projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Btrfs: create UUID tree if required
[linux-2.6-block.git] / fs / btrfs / ioctl.c
CommitLineData
f46b5a66
CH		 1/*
2 * Copyright (C) 2007 Oracle. All rights reserved.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public
6 * License v2 as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public
14 * License along with this program; if not, write to the
15 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
16 * Boston, MA 021110-1307, USA.
17 */
18
19#include <linux/kernel.h>
20#include <linux/bio.h>
21#include <linux/buffer_head.h>
22#include <linux/file.h>
23#include <linux/fs.h>
cb8e7090 24#include <linux/fsnotify.h>
f46b5a66
CH		 25#include <linux/pagemap.h>
26#include <linux/highmem.h>
27#include <linux/time.h>
28#include <linux/init.h>
29#include <linux/string.h>
f46b5a66 30#include <linux/backing-dev.h>
cb8e7090 31#include <linux/mount.h>
f46b5a66 32#include <linux/mpage.h>
cb8e7090 33#include <linux/namei.h>
f46b5a66
CH		 34#include <linux/swap.h>
35#include <linux/writeback.h>
36#include <linux/statfs.h>
37#include <linux/compat.h>
38#include <linux/bit_spinlock.h>
cb8e7090 39#include <linux/security.h>
f46b5a66 40#include <linux/xattr.h>
7ea394f1 41#include <linux/vmalloc.h>
5a0e3ad6 42#include <linux/slab.h>
f7039b1d 43#include <linux/blkdev.h>
8ea05e3a 44#include <linux/uuid.h>
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
d5c12070 370static noinline int create_subvol(struct inode *dir,
cb8e7090 371 struct dentry *dentry,
72fd032e 372 char *name, int namelen,
6f72c7e2 373 u64 *async_transid,
8696c533 374 struct btrfs_qgroup_inherit *inherit)
f46b5a66
CH		 375{
376 struct btrfs_trans_handle *trans;
377 struct btrfs_key key;
378 struct btrfs_root_item root_item;
379 struct btrfs_inode_item *inode_item;
380 struct extent_buffer *leaf;
d5c12070 381 struct btrfs_root *root = BTRFS_I(dir)->root;
76dda93c 382 struct btrfs_root *new_root;
d5c12070 383 struct btrfs_block_rsv block_rsv;
8ea05e3a 384 struct timespec cur_time = CURRENT_TIME;
f46b5a66
CH		 385 int ret;
386 int err;
387 u64 objectid;
388 u64 new_dirid = BTRFS_FIRST_FREE_OBJECTID;
3de4586c 389 u64 index = 0;
d5c12070 390 u64 qgroup_reserved;
8ea05e3a 391 uuid_le new_uuid;
f46b5a66 392
581bb050 393 ret = btrfs_find_free_objectid(root->fs_info->tree_root, &objectid);
2fbe8c8a 394 if (ret)
a22285a6 395 return ret;
6a912213 396
d5c12070 397 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
9ed74f2d 398 /*
d5c12070
MX		 399 * The same as the snapshot creation, please see the comment
400 * of create_snapshot().
9ed74f2d 401 */
d5c12070 402 ret = btrfs_subvolume_reserve_metadata(root, &block_rsv,
ee3441b4 403 7, &qgroup_reserved, false);
d5c12070
MX		 404 if (ret)
405 return ret;
406
407 trans = btrfs_start_transaction(root, 0);
408 if (IS_ERR(trans)) {
409 ret = PTR_ERR(trans);
410 goto out;
411 }
412 trans->block_rsv = &block_rsv;
413 trans->bytes_reserved = block_rsv.size;
f46b5a66 414
8696c533 415 ret = btrfs_qgroup_inherit(trans, root->fs_info, 0, objectid, inherit);
6f72c7e2
AJ		 416 if (ret)
417 goto fail;
418
5d4f98a2 419 leaf = btrfs_alloc_free_block(trans, root, root->leafsize,
5581a51a 420 0, objectid, NULL, 0, 0, 0);
8e8a1e31
JB		 421 if (IS_ERR(leaf)) {
422 ret = PTR_ERR(leaf);
423 goto fail;
424 }
f46b5a66 425
5d4f98a2 426 memset_extent_buffer(leaf, 0, 0, sizeof(struct btrfs_header));
f46b5a66
CH		 427 btrfs_set_header_bytenr(leaf, leaf->start);
428 btrfs_set_header_generation(leaf, trans->transid);
5d4f98a2 429 btrfs_set_header_backref_rev(leaf, BTRFS_MIXED_BACKREF_REV);
f46b5a66
CH		 430 btrfs_set_header_owner(leaf, objectid);
431
432 write_extent_buffer(leaf, root->fs_info->fsid,
433 (unsigned long)btrfs_header_fsid(leaf),
434 BTRFS_FSID_SIZE);
5d4f98a2
YZ		 435 write_extent_buffer(leaf, root->fs_info->chunk_tree_uuid,
436 (unsigned long)btrfs_header_chunk_tree_uuid(leaf),
437 BTRFS_UUID_SIZE);
f46b5a66
CH		 438 btrfs_mark_buffer_dirty(leaf);
439
8ea05e3a
AB		 440 memset(&root_item, 0, sizeof(root_item));
441
f46b5a66 442 inode_item = &root_item.inode;
3cae210f
QW		 443 btrfs_set_stack_inode_generation(inode_item, 1);
444 btrfs_set_stack_inode_size(inode_item, 3);
445 btrfs_set_stack_inode_nlink(inode_item, 1);
446 btrfs_set_stack_inode_nbytes(inode_item, root->leafsize);
447 btrfs_set_stack_inode_mode(inode_item, S_IFDIR | 0755);
f46b5a66 448
3cae210f
QW		 449 btrfs_set_root_flags(&root_item, 0);
450 btrfs_set_root_limit(&root_item, 0);
451 btrfs_set_stack_inode_flags(inode_item, BTRFS_INODE_ROOT_ITEM_INIT);
08fe4db1 452
f46b5a66 453 btrfs_set_root_bytenr(&root_item, leaf->start);
84234f3a 454 btrfs_set_root_generation(&root_item, trans->transid);
f46b5a66
CH		 455 btrfs_set_root_level(&root_item, 0);
456 btrfs_set_root_refs(&root_item, 1);
86b9f2ec 457 btrfs_set_root_used(&root_item, leaf->len);
80ff3856 458 btrfs_set_root_last_snapshot(&root_item, 0);
f46b5a66 459
8ea05e3a
AB		 460 btrfs_set_root_generation_v2(&root_item,
461 btrfs_root_generation(&root_item));
462 uuid_le_gen(&new_uuid);
463 memcpy(root_item.uuid, new_uuid.b, BTRFS_UUID_SIZE);
3cae210f
QW		 464 btrfs_set_stack_timespec_sec(&root_item.otime, cur_time.tv_sec);
465 btrfs_set_stack_timespec_nsec(&root_item.otime, cur_time.tv_nsec);
8ea05e3a
AB		 466 root_item.ctime = root_item.otime;
467 btrfs_set_root_ctransid(&root_item, trans->transid);
468 btrfs_set_root_otransid(&root_item, trans->transid);
f46b5a66 469
925baedd 470 btrfs_tree_unlock(leaf);
f46b5a66
CH		 471 free_extent_buffer(leaf);
472 leaf = NULL;
473
474 btrfs_set_root_dirid(&root_item, new_dirid);
475
476 key.objectid = objectid;
5d4f98a2 477 key.offset = 0;
f46b5a66
CH		 478 btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
479 ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
480 &root_item);
481 if (ret)
482 goto fail;
483
76dda93c
YZ		 484 key.offset = (u64)-1;
485 new_root = btrfs_read_fs_root_no_name(root->fs_info, &key);
79787eaa
JM		 486 if (IS_ERR(new_root)) {
487 btrfs_abort_transaction(trans, root, PTR_ERR(new_root));
488 ret = PTR_ERR(new_root);
489 goto fail;
490 }
76dda93c
YZ		 491
492 btrfs_record_root_in_trans(trans, new_root);
493
d82a6f1d 494 ret = btrfs_create_subvol_root(trans, new_root, new_dirid);
ce598979
MF		 495 if (ret) {
496 /* We potentially lose an unused inode item here */
79787eaa 497 btrfs_abort_transaction(trans, root, ret);
ce598979
MF		 498 goto fail;
499 }
500
f46b5a66
CH		 501 /*
502 * insert the directory item
503 */
3de4586c 504 ret = btrfs_set_inode_index(dir, &index);
79787eaa
JM		 505 if (ret) {
506 btrfs_abort_transaction(trans, root, ret);
507 goto fail;
508 }
3de4586c
CM		 509
510 ret = btrfs_insert_dir_item(trans, root,
16cdcec7 511 name, namelen, dir, &key,
3de4586c 512 BTRFS_FT_DIR, index);
79787eaa
JM		 513 if (ret) {
514 btrfs_abort_transaction(trans, root, ret);
f46b5a66 515 goto fail;
79787eaa 516 }
0660b5af 517
52c26179
YZ		 518 btrfs_i_size_write(dir, dir->i_size + namelen * 2);
519 ret = btrfs_update_inode(trans, root, dir);
520 BUG_ON(ret);
521
0660b5af 522 ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
4df27c4d 523 objectid, root->root_key.objectid,
33345d01 524 btrfs_ino(dir), index, name, namelen);
0660b5af 525
76dda93c 526 BUG_ON(ret);
f46b5a66 527
f46b5a66 528fail:
d5c12070
MX		 529 trans->block_rsv = NULL;
530 trans->bytes_reserved = 0;
72fd032e
SW		 531 if (async_transid) {
532 *async_transid = trans->transid;
533 err = btrfs_commit_transaction_async(trans, root, 1);
00d71c9c
MX		 534 if (err)
535 err = btrfs_commit_transaction(trans, root);
72fd032e
SW		 536 } else {
537 err = btrfs_commit_transaction(trans, root);
538 }
f46b5a66
CH		 539 if (err && !ret)
540 ret = err;
1a65e24b
CM		 541
542 if (!ret)
543 d_instantiate(dentry, btrfs_lookup_dentry(dir, dentry));
d5c12070
MX		 544out:
545 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
f46b5a66
CH		 546 return ret;
547}
548
e9662f70
MX		 549static int create_snapshot(struct btrfs_root *root, struct inode *dir,
550 struct dentry *dentry, char *name, int namelen,
551 u64 *async_transid, bool readonly,
552 struct btrfs_qgroup_inherit *inherit)
f46b5a66 553{
2e4bfab9 554 struct inode *inode;
f46b5a66
CH		 555 struct btrfs_pending_snapshot *pending_snapshot;
556 struct btrfs_trans_handle *trans;
2e4bfab9 557 int ret;
f46b5a66
CH		 558
559 if (!root->ref_cows)
560 return -EINVAL;
561
6a03843d
MX		 562 ret = btrfs_start_delalloc_inodes(root, 0);
563 if (ret)
564 return ret;
565
566 btrfs_wait_ordered_extents(root, 0);
567
3de4586c 568 pending_snapshot = kzalloc(sizeof(*pending_snapshot), GFP_NOFS);
a22285a6
YZ		 569 if (!pending_snapshot)
570 return -ENOMEM;
571
66d8f3dd
MX		 572 btrfs_init_block_rsv(&pending_snapshot->block_rsv,
573 BTRFS_BLOCK_RSV_TEMP);
d5c12070
MX		 574 /*
575 * 1 - parent dir inode
576 * 2 - dir entries
577 * 1 - root item
578 * 2 - root ref/backref
579 * 1 - root of snapshot
580 */
581 ret = btrfs_subvolume_reserve_metadata(BTRFS_I(dir)->root,
582 &pending_snapshot->block_rsv, 7,
ee3441b4
JM		 583 &pending_snapshot->qgroup_reserved,
584 false);
d5c12070
MX		 585 if (ret)
586 goto out;
587
3de4586c 588 pending_snapshot->dentry = dentry;
f46b5a66 589 pending_snapshot->root = root;
b83cc969 590 pending_snapshot->readonly = readonly;
e9662f70 591 pending_snapshot->dir = dir;
8696c533 592 pending_snapshot->inherit = inherit;
a22285a6 593
d5c12070 594 trans = btrfs_start_transaction(root, 0);
a22285a6
YZ		 595 if (IS_ERR(trans)) {
596 ret = PTR_ERR(trans);
597 goto fail;
598 }
599
8351583e 600 spin_lock(&root->fs_info->trans_lock);
f46b5a66
CH		 601 list_add(&pending_snapshot->list,
602 &trans->transaction->pending_snapshots);
8351583e 603 spin_unlock(&root->fs_info->trans_lock);
72fd032e
SW		 604 if (async_transid) {
605 *async_transid = trans->transid;
606 ret = btrfs_commit_transaction_async(trans,
607 root->fs_info->extent_root, 1);
00d71c9c
MX		 608 if (ret)
609 ret = btrfs_commit_transaction(trans, root);
72fd032e
SW		 610 } else {
611 ret = btrfs_commit_transaction(trans,
612 root->fs_info->extent_root);
613 }
aec8030a 614 if (ret)
c37b2b62 615 goto fail;
a22285a6
YZ		 616
617 ret = pending_snapshot->error;
618 if (ret)
619 goto fail;
620
66b4ffd1
JB		 621 ret = btrfs_orphan_cleanup(pending_snapshot->snap);
622 if (ret)
623 goto fail;
f46b5a66 624
2fbe8c8a 625 inode = btrfs_lookup_dentry(dentry->d_parent->d_inode, dentry);
2e4bfab9
YZ		 626 if (IS_ERR(inode)) {
627 ret = PTR_ERR(inode);
628 goto fail;
629 }
630 BUG_ON(!inode);
631 d_instantiate(dentry, inode);
632 ret = 0;
633fail:
d5c12070
MX		 634 btrfs_subvolume_release_metadata(BTRFS_I(dir)->root,
635 &pending_snapshot->block_rsv,
636 pending_snapshot->qgroup_reserved);
637out:
a22285a6 638 kfree(pending_snapshot);
f46b5a66
CH		 639 return ret;
640}
641
4260f7c7
SW		 642/* copy of check_sticky in fs/namei.c()
643* It's inline, so penalty for filesystems that don't use sticky bit is
644* minimal.
645*/
646static inline int btrfs_check_sticky(struct inode *dir, struct inode *inode)
647{
2f2f43d3 648 kuid_t fsuid = current_fsuid();
4260f7c7
SW		 649
650 if (!(dir->i_mode & S_ISVTX))
651 return 0;
2f2f43d3 652 if (uid_eq(inode->i_uid, fsuid))
4260f7c7 653 return 0;
2f2f43d3 654 if (uid_eq(dir->i_uid, fsuid))
4260f7c7
SW		 655 return 0;
656 return !capable(CAP_FOWNER);
657}
658
659/* copy of may_delete in fs/namei.c()
660 * Check whether we can remove a link victim from directory dir, check
661 * whether the type of victim is right.
662 * 1. We can't do it if dir is read-only (done in permission())
663 * 2. We should have write and exec permissions on dir
664 * 3. We can't remove anything from append-only dir
665 * 4. We can't do anything with immutable dir (done in permission())
666 * 5. If the sticky bit on dir is set we should either
667 * a. be owner of dir, or
668 * b. be owner of victim, or
669 * c. have CAP_FOWNER capability
670 * 6. If the victim is append-only or immutable we can't do antyhing with
671 * links pointing to it.
672 * 7. If we were asked to remove a directory and victim isn't one - ENOTDIR.
673 * 8. If we were asked to remove a non-directory and victim isn't one - EISDIR.
674 * 9. We can't remove a root or mountpoint.
675 * 10. We don't allow removal of NFS sillyrenamed files; it's handled by
676 * nfs_async_unlink().
677 */
678
679static int btrfs_may_delete(struct inode *dir,struct dentry *victim,int isdir)
680{
681 int error;
682
683 if (!victim->d_inode)
684 return -ENOENT;
685
686 BUG_ON(victim->d_parent->d_inode != dir);
4fa6b5ec 687 audit_inode_child(dir, victim, AUDIT_TYPE_CHILD_DELETE);
4260f7c7
SW		 688
689 error = inode_permission(dir, MAY_WRITE | MAY_EXEC);
690 if (error)
691 return error;
692 if (IS_APPEND(dir))
693 return -EPERM;
694 if (btrfs_check_sticky(dir, victim->d_inode)||
695 IS_APPEND(victim->d_inode)||
696 IS_IMMUTABLE(victim->d_inode) || IS_SWAPFILE(victim->d_inode))
697 return -EPERM;
698 if (isdir) {
699 if (!S_ISDIR(victim->d_inode->i_mode))
700 return -ENOTDIR;
701 if (IS_ROOT(victim))
702 return -EBUSY;
703 } else if (S_ISDIR(victim->d_inode->i_mode))
704 return -EISDIR;
705 if (IS_DEADDIR(dir))
706 return -ENOENT;
707 if (victim->d_flags & DCACHE_NFSFS_RENAMED)
708 return -EBUSY;
709 return 0;
710}
711
cb8e7090
CH		 712/* copy of may_create in fs/namei.c() */
713static inline int btrfs_may_create(struct inode *dir, struct dentry *child)
714{
715 if (child->d_inode)
716 return -EEXIST;
717 if (IS_DEADDIR(dir))
718 return -ENOENT;
719 return inode_permission(dir, MAY_WRITE | MAY_EXEC);
720}
721
722/*
723 * Create a new subvolume below @parent. This is largely modeled after
724 * sys_mkdirat and vfs_mkdir, but we only do a single component lookup
725 * inside this filesystem so it's quite a bit simpler.
726 */
76dda93c
YZ		 727static noinline int btrfs_mksubvol(struct path *parent,
728 char *name, int namelen,
72fd032e 729 struct btrfs_root *snap_src,
6f72c7e2 730 u64 *async_transid, bool readonly,
8696c533 731 struct btrfs_qgroup_inherit *inherit)
cb8e7090 732{
76dda93c 733 struct inode *dir = parent->dentry->d_inode;
cb8e7090
CH		 734 struct dentry *dentry;
735 int error;
736
5c50c9b8
DS		 737 error = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
738 if (error == -EINTR)
739 return error;
cb8e7090
CH		 740
741 dentry = lookup_one_len(name, parent->dentry, namelen);
742 error = PTR_ERR(dentry);
743 if (IS_ERR(dentry))
744 goto out_unlock;
745
746 error = -EEXIST;
747 if (dentry->d_inode)
748 goto out_dput;
749
76dda93c 750 error = btrfs_may_create(dir, dentry);
cb8e7090 751 if (error)
a874a63e 752 goto out_dput;
cb8e7090 753
9c52057c
CM		 754 /*
755 * even if this name doesn't exist, we may get hash collisions.
756 * check for them now when we can safely fail
757 */
758 error = btrfs_check_dir_item_collision(BTRFS_I(dir)->root,
759 dir->i_ino, name,
760 namelen);
761 if (error)
762 goto out_dput;
763
76dda93c
YZ		 764 down_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
765
766 if (btrfs_root_refs(&BTRFS_I(dir)->root->root_item) == 0)
767 goto out_up_read;
768
3de4586c 769 if (snap_src) {
e9662f70 770 error = create_snapshot(snap_src, dir, dentry, name, namelen,
6f72c7e2 771 async_transid, readonly, inherit);
3de4586c 772 } else {
d5c12070
MX		 773 error = create_subvol(dir, dentry, name, namelen,
774 async_transid, inherit);
3de4586c 775 }
76dda93c
YZ		 776 if (!error)
777 fsnotify_mkdir(dir, dentry);
778out_up_read:
779 up_read(&BTRFS_I(dir)->root->fs_info->subvol_sem);
cb8e7090
CH		 780out_dput:
781 dput(dentry);
782out_unlock:
76dda93c 783 mutex_unlock(&dir->i_mutex);
cb8e7090
CH		 784 return error;
785}
786
4cb5300b
CM		 787/*
788 * When we're defragging a range, we don't want to kick it off again
789 * if it is really just waiting for delalloc to send it down.
