherited.
6cbff00f
CH		 130 */
131void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
132{
0b4dcea5
CM		 133 unsigned int flags;
134
135 if (!dir)
136 return;
137
138 flags = BTRFS_I(dir)->flags;
6cbff00f 139
e27425d6
JB		 140 if (flags & BTRFS_INODE_NOCOMPRESS) {
141 BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
142 BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
143 } else if (flags & BTRFS_INODE_COMPRESS) {
144 BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
145 BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
146 }
147
213490b3 148 if (flags & BTRFS_INODE_NODATACOW) {
e27425d6 149 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
213490b3
LB		 150 if (S_ISREG(inode->i_mode))
151 BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
152 }
6cbff00f 153
6cbff00f
CH		 154 btrfs_update_iflags(inode);
155}
156
157static int btrfs_ioctl_getflags(struct file *file, void __user *arg)
158{
496ad9aa 159 struct btrfs_inode *ip = BTRFS_I(file_inode(file));
6cbff00f
CH		 160 unsigned int flags = btrfs_flags_to_ioctl(ip->flags);
161
162 if (copy_to_user(arg, &flags, sizeof(flags)))
163 return -EFAULT;
164 return 0;
165}
166
75e7cb7f
LB		 167static int check_flags(unsigned int flags)
168{
169 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
170 FS_NOATIME_FL | FS_NODUMP_FL | \
171 FS_SYNC_FL | FS_DIRSYNC_FL | \
e1e8fb6a
LZ		 172 FS_NOCOMP_FL | FS_COMPR_FL |
173 FS_NOCOW_FL))
75e7cb7f
LB		 174 return -EOPNOTSUPP;
175
176 if ((flags & FS_NOCOMP_FL) && (flags & FS_COMPR_FL))
177 return -EINVAL;
178
75e7cb7f
LB		 179 return 0;
180}
181
6cbff00f
CH		 182static int btrfs_ioctl_setflags(struct file *file, void __user *arg)
183{
496ad9aa 184 struct inode *inode = file_inode(file);
6cbff00f
CH		 185 struct btrfs_inode *ip = BTRFS_I(inode);
186 struct btrfs_root *root = ip->root;
187 struct btrfs_trans_handle *trans;
188 unsigned int flags, oldflags;
189 int ret;
f062abf0
LZ		 190 u64 ip_oldflags;
191 unsigned int i_oldflags;
7e97b8da 192 umode_t mode;
6cbff00f 193
b83cc969
LZ		 194 if (btrfs_root_readonly(root))
195 return -EROFS;
196
6cbff00f
CH		 197 if (copy_from_user(&flags, arg, sizeof(flags)))
198 return -EFAULT;
199
75e7cb7f
LB		 200 ret = check_flags(flags);
201 if (ret)
202 return ret;
f46b5a66 203
2e149670 204 if (!inode_owner_or_capable(inode))
6cbff00f
CH		 205 return -EACCES;
206
e7848683
JK		 207 ret = mnt_want_write_file(file);
208 if (ret)
209 return ret;
210
6cbff00f
CH		 211 mutex_lock(&inode->i_mutex);
212
f062abf0
LZ		 213 ip_oldflags = ip->flags;
214 i_oldflags = inode->i_flags;
7e97b8da 215 mode = inode->i_mode;
f062abf0 216
6cbff00f
CH		 217 flags = btrfs_mask_flags(inode->i_mode, flags);
218 oldflags = btrfs_flags_to_ioctl(ip->flags);
219 if ((flags ^ oldflags) & (FS_APPEND_FL | FS_IMMUTABLE_FL)) {
220 if (!capable(CAP_LINUX_IMMUTABLE)) {
221 ret = -EPERM;
222 goto out_unlock;
223 }
224 }
225
6cbff00f
CH		 226 if (flags & FS_SYNC_FL)
227 ip->flags |= BTRFS_INODE_SYNC;
228 else
229 ip->flags &= ~BTRFS_INODE_SYNC;
230 if (flags & FS_IMMUTABLE_FL)
231 ip->flags |= BTRFS_INODE_IMMUTABLE;
232 else
233 ip->flags &= ~BTRFS_INODE_IMMUTABLE;
234 if (flags & FS_APPEND_FL)
235 ip->flags |= BTRFS_INODE_APPEND;
236 else
237 ip->flags &= ~BTRFS_INODE_APPEND;
238 if (flags & FS_NODUMP_FL)
239 ip->flags |= BTRFS_INODE_NODUMP;
240 else
241 ip->flags &= ~BTRFS_INODE_NODUMP;
242 if (flags & FS_NOATIME_FL)
243 ip->flags |= BTRFS_INODE_NOATIME;
244 else
245 ip->flags &= ~BTRFS_INODE_NOATIME;
246 if (flags & FS_DIRSYNC_FL)
247 ip->flags |= BTRFS_INODE_DIRSYNC;
248 else
249 ip->flags &= ~BTRFS_INODE_DIRSYNC;
7e97b8da
DS		 250 if (flags & FS_NOCOW_FL) {
251 if (S_ISREG(mode)) {
252 /*
253 * It's safe to turn csums off here, no extents exist.
254 * Otherwise we want the flag to reflect the real COW
255 * status of the file and will not set it.
256 */
257 if (inode->i_size == 0)
258 ip->flags |= BTRFS_INODE_NODATACOW
259 | BTRFS_INODE_NODATASUM;
260 } else {
261 ip->flags |= BTRFS_INODE_NODATACOW;
262 }
263 } else {
264 /*
265 * Revert back under same assuptions as above
266 */
267 if (S_ISREG(mode)) {
268 if (inode->i_size == 0)
269 ip->flags &= ~(BTRFS_INODE_NODATACOW
270 | BTRFS_INODE_NODATASUM);
271 } else {
272 ip->flags &= ~BTRFS_INODE_NODATACOW;
273 }
274 }
6cbff00f 275
75e7cb7f
LB		 276 /*
277 * The COMPRESS flag can only be changed by users, while the NOCOMPRESS
278 * flag may be changed automatically if compression code won't make
279 * things smaller.
280 */
281 if (flags & FS_NOCOMP_FL) {
282 ip->flags &= ~BTRFS_INODE_COMPRESS;
283 ip->flags |= BTRFS_INODE_NOCOMPRESS;
284 } else if (flags & FS_COMPR_FL) {
285 ip->flags |= BTRFS_INODE_COMPRESS;
286 ip->flags &= ~BTRFS_INODE_NOCOMPRESS;
ebcb904d
LZ		 287 } else {
288 ip->flags &= ~(BTRFS_INODE_COMPRESS | BTRFS_INODE_NOCOMPRESS);
75e7cb7f 289 }
6cbff00f 290
4da6f1a3 291 trans = btrfs_start_transaction(root, 1);
f062abf0
LZ		 292 if (IS_ERR(trans)) {
293 ret = PTR_ERR(trans);
294 goto out_drop;
295 }
6cbff00f 296
306424cc 297 btrfs_update_iflags(inode);
0c4d2d95 298 inode_inc_iversion(inode);
306424cc 299 inode->i_ctime = CURRENT_TIME;
6cbff00f 300 ret = btrfs_update_inode(trans, root, inode);
6cbff00f 301
6cbff00f 302 btrfs_end_transaction(trans, root);
f062abf0
LZ		 303 out_drop:
304 if (ret) {
305 ip->flags = ip_oldflags;
306 inode->i_flags = i_oldflags;
307 }
6cbff00f 308
6cbff00f
CH		 309 out_unlock:
310 mutex_unlock(&inode->i_mutex);
e7848683 311 mnt_drop_write_file(file);
2d4e6f6a 312 return ret;
6cbff00f
CH		 313}
314
315static int btrfs_ioctl_getversion(struct file *file, int __user *arg)
316{
496ad9aa 317 struct inode *inode = file_inode(file);
6cbff00f
CH		 318
319 return put_user(inode->i_generation, arg);
320}
f46b5a66 321
f7039b1d
LD		 322static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
323{
815745cf 324 struct btrfs_fs_info *fs_info = btrfs_sb(fdentry(file)->d_sb);
f7039b1d
LD		 325 struct btrfs_device *device;
326 struct request_queue *q;
327 struct fstrim_range range;
328 u64 minlen = ULLONG_MAX;
329 u64 num_devices = 0;
815745cf 330 u64 total_bytes = btrfs_super_total_bytes(fs_info->super_copy);
f7039b1d
LD		 331 int ret;
332
333 if (!capable(CAP_SYS_ADMIN))
334 return -EPERM;
335
1f78160c
XG		 336 rcu_read_lock();
337 list_for_each_entry_rcu(device, &fs_info->fs_devices->devices,
338 dev_list) {
f7039b1d
LD		 339 if (!device->bdev)
340 continue;
341 q = bdev_get_queue(device->bdev);
342 if (blk_queue_discard(q)) {
343 num_devices++;
344 minlen = min((u64)q->limits.discard_granularity,
345 minlen);
346 }
347 }
1f78160c 348 rcu_read_unlock();
f4c697e6 349
f7039b1d
LD		 350 if (!num_devices)
351 return -EOPNOTSUPP;
f7039b1d
LD		 352 if (copy_from_user(&range, arg, sizeof(range)))
353 return -EFAULT;
e515c18b
LC		 354 if (range.start > total_bytes ||
355 range.len < fs_info->sb->s_blocksize)
f4c697e6 356 return -EINVAL;
f7039b1d 357
f4c697e6 358 range.len = min(range.len, total_bytes - range.start);
f7039b1d 359 range.minlen = max(range.minlen, minlen);
815745cf 360 ret = btrfs_trim_fs(fs_info->tree_root, &range);
f7039b1d
LD		 361 if (ret < 0)
362 return ret;
363
364 if (copy_to_user(arg, &range, sizeof(range)))
365 return -EFAULT;
366
367 return 0;
368}
369
dd5f9615
SB		 370int btrfs_is_empty_uuid(u8 *uuid)
371{
372 static char empty_uuid[BTRFS_UUID_SIZE] = {0};
373
374 return !memcmp(uuid, empty_uuid, BTRFS_UUID_SIZE);
375}
376
d5c12070 377static noinline int create_subvol(struct inode *dir,
cb8e7090 378 struct dentry *dentry,
72fd032e 379 char *name, int namelen,
6f72c7e2 380 u64 *async_transid,
8696c533 381 struct btrfs_qgroup_inherit *inherit)
f46b5a66
CH		 382{
383 struct btrfs_trans_handle *trans;
384 struct btrfs_key key;
385 struct btrfs_root_item root_item;
386 struct btrfs_inode_item *inode_item;
387 struct extent_buffer *leaf;
d5c12070 388 struct btrfs_root *root = BTRFS_I(dir)->root;
76dda93c 389 struct btrfs_root *new_root;
d5c12070 390 struct btrfs_block_rsv block_rsv;
8ea05e3a 391 struct timespec cur_time = CURRENT_TIME;
f46b5a66
CH		 392 int ret;
393 int err;
394 u64 objectid;
395 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
3de4586c 396 u64 index = 0;
d5c12070 397 u64 qgroup_reserved;
8ea05e3a 398 uuid_le new_uuid;
f46b5a66 399
581bb050 400 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
2fbe8c8a 401 if (ret)
a22285a6 402 return ret;
6a912213 403
d5c12070 404 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
9ed74f2d 405 /*
d5c12070
MX		 406 * The same as the snapshot creation, please see the comment
407 * of create_snapshot().
9ed74f2d 408 */
d5c12070 409 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
dd5f9615 410 8, &qgroup_reserved, false);
d5c12070
MX		 411 if (ret)
412 return ret;
413
414 trans = btrfs_start_transaction(root, 0);
415 if (IS_ERR(trans)) {
416 ret = PTR_ERR(trans);
417 goto out;
418 }
419 trans->block_rsv = &block_rsv;
420 trans->bytes_reserved = block_rsv.size;
f46b5a66 421
8696c533 422 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
6f72c7e2
AJ		 423 if (ret)
424 goto fail;
425
5d4f98a2 426 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
5581a51a 427 0, objectid, NULL, 0, 0, 0);
8e8a1e31
JB		 428 if (IS_ERR(leaf)) {
429 ret = PTR_ERR(leaf);
430 goto fail;
431 }
f46b5a66 432
5d4f98a2 433 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
f46b5a66
CH		 434 btrfs_set_header_bytenr(leaf, leaf->start);
435 btrfs_set_header_generation(leaf, trans->transid);
5d4f98a2 436 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
f46b5a66
CH		 437 btrfs_set_header_owner(leaf, objectid);
438
fba6aa75 439 write_extent_buffer(leaf, root->fs_info->fsid, btrfs_header_fsid(leaf),
f46b5a66 440 BTRFS_FSID_SIZE);
5d4f98a2 441 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
b308bc2f 442 btrfs_header_chunk_tree_uuid(leaf),
5d4f98a2 443 BTRFS_UUID_SIZE);
f46b5a66
CH		 444 btrfs_mark_buffer_dirty(leaf);
445
8ea05e3a
AB		 446 memset(&root_item, 0, sizeof(root_item));
447
f46b5a66 448 inode_item = &root_item.inode;
3cae210f
QW		 449 btrfs_set_stack_inode_generation(inode_item, 1);
450 btrfs_set_stack_inode_size(inode_item, 3);
451 btrfs_set_stack_inode_nlink(inode_item, 1);
452 btrfs_set_stack_inode_nbytes(inode_item, root->leafsize);
453 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
f46b5a66 454
3cae210f
QW		 455 btrfs_set_root_flags(&root_item, 0);
456 btrfs_set_root_limit(&root_item, 0);
457 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
08fe4db1 458
f46b5a66 459 btrfs_set_root_bytenr(&root_item, leaf->start);
84234f3a 460 btrfs_set_root_generation(&root_item, trans->transid);
f46b5a66
CH		 461 btrfs_set_root_level(&root_item, 0);
462 btrfs_set_root_refs(&root_item, 1);
86b9f2ec 463 btrfs_set_root_used(&root_item, leaf->len);
80ff3856 464 btrfs_set_root_last_snapshot(&root_item, 0);
f46b5a66 465
8ea05e3a
AB		 466 btrfs_set_root_generation_v2(&root_item,
467 btrfs_root_generation(&root_item));
468 uuid_le_gen(&new_uuid);
469 memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
3cae210f
QW		 470 btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec);
471 btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec);
8ea05e3a
AB		 472 root_item.ctime = root_item.otime;
473 btrfs_set_root_ctransid(&root_item, trans->transid);
474 btrfs_set_root_otransid(&root_item, trans->transid);
f46b5a66 475
925baedd 476 btrfs_tree_unlock(leaf);
f46b5a66
CH		 477 free_extent_buffer(leaf);
478 leaf = NULL;
479
480 btrfs_set_root_dirid(&root_item, new_dirid);
481
482 key.objectid = objectid;
5d4f98a2 483 key.offset = 0;
f46b5a66
CH		 484 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
485 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
486 &root_item);
487 if (ret)
488 goto fail;
489
76dda93c
YZ		 490 key.offset = (u64)-1;
491 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
79787eaa
JM		 492 if (IS_ERR(new_root)) {
493 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
494 ret = PTR_ERR(new_root);
495 goto fail;
496 }
76dda93c
YZ		 497
498 btrfs_record_root_in_trans(trans, new_root);
499
d82a6f1d 500 ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
ce598979
MF		 501 if (ret) {
502 /* We potentially lose an unused inode item here */
79787eaa 503 btrfs_abort_transaction(trans, root, ret);
ce598979
MF		 504 goto fail;
505 }
506
f46b5a66
CH		 507 /*
508 * insert the directory item
509 */
3de4586c 510 ret = btrfs_set_inode_index(dir, &index);
79787eaa
JM		 511 if (ret) {
512 btrfs_abort_transaction(trans, root, ret);
513 goto fail;
514 }
3de4586c
CM		 515
516 ret = btrfs_insert_dir_item(trans, root,
16cdcec7 517 name, namelen, dir, &key,
3de4586c 518 BTRFS_FT_DIR, index);
79787eaa
JM		 519 if (ret) {
520 btrfs_abort_transaction(trans, root, ret);
f46b5a66 521 goto fail;
79787eaa 522 }
0660b5af 523
52c26179
YZ		 524 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
525 ret = btrfs_update_inode(trans, root, dir);
526 BUG_ON(ret);
527
0660b5af 528 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
4df27c4d 529 objectid, root->root_key.objectid,
33345d01 530 btrfs_ino(dir), index, name, namelen);
76dda93c 531 BUG_ON(ret);
f46b5a66 532
dd5f9615
SB		 533 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
534 root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
535 objectid);
536 if (ret)
537 btrfs_abort_transaction(trans, root, ret);
538
f46b5a66 539fail:
d5c12070
MX		 540 trans->block_rsv = NULL;
541 trans->bytes_reserved = 0;
72fd032e
SW		 542 if (async_transid) {
543 *async_transid = trans->transid;
544 err = btrfs_commit_transaction_async(trans, root, 1);
00d71c9c
MX		 545 if (err)
546 err = btrfs_commit_transaction(trans, root);
72fd032e
SW		 547 } else {
548 err = btrfs_commit_transaction(trans, root);
549 }
f46b5a66
CH		 550 if (err && !ret)
551 ret = err;
1a65e24b
CM		 552
553 if (!ret)
554 d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
d5c12070
MX		 555out:
556 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
f46b5a66
CH		 557 return ret;
558}
559
e9662f70
MX		 560static int create_snapshot(struct btrfs_root *root, struct inode *dir,
561 struct dentry *dentry, char *name, int namelen,
562 u64 *async_transid, bool readonly,
563 struct btrfs_qgroup_inherit *inherit)
f46b5a66 564{
2e4bfab9 565 struct inode *inode;
f46b5a66
CH		 566 struct btrfs_pending_snapshot *pending_snapshot;
567 struct btrfs_trans_handle *trans;
2e4bfab9 568 int ret;
f46b5a66
CH		 569
570 if (!root->ref_cows)
571 return -EINVAL;
572
6a03843d
MX		 573 ret = btrfs_start_delalloc_inodes(root, 0);
574 if (ret)
575 return ret;
576
577 btrfs_wait_ordered_extents(root, 0);
578
3de4586c 579 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
a22285a6
YZ		 580 if (!pending_snapshot)
581 return -ENOMEM;
582
66d8f3dd
MX		 583 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
584 BTRFS_BLOCK_RSV_TEMP);
d5c12070
MX		 585 /*
586 * 1 - parent dir inode
587 * 2 - dir entries
588 * 1 - root item
589 * 2 - root ref/backref
590 * 1 - root of snapshot
dd5f9615 591 * 1 - UUID item
d5c12070
MX		 592 */
593 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
dd5f9615 594 &pending_snapshot->block_rsv, 8,
ee3441b4
JM		 595 &pending_snapshot->qgroup_reserved,
596 false);
d5c12070
MX		 597 if (ret)
598 goto out;
599
3de4586c 600 pending_snapshot->dentry = dentry;
f46b5a66 601 pending_snapshot->root = root;
b83cc969 602 pending_snapshot->readonly = readonly;
e9662f70 603 pending_snapshot->dir = dir;
8696c533 604 pending_snapshot->inherit = inherit;
a22285a6 605
d5c12070 606 trans = btrfs_start_transaction(root, 0);
a22285a6
YZ		 607 if (IS_ERR(trans)) {
608 ret = PTR_ERR(trans);
609 goto fail;
610 }
611
8351583e 612 spin_lock(&root->fs_info->trans_lock);
f46b5a66
CH		 613 list_add(&pending_snapshot->list,
614 &trans->transaction->pending_snapshots);
8351583e 615 spin_unlock(&root->fs_info->trans_lock);
72fd032e
SW		 616 if (async_transid) {
617 *async_transid = trans->transid;
618 ret = btrfs_commit_transaction_async(trans,
619 root->fs_info->extent_root, 1);
00d71c9c
MX		 620 if (ret)
621 ret = btrfs_commit_transaction(trans, root);
72fd032e
SW		 622 } else {
623 ret = btrfs_commit_transaction(trans,
624 root->fs_info->extent_root);
625 }
aec8030a 626 if (ret)
c37b2b62 627 goto fail;
a22285a6
YZ		 628
629 ret = pending_snapshot->error;
630 if (ret)
631 goto fail;
632
66b4ffd1
JB		 633 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
634 if (ret)
635 goto fail;
f46b5a66 636
2fbe8c8a 637 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
2e4bfab9
YZ		 638 if (IS_ERR(inode)) {
639 ret = PTR_ERR(inode);
640 goto fail;
641 }
642 BUG_ON(!inode);
643 d_instantiate(dentry, inode);
644 ret = 0;
645fail:
d5c12070
MX		 646 btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
647 &pending_snapshot->block_rsv,
648 pending_snapshot->qgroup_reserved);
649out:
a22285a6 650 kfree(pending_snapshot);
f46b5a66
CH		 651 return ret;
652}
653
4260f7c7
SW		 654/* copy of check_sticky in fs/namei.c()
655* It's inline, so penalty for filesystems that don't use sticky bit is
656* minimal.