790 * If we find a nice big extent or delalloc range for the bytes in the
791 * file you want to defrag, we return 0 to let you know to skip this
792 * part of the file
793 */
794static int check_defrag_in_cache(struct inode *inode, u64 offset, int thresh)
795{
796 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
797 struct extent_map *em = NULL;
798 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
799 u64 end;
800
801 read_lock(&em_tree->lock);
802 em = lookup_extent_mapping(em_tree, offset, PAGE_CACHE_SIZE);
803 read_unlock(&em_tree->lock);
804
805 if (em) {
806 end = extent_map_end(em);
807 free_extent_map(em);
808 if (end - offset > thresh)
809 return 0;
810 }
811 /* if we already have a nice delalloc here, just stop */
812 thresh /= 2;
813 end = count_range_bits(io_tree, &offset, offset + thresh,
814 thresh, EXTENT_DELALLOC, 1);
815 if (end >= thresh)
816 return 0;
817 return 1;
818}
819
820/*
821 * helper function to walk through a file and find extents
822 * newer than a specific transid, and smaller than thresh.
823 *
824 * This is used by the defragging code to find new and small
825 * extents
826 */
827static int find_new_extents(struct btrfs_root *root,
828 struct inode *inode, u64 newer_than,
829 u64 *off, int thresh)
830{
831 struct btrfs_path *path;
832 struct btrfs_key min_key;
833 struct btrfs_key max_key;
834 struct extent_buffer *leaf;
835 struct btrfs_file_extent_item *extent;
836 int type;
837 int ret;
a4689d2b 838 u64 ino = btrfs_ino(inode);
4cb5300b
CM		 839
840 path = btrfs_alloc_path();
841 if (!path)
842 return -ENOMEM;
843
a4689d2b 844 min_key.objectid = ino;
4cb5300b
CM		 845 min_key.type = BTRFS_EXTENT_DATA_KEY;
846 min_key.offset = *off;
847
a4689d2b 848 max_key.objectid = ino;
4cb5300b
CM		 849 max_key.type = (u8)-1;
850 max_key.offset = (u64)-1;
851
852 path->keep_locks = 1;
853
854 while(1) {
855 ret = btrfs_search_forward(root, &min_key, &max_key,
de78b51a 856 path, newer_than);
4cb5300b
CM		 857 if (ret != 0)
858 goto none;
a4689d2b 859 if (min_key.objectid != ino)
4cb5300b
CM		 860 goto none;
861 if (min_key.type != BTRFS_EXTENT_DATA_KEY)
862 goto none;
863
864 leaf = path->nodes[0];
865 extent = btrfs_item_ptr(leaf, path->slots[0],
866 struct btrfs_file_extent_item);
867
868 type = btrfs_file_extent_type(leaf, extent);
869 if (type == BTRFS_FILE_EXTENT_REG &&
870 btrfs_file_extent_num_bytes(leaf, extent) < thresh &&
871 check_defrag_in_cache(inode, min_key.offset, thresh)) {
872 *off = min_key.offset;
873 btrfs_free_path(path);
874 return 0;
875 }
876
877 if (min_key.offset == (u64)-1)
878 goto none;
879
880 min_key.offset++;
881 btrfs_release_path(path);
882 }
883none:
884 btrfs_free_path(path);
885 return -ENOENT;
886}
887
6c282eb4 888static struct extent_map *defrag_lookup_extent(struct inode *inode, u64 start)
17ce6ef8
LB		 889{
890 struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
6c282eb4
LZ		 891 struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
892 struct extent_map *em;
893 u64 len = PAGE_CACHE_SIZE;
17ce6ef8 894
6c282eb4
LZ		 895 /*
896 * hopefully we have this extent in the tree already, try without
897 * the full extent lock
898 */
17ce6ef8 899 read_lock(&em_tree->lock);
6c282eb4 900 em = lookup_extent_mapping(em_tree, start, len);
17ce6ef8
LB		 901 read_unlock(&em_tree->lock);
902
6c282eb4
LZ		 903 if (!em) {
904 /* get the big lock and read metadata off disk */
905 lock_extent(io_tree, start, start + len - 1);
906 em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
907 unlock_extent(io_tree, start, start + len - 1);
908
909 if (IS_ERR(em))
910 return NULL;
911 }
912
913 return em;
914}
17ce6ef8 915
6c282eb4
LZ		 916static bool defrag_check_next_extent(struct inode *inode, struct extent_map *em)
917{
918 struct extent_map *next;
919 bool ret = true;
920
921 /* this is the last extent */
922 if (em->start + em->len >= i_size_read(inode))
923 return false;
924
925 next = defrag_lookup_extent(inode, em->start + em->len);
926 if (!next || next->block_start >= EXTENT_MAP_LAST_BYTE)
927 ret = false;
928
929 free_extent_map(next);
17ce6ef8
LB		 930 return ret;
931}
932
6c282eb4 933static int should_defrag_range(struct inode *inode, u64 start, int thresh,
a43a2111
AM		 934 u64 *last_len, u64 *skip, u64 *defrag_end,
935 int compress)
940100a4 936{
6c282eb4 937 struct extent_map *em;
940100a4 938 int ret = 1;
6c282eb4 939 bool next_mergeable = true;
940100a4
CM		 940
941 /*
008873ea 942 * make sure that once we start defragging an extent, we keep on
940100a4
CM		 943 * defragging it
944 */
945 if (start < *defrag_end)
946 return 1;
947
948 *skip = 0;
949
6c282eb4
LZ		 950 em = defrag_lookup_extent(inode, start);
951 if (!em)
952 return 0;
940100a4
CM		 953
954 /* this will cover holes, and inline extents */
17ce6ef8 955 if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
940100a4 956 ret = 0;
17ce6ef8
LB		 957 goto out;
958 }
959
6c282eb4 960 next_mergeable = defrag_check_next_extent(inode, em);
940100a4
CM		 961
962 /*
6c282eb4
LZ		 963 * we hit a real extent, if it is big or the next extent is not a
964 * real extent, don't bother defragging it
940100a4 965 */
a43a2111 966 if (!compress && (*last_len == 0 || *last_len >= thresh) &&
6c282eb4 967 (em->len >= thresh || !next_mergeable))
940100a4 968 ret = 0;
17ce6ef8 969out:
940100a4
CM		 970 /*
971 * last_len ends up being a counter of how many bytes we've defragged.
972 * every time we choose not to defrag an extent, we reset *last_len
973 * so that the next tiny extent will force a defrag.
974 *
975 * The end result of this is that tiny extents before a single big
976 * extent will force at least part of that big extent to be defragged.
977 */
978 if (ret) {
940100a4
CM		 979 *defrag_end = extent_map_end(em);
980 } else {
981 *last_len = 0;
982 *skip = extent_map_end(em);
983 *defrag_end = 0;
984 }
985
986 free_extent_map(em);
987 return ret;
988}
989
4cb5300b
CM		 990/*
991 * it doesn't do much good to defrag one or two pages
992 * at a time. This pulls in a nice chunk of pages
993 * to COW and defrag.
994 *
995 * It also makes sure the delalloc code has enough
996 * dirty data to avoid making new small extents as part
997 * of the defrag
998 *
999 * It's a good idea to start RA on this range
1000 * before calling this.
1001 */
1002static int cluster_pages_for_defrag(struct inode *inode,
1003 struct page **pages,
1004 unsigned long start_index,
1005 int num_pages)
f46b5a66 1006{
4cb5300b
CM		 1007 unsigned long file_end;
1008 u64 isize = i_size_read(inode);
1009 u64 page_start;
1010 u64 page_end;
1f12bd06 1011 u64 page_cnt;
4cb5300b
CM		 1012 int ret;
1013 int i;
1014 int i_done;
3eaa2885 1015 struct btrfs_ordered_extent *ordered;
4cb5300b 1016 struct extent_state *cached_state = NULL;
600a45e1 1017 struct extent_io_tree *tree;
3b16a4e3 1018 gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
4cb5300b 1019
4cb5300b 1020 file_end = (isize - 1) >> PAGE_CACHE_SHIFT;
1f12bd06
LB		 1021 if (!isize || start_index > file_end)
1022 return 0;
1023
1024 page_cnt = min_t(u64, (u64)num_pages, (u64)file_end - start_index + 1);
4cb5300b
CM		 1025
1026 ret = btrfs_delalloc_reserve_space(inode,
1f12bd06 1027 page_cnt << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1028 if (ret)
1029 return ret;
4cb5300b 1030 i_done = 0;
600a45e1 1031 tree = &BTRFS_I(inode)->io_tree;
4cb5300b
CM		 1032
1033 /* step one, lock all the pages */
1f12bd06 1034 for (i = 0; i < page_cnt; i++) {
4cb5300b 1035 struct page *page;
600a45e1 1036again:
a94733d0 1037 page = find_or_create_page(inode->i_mapping,
600a45e1 1038 start_index + i, mask);
4cb5300b
CM		 1039 if (!page)
1040 break;
1041
600a45e1
MX		 1042 page_start = page_offset(page);
1043 page_end = page_start + PAGE_CACHE_SIZE - 1;
1044 while (1) {
d0082371 1045 lock_extent(tree, page_start, page_end);
600a45e1
MX		 1046 ordered = btrfs_lookup_ordered_extent(inode,
1047 page_start);
d0082371 1048 unlock_extent(tree, page_start, page_end);
600a45e1
MX		 1049 if (!ordered)
1050 break;
1051
1052 unlock_page(page);
1053 btrfs_start_ordered_extent(inode, ordered, 1);
1054 btrfs_put_ordered_extent(ordered);
1055 lock_page(page);
1f12bd06
LB		 1056 /*
1057 * we unlocked the page above, so we need check if
1058 * it was released or not.
1059 */
1060 if (page->mapping != inode->i_mapping) {
1061 unlock_page(page);
1062 page_cache_release(page);
1063 goto again;
1064 }
600a45e1
MX		 1065 }
1066
4cb5300b
CM		 1067 if (!PageUptodate(page)) {
1068 btrfs_readpage(NULL, page);
1069 lock_page(page);
1070 if (!PageUptodate(page)) {
1071 unlock_page(page);
1072 page_cache_release(page);
1073 ret = -EIO;
1074 break;
1075 }
1076 }
600a45e1 1077
600a45e1
MX		 1078 if (page->mapping != inode->i_mapping) {
1079 unlock_page(page);
1080 page_cache_release(page);
1081 goto again;
1082 }
1083
4cb5300b
CM		 1084 pages[i] = page;
1085 i_done++;
1086 }
1087 if (!i_done || ret)
1088 goto out;
1089
1090 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1091 goto out;
1092
1093 /*
1094 * so now we have a nice long stream of locked
1095 * and up to date pages, lets wait on them
1096 */
1097 for (i = 0; i < i_done; i++)
1098 wait_on_page_writeback(pages[i]);
1099
1100 page_start = page_offset(pages[0]);
1101 page_end = page_offset(pages[i_done - 1]) + PAGE_CACHE_SIZE;
1102
1103 lock_extent_bits(&BTRFS_I(inode)->io_tree,
d0082371 1104 page_start, page_end - 1, 0, &cached_state);
4cb5300b
CM		 1105 clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start,
1106 page_end - 1, EXTENT_DIRTY | EXTENT_DELALLOC |
9e8a4a8b
LB		 1107 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 0, 0,
1108 &cached_state, GFP_NOFS);
4cb5300b 1109
1f12bd06 1110 if (i_done != page_cnt) {
9e0baf60
JB		 1111 spin_lock(&BTRFS_I(inode)->lock);
1112 BTRFS_I(inode)->outstanding_extents++;
1113 spin_unlock(&BTRFS_I(inode)->lock);
4cb5300b 1114 btrfs_delalloc_release_space(inode,
1f12bd06 1115 (page_cnt - i_done) << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1116 }
1117
1118
9e8a4a8b
LB		 1119 set_extent_defrag(&BTRFS_I(inode)->io_tree, page_start, page_end - 1,
1120 &cached_state, GFP_NOFS);
4cb5300b
CM		 1121
1122 unlock_extent_cached(&BTRFS_I(inode)->io_tree,
1123 page_start, page_end - 1, &cached_state,
1124 GFP_NOFS);
1125
1126 for (i = 0; i < i_done; i++) {
1127 clear_page_dirty_for_io(pages[i]);
1128 ClearPageChecked(pages[i]);
1129 set_page_extent_mapped(pages[i]);
1130 set_page_dirty(pages[i]);
1131 unlock_page(pages[i]);
1132 page_cache_release(pages[i]);
1133 }
1134 return i_done;
1135out:
1136 for (i = 0; i < i_done; i++) {
1137 unlock_page(pages[i]);
1138 page_cache_release(pages[i]);
1139 }
1f12bd06 1140 btrfs_delalloc_release_space(inode, page_cnt << PAGE_CACHE_SHIFT);
4cb5300b
CM		 1141 return ret;
1142
1143}
1144
1145int btrfs_defrag_file(struct inode *inode, struct file *file,
1146 struct btrfs_ioctl_defrag_range_args *range,
1147 u64 newer_than, unsigned long max_to_defrag)
1148{
1149 struct btrfs_root *root = BTRFS_I(inode)->root;
4cb5300b 1150 struct file_ra_state *ra = NULL;
f46b5a66 1151 unsigned long last_index;
151a31b2 1152 u64 isize = i_size_read(inode);
940100a4
CM		 1153 u64 last_len = 0;
1154 u64 skip = 0;
1155 u64 defrag_end = 0;
4cb5300b 1156 u64 newer_off = range->start;
f46b5a66 1157 unsigned long i;
008873ea 1158 unsigned long ra_index = 0;
f46b5a66 1159 int ret;
4cb5300b 1160 int defrag_count = 0;
1a419d85 1161 int compress_type = BTRFS_COMPRESS_ZLIB;
4cb5300b 1162 int extent_thresh = range->extent_thresh;
008873ea
LZ		 1163 int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
1164 int cluster = max_cluster;
4cb5300b
CM		 1165 u64 new_align = ~((u64)128 * 1024 - 1);
1166 struct page **pages = NULL;
1167
0abd5b17
LB		 1168 if (isize == 0)
1169 return 0;
1170
1171 if (range->start >= isize)
1172 return -EINVAL;
1a419d85
LZ		 1173
1174 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS) {
1175 if (range->compress_type > BTRFS_COMPRESS_TYPES)
1176 return -EINVAL;
1177 if (range->compress_type)
1178 compress_type = range->compress_type;
1179 }
f46b5a66 1180
0abd5b17
LB		 1181 if (extent_thresh == 0)
1182 extent_thresh = 256 * 1024;
940100a4 1183
4cb5300b
CM		 1184 /*
1185 * if we were not given a file, allocate a readahead
1186 * context
1187 */
1188 if (!file) {
1189 ra = kzalloc(sizeof(*ra), GFP_NOFS);
1190 if (!ra)
1191 return -ENOMEM;
1192 file_ra_state_init(ra, inode->i_mapping);
1193 } else {
1194 ra = &file->f_ra;
1195 }
1196
008873ea 1197 pages = kmalloc(sizeof(struct page *) * max_cluster,
4cb5300b
CM		 1198 GFP_NOFS);
1199 if (!pages) {
1200 ret = -ENOMEM;
1201 goto out_ra;
1202 }
1203
1204 /* find the last page to defrag */
1e701a32 1205 if (range->start + range->len > range->start) {
151a31b2 1206 last_index = min_t(u64, isize - 1,
1e701a32
CM		 1207 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
1208 } else {
151a31b2 1209 last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
1e701a32
CM		 1210 }
1211
4cb5300b
CM		 1212 if (newer_than) {
1213 ret = find_new_extents(root, inode, newer_than,
1214 &newer_off, 64 * 1024);
1215 if (!ret) {
1216 range->start = newer_off;
1217 /*
1218 * we always align our defrag to help keep
1219 * the extents in the file evenly spaced
1220 */
1221 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
4cb5300b
CM		 1222 } else
1223 goto out_ra;
1224 } else {
1225 i = range->start >> PAGE_CACHE_SHIFT;
1226 }
1227 if (!max_to_defrag)
7ec31b54 1228 max_to_defrag = last_index + 1;
4cb5300b 1229
2a0f7f57
LZ		 1230 /*
1231 * make writeback starts from i, so the defrag range can be
1232 * written sequentially.