657*/
658static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
659{
2f2f43d3 660 kuid_t fsuid = current_fsuid();
4260f7c7
SW		 661
662 if (!(dir->i_mode & S_ISVTX))
663 return 0;
2f2f43d3 664 if (uid_eq(inode->i_uid, fsuid))
4260f7c7 665 return 0;
2f2f43d3 666 if (uid_eq(dir->i_uid, fsuid))
4260f7c7
SW		 667 return 0;
668 return !capable(CAP_FOWNER);
669}
670
671/* copy of may_delete in fs/namei.c()
672 * Check whether we can remove a link victim from directory dir, check
673 * whether the type of victim is right.
674 * 1. We can't do it if dir is read-only (done in permission())
675 * 2. We should have write and exec permissions on dir
676 * 3. We can't remove anything from append-only dir
677 * 4. We can't do anything with immutable dir (done in permission())
678 * 5. If the sticky bit on dir is set we should either
679 * a. be owner of dir, or
680 * b. be owner of victim, or
681 * c. have CAP_FOWNER capability
682 * 6. If the victim is append-only or immutable we can't do antyhing with
683 * links pointing to it.
684 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
685 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
686 * 9. We can't remove a root or mountpoint.
687 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
688 * nfs_async_unlink().
689 */
690
691static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
692{
693 int error;
694
695 if (!victim->d_inode)
696 return -ENOENT;
697
698 BUG_ON(victim->d_parent->d_inode != dir);
4fa6b5ec 699 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
4260f7c7
SW		 700
701 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
702 if (error)
703 return error;
704 if (IS_APPEND(dir))
705 return -EPERM;
706 if (btrfs_check_sticky(dir, victim->d_inode)||
707 IS_APPEND(victim->d_inode)||
708 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
709 return -EPERM;
710 if (isdir) {
711 if (!S_ISDIR(victim->d_inode->i_mode))
712 return -ENOTDIR;
713 if (IS_ROOT(victim))
714 return -EBUSY;
715 } else if (S_ISDIR(victim->d_inode->i_mode))
716 return -EISDIR;
717 if (IS_DEADDIR(dir))
718 return -ENOENT;
719 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
720 return -EBUSY;
721 return 0;
722}
723
cb8e7090
CH		 724/* copy of may_create in fs/namei.c() */
725static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
726{
727 if (child->d_inode)
728 return -EEXIST;
729 if (IS_DEADDIR(dir))
730 return -ENOENT;
731 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
732}
733
734/*
735 * Create a new subvolume below @parent. This is largely modeled after
736 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
737 * inside this filesystem so it's quite a bit simpler.
738 */
76dda93c
YZ		 739static noinline int btrfs_mksubvol(struct path *parent,
740 char *name, int namelen,
72fd032e 741 struct btrfs_root *snap_src,
6f72c7e2 742 u64 *async_transid, bool readonly,
8696c533 743 struct btrfs_qgroup_inherit *inherit)
cb8e7090 744{
76dda93c 745 struct inode *dir = parent->dentry->d_inode;
cb8e7090
CH		 746 struct dentry *dentry;
747 int error;
748
5c50c9b8
DS		 749 error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
750 if (error == -EINTR)
751 return error;
cb8e7090
CH		 752
753 dentry = lookup_one_len(name, parent->dentry, namelen);
754 error = PTR_ERR(dentry);
755 if (IS_ERR(dentry))
756 goto out_unlock;
757
758 error = -EEXIST;
759 if (dentry->d_inode)
760 goto out_dput;
761
76dda93c 762 error = btrfs_may_create(dir, dentry);
cb8e7090 763 if (error)
a874a63e 764 goto out_dput;
cb8e7090 765
9c52057c
CM		 766 /*
767 * even if this name doesn't exist, we may get hash collisions.
768 * check for them now when we can safely fail
769 */
770 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
771 dir->i_ino, name,
772 namelen);
773 if (error)
774 goto out_dput;
775
76dda93c
YZ		 776 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
777
778 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
779 goto out_up_read;
780
3de4586c 781 if (snap_src) {
e9662f70 782 error = create_snapshot(snap_src, dir, dentry, name, namelen,
6f72c7e2 783 async_transid, readonly, inherit);
3de4586c 784 } else {
d5c12070
MX		 785 error = create_subvol(dir, dentry, name, namelen,
786 async_transid, inherit);
3de4586c 787 }
76dda93c
YZ		 788 if (!error)
789 fsnotify_mkdir(dir, dentry);
790out_up_read:
791 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
cb8e7090
CH		 792out_dput:
793 dput(dentry);
794out_unlock:
76dda93c 795 mutex_unlock(&dir->i_mutex);
cb8e7090
CH		 796 return error;
797}
798
4cb5300b
CM		 799/*
800 * When we're defragging a range, we don't want to kick it off again
801 * if it is really just waiting for delalloc to send it down.
802 * If we find a nice big extent or delalloc range for the bytes in the
803 * file you want to defrag, we return 0 to let you know to skip this
804 * part of the file
805 */
806static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
807{
808 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
809 struct extent_map *em = NULL;
810 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
811 u64 end;
812
813 read_lock(&em_tree->lock);
814 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
815 read_unlock(&em_tree->lock);
816
817 if (em) {
818 end = extent_map_end(em);
819 free_extent_map(em);
820 if (end - offset > thresh)
821 return 0;
822 }
823 /* if we already have a nice delalloc here, just stop */
824 thresh /= 2;
825 end = count_range_bits(io_tree, &offset, offset + thresh,
826 thresh, EXTENT_DELALLOC, 1);
827 if (end >= thresh)
828 return 0;
829 return 1;
830}
831
832/*
833 * helper function to walk through a file and find extents
834 * newer than a specific transid, and smaller than thresh.
835 *
836 * This is used by the defragging code to find new and small
837 * extents
838 */
839static int find_new_extents(struct btrfs_root *root,
840 struct inode *inode, u64 newer_than,
841 u64 *off, int thresh)
842{
843 struct btrfs_path *path;
844 struct btrfs_key min_key;
845 struct btrfs_key max_key;
846 struct extent_buffer *leaf;
847 struct btrfs_file_extent_item *extent;
848 int type;
849 int ret;
a4689d2b 850 u64 ino = btrfs_ino(inode);
4cb5300b
CM		 851
852 path = btrfs_alloc_path();
853 if (!path)
854 return -ENOMEM;
855
a4689d2b 856 min_key.objectid = ino;
4cb5300b
CM		 857 min_key.type = BTRFS_EXTENT_DATA_KEY;
858 min_key.offset = *off;
859
a4689d2b 860 max_key.objectid = ino;
4cb5300b
CM		 861 max_key.type = (u8)-1;
862 max_key.offset = (u64)-1;
863
864 path->keep_locks = 1;
865
866 while(1) {
867 ret = btrfs_search_forward(root, &min_key, &max_key,
de78b51a 868 path, newer_than);
4cb5300b
CM		 869 if (ret != 0)
870 goto none;
a4689d2b 871 if (min_key.objectid != ino)
4cb5300b
CM		 872 goto none;
873 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
874 goto none;
875
876 leaf = path->nodes[0];
877 extent = btrfs_item_ptr(leaf, path->slots[0],
878 struct btrfs_file_extent_item);
879
880 type = btrfs_file_extent_type(leaf, extent);
881 if (type == BTRFS_FILE_EXTENT_REG &&
882 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
883 check_defrag_in_cache(inode, min_key.offset, thresh)) {
884 *off = min_key.offset;
885 btrfs_free_path(path);
886 return 0;
887 }
888
889 if (min_key.offset == (u64)-1)
890 goto none;
891
892 min_key.offset++;
893 btrfs_release_path(path);
894 }
895none:
896 btrfs_free_path(path);
897 return -ENOENT;
898}
899
6c282eb4 900static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
17ce6ef8
LB		 901{
902 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
6c282eb4
LZ		 903 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
904 struct extent_map *em;
905 u64 len = PAGE_CACHE_SIZE;
17ce6ef8 906
6c282eb4
LZ		 907 /*
908 * hopefully we have this extent in the tree already, try without
909 * the full extent lock
910 */
17ce6ef8 911 read_lock(&em_tree->lock);
6c282eb4 912 em = lookup_extent_mapping(em_tree, start, len);
17ce6ef8
LB		 913 read_unlock(&em_tree->lock);
914
6c282eb4
LZ		 915 if (!em) {
916 /* get the big lock and read metadata off disk */
917 lock_extent(io_tree, start, start + len - 1);
918 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
919 unlock_extent(io_tree, start, start + len - 1);
920
921 if (IS_ERR(em))
922 return NULL;
923 }
924
925 return em;
926}
17ce6ef8 927
6c282eb4
LZ		 928static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
929{
930 struct extent_map *next;
931 bool ret = true;
932
933 /* this is the last extent */
934 if (em->start + em->len >= i_size_read(inode))
935 return false;
936
937 next = defrag_lookup_extent(inode, em->start + em->len);
938 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
939 ret = false;
940
941 free_extent_map(next);
17ce6ef8
LB		 942 return ret;
943}
944
6c282eb4 945static int should_defrag_range(struct inode *inode, u64 start, int thresh,
a43a2111
AM		 946 u64 *last_len, u64 *skip, u64 *defrag_end,
947 int compress)
940100a4 948{
6c282eb4 949 struct extent_map *em;
940100a4 950 int ret = 1;
6c282eb4 951 bool next_mergeable = true;
940100a4
CM		 952
953 /*
008873ea 954 * make sure that once we start defragging an extent, we keep on
940100a4
CM		 955 * defragging it
956 */
957 if (start < *defrag_end)
958 return 1;
959
960 *skip = 0;
961
6c282eb4
LZ		 962 em = defrag_lookup_extent(inode, start);
963 if (!em)
964 return 0;
940100a4
CM		 965
966 /* this will cover holes, and inline extents */
17ce6ef8 967 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
940100a4 968 ret = 0;
17ce6ef8
LB		 969 goto out;
970 }
971
6c282eb4 972 next_mergeable = defrag_check_next_extent(inode, em);
940100a4
CM		 973
974 /*
6c282eb4
LZ		 975 * we hit a real extent, if it is big or the next extent is not a
976 * real extent, don't bother defragging it
940100a4 977 */
a43a2111 978 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
6c282eb4 979 (em->len >= thresh || !next_mergeable))
940100a4 980 ret = 0;
17ce6ef8 981out:
940100a4
CM		 982 /*
983 * last_len ends up being a counter of how many bytes we've defragged.
984 * every time we choose not to defrag an extent, we reset *last_len
985 * so that the next tiny extent will force a defrag.
986 *
987 * The end result of this is that tiny extents before a single big
988 * extent will force at least part of that big extent to be defragged.
989 */
990 if (ret) {
940100a4
CM		 991 *defrag_end = extent_map_end(em);
992 } else {
993 *last_len = 0;
994 *skip = extent_map_end(em);
995 *defrag_end = 0;
996 }
997
998 free_extent_map(em);
999 return ret;
1000}
1001
4cb5300b
CM		 1002/*
1003 * it doesn't do much good to defrag one or two pages
1004 * at a time. This pulls in a nice chunk of pages
1005 * to COW and defrag.
1006 *
1007 * It also makes sure the delalloc code has enough
1008 * dirty data to avoid making new small extents as part
1009 * of the defrag
1010 *
1011 * It's a good idea to start RA on this range
1012 * before calling this.
1013 */
1014static int cluster_pages_for_defrag(struct inode *inode,
1015 struct page **pages,
1016 unsigned long start_index,
1017 int num_pages)
f46b5a66 1018{
4cb5300b
CM		 1019 unsigned long file_end;
1020 u64 isize = i_size_read(inode);
1021 u64 page_start;
1022 u64 page_end;
1f12bd06 1023 u64 page_cnt;
4cb5300b
CM		 1024 int ret;
1025 int i;
1026 int i_done;
3eaa2885 1027 struct btrfs_ordered_extent *ordered;
4cb5300b 1028 struct extent_state *cached_state = NULL;
600a45e1 1029 struct extent_io_tree *tree;
3b16a4e3 1030 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
4cb5300b 1031
4cb5300b 1032 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1f12bd06
LB		 1033 if (!isize || start_index > file_end)
1034 return 0;
1035
1036 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
4cb5300b
CM		 1037
1038 ret = btrfs_delalloc_reserve_space(inode,
1f12bd06 1039 page_cnt << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1040 if (ret)
1041 return ret;
4cb5300b 1042 i_done = 0;
600a45e1 1043 tree = &BTRFS_I(inode)->io_tree;
4cb5300b
CM		 1044
1045 /* step one, lock all the pages */
1f12bd06 1046 for (i = 0; i < page_cnt; i++) {
4cb5300b 1047 struct page *page;
600a45e1 1048again:
a94733d0 1049 page = find_or_create_page(inode->i_mapping,
600a45e1 1050 start_index + i, mask);
4cb5300b
CM		 1051 if (!page)
1052 break;
1053
600a45e1
MX		 1054 page_start = page_offset(page);
1055 page_end = page_start + PAGE_CACHE_SIZE - 1;
1056 while (1) {
d0082371 1057 lock_extent(tree, page_start, page_end);
600a45e1
MX		 1058 ordered = btrfs_lookup_ordered_extent(inode,
1059 page_start);
d0082371 1060 unlock_extent(tree, page_start, page_end);
600a45e1
MX		 1061 if (!ordered)
1062 break;
1063
1064 unlock_page(page);
1065 btrfs_start_ordered_extent(inode, ordered, 1);
1066 btrfs_put_ordered_extent(ordered);
1067 lock_page(page);
1f12bd06
LB		 1068 /*
1069 * we unlocked the page above, so we need check if
1070 * it was released or not.
1071 */
1072 if (page->mapping != inode->i_mapping) {
1073 unlock_page(page);
1074 page_cache_release(page);
1075 goto again;
1076 }
600a45e1
MX		 1077 }
1078
4cb5300b
CM		 1079 if (!PageUptodate(page)) {
1080 btrfs_readpage(NULL, page);
1081 lock_page(page);
1082 if (!PageUptodate(page)) {
1083 unlock_page(page);
1084 page_cache_release(page);
1085 ret = -EIO;
1086 break;
1087 }
1088 }
600a45e1 1089
600a45e1
MX		 1090 if (page->mapping != inode->i_mapping) {
1091 unlock_page(page);
1092 page_cache_release(page);
1093 goto again;
1094 }
1095
4cb5300b
CM		 1096 pages[i] = page;
1097 i_done++;
1098 }
1099 if (!i_done || ret)
1100 goto out;
1101
1102 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1103 goto out;
1104
1105 /*
1106 * so now we have a nice long stream of locked
1107 * and up to date pages, lets wait on them
1108 */
1109 for (i = 0; i < i_done; i++)
1110 wait_on_page_writeback(pages[i]);
1111
1112 page_start = page_offset(pages[0]);
1113 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1114
1115 lock_extent_bits(&BTRFS_I(inode)->io_tree,
d0082371 1116 page_start, page_end - 1, 0, &cached_state);
4cb5300b
CM		 1117 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1118 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
9e8a4a8b
LB		 1119 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1120 &cached_state, GFP_NOFS);
4cb5300b 1121
1f12bd06 1122 if (i_done != page_cnt) {
9e0baf60
JB		 1123 spin_lock(&BTRFS_I(inode)->lock);
1124 BTRFS_I(inode)->outstanding_extents++;
1125 spin_unlock(&BTRFS_I(inode)->lock);
4cb5300b 1126 btrfs_delalloc_release_space(inode,
1f12bd06 1127 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1128 }
1129
1130
9e8a4a8b
LB		 1131 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1132 &cached_state, GFP_NOFS);
4cb5300b
CM		 1133
1134 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1135 page_start, page_end - 1, &cached_state,
1136 GFP_NOFS);
1137
1138 for (i = 0; i < i_done; i++) {
1139 clear_page_dirty_for_io(pages[i]);
1140 ClearPageChecked(pages[i]);
1141 set_page_extent_mapped(pages[i]);
1142 set_page_dirty(pages[i]);
1143 unlock_page(pages[i]);
1144 page_cache_release(pages[i]);
1145 }
1146 return i_done;
1147out:
1148 for (i = 0; i < i_done; i++) {
1149 unlock_page(pages[i]);
1150 page_cache_release(pages[i]);
1151 }
1f12bd06 1152 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1153 return ret;
1154
1155}
1156
1157int btrfs_defrag_file(struct inode *inode, struct file *file,
1158 struct btrfs_ioctl_defrag_range_args *range,
1159 u64 newer_than, unsigned long max_to_defrag)
1160{
1161 struct btrfs_root *root = BTRFS_I(inode)->root;
4cb5300b 1162 struct file_ra_state *ra = NULL;
f46b5a66 1163 unsigned long last_index;
151a31b2 1164 u64 isize = i_size_read(inode);
940100a4
CM		 1165 u64 last_len = 0;
1166 u64 skip = 0;
1167 u64 defrag_end = 0;
4cb5300b 1168 u64 newer_off = range->start;
f46b5a66 1169 unsigned long i;
008873ea 1170 unsigned long ra_index = 0;
f46b5a66 1171 int ret;
4cb5300b 1172 int defrag_count = 0;
1a419d85 1173 int compress_type = BTRFS_COMPRESS_ZLIB;
4cb5300b 1174 int extent_thresh = range->extent_thresh;
008873ea
LZ		 1175 int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1176 int cluster = max_cluster;
4cb5300b
CM		 1177 u64 new_align = ~((u64)128 * 1024 - 1);
1178 struct page **pages = NULL;
1179
0abd5b17
LB		 1180 if (isize == 0)
1181 return 0;
1182
1183 if (range->start >= isize)
1184 return -EINVAL;
1a419d85
LZ		 1185
1186 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1187 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1188 return -EINVAL;
1189 if (range->compress_type)
1190 compress_type = range->compress_type;
1191 }
f46b5a66 1192
0abd5b17
LB		 1193 if (extent_thresh == 0)
1194 extent_thresh = 256 * 1024;
940100a4 1195
4cb5300b
CM		 1196 /*
1197 * if we were not given a file, allocate a readahead
1198 * context
1199 */
1200 if (!file) {
1201 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1202 if (!ra)
1203 return -ENOMEM;
1204 file_ra_state_init(ra, inode->i_mapping);
1205 } else {
1206 ra = &file->f_ra;
1207 }
1208
008873ea 1209 pages = kmalloc(sizeof(struct page *) * max_cluster,
4cb5300b
CM		 1210 GFP_NOFS);
1211 if (!pages) {
1212 ret = -ENOMEM;
1213 goto out_ra;
1214 }
1215
1216 /* find the last page to defrag */
1e701a32 1217 if (range->start + range->len > range->start) {
151a31b2 1218 last_index = min_t(u64, isize - 1,
1e701a32
CM		 1219 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1220 } else {
151a31b2 1221 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1e701a32
CM		 1222 }
1223
4cb5300b
CM		 1224 if (newer_than) {
1225 ret = find_new_extents(root, inode, newer_than,
1226 &newer_off, 64 * 1024);
1227 if (!ret) {
1228 range->start = newer_off;
1229 /*
1230 * we always align our defrag to help keep
1231 * the extents in the file evenly spaced
1232 */
1233 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
4cb5300b
CM		 1234 } else
1235 goto out_ra;
1236 } else {
1237 i = range->start >> PAGE_CACHE_SHIFT;
1238 }
1239 if (!max_to_defrag)
7ec31b54 1240 max_to_defrag = last_index + 1;
4cb5300b 1241
2a0f7f57
LZ		 1242 /*
1243 * make writeback starts from i, so the defrag range can be
1244 * written sequentially.
1245 */
1246 if (i < inode->i_mapping->writeback_index)
1247 inode->i_mapping->writeback_index = i;
1248
f7f43cc8
CM		 1249 while (i <= last_index && defrag_count < max_to_defrag &&
1250 (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1251 PAGE_CACHE_SHIFT)) {
4cb5300b
CM		 1252 /*
1253 * make sure we stop running if someone unmounts
1254 * the FS
1255 */
1256 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1257 break;
1258
210549eb
DS		 1259 if (btrfs_defrag_cancelled(root->fs_info)) {
1260 printk(KERN_DEBUG "btrfs: defrag_file cancelled\n");
1261 ret = -EAGAIN;
1262 break;
1263 }
1264
66c26892 1265 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
6c282eb4 1266 extent_thresh, &last_len, &skip,
a43a2111
AM		 1267 &defrag_end, range->flags &
1268 BTRFS_DEFRAG_RANGE_COMPRESS)) {
940100a4
CM		 1269 unsigned long next;
1270 /*
1271 * the should_defrag function tells us how much to skip
1272 * bump our counter by the suggested amount
1273 */
1274 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1275 i = max(i + 1, next);
1276 continue;
1277 }
008873ea
LZ		 1278
1279 if (!newer_than) {
1280 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1281 PAGE_CACHE_SHIFT) - i;
1282 cluster = min(cluster, max_cluster);
1283 } else {
1284 cluster = max_cluster;
1285 }
1286
008873ea
LZ		 1287 if (i + cluster > ra_index) {
1288 ra_index = max(i, ra_index);
1289 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1290 cluster);
1291 ra_index += max_cluster;
1292 }
940100a4 1293
ecb8bea8 1294 mutex_lock(&inode->i_mutex);
633085c7
FDBM		 1295 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1296 BTRFS_I(inode)->force_compress = compress_type;
008873ea 1297 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
ecb8bea8
LB		 1298 if (ret < 0) {
1299 mutex_unlock(&inode->i_mutex);
4cb5300b 1300 goto out_ra;
ecb8bea8 1301 }
4cb5300b
CM		 1302
1303 defrag_count += ret;
d0e1d66b 1304 balance_dirty_pages_ratelimited(inode->i_mapping);
ecb8bea8 1305 mutex_unlock(&inode->i_mutex);
4cb5300b
CM		 1306
1307 if (newer_than) {
1308 if (newer_off == (u64)-1)
1309 break;
1310
e1f041e1
LB		 1311 if (ret > 0)
1312 i += ret;
1313
4cb5300b
CM		 1314 newer_off = max(newer_off + 1,
1315 (u64)i << PAGE_CACHE_SHIFT);
1316
1317 ret = find_new_extents(root, inode,
1318 newer_than, &newer_off,
1319 64 * 1024);
1320 if (!ret) {
1321 range->start = newer_off;
1322 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
4cb5300b
CM		 1323 } else {
1324 break;
f46b5a66 1325 }
4cb5300b 1326 } else {
008873ea 1327 if (ret > 0) {
cbcc8326 1328 i += ret;
008873ea
LZ		 1329 last_len += ret << PAGE_CACHE_SHIFT;
1330 } else {
cbcc8326 1331 i++;
008873ea
LZ		 1332 last_len = 0;
1333 }
f46b5a66 1334 }
f46b5a66
CH		 1335 }
1336
1e701a32
CM		 1337 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1338 filemap_flush(inode->i_mapping);
1339
1340 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1341 /* the filemap_flush will queue IO into the worker threads, but
1342 * we have to make sure the IO is actually started and that
1343 * ordered extents get created before we return
1344 */
1345 atomic_inc(&root->fs_info->async_submit_draining);
1346 while (atomic_read(&root->fs_info->nr_async_submits) ||
1347 atomic_read(&root->fs_info->async_delalloc_pages)) {
1348 wait_event(root->fs_info->async_submit_wait,
1349 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1350 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1351 }
1352 atomic_dec(&root->fs_info->async_submit_draining);
1e701a32
CM		 1353 }
1354
1a419d85 1355 if (range->compress_type == BTRFS_COMPRESS_LZO) {
2b0ce2c2 1356 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1a419d85
LZ		 1357 }
1358
60ccf82f 1359 ret = defrag_count;
940100a4 1360
4cb5300b 1361out_ra:
633085c7
FDBM		 1362 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1363 mutex_lock(&inode->i_mutex);
1364 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1365 mutex_unlock(&inode->i_mutex);
1366 }
4cb5300b
CM		 1367 if (!file)
1368 kfree(ra);
1369 kfree(pages);
940100a4 1370 return ret;
f46b5a66
CH		 1371}
1372
198605a8 1373static noinline int btrfs_ioctl_resize(struct file *file,
76dda93c 1374 void __user *arg)
f46b5a66
CH		 1375{
1376 u64 new_size;
1377 u64 old_size;
1378 u64 devid = 1;
496ad9aa 1379 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
f46b5a66
CH		 1380 struct btrfs_ioctl_vol_args *vol_args;
1381 struct btrfs_trans_handle *trans;
1382 struct btrfs_device *device = NULL;
1383 char *sizestr;
1384 char *devstr = NULL;
1385 int ret = 0;
f46b5a66
CH		 1386 int mod = 0;
1387
e441d54d
CM		 1388 if (!capable(CAP_SYS_ADMIN))
1389 return -EPERM;
1390
198605a8
MX		 1391 ret = mnt_want_write_file(file);
1392 if (ret)
1393 return ret;
1394
5ac00add
SB		 1395 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1396 1)) {
1397 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
97547676 1398 mnt_drop_write_file(file);
2c0c9da0 1399 return -EINVAL;
c9e9f97b
ID		 1400 }
1401
5ac00add 1402 mutex_lock(&root->fs_info->volume_mutex);
dae7b665 1403 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 1404 if (IS_ERR(vol_args)) {
1405 ret = PTR_ERR(vol_args);
1406 goto out;
1407 }
5516e595
MF		 1408
1409 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 1410
f46b5a66
CH		 1411 sizestr = vol_args->name;
1412 devstr = strchr(sizestr, ':');
1413 if (devstr) {
1414 char *end;
1415 sizestr = devstr + 1;
1416 *devstr = '\0';
1417 devstr = vol_args->name;
1418 devid = simple_strtoull(devstr, &end, 10);
dfd79829
MX		 1419 if (!devid) {
1420 ret = -EINVAL;
1421 goto out_free;
1422 }
c1c9ff7c 1423 printk(KERN_INFO "btrfs: resizing devid %llu\n", devid);
f46b5a66 1424 }
dba60f3f 1425
aa1b8cd4 1426 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
f46b5a66 1427 if (!device) {
5bb14682 1428 printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
c1c9ff7c 1429 devid);
dfd79829 1430 ret = -ENODEV;
c9e9f97b 1431 goto out_free;
f46b5a66 1432 }
dba60f3f
MX		 1433
1434 if (!device->writeable) {
4e42ae1b 1435 printk(KERN_INFO "btrfs: resizer unable to apply on "
dba60f3f 1436 "readonly device %llu\n",
c1c9ff7c 1437 devid);
dfd79829 1438 ret = -EPERM;
4e42ae1b
LB		 1439 goto out_free;
1440 }
1441
f46b5a66
CH		 1442 if (!strcmp(sizestr, "max"))
1443 new_size = device->bdev->bd_inode->i_size;
1444 else {
1445 if (sizestr[0] == '-') {
1446 mod = -1;
1447 sizestr++;
1448 } else if (sizestr[0] == '+') {
1449 mod = 1;
1450 sizestr++;
1451 }
91748467 1452 new_size = memparse(sizestr, NULL);
f46b5a66
CH		 1453 if (new_size == 0) {
1454 ret = -EINVAL;
c9e9f97b 1455 goto out_free;
f46b5a66
CH		 1456 }
1457 }
1458
63a212ab 1459 if (device->is_tgtdev_for_dev_replace) {
dfd79829 1460 ret = -EPERM;
63a212ab
SB		 1461 goto out_free;
1462 }
1463
f46b5a66
CH		 1464 old_size = device->total_bytes;
1465
1466 if (mod < 0) {
1467 if (new_size > old_size) {
1468 ret = -EINVAL;
c9e9f97b 1469 goto out_free;
f46b5a66
CH		 1470 }
1471 new_size = old_size - new_size;
1472 } else if (mod > 0) {
1473 new_size = old_size + new_size;
1474 }
1475
1476 if (new_size < 256 * 1024 * 1024) {
1477 ret = -EINVAL;
c9e9f97b 1478 goto out_free;
f46b5a66
CH		 1479 }
1480 if (new_size > device->bdev->bd_inode->i_size) {
1481 ret = -EFBIG;
c9e9f97b 1482 goto out_free;
f46b5a66
CH		 1483 }
1484
1485 do_div(new_size, root->sectorsize);
1486 new_size *= root->sectorsize;
1487
606686ee 1488 printk_in_rcu(KERN_INFO "btrfs: new size for %s is %llu\n",
c1c9ff7c 1489 rcu_str_deref(device->name), new_size);
f46b5a66
CH		 1490
1491 if (new_size > old_size) {
a22285a6 1492 trans = btrfs_start_transaction(root, 0);
98d5dc13
TI		 1493 if (IS_ERR(trans)) {
1494 ret = PTR_ERR(trans);
c9e9f97b 1495 goto out_free;
98d5dc13 1496 }
f46b5a66
CH		 1497 ret = btrfs_grow_device(trans, device, new_size);
1498 btrfs_commit_transaction(trans, root);
ece7d20e 1499 } else if (new_size < old_size) {
f46b5a66 1500 ret = btrfs_shrink_device(device, new_size);
0253f40e 1501 } /* equal, nothing need to do */
f46b5a66 1502
c9e9f97b 1503out_free:
f46b5a66 1504 kfree(vol_args);
c9e9f97b
ID		 1505out:
1506 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 1507 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
18f39c41 1508 mnt_drop_write_file(file);
f46b5a66
CH		 1509 return ret;
1510}
1511
72fd032e 1512static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
6f72c7e2
AJ		 1513 char *name, unsigned long fd, int subvol,
1514 u64 *transid, bool readonly,
8696c533 1515 struct btrfs_qgroup_inherit *inherit)
f46b5a66 1516{
f46b5a66 1517 int namelen;
3de4586c 1518 int ret = 0;
f46b5a66 1519
a874a63e
LB		 1520 ret = mnt_want_write_file(file);
1521 if (ret)
1522 goto out;
1523
72fd032e
SW		 1524 namelen = strlen(name);
1525 if (strchr(name, '/')) {
f46b5a66 1526 ret = -EINVAL;
a874a63e 1527 goto out_drop_write;
f46b5a66
CH		 1528 }
1529
16780cab
CM		 1530 if (name[0] == '.' &&
1531 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1532 ret = -EEXIST;
a874a63e 1533 goto out_drop_write;
16780cab
CM		 1534 }
1535
3de4586c 1536 if (subvol) {
72fd032e 1537 ret = btrfs_mksubvol(&file->f_path, name, namelen,
6f72c7e2 1538 NULL, transid, readonly, inherit);
cb8e7090 1539 } else {
2903ff01 1540 struct fd src = fdget(fd);
3de4586c 1541 struct inode *src_inode;
2903ff01 1542 if (!src.file) {
3de4586c 1543 ret = -EINVAL;
a874a63e 1544 goto out_drop_write;
3de4586c
CM		 1545 }
1546
496ad9aa
AV		 1547 src_inode = file_inode(src.file);
1548 if (src_inode->i_sb != file_inode(file)->i_sb) {
d397712b
CM		 1549 printk(KERN_INFO "btrfs: Snapshot src from "
1550 "another FS\n");
3de4586c 1551 ret = -EINVAL;
ecd18815
AV		 1552 } else {
1553 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1554 BTRFS_I(src_inode)->root,
1555 transid, readonly, inherit);
3de4586c 1556 }
2903ff01 1557 fdput(src);
cb8e7090 1558 }
a874a63e
LB		 1559out_drop_write:
1560 mnt_drop_write_file(file);
f46b5a66 1561out:
72fd032e
SW		 1562 return ret;
1563}
1564
1565static noinline int btrfs_ioctl_snap_create(struct file *file,
fa0d2b9b 1566 void __user *arg, int subvol)
72fd032e 1567{
fa0d2b9b 1568 struct btrfs_ioctl_vol_args *vol_args;
72fd032e
SW		 1569 int ret;
1570
fa0d2b9b
LZ		 1571 vol_args = memdup_user(arg, sizeof(*vol_args));
1572 if (IS_ERR(vol_args))
1573 return PTR_ERR(vol_args);
1574 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
72fd032e 1575
fa0d2b9b 1576 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
b83cc969 1577 vol_args->fd, subvol,
6f72c7e2 1578 NULL, false, NULL);
fdfb1e4f 1579
fa0d2b9b
LZ		 1580 kfree(vol_args);
1581 return ret;
1582}
fdfb1e4f 1583
fa0d2b9b
LZ		 1584static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1585 void __user *arg, int subvol)
1586{
1587 struct btrfs_ioctl_vol_args_v2 *vol_args;
1588 int ret;
1589 u64 transid = 0;
1590 u64 *ptr = NULL;
b83cc969 1591 bool readonly = false;
6f72c7e2 1592 struct btrfs_qgroup_inherit *inherit = NULL;
75eaa0e2 1593
fa0d2b9b
LZ		 1594 vol_args = memdup_user(arg, sizeof(*vol_args));
1595 if (IS_ERR(vol_args))
1596 return PTR_ERR(vol_args);
1597 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
75eaa0e2 1598
b83cc969 1599 if (vol_args->flags &
6f72c7e2
AJ		 1600 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1601 BTRFS_SUBVOL_QGROUP_INHERIT)) {
b83cc969 1602 ret = -EOPNOTSUPP;
fa0d2b9b 1603 goto out;
72fd032e 1604 }
fa0d2b9b
LZ		 1605
1606 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1607 ptr = &transid;
b83cc969
LZ		 1608 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1609 readonly = true;
6f72c7e2
AJ		 1610 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1611 if (vol_args->size > PAGE_CACHE_SIZE) {
1612 ret = -EINVAL;
1613 goto out;
1614 }
1615 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1616 if (IS_ERR(inherit)) {
1617 ret = PTR_ERR(inherit);
1618 goto out;
1619 }
1620 }
fa0d2b9b
LZ		 1621
1622 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
6f72c7e2 1623 vol_args->fd, subvol, ptr,
8696c533 1624 readonly, inherit);
fa0d2b9b
LZ		 1625
1626 if (ret == 0 && ptr &&
1627 copy_to_user(arg +
1628 offsetof(struct btrfs_ioctl_vol_args_v2,
1629 transid), ptr, sizeof(*ptr)))
1630 ret = -EFAULT;
fdfb1e4f 1631out:
f46b5a66 1632 kfree(vol_args);
6f72c7e2 1633 kfree(inherit);
f46b5a66
CH		 1634 return ret;
1635}
1636
0caa102d
LZ		 1637static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1638 void __user *arg)
1639{
496ad9aa 1640 struct inode *inode = file_inode(file);
0caa102d
LZ		 1641 struct btrfs_root *root = BTRFS_I(inode)->root;
1642 int ret = 0;
1643 u64 flags = 0;
1644
33345d01 1645 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
0caa102d
LZ		 1646 return -EINVAL;
1647
1648 down_read(&root->fs_info->subvol_sem);
1649 if (btrfs_root_readonly(root))
1650 flags |= BTRFS_SUBVOL_RDONLY;
1651 up_read(&root->fs_info->subvol_sem);
1652
1653 if (copy_to_user(arg, &flags, sizeof(flags)))
1654 ret = -EFAULT;
1655
1656 return ret;
1657}
1658
1659static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1660 void __user *arg)
1661{
496ad9aa 1662 struct inode *inode = file_inode(file);
0caa102d
LZ		 1663 struct btrfs_root *root = BTRFS_I(inode)->root;
1664 struct btrfs_trans_handle *trans;
1665 u64 root_flags;
1666 u64 flags;
1667 int ret = 0;
1668
b9ca0664
LB		 1669 ret = mnt_want_write_file(file);
1670 if (ret)
1671 goto out;
0caa102d 1672
b9ca0664
LB		 1673 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1674 ret = -EINVAL;
1675 goto out_drop_write;
1676 }
0caa102d 1677
b9ca0664
LB		 1678 if (copy_from_user(&flags, arg, sizeof(flags))) {
1679 ret = -EFAULT;
1680 goto out_drop_write;
1681 }
0caa102d 1682
b9ca0664
LB		 1683 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1684 ret = -EINVAL;
1685 goto out_drop_write;
1686 }
0caa102d 1687
b9ca0664
LB		 1688 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1689 ret = -EOPNOTSUPP;
1690 goto out_drop_write;
1691 }
0caa102d 1692
b9ca0664
LB		 1693 if (!inode_owner_or_capable(inode)) {
1694 ret = -EACCES;
1695 goto out_drop_write;
1696 }
b4dc2b8c 1697
0caa102d
LZ		 1698 down_write(&root->fs_info->subvol_sem);
1699
1700 /* nothing to do */
1701 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