1233 */
1234 if (i < inode->i_mapping->writeback_index)
1235 inode->i_mapping->writeback_index = i;
1236
f7f43cc8
CM		 1237 while (i <= last_index && defrag_count < max_to_defrag &&
1238 (i < (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
1239 PAGE_CACHE_SHIFT)) {
4cb5300b
CM		 1240 /*
1241 * make sure we stop running if someone unmounts
1242 * the FS
1243 */
1244 if (!(inode->i_sb->s_flags & MS_ACTIVE))
1245 break;
1246
210549eb
DS		 1247 if (btrfs_defrag_cancelled(root->fs_info)) {
1248 printk(KERN_DEBUG "btrfs: defrag_file cancelled\n");
1249 ret = -EAGAIN;
1250 break;
1251 }
1252
66c26892 1253 if (!should_defrag_range(inode, (u64)i << PAGE_CACHE_SHIFT,
6c282eb4 1254 extent_thresh, &last_len, &skip,
a43a2111
AM		 1255 &defrag_end, range->flags &
1256 BTRFS_DEFRAG_RANGE_COMPRESS)) {
940100a4
CM		 1257 unsigned long next;
1258 /*
1259 * the should_defrag function tells us how much to skip
1260 * bump our counter by the suggested amount
1261 */
1262 next = (skip + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
1263 i = max(i + 1, next);
1264 continue;
1265 }
008873ea
LZ		 1266
1267 if (!newer_than) {
1268 cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
1269 PAGE_CACHE_SHIFT) - i;
1270 cluster = min(cluster, max_cluster);
1271 } else {
1272 cluster = max_cluster;
1273 }
1274
1e701a32 1275 if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
1a419d85 1276 BTRFS_I(inode)->force_compress = compress_type;
940100a4 1277
008873ea
LZ		 1278 if (i + cluster > ra_index) {
1279 ra_index = max(i, ra_index);
1280 btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
1281 cluster);
1282 ra_index += max_cluster;
1283 }
940100a4 1284
ecb8bea8 1285 mutex_lock(&inode->i_mutex);
008873ea 1286 ret = cluster_pages_for_defrag(inode, pages, i, cluster);
ecb8bea8
LB		 1287 if (ret < 0) {
1288 mutex_unlock(&inode->i_mutex);
4cb5300b 1289 goto out_ra;
ecb8bea8 1290 }
4cb5300b
CM		 1291
1292 defrag_count += ret;
d0e1d66b 1293 balance_dirty_pages_ratelimited(inode->i_mapping);
ecb8bea8 1294 mutex_unlock(&inode->i_mutex);
4cb5300b
CM		 1295
1296 if (newer_than) {
1297 if (newer_off == (u64)-1)
1298 break;
1299
e1f041e1
LB		 1300 if (ret > 0)
1301 i += ret;
1302
4cb5300b
CM		 1303 newer_off = max(newer_off + 1,
1304 (u64)i << PAGE_CACHE_SHIFT);
1305
1306 ret = find_new_extents(root, inode,
1307 newer_than, &newer_off,
1308 64 * 1024);
1309 if (!ret) {
1310 range->start = newer_off;
1311 i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
4cb5300b
CM		 1312 } else {
1313 break;
f46b5a66 1314 }
4cb5300b 1315 } else {
008873ea 1316 if (ret > 0) {
cbcc8326 1317 i += ret;
008873ea
LZ		 1318 last_len += ret << PAGE_CACHE_SHIFT;
1319 } else {
cbcc8326 1320 i++;
008873ea
LZ		 1321 last_len = 0;
1322 }
f46b5a66 1323 }
f46b5a66
CH		 1324 }
1325
1e701a32
CM		 1326 if ((range->flags & BTRFS_DEFRAG_RANGE_START_IO))
1327 filemap_flush(inode->i_mapping);
1328
1329 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
1330 /* the filemap_flush will queue IO into the worker threads, but
1331 * we have to make sure the IO is actually started and that
1332 * ordered extents get created before we return
1333 */
1334 atomic_inc(&root->fs_info->async_submit_draining);
1335 while (atomic_read(&root->fs_info->nr_async_submits) ||
1336 atomic_read(&root->fs_info->async_delalloc_pages)) {
1337 wait_event(root->fs_info->async_submit_wait,
1338 (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
1339 atomic_read(&root->fs_info->async_delalloc_pages) == 0));
1340 }
1341 atomic_dec(&root->fs_info->async_submit_draining);
1342
1343 mutex_lock(&inode->i_mutex);
261507a0 1344 BTRFS_I(inode)->force_compress = BTRFS_COMPRESS_NONE;
1e701a32
CM		 1345 mutex_unlock(&inode->i_mutex);
1346 }
1347
1a419d85 1348 if (range->compress_type == BTRFS_COMPRESS_LZO) {
2b0ce2c2 1349 btrfs_set_fs_incompat(root->fs_info, COMPRESS_LZO);
1a419d85
LZ		 1350 }
1351
60ccf82f 1352 ret = defrag_count;
940100a4 1353
4cb5300b
CM		 1354out_ra:
1355 if (!file)
1356 kfree(ra);
1357 kfree(pages);
940100a4 1358 return ret;
f46b5a66
CH		 1359}
1360
198605a8 1361static noinline int btrfs_ioctl_resize(struct file *file,
76dda93c 1362 void __user *arg)
f46b5a66
CH		 1363{
1364 u64 new_size;
1365 u64 old_size;
1366 u64 devid = 1;
496ad9aa 1367 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
f46b5a66
CH		 1368 struct btrfs_ioctl_vol_args *vol_args;
1369 struct btrfs_trans_handle *trans;
1370 struct btrfs_device *device = NULL;
1371 char *sizestr;
1372 char *devstr = NULL;
1373 int ret = 0;
f46b5a66
CH		 1374 int mod = 0;
1375
e441d54d
CM		 1376 if (!capable(CAP_SYS_ADMIN))
1377 return -EPERM;
1378
198605a8
MX		 1379 ret = mnt_want_write_file(file);
1380 if (ret)
1381 return ret;
1382
5ac00add
SB		 1383 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
1384 1)) {
1385 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
97547676 1386 mnt_drop_write_file(file);
2c0c9da0 1387 return -EINVAL;
c9e9f97b
ID		 1388 }
1389
5ac00add 1390 mutex_lock(&root->fs_info->volume_mutex);
dae7b665 1391 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 1392 if (IS_ERR(vol_args)) {
1393 ret = PTR_ERR(vol_args);
1394 goto out;
1395 }
5516e595
MF		 1396
1397 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 1398
f46b5a66
CH		 1399 sizestr = vol_args->name;
1400 devstr = strchr(sizestr, ':');
1401 if (devstr) {
1402 char *end;
1403 sizestr = devstr + 1;
1404 *devstr = '\0';
1405 devstr = vol_args->name;
1406 devid = simple_strtoull(devstr, &end, 10);
dfd79829
MX		 1407 if (!devid) {
1408 ret = -EINVAL;
1409 goto out_free;
1410 }
5bb14682 1411 printk(KERN_INFO "btrfs: resizing devid %llu\n",
21380931 1412 (unsigned long long)devid);
f46b5a66 1413 }
dba60f3f 1414
aa1b8cd4 1415 device = btrfs_find_device(root->fs_info, devid, NULL, NULL);
f46b5a66 1416 if (!device) {
5bb14682 1417 printk(KERN_INFO "btrfs: resizer unable to find device %llu\n",
21380931 1418 (unsigned long long)devid);
dfd79829 1419 ret = -ENODEV;
c9e9f97b 1420 goto out_free;
f46b5a66 1421 }
dba60f3f
MX		 1422
1423 if (!device->writeable) {
4e42ae1b 1424 printk(KERN_INFO "btrfs: resizer unable to apply on "
dba60f3f 1425 "readonly device %llu\n",
a8c4a33b 1426 (unsigned long long)devid);
dfd79829 1427 ret = -EPERM;
4e42ae1b
LB		 1428 goto out_free;
1429 }
1430
f46b5a66
CH		 1431 if (!strcmp(sizestr, "max"))
1432 new_size = device->bdev->bd_inode->i_size;
1433 else {
1434 if (sizestr[0] == '-') {
1435 mod = -1;
1436 sizestr++;
1437 } else if (sizestr[0] == '+') {
1438 mod = 1;
1439 sizestr++;
1440 }
91748467 1441 new_size = memparse(sizestr, NULL);
f46b5a66
CH		 1442 if (new_size == 0) {
1443 ret = -EINVAL;
c9e9f97b 1444 goto out_free;
f46b5a66
CH		 1445 }
1446 }
1447
63a212ab 1448 if (device->is_tgtdev_for_dev_replace) {
dfd79829 1449 ret = -EPERM;
63a212ab
SB		 1450 goto out_free;
1451 }
1452
f46b5a66
CH		 1453 old_size = device->total_bytes;
1454
1455 if (mod < 0) {
1456 if (new_size > old_size) {
1457 ret = -EINVAL;
c9e9f97b 1458 goto out_free;
f46b5a66
CH		 1459 }
1460 new_size = old_size - new_size;
1461 } else if (mod > 0) {
1462 new_size = old_size + new_size;
1463 }
1464
1465 if (new_size < 256 * 1024 * 1024) {
1466 ret = -EINVAL;
c9e9f97b 1467 goto out_free;
f46b5a66
CH		 1468 }
1469 if (new_size > device->bdev->bd_inode->i_size) {
1470 ret = -EFBIG;
c9e9f97b 1471 goto out_free;
f46b5a66
CH		 1472 }
1473
1474 do_div(new_size, root->sectorsize);
1475 new_size *= root->sectorsize;
1476
606686ee
JB		 1477 printk_in_rcu(KERN_INFO "btrfs: new size for %s is %llu\n",
1478 rcu_str_deref(device->name),
1479 (unsigned long long)new_size);
f46b5a66
CH		 1480
1481 if (new_size > old_size) {
a22285a6 1482 trans = btrfs_start_transaction(root, 0);
98d5dc13
TI		 1483 if (IS_ERR(trans)) {
1484 ret = PTR_ERR(trans);
c9e9f97b 1485 goto out_free;
98d5dc13 1486 }
f46b5a66
CH		 1487 ret = btrfs_grow_device(trans, device, new_size);
1488 btrfs_commit_transaction(trans, root);
ece7d20e 1489 } else if (new_size < old_size) {
f46b5a66 1490 ret = btrfs_shrink_device(device, new_size);
0253f40e 1491 } /* equal, nothing need to do */
f46b5a66 1492
c9e9f97b 1493out_free:
f46b5a66 1494 kfree(vol_args);
c9e9f97b
ID		 1495out:
1496 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 1497 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
18f39c41 1498 mnt_drop_write_file(file);
f46b5a66
CH		 1499 return ret;
1500}
1501
72fd032e 1502static noinline int btrfs_ioctl_snap_create_transid(struct file *file,
6f72c7e2
AJ		 1503 char *name, unsigned long fd, int subvol,
1504 u64 *transid, bool readonly,
8696c533 1505 struct btrfs_qgroup_inherit *inherit)
f46b5a66 1506{
f46b5a66 1507 int namelen;
3de4586c 1508 int ret = 0;
f46b5a66 1509
a874a63e
LB		 1510 ret = mnt_want_write_file(file);
1511 if (ret)
1512 goto out;
1513
72fd032e
SW		 1514 namelen = strlen(name);
1515 if (strchr(name, '/')) {
f46b5a66 1516 ret = -EINVAL;
a874a63e 1517 goto out_drop_write;
f46b5a66
CH		 1518 }
1519
16780cab
CM		 1520 if (name[0] == '.' &&
1521 (namelen == 1 || (name[1] == '.' && namelen == 2))) {
1522 ret = -EEXIST;
a874a63e 1523 goto out_drop_write;
16780cab
CM		 1524 }
1525
3de4586c 1526 if (subvol) {
72fd032e 1527 ret = btrfs_mksubvol(&file->f_path, name, namelen,
6f72c7e2 1528 NULL, transid, readonly, inherit);
cb8e7090 1529 } else {
2903ff01 1530 struct fd src = fdget(fd);
3de4586c 1531 struct inode *src_inode;
2903ff01 1532 if (!src.file) {
3de4586c 1533 ret = -EINVAL;
a874a63e 1534 goto out_drop_write;
3de4586c
CM		 1535 }
1536
496ad9aa
AV		 1537 src_inode = file_inode(src.file);
1538 if (src_inode->i_sb != file_inode(file)->i_sb) {
d397712b
CM		 1539 printk(KERN_INFO "btrfs: Snapshot src from "
1540 "another FS\n");
3de4586c 1541 ret = -EINVAL;
ecd18815
AV		 1542 } else {
1543 ret = btrfs_mksubvol(&file->f_path, name, namelen,
1544 BTRFS_I(src_inode)->root,
1545 transid, readonly, inherit);
3de4586c 1546 }
2903ff01 1547 fdput(src);
cb8e7090 1548 }
a874a63e
LB		 1549out_drop_write:
1550 mnt_drop_write_file(file);
f46b5a66 1551out:
72fd032e
SW		 1552 return ret;
1553}
1554
1555static noinline int btrfs_ioctl_snap_create(struct file *file,
fa0d2b9b 1556 void __user *arg, int subvol)
72fd032e 1557{
fa0d2b9b 1558 struct btrfs_ioctl_vol_args *vol_args;
72fd032e
SW		 1559 int ret;
1560
fa0d2b9b
LZ		 1561 vol_args = memdup_user(arg, sizeof(*vol_args));
1562 if (IS_ERR(vol_args))
1563 return PTR_ERR(vol_args);
1564 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
72fd032e 1565
fa0d2b9b 1566 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
b83cc969 1567 vol_args->fd, subvol,
6f72c7e2 1568 NULL, false, NULL);
fdfb1e4f 1569
fa0d2b9b
LZ		 1570 kfree(vol_args);
1571 return ret;
1572}
fdfb1e4f 1573
fa0d2b9b
LZ		 1574static noinline int btrfs_ioctl_snap_create_v2(struct file *file,
1575 void __user *arg, int subvol)
1576{
1577 struct btrfs_ioctl_vol_args_v2 *vol_args;
1578 int ret;
1579 u64 transid = 0;
1580 u64 *ptr = NULL;
b83cc969 1581 bool readonly = false;
6f72c7e2 1582 struct btrfs_qgroup_inherit *inherit = NULL;
75eaa0e2 1583
fa0d2b9b
LZ		 1584 vol_args = memdup_user(arg, sizeof(*vol_args));
1585 if (IS_ERR(vol_args))
1586 return PTR_ERR(vol_args);
1587 vol_args->name[BTRFS_SUBVOL_NAME_MAX] = '\0';
75eaa0e2 1588
b83cc969 1589 if (vol_args->flags &
6f72c7e2
AJ		 1590 ~(BTRFS_SUBVOL_CREATE_ASYNC | BTRFS_SUBVOL_RDONLY |
1591 BTRFS_SUBVOL_QGROUP_INHERIT)) {
b83cc969 1592 ret = -EOPNOTSUPP;
fa0d2b9b 1593 goto out;
72fd032e 1594 }
fa0d2b9b
LZ		 1595
1596 if (vol_args->flags & BTRFS_SUBVOL_CREATE_ASYNC)
1597 ptr = &transid;
b83cc969
LZ		 1598 if (vol_args->flags & BTRFS_SUBVOL_RDONLY)
1599 readonly = true;
6f72c7e2
AJ		 1600 if (vol_args->flags & BTRFS_SUBVOL_QGROUP_INHERIT) {
1601 if (vol_args->size > PAGE_CACHE_SIZE) {
1602 ret = -EINVAL;
1603 goto out;
1604 }
1605 inherit = memdup_user(vol_args->qgroup_inherit, vol_args->size);
1606 if (IS_ERR(inherit)) {
1607 ret = PTR_ERR(inherit);
1608 goto out;
1609 }
1610 }
fa0d2b9b
LZ		 1611
1612 ret = btrfs_ioctl_snap_create_transid(file, vol_args->name,
6f72c7e2 1613 vol_args->fd, subvol, ptr,
8696c533 1614 readonly, inherit);
fa0d2b9b
LZ		 1615
1616 if (ret == 0 && ptr &&
1617 copy_to_user(arg +
1618 offsetof(struct btrfs_ioctl_vol_args_v2,
1619 transid), ptr, sizeof(*ptr)))
1620 ret = -EFAULT;
fdfb1e4f 1621out:
f46b5a66 1622 kfree(vol_args);
6f72c7e2 1623 kfree(inherit);
f46b5a66
CH		 1624 return ret;
1625}
1626
0caa102d
LZ		 1627static noinline int btrfs_ioctl_subvol_getflags(struct file *file,
1628 void __user *arg)
1629{
496ad9aa 1630 struct inode *inode = file_inode(file);
0caa102d
LZ		 1631 struct btrfs_root *root = BTRFS_I(inode)->root;
1632 int ret = 0;
1633 u64 flags = 0;
1634
33345d01 1635 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID)
0caa102d
LZ		 1636 return -EINVAL;
1637
1638 down_read(&root->fs_info->subvol_sem);
1639 if (btrfs_root_readonly(root))
1640 flags |= BTRFS_SUBVOL_RDONLY;
1641 up_read(&root->fs_info->subvol_sem);
1642
1643 if (copy_to_user(arg, &flags, sizeof(flags)))
1644 ret = -EFAULT;
1645
1646 return ret;
1647}
1648
1649static noinline int btrfs_ioctl_subvol_setflags(struct file *file,
1650 void __user *arg)
1651{
496ad9aa 1652 struct inode *inode = file_inode(file);
0caa102d
LZ		 1653 struct btrfs_root *root = BTRFS_I(inode)->root;
1654 struct btrfs_trans_handle *trans;
1655 u64 root_flags;
1656 u64 flags;
1657 int ret = 0;
1658
b9ca0664
LB		 1659 ret = mnt_want_write_file(file);
1660 if (ret)
1661 goto out;
0caa102d 1662
b9ca0664
LB		 1663 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
1664 ret = -EINVAL;
1665 goto out_drop_write;
1666 }
0caa102d 1667
b9ca0664
LB		 1668 if (copy_from_user(&flags, arg, sizeof(flags))) {
1669 ret = -EFAULT;
1670 goto out_drop_write;
1671 }
0caa102d 1672
b9ca0664
LB		 1673 if (flags & BTRFS_SUBVOL_CREATE_ASYNC) {
1674 ret = -EINVAL;
1675 goto out_drop_write;
1676 }
0caa102d 1677
b9ca0664
LB		 1678 if (flags & ~BTRFS_SUBVOL_RDONLY) {
1679 ret = -EOPNOTSUPP;
1680 goto out_drop_write;
1681 }
0caa102d 1682
b9ca0664
LB		 1683 if (!inode_owner_or_capable(inode)) {
1684 ret = -EACCES;
1685 goto out_drop_write;
1686 }
b4dc2b8c 1687
0caa102d
LZ		 1688 down_write(&root->fs_info->subvol_sem);
1689
1690 /* nothing to do */
1691 if (!!(flags & BTRFS_SUBVOL_RDONLY) == btrfs_root_readonly(root))
b9ca0664 1692 goto out_drop_sem;
0caa102d
LZ		 1693
1694 root_flags = btrfs_root_flags(&root->root_item);
1695 if (flags & BTRFS_SUBVOL_RDONLY)
1696 btrfs_set_root_flags(&root->root_item,
1697 root_flags | BTRFS_ROOT_SUBVOL_RDONLY);
1698 else
1699 btrfs_set_root_flags(&root->root_item,
1700 root_flags & ~BTRFS_ROOT_SUBVOL_RDONLY);
1701
1702 trans = btrfs_start_transaction(root, 1);
1703 if (IS_ERR(trans)) {
1704 ret = PTR_ERR(trans);
1705 goto out_reset;
1706 }
1707
b4dc2b8c 1708 ret = btrfs_update_root(trans, root->fs_info->tree_root,
0caa102d
LZ		 1709 &root->root_key, &root->root_item);
1710
1711 btrfs_commit_transaction(trans, root);
1712out_reset:
1713 if (ret)
1714 btrfs_set_root_flags(&root->root_item, root_flags);
b9ca0664 1715out_drop_sem:
0caa102d 1716 up_write(&root->fs_info->subvol_sem);
b9ca0664
LB		 1717out_drop_write:
1718 mnt_drop_write_file(file);
1719out:
0caa102d
LZ		 1720 return ret;
1721}
1722
76dda93c
YZ		 1723/*
1724 * helper to check if the subvolume references other subvolumes
1725 */
1726static noinline int may_destroy_subvol(struct btrfs_root *root)
1727{
1728 struct btrfs_path *path;
175a2b87 1729 struct btrfs_dir_item *di;
76dda93c 1730 struct btrfs_key key;
175a2b87 1731 u64 dir_id;
76dda93c
YZ		 1732 int ret;
1733
1734 path = btrfs_alloc_path();
1735 if (!path)
1736 return -ENOMEM;
1737
175a2b87
JB		 1738 /* Make sure this root isn't set as the default subvol */
1739 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
1740 di = btrfs_lookup_dir_item(NULL, root->fs_info->tree_root, path,
1741 dir_id, "default", 7, 0);
1742 if (di && !IS_ERR(di)) {
1743 btrfs_dir_item_key_to_cpu(path->nodes[0], di, &key);
1744 if (key.objectid == root->root_key.objectid) {
1745 ret = -ENOTEMPTY;
1746 goto out;
1747 }
1748 btrfs_release_path(path);
1749 }
1750
76dda93c
YZ		 1751 key.objectid = root->root_key.objectid;
1752 key.type = BTRFS_ROOT_REF_KEY;
1753 key.offset = (u64)-1;
1754
1755 ret = btrfs_search_slot(NULL, root->fs_info->tree_root,
1756 &key, path, 0, 0);
1757 if (ret < 0)
1758 goto out;
1759 BUG_ON(ret == 0);
1760
1761 ret = 0;
1762 if (path->slots[0] > 0) {
1763 path->slots[0]--;
1764 btrfs_item_key_to_cpu(path->nodes[0], &key, path->slots[0]);
1765 if (key.objectid == root->root_key.objectid &&
1766 key.type == BTRFS_ROOT_REF_KEY)
1767 ret = -ENOTEMPTY;
1768 }
1769out:
1770 btrfs_free_path(path);
1771 return ret;
1772}
1773
ac8e9819
CM		 1774static noinline int key_in_sk(struct btrfs_key *key,
1775 struct btrfs_ioctl_search_key *sk)
1776{
abc6e134
CM		 1777 struct btrfs_key test;
1778 int ret;
1779
1780 test.objectid = sk->min_objectid;
1781 test.type = sk->min_type;
1782 test.offset = sk->min_offset;
1783
1784 ret = btrfs_comp_cpu_keys(key, &test);
1785 if (ret < 0)
ac8e9819 1786 return 0;
abc6e134
CM		 1787
1788 test.objectid = sk->max_objectid;
1789 test.type = sk->max_type;
1790 test.offset = sk->max_offset;
1791
1792 ret = btrfs_comp_cpu_keys(key, &test);
1793 if (ret > 0)
ac8e9819
CM		 1794 return 0;
1795 return 1;
1796}
1797
1798static noinline int copy_to_sk(struct btrfs_root *root,
1799 struct btrfs_path *path,
1800 struct btrfs_key *key,
1801 struct btrfs_ioctl_search_key *sk,
1802 char *buf,
1803 unsigned long *sk_offset,
1804 int *num_found)
1805{
1806 u64 found_transid;
1807 struct extent_buffer *leaf;
1808 struct btrfs_ioctl_search_header sh;
1809 unsigned long item_off;
1810 unsigned long item_len;
1811 int nritems;
1812 int i;
1813 int slot;
ac8e9819
CM		 1814 int ret = 0;
1815
1816 leaf = path->nodes[0];
1817 slot = path->slots[0];
1818 nritems = btrfs_header_nritems(leaf);
1819
1820 if (btrfs_header_generation(leaf) > sk->max_transid) {
1821 i = nritems;
1822 goto advance_key;
1823 }
1824 found_transid = btrfs_header_generation(leaf);
1825
1826 for (i = slot; i < nritems; i++) {
1827 item_off = btrfs_item_ptr_offset(leaf, i);
1828 item_len = btrfs_item_size_nr(leaf, i);
1829
03b71c6c
GP		 1830 btrfs_item_key_to_cpu(leaf, key, i);
1831 if (!key_in_sk(key, sk))
1832 continue;
1833
1834 if (sizeof(sh) + item_len > BTRFS_SEARCH_ARGS_BUFSIZE)
ac8e9819
CM		 1835 item_len = 0;
1836
1837 if (sizeof(sh) + item_len + *sk_offset >
1838 BTRFS_SEARCH_ARGS_BUFSIZE) {
1839 ret = 1;
1840 goto overflow;
1841 }
1842
ac8e9819
CM		 1843 sh.objectid = key->objectid;
1844 sh.offset = key->offset;
1845 sh.type = key->type;
1846 sh.len = item_len;
1847 sh.transid = found_transid;
1848
1849 /* copy search result header */
1850 memcpy(buf + *sk_offset, &sh, sizeof(sh));
1851 *sk_offset += sizeof(sh);
1852
1853 if (item_len) {
1854 char *p = buf + *sk_offset;
1855 /* copy the item */
1856 read_extent_buffer(leaf, p,
1857 item_off, item_len);
1858 *sk_offset += item_len;
ac8e9819 1859 }
e2156867 1860 (*num_found)++;
ac8e9819
CM		 1861
1862 if (*num_found >= sk->nr_items)
1863 break;
1864 }
1865advance_key:
abc6e134
CM		 1866 ret = 0;
1867 if (key->offset < (u64)-1 && key->offset < sk->max_offset)
ac8e9819 1868 key->offset++;
abc6e134
CM		 1869 else if (key->type < (u8)-1 && key->type < sk->max_type) {
1870 key->offset = 0;
ac8e9819 1871 key->type++;
abc6e134
CM		 1872 } else if (key->objectid < (u64)-1 && key->objectid < sk->max_objectid) {
1873 key->offset = 0;
1874 key->type = 0;
ac8e9819 1875 key->objectid++;
abc6e134
CM		 1876 } else
1877 ret = 1;
ac8e9819 1878overflow:
ac8e9819
CM		 1879 return ret;
1880}
1881
1882static noinline int search_ioctl(struct inode *inode,
1883 struct btrfs_ioctl_search_args *args)
1884{
1885 struct btrfs_root *root;
1886 struct btrfs_key key;
1887 struct btrfs_key max_key;
1888 struct btrfs_path *path;
1889 struct btrfs_ioctl_search_key *sk = &args->key;
1890 struct btrfs_fs_info *info = BTRFS_I(inode)->root->fs_info;
1891 int ret;
1892 int num_found = 0;
1893 unsigned long sk_offset = 0;
1894
1895 path = btrfs_alloc_path();
1896 if (!path)
1897 return -ENOMEM;
1898
1899 if (sk->tree_id == 0) {
1900 /* search the root of the inode that was passed */
1901 root = BTRFS_I(inode)->root;
1902 } else {
1903 key.objectid = sk->tree_id;
1904 key.type = BTRFS_ROOT_ITEM_KEY;
1905 key.offset = (u64)-1;
1906 root = btrfs_read_fs_root_no_name(info, &key);
1907 if (IS_ERR(root)) {
1908 printk(KERN_ERR "could not find root %llu\n",
1909 sk->tree_id);
1910 btrfs_free_path(path);
1911 return -ENOENT;
1912 }
1913 }
1914
1915 key.objectid = sk->min_objectid;
1916 key.type = sk->min_type;
1917 key.offset = sk->min_offset;
1918
1919 max_key.objectid = sk->max_objectid;
1920 max_key.type = sk->max_type;
1921 max_key.offset = sk->max_offset;
1922
1923 path->keep_locks = 1;
1924
1925 while(1) {
de78b51a 1926 ret = btrfs_search_forward(root, &key, &max_key, path,
ac8e9819
CM		 1927 sk->min_transid);
1928 if (ret != 0) {
1929 if (ret > 0)
1930 ret = 0;
1931 goto err;
1932 }
1933 ret = copy_to_sk(root, path, &key, sk, args->buf,
1934 &sk_offset, &num_found);
b3b4aa74 1935 btrfs_release_path(path);
ac8e9819
CM		 1936 if (ret || num_found >= sk->nr_items)
1937 break;
1938
1939 }
1940 ret = 0;
1941err:
1942 sk->nr_items = num_found;
1943 btrfs_free_path(path);
1944 return ret;
1945}
1946
1947static noinline int btrfs_ioctl_tree_search(struct file *file,
1948 void __user *argp)
1949{
1950 struct btrfs_ioctl_search_args *args;
1951 struct inode *inode;
1952 int ret;
1953
1954 if (!capable(CAP_SYS_ADMIN))
1955 return -EPERM;
1956
2354d08f
JL		 1957 args = memdup_user(argp, sizeof(*args));
1958 if (IS_ERR(args))
1959 return PTR_ERR(args);
ac8e9819 1960
496ad9aa 1961 inode = file_inode(file);
ac8e9819
CM		 1962 ret = search_ioctl(inode, args);
1963 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
1964 ret = -EFAULT;
1965 kfree(args);
1966 return ret;
1967}
1968
98d377a0 1969/*
ac8e9819
CM		 1970 * Search INODE_REFs to identify path name of 'dirid' directory
1971 * in a 'tree_id' tree. and sets path name to 'name'.
1972 */
98d377a0
TH		 1973static noinline int btrfs_search_path_in_tree(struct btrfs_fs_info *info,
1974 u64 tree_id, u64 dirid, char *name)
1975{
1976 struct btrfs_root *root;
1977 struct btrfs_key key;
ac8e9819 1978 char *ptr;
98d377a0
TH		 1979 int ret = -1;
1980 int slot;
1981 int len;
1982 int total_len = 0;
1983 struct btrfs_inode_ref *iref;
1984 struct extent_buffer *l;
1985 struct btrfs_path *path;
1986
1987 if (dirid == BTRFS_FIRST_FREE_OBJECTID) {
1988 name[0]='\0';
1989 return 0;
1990 }
1991
1992 path = btrfs_alloc_path();
1993 if (!path)
1994 return -ENOMEM;
1995
ac8e9819 1996 ptr = &name[BTRFS_INO_LOOKUP_PATH_MAX];
98d377a0
TH		 1997
1998 key.objectid = tree_id;
1999 key.type = BTRFS_ROOT_ITEM_KEY;
2000 key.offset = (u64)-1;
2001 root = btrfs_read_fs_root_no_name(info, &key);
2002 if (IS_ERR(root)) {
2003 printk(KERN_ERR "could not find root %llu\n", tree_id);
8ad6fcab
CM		 2004 ret = -ENOENT;
2005 goto out;
98d377a0
TH		 2006 }
2007
2008 key.objectid = dirid;
2009 key.type = BTRFS_INODE_REF_KEY;
8ad6fcab 2010 key.offset = (u64)-1;
98d377a0
TH		 2011
2012 while(1) {
2013 ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
2014 if (ret < 0)
2015 goto out;
18674c6c
FDBM		 2016 else if (ret > 0) {
2017 ret = btrfs_previous_item(root, path, dirid,
2018 BTRFS_INODE_REF_KEY);
2019 if (ret < 0)
2020 goto out;
2021 else if (ret > 0) {
2022 ret = -ENOENT;
2023 goto out;
2024 }
2025 }
98d377a0
TH		 2026
2027 l = path->nodes[0];
2028 slot = path->slots[0];
2029 btrfs_item_key_to_cpu(l, &key, slot);
2030
98d377a0
TH		 2031 iref = btrfs_item_ptr(l, slot, struct btrfs_inode_ref);
2032 len = btrfs_inode_ref_name_len(l, iref);
2033 ptr -= len + 1;
2034 total_len += len + 1;
a696cf35
FDBM		 2035 if (ptr < name) {
2036 ret = -ENAMETOOLONG;
98d377a0 2037 goto out;
a696cf35 2038 }
98d377a0
TH		 2039
2040 *(ptr + len) = '/';
2041 read_extent_buffer(l, ptr,(unsigned long)(iref + 1), len);
2042
2043 if (key.offset == BTRFS_FIRST_FREE_OBJECTID)
2044 break;
2045
b3b4aa74 2046 btrfs_release_path(path);
98d377a0 2047 key.objectid = key.offset;
8ad6fcab 2048 key.offset = (u64)-1;
98d377a0 2049 dirid = key.objectid;
98d377a0 2050 }
77906a50 2051 memmove(name, ptr, total_len);
98d377a0
TH		 2052 name[total_len]='\0';
2053 ret = 0;
2054out:
2055 btrfs_free_path(path);
ac8e9819
CM		 2056 return ret;
2057}
2058
2059static noinline int btrfs_ioctl_ino_lookup(struct file *file,
2060 void __user *argp)
2061{
2062 struct btrfs_ioctl_ino_lookup_args *args;
2063 struct inode *inode;
2064 int ret;
2065
2066 if (!capable(CAP_SYS_ADMIN))
2067 return -EPERM;
2068
2354d08f
JL		 2069 args = memdup_user(argp, sizeof(*args));
2070 if (IS_ERR(args))
2071 return PTR_ERR(args);
c2b96929 2072
496ad9aa 2073 inode = file_inode(file);
ac8e9819 2074
1b53ac4d
CM		 2075 if (args->treeid == 0)
2076 args->treeid = BTRFS_I(inode)->root->root_key.objectid;
2077
ac8e9819
CM		 2078 ret = btrfs_search_path_in_tree(BTRFS_I(inode)->root->fs_info,
2079 args->treeid, args->objectid,
2080 args->name);
2081
2082 if (ret == 0 && copy_to_user(argp, args, sizeof(*args)))
2083 ret = -EFAULT;
2084
2085 kfree(args);
98d377a0
TH		 2086 return ret;
2087}
2088
76dda93c
YZ		 2089static noinline int btrfs_ioctl_snap_destroy(struct file *file,
2090 void __user *arg)
2091{
2092 struct dentry *parent = fdentry(file);
2093 struct dentry *dentry;
2094 struct inode *dir = parent->d_inode;
2095 struct inode *inode;
2096 struct btrfs_root *root = BTRFS_I(dir)->root;
2097 struct btrfs_root *dest = NULL;
2098 struct btrfs_ioctl_vol_args *vol_args;
2099 struct btrfs_trans_handle *trans;
c58aaad2
MX		 2100 struct btrfs_block_rsv block_rsv;
2101 u64 qgroup_reserved;
76dda93c
YZ		 2102 int namelen;
2103 int ret;
2104 int err = 0;
2105
76dda93c
YZ		 2106 vol_args = memdup_user(arg, sizeof(*vol_args));
2107 if (IS_ERR(vol_args))
2108 return PTR_ERR(vol_args);
2109
2110 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
2111 namelen = strlen(vol_args->name);
2112 if (strchr(vol_args->name, '/') ||
2113 strncmp(vol_args->name, "..", namelen) == 0) {
2114 err = -EINVAL;
2115 goto out;
2116 }
2117
a561be71 2118 err = mnt_want_write_file(file);
76dda93c
YZ		 2119 if (err)
2120 goto out;
2121
5c50c9b8
DS		 2122 err = mutex_lock_killable_nested(&dir->i_mutex, I_MUTEX_PARENT);
2123 if (err == -EINTR)
2124 goto out;
76dda93c
YZ		 2125 dentry = lookup_one_len(vol_args->name, parent, namelen);
2126 if (IS_ERR(dentry)) {
2127 err = PTR_ERR(dentry);
2128 goto out_unlock_dir;
2129 }
2130
2131 if (!dentry->d_inode) {
2132 err = -ENOENT;
2133 goto out_dput;
2134 }
2135
2136 inode = dentry->d_inode;
4260f7c7
SW		 2137 dest = BTRFS_I(inode)->root;
2138 if (!capable(CAP_SYS_ADMIN)){
2139 /*
2140 * Regular user. Only allow this with a special mount
2141 * option, when the user has write+exec access to the
2142 * subvol root, and when rmdir(2) would have been
2143 * allowed.
2144 *
2145 * Note that this is _not_ check that the subvol is
2146 * empty or doesn't contain data that we wouldn't
2147 * otherwise be able to delete.
2148 *
2149 * Users who want to delete empty subvols should try
2150 * rmdir(2).
2151 */
2152 err = -EPERM;
2153 if (!btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
2154 goto out_dput;
2155
2156 /*
2157 * Do not allow deletion if the parent dir is the same
2158 * as the dir to be deleted. That means the ioctl
2159 * must be called on the dentry referencing the root
2160 * of the subvol, not a random directory contained
2161 * within it.
2162 */
2163 err = -EINVAL;
2164 if (root == dest)
2165 goto out_dput;
2166
2167 err = inode_permission(inode, MAY_WRITE | MAY_EXEC);
2168 if (err)
2169 goto out_dput;
4260f7c7
SW		 2170 }
2171
5c39da5b
MX		 2172 /* check if subvolume may be deleted by a user */
2173 err = btrfs_may_delete(dir, dentry, 1);
2174 if (err)
2175 goto out_dput;
2176
33345d01 2177 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
76dda93c
YZ		 2178 err = -EINVAL;
2179 goto out_dput;
2180 }
2181
76dda93c
YZ		 2182 mutex_lock(&inode->i_mutex);
2183 err = d_invalidate(dentry);
2184 if (err)
2185 goto out_unlock;
2186
2187 down_write(&root->fs_info->subvol_sem);
2188
2189 err = may_destroy_subvol(dest);
2190 if (err)
2191 goto out_up_write;
2192
c58aaad2
MX		 2193 btrfs_init_block_rsv(&block_rsv, BTRFS_BLOCK_RSV_TEMP);
2194 /*
2195 * One for dir inode, two for dir entries, two for root
2196 * ref/backref.