b9ca0664 1702 goto out_drop_sem;
0caa102d
LZ		 1703
1704 root_flags = btrfs_root_flags(&root->root_item);
1705 if (flags & BTRFS_SUBVOL_RDONLY)
1706 btrfs_set_root_flags(&root->root_item,
1707 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1708 else
1709 btrfs_set_root_flags(&root->root_item,
1710 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1711
1712 trans = btrfs_start_transaction(root, 1);
1713 if (IS_ERR(trans)) {
1714 ret = PTR_ERR(trans);
1715 goto out_reset;
1716 }
1717
b4dc2b8c 1718 ret = btrfs_update_root(trans, root->fs_info->tree_root,
0caa102d
LZ		 1719 &root->root_key, &root->root_item);
1720
1721 btrfs_commit_transaction(trans, root);
1722out_reset:
1723 if (ret)
1724 btrfs_set_root_flags(&root->root_item, root_flags);
b9ca0664 1725out_drop_sem:
0caa102d 1726 up_write(&root->fs_info->subvol_sem);
b9ca0664
LB		 1727out_drop_write:
1728 mnt_drop_write_file(file);
1729out:
0caa102d
LZ		 1730 return ret;
1731}
1732
76dda93c
YZ		 1733/*
1734 * helper to check if the subvolume references other subvolumes
1735 */
1736static noinline int may_destroy_subvol(struct btrfs_root *root)
1737{
1738 struct btrfs_path *path;
175a2b87 1739 struct btrfs_dir_item *di;
76dda93c 1740 struct btrfs_key key;
175a2b87 1741 u64 dir_id;
76dda93c
YZ		 1742 int ret;
1743
1744 path = btrfs_alloc_path();
1745 if (!path)
1746 return -ENOMEM;
1747
175a2b87
JB		 1748 /* Make sure this root isn't set as the default subvol */
1749 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1750 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1751 dir_id, "default", 7, 0);
1752 if (di && !IS_ERR(di)) {
1753 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1754 if (key.objectid == root->root_key.objectid) {
1755 ret = -ENOTEMPTY;
1756 goto out;
1757 }
1758 btrfs_release_path(path);
1759 }
1760
76dda93c
YZ		 1761 key.objectid = root->root_key.objectid;
1762 key.type = BTRFS_ROOT_REF_KEY;
1763 key.offset = (u64)-1;
1764
1765 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1766 &key, path, 0, 0);
1767 if (ret < 0)
1768 goto out;
1769 BUG_ON(ret == 0);
1770
1771 ret = 0;
1772 if (path->slots[0] > 0) {
1773 path->slots[0]--;
1774 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1775 if (key.objectid == root->root_key.objectid &&
1776 key.type == BTRFS_ROOT_REF_KEY)
1777 ret = -ENOTEMPTY;
1778 }
1779out:
1780 btrfs_free_path(path);
1781 return ret;
1782}
1783
ac8e9819
CM		 1784static noinline int key_in_sk(struct btrfs_key *key,
1785 struct btrfs_ioctl_search_key *sk)
1786{
abc6e134
CM		 1787 struct btrfs_key test;
1788 int ret;
1789
1790 test.objectid = sk->min_objectid;
1791 test.type = sk->min_type;
1792 test.offset = sk->min_offset;
1793
1794 ret = btrfs_comp_cpu_keys(key, &test);
1795 if (ret < 0)
ac8e9819 1796 return 0;
abc6e134
CM		 1797
1798 test.objectid = sk->max_objectid;
1799 test.type = sk->max_type;
1800 test.offset = sk->max_offset;
1801
1802 ret = btrfs_comp_cpu_keys(key, &test);
1803 if (ret > 0)
ac8e9819
CM		 1804 return 0;
1805 return 1;
1806}
1807
1808static noinline int copy_to_sk(struct btrfs_root *root,
1809 struct btrfs_path *path,
1810 struct btrfs_key *key,
1811 struct btrfs_ioctl_search_key *sk,
1812 char *buf,
1813 unsigned long *sk_offset,
1814 int *num_found)
1815{
1816 u64 found_transid;
1817 struct extent_buffer *leaf;
1818 struct btrfs_ioctl_search_header sh;
1819 unsigned long item_off;
1820 unsigned long item_len;
1821 int nritems;
1822 int i;
1823 int slot;
ac8e9819
CM		 1824 int ret = 0;
1825
1826 leaf = path->nodes[0];
1827 slot = path->slots[0];
1828 nritems = btrfs_header_nritems(leaf);
1829
1830 if (btrfs_header_generation(leaf) > sk->max_transid) {
1831 i = nritems;
1832 goto advance_key;
1833 }
1834 found_transid = btrfs_header_generation(leaf);
1835
1836 for (i = slot; i < nritems; i++) {
1837 item_off = btrfs_item_ptr_offset(leaf, i);
1838 item_len = btrfs_item_size_nr(leaf, i);
1839
03b71c6c
GP		 1840 btrfs_item_key_to_cpu(leaf, key, i);
1841 if (!key_in_sk(key, sk))
1842 continue;
1843
1844 if (sizeof(sh) + item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
ac8e9819
CM		 1845 item_len = 0;
1846
1847 if (sizeof(sh) + item_len + *sk_offset >
1848 BTRFS_SEARCH_ARGS_BUFSIZE) {
1849 ret = 1;
1850 goto overflow;
1851 }
1852
ac8e9819
CM		 1853 sh.objectid = key->objectid;
1854 sh.offset = key->offset;
1855 sh.type = key->type;
1856 sh.len = item_len;
1857 sh.transid = found_transid;
1858
1859 /* copy search result header */
1860 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1861 *sk_offset += sizeof(sh);
1862
1863 if (item_len) {
1864 char *p = buf + *sk_offset;
1865 /* copy the item */
1866 read_extent_buffer(leaf, p,
1867 item_off, item_len);
1868 *sk_offset += item_len;
ac8e9819 1869 }
e2156867 1870 (*num_found)++;
ac8e9819
CM		 1871
1872 if (*num_found >= sk->nr_items)
1873 break;
1874 }
1875advance_key:
abc6e134
CM		 1876 ret = 0;
1877 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
ac8e9819 1878 key->offset++;
abc6e134
CM		 1879 else if (key->type < (u8)-1 && key->type < sk->max_type) {
1880 key->offset = 0;
ac8e9819 1881 key->type++;
abc6e134
CM		 1882 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1883 key->offset = 0;
1884 key->type = 0;
ac8e9819 1885 key->objectid++;
abc6e134
CM		 1886 } else
1887 ret = 1;
ac8e9819 1888overflow:
ac8e9819
CM		 1889 return ret;
1890}
1891
1892static noinline int search_ioctl(struct inode *inode,
1893 struct btrfs_ioctl_search_args *args)
1894{
1895 struct btrfs_root *root;
1896 struct btrfs_key key;
1897 struct btrfs_key max_key;
1898 struct btrfs_path *path;
1899 struct btrfs_ioctl_search_key *sk = &args->key;
1900 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1901 int ret;
1902 int num_found = 0;
1903 unsigned long sk_offset = 0;
1904
1905 path = btrfs_alloc_path();
1906 if (!path)
1907 return -ENOMEM;
1908
1909 if (sk->tree_id == 0) {
1910 /* search the root of the inode that was passed */
1911 root = BTRFS_I(inode)->root;
1912 } else {
1913 key.objectid = sk->tree_id;
1914 key.type = BTRFS_ROOT_ITEM_KEY;
1915 key.offset = (u64)-1;
1916 root = btrfs_read_fs_root_no_name(info, &key);
1917 if (IS_ERR(root)) {
1918 printk(KERN_ERR "could not find root %llu\n",
1919 sk->tree_id);
1920 btrfs_free_path(path);
1921 return -ENOENT;
1922 }
1923 }
1924
1925 key.objectid = sk->min_objectid;
1926 key.type = sk->min_type;
1927 key.offset = sk->min_offset;
1928
1929 max_key.objectid = sk->max_objectid;
1930 max_key.type = sk->max_type;
1931 max_key.offset = sk->max_offset;
1932
1933 path->keep_locks = 1;
1934
1935 while(1) {
de78b51a 1936 ret = btrfs_search_forward(root, &key, &max_key, path,
ac8e9819
CM		 1937 sk->min_transid);
1938 if (ret != 0) {
1939 if (ret > 0)
1940 ret = 0;
1941 goto err;
1942 }
1943 ret = copy_to_sk(root, path, &key, sk, args->buf,
1944 &sk_offset, &num_found);
b3b4aa74 1945 btrfs_release_path(path);
ac8e9819
CM		 1946 if (ret || num_found >= sk->nr_items)
1947 break;
1948
1949 }
1950 ret = 0;
1951err:
1952 sk->nr_items = num_found;
1953 btrfs_free_path(path);
1954 return ret;
1955}
1956
1957static noinline int btrfs_ioctl_tree_search(struct file *file,
1958 void __user *argp)
1959{
1960 struct btrfs_ioctl_search_args *args;
1961 struct inode *inode;
1962 int ret;
1963
1964 if (!capable(CAP_SYS_ADMIN))
1965 return -EPERM;
1966
2354d08f
JL		 1967 args = memdup_user(argp, sizeof(*args));
1968 if (IS_ERR(args))
1969 return PTR_ERR(args);
ac8e9819 1970
496ad9aa 1971 inode = file_inode(file);
ac8e9819
CM		 1972 ret = search_ioctl(inode, args);
1973 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1974 ret = -EFAULT;
1975 kfree(args);
1976 return ret;
1977}
1978
98d377a0 1979/*
ac8e9819
CM		 1980 * Search INODE_REFs to identify path name of 'dirid' directory
1981 * in a 'tree_id' tree. and sets path name to 'name'.
1982 */
98d377a0
TH		 1983static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1984 u64 tree_id, u64 dirid, char *name)
1985{
1986 struct btrfs_root *root;
1987 struct btrfs_key key;
ac8e9819 1988 char *ptr;
98d377a0
TH		 1989 int ret = -1;
1990 int slot;
1991 int len;
1992 int total_len = 0;
1993 struct btrfs_inode_ref *iref;
1994 struct extent_buffer *l;
1995 struct btrfs_path *path;
1996
1997 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1998 name[0]='\0';
1999 return 0;
2000 }
2001
2002 path = btrfs_alloc_path();
2003 if (!path)
2004 return -ENOMEM;
2005
ac8e9819 2006 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
98d377a0
TH		 2007
2008 key.objectid = tree_id;
2009 key.type = BTRFS_ROOT_ITEM_KEY;
2010 key.offset = (u64)-1;
2011 root = btrfs_read_fs_root_no_name(info, &key);
2012 if (IS_ERR(root)) {
2013 printk(KERN_ERR "could not find root %llu\n", tree_id);
8ad6fcab
CM		 2014 ret = -ENOENT;
2015 goto out;
98d377a0
TH		 2016 }
2017
2018 key.objectid = dirid;
2019 key.type = BTRFS_INODE_REF_KEY;
8ad6fcab 2020 key.offset = (u64)-1;
98d377a0
TH		 2021
2022 while(1) {
2023 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2024 if (ret < 0)
2025 goto out;
18674c6c
FDBM		 2026 else if (ret > 0) {
2027 ret = btrfs_previous_item(root, path, dirid,
2028 BTRFS_INODE_REF_KEY);
2029 if (ret < 0)
2030 goto out;
2031 else if (ret > 0) {
2032 ret = -ENOENT;
2033 goto out;
2034 }
2035 }
98d377a0
TH		 2036
2037 l = path->nodes[0];
2038 slot = path->slots[0];
2039 btrfs_item_key_to_cpu(l, &key, slot);
2040
98d377a0
TH		 2041 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2042 len = btrfs_inode_ref_name_len(l, iref);
2043 ptr -= len + 1;
2044 total_len += len + 1;
a696cf35
FDBM		 2045 if (ptr < name) {
2046 ret = -ENAMETOOLONG;
98d377a0 2047 goto out;
a696cf35 2048 }
98d377a0
TH		 2049
2050 *(ptr + len) = '/';
2051 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
2052
2053 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2054 break;
2055
b3b4aa74 2056 btrfs_release_path(path);
98d377a0 2057 key.objectid = key.offset;
8ad6fcab 2058 key.offset = (u64)-1;
98d377a0 2059 dirid = key.objectid;
98d377a0 2060 }
77906a50 2061 memmove(name, ptr, total_len);
98d377a0
TH		 2062 name[total_len]='\0';
2063 ret = 0;
2064out:
2065 btrfs_free_path(path);
ac8e9819
CM		 2066 return ret;
2067}
2068
2069static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2070 void __user *argp)
2071{
2072 struct btrfs_ioctl_ino_lookup_args *args;
2073 struct inode *inode;
2074 int ret;
2075
2076 if (!capable(CAP_SYS_ADMIN))
2077 return -EPERM;
2078
2354d08f
JL		 2079 args = memdup_user(argp, sizeof(*args));
2080 if (IS_ERR(args))
2081 return PTR_ERR(args);
c2b96929 2082
496ad9aa 2083 inode = file_inode(file);
ac8e9819 2084
1b53ac4d
CM		 2085 if (args->treeid == 0)
2086 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2087
ac8e9819
CM		 2088 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2089 args->treeid, args->objectid,
2090 args->name);
2091
2092 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2093 ret = -EFAULT;
2094
2095 kfree(args);
98d377a0
TH		 2096 return ret;
2097}
2098
76dda93c
YZ		 2099static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2100 void __user *arg)
2101{
2102 struct dentry *parent = fdentry(file);
2103 struct dentry *dentry;
2104 struct inode *dir = parent->d_inode;
2105 struct inode *inode;
2106 struct btrfs_root *root = BTRFS_I(dir)->root;
2107 struct btrfs_root *dest = NULL;
2108 struct btrfs_ioctl_vol_args *vol_args;
2109 struct btrfs_trans_handle *trans;
c58aaad2
MX		 2110 struct btrfs_block_rsv block_rsv;
2111 u64 qgroup_reserved;
76dda93c
YZ		 2112 int namelen;
2113 int ret;
2114 int err = 0;
2115
76dda93c
YZ		 2116 vol_args = memdup_user(arg, sizeof(*vol_args));
2117 if (IS_ERR(vol_args))
2118 return PTR_ERR(vol_args);
2119
2120 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2121 namelen = strlen(vol_args->name);
2122 if (strchr(vol_args->name, '/') ||
2123 strncmp(vol_args->name, "..", namelen) == 0) {
2124 err = -EINVAL;
2125 goto out;
2126 }
2127
a561be71 2128 err = mnt_want_write_file(file);
76dda93c
YZ		 2129 if (err)
2130 goto out;
2131
5c50c9b8
DS		 2132 err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2133 if (err == -EINTR)
2134 goto out;
76dda93c
YZ		 2135 dentry = lookup_one_len(vol_args->name, parent, namelen);
2136 if (IS_ERR(dentry)) {
2137 err = PTR_ERR(dentry);
2138 goto out_unlock_dir;
2139 }
2140
2141 if (!dentry->d_inode) {
2142 err = -ENOENT;
2143 goto out_dput;
2144 }
2145
2146 inode = dentry->d_inode;
4260f7c7
SW		 2147 dest = BTRFS_I(inode)->root;
2148 if (!capable(CAP_SYS_ADMIN)){
2149 /*
2150 * Regular user. Only allow this with a special mount
2151 * option, when the user has write+exec access to the
2152 * subvol root, and when rmdir(2) would have been
2153 * allowed.
2154 *
2155 * Note that this is _not_ check that the subvol is
2156 * empty or doesn't contain data that we wouldn't
2157 * otherwise be able to delete.
2158 *
2159 * Users who want to delete empty subvols should try
2160 * rmdir(2).
2161 */
2162 err = -EPERM;
2163 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2164 goto out_dput;
2165
2166 /*
2167 * Do not allow deletion if the parent dir is the same
2168 * as the dir to be deleted. That means the ioctl
2169 * must be called on the dentry referencing the root
2170 * of the subvol, not a random directory contained
2171 * within it.
2172 */
2173 err = -EINVAL;
2174 if (root == dest)
2175 goto out_dput;
2176
2177 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2178 if (err)
2179 goto out_dput;
4260f7c7
SW		 2180 }
2181
5c39da5b
MX		 2182 /* check if subvolume may be deleted by a user */
2183 err = btrfs_may_delete(dir, dentry, 1);
2184 if (err)
2185 goto out_dput;
2186
33345d01 2187 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
76dda93c
YZ		 2188 err = -EINVAL;
2189 goto out_dput;
2190 }
2191
76dda93c
YZ		 2192 mutex_lock(&inode->i_mutex);
2193 err = d_invalidate(dentry);
2194 if (err)
2195 goto out_unlock;
2196
2197 down_write(&root->fs_info->subvol_sem);
2198
2199 err = may_destroy_subvol(dest);
2200 if (err)
2201 goto out_up_write;
2202
c58aaad2
MX		 2203 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2204 /*
2205 * One for dir inode, two for dir entries, two for root
2206 * ref/backref.