2197 */
2198 err = btrfs_subvolume_reserve_metadata(root, &block_rsv,
ee3441b4 2199 5, &qgroup_reserved, true);
c58aaad2
MX		 2200 if (err)
2201 goto out_up_write;
2202
a22285a6
YZ		 2203 trans = btrfs_start_transaction(root, 0);
2204 if (IS_ERR(trans)) {
2205 err = PTR_ERR(trans);
c58aaad2 2206 goto out_release;
a22285a6 2207 }
c58aaad2
MX		 2208 trans->block_rsv = &block_rsv;
2209 trans->bytes_reserved = block_rsv.size;
a22285a6 2210
76dda93c
YZ		 2211 ret = btrfs_unlink_subvol(trans, root, dir,
2212 dest->root_key.objectid,
2213 dentry->d_name.name,
2214 dentry->d_name.len);
79787eaa
JM		 2215 if (ret) {
2216 err = ret;
2217 btrfs_abort_transaction(trans, root, ret);
2218 goto out_end_trans;
2219 }
76dda93c
YZ		 2220
2221 btrfs_record_root_in_trans(trans, dest);
2222
2223 memset(&dest->root_item.drop_progress, 0,
2224 sizeof(dest->root_item.drop_progress));
2225 dest->root_item.drop_level = 0;
2226 btrfs_set_root_refs(&dest->root_item, 0);
2227
d68fc57b
YZ		 2228 if (!xchg(&dest->orphan_item_inserted, 1)) {
2229 ret = btrfs_insert_orphan_item(trans,
2230 root->fs_info->tree_root,
2231 dest->root_key.objectid);
79787eaa
JM		 2232 if (ret) {
2233 btrfs_abort_transaction(trans, root, ret);
2234 err = ret;
2235 goto out_end_trans;
2236 }
d68fc57b 2237 }
79787eaa 2238out_end_trans:
c58aaad2
MX		 2239 trans->block_rsv = NULL;
2240 trans->bytes_reserved = 0;
531cb13f 2241 ret = btrfs_end_transaction(trans, root);
79787eaa
JM		 2242 if (ret && !err)
2243 err = ret;
76dda93c 2244 inode->i_flags |= S_DEAD;
c58aaad2
MX		 2245out_release:
2246 btrfs_subvolume_release_metadata(root, &block_rsv, qgroup_reserved);
76dda93c
YZ		 2247out_up_write:
2248 up_write(&root->fs_info->subvol_sem);
2249out_unlock:
2250 mutex_unlock(&inode->i_mutex);
2251 if (!err) {
efefb143 2252 shrink_dcache_sb(root->fs_info->sb);
76dda93c
YZ		 2253 btrfs_invalidate_inodes(dest);
2254 d_delete(dentry);
fa6ac876
LB		 2255
2256 /* the last ref */
2257 if (dest->cache_inode) {
2258 iput(dest->cache_inode);
2259 dest->cache_inode = NULL;
2260 }
76dda93c
YZ		 2261 }
2262out_dput:
2263 dput(dentry);
2264out_unlock_dir:
2265 mutex_unlock(&dir->i_mutex);
2a79f17e 2266 mnt_drop_write_file(file);
76dda93c
YZ		 2267out:
2268 kfree(vol_args);
2269 return err;
2270}
2271
1e701a32 2272static int btrfs_ioctl_defrag(struct file *file, void __user *argp)
f46b5a66 2273{
496ad9aa 2274 struct inode *inode = file_inode(file);
f46b5a66 2275 struct btrfs_root *root = BTRFS_I(inode)->root;
1e701a32 2276 struct btrfs_ioctl_defrag_range_args *range;
c146afad
YZ		 2277 int ret;
2278
25122d15
ID		 2279 ret = mnt_want_write_file(file);
2280 if (ret)
2281 return ret;
b83cc969 2282
25122d15
ID		 2283 if (btrfs_root_readonly(root)) {
2284 ret = -EROFS;
2285 goto out;
5ac00add 2286 }
f46b5a66
CH		 2287
2288 switch (inode->i_mode & S_IFMT) {
2289 case S_IFDIR:
e441d54d
CM		 2290 if (!capable(CAP_SYS_ADMIN)) {
2291 ret = -EPERM;
2292 goto out;
2293 }
de78b51a 2294 ret = btrfs_defrag_root(root);
8929ecfa
YZ		 2295 if (ret)
2296 goto out;
de78b51a 2297 ret = btrfs_defrag_root(root->fs_info->extent_root);
f46b5a66
CH		 2298 break;
2299 case S_IFREG:
e441d54d
CM		 2300 if (!(file->f_mode & FMODE_WRITE)) {
2301 ret = -EINVAL;
2302 goto out;
2303 }
1e701a32
CM		 2304
2305 range = kzalloc(sizeof(*range), GFP_KERNEL);
2306 if (!range) {
2307 ret = -ENOMEM;
2308 goto out;
2309 }
2310
2311 if (argp) {
2312 if (copy_from_user(range, argp,
2313 sizeof(*range))) {
2314 ret = -EFAULT;
2315 kfree(range);
683be16e 2316 goto out;
1e701a32
CM		 2317 }
2318 /* compression requires us to start the IO */
2319 if ((range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)) {
2320 range->flags |= BTRFS_DEFRAG_RANGE_START_IO;
2321 range->extent_thresh = (u32)-1;
2322 }
2323 } else {
2324 /* the rest are all set to zero by kzalloc */
2325 range->len = (u64)-1;
2326 }
496ad9aa 2327 ret = btrfs_defrag_file(file_inode(file), file,
4cb5300b
CM		 2328 range, 0, 0);
2329 if (ret > 0)
2330 ret = 0;
1e701a32 2331 kfree(range);
f46b5a66 2332 break;
8929ecfa
YZ		 2333 default:
2334 ret = -EINVAL;
f46b5a66 2335 }
e441d54d 2336out:
25122d15 2337 mnt_drop_write_file(file);
e441d54d 2338 return ret;
f46b5a66
CH		 2339}
2340
b2950863 2341static long btrfs_ioctl_add_dev(struct btrfs_root *root, void __user *arg)
f46b5a66
CH		 2342{
2343 struct btrfs_ioctl_vol_args *vol_args;
2344 int ret;
2345
e441d54d
CM		 2346 if (!capable(CAP_SYS_ADMIN))
2347 return -EPERM;
2348
5ac00add
SB		 2349 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2350 1)) {
2351 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
2c0c9da0 2352 return -EINVAL;
c9e9f97b
ID		 2353 }
2354
5ac00add 2355 mutex_lock(&root->fs_info->volume_mutex);
dae7b665 2356 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 2357 if (IS_ERR(vol_args)) {
2358 ret = PTR_ERR(vol_args);
2359 goto out;
2360 }
f46b5a66 2361
5516e595 2362 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66
CH		 2363 ret = btrfs_init_new_device(root, vol_args->name);
2364
f46b5a66 2365 kfree(vol_args);
c9e9f97b
ID		 2366out:
2367 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 2368 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
f46b5a66
CH		 2369 return ret;
2370}
2371
da24927b 2372static long btrfs_ioctl_rm_dev(struct file *file, void __user *arg)
f46b5a66 2373{
496ad9aa 2374 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
f46b5a66
CH		 2375 struct btrfs_ioctl_vol_args *vol_args;
2376 int ret;
2377
e441d54d
CM		 2378 if (!capable(CAP_SYS_ADMIN))
2379 return -EPERM;
2380
da24927b
MX		 2381 ret = mnt_want_write_file(file);
2382 if (ret)
2383 return ret;
c146afad 2384
dae7b665 2385 vol_args = memdup_user(arg, sizeof(*vol_args));
c9e9f97b
ID		 2386 if (IS_ERR(vol_args)) {
2387 ret = PTR_ERR(vol_args);
2388 goto out;
2389 }
f46b5a66 2390
5516e595 2391 vol_args->name[BTRFS_PATH_NAME_MAX] = '\0';
f46b5a66 2392
183860f6
AJ		 2393 if (atomic_xchg(&root->fs_info->mutually_exclusive_operation_running,
2394 1)) {
2395 ret = BTRFS_ERROR_DEV_EXCL_RUN_IN_PROGRESS;
2396 goto out;
2397 }
2398
2399 mutex_lock(&root->fs_info->volume_mutex);
2400 ret = btrfs_rm_device(root, vol_args->name);
c9e9f97b 2401 mutex_unlock(&root->fs_info->volume_mutex);
5ac00add 2402 atomic_set(&root->fs_info->mutually_exclusive_operation_running, 0);
183860f6
AJ		 2403
2404out:
2405 kfree(vol_args);
4ac20c70 2406 mnt_drop_write_file(file);
f46b5a66
CH		 2407 return ret;
2408}
2409
475f6387
JS		 2410static long btrfs_ioctl_fs_info(struct btrfs_root *root, void __user *arg)
2411{
027ed2f0 2412 struct btrfs_ioctl_fs_info_args *fi_args;
475f6387
JS		 2413 struct btrfs_device *device;
2414 struct btrfs_device *next;
2415 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
027ed2f0 2416 int ret = 0;
475f6387
JS		 2417
2418 if (!capable(CAP_SYS_ADMIN))
2419 return -EPERM;
2420
027ed2f0
LZ		 2421 fi_args = kzalloc(sizeof(*fi_args), GFP_KERNEL);
2422 if (!fi_args)
2423 return -ENOMEM;
2424
2425 fi_args->num_devices = fs_devices->num_devices;
2426 memcpy(&fi_args->fsid, root->fs_info->fsid, sizeof(fi_args->fsid));
475f6387
JS		 2427
2428 mutex_lock(&fs_devices->device_list_mutex);
2429 list_for_each_entry_safe(device, next, &fs_devices->devices, dev_list) {
027ed2f0
LZ		 2430 if (device->devid > fi_args->max_id)
2431 fi_args->max_id = device->devid;
475f6387
JS		 2432 }
2433 mutex_unlock(&fs_devices->device_list_mutex);
2434
027ed2f0
LZ		 2435 if (copy_to_user(arg, fi_args, sizeof(*fi_args)))
2436 ret = -EFAULT;
475f6387 2437
027ed2f0
LZ		 2438 kfree(fi_args);
2439 return ret;
475f6387
JS		 2440}
2441
2442static long btrfs_ioctl_dev_info(struct btrfs_root *root, void __user *arg)
2443{
2444 struct btrfs_ioctl_dev_info_args *di_args;
2445 struct btrfs_device *dev;
2446 struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
2447 int ret = 0;
2448 char *s_uuid = NULL;
2449 char empty_uuid[BTRFS_UUID_SIZE] = {0};
2450
2451 if (!capable(CAP_SYS_ADMIN))
2452 return -EPERM;
2453
2454 di_args = memdup_user(arg, sizeof(*di_args));
2455 if (IS_ERR(di_args))
2456 return PTR_ERR(di_args);
2457
2458 if (memcmp(empty_uuid, di_args->uuid, BTRFS_UUID_SIZE) != 0)
2459 s_uuid = di_args->uuid;
2460
2461 mutex_lock(&fs_devices->device_list_mutex);
aa1b8cd4 2462 dev = btrfs_find_device(root->fs_info, di_args->devid, s_uuid, NULL);
475f6387
JS		 2463
2464 if (!dev) {
2465 ret = -ENODEV;
2466 goto out;
2467 }
2468
2469 di_args->devid = dev->devid;
2470 di_args->bytes_used = dev->bytes_used;
2471 di_args->total_bytes = dev->total_bytes;
2472 memcpy(di_args->uuid, dev->uuid, sizeof(di_args->uuid));
a27202fb 2473 if (dev->name) {
606686ee
JB		 2474 struct rcu_string *name;
2475
2476 rcu_read_lock();
2477 name = rcu_dereference(dev->name);
2478 strncpy(di_args->path, name->str, sizeof(di_args->path));
2479 rcu_read_unlock();
a27202fb
JM		 2480 di_args->path[sizeof(di_args->path) - 1] = 0;
2481 } else {
99ba55ad 2482 di_args->path[0] = '\0';
a27202fb 2483 }
475f6387
JS		 2484
2485out:
55793c0d 2486 mutex_unlock(&fs_devices->device_list_mutex);
475f6387
JS		 2487 if (ret == 0 && copy_to_user(arg, di_args, sizeof(*di_args)))
2488 ret = -EFAULT;
2489
2490 kfree(di_args);
2491 return ret;
2492}
2493
416161db
MF		 2494static struct page *extent_same_get_page(struct inode *inode, u64 off)
2495{
2496 struct page *page;
2497 pgoff_t index;
2498 struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
2499
2500 index = off >> PAGE_CACHE_SHIFT;
2501
2502 page = grab_cache_page(inode->i_mapping, index);
2503 if (!page)
2504 return NULL;
2505
2506 if (!PageUptodate(page)) {
2507 if (extent_read_full_page_nolock(tree, page, btrfs_get_extent,
2508 0))
2509 return NULL;
2510 lock_page(page);
2511 if (!PageUptodate(page)) {
2512 unlock_page(page);
2513 page_cache_release(page);
2514 return NULL;
2515 }
2516 }
2517 unlock_page(page);
2518
2519 return page;
2520}
2521
77fe20dc
MF		 2522static inline void lock_extent_range(struct inode *inode, u64 off, u64 len)
2523{
2524 /* do any pending delalloc/csum calc on src, one way or
2525 another, and lock file content */
2526 while (1) {
2527 struct btrfs_ordered_extent *ordered;
2528 lock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2529 ordered = btrfs_lookup_first_ordered_extent(inode,
2530 off + len - 1);
2531 if (!ordered &&
2532 !test_range_bit(&BTRFS_I(inode)->io_tree, off,
2533 off + len - 1, EXTENT_DELALLOC, 0, NULL))
2534 break;
2535 unlock_extent(&BTRFS_I(inode)->io_tree, off, off + len - 1);
2536 if (ordered)
2537 btrfs_put_ordered_extent(ordered);
2538 btrfs_wait_ordered_range(inode, off, len);
2539 }
2540}
2541
416161db
MF		 2542static void btrfs_double_unlock(struct inode *inode1, u64 loff1,
2543 struct inode *inode2, u64 loff2, u64 len)
2544{
2545 unlock_extent(&BTRFS_I(inode1)->io_tree, loff1, loff1 + len - 1);
2546 unlock_extent(&BTRFS_I(inode2)->io_tree, loff2, loff2 + len - 1);
2547
2548 mutex_unlock(&inode1->i_mutex);
2549 mutex_unlock(&inode2->i_mutex);
2550}
2551
2552static void btrfs_double_lock(struct inode *inode1, u64 loff1,
2553 struct inode *inode2, u64 loff2, u64 len)
2554{
2555 if (inode1 < inode2) {
2556 swap(inode1, inode2);
2557 swap(loff1, loff2);
2558 }
2559
2560 mutex_lock_nested(&inode1->i_mutex, I_MUTEX_PARENT);
2561 lock_extent_range(inode1, loff1, len);
2562 if (inode1 != inode2) {
2563 mutex_lock_nested(&inode2->i_mutex, I_MUTEX_CHILD);
2564 lock_extent_range(inode2, loff2, len);
2565 }
2566}
2567
2568static int btrfs_cmp_data(struct inode *src, u64 loff, struct inode *dst,
2569 u64 dst_loff, u64 len)
2570{
2571 int ret = 0;
2572 struct page *src_page, *dst_page;
2573 unsigned int cmp_len = PAGE_CACHE_SIZE;
2574 void *addr, *dst_addr;
2575
2576 while (len) {
2577 if (len < PAGE_CACHE_SIZE)
2578 cmp_len = len;
2579
2580 src_page = extent_same_get_page(src, loff);
2581 if (!src_page)
2582 return -EINVAL;
2583 dst_page = extent_same_get_page(dst, dst_loff);
2584 if (!dst_page) {
2585 page_cache_release(src_page);
2586 return -EINVAL;
2587 }
2588 addr = kmap_atomic(src_page);
2589 dst_addr = kmap_atomic(dst_page);
2590
2591 flush_dcache_page(src_page);
2592 flush_dcache_page(dst_page);
2593
2594 if (memcmp(addr, dst_addr, cmp_len))
2595 ret = BTRFS_SAME_DATA_DIFFERS;
2596
2597 kunmap_atomic(addr);
2598 kunmap_atomic(dst_addr);
2599 page_cache_release(src_page);
2600 page_cache_release(dst_page);
2601
2602 if (ret)
2603 break;
2604
2605 loff += cmp_len;
2606 dst_loff += cmp_len;
2607 len -= cmp_len;
2608 }
2609
2610 return ret;
2611}
2612
2613static int extent_same_check_offsets(struct inode *inode, u64 off, u64 len)
2614{
2615 u64 bs = BTRFS_I(inode)->root->fs_info->sb->s_blocksize;
2616
2617 if (off + len > inode->i_size || off + len < off)
2618 return -EINVAL;
2619 /* Check that we are block aligned - btrfs_clone() requires this */
2620 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs))
2621 return -EINVAL;
2622
2623 return 0;
2624}
2625
2626static int btrfs_extent_same(struct inode *src, u64 loff, u64 len,
2627 struct inode *dst, u64 dst_loff)
2628{
2629 int ret;
2630
2631 /*
2632 * btrfs_clone() can't handle extents in the same file
2633 * yet. Once that works, we can drop this check and replace it
2634 * with a check for the same inode, but overlapping extents.
2635 */
2636 if (src == dst)
2637 return -EINVAL;
2638
2639 btrfs_double_lock(src, loff, dst, dst_loff, len);
2640
2641 ret = extent_same_check_offsets(src, loff, len);
2642 if (ret)
2643 goto out_unlock;
2644
2645 ret = extent_same_check_offsets(dst, dst_loff, len);
2646 if (ret)
2647 goto out_unlock;
2648
2649 /* don't make the dst file partly checksummed */
2650 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
2651 (BTRFS_I(dst)->flags & BTRFS_INODE_NODATASUM)) {
2652 ret = -EINVAL;
2653 goto out_unlock;
2654 }
2655
2656 ret = btrfs_cmp_data(src, loff, dst, dst_loff, len);
2657 if (ret == 0)
2658 ret = btrfs_clone(src, dst, loff, len, len, dst_loff);
2659
2660out_unlock:
2661 btrfs_double_unlock(src, loff, dst, dst_loff, len);
2662
2663 return ret;
2664}
2665
2666#define BTRFS_MAX_DEDUPE_LEN (16 * 1024 * 1024)
2667
2668static long btrfs_ioctl_file_extent_same(struct file *file,
2669 void __user *argp)
2670{
2671 struct btrfs_ioctl_same_args *args = argp;
2672 struct btrfs_ioctl_same_args same;
2673 struct btrfs_ioctl_same_extent_info info;
2674 struct inode *src = file->f_dentry->d_inode;
2675 struct file *dst_file = NULL;
2676 struct inode *dst;
2677 u64 off;
2678 u64 len;
2679 int i;
2680 int ret;
2681 u64 bs = BTRFS_I(src)->root->fs_info->sb->s_blocksize;
2682 bool is_admin = capable(CAP_SYS_ADMIN);
2683
2684 if (!(file->f_mode & FMODE_READ))
2685 return -EINVAL;
2686
2687 ret = mnt_want_write_file(file);
2688 if (ret)
2689 return ret;
2690
2691 if (copy_from_user(&same,
2692 (struct btrfs_ioctl_same_args __user *)argp,
2693 sizeof(same))) {
2694 ret = -EFAULT;
2695 goto out;
2696 }
2697
2698 off = same.logical_offset;
2699 len = same.length;
2700
2701 /*
2702 * Limit the total length we will dedupe for each operation.
2703 * This is intended to bound the total time spent in this
2704 * ioctl to something sane.
2705 */
2706 if (len > BTRFS_MAX_DEDUPE_LEN)
2707 len = BTRFS_MAX_DEDUPE_LEN;
2708
2709 if (WARN_ON_ONCE(bs < PAGE_CACHE_SIZE)) {
2710 /*
2711 * Btrfs does not support blocksize < page_size. As a
2712 * result, btrfs_cmp_data() won't correctly handle
2713 * this situation without an update.
2714 */
2715 ret = -EINVAL;
2716 goto out;
2717 }
2718
2719 ret = -EISDIR;
2720 if (S_ISDIR(src->i_mode))
2721 goto out;
2722
2723 ret = -EACCES;
2724 if (!S_ISREG(src->i_mode))
2725 goto out;
2726
2727 ret = 0;
2728 for (i = 0; i < same.dest_count; i++) {
2729 if (copy_from_user(&info, &args->info[i], sizeof(info))) {
2730 ret = -EFAULT;
2731 goto out;
2732 }
2733
2734 info.bytes_deduped = 0;
2735
2736 dst_file = fget(info.fd);
2737 if (!dst_file) {
2738 info.status = -EBADF;
2739 goto next;
2740 }
2741
2742 if (!(is_admin || (dst_file->f_mode & FMODE_WRITE))) {
2743 info.status = -EINVAL;
2744 goto next;
2745 }
2746
2747 info.status = -EXDEV;
2748 if (file->f_path.mnt != dst_file->f_path.mnt)
2749 goto next;
2750
2751 dst = dst_file->f_dentry->d_inode;
2752 if (src->i_sb != dst->i_sb)
2753 goto next;
2754
2755 if (S_ISDIR(dst->i_mode)) {
2756 info.status = -EISDIR;
2757 goto next;
2758 }
2759
2760 if (!S_ISREG(dst->i_mode)) {
2761 info.status = -EACCES;
2762 goto next;
2763 }
2764
2765 info.status = btrfs_extent_same(src, off, len, dst,
2766 info.logical_offset);
2767 if (info.status == 0)
2768 info.bytes_deduped += len;
2769
2770next:
2771 if (dst_file)
2772 fput(dst_file);
2773
2774 if (__put_user_unaligned(info.status, &args->info[i].status) ||
2775 __put_user_unaligned(info.bytes_deduped,
2776 &args->info[i].bytes_deduped)) {
2777 ret = -EFAULT;
2778 goto out;
2779 }
2780 }
2781
2782out:
2783 mnt_drop_write_file(file);
2784 return ret;
2785}
2786
32b7c687
MF		 2787/**
2788 * btrfs_clone() - clone a range from inode file to another
2789 *
2790 * @src: Inode to clone from
2791 * @inode: Inode to clone to
2792 * @off: Offset within source to start clone from
2793 * @olen: Original length, passed by user, of range to clone
2794 * @olen_aligned: Block-aligned value of olen, extent_same uses
2795 * identical values here
2796 * @destoff: Offset within @inode to start clone
2797 */
2798static int btrfs_clone(struct inode *src, struct inode *inode,
2799 u64 off, u64 olen, u64 olen_aligned, u64 destoff)
f46b5a66 2800{
f46b5a66 2801 struct btrfs_root *root = BTRFS_I(inode)->root;
32b7c687 2802 struct btrfs_path *path = NULL;
f46b5a66 2803 struct extent_buffer *leaf;
32b7c687
MF		 2804 struct btrfs_trans_handle *trans;
2805 char *buf = NULL;
ae01a0ab 2806 struct btrfs_key key;
f46b5a66
CH		 2807 u32 nritems;
2808 int slot;
ae01a0ab 2809 int ret;
32b7c687 2810 u64 len = olen_aligned;
ae01a0ab
YZ		 2811
2812 ret = -ENOMEM;
2813 buf = vmalloc(btrfs_level_size(root, 0));
2814 if (!buf)
32b7c687 2815 return ret;
ae01a0ab
YZ		 2816
2817 path = btrfs_alloc_path();
2818 if (!path) {
2819 vfree(buf);
32b7c687 2820 return ret;
d525e8ab
LZ		 2821 }
2822
32b7c687 2823 path->reada = 2;
c5c9cd4d 2824 /* clone data */
33345d01 2825 key.objectid = btrfs_ino(src);
ae01a0ab
YZ		 2826 key.type = BTRFS_EXTENT_DATA_KEY;
2827 key.offset = 0;
f46b5a66
CH		 2828
2829 while (1) {
2830 /*
2831 * note the key will change type as we walk through the
2832 * tree.