2207 */
2208 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
ee3441b4 2209 5, &qgroup_reserved, true);
c58aaad2
MX		 2210 if (err)
2211 goto out_up_write;
2212
a22285a6
YZ		 2213 trans = btrfs_start_transaction(root, 0);
2214 if (IS_ERR(trans)) {
2215 err = PTR_ERR(trans);
c58aaad2 2216 goto out_release;
a22285a6 2217 }
c58aaad2
MX		 2218 trans->block_rsv = &block_rsv;
2219 trans->bytes_reserved = block_rsv.size;
a22285a6 2220
76dda93c
YZ		 2221 ret = btrfs_unlink_subvol(trans, root, dir,
2222 dest->root_key.objectid,
2223 dentry->d_name.name,
2224 dentry->d_name.len);
79787eaa
JM		 2225 if (ret) {
2226 err = ret;
2227 btrfs_abort_transaction(trans, root, ret);
2228 goto out_end_trans;
2229 }
76dda93c
YZ		 2230
2231 btrfs_record_root_in_trans(trans, dest);
2232
2233 memset(&dest->root_item.drop_progress, 0,
2234 sizeof(dest->root_item.drop_progress));
2235 dest->root_item.drop_level = 0;
2236 btrfs_set_root_refs(&dest->root_item, 0);
2237
d68fc57b
YZ		 2238 if (!xchg(&dest->orphan_item_inserted, 1)) {
2239 ret = btrfs_insert_orphan_item(trans,
2240 root->fs_info->tree_root,
2241 dest->root_key.objectid);
79787eaa
JM		 2242 if (ret) {
2243 btrfs_abort_transaction(trans, root, ret);
2244 err = ret;
2245 goto out_end_trans;
2246 }
d68fc57b 2247 }
dd5f9615
SB		 2248
2249 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2250 dest->root_item.uuid, BTRFS_UUID_KEY_SUBVOL,
2251 dest->root_key.objectid);
2252 if (ret && ret != -ENOENT) {
2253 btrfs_abort_transaction(trans, root, ret);
2254 err = ret;
2255 goto out_end_trans;
2256 }
2257 if (!btrfs_is_empty_uuid(dest->root_item.received_uuid)) {
2258 ret = btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
2259 dest->root_item.received_uuid,
2260 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
2261 dest->root_key.objectid);
2262 if (ret && ret != -ENOENT) {
2263 btrfs_abort_transaction(trans, root, ret);
2264 err = ret;
2265 goto out_end_trans;
2266 }
2267 }
2268
79787eaa 2269out_end_trans:
c58aaad2
MX		 2270 trans->block_rsv = NULL;
2271 trans->bytes_reserved = 0;
531cb13f 2272 ret = btrfs_end_transaction(trans, root);
79787eaa
JM		 2273 if (ret && !err)
2274 err = ret;
76dda93c 2275 inode->i_flags |= S_DEAD;
c58aaad2
MX		 2276out_release:
2277 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
76dda93c
YZ		 2278out_up_write:
2279 up_write(&root->fs_info->subvol_sem);
2280out_unlock:
2281 mutex_unlock(&inode->i_mutex);
2282 if (!err) {
efefb143 2283 shrink_dcache_sb(root->fs_info->sb);
76dda93c
YZ		 2284 btrfs_invalidate_inodes(dest);
2285 d_delete(dentry);
fa6ac876
LB		 2286
2287 /* the last ref */
2288 if (dest->cache_inode) {
2289 iput(dest->cache_inode);
2290 dest->cache_inode = NULL;
2291 }
76dda93c
YZ		 2292 }
2293out_dput:
2294 dput(dentry);
2295out_unlock_dir:
2296 mutex_unlock(&dir->i_mutex);
2a79f17e 2297 mnt_drop_write_file(file);
76dda93c
YZ		 2298out:
2299 kfree(vol_args);
2300 return err;
2301}
2302
1e701a32 2303static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
f46b5a66 2304{
496ad9aa 2305 struct inode *inode = file_inode(file);
f46b5a66 2306 struct btrfs_root *root = BTRFS_I(inode)->root;
1e701a32 2307 struct btrfs_ioctl_defrag_range_args *range;
c146afad
YZ		 2308 int ret;
2309
25122d15
ID		 2310 ret = mnt_want_write_file(file);
2311 if (ret)
2312 return ret;
b83cc969 2313
25122d15
ID		 2314 if (btrfs_root_readonly(root)) {
2315 ret = -EROFS;
2316 goto out;
5ac00add 2317 }
f46b5a66
CH		 2318
2319 switch (inode->i_mode & S_IFMT) {
2320 case S_IFDIR:
e441d54d
CM		 2321 if (!capable(CAP_SYS_ADMIN)) {
2322 ret = -EPERM;
2323 goto out;
2324 }
de78b51a 2325 ret = btrfs_defrag_root(root);
8929ecfa
YZ		 2326 if (ret)
2327 goto out;
de78b51a 2328 ret = btrfs_defrag_root(root->fs_info->extent_root);
f46b5a66
CH		 2329 break;
2330 case S_IFREG:
e441d54d
CM		 2331 if (!(file->f_mode & FMODE_WRITE)) {
2332 ret = -EINVAL;
2333 goto out;
2334 }
1e701a32
CM		 2335
2336 range = kzalloc(sizeof(*range), GFP_KERNEL);
2337 if (!range) {
2338 ret = -ENOMEM;
2339 goto out;
2340 }
2341
2342 if (argp) {
2343 if (copy_from_user(range, argp,
2344 sizeof(*range))) {
2345 ret = -EFAULT;
2346 kfree(range);
683be16e 2347 goto out;
1e701a32
CM		 2348 }
2349 /* compression requires us to start the IO */
2350 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2351 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2352 range->extent_thresh = (u32)-1;
2353 }
2354 } else {
2355 /* the rest are all set to zero by kzalloc */
2356 range->len = (u64)-1;
2357 }
496ad9aa 2358 ret = btrfs_defrag_file(file_inode(file), file,
4cb5300b
CM		 2359 range, 0, 0);
2360 if (ret > 0)
2361 ret = 0;
1e701a32 2362 kfree(range);
f46b5a66 2363 break;
8929ecfa
YZ		 2364 default:
2365 ret = -EINVAL;
f46b5a66 2366 }
e441d54d 2367out:
25122d15 2368 mnt_drop_write_file(file);
e441d54d 2369 return ret;
f46b5a66
CH		 2370}
2371
b2950863 2372static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
f46b5a66
CH		 2373{
2374 struct btrfs_ioctl_vol_args *vol_args;
2375 int ret;
2376
e441d54d
CM		 2377 if (!capable(CAP_SYS_ADMIN))
2378 return -EPERM;
2379
5ac00add
SB		 2380 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2381 1)) {
2382 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2c0c9da0 2383 return -EINVAL;
c9e9f97b
ID		 2384 }
2385
5ac00add 2386 mutex_lock(&root->fs_info->volume_mutex);
dae7b665 2387 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 2388 if (IS_ERR(vol_args)) {
2389 ret = PTR_ERR(vol_args);
2390 goto out;
2391 }
f46b5a66 2392
5516e595 2393 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66
CH		 2394 ret = btrfs_init_new_device(root, vol_args->name);
2395
f46b5a66 2396 kfree(vol_args);
c9e9f97b
ID		 2397out:
2398 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 2399 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
f46b5a66
CH		 2400 return ret;
2401}
2402
da24927b 2403static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
f46b5a66 2404{
496ad9aa 2405 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
f46b5a66
CH		 2406 struct btrfs_ioctl_vol_args *vol_args;
2407 int ret;
2408
e441d54d
CM		 2409 if (!capable(CAP_SYS_ADMIN))
2410 return -EPERM;
2411
da24927b
MX		 2412 ret = mnt_want_write_file(file);
2413 if (ret)
2414 return ret;
c146afad 2415
dae7b665 2416 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 2417 if (IS_ERR(vol_args)) {
2418 ret = PTR_ERR(vol_args);
2419 goto out;
2420 }
f46b5a66 2421
5516e595 2422 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 2423
183860f6
AJ		 2424 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2425 1)) {
2426 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2427 goto out;
2428 }
2429
2430 mutex_lock(&root->fs_info->volume_mutex);
2431 ret = btrfs_rm_device(root, vol_args->name);
c9e9f97b 2432 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 2433 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
183860f6
AJ		 2434
2435out:
2436 kfree(vol_args);
4ac20c70 2437 mnt_drop_write_file(file);
f46b5a66
CH		 2438 return ret;
2439}
2440
475f6387
JS		 2441static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2442{
027ed2f0 2443 struct btrfs_ioctl_fs_info_args *fi_args;
475f6387
JS		 2444 struct btrfs_device *device;
2445 struct btrfs_device *next;
2446 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
027ed2f0 2447 int ret = 0;
475f6387
JS		 2448
2449 if (!capable(CAP_SYS_ADMIN))
2450 return -EPERM;
2451
027ed2f0
LZ		 2452 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2453 if (!fi_args)
2454 return -ENOMEM;
2455
2456 fi_args->num_devices = fs_devices->num_devices;
2457 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
475f6387
JS		 2458
2459 mutex_lock(&fs_devices->device_list_mutex);
2460 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
027ed2f0
LZ		 2461 if (device->devid > fi_args->max_id)
2462 fi_args->max_id = device->devid;
475f6387
JS		 2463 }
2464 mutex_unlock(&fs_devices->device_list_mutex);
2465
027ed2f0
LZ		 2466 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2467 ret = -EFAULT;
475f6387 2468
027ed2f0
LZ		 2469 kfree(fi_args);
2470 return ret;
475f6387
JS		 2471}
2472
2473static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2474{
2475 struct btrfs_ioctl_dev_info_args *di_args;
2476 struct btrfs_device *dev;
2477 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2478 int ret = 0;
2479 char *s_uuid = NULL;
475f6387
JS		 2480
2481 if (!capable(CAP_SYS_ADMIN))
2482 return -EPERM;
2483
2484 di_args = memdup_user(arg, sizeof(*di_args));
2485 if (IS_ERR(di_args))
2486 return PTR_ERR(di_args);
2487
dd5f9615 2488 if (!btrfs_is_empty_uuid(di_args->uuid))
475f6387
JS		 2489 s_uuid = di_args->uuid;
2490
2491 mutex_lock(&fs_devices->device_list_mutex);
aa1b8cd4 2492 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
475f6387
JS		 2493
2494 if (!dev) {
2495 ret = -ENODEV;
2496 goto out;
2497 }
2498
2499 di_args->devid = dev->devid;
2500 di_args->bytes_used = dev->bytes_used;
2501 di_args->total_bytes = dev->total_bytes;
2502 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
a27202fb 2503 if (dev->name) {
606686ee
JB		 2504 struct rcu_string *name;
2505
2506 rcu_read_lock();
2507 name = rcu_dereference(dev->name);
2508 strncpy(di_args->path, name->str, sizeof(di_args->path));
2509 rcu_read_unlock();
a27202fb
JM		 2510 di_args->path[sizeof(di_args->path) - 1] = 0;
2511 } else {
99ba55ad 2512 di_args->path[0] = '\0';
a27202fb 2513 }
475f6387
JS		 2514
2515out:
55793c0d 2516 mutex_unlock(&fs_devices->device_list_mutex);
475f6387
JS		 2517 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2518 ret = -EFAULT;
2519
2520 kfree(di_args);
2521 return ret;
2522}
2523
416161db
MF		 2524static struct page *extent_same_get_page(struct inode *inode, u64 off)
2525{
2526 struct page *page;
2527 pgoff_t index;
2528 struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
2529
2530 index = off >> PAGE_CACHE_SHIFT;
2531
2532 page = grab_cache_page(inode->i_mapping, index);
2533 if (!page)
2534 return NULL;
2535
2536 if (!PageUptodate(page)) {
2537 if (extent_read_full_page_nolock(tree, page, btrfs_get_extent,
2538 0))
2539 return NULL;
2540 lock_page(page);
2541 if (!PageUptodate(page)) {
2542 unlock_page(page);
2543 page_cache_release(page);
2544 return NULL;
2545 }
2546 }
2547 unlock_page(page);
2548
2549 return page;
2550}
2551
77fe20dc
MF		 2552static inline void lock_extent_range(struct inode *inode, u64 off, u64 len)
2553{
2554 /* do any pending delalloc/csum calc on src, one way or
2555 another, and lock file content */
2556 while (1) {
2557 struct btrfs_ordered_extent *ordered;
2558 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2559 ordered = btrfs_lookup_first_ordered_extent(inode,
2560 off + len - 1);
2561 if (!ordered &&
2562 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2563 off + len - 1, EXTENT_DELALLOC, 0, NULL))
2564 break;
2565 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2566 if (ordered)
2567 btrfs_put_ordered_extent(ordered);
2568 btrfs_wait_ordered_range(inode, off, len);
2569 }
2570}
2571
416161db
MF		 2572static void btrfs_double_unlock(struct inode *inode1, u64 loff1,
2573 struct inode *inode2, u64 loff2, u64 len)
2574{
2575 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2576 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2577
2578 mutex_unlock(&inode1->i_mutex);
2579 mutex_unlock(&inode2->i_mutex);
2580}
2581
2582static void btrfs_double_lock(struct inode *inode1, u64 loff1,
2583 struct inode *inode2, u64 loff2, u64 len)
2584{
2585 if (inode1 < inode2) {
2586 swap(inode1, inode2);
2587 swap(loff1, loff2);
2588 }
2589
2590 mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT);
2591 lock_extent_range(inode1, loff1, len);
2592 if (inode1 != inode2) {
2593 mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD);
2594 lock_extent_range(inode2, loff2, len);
2595 }
2596}
2597
2598static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
2599 u64 dst_loff, u64 len)
2600{
2601 int ret = 0;
2602 struct page *src_page, *dst_page;
2603 unsigned int cmp_len = PAGE_CACHE_SIZE;
2604 void *addr, *dst_addr;
2605
2606 while (len) {
2607 if (len < PAGE_CACHE_SIZE)
2608 cmp_len = len;
2609
2610 src_page = extent_same_get_page(src, loff);
2611 if (!src_page)
2612 return -EINVAL;
2613 dst_page = extent_same_get_page(dst, dst_loff);
2614 if (!dst_page) {
2615 page_cache_release(src_page);
2616 return -EINVAL;
2617 }
2618 addr = kmap_atomic(src_page);
2619 dst_addr = kmap_atomic(dst_page);
2620
2621 flush_dcache_page(src_page);
2622 flush_dcache_page(dst_page);
2623
2624 if (memcmp(addr, dst_addr, cmp_len))
2625 ret = BTRFS_SAME_DATA_DIFFERS;
2626
2627 kunmap_atomic(addr);
2628 kunmap_atomic(dst_addr);
2629 page_cache_release(src_page);
2630 page_cache_release(dst_page);
2631
2632 if (ret)
2633 break;
2634
2635 loff += cmp_len;
2636 dst_loff += cmp_len;
2637 len -= cmp_len;
2638 }
2639
2640 return ret;
2641}
2642
2643static int extent_same_check_offsets(struct inode *inode, u64 off, u64 len)
2644{
2645 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
2646
2647 if (off + len > inode->i_size || off + len < off)
2648 return -EINVAL;
2649 /* Check that we are block aligned - btrfs_clone() requires this */
2650 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
2651 return -EINVAL;
2652
2653 return 0;
2654}
2655
2656static int btrfs_extent_same(struct inode *src, u64 loff, u64 len,
2657 struct inode *dst, u64 dst_loff)
2658{
2659 int ret;
2660
2661 /*
2662 * btrfs_clone() can't handle extents in the same file
2663 * yet. Once that works, we can drop this check and replace it
2664 * with a check for the same inode, but overlapping extents.
2665 */
2666 if (src == dst)
2667 return -EINVAL;
2668
2669 btrfs_double_lock(src, loff, dst, dst_loff, len);
2670
2671 ret = extent_same_check_offsets(src, loff, len);
2672 if (ret)
2673 goto out_unlock;
2674
2675 ret = extent_same_check_offsets(dst, dst_loff, len);
2676 if (ret)
2677 goto out_unlock;
2678
2679 /* don't make the dst file partly checksummed */
2680 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2681 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
2682 ret = -EINVAL;
2683 goto out_unlock;
2684 }
2685
2686 ret = btrfs_cmp_data(src, loff, dst, dst_loff, len);
2687 if (ret == 0)
2688 ret = btrfs_clone(src, dst, loff, len, len, dst_loff);
2689
2690out_unlock:
2691 btrfs_double_unlock(src, loff, dst, dst_loff, len);
2692
2693 return ret;
2694}
2695
2696#define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024)
2697
2698static long btrfs_ioctl_file_extent_same(struct file *file,
2699 void __user *argp)
2700{
2701 struct btrfs_ioctl_same_args *args = argp;
2702 struct btrfs_ioctl_same_args same;
2703 struct btrfs_ioctl_same_extent_info info;
2704 struct inode *src = file->f_dentry->d_inode;
2705 struct file *dst_file = NULL;
2706 struct inode *dst;
2707 u64 off;
2708 u64 len;
2709 int i;
2710 int ret;
2711 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
2712 bool is_admin = capable(CAP_SYS_ADMIN);
2713
2714 if (!(file->f_mode & FMODE_READ))
2715 return -EINVAL;
2716
2717 ret = mnt_want_write_file(file);
2718 if (ret)
2719 return ret;
2720
2721 if (copy_from_user(&same,
2722 (struct btrfs_ioctl_same_args __user *)argp,
2723 sizeof(same))) {
2724 ret = -EFAULT;
2725 goto out;
2726 }
2727
2728 off = same.logical_offset;
2729 len = same.length;
2730
2731 /*
2732 * Limit the total length we will dedupe for each operation.
2733 * This is intended to bound the total time spent in this
2734 * ioctl to something sane.
2735 */
2736 if (len > BTRFS_MAX_DEDUPE_LEN)
2737 len = BTRFS_MAX_DEDUPE_LEN;
2738
2739 if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) {
2740 /*
2741 * Btrfs does not support blocksize < page_size. As a
2742 * result, btrfs_cmp_data() won't correctly handle
2743 * this situation without an update.
2744 */
2745 ret = -EINVAL;
2746 goto out;
2747 }
2748
2749 ret = -EISDIR;
2750 if (S_ISDIR(src->i_mode))
2751 goto out;
2752
2753 ret = -EACCES;
2754 if (!S_ISREG(src->i_mode))
2755 goto out;
2756
2757 ret = 0;
2758 for (i = 0; i < same.dest_count; i++) {
2759 if (copy_from_user(&info, &args->info[i], sizeof(info))) {
2760 ret = -EFAULT;
2761 goto out;
2762 }
2763
2764 info.bytes_deduped = 0;
2765
2766 dst_file = fget(info.fd);
2767 if (!dst_file) {
2768 info.status = -EBADF;
2769 goto next;
2770 }
2771
2772 if (!(is_admin || (dst_file->f_mode & FMODE_WRITE))) {
2773 info.status = -EINVAL;
2774 goto next;
2775 }
2776
2777 info.status = -EXDEV;
2778 if (file->f_path.mnt != dst_file->f_path.mnt)
2779 goto next;
2780
2781 dst = dst_file->f_dentry->d_inode;
2782 if (src->i_sb != dst->i_sb)
2783 goto next;
2784
2785 if (S_ISDIR(dst->i_mode)) {
2786 info.status = -EISDIR;
2787 goto next;
2788 }
2789
2790 if (!S_ISREG(dst->i_mode)) {
2791 info.status = -EACCES;
2792 goto next;
2793 }
2794
2795 info.status = btrfs_extent_same(src, off, len, dst,
2796 info.logical_offset);
2797 if (info.status == 0)
2798 info.bytes_deduped += len;
2799
2800next:
2801 if (dst_file)
2802 fput(dst_file);
2803
2804 if (__put_user_unaligned(info.status, &args->info[i].status) ||
2805 __put_user_unaligned(info.bytes_deduped,
2806 &args->info[i].bytes_deduped)) {
2807 ret = -EFAULT;
2808 goto out;
2809 }
2810 }
2811
2812out:
2813 mnt_drop_write_file(file);
2814 return ret;
2815}
2816
32b7c687
MF		 2817/**
2818 * btrfs_clone() - clone a range from inode file to another
2819 *
2820 * @src: Inode to clone from
2821 * @inode: Inode to clone to
2822 * @off: Offset within source to start clone from
2823 * @olen: Original length, passed by user, of range to clone
2824 * @olen_aligned: Block-aligned value of olen, extent_same uses
2825 * identical values here
2826 * @destoff: Offset within @inode to start clone
2827 */
2828static int btrfs_clone(struct inode *src, struct inode *inode,
2829 u64 off, u64 olen, u64 olen_aligned, u64 destoff)
f46b5a66 2830{
f46b5a66 2831 struct btrfs_root *root = BTRFS_I(inode)->root;
32b7c687 2832 struct btrfs_path *path = NULL;
f46b5a66 2833 struct extent_buffer *leaf;
32b7c687
MF		 2834 struct btrfs_trans_handle *trans;
2835 char *buf = NULL;
ae01a0ab 2836 struct btrfs_key key;
f46b5a66
CH		 2837 u32 nritems;
2838 int slot;
ae01a0ab 2839 int ret;
32b7c687 2840 u64 len = olen_aligned;
ae01a0ab
YZ		 2841
2842 ret = -ENOMEM;
2843 buf = vmalloc(btrfs_level_size(root, 0));
2844 if (!buf)
32b7c687 2845 return ret;
ae01a0ab
YZ		 2846
2847 path = btrfs_alloc_path();
2848 if (!path) {
2849 vfree(buf);
32b7c687 2850 return ret;
d525e8ab
LZ		 2851 }
2852
32b7c687 2853 path->reada = 2;
c5c9cd4d 2854 /* clone data */
33345d01 2855 key.objectid = btrfs_ino(src);
ae01a0ab
YZ		 2856 key.type = BTRFS_EXTENT_DATA_KEY;
2857 key.offset = 0;
f46b5a66
CH		 2858
2859 while (1) {
2860 /*
2861 * note the key will change type as we walk through the
2862 * tree.