2833 */
362a20c5
DS		 2834 ret = btrfs_search_slot(NULL, BTRFS_I(src)->root, &key, path,
2835 0, 0);
f46b5a66
CH		 2836 if (ret < 0)
2837 goto out;
2838
ae01a0ab
YZ		 2839 nritems = btrfs_header_nritems(path->nodes[0]);
2840 if (path->slots[0] >= nritems) {
362a20c5 2841 ret = btrfs_next_leaf(BTRFS_I(src)->root, path);
f46b5a66
CH		 2842 if (ret < 0)
2843 goto out;
2844 if (ret > 0)
2845 break;
ae01a0ab 2846 nritems = btrfs_header_nritems(path->nodes[0]);
f46b5a66
CH		 2847 }
2848 leaf = path->nodes[0];
2849 slot = path->slots[0];
f46b5a66 2850
ae01a0ab 2851 btrfs_item_key_to_cpu(leaf, &key, slot);
d20f7043 2852 if (btrfs_key_type(&key) > BTRFS_EXTENT_DATA_KEY ||
33345d01 2853 key.objectid != btrfs_ino(src))
f46b5a66
CH		 2854 break;
2855
c5c9cd4d
SW		 2856 if (btrfs_key_type(&key) == BTRFS_EXTENT_DATA_KEY) {
2857 struct btrfs_file_extent_item *extent;
2858 int type;
31840ae1
ZY		 2859 u32 size;
2860 struct btrfs_key new_key;
c5c9cd4d
SW		 2861 u64 disko = 0, diskl = 0;
2862 u64 datao = 0, datal = 0;
2863 u8 comp;
b5384d48 2864 u64 endoff;
31840ae1
ZY		 2865
2866 size = btrfs_item_size_nr(leaf, slot);
2867 read_extent_buffer(leaf, buf,
2868 btrfs_item_ptr_offset(leaf, slot),
2869 size);
c5c9cd4d
SW		 2870
2871 extent = btrfs_item_ptr(leaf, slot,
2872 struct btrfs_file_extent_item);
2873 comp = btrfs_file_extent_compression(leaf, extent);
2874 type = btrfs_file_extent_type(leaf, extent);
c8a894d7
CM		 2875 if (type == BTRFS_FILE_EXTENT_REG ||
2876 type == BTRFS_FILE_EXTENT_PREALLOC) {
d397712b
CM		 2877 disko = btrfs_file_extent_disk_bytenr(leaf,
2878 extent);
2879 diskl = btrfs_file_extent_disk_num_bytes(leaf,
2880 extent);
c5c9cd4d 2881 datao = btrfs_file_extent_offset(leaf, extent);
d397712b
CM		 2882 datal = btrfs_file_extent_num_bytes(leaf,
2883 extent);
c5c9cd4d
SW		 2884 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2885 /* take upper bound, may be compressed */
2886 datal = btrfs_file_extent_ram_bytes(leaf,
2887 extent);
2888 }
b3b4aa74 2889 btrfs_release_path(path);
31840ae1 2890
050006a7 2891 if (key.offset + datal <= off ||
aa42ffd9 2892 key.offset >= off + len - 1)
c5c9cd4d
SW		 2893 goto next;
2894
31840ae1 2895 memcpy(&new_key, &key, sizeof(new_key));
33345d01 2896 new_key.objectid = btrfs_ino(inode);
4d728ec7
LZ		 2897 if (off <= key.offset)
2898 new_key.offset = key.offset + destoff - off;
2899 else
2900 new_key.offset = destoff;
31840ae1 2901
b6f3409b
SW		 2902 /*
2903 * 1 - adjusting old extent (we may have to split it)
2904 * 1 - add new extent
2905 * 1 - inode update
2906 */
2907 trans = btrfs_start_transaction(root, 3);
a22285a6
YZ		 2908 if (IS_ERR(trans)) {
2909 ret = PTR_ERR(trans);
2910 goto out;
2911 }
2912
c8a894d7
CM		 2913 if (type == BTRFS_FILE_EXTENT_REG ||
2914 type == BTRFS_FILE_EXTENT_PREALLOC) {
d72c0842
LZ		 2915 /*
2916 * a | --- range to clone ---| b
2917 * | ------------- extent ------------- |
2918 */
2919
2920 /* substract range b */
2921 if (key.offset + datal > off + len)
2922 datal = off + len - key.offset;
2923
2924 /* substract range a */
a22285a6
YZ		 2925 if (off > key.offset) {
2926 datao += off - key.offset;
2927 datal -= off - key.offset;
2928 }
2929
5dc562c5 2930 ret = btrfs_drop_extents(trans, root, inode,
a22285a6
YZ		 2931 new_key.offset,
2932 new_key.offset + datal,
2671485d 2933 1);
79787eaa
JM		 2934 if (ret) {
2935 btrfs_abort_transaction(trans, root,
2936 ret);
2937 btrfs_end_transaction(trans, root);
2938 goto out;
2939 }
a22285a6 2940
c5c9cd4d
SW		 2941 ret = btrfs_insert_empty_item(trans, root, path,
2942 &new_key, size);
79787eaa
JM		 2943 if (ret) {
2944 btrfs_abort_transaction(trans, root,
2945 ret);
2946 btrfs_end_transaction(trans, root);
2947 goto out;
2948 }
c5c9cd4d
SW		 2949
2950 leaf = path->nodes[0];
2951 slot = path->slots[0];
2952 write_extent_buffer(leaf, buf,
31840ae1
ZY		 2953 btrfs_item_ptr_offset(leaf, slot),
2954 size);
ae01a0ab 2955
c5c9cd4d 2956 extent = btrfs_item_ptr(leaf, slot,
f46b5a66 2957 struct btrfs_file_extent_item);
c5c9cd4d 2958
c5c9cd4d
SW		 2959 /* disko == 0 means it's a hole */
2960 if (!disko)
2961 datao = 0;
c5c9cd4d
SW		 2962
2963 btrfs_set_file_extent_offset(leaf, extent,
2964 datao);
2965 btrfs_set_file_extent_num_bytes(leaf, extent,
2966 datal);
2967 if (disko) {
2968 inode_add_bytes(inode, datal);
ae01a0ab 2969 ret = btrfs_inc_extent_ref(trans, root,
5d4f98a2
YZ		 2970 disko, diskl, 0,
2971 root->root_key.objectid,
33345d01 2972 btrfs_ino(inode),
66d7e7f0
AJ		 2973 new_key.offset - datao,
2974 0);
79787eaa
JM		 2975 if (ret) {
2976 btrfs_abort_transaction(trans,
2977 root,
2978 ret);
2979 btrfs_end_transaction(trans,
2980 root);
2981 goto out;
2982
2983 }
f46b5a66 2984 }
c5c9cd4d
SW		 2985 } else if (type == BTRFS_FILE_EXTENT_INLINE) {
2986 u64 skip = 0;
2987 u64 trim = 0;
2988 if (off > key.offset) {
2989 skip = off - key.offset;
2990 new_key.offset += skip;
2991 }
d397712b 2992
aa42ffd9
LB		 2993 if (key.offset + datal > off + len)
2994 trim = key.offset + datal - (off + len);
d397712b 2995
c5c9cd4d 2996 if (comp && (skip || trim)) {
c5c9cd4d 2997 ret = -EINVAL;
a22285a6 2998 btrfs_end_transaction(trans, root);
c5c9cd4d
SW		 2999 goto out;
3000 }
3001 size -= skip + trim;
3002 datal -= skip + trim;
a22285a6 3003
5dc562c5 3004 ret = btrfs_drop_extents(trans, root, inode,
a22285a6
YZ		 3005 new_key.offset,
3006 new_key.offset + datal,
2671485d 3007 1);
79787eaa
JM		 3008 if (ret) {
3009 btrfs_abort_transaction(trans, root,
3010 ret);
3011 btrfs_end_transaction(trans, root);
3012 goto out;
3013 }
a22285a6 3014
c5c9cd4d
SW		 3015 ret = btrfs_insert_empty_item(trans, root, path,
3016 &new_key, size);
79787eaa
JM		 3017 if (ret) {
3018 btrfs_abort_transaction(trans, root,
3019 ret);
3020 btrfs_end_transaction(trans, root);
3021 goto out;
3022 }
c5c9cd4d
SW		 3023
3024 if (skip) {
d397712b
CM		 3025 u32 start =
3026 btrfs_file_extent_calc_inline_size(0);
c5c9cd4d
SW		 3027 memmove(buf+start, buf+start+skip,
3028 datal);
3029 }
3030
3031 leaf = path->nodes[0];
3032 slot = path->slots[0];
3033 write_extent_buffer(leaf, buf,
3034 btrfs_item_ptr_offset(leaf, slot),
3035 size);
3036 inode_add_bytes(inode, datal);
f46b5a66 3037 }
c5c9cd4d
SW		 3038
3039 btrfs_mark_buffer_dirty(leaf);
b3b4aa74 3040 btrfs_release_path(path);
c5c9cd4d 3041
0c4d2d95 3042 inode_inc_iversion(inode);
a22285a6 3043 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
b5384d48
SW		 3044
3045 /*
3046 * we round up to the block size at eof when
3047 * determining which extents to clone above,
3048 * but shouldn't round up the file size
3049 */
3050 endoff = new_key.offset + datal;
5f3888ff
LZ		 3051 if (endoff > destoff+olen)
3052 endoff = destoff+olen;
b5384d48
SW		 3053 if (endoff > inode->i_size)
3054 btrfs_i_size_write(inode, endoff);
3055
a22285a6 3056 ret = btrfs_update_inode(trans, root, inode);
79787eaa
JM		 3057 if (ret) {
3058 btrfs_abort_transaction(trans, root, ret);
3059 btrfs_end_transaction(trans, root);
3060 goto out;
3061 }
3062 ret = btrfs_end_transaction(trans, root);
a22285a6 3063 }
d397712b 3064next:
b3b4aa74 3065 btrfs_release_path(path);
f46b5a66 3066 key.offset++;
f46b5a66 3067 }
f46b5a66 3068 ret = 0;
32b7c687 3069
f46b5a66 3070out:
b3b4aa74 3071 btrfs_release_path(path);
32b7c687
MF		 3072 btrfs_free_path(path);
3073 vfree(buf);
3074 return ret;
3075}
3076
3077static noinline long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
3078 u64 off, u64 olen, u64 destoff)
3079{
3080 struct inode *inode = fdentry(file)->d_inode;
3081 struct btrfs_root *root = BTRFS_I(inode)->root;
3082 struct fd src_file;
3083 struct inode *src;
3084 int ret;
3085 u64 len = olen;
3086 u64 bs = root->fs_info->sb->s_blocksize;
3087 int same_inode = 0;
3088
3089 /*
3090 * TODO:
3091 * - split compressed inline extents. annoying: we need to
3092 * decompress into destination's address_space (the file offset
3093 * may change, so source mapping won't do), then recompress (or
3094 * otherwise reinsert) a subrange.
3095 * - allow ranges within the same file to be cloned (provided
3096 * they don't overlap)?
3097 */
3098
3099 /* the destination must be opened for writing */
3100 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND))
3101 return -EINVAL;
3102
3103 if (btrfs_root_readonly(root))
3104 return -EROFS;
3105
3106 ret = mnt_want_write_file(file);
3107 if (ret)
3108 return ret;
3109
3110 src_file = fdget(srcfd);
3111 if (!src_file.file) {
3112 ret = -EBADF;
3113 goto out_drop_write;
3114 }
3115
3116 ret = -EXDEV;
3117 if (src_file.file->f_path.mnt != file->f_path.mnt)
3118 goto out_fput;
3119
3120 src = file_inode(src_file.file);
3121
3122 ret = -EINVAL;
3123 if (src == inode)
3124 same_inode = 1;
3125
3126 /* the src must be open for reading */
3127 if (!(src_file.file->f_mode & FMODE_READ))
3128 goto out_fput;
3129
3130 /* don't make the dst file partly checksummed */
3131 if ((BTRFS_I(src)->flags & BTRFS_INODE_NODATASUM) !=
3132 (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM))
3133 goto out_fput;
3134
3135 ret = -EISDIR;
3136 if (S_ISDIR(src->i_mode) || S_ISDIR(inode->i_mode))
3137 goto out_fput;
3138
3139 ret = -EXDEV;
3140 if (src->i_sb != inode->i_sb)
3141 goto out_fput;
3142
3143 if (!same_inode) {
3144 if (inode < src) {
3145 mutex_lock_nested(&inode->i_mutex, I_MUTEX_PARENT);
3146 mutex_lock_nested(&src->i_mutex, I_MUTEX_CHILD);
3147 } else {
3148 mutex_lock_nested(&src->i_mutex, I_MUTEX_PARENT);
3149 mutex_lock_nested(&inode->i_mutex, I_MUTEX_CHILD);
3150 }
3151 } else {
3152 mutex_lock(&src->i_mutex);
3153 }
3154
3155 /* determine range to clone */
3156 ret = -EINVAL;
3157 if (off + len > src->i_size || off + len < off)
3158 goto out_unlock;
3159 if (len == 0)
3160 olen = len = src->i_size - off;
3161 /* if we extend to eof, continue to block boundary */
3162 if (off + len == src->i_size)
3163 len = ALIGN(src->i_size, bs) - off;
3164
3165 /* verify the end result is block aligned */
3166 if (!IS_ALIGNED(off, bs) || !IS_ALIGNED(off + len, bs) ||
3167 !IS_ALIGNED(destoff, bs))
3168 goto out_unlock;
3169
3170 /* verify if ranges are overlapped within the same file */
3171 if (same_inode) {
3172 if (destoff + len > off && destoff < off + len)
3173 goto out_unlock;
3174 }
3175
3176 if (destoff > inode->i_size) {
3177 ret = btrfs_cont_expand(inode, inode->i_size, destoff);
3178 if (ret)
3179 goto out_unlock;
3180 }
3181
3182 /* truncate page cache pages from target inode range */
3183 truncate_inode_pages_range(&inode->i_data, destoff,
3184 PAGE_CACHE_ALIGN(destoff + len) - 1);
3185
3186 lock_extent_range(src, off, len);
3187
3188 ret = btrfs_clone(src, inode, off, olen, len, destoff);
3189
aa42ffd9 3190 unlock_extent(&BTRFS_I(src)->io_tree, off, off + len - 1);
f46b5a66
CH		 3191out_unlock:
3192 mutex_unlock(&src->i_mutex);
a96fbc72
LB		 3193 if (!same_inode)
3194 mutex_unlock(&inode->i_mutex);
f46b5a66 3195out_fput:
2903ff01 3196 fdput(src_file);
ab67b7c1 3197out_drop_write:
2a79f17e 3198 mnt_drop_write_file(file);
f46b5a66
CH		 3199 return ret;
3200}
3201
7a865e8a 3202static long btrfs_ioctl_clone_range(struct file *file, void __user *argp)
c5c9cd4d
SW		 3203{
3204 struct btrfs_ioctl_clone_range_args args;
3205
7a865e8a 3206 if (copy_from_user(&args, argp, sizeof(args)))
c5c9cd4d
SW		 3207 return -EFAULT;
3208 return btrfs_ioctl_clone(file, args.src_fd, args.src_offset,
3209 args.src_length, args.dest_offset);
3210}
3211
f46b5a66
CH		 3212/*
3213 * there are many ways the trans_start and trans_end ioctls can lead
3214 * to deadlocks. They should only be used by applications that
3215 * basically own the machine, and have a very in depth understanding
3216 * of all the possible deadlocks and enospc problems.
3217 */
b2950863 3218static long btrfs_ioctl_trans_start(struct file *file)
f46b5a66 3219{
496ad9aa 3220 struct inode *inode = file_inode(file);
f46b5a66
CH		 3221 struct btrfs_root *root = BTRFS_I(inode)->root;
3222 struct btrfs_trans_handle *trans;
1ab86aed 3223 int ret;
f46b5a66 3224
1ab86aed 3225 ret = -EPERM;
df5b5520 3226 if (!capable(CAP_SYS_ADMIN))
1ab86aed 3227 goto out;
df5b5520 3228
1ab86aed
SW		 3229 ret = -EINPROGRESS;
3230 if (file->private_data)
f46b5a66 3231 goto out;
9ca9ee09 3232
b83cc969
LZ		 3233 ret = -EROFS;
3234 if (btrfs_root_readonly(root))
3235 goto out;
3236
a561be71 3237 ret = mnt_want_write_file(file);
c146afad
YZ		 3238 if (ret)
3239 goto out;
3240
a4abeea4 3241 atomic_inc(&root->fs_info->open_ioctl_trans);
9ca9ee09 3242
1ab86aed 3243 ret = -ENOMEM;
7a7eaa40 3244 trans = btrfs_start_ioctl_transaction(root);
abd30bb0 3245 if (IS_ERR(trans))
1ab86aed
SW		 3246 goto out_drop;
3247
3248 file->private_data = trans;
3249 return 0;
3250
3251out_drop:
a4abeea4 3252 atomic_dec(&root->fs_info->open_ioctl_trans);
2a79f17e 3253 mnt_drop_write_file(file);
f46b5a66 3254out:
f46b5a66
CH		 3255 return ret;
3256}
3257
6ef5ed0d
JB		 3258static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
3259{
496ad9aa 3260 struct inode *inode = file_inode(file);
6ef5ed0d
JB		 3261 struct btrfs_root *root = BTRFS_I(inode)->root;
3262 struct btrfs_root *new_root;
3263 struct btrfs_dir_item *di;
3264 struct btrfs_trans_handle *trans;
3265 struct btrfs_path *path;
3266 struct btrfs_key location;
3267 struct btrfs_disk_key disk_key;
6ef5ed0d
JB		 3268 u64 objectid = 0;
3269 u64 dir_id;
3c04ce01 3270 int ret;
6ef5ed0d
JB		 3271
3272 if (!capable(CAP_SYS_ADMIN))
3273 return -EPERM;
3274
3c04ce01
MX		 3275 ret = mnt_want_write_file(file);
3276 if (ret)
3277 return ret;
3278
3279 if (copy_from_user(&objectid, argp, sizeof(objectid))) {
3280 ret = -EFAULT;
3281 goto out;
3282 }
6ef5ed0d
JB		 3283
3284 if (!objectid)
3285 objectid = root->root_key.objectid;
3286
3287 location.objectid = objectid;
3288 location.type = BTRFS_ROOT_ITEM_KEY;
3289 location.offset = (u64)-1;
3290
3291 new_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
3c04ce01
MX		 3292 if (IS_ERR(new_root)) {
3293 ret = PTR_ERR(new_root);
3294 goto out;
3295 }
6ef5ed0d 3296
6ef5ed0d 3297 path = btrfs_alloc_path();
3c04ce01
MX		 3298 if (!path) {
3299 ret = -ENOMEM;
3300 goto out;
3301 }
6ef5ed0d
JB		 3302 path->leave_spinning = 1;
3303
3304 trans = btrfs_start_transaction(root, 1);
98d5dc13 3305 if (IS_ERR(trans)) {
6ef5ed0d 3306 btrfs_free_path(path);
3c04ce01
MX		 3307 ret = PTR_ERR(trans);
3308 goto out;
6ef5ed0d
JB		 3309 }
3310
6c41761f 3311 dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
6ef5ed0d
JB		 3312 di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
3313 dir_id, "default", 7, 1);
cf1e99a4 3314 if (IS_ERR_OR_NULL(di)) {
6ef5ed0d
JB		 3315 btrfs_free_path(path);
3316 btrfs_end_transaction(trans, root);
3317 printk(KERN_ERR "Umm, you don't have the default dir item, "
3318 "this isn't going to work\n");
3c04ce01
MX		 3319 ret = -ENOENT;
3320 goto out;
6ef5ed0d
JB		 3321 }
3322
3323 btrfs_cpu_key_to_disk(&disk_key, &new_root->root_key);
3324 btrfs_set_dir_item_key(path->nodes[0], di, &disk_key);
3325 btrfs_mark_buffer_dirty(path->nodes[0]);
3326 btrfs_free_path(path);
3327
2b0ce2c2 3328 btrfs_set_fs_incompat(root->fs_info, DEFAULT_SUBVOL);
6ef5ed0d 3329 btrfs_end_transaction(trans, root);
3c04ce01
MX		 3330out:
3331 mnt_drop_write_file(file);
3332 return ret;
6ef5ed0d
JB		 3333}
3334
5af3e8cc
SB		 3335void btrfs_get_block_group_info(struct list_head *groups_list,
3336 struct btrfs_ioctl_space_info *space)
bf5fc093
JB		 3337{
3338 struct btrfs_block_group_cache *block_group;
3339
3340 space->total_bytes = 0;
3341 space->used_bytes = 0;
3342 space->flags = 0;
3343 list_for_each_entry(block_group, groups_list, list) {
3344 space->flags = block_group->flags;
3345 space->total_bytes += block_group->key.offset;
3346 space->used_bytes +=
3347 btrfs_block_group_used(&block_group->item);
3348 }
3349}
3350
48a3b636 3351static long btrfs_ioctl_space_info(struct btrfs_root *root, void __user *arg)
1406e432
JB		 3352{
3353 struct btrfs_ioctl_space_args space_args;
3354 struct btrfs_ioctl_space_info space;
3355 struct btrfs_ioctl_space_info *dest;
7fde62bf 3356 struct btrfs_ioctl_space_info *dest_orig;
13f2696f 3357 struct btrfs_ioctl_space_info __user *user_dest;
1406e432 3358 struct btrfs_space_info *info;
bf5fc093
JB		 3359 u64 types[] = {BTRFS_BLOCK_GROUP_DATA,
3360 BTRFS_BLOCK_GROUP_SYSTEM,
3361 BTRFS_BLOCK_GROUP_METADATA,
3362 BTRFS_BLOCK_GROUP_DATA | BTRFS_BLOCK_GROUP_METADATA};
3363 int num_types = 4;
7fde62bf 3364 int alloc_size;
1406e432 3365 int ret = 0;
51788b1b 3366 u64 slot_count = 0;
bf5fc093 3367 int i, c;
1406e432
JB		 3368
3369 if (copy_from_user(&space_args,
3370 (struct btrfs_ioctl_space_args __user *)arg,
3371 sizeof(space_args)))
3372 return -EFAULT;
3373
bf5fc093
JB		 3374 for (i = 0; i < num_types; i++) {
3375 struct btrfs_space_info *tmp;
3376
3377 info = NULL;
3378 rcu_read_lock();
3379 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3380 list) {
3381 if (tmp->flags == types[i]) {
3382 info = tmp;
3383 break;
3384 }
3385 }
3386 rcu_read_unlock();
3387
3388 if (!info)
3389 continue;
3390
3391 down_read(&info->groups_sem);
3392 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3393 if (!list_empty(&info->block_groups[c]))
3394 slot_count++;
3395 }
3396 up_read(&info->groups_sem);
3397 }
7fde62bf
CM		 3398
3399 /* space_slots == 0 means they are asking for a count */
3400 if (space_args.space_slots == 0) {
3401 space_args.total_spaces = slot_count;
3402 goto out;
3403 }
bf5fc093 3404
51788b1b 3405 slot_count = min_t(u64, space_args.space_slots, slot_count);
bf5fc093 3406
7fde62bf 3407 alloc_size = sizeof(*dest) * slot_count;
bf5fc093 3408
7fde62bf
CM		 3409 /* we generally have at most 6 or so space infos, one for each raid
3410 * level. So, a whole page should be more than enough for everyone
3411 */
3412 if (alloc_size > PAGE_CACHE_SIZE)
3413 return -ENOMEM;
3414
1406e432 3415 space_args.total_spaces = 0;
7fde62bf
CM		 3416 dest = kmalloc(alloc_size, GFP_NOFS);
3417 if (!dest)
3418 return -ENOMEM;
3419 dest_orig = dest;
1406e432 3420
7fde62bf 3421 /* now we have a buffer to copy into */
bf5fc093
JB		 3422 for (i = 0; i < num_types; i++) {
3423 struct btrfs_space_info *tmp;
3424
51788b1b
DR		 3425 if (!slot_count)
3426 break;
3427
bf5fc093
JB		 3428 info = NULL;
3429 rcu_read_lock();
3430 list_for_each_entry_rcu(tmp, &root->fs_info->space_info,
3431 list) {
3432 if (tmp->flags == types[i]) {
3433 info = tmp;
3434 break;
3435 }
3436 }
3437 rcu_read_unlock();
7fde62bf 3438
bf5fc093
JB		 3439 if (!info)
3440 continue;
3441 down_read(&info->groups_sem);
3442 for (c = 0; c < BTRFS_NR_RAID_TYPES; c++) {
3443 if (!list_empty(&info->block_groups[c])) {
5af3e8cc
SB		 3444 btrfs_get_block_group_info(
3445 &info->block_groups[c], &space);
bf5fc093
JB		 3446 memcpy(dest, &space, sizeof(space));
3447 dest++;
3448 space_args.total_spaces++;
51788b1b 3449 slot_count--;
bf5fc093 3450 }
51788b1b
DR		 3451 if (!slot_count)
3452 break;
bf5fc093
JB		 3453 }
3454 up_read(&info->groups_sem);
1406e432 3455 }
1406e432 3456
2eec6c81 3457 user_dest = (struct btrfs_ioctl_space_info __user *)
7fde62bf
CM		 3458 (arg + sizeof(struct btrfs_ioctl_space_args));
3459
3460 if (copy_to_user(user_dest, dest_orig, alloc_size))
3461 ret = -EFAULT;
3462
3463 kfree(dest_orig);
3464out:
3465 if (ret == 0 && copy_to_user(arg, &space_args, sizeof(space_args)))
1406e432
JB		 3466 ret = -EFAULT;
3467
3468 return ret;
3469}
3470
f46b5a66
CH		 3471/*
3472 * there are many ways the trans_start and trans_end ioctls can lead
3473 * to deadlocks. They should only be used by applications that
3474 * basically own the machine, and have a very in depth understanding
3475 * of all the possible deadlocks and enospc problems.