2863 */
362a20c5
DS		 2864 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
2865 0, 0);
f46b5a66
CH		 2866 if (ret < 0)
2867 goto out;
2868
ae01a0ab
YZ		 2869 nritems = btrfs_header_nritems(path->nodes[0]);
2870 if (path->slots[0] >= nritems) {
362a20c5 2871 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
f46b5a66
CH		 2872 if (ret < 0)
2873 goto out;
2874 if (ret > 0)
2875 break;
ae01a0ab 2876 nritems = btrfs_header_nritems(path->nodes[0]);
f46b5a66
CH		 2877 }
2878 leaf = path->nodes[0];
2879 slot = path->slots[0];
f46b5a66 2880
ae01a0ab 2881 btrfs_item_key_to_cpu(leaf, &key, slot);
d20f7043 2882 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
33345d01 2883 key.objectid != btrfs_ino(src))
f46b5a66
CH		 2884 break;
2885
c5c9cd4d
SW		 2886 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2887 struct btrfs_file_extent_item *extent;
2888 int type;
31840ae1
ZY		 2889 u32 size;
2890 struct btrfs_key new_key;
c5c9cd4d
SW		 2891 u64 disko = 0, diskl = 0;
2892 u64 datao = 0, datal = 0;
2893 u8 comp;
b5384d48 2894 u64 endoff;
31840ae1
ZY		 2895
2896 size = btrfs_item_size_nr(leaf, slot);
2897 read_extent_buffer(leaf, buf,
2898 btrfs_item_ptr_offset(leaf, slot),
2899 size);
c5c9cd4d
SW		 2900
2901 extent = btrfs_item_ptr(leaf, slot,
2902 struct btrfs_file_extent_item);
2903 comp = btrfs_file_extent_compression(leaf, extent);
2904 type = btrfs_file_extent_type(leaf, extent);
c8a894d7
CM		 2905 if (type == BTRFS_FILE_EXTENT_REG ||
2906 type == BTRFS_FILE_EXTENT_PREALLOC) {
d397712b
CM		 2907 disko = btrfs_file_extent_disk_bytenr(leaf,
2908 extent);
2909 diskl = btrfs_file_extent_disk_num_bytes(leaf,
2910 extent);
c5c9cd4d 2911 datao = btrfs_file_extent_offset(leaf, extent);
d397712b
CM		 2912 datal = btrfs_file_extent_num_bytes(leaf,
2913 extent);
c5c9cd4d
SW		 2914 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2915 /* take upper bound, may be compressed */
2916 datal = btrfs_file_extent_ram_bytes(leaf,
2917 extent);
2918 }
b3b4aa74 2919 btrfs_release_path(path);
31840ae1 2920
050006a7 2921 if (key.offset + datal <= off ||
aa42ffd9 2922 key.offset >= off + len - 1)
c5c9cd4d
SW		 2923 goto next;
2924
31840ae1 2925 memcpy(&new_key, &key, sizeof(new_key));
33345d01 2926 new_key.objectid = btrfs_ino(inode);
4d728ec7
LZ		 2927 if (off <= key.offset)
2928 new_key.offset = key.offset + destoff - off;
2929 else
2930 new_key.offset = destoff;
31840ae1 2931
b6f3409b
SW		 2932 /*
2933 * 1 - adjusting old extent (we may have to split it)
2934 * 1 - add new extent
2935 * 1 - inode update
2936 */
2937 trans = btrfs_start_transaction(root, 3);
a22285a6
YZ		 2938 if (IS_ERR(trans)) {
2939 ret = PTR_ERR(trans);
2940 goto out;
2941 }
2942
c8a894d7
CM		 2943 if (type == BTRFS_FILE_EXTENT_REG ||
2944 type == BTRFS_FILE_EXTENT_PREALLOC) {
d72c0842
LZ		 2945 /*
2946 * a | --- range to clone ---| b
2947 * | ------------- extent ------------- |
2948 */
2949
2950 /* substract range b */
2951 if (key.offset + datal > off + len)
2952 datal = off + len - key.offset;
2953
2954 /* substract range a */
a22285a6
YZ		 2955 if (off > key.offset) {
2956 datao += off - key.offset;
2957 datal -= off - key.offset;
2958 }
2959
5dc562c5 2960 ret = btrfs_drop_extents(trans, root, inode,
a22285a6
YZ		 2961 new_key.offset,
2962 new_key.offset + datal,
2671485d 2963 1);
79787eaa
JM		 2964 if (ret) {
2965 btrfs_abort_transaction(trans, root,
2966 ret);
2967 btrfs_end_transaction(trans, root);
2968 goto out;
2969 }
a22285a6 2970
c5c9cd4d
SW		 2971 ret = btrfs_insert_empty_item(trans, root, path,
2972 &new_key, size);
79787eaa
JM		 2973 if (ret) {
2974 btrfs_abort_transaction(trans, root,
2975 ret);
2976 btrfs_end_transaction(trans, root);
2977 goto out;
2978 }
c5c9cd4d
SW		 2979
2980 leaf = path->nodes[0];
2981 slot = path->slots[0];
2982 write_extent_buffer(leaf, buf,
31840ae1
ZY		 2983 btrfs_item_ptr_offset(leaf, slot),
2984 size);
ae01a0ab 2985
c5c9cd4d 2986 extent = btrfs_item_ptr(leaf, slot,
f46b5a66 2987 struct btrfs_file_extent_item);
c5c9cd4d 2988
c5c9cd4d
SW		 2989 /* disko == 0 means it's a hole */
2990 if (!disko)
2991 datao = 0;
c5c9cd4d
SW		 2992
2993 btrfs_set_file_extent_offset(leaf, extent,
2994 datao);
2995 btrfs_set_file_extent_num_bytes(leaf, extent,
2996 datal);
2997 if (disko) {
2998 inode_add_bytes(inode, datal);
ae01a0ab 2999 ret = btrfs_inc_extent_ref(trans, root,
5d4f98a2
YZ		 3000 disko, diskl, 0,
3001 root->root_key.objectid,
33345d01 3002 btrfs_ino(inode),
66d7e7f0
AJ		 3003 new_key.offset - datao,
3004 0);
79787eaa
JM		 3005 if (ret) {
3006 btrfs_abort_transaction(trans,
3007 root,
3008 ret);
3009 btrfs_end_transaction(trans,
3010 root);
3011 goto out;
3012
3013 }
f46b5a66 3014 }
c5c9cd4d
SW		 3015 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
3016 u64 skip = 0;
3017 u64 trim = 0;
3018 if (off > key.offset) {
3019 skip = off - key.offset;
3020 new_key.offset += skip;
3021 }
d397712b 3022
aa42ffd9
LB		 3023 if (key.offset + datal > off + len)
3024 trim = key.offset + datal - (off + len);
d397712b 3025
c5c9cd4d 3026 if (comp && (skip || trim)) {
c5c9cd4d 3027 ret = -EINVAL;
a22285a6 3028 btrfs_end_transaction(trans, root);
c5c9cd4d
SW		 3029 goto out;
3030 }
3031 size -= skip + trim;
3032 datal -= skip + trim;
a22285a6 3033
5dc562c5 3034 ret = btrfs_drop_extents(trans, root, inode,
a22285a6
YZ		 3035 new_key.offset,
3036 new_key.offset + datal,
2671485d 3037 1);
79787eaa
JM		 3038 if (ret) {
3039 btrfs_abort_transaction(trans, root,
3040 ret);
3041 btrfs_end_transaction(trans, root);
3042 goto out;
3043 }
a22285a6 3044
c5c9cd4d
SW		 3045 ret = btrfs_insert_empty_item(trans, root, path,
3046 &new_key, size);
79787eaa
JM		 3047 if (ret) {
3048 btrfs_abort_transaction(trans, root,
3049 ret);
3050 btrfs_end_transaction(trans, root);
3051 goto out;
3052 }
c5c9cd4d
SW		 3053
3054 if (skip) {
d397712b
CM		 3055 u32 start =
3056 btrfs_file_extent_calc_inline_size(0);
c5c9cd4d
SW		 3057 memmove(buf+start, buf+start+skip,
3058 datal);
3059 }
3060
3061 leaf = path->nodes[0];
3062 slot = path->slots[0];
3063 write_extent_buffer(leaf, buf,
3064 btrfs_item_ptr_offset(leaf, slot),
3065 size);
3066 inode_add_bytes(inode, datal);
f46b5a66 3067 }
c5c9cd4d
SW		 3068
3069 btrfs_mark_buffer_dirty(leaf);
b3b4aa74 3070 btrfs_release_path(path);
c5c9cd4d 3071
0c4d2d95 3072 inode_inc_iversion(inode);
a22285a6 3073 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
b5384d48
SW		 3074
3075 /*
3076 * we round up to the block size at eof when
3077 * determining which extents to clone above,
3078 * but shouldn't round up the file size
3079 */
3080 endoff = new_key.offset + datal;
5f3888ff
LZ		 3081 if (endoff > destoff+olen)
3082 endoff = destoff+olen;
b5384d48
SW		 3083 if (endoff > inode->i_size)
3084 btrfs_i_size_write(inode, endoff);
3085
a22285a6 3086 ret = btrfs_update_inode(trans, root, inode);
79787eaa
JM		 3087 if (ret) {
3088 btrfs_abort_transaction(trans, root, ret);
3089 btrfs_end_transaction(trans, root);
3090 goto out;
3091 }
3092 ret = btrfs_end_transaction(trans, root);
a22285a6 3093 }
d397712b 3094next:
b3b4aa74 3095 btrfs_release_path(path);
f46b5a66 3096 key.offset++;
f46b5a66 3097 }
f46b5a66 3098 ret = 0;
32b7c687 3099
f46b5a66 3100out:
b3b4aa74 3101 btrfs_release_path(path);
32b7c687
MF		 3102 btrfs_free_path(path);
3103 vfree(buf);
3104 return ret;
3105}
3106
3107static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
3108 u64 off, u64 olen, u64 destoff)
3109{
3110 struct inode *inode = fdentry(file)->d_inode;
3111 struct btrfs_root *root = BTRFS_I(inode)->root;
3112 struct fd src_file;
3113 struct inode *src;
3114 int ret;
3115 u64 len = olen;
3116 u64 bs = root->fs_info->sb->s_blocksize;
3117 int same_inode = 0;
3118
3119 /*
3120 * TODO:
3121 * - split compressed inline extents. annoying: we need to
3122 * decompress into destination's address_space (the file offset
3123 * may change, so source mapping won't do), then recompress (or
3124 * otherwise reinsert) a subrange.
3125 * - allow ranges within the same file to be cloned (provided
3126 * they don't overlap)?
3127 */
3128
3129 /* the destination must be opened for writing */
3130 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
3131 return -EINVAL;
3132
3133 if (btrfs_root_readonly(root))
3134 return -EROFS;
3135
3136 ret = mnt_want_write_file(file);
3137 if (ret)
3138 return ret;
3139
3140 src_file = fdget(srcfd);
3141 if (!src_file.file) {
3142 ret = -EBADF;
3143 goto out_drop_write;
3144 }
3145
3146 ret = -EXDEV;
3147 if (src_file.file->f_path.mnt != file->f_path.mnt)
3148 goto out_fput;
3149
3150 src = file_inode(src_file.file);
3151
3152 ret = -EINVAL;
3153 if (src == inode)
3154 same_inode = 1;
3155
3156 /* the src must be open for reading */
3157 if (!(src_file.file->f_mode & FMODE_READ))
3158 goto out_fput;
3159
3160 /* don't make the dst file partly checksummed */
3161 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3162 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3163 goto out_fput;
3164
3165 ret = -EISDIR;
3166 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3167 goto out_fput;
3168
3169 ret = -EXDEV;
3170 if (src->i_sb != inode->i_sb)
3171 goto out_fput;
3172
3173 if (!same_inode) {
3174 if (inode < src) {
3175 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
3176 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
3177 } else {
3178 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
3179 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
3180 }
3181 } else {
3182 mutex_lock(&src->i_mutex);
3183 }
3184
3185 /* determine range to clone */
3186 ret = -EINVAL;
3187 if (off + len > src->i_size || off + len < off)
3188 goto out_unlock;
3189 if (len == 0)
3190 olen = len = src->i_size - off;
3191 /* if we extend to eof, continue to block boundary */
3192 if (off + len == src->i_size)
3193 len = ALIGN(src->i_size, bs) - off;
3194
3195 /* verify the end result is block aligned */
3196 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3197 !IS_ALIGNED(destoff, bs))
3198 goto out_unlock;
3199
3200 /* verify if ranges are overlapped within the same file */
3201 if (same_inode) {
3202 if (destoff + len > off && destoff < off + len)
3203 goto out_unlock;
3204 }
3205
3206 if (destoff > inode->i_size) {
3207 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3208 if (ret)
3209 goto out_unlock;
3210 }
3211
3212 /* truncate page cache pages from target inode range */
3213 truncate_inode_pages_range(&inode->i_data, destoff,
3214 PAGE_CACHE_ALIGN(destoff + len) - 1);
3215
3216 lock_extent_range(src, off, len);
3217
3218 ret = btrfs_clone(src, inode, off, olen, len, destoff);
3219
aa42ffd9 3220 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
f46b5a66
CH		 3221out_unlock:
3222 mutex_unlock(&src->i_mutex);
a96fbc72
LB		 3223 if (!same_inode)
3224 mutex_unlock(&inode->i_mutex);
f46b5a66 3225out_fput:
2903ff01 3226 fdput(src_file);
ab67b7c1 3227out_drop_write:
2a79f17e 3228 mnt_drop_write_file(file);
f46b5a66
CH		 3229 return ret;
3230}
3231
7a865e8a 3232static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
c5c9cd4d
SW		 3233{
3234 struct btrfs_ioctl_clone_range_args args;
3235
7a865e8a 3236 if (copy_from_user(&args, argp, sizeof(args)))
c5c9cd4d
SW		 3237 return -EFAULT;
3238 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
3239 args.src_length, args.dest_offset);
3240}
3241
f46b5a66
CH		 3242/*
3243 * there are many ways the trans_start and trans_end ioctls can lead
3244 * to deadlocks. They should only be used by applications that
3245 * basically own the machine, and have a very in depth understanding
3246 * of all the possible deadlocks and enospc problems.
3247 */
b2950863 3248static long btrfs_ioctl_trans_start(struct file *file)
f46b5a66 3249{
496ad9aa 3250 struct inode *inode = file_inode(file);
f46b5a66
CH		 3251 struct btrfs_root *root = BTRFS_I(inode)->root;
3252 struct btrfs_trans_handle *trans;
1ab86aed 3253 int ret;
f46b5a66 3254
1ab86aed 3255 ret = -EPERM;
df5b5520 3256 if (!capable(CAP_SYS_ADMIN))
1ab86aed 3257 goto out;
df5b5520 3258
1ab86aed
SW		 3259 ret = -EINPROGRESS;
3260 if (file->private_data)
f46b5a66 3261 goto out;
9ca9ee09 3262
b83cc969
LZ		 3263 ret = -EROFS;
3264 if (btrfs_root_readonly(root))
3265 goto out;
3266
a561be71 3267 ret = mnt_want_write_file(file);
c146afad
YZ		 3268 if (ret)
3269 goto out;
3270
a4abeea4 3271 atomic_inc(&root->fs_info->open_ioctl_trans);
9ca9ee09 3272
1ab86aed 3273 ret = -ENOMEM;
7a7eaa40 3274 trans = btrfs_start_ioctl_transaction(root);
abd30bb0 3275 if (IS_ERR(trans))
1ab86aed
SW		 3276 goto out_drop;
3277
3278 file->private_data = trans;
3279 return 0;
3280
3281out_drop:
a4abeea4 3282 atomic_dec(&root->fs_info->open_ioctl_trans);
2a79f17e 3283 mnt_drop_write_file(file);
f46b5a66 3284out:
f46b5a66
CH		 3285 return ret;
3286}
3287
6ef5ed0d
JB		 3288static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3289{
496ad9aa 3290 struct inode *inode = file_inode(file);
6ef5ed0d
JB		 3291 struct btrfs_root *root = BTRFS_I(inode)->root;
3292 struct btrfs_root *new_root;
3293 struct btrfs_dir_item *di;
3294 struct btrfs_trans_handle *trans;
3295 struct btrfs_path *path;
3296 struct btrfs_key location;
3297 struct btrfs_disk_key disk_key;
6ef5ed0d
JB		 3298 u64 objectid = 0;
3299 u64 dir_id;
3c04ce01 3300 int ret;
6ef5ed0d
JB		 3301
3302 if (!capable(CAP_SYS_ADMIN))
3303 return -EPERM;
3304
3c04ce01
MX		 3305 ret = mnt_want_write_file(file);
3306 if (ret)
3307 return ret;
3308
3309 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3310 ret = -EFAULT;
3311 goto out;
3312 }
6ef5ed0d
JB		 3313
3314 if (!objectid)
3315 objectid = root->root_key.objectid;
3316
3317 location.objectid = objectid;
3318 location.type = BTRFS_ROOT_ITEM_KEY;
3319 location.offset = (u64)-1;
3320
3321 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
3c04ce01
MX		 3322 if (IS_ERR(new_root)) {
3323 ret = PTR_ERR(new_root);
3324 goto out;
3325 }
6ef5ed0d 3326
6ef5ed0d 3327 path = btrfs_alloc_path();
3c04ce01
MX		 3328 if (!path) {
3329 ret = -ENOMEM;
3330 goto out;
3331 }
6ef5ed0d
JB		 3332 path->leave_spinning = 1;
3333
3334 trans = btrfs_start_transaction(root, 1);
98d5dc13 3335 if (IS_ERR(trans)) {
6ef5ed0d 3336 btrfs_free_path(path);
3c04ce01
MX		 3337 ret = PTR_ERR(trans);
3338 goto out;
6ef5ed0d
JB		 3339 }
3340
6c41761f 3341 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
6ef5ed0d
JB		 3342 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
3343 dir_id, "default", 7, 1);
cf1e99a4 3344 if (IS_ERR_OR_NULL(di)) {
6ef5ed0d
JB		 3345 btrfs_free_path(path);
3346 btrfs_end_transaction(trans, root);
3347 printk(KERN_ERR "Umm, you don't have the default dir item, "
3348 "this isn't going to work\n");
3c04ce01
MX		 3349 ret = -ENOENT;
3350 goto out;
6ef5ed0d
JB		 3351 }
3352
3353 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3354 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3355 btrfs_mark_buffer_dirty(path->nodes[0]);
3356 btrfs_free_path(path);
3357
2b0ce2c2 3358 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
6ef5ed0d 3359 btrfs_end_transaction(trans, root);
3c04ce01
MX		 3360out:
3361 mnt_drop_write_file(file);
3362 return ret;
6ef5ed0d
JB		 3363}
3364
5af3e8cc
SB		 3365void btrfs_get_block_group_info(struct list_head *groups_list,
3366 struct btrfs_ioctl_space_info *space)
bf5fc093
JB		 3367{
3368 struct btrfs_block_group_cache *block_group;
3369
3370 space->total_bytes = 0;
3371 space->used_bytes = 0;
3372 space->flags = 0;
3373 list_for_each_entry(block_group, groups_list, list) {
3374 space->flags = block_group->flags;
3375 space->total_bytes += block_group->key.offset;
3376 space->used_bytes +=
3377 btrfs_block_group_used(&block_group->item);
3378 }
3379}
3380
48a3b636 3381static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
1406e432
JB		 3382{
3383 struct btrfs_ioctl_space_args space_args;
3384 struct btrfs_ioctl_space_info space;
3385 struct btrfs_ioctl_space_info *dest;
7fde62bf 3386 struct btrfs_ioctl_space_info *dest_orig;
13f2696f 3387 struct btrfs_ioctl_space_info __user *user_dest;
1406e432 3388 struct btrfs_space_info *info;
bf5fc093
JB		 3389 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3390 BTRFS_BLOCK_GROUP_SYSTEM,
3391 BTRFS_BLOCK_GROUP_METADATA,
3392 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3393 int num_types = 4;
7fde62bf 3394 int alloc_size;
1406e432 3395 int ret = 0;
51788b1b 3396 u64 slot_count = 0;
bf5fc093 3397 int i, c;
1406e432
JB		 3398
3399 if (copy_from_user(&space_args,
3400 (struct btrfs_ioctl_space_args __user *)arg,
3401 sizeof(space_args)))
3402 return -EFAULT;
3403
bf5fc093
JB		 3404 for (i = 0; i < num_types; i++) {
3405 struct btrfs_space_info *tmp;
3406
3407 info = NULL;
3408 rcu_read_lock();
3409 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3410 list) {
3411 if (tmp->flags == types[i]) {
3412 info = tmp;
3413 break;
3414 }
3415 }
3416 rcu_read_unlock();
3417
3418 if (!info)
3419 continue;
3420
3421 down_read(&info->groups_sem);
3422 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3423 if (!list_empty(&info->block_groups[c]))
3424 slot_count++;
3425 }
3426 up_read(&info->groups_sem);
3427 }
7fde62bf
CM		 3428
3429 /* space_slots == 0 means they are asking for a count */
3430 if (space_args.space_slots == 0) {
3431 space_args.total_spaces = slot_count;
3432 goto out;
3433 }
bf5fc093 3434
51788b1b 3435 slot_count = min_t(u64, space_args.space_slots, slot_count);
bf5fc093 3436
7fde62bf 3437 alloc_size = sizeof(*dest) * slot_count;
bf5fc093 3438
7fde62bf
CM		 3439 /* we generally have at most 6 or so space infos, one for each raid
3440 * level. So, a whole page should be more than enough for everyone
3441 */
3442 if (alloc_size > PAGE_CACHE_SIZE)
3443 return -ENOMEM;
3444
1406e432 3445 space_args.total_spaces = 0;
7fde62bf
CM		 3446 dest = kmalloc(alloc_size, GFP_NOFS);
3447 if (!dest)
3448 return -ENOMEM;
3449 dest_orig = dest;
1406e432 3450
7fde62bf 3451 /* now we have a buffer to copy into */
bf5fc093
JB		 3452 for (i = 0; i < num_types; i++) {
3453 struct btrfs_space_info *tmp;
3454
51788b1b
DR		 3455 if (!slot_count)
3456 break;
3457
bf5fc093
JB		 3458 info = NULL;
3459 rcu_read_lock();
3460 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3461 list) {
3462 if (tmp->flags == types[i]) {
3463 info = tmp;
3464 break;
3465 }
3466 }
3467 rcu_read_unlock();
7fde62bf 3468
bf5fc093
JB		 3469 if (!info)
3470 continue;
3471 down_read(&info->groups_sem);
3472 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3473 if (!list_empty(&info->block_groups[c])) {
5af3e8cc
SB		 3474 btrfs_get_block_group_info(
3475 &info->block_groups[c], &space);
bf5fc093
JB		 3476 memcpy(dest, &space, sizeof(space));
3477 dest++;
3478 space_args.total_spaces++;
51788b1b 3479 slot_count--;
bf5fc093 3480 }
51788b1b
DR		 3481 if (!slot_count)
3482 break;
bf5fc093
JB		 3483 }
3484 up_read(&info->groups_sem);
1406e432 3485 }
1406e432 3486
2eec6c81 3487 user_dest = (struct btrfs_ioctl_space_info __user *)
7fde62bf
CM		 3488 (arg + sizeof(struct btrfs_ioctl_space_args));
3489
3490 if (copy_to_user(user_dest, dest_orig, alloc_size))
3491 ret = -EFAULT;
3492
3493 kfree(dest_orig);
3494out:
3495 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
1406e432
JB		 3496 ret = -EFAULT;
3497
3498 return ret;
3499}
3500
f46b5a66
CH		 3501/*
3502 * there are many ways the trans_start and trans_end ioctls can lead
3503 * to deadlocks. They should only be used by applications that
3504 * basically own the machine, and have a very in depth understanding
3505 * of all the possible deadlocks and enospc problems.