3476 */
3477long btrfs_ioctl_trans_end(struct file *file)
3478{
496ad9aa 3479 struct inode *inode = file_inode(file);
f46b5a66
CH		 3480 struct btrfs_root *root = BTRFS_I(inode)->root;
3481 struct btrfs_trans_handle *trans;
f46b5a66 3482
f46b5a66 3483 trans = file->private_data;
1ab86aed
SW		 3484 if (!trans)
3485 return -EINVAL;
b214107e 3486 file->private_data = NULL;
9ca9ee09 3487
1ab86aed
SW		 3488 btrfs_end_transaction(trans, root);
3489
a4abeea4 3490 atomic_dec(&root->fs_info->open_ioctl_trans);
9ca9ee09 3491
2a79f17e 3492 mnt_drop_write_file(file);
1ab86aed 3493 return 0;
f46b5a66
CH		 3494}
3495
9a8c28be
MX		 3496static noinline long btrfs_ioctl_start_sync(struct btrfs_root *root,
3497 void __user *argp)
46204592 3498{
46204592
SW		 3499 struct btrfs_trans_handle *trans;
3500 u64 transid;
db5b493a 3501 int ret;
46204592 3502
d4edf39b 3503 trans = btrfs_attach_transaction_barrier(root);
ff7c1d33
MX		 3504 if (IS_ERR(trans)) {
3505 if (PTR_ERR(trans) != -ENOENT)
3506 return PTR_ERR(trans);
3507
3508 /* No running transaction, don't bother */
3509 transid = root->fs_info->last_trans_committed;
3510 goto out;
3511 }
46204592 3512 transid = trans->transid;
db5b493a 3513 ret = btrfs_commit_transaction_async(trans, root, 0);
8b2b2d3c
TI		 3514 if (ret) {
3515 btrfs_end_transaction(trans, root);
db5b493a 3516 return ret;
8b2b2d3c 3517 }
ff7c1d33 3518out:
46204592
SW		 3519 if (argp)
3520 if (copy_to_user(argp, &transid, sizeof(transid)))
3521 return -EFAULT;
3522 return 0;
3523}
3524
9a8c28be
MX		 3525static noinline long btrfs_ioctl_wait_sync(struct btrfs_root *root,
3526 void __user *argp)
46204592 3527{
46204592
SW		 3528 u64 transid;
3529
3530 if (argp) {
3531 if (copy_from_user(&transid, argp, sizeof(transid)))
3532 return -EFAULT;
3533 } else {
3534 transid = 0; /* current trans */
3535 }
3536 return btrfs_wait_for_commit(root, transid);
3537}
3538
b8e95489 3539static long btrfs_ioctl_scrub(struct file *file, void __user *arg)
475f6387 3540{
496ad9aa 3541 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
475f6387 3542 struct btrfs_ioctl_scrub_args *sa;
b8e95489 3543 int ret;
475f6387
JS		 3544
3545 if (!capable(CAP_SYS_ADMIN))
3546 return -EPERM;
3547
3548 sa = memdup_user(arg, sizeof(*sa));
3549 if (IS_ERR(sa))
3550 return PTR_ERR(sa);
3551
b8e95489
MX		 3552 if (!(sa->flags & BTRFS_SCRUB_READONLY)) {
3553 ret = mnt_want_write_file(file);
3554 if (ret)
3555 goto out;
3556 }
3557
aa1b8cd4 3558 ret = btrfs_scrub_dev(root->fs_info, sa->devid, sa->start, sa->end,
63a212ab
SB		 3559 &sa->progress, sa->flags & BTRFS_SCRUB_READONLY,
3560 0);
475f6387
JS		 3561
3562 if (copy_to_user(arg, sa, sizeof(*sa)))
3563 ret = -EFAULT;
3564
b8e95489
MX		 3565 if (!(sa->flags & BTRFS_SCRUB_READONLY))
3566 mnt_drop_write_file(file);
3567out:
475f6387
JS		 3568 kfree(sa);
3569 return ret;
3570}
3571
3572static long btrfs_ioctl_scrub_cancel(struct btrfs_root *root, void __user *arg)
3573{
3574 if (!capable(CAP_SYS_ADMIN))
3575 return -EPERM;
3576
aa1b8cd4 3577 return btrfs_scrub_cancel(root->fs_info);
475f6387
JS		 3578}
3579
3580static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
3581 void __user *arg)
3582{
3583 struct btrfs_ioctl_scrub_args *sa;
3584 int ret;
3585
3586 if (!capable(CAP_SYS_ADMIN))
3587 return -EPERM;
3588
3589 sa = memdup_user(arg, sizeof(*sa));
3590 if (IS_ERR(sa))
3591 return PTR_ERR(sa);
3592
3593 ret = btrfs_scrub_progress(root, sa->devid, &sa->progress);
3594
3595 if (copy_to_user(arg, sa, sizeof(*sa)))
3596 ret = -EFAULT;
3597
3598 kfree(sa);
3599 return ret;
3600}
3601
c11d2c23 3602static long btrfs_ioctl_get_dev_stats(struct btrfs_root *root,
b27f7c0c 3603 void __user *arg)
c11d2c23
SB		 3604{
3605 struct btrfs_ioctl_get_dev_stats *sa;
3606 int ret;
3607
c11d2c23
SB		 3608 sa = memdup_user(arg, sizeof(*sa));
3609 if (IS_ERR(sa))
3610 return PTR_ERR(sa);
3611
b27f7c0c
DS		 3612 if ((sa->flags & BTRFS_DEV_STATS_RESET) && !capable(CAP_SYS_ADMIN)) {
3613 kfree(sa);
3614 return -EPERM;
3615 }
3616
3617 ret = btrfs_get_dev_stats(root, sa);
c11d2c23
SB		 3618
3619 if (copy_to_user(arg, sa, sizeof(*sa)))
3620 ret = -EFAULT;
3621
3622 kfree(sa);
3623 return ret;
3624}
3625
3f6bcfbd
SB		 3626static long btrfs_ioctl_dev_replace(struct btrfs_root *root, void __user *arg)
3627{
3628 struct btrfs_ioctl_dev_replace_args *p;
3629 int ret;
3630
3631 if (!capable(CAP_SYS_ADMIN))
3632 return -EPERM;
3633
3634 p = memdup_user(arg, sizeof(*p));
3635 if (IS_ERR(p))
3636 return PTR_ERR(p);
3637
3638 switch (p->cmd) {
3639 case BTRFS_IOCTL_DEV_REPLACE_CMD_START:
3640 if (atomic_xchg(
3641 &root->fs_info->mutually_exclusive_operation_running,
3642 1)) {
3643 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3644 ret = -EINPROGRESS;
3645 } else {
3646 ret = btrfs_dev_replace_start(root, p);
3647 atomic_set(
3648 &root->fs_info->mutually_exclusive_operation_running,
3649 0);
3650 }
3651 break;
3652 case BTRFS_IOCTL_DEV_REPLACE_CMD_STATUS:
3653 btrfs_dev_replace_status(root->fs_info, p);
3654 ret = 0;
3655 break;
3656 case BTRFS_IOCTL_DEV_REPLACE_CMD_CANCEL:
3657 ret = btrfs_dev_replace_cancel(root->fs_info, p);
3658 break;
3659 default:
3660 ret = -EINVAL;
3661 break;
3662 }
3663
3664 if (copy_to_user(arg, p, sizeof(*p)))
3665 ret = -EFAULT;
3666
3667 kfree(p);
3668 return ret;
3669}
3670
d7728c96
JS		 3671static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
3672{
3673 int ret = 0;
3674 int i;
740c3d22 3675 u64 rel_ptr;
d7728c96 3676 int size;
806468f8 3677 struct btrfs_ioctl_ino_path_args *ipa = NULL;
d7728c96
JS		 3678 struct inode_fs_paths *ipath = NULL;
3679 struct btrfs_path *path;
3680
82b22ac8 3681 if (!capable(CAP_DAC_READ_SEARCH))
d7728c96
JS		 3682 return -EPERM;
3683
3684 path = btrfs_alloc_path();
3685 if (!path) {
3686 ret = -ENOMEM;
3687 goto out;
3688 }
3689
3690 ipa = memdup_user(arg, sizeof(*ipa));
3691 if (IS_ERR(ipa)) {
3692 ret = PTR_ERR(ipa);
3693 ipa = NULL;
3694 goto out;
3695 }
3696
3697 size = min_t(u32, ipa->size, 4096);
3698 ipath = init_ipath(size, root, path);
3699 if (IS_ERR(ipath)) {
3700 ret = PTR_ERR(ipath);
3701 ipath = NULL;
3702 goto out;
3703 }
3704
3705 ret = paths_from_inode(ipa->inum, ipath);
3706 if (ret < 0)
3707 goto out;
3708
3709 for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
745c4d8e
JM		 3710 rel_ptr = ipath->fspath->val[i] -
3711 (u64)(unsigned long)ipath->fspath->val;
740c3d22 3712 ipath->fspath->val[i] = rel_ptr;
d7728c96
JS		 3713 }
3714
745c4d8e
JM		 3715 ret = copy_to_user((void *)(unsigned long)ipa->fspath,
3716 (void *)(unsigned long)ipath->fspath, size);
d7728c96
JS		 3717 if (ret) {
3718 ret = -EFAULT;
3719 goto out;
3720 }
3721
3722out:
3723 btrfs_free_path(path);
3724 free_ipath(ipath);
3725 kfree(ipa);
3726
3727 return ret;
3728}
3729
3730static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
3731{
3732 struct btrfs_data_container *inodes = ctx;
3733 const size_t c = 3 * sizeof(u64);
3734
3735 if (inodes->bytes_left >= c) {
3736 inodes->bytes_left -= c;
3737 inodes->val[inodes->elem_cnt] = inum;
3738 inodes->val[inodes->elem_cnt + 1] = offset;
3739 inodes->val[inodes->elem_cnt + 2] = root;
3740 inodes->elem_cnt += 3;
3741 } else {
3742 inodes->bytes_missing += c - inodes->bytes_left;
3743 inodes->bytes_left = 0;
3744 inodes->elem_missed += 3;
3745 }
3746
3747 return 0;
3748}
3749
3750static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
3751 void __user *arg)
3752{
3753 int ret = 0;
3754 int size;
d7728c96
JS		 3755 struct btrfs_ioctl_logical_ino_args *loi;
3756 struct btrfs_data_container *inodes = NULL;
3757 struct btrfs_path *path = NULL;
d7728c96
JS		 3758
3759 if (!capable(CAP_SYS_ADMIN))
3760 return -EPERM;
3761
3762 loi = memdup_user(arg, sizeof(*loi));
3763 if (IS_ERR(loi)) {
3764 ret = PTR_ERR(loi);
3765 loi = NULL;
3766 goto out;
3767 }
3768
3769 path = btrfs_alloc_path();
3770 if (!path) {
3771 ret = -ENOMEM;
3772 goto out;
3773 }
3774
425d17a2 3775 size = min_t(u32, loi->size, 64 * 1024);
d7728c96
JS		 3776 inodes = init_data_container(size);
3777 if (IS_ERR(inodes)) {
3778 ret = PTR_ERR(inodes);
3779 inodes = NULL;
3780 goto out;
3781 }
3782
df031f07
LB		 3783 ret = iterate_inodes_from_logical(loi->logical, root->fs_info, path,
3784 build_ino_list, inodes);
3785 if (ret == -EINVAL)
d7728c96
JS		 3786 ret = -ENOENT;
3787 if (ret < 0)
3788 goto out;
3789
745c4d8e
JM		 3790 ret = copy_to_user((void *)(unsigned long)loi->inodes,
3791 (void *)(unsigned long)inodes, size);
d7728c96
JS		 3792 if (ret)
3793 ret = -EFAULT;
3794
3795out:
3796 btrfs_free_path(path);
425d17a2 3797 vfree(inodes);
d7728c96
JS		 3798 kfree(loi);
3799
3800 return ret;
3801}
3802
19a39dce 3803void update_ioctl_balance_args(struct btrfs_fs_info *fs_info, int lock,
c9e9f97b
ID		 3804 struct btrfs_ioctl_balance_args *bargs)
3805{
3806 struct btrfs_balance_control *bctl = fs_info->balance_ctl;
3807
3808 bargs->flags = bctl->flags;
3809
837d5b6e
ID		 3810 if (atomic_read(&fs_info->balance_running))
3811 bargs->state |= BTRFS_BALANCE_STATE_RUNNING;
3812 if (atomic_read(&fs_info->balance_pause_req))
3813 bargs->state |= BTRFS_BALANCE_STATE_PAUSE_REQ;
a7e99c69
ID		 3814 if (atomic_read(&fs_info->balance_cancel_req))
3815 bargs->state |= BTRFS_BALANCE_STATE_CANCEL_REQ;
837d5b6e 3816
c9e9f97b
ID		 3817 memcpy(&bargs->data, &bctl->data, sizeof(bargs->data));
3818 memcpy(&bargs->meta, &bctl->meta, sizeof(bargs->meta));
3819 memcpy(&bargs->sys, &bctl->sys, sizeof(bargs->sys));
19a39dce
ID		 3820
3821 if (lock) {
3822 spin_lock(&fs_info->balance_lock);
3823 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3824 spin_unlock(&fs_info->balance_lock);
3825 } else {
3826 memcpy(&bargs->stat, &bctl->stat, sizeof(bargs->stat));
3827 }
c9e9f97b
ID		 3828}
3829
9ba1f6e4 3830static long btrfs_ioctl_balance(struct file *file, void __user *arg)
c9e9f97b 3831{
496ad9aa 3832 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
c9e9f97b
ID		 3833 struct btrfs_fs_info *fs_info = root->fs_info;
3834 struct btrfs_ioctl_balance_args *bargs;
3835 struct btrfs_balance_control *bctl;
ed0fb78f 3836 bool need_unlock; /* for mut. excl. ops lock */
c9e9f97b
ID		 3837 int ret;
3838
3839 if (!capable(CAP_SYS_ADMIN))
3840 return -EPERM;
3841
e54bfa31 3842 ret = mnt_want_write_file(file);
9ba1f6e4
LB		 3843 if (ret)
3844 return ret;
3845
ed0fb78f
ID		 3846again:
3847 if (!atomic_xchg(&fs_info->mutually_exclusive_operation_running, 1)) {
3848 mutex_lock(&fs_info->volume_mutex);
3849 mutex_lock(&fs_info->balance_mutex);
3850 need_unlock = true;
3851 goto locked;
3852 }
3853
3854 /*
3855 * mut. excl. ops lock is locked. Three possibilites:
3856 * (1) some other op is running
3857 * (2) balance is running
3858 * (3) balance is paused -- special case (think resume)
3859 */
c9e9f97b 3860 mutex_lock(&fs_info->balance_mutex);
ed0fb78f
ID		 3861 if (fs_info->balance_ctl) {
3862 /* this is either (2) or (3) */
3863 if (!atomic_read(&fs_info->balance_running)) {
3864 mutex_unlock(&fs_info->balance_mutex);
3865 if (!mutex_trylock(&fs_info->volume_mutex))
3866 goto again;
3867 mutex_lock(&fs_info->balance_mutex);
3868
3869 if (fs_info->balance_ctl &&
3870 !atomic_read(&fs_info->balance_running)) {
3871 /* this is (3) */
3872 need_unlock = false;
3873 goto locked;
3874 }
3875
3876 mutex_unlock(&fs_info->balance_mutex);
3877 mutex_unlock(&fs_info->volume_mutex);
3878 goto again;
3879 } else {
3880 /* this is (2) */
3881 mutex_unlock(&fs_info->balance_mutex);
3882 ret = -EINPROGRESS;
3883 goto out;
3884 }
3885 } else {
3886 /* this is (1) */
3887 mutex_unlock(&fs_info->balance_mutex);
3888 pr_info("btrfs: dev add/delete/balance/replace/resize operation in progress\n");
3889 ret = -EINVAL;
3890 goto out;
3891 }
3892
3893locked:
3894 BUG_ON(!atomic_read(&fs_info->mutually_exclusive_operation_running));
c9e9f97b
ID		 3895
3896 if (arg) {
3897 bargs = memdup_user(arg, sizeof(*bargs));
3898 if (IS_ERR(bargs)) {
3899 ret = PTR_ERR(bargs);
ed0fb78f 3900 goto out_unlock;
c9e9f97b 3901 }
de322263
ID		 3902
3903 if (bargs->flags & BTRFS_BALANCE_RESUME) {
3904 if (!fs_info->balance_ctl) {
3905 ret = -ENOTCONN;
3906 goto out_bargs;
3907 }
3908
3909 bctl = fs_info->balance_ctl;
3910 spin_lock(&fs_info->balance_lock);
3911 bctl->flags |= BTRFS_BALANCE_RESUME;
3912 spin_unlock(&fs_info->balance_lock);
3913
3914 goto do_balance;
3915 }
c9e9f97b
ID		 3916 } else {
3917 bargs = NULL;
3918 }
3919
ed0fb78f 3920 if (fs_info->balance_ctl) {
837d5b6e
ID		 3921 ret = -EINPROGRESS;
3922 goto out_bargs;
3923 }
3924
c9e9f97b
ID		 3925 bctl = kzalloc(sizeof(*bctl), GFP_NOFS);
3926 if (!bctl) {
3927 ret = -ENOMEM;
3928 goto out_bargs;
3929 }
3930
3931 bctl->fs_info = fs_info;
3932 if (arg) {
3933 memcpy(&bctl->data, &bargs->data, sizeof(bctl->data));
3934 memcpy(&bctl->meta, &bargs->meta, sizeof(bctl->meta));
3935 memcpy(&bctl->sys, &bargs->sys, sizeof(bctl->sys));
3936
3937 bctl->flags = bargs->flags;
f43ffb60
ID		 3938 } else {
3939 /* balance everything - no filters */
3940 bctl->flags |= BTRFS_BALANCE_TYPE_MASK;
c9e9f97b
ID		 3941 }
3942
de322263 3943do_balance:
c9e9f97b 3944 /*
ed0fb78f
ID		 3945 * Ownership of bctl and mutually_exclusive_operation_running
3946 * goes to to btrfs_balance. bctl is freed in __cancel_balance,
3947 * or, if restriper was paused all the way until unmount, in
3948 * free_fs_info. mutually_exclusive_operation_running is
3949 * cleared in __cancel_balance.