3506 */
3507long btrfs_ioctl_trans_end(struct file *file)
3508{
496ad9aa 3509 struct inode *inode = file_inode(file);
f46b5a66
CH		 3510 struct btrfs_root *root = BTRFS_I(inode)->root;
3511 struct btrfs_trans_handle *trans;
f46b5a66 3512
f46b5a66 3513 trans = file->private_data;
1ab86aed
SW		 3514 if (!trans)
3515 return -EINVAL;
b214107e 3516 file->private_data = NULL;
9ca9ee09 3517
1ab86aed
SW		 3518 btrfs_end_transaction(trans, root);
3519
a4abeea4 3520 atomic_dec(&root->fs_info->open_ioctl_trans);
9ca9ee09 3521
2a79f17e 3522 mnt_drop_write_file(file);
1ab86aed 3523 return 0;
f46b5a66
CH		 3524}
3525
9a8c28be
MX		 3526static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3527 void __user *argp)
46204592 3528{
46204592
SW		 3529 struct btrfs_trans_handle *trans;
3530 u64 transid;
db5b493a 3531 int ret;
46204592 3532
d4edf39b 3533 trans = btrfs_attach_transaction_barrier(root);
ff7c1d33
MX		 3534 if (IS_ERR(trans)) {
3535 if (PTR_ERR(trans) != -ENOENT)
3536 return PTR_ERR(trans);
3537
3538 /* No running transaction, don't bother */
3539 transid = root->fs_info->last_trans_committed;
3540 goto out;
3541 }
46204592 3542 transid = trans->transid;
db5b493a 3543 ret = btrfs_commit_transaction_async(trans, root, 0);
8b2b2d3c
TI		 3544 if (ret) {
3545 btrfs_end_transaction(trans, root);
db5b493a 3546 return ret;
8b2b2d3c 3547 }
ff7c1d33 3548out:
46204592
SW		 3549 if (argp)
3550 if (copy_to_user(argp, &transid, sizeof(transid)))
3551 return -EFAULT;
3552 return 0;
3553}
3554
9a8c28be
MX		 3555static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
3556 void __user *argp)
46204592 3557{
46204592
SW		 3558 u64 transid;
3559
3560 if (argp) {
3561 if (copy_from_user(&transid, argp, sizeof(transid)))
3562 return -EFAULT;
3563 } else {
3564 transid = 0; /* current trans */
3565 }
3566 return btrfs_wait_for_commit(root, transid);
3567}
3568
b8e95489 3569static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
475f6387 3570{
496ad9aa 3571 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
475f6387 3572 struct btrfs_ioctl_scrub_args *sa;
b8e95489 3573 int ret;
475f6387
JS		 3574
3575 if (!capable(CAP_SYS_ADMIN))
3576 return -EPERM;
3577
3578 sa = memdup_user(arg, sizeof(*sa));
3579 if (IS_ERR(sa))
3580 return PTR_ERR(sa);
3581
b8e95489
MX		 3582 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3583 ret = mnt_want_write_file(file);
3584 if (ret)
3585 goto out;
3586 }
3587
aa1b8cd4 3588 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
63a212ab
SB		 3589 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3590 0);
475f6387
JS		 3591
3592 if (copy_to_user(arg, sa, sizeof(*sa)))
3593 ret = -EFAULT;
3594
b8e95489
MX		 3595 if (!(sa->flags & BTRFS_SCRUB_READONLY))
3596 mnt_drop_write_file(file);
3597out:
475f6387
JS		 3598 kfree(sa);
3599 return ret;
3600}
3601
3602static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3603{
3604 if (!capable(CAP_SYS_ADMIN))
3605 return -EPERM;
3606
aa1b8cd4 3607 return btrfs_scrub_cancel(root->fs_info);
475f6387
JS		 3608}
3609
3610static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3611 void __user *arg)
3612{
3613 struct btrfs_ioctl_scrub_args *sa;
3614 int ret;
3615
3616 if (!capable(CAP_SYS_ADMIN))
3617 return -EPERM;
3618
3619 sa = memdup_user(arg, sizeof(*sa));
3620 if (IS_ERR(sa))
3621 return PTR_ERR(sa);
3622
3623 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3624
3625 if (copy_to_user(arg, sa, sizeof(*sa)))
3626 ret = -EFAULT;
3627
3628 kfree(sa);
3629 return ret;
3630}
3631
c11d2c23 3632static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
b27f7c0c 3633 void __user *arg)
c11d2c23
SB		 3634{
3635 struct btrfs_ioctl_get_dev_stats *sa;
3636 int ret;
3637
c11d2c23
SB		 3638 sa = memdup_user(arg, sizeof(*sa));
3639 if (IS_ERR(sa))
3640 return PTR_ERR(sa);
3641
b27f7c0c
DS		 3642 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3643 kfree(sa);
3644 return -EPERM;
3645 }
3646
3647 ret = btrfs_get_dev_stats(root, sa);
c11d2c23
SB		 3648
3649 if (copy_to_user(arg, sa, sizeof(*sa)))
3650 ret = -EFAULT;
3651
3652 kfree(sa);
3653 return ret;
3654}
3655
3f6bcfbd
SB		 3656static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
3657{
3658 struct btrfs_ioctl_dev_replace_args *p;
3659 int ret;
3660
3661 if (!capable(CAP_SYS_ADMIN))
3662 return -EPERM;
3663
3664 p = memdup_user(arg, sizeof(*p));
3665 if (IS_ERR(p))
3666 return PTR_ERR(p);
3667
3668 switch (p->cmd) {
3669 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
bbb651e4
SB		 3670 if (root->fs_info->sb->s_flags & MS_RDONLY)
3671 return -EROFS;
3672
3f6bcfbd
SB		 3673 if (atomic_xchg(
3674 &root->fs_info->mutually_exclusive_operation_running,
3675 1)) {
3676 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3677 ret = -EINPROGRESS;
3678 } else {
3679 ret = btrfs_dev_replace_start(root, p);
3680 atomic_set(
3681 &root->fs_info->mutually_exclusive_operation_running,
3682 0);
3683 }
3684 break;
3685 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3686 btrfs_dev_replace_status(root->fs_info, p);
3687 ret = 0;
3688 break;
3689 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3690 ret = btrfs_dev_replace_cancel(root->fs_info, p);
3691 break;
3692 default:
3693 ret = -EINVAL;
3694 break;
3695 }
3696
3697 if (copy_to_user(arg, p, sizeof(*p)))
3698 ret = -EFAULT;
3699
3700 kfree(p);
3701 return ret;
3702}
3703
d7728c96
JS		 3704static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3705{
3706 int ret = 0;
3707 int i;
740c3d22 3708 u64 rel_ptr;
d7728c96 3709 int size;
806468f8 3710 struct btrfs_ioctl_ino_path_args *ipa = NULL;
d7728c96
JS		 3711 struct inode_fs_paths *ipath = NULL;
3712 struct btrfs_path *path;
3713
82b22ac8 3714 if (!capable(CAP_DAC_READ_SEARCH))
d7728c96
JS		 3715 return -EPERM;
3716
3717 path = btrfs_alloc_path();
3718 if (!path) {
3719 ret = -ENOMEM;
3720 goto out;
3721 }
3722
3723 ipa = memdup_user(arg, sizeof(*ipa));
3724 if (IS_ERR(ipa)) {
3725 ret = PTR_ERR(ipa);
3726 ipa = NULL;
3727 goto out;
3728 }
3729
3730 size = min_t(u32, ipa->size, 4096);
3731 ipath = init_ipath(size, root, path);
3732 if (IS_ERR(ipath)) {
3733 ret = PTR_ERR(ipath);
3734 ipath = NULL;
3735 goto out;
3736 }
3737
3738 ret = paths_from_inode(ipa->inum, ipath);
3739 if (ret < 0)
3740 goto out;
3741
3742 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
745c4d8e
JM		 3743 rel_ptr = ipath->fspath->val[i] -
3744 (u64)(unsigned long)ipath->fspath->val;
740c3d22 3745 ipath->fspath->val[i] = rel_ptr;
d7728c96
JS		 3746 }
3747
745c4d8e
JM		 3748 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3749 (void *)(unsigned long)ipath->fspath, size);
d7728c96
JS		 3750 if (ret) {
3751 ret = -EFAULT;
3752 goto out;
3753 }
3754
3755out:
3756 btrfs_free_path(path);
3757 free_ipath(ipath);
3758 kfree(ipa);
3759
3760 return ret;
3761}
3762
3763static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3764{
3765 struct btrfs_data_container *inodes = ctx;
3766 const size_t c = 3 * sizeof(u64);
3767
3768 if (inodes->bytes_left >= c) {
3769 inodes->bytes_left -= c;
3770 inodes->val[inodes->elem_cnt] = inum;
3771 inodes->val[inodes->elem_cnt + 1] = offset;
3772 inodes->val[inodes->elem_cnt + 2] = root;
3773 inodes->elem_cnt += 3;
3774 } else {
3775 inodes->bytes_missing += c - inodes->bytes_left;
3776 inodes->bytes_left = 0;
3777 inodes->elem_missed += 3;
3778 }
3779
3780 return 0;
3781}
3782
3783static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3784 void __user *arg)
3785{
3786 int ret = 0;
3787 int size;
d7728c96
JS		 3788 struct btrfs_ioctl_logical_ino_args *loi;
3789 struct btrfs_data_container *inodes = NULL;
3790 struct btrfs_path *path = NULL;
d7728c96
JS		 3791
3792 if (!capable(CAP_SYS_ADMIN))
3793 return -EPERM;
3794
3795 loi = memdup_user(arg, sizeof(*loi));
3796 if (IS_ERR(loi)) {
3797 ret = PTR_ERR(loi);
3798 loi = NULL;
3799 goto out;
3800 }
3801
3802 path = btrfs_alloc_path();
3803 if (!path) {
3804 ret = -ENOMEM;
3805 goto out;
3806 }
3807
425d17a2 3808 size = min_t(u32, loi->size, 64 * 1024);
d7728c96
JS		 3809 inodes = init_data_container(size);
3810 if (IS_ERR(inodes)) {
3811 ret = PTR_ERR(inodes);
3812 inodes = NULL;
3813 goto out;
3814 }
3815
df031f07
LB		 3816 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
3817 build_ino_list, inodes);
3818 if (ret == -EINVAL)
d7728c96
JS		 3819 ret = -ENOENT;
3820 if (ret < 0)
3821 goto out;
3822
745c4d8e
JM		 3823 ret = copy_to_user((void *)(unsigned long)loi->inodes,
3824 (void *)(unsigned long)inodes, size);
d7728c96
JS		 3825 if (ret)
3826 ret = -EFAULT;
3827
3828out:
3829 btrfs_free_path(path);
425d17a2 3830 vfree(inodes);
d7728c96
JS		 3831 kfree(loi);
3832
3833 return ret;
3834}
3835
19a39dce 3836void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
c9e9f97b
ID		 3837 struct btrfs_ioctl_balance_args *bargs)
3838{
3839 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3840
3841 bargs->flags = bctl->flags;
3842
837d5b6e
ID		 3843 if (atomic_read(&fs_info->balance_running))
3844 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3845 if (atomic_read(&fs_info->balance_pause_req))
3846 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
a7e99c69
ID		 3847 if (atomic_read(&fs_info->balance_cancel_req))
3848 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
837d5b6e 3849
c9e9f97b
ID		 3850 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3851 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3852 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
19a39dce
ID		 3853
3854 if (lock) {
3855 spin_lock(&fs_info->balance_lock);
3856 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3857 spin_unlock(&fs_info->balance_lock);
3858 } else {
3859 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3860 }
c9e9f97b
ID		 3861}
3862
9ba1f6e4 3863static long btrfs_ioctl_balance(struct file *file, void __user *arg)
c9e9f97b 3864{
496ad9aa 3865 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
c9e9f97b
ID		 3866 struct btrfs_fs_info *fs_info = root->fs_info;
3867 struct btrfs_ioctl_balance_args *bargs;
3868 struct btrfs_balance_control *bctl;
ed0fb78f 3869 bool need_unlock; /* for mut. excl. ops lock */
c9e9f97b
ID		 3870 int ret;
3871
3872 if (!capable(CAP_SYS_ADMIN))
3873 return -EPERM;
3874
e54bfa31 3875 ret = mnt_want_write_file(file);
9ba1f6e4
LB		 3876 if (ret)
3877 return ret;
3878
ed0fb78f
ID		 3879again:
3880 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
3881 mutex_lock(&fs_info->volume_mutex);
3882 mutex_lock(&fs_info->balance_mutex);
3883 need_unlock = true;
3884 goto locked;
3885 }
3886
3887 /*
3888 * mut. excl. ops lock is locked. Three possibilites:
3889 * (1) some other op is running
3890 * (2) balance is running
3891 * (3) balance is paused -- special case (think resume)
3892 */
c9e9f97b 3893 mutex_lock(&fs_info->balance_mutex);
ed0fb78f
ID		 3894 if (fs_info->balance_ctl) {
3895 /* this is either (2) or (3) */
3896 if (!atomic_read(&fs_info->balance_running)) {
3897 mutex_unlock(&fs_info->balance_mutex);
3898 if (!mutex_trylock(&fs_info->volume_mutex))
3899 goto again;
3900 mutex_lock(&fs_info->balance_mutex);
3901
3902 if (fs_info->balance_ctl &&
3903 !atomic_read(&fs_info->balance_running)) {
3904 /* this is (3) */
3905 need_unlock = false;
3906 goto locked;
3907 }
3908
3909 mutex_unlock(&fs_info->balance_mutex);
3910 mutex_unlock(&fs_info->volume_mutex);
3911 goto again;
3912 } else {
3913 /* this is (2) */
3914 mutex_unlock(&fs_info->balance_mutex);
3915 ret = -EINPROGRESS;
3916 goto out;
3917 }
3918 } else {
3919 /* this is (1) */
3920 mutex_unlock(&fs_info->balance_mutex);
3921 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3922 ret = -EINVAL;
3923 goto out;
3924 }
3925
3926locked:
3927 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
c9e9f97b
ID		 3928
3929 if (arg) {
3930 bargs = memdup_user(arg, sizeof(*bargs));
3931 if (IS_ERR(bargs)) {
3932 ret = PTR_ERR(bargs);
ed0fb78f 3933 goto out_unlock;
c9e9f97b 3934 }
de322263
ID		 3935
3936 if (bargs->flags & BTRFS_BALANCE_RESUME) {
3937 if (!fs_info->balance_ctl) {
3938 ret = -ENOTCONN;
3939 goto out_bargs;
3940 }
3941
3942 bctl = fs_info->balance_ctl;
3943 spin_lock(&fs_info->balance_lock);
3944 bctl->flags |= BTRFS_BALANCE_RESUME;
3945 spin_unlock(&fs_info->balance_lock);
3946
3947 goto do_balance;
3948 }
c9e9f97b
ID		 3949 } else {
3950 bargs = NULL;
3951 }
3952
ed0fb78f 3953 if (fs_info->balance_ctl) {
837d5b6e
ID		 3954 ret = -EINPROGRESS;
3955 goto out_bargs;
3956 }
3957
c9e9f97b
ID		 3958 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3959 if (!bctl) {
3960 ret = -ENOMEM;
3961 goto out_bargs;
3962 }
3963
3964 bctl->fs_info = fs_info;
3965 if (arg) {
3966 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3967 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3968 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3969
3970 bctl->flags = bargs->flags;
f43ffb60
ID		 3971 } else {
3972 /* balance everything - no filters */
3973 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
c9e9f97b
ID		 3974 }
3975
de322263 3976do_balance:
c9e9f97b 3977 /*
ed0fb78f
ID		 3978 * Ownership of bctl and mutually_exclusive_operation_running
3979 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
3980 * or, if restriper was paused all the way until unmount, in
3981 * free_fs_info. mutually_exclusive_operation_running is
3982 * cleared in __cancel_balance.
c9e9f97b 3983 */
ed0fb78f
ID		 3984 need_unlock = false;
3985
3986 ret = btrfs_balance(bctl, bargs);
3987
c9e9f97b
ID		 3988 if (arg) {
3989 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3990 ret = -EFAULT;
3991 }
3992
3993out_bargs:
3994 kfree(bargs);
ed0fb78f 3995out_unlock:
c9e9f97b
ID		 3996 mutex_unlock(&fs_info->balance_mutex);
3997 mutex_unlock(&fs_info->volume_mutex);
ed0fb78f
ID		 3998 if (need_unlock)
3999 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
4000out:
e54bfa31 4001 mnt_drop_write_file(file);
c9e9f97b
ID		 4002 return ret;
4003}
4004
837d5b6e
ID		 4005static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
4006{
4007 if (!capable(CAP_SYS_ADMIN))
4008 return -EPERM;
4009
4010 switch (cmd) {
4011 case BTRFS_BALANCE_CTL_PAUSE:
4012 return btrfs_pause_balance(root->fs_info);
a7e99c69
ID		 4013 case BTRFS_BALANCE_CTL_CANCEL:
4014 return btrfs_cancel_balance(root->fs_info);
837d5b6e
ID		 4015 }
4016
4017 return -EINVAL;
4018}
4019
19a39dce
ID		 4020static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
4021 void __user *arg)
4022{
4023 struct btrfs_fs_info *fs_info = root->fs_info;
4024 struct btrfs_ioctl_balance_args *bargs;
4025 int ret = 0;
4026
4027 if (!capable(CAP_SYS_ADMIN))
4028 return -EPERM;
4029
4030 mutex_lock(&fs_info->balance_mutex);
4031 if (!fs_info->balance_ctl) {
4032 ret = -ENOTCONN;
4033 goto out;
4034 }
4035
4036 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
4037 if (!bargs) {
4038 ret = -ENOMEM;
4039 goto out;
4040 }
4041
4042 update_ioctl_balance_args(fs_info, 1, bargs);
4043
4044 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4045 ret = -EFAULT;
4046
4047 kfree(bargs);
4048out:
4049 mutex_unlock(&fs_info->balance_mutex);
4050 return ret;
4051}
4052
905b0dda 4053static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
5d13a37b 4054{
496ad9aa 4055 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4056 struct btrfs_ioctl_quota_ctl_args *sa;
4057 struct btrfs_trans_handle *trans = NULL;
4058 int ret;
4059 int err;
4060
4061 if (!capable(CAP_SYS_ADMIN))
4062 return -EPERM;
4063
905b0dda
MX		 4064 ret = mnt_want_write_file(file);
4065 if (ret)
4066 return ret;
5d13a37b
AJ		 4067
4068 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4069 if (IS_ERR(sa)) {
4070 ret = PTR_ERR(sa);
4071 goto drop_write;
4072 }
5d13a37b 4073
7708f029 4074 down_write(&root->fs_info->subvol_sem);
2f232036
JS		 4075 trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4076 if (IS_ERR(trans)) {
4077 ret = PTR_ERR(trans);
4078 goto out;
5d13a37b
AJ		 4079 }
4080
4081 switch (sa->cmd) {
4082 case BTRFS_QUOTA_CTL_ENABLE:
4083 ret = btrfs_quota_enable(trans, root->fs_info);
4084 break;
4085 case BTRFS_QUOTA_CTL_DISABLE:
4086 ret = btrfs_quota_disable(trans, root->fs_info);
4087 break;
5d13a37b
AJ		 4088 default:
4089 ret = -EINVAL;
4090 break;
4091 }
4092
2f232036
JS		 4093 err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4094 if (err && !ret)
4095 ret = err;
5d13a37b
AJ		 4096out:
4097 kfree(sa);
7708f029 4098 up_write(&root->fs_info->subvol_sem);
905b0dda
MX		 4099drop_write:
4100 mnt_drop_write_file(file);
5d13a37b
AJ		 4101 return ret;
4102}
4103
905b0dda 4104static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
5d13a37b 4105{
496ad9aa 4106 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4107 struct btrfs_ioctl_qgroup_assign_args *sa;
4108 struct btrfs_trans_handle *trans;
4109 int ret;
4110 int err;
4111
4112 if (!capable(CAP_SYS_ADMIN))
4113 return -EPERM;
4114
905b0dda
MX		 4115 ret = mnt_want_write_file(file);
4116 if (ret)
4117 return ret;
5d13a37b
AJ		 4118
4119 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4120 if (IS_ERR(sa)) {
4121 ret = PTR_ERR(sa);
4122 goto drop_write;
4123 }
5d13a37b
AJ		 4124
4125 trans = btrfs_join_transaction(root);
4126 if (IS_ERR(trans)) {
4127 ret = PTR_ERR(trans);
4128 goto out;
4129 }
4130
4131 /* FIXME: check if the IDs really exist */
4132 if (sa->assign) {
4133 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4134 sa->src, sa->dst);
4135 } else {
4136 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4137 sa->src, sa->dst);
4138 }
4139
4140 err = btrfs_end_transaction(trans, root);
4141 if (err && !ret)
4142 ret = err;
4143
4144out:
4145 kfree(sa);
905b0dda
MX		 4146drop_write:
4147 mnt_drop_write_file(file);
5d13a37b
AJ		 4148 return ret;
4149}
4150
905b0dda 4151static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
5d13a37b 4152{
496ad9aa 4153 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4154 struct btrfs_ioctl_qgroup_create_args *sa;
4155 struct btrfs_trans_handle *trans;
4156 int ret;
4157 int err;
4158
4159 if (!capable(CAP_SYS_ADMIN))
4160 return -EPERM;
4161
905b0dda
MX		 4162 ret = mnt_want_write_file(file);
4163 if (ret)
4164 return ret;
5d13a37b
AJ		 4165
4166 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4167 if (IS_ERR(sa)) {
4168 ret = PTR_ERR(sa);
4169 goto drop_write;
4170 }
5d13a37b 4171
d86e56cf
MX		 4172 if (!sa->qgroupid) {
4173 ret = -EINVAL;
4174 goto out;
4175 }
4176
5d13a37b
AJ		 4177 trans = btrfs_join_transaction(root);
4178 if (IS_ERR(trans)) {
4179 ret = PTR_ERR(trans);
4180 goto out;
4181 }
4182
4183 /* FIXME: check if the IDs really exist */
4184 if (sa->create) {
4185 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
4186 NULL);
4187 } else {
4188 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4189 }
4190
4191 err = btrfs_end_transaction(trans, root);
4192 if (err && !ret)
4193 ret = err;
4194
4195out:
4196 kfree(sa);
905b0dda
MX		 4197drop_write:
4198 mnt_drop_write_file(file);
5d13a37b
AJ		 4199 return ret;
4200}
4201
905b0dda 4202static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
5d13a37b 4203{
496ad9aa 4204 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4205 struct btrfs_ioctl_qgroup_limit_args *sa;
4206 struct btrfs_trans_handle *trans;
4207 int ret;
4208 int err;
4209 u64 qgroupid;
4210
4211 if (!capable(CAP_SYS_ADMIN))
4212 return -EPERM;
4213
905b0dda
MX		 4214 ret = mnt_want_write_file(file);
4215 if (ret)
4216 return ret;
5d13a37b
AJ		 4217
4218 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4219 if (IS_ERR(sa)) {
4220 ret = PTR_ERR(sa);
4221 goto drop_write;
4222 }
5d13a37b
AJ		 4223
4224 trans = btrfs_join_transaction(root);
4225 if (IS_ERR(trans)) {
4226 ret = PTR_ERR(trans);
4227 goto out;
4228 }
4229
4230 qgroupid = sa->qgroupid;
4231 if (!qgroupid) {
4232 /* take the current subvol as qgroup */
4233 qgroupid = root->root_key.objectid;
4234 }
4235
4236 /* FIXME: check if the IDs really exist */
4237 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
4238
4239 err = btrfs_end_transaction(trans, root);
4240 if (err && !ret)
4241 ret = err;
4242
4243out:
4244 kfree(sa);
905b0dda
MX		 4245drop_write:
4246 mnt_drop_write_file(file);
5d13a37b
AJ		 4247 return ret;
4248}
4249
2f232036
JS		 4250static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4251{
6d0379ec 4252 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2f232036
JS		 4253 struct btrfs_ioctl_quota_rescan_args *qsa;
4254 int ret;
4255
4256 if (!capable(CAP_SYS_ADMIN))
4257 return -EPERM;
4258
4259 ret = mnt_want_write_file(file);
4260 if (ret)
4261 return ret;
4262
4263 qsa = memdup_user(arg, sizeof(*qsa));
4264 if (IS_ERR(qsa)) {
4265 ret = PTR_ERR(qsa);
4266 goto drop_write;
4267 }
4268
4269 if (qsa->flags) {
4270 ret = -EINVAL;
4271 goto out;
4272 }
4273
4274 ret = btrfs_qgroup_rescan(root->fs_info);
4275
4276out:
4277 kfree(qsa);
4278drop_write:
4279 mnt_drop_write_file(file);
4280 return ret;
4281}
4282
4283static long btrfs_ioctl_quota_rescan_status(struct file *file, void __user *arg)
4284{
6d0379ec 4285 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2f232036
JS		 4286 struct btrfs_ioctl_quota_rescan_args *qsa;
4287 int ret = 0;
4288
4289 if (!capable(CAP_SYS_ADMIN))
4290 return -EPERM;
4291
4292 qsa = kzalloc(sizeof(*qsa), GFP_NOFS);
4293 if (!qsa)
4294 return -ENOMEM;
4295
4296 if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4297 qsa->flags = 1;
4298 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
4299 }
4300
4301 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4302 ret = -EFAULT;
4303
4304 kfree(qsa);
4305 return ret;
4306}
4307
57254b6e
JS		 4308static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
4309{
4310 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
4311
4312 if (!capable(CAP_SYS_ADMIN))
4313 return -EPERM;
4314
4315 return btrfs_qgroup_wait_for_completion(root->fs_info);
4316}
4317
8ea05e3a
AB		 4318static long btrfs_ioctl_set_received_subvol(struct file *file,
4319 void __user *arg)
4320{
4321 struct btrfs_ioctl_received_subvol_args *sa = NULL;
496ad9aa 4322 struct inode *inode = file_inode(file);
8ea05e3a
AB		 4323 struct btrfs_root *root = BTRFS_I(inode)->root;
4324 struct btrfs_root_item *root_item = &root->root_item;
4325 struct btrfs_trans_handle *trans;
4326 struct timespec ct = CURRENT_TIME;
4327 int ret = 0;
dd5f9615 4328 int received_uuid_changed;
8ea05e3a
AB		 4329
4330 ret = mnt_want_write_file(file);
4331 if (ret < 0)
4332 return ret;
4333
4334 down_write(&root->fs_info->subvol_sem);
4335
4336 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
4337 ret = -EINVAL;
4338 goto out;
4339 }
4340
4341 if (btrfs_root_readonly(root)) {
4342 ret = -EROFS;
4343 goto out;
4344 }
4345
4346 if (!inode_owner_or_capable(inode)) {
4347 ret = -EACCES;
4348 goto out;
4349 }
4350
4351 sa = memdup_user(arg, sizeof(*sa));
4352 if (IS_ERR(sa)) {
4353 ret = PTR_ERR(sa);
4354 sa = NULL;
4355 goto out;
4356 }
4357
dd5f9615
SB		 4358 /*
4359 * 1 - root item
4360 * 2 - uuid items (received uuid + subvol uuid)
4361 */
4362 trans = btrfs_start_transaction(root, 3);
8ea05e3a
AB		 4363 if (IS_ERR(trans)) {
4364 ret = PTR_ERR(trans);
4365 trans = NULL;
4366 goto out;
4367 }
4368
4369 sa->rtransid = trans->transid;
4370 sa->rtime.sec = ct.tv_sec;
4371 sa->rtime.nsec = ct.tv_nsec;
4372
dd5f9615
SB		 4373 received_uuid_changed = memcmp(root_item->received_uuid, sa->uuid,
4374 BTRFS_UUID_SIZE);
4375 if (received_uuid_changed &&
4376 !btrfs_is_empty_uuid(root_item->received_uuid))
4377 btrfs_uuid_tree_rem(trans, root->fs_info->uuid_root,
4378 root_item->received_uuid,
4379 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4380 root->root_key.objectid);
8ea05e3a
AB		 4381 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4382 btrfs_set_root_stransid(root_item, sa->stransid);
4383 btrfs_set_root_rtransid(root_item, sa->rtransid);
3cae210f
QW		 4384 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4385 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4386 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4387 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
8ea05e3a
AB		 4388
4389 ret = btrfs_update_root(trans, root->fs_info->tree_root,
4390 &root->root_key, &root->root_item);
4391 if (ret < 0) {
4392 btrfs_end_transaction(trans, root);
8ea05e3a 4393 goto out;
dd5f9615
SB		 4394 }
4395 if (received_uuid_changed && !btrfs_is_empty_uuid(sa->uuid)) {
4396 ret = btrfs_uuid_tree_add(trans, root->fs_info->uuid_root,
4397 sa->uuid,
4398 BTRFS_UUID_KEY_RECEIVED_SUBVOL,
4399 root->root_key.objectid);
4400 if (ret < 0 && ret != -EEXIST) {
4401 btrfs_abort_transaction(trans, root, ret);
8ea05e3a 4402 goto out;
dd5f9615
SB		 4403 }
4404 }
4405 ret = btrfs_commit_transaction(trans, root);
4406 if (ret < 0) {
4407 btrfs_abort_transaction(trans, root, ret);
4408 goto out;
8ea05e3a
AB		 4409 }
4410
4411 ret = copy_to_user(arg, sa, sizeof(*sa));
4412 if (ret)
4413 ret = -EFAULT;
4414
4415out:
4416 kfree(sa);
4417 up_write(&root->fs_info->subvol_sem);
4418 mnt_drop_write_file(file);
4419 return ret;
4420}
4421
867ab667 4422static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
4423{
6d0379ec 4424 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
a1b83ac5 4425 size_t len;
867ab667 4426 int ret;
a1b83ac5
AJ		 4427 char label[BTRFS_LABEL_SIZE];
4428
4429 spin_lock(&root->fs_info->super_lock);
4430 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4431 spin_unlock(&root->fs_info->super_lock);
4432
4433 len = strnlen(label, BTRFS_LABEL_SIZE);
867ab667 4434
4435 if (len == BTRFS_LABEL_SIZE) {
4436 pr_warn("btrfs: label is too long, return the first %zu bytes\n",
4437 --len);
4438 }
4439
867ab667 4440 ret = copy_to_user(arg, label, len);
867ab667 4441
4442 return ret ? -EFAULT : 0;
4443}
4444
a8bfd4ab 4445static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4446{
6d0379ec 4447 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
a8bfd4ab 4448 struct btrfs_super_block *super_block = root->fs_info->super_copy;
4449 struct btrfs_trans_handle *trans;
4450 char label[BTRFS_LABEL_SIZE];
4451 int ret;
4452
4453 if (!capable(CAP_SYS_ADMIN))
4454 return -EPERM;
4455
4456 if (copy_from_user(label, arg, sizeof(label)))
4457 return -EFAULT;
4458
4459 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
4460 pr_err("btrfs: unable to set label with more than %d bytes\n",
4461 BTRFS_LABEL_SIZE - 1);
4462 return -EINVAL;
4463 }
4464
4465 ret = mnt_want_write_file(file);
4466 if (ret)
4467 return ret;
4468
a8bfd4ab 4469 trans = btrfs_start_transaction(root, 0);
4470 if (IS_ERR(trans)) {
4471 ret = PTR_ERR(trans);
4472 goto out_unlock;
4473 }
4474
a1b83ac5 4475 spin_lock(&root->fs_info->super_lock);
a8bfd4ab 4476 strcpy(super_block->label, label);
a1b83ac5 4477 spin_unlock(&root->fs_info->super_lock);
a8bfd4ab 4478 ret = btrfs_end_transaction(trans, root);
4479
4480out_unlock:
a8bfd4ab 4481 mnt_drop_write_file(file);
4482 return ret;
4483}
4484
f46b5a66
CH		 4485long btrfs_ioctl(struct file *file, unsigned int
4486 cmd, unsigned long arg)
4487{
496ad9aa 4488 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4bcabaa3 4489 void __user *argp = (void __user *)arg;
f46b5a66
CH		 4490
4491 switch (cmd) {
6cbff00f
CH		 4492 case FS_IOC_GETFLAGS:
4493 return btrfs_ioctl_getflags(file, argp);
4494 case FS_IOC_SETFLAGS:
4495 return btrfs_ioctl_setflags(file, argp);
4496 case FS_IOC_GETVERSION:
4497 return btrfs_ioctl_getversion(file, argp);
f7039b1d
LD		 4498 case FITRIM:
4499 return btrfs_ioctl_fitrim(file, argp);
f46b5a66 4500 case BTRFS_IOC_SNAP_CREATE:
fa0d2b9b 4501 return btrfs_ioctl_snap_create(file, argp, 0);
fdfb1e4f 4502 case BTRFS_IOC_SNAP_CREATE_V2:
fa0d2b9b 4503 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3de4586c 4504 case BTRFS_IOC_SUBVOL_CREATE:
fa0d2b9b 4505 return btrfs_ioctl_snap_create(file, argp, 1);
6f72c7e2
AJ		 4506 case BTRFS_IOC_SUBVOL_CREATE_V2:
4507 return btrfs_ioctl_snap_create_v2(file, argp, 1);
76dda93c
YZ		 4508 case BTRFS_IOC_SNAP_DESTROY:
4509 return btrfs_ioctl_snap_destroy(file, argp);
0caa102d
LZ		 4510 case BTRFS_IOC_SUBVOL_GETFLAGS:
4511 return btrfs_ioctl_subvol_getflags(file, argp);
4512 case BTRFS_IOC_SUBVOL_SETFLAGS:
4513 return btrfs_ioctl_subvol_setflags(file, argp);
6ef5ed0d
JB		 4514 case BTRFS_IOC_DEFAULT_SUBVOL:
4515 return btrfs_ioctl_default_subvol(file, argp);
f46b5a66 4516 case BTRFS_IOC_DEFRAG:
1e701a32
CM		 4517 return btrfs_ioctl_defrag(file, NULL);
4518 case BTRFS_IOC_DEFRAG_RANGE:
4519 return btrfs_ioctl_defrag(file, argp);
f46b5a66 4520 case BTRFS_IOC_RESIZE:
198605a8 4521 return btrfs_ioctl_resize(file, argp);
f46b5a66 4522 case BTRFS_IOC_ADD_DEV:
4bcabaa3 4523 return btrfs_ioctl_add_dev(root, argp);
f46b5a66 4524 case BTRFS_IOC_RM_DEV:
da24927b 4525 return btrfs_ioctl_rm_dev(file, argp);
475f6387
JS		 4526 case BTRFS_IOC_FS_INFO:
4527 return btrfs_ioctl_fs_info(root, argp);
4528 case BTRFS_IOC_DEV_INFO:
4529 return btrfs_ioctl_dev_info(root, argp);
f46b5a66 4530 case BTRFS_IOC_BALANCE:
9ba1f6e4 4531 return btrfs_ioctl_balance(file, NULL);
f46b5a66 4532 case BTRFS_IOC_CLONE:
c5c9cd4d
SW		 4533 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
4534 case BTRFS_IOC_CLONE_RANGE:
7a865e8a 4535 return btrfs_ioctl_clone_range(file, argp);
f46b5a66
CH		 4536 case BTRFS_IOC_TRANS_START:
4537 return btrfs_ioctl_trans_start(file);
4538 case BTRFS_IOC_TRANS_END:
4539 return btrfs_ioctl_trans_end(file);
ac8e9819
CM		 4540 case BTRFS_IOC_TREE_SEARCH:
4541 return btrfs_ioctl_tree_search(file, argp);
4542 case BTRFS_IOC_INO_LOOKUP:
4543 return btrfs_ioctl_ino_lookup(file, argp);
d7728c96
JS		 4544 case BTRFS_IOC_INO_PATHS:
4545 return btrfs_ioctl_ino_to_path(root, argp);
4546 case BTRFS_IOC_LOGICAL_INO:
4547 return btrfs_ioctl_logical_to_ino(root, argp);
1406e432
JB		 4548 case BTRFS_IOC_SPACE_INFO:
4549 return btrfs_ioctl_space_info(root, argp);
f46b5a66
CH		 4550 case BTRFS_IOC_SYNC:
4551 btrfs_sync_fs(file->f_dentry->d_sb, 1);
4552 return 0;
46204592 4553 case BTRFS_IOC_START_SYNC:
9a8c28be 4554 return btrfs_ioctl_start_sync(root, argp);
46204592 4555 case BTRFS_IOC_WAIT_SYNC:
9a8c28be 4556 return btrfs_ioctl_wait_sync(root, argp);
475f6387 4557 case BTRFS_IOC_SCRUB:
b8e95489 4558 return btrfs_ioctl_scrub(file, argp);
475f6387
JS		 4559 case BTRFS_IOC_SCRUB_CANCEL:
4560 return btrfs_ioctl_scrub_cancel(root, argp);
4561 case BTRFS_IOC_SCRUB_PROGRESS:
4562 return btrfs_ioctl_scrub_progress(root, argp);
c9e9f97b 4563 case BTRFS_IOC_BALANCE_V2:
9ba1f6e4 4564 return btrfs_ioctl_balance(file, argp);
837d5b6e
ID		 4565 case BTRFS_IOC_BALANCE_CTL:
4566 return btrfs_ioctl_balance_ctl(root, arg);
19a39dce
ID		 4567 case BTRFS_IOC_BALANCE_PROGRESS:
4568 return btrfs_ioctl_balance_progress(root, argp);
8ea05e3a
AB		 4569 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
4570 return btrfs_ioctl_set_received_subvol(file, argp);
31db9f7c
AB		 4571 case BTRFS_IOC_SEND:
4572 return btrfs_ioctl_send(file, argp);
c11d2c23 4573 case BTRFS_IOC_GET_DEV_STATS:
b27f7c0c 4574 return btrfs_ioctl_get_dev_stats(root, argp);
5d13a37b 4575 case BTRFS_IOC_QUOTA_CTL:
905b0dda 4576 return btrfs_ioctl_quota_ctl(file, argp);
5d13a37b 4577 case BTRFS_IOC_QGROUP_ASSIGN:
905b0dda 4578 return btrfs_ioctl_qgroup_assign(file, argp);
5d13a37b 4579 case BTRFS_IOC_QGROUP_CREATE:
905b0dda 4580 return btrfs_ioctl_qgroup_create(file, argp);
5d13a37b 4581 case BTRFS_IOC_QGROUP_LIMIT:
905b0dda 4582 return btrfs_ioctl_qgroup_limit(file, argp);
2f232036
JS		 4583 case BTRFS_IOC_QUOTA_RESCAN:
4584 return btrfs_ioctl_quota_rescan(file, argp);
4585 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
4586 return btrfs_ioctl_quota_rescan_status(file, argp);
57254b6e
JS		 4587 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
4588 return btrfs_ioctl_quota_rescan_wait(file, argp);
3f6bcfbd
SB		 4589 case BTRFS_IOC_DEV_REPLACE:
4590 return btrfs_ioctl_dev_replace(root, argp);
867ab667 4591 case BTRFS_IOC_GET_FSLABEL:
4592 return btrfs_ioctl_get_fslabel(file, argp);
a8bfd4ab 4593 case BTRFS_IOC_SET_FSLABEL:
4594 return btrfs_ioctl_set_fslabel(file, argp);
416161db
MF		 4595 case BTRFS_IOC_FILE_EXTENT_SAME:
4596 return btrfs_ioctl_file_extent_same(file, argp);
f46b5a66
CH		 4597 }
4598
4599 return -ENOTTY;
4600}
Ubuntu Hirsute Linux kernel mirror