c9e9f97b 3950 */
ed0fb78f
ID		 3951 need_unlock = false;
3952
3953 ret = btrfs_balance(bctl, bargs);
3954
c9e9f97b
ID		 3955 if (arg) {
3956 if (copy_to_user(arg, bargs, sizeof(*bargs)))
3957 ret = -EFAULT;
3958 }
3959
3960out_bargs:
3961 kfree(bargs);
ed0fb78f 3962out_unlock:
c9e9f97b
ID		 3963 mutex_unlock(&fs_info->balance_mutex);
3964 mutex_unlock(&fs_info->volume_mutex);
ed0fb78f
ID		 3965 if (need_unlock)
3966 atomic_set(&fs_info->mutually_exclusive_operation_running, 0);
3967out:
e54bfa31 3968 mnt_drop_write_file(file);
c9e9f97b
ID		 3969 return ret;
3970}
3971
837d5b6e
ID		 3972static long btrfs_ioctl_balance_ctl(struct btrfs_root *root, int cmd)
3973{
3974 if (!capable(CAP_SYS_ADMIN))
3975 return -EPERM;
3976
3977 switch (cmd) {
3978 case BTRFS_BALANCE_CTL_PAUSE:
3979 return btrfs_pause_balance(root->fs_info);
a7e99c69
ID		 3980 case BTRFS_BALANCE_CTL_CANCEL:
3981 return btrfs_cancel_balance(root->fs_info);
837d5b6e
ID		 3982 }
3983
3984 return -EINVAL;
3985}
3986
19a39dce
ID		 3987static long btrfs_ioctl_balance_progress(struct btrfs_root *root,
3988 void __user *arg)
3989{
3990 struct btrfs_fs_info *fs_info = root->fs_info;
3991 struct btrfs_ioctl_balance_args *bargs;
3992 int ret = 0;
3993
3994 if (!capable(CAP_SYS_ADMIN))
3995 return -EPERM;
3996
3997 mutex_lock(&fs_info->balance_mutex);
3998 if (!fs_info->balance_ctl) {
3999 ret = -ENOTCONN;
4000 goto out;
4001 }
4002
4003 bargs = kzalloc(sizeof(*bargs), GFP_NOFS);
4004 if (!bargs) {
4005 ret = -ENOMEM;
4006 goto out;
4007 }
4008
4009 update_ioctl_balance_args(fs_info, 1, bargs);
4010
4011 if (copy_to_user(arg, bargs, sizeof(*bargs)))
4012 ret = -EFAULT;
4013
4014 kfree(bargs);
4015out:
4016 mutex_unlock(&fs_info->balance_mutex);
4017 return ret;
4018}
4019
905b0dda 4020static long btrfs_ioctl_quota_ctl(struct file *file, void __user *arg)
5d13a37b 4021{
496ad9aa 4022 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4023 struct btrfs_ioctl_quota_ctl_args *sa;
4024 struct btrfs_trans_handle *trans = NULL;
4025 int ret;
4026 int err;
4027
4028 if (!capable(CAP_SYS_ADMIN))
4029 return -EPERM;
4030
905b0dda
MX		 4031 ret = mnt_want_write_file(file);
4032 if (ret)
4033 return ret;
5d13a37b
AJ		 4034
4035 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4036 if (IS_ERR(sa)) {
4037 ret = PTR_ERR(sa);
4038 goto drop_write;
4039 }
5d13a37b 4040
7708f029 4041 down_write(&root->fs_info->subvol_sem);
2f232036
JS		 4042 trans = btrfs_start_transaction(root->fs_info->tree_root, 2);
4043 if (IS_ERR(trans)) {
4044 ret = PTR_ERR(trans);
4045 goto out;
5d13a37b
AJ		 4046 }
4047
4048 switch (sa->cmd) {
4049 case BTRFS_QUOTA_CTL_ENABLE:
4050 ret = btrfs_quota_enable(trans, root->fs_info);
4051 break;
4052 case BTRFS_QUOTA_CTL_DISABLE:
4053 ret = btrfs_quota_disable(trans, root->fs_info);
4054 break;
5d13a37b
AJ		 4055 default:
4056 ret = -EINVAL;
4057 break;
4058 }
4059
2f232036
JS		 4060 err = btrfs_commit_transaction(trans, root->fs_info->tree_root);
4061 if (err && !ret)
4062 ret = err;
5d13a37b
AJ		 4063out:
4064 kfree(sa);
7708f029 4065 up_write(&root->fs_info->subvol_sem);
905b0dda
MX		 4066drop_write:
4067 mnt_drop_write_file(file);
5d13a37b
AJ		 4068 return ret;
4069}
4070
905b0dda 4071static long btrfs_ioctl_qgroup_assign(struct file *file, void __user *arg)
5d13a37b 4072{
496ad9aa 4073 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4074 struct btrfs_ioctl_qgroup_assign_args *sa;
4075 struct btrfs_trans_handle *trans;
4076 int ret;
4077 int err;
4078
4079 if (!capable(CAP_SYS_ADMIN))
4080 return -EPERM;
4081
905b0dda
MX		 4082 ret = mnt_want_write_file(file);
4083 if (ret)
4084 return ret;
5d13a37b
AJ		 4085
4086 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4087 if (IS_ERR(sa)) {
4088 ret = PTR_ERR(sa);
4089 goto drop_write;
4090 }
5d13a37b
AJ		 4091
4092 trans = btrfs_join_transaction(root);
4093 if (IS_ERR(trans)) {
4094 ret = PTR_ERR(trans);
4095 goto out;
4096 }
4097
4098 /* FIXME: check if the IDs really exist */
4099 if (sa->assign) {
4100 ret = btrfs_add_qgroup_relation(trans, root->fs_info,
4101 sa->src, sa->dst);
4102 } else {
4103 ret = btrfs_del_qgroup_relation(trans, root->fs_info,
4104 sa->src, sa->dst);
4105 }
4106
4107 err = btrfs_end_transaction(trans, root);
4108 if (err && !ret)
4109 ret = err;
4110
4111out:
4112 kfree(sa);
905b0dda
MX		 4113drop_write:
4114 mnt_drop_write_file(file);
5d13a37b
AJ		 4115 return ret;
4116}
4117
905b0dda 4118static long btrfs_ioctl_qgroup_create(struct file *file, void __user *arg)
5d13a37b 4119{
496ad9aa 4120 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4121 struct btrfs_ioctl_qgroup_create_args *sa;
4122 struct btrfs_trans_handle *trans;
4123 int ret;
4124 int err;
4125
4126 if (!capable(CAP_SYS_ADMIN))
4127 return -EPERM;
4128
905b0dda
MX		 4129 ret = mnt_want_write_file(file);
4130 if (ret)
4131 return ret;
5d13a37b
AJ		 4132
4133 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4134 if (IS_ERR(sa)) {
4135 ret = PTR_ERR(sa);
4136 goto drop_write;
4137 }
5d13a37b 4138
d86e56cf
MX		 4139 if (!sa->qgroupid) {
4140 ret = -EINVAL;
4141 goto out;
4142 }
4143
5d13a37b
AJ		 4144 trans = btrfs_join_transaction(root);
4145 if (IS_ERR(trans)) {
4146 ret = PTR_ERR(trans);
4147 goto out;
4148 }
4149
4150 /* FIXME: check if the IDs really exist */
4151 if (sa->create) {
4152 ret = btrfs_create_qgroup(trans, root->fs_info, sa->qgroupid,
4153 NULL);
4154 } else {
4155 ret = btrfs_remove_qgroup(trans, root->fs_info, sa->qgroupid);
4156 }
4157
4158 err = btrfs_end_transaction(trans, root);
4159 if (err && !ret)
4160 ret = err;
4161
4162out:
4163 kfree(sa);
905b0dda
MX		 4164drop_write:
4165 mnt_drop_write_file(file);
5d13a37b
AJ		 4166 return ret;
4167}
4168
905b0dda 4169static long btrfs_ioctl_qgroup_limit(struct file *file, void __user *arg)
5d13a37b 4170{
496ad9aa 4171 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
5d13a37b
AJ		 4172 struct btrfs_ioctl_qgroup_limit_args *sa;
4173 struct btrfs_trans_handle *trans;
4174 int ret;
4175 int err;
4176 u64 qgroupid;
4177
4178 if (!capable(CAP_SYS_ADMIN))
4179 return -EPERM;
4180
905b0dda
MX		 4181 ret = mnt_want_write_file(file);
4182 if (ret)
4183 return ret;
5d13a37b
AJ		 4184
4185 sa = memdup_user(arg, sizeof(*sa));
905b0dda
MX		 4186 if (IS_ERR(sa)) {
4187 ret = PTR_ERR(sa);
4188 goto drop_write;
4189 }
5d13a37b
AJ		 4190
4191 trans = btrfs_join_transaction(root);
4192 if (IS_ERR(trans)) {
4193 ret = PTR_ERR(trans);
4194 goto out;
4195 }
4196
4197 qgroupid = sa->qgroupid;
4198 if (!qgroupid) {
4199 /* take the current subvol as qgroup */
4200 qgroupid = root->root_key.objectid;
4201 }
4202
4203 /* FIXME: check if the IDs really exist */
4204 ret = btrfs_limit_qgroup(trans, root->fs_info, qgroupid, &sa->lim);
4205
4206 err = btrfs_end_transaction(trans, root);
4207 if (err && !ret)
4208 ret = err;
4209
4210out:
4211 kfree(sa);
905b0dda
MX		 4212drop_write:
4213 mnt_drop_write_file(file);
5d13a37b
AJ		 4214 return ret;
4215}
4216
2f232036
JS		 4217static long btrfs_ioctl_quota_rescan(struct file *file, void __user *arg)
4218{
6d0379ec 4219 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
2f232036
JS		 4220 struct btrfs_ioctl_quota_rescan_args *qsa;
4221 int ret;
4222
4223 if (!capable(CAP_SYS_ADMIN))
4224 return -EPERM;
4225
4226 ret = mnt_want_write_file(file);
4227 if (ret)
4228 return ret;
4229
4230 qsa = memdup_user(arg, sizeof(*qsa));
4231 if (IS_ERR(qsa)) {
4232 ret = PTR_ERR(qsa);
4233 goto drop_write;
4234 }
4235
4236 if (qsa->flags) {
4237 ret = -EINVAL;
4238 goto out;
4239 }
4240
4241 ret = btrfs_qgroup_rescan(root->fs_info);
4242
4243out:
4244 kfree(qsa);
4245drop_write:
4246 mnt_drop_write_file(file);
4247 return ret;
4248}
4249
4250static long btrfs_ioctl_quota_rescan_status(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 = 0;
4255
4256 if (!capable(CAP_SYS_ADMIN))
4257 return -EPERM;
4258
4259 qsa = kzalloc(sizeof(*qsa), GFP_NOFS);
4260 if (!qsa)
4261 return -ENOMEM;
4262
4263 if (root->fs_info->qgroup_flags & BTRFS_QGROUP_STATUS_FLAG_RESCAN) {
4264 qsa->flags = 1;
4265 qsa->progress = root->fs_info->qgroup_rescan_progress.objectid;
4266 }
4267
4268 if (copy_to_user(arg, qsa, sizeof(*qsa)))
4269 ret = -EFAULT;
4270
4271 kfree(qsa);
4272 return ret;
4273}
4274
57254b6e
JS		 4275static long btrfs_ioctl_quota_rescan_wait(struct file *file, void __user *arg)
4276{
4277 struct btrfs_root *root = BTRFS_I(fdentry(file)->d_inode)->root;
4278
4279 if (!capable(CAP_SYS_ADMIN))
4280 return -EPERM;
4281
4282 return btrfs_qgroup_wait_for_completion(root->fs_info);
4283}
4284
8ea05e3a
AB		 4285static long btrfs_ioctl_set_received_subvol(struct file *file,
4286 void __user *arg)
4287{
4288 struct btrfs_ioctl_received_subvol_args *sa = NULL;
496ad9aa 4289 struct inode *inode = file_inode(file);
8ea05e3a
AB		 4290 struct btrfs_root *root = BTRFS_I(inode)->root;
4291 struct btrfs_root_item *root_item = &root->root_item;
4292 struct btrfs_trans_handle *trans;
4293 struct timespec ct = CURRENT_TIME;
4294 int ret = 0;
4295
4296 ret = mnt_want_write_file(file);
4297 if (ret < 0)
4298 return ret;
4299
4300 down_write(&root->fs_info->subvol_sem);
4301
4302 if (btrfs_ino(inode) != BTRFS_FIRST_FREE_OBJECTID) {
4303 ret = -EINVAL;
4304 goto out;
4305 }
4306
4307 if (btrfs_root_readonly(root)) {
4308 ret = -EROFS;
4309 goto out;
4310 }
4311
4312 if (!inode_owner_or_capable(inode)) {
4313 ret = -EACCES;
4314 goto out;
4315 }
4316
4317 sa = memdup_user(arg, sizeof(*sa));
4318 if (IS_ERR(sa)) {
4319 ret = PTR_ERR(sa);
4320 sa = NULL;
4321 goto out;
4322 }
4323
4324 trans = btrfs_start_transaction(root, 1);
4325 if (IS_ERR(trans)) {
4326 ret = PTR_ERR(trans);
4327 trans = NULL;
4328 goto out;
4329 }
4330
4331 sa->rtransid = trans->transid;
4332 sa->rtime.sec = ct.tv_sec;
4333 sa->rtime.nsec = ct.tv_nsec;
4334
4335 memcpy(root_item->received_uuid, sa->uuid, BTRFS_UUID_SIZE);
4336 btrfs_set_root_stransid(root_item, sa->stransid);
4337 btrfs_set_root_rtransid(root_item, sa->rtransid);
3cae210f
QW		 4338 btrfs_set_stack_timespec_sec(&root_item->stime, sa->stime.sec);
4339 btrfs_set_stack_timespec_nsec(&root_item->stime, sa->stime.nsec);
4340 btrfs_set_stack_timespec_sec(&root_item->rtime, sa->rtime.sec);
4341 btrfs_set_stack_timespec_nsec(&root_item->rtime, sa->rtime.nsec);
8ea05e3a
AB		 4342
4343 ret = btrfs_update_root(trans, root->fs_info->tree_root,
4344 &root->root_key, &root->root_item);
4345 if (ret < 0) {
4346 btrfs_end_transaction(trans, root);
4347 trans = NULL;
4348 goto out;
4349 } else {
4350 ret = btrfs_commit_transaction(trans, root);
4351 if (ret < 0)
4352 goto out;
4353 }
4354
4355 ret = copy_to_user(arg, sa, sizeof(*sa));
4356 if (ret)
4357 ret = -EFAULT;
4358
4359out:
4360 kfree(sa);
4361 up_write(&root->fs_info->subvol_sem);
4362 mnt_drop_write_file(file);
4363 return ret;
4364}
4365
867ab667 4366static int btrfs_ioctl_get_fslabel(struct file *file, void __user *arg)
4367{
6d0379ec 4368 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
a1b83ac5 4369 size_t len;
867ab667 4370 int ret;
a1b83ac5
AJ		 4371 char label[BTRFS_LABEL_SIZE];
4372
4373 spin_lock(&root->fs_info->super_lock);
4374 memcpy(label, root->fs_info->super_copy->label, BTRFS_LABEL_SIZE);
4375 spin_unlock(&root->fs_info->super_lock);
4376
4377 len = strnlen(label, BTRFS_LABEL_SIZE);
867ab667 4378
4379 if (len == BTRFS_LABEL_SIZE) {
4380 pr_warn("btrfs: label is too long, return the first %zu bytes\n",
4381 --len);
4382 }
4383
867ab667 4384 ret = copy_to_user(arg, label, len);
867ab667 4385
4386 return ret ? -EFAULT : 0;
4387}
4388
a8bfd4ab 4389static int btrfs_ioctl_set_fslabel(struct file *file, void __user *arg)
4390{
6d0379ec 4391 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
a8bfd4ab 4392 struct btrfs_super_block *super_block = root->fs_info->super_copy;
4393 struct btrfs_trans_handle *trans;
4394 char label[BTRFS_LABEL_SIZE];
4395 int ret;
4396
4397 if (!capable(CAP_SYS_ADMIN))
4398 return -EPERM;
4399
4400 if (copy_from_user(label, arg, sizeof(label)))
4401 return -EFAULT;
4402
4403 if (strnlen(label, BTRFS_LABEL_SIZE) == BTRFS_LABEL_SIZE) {
4404 pr_err("btrfs: unable to set label with more than %d bytes\n",
4405 BTRFS_LABEL_SIZE - 1);
4406 return -EINVAL;
4407 }
4408
4409 ret = mnt_want_write_file(file);
4410 if (ret)
4411 return ret;
4412
a8bfd4ab 4413 trans = btrfs_start_transaction(root, 0);
4414 if (IS_ERR(trans)) {
4415 ret = PTR_ERR(trans);
4416 goto out_unlock;
4417 }
4418
a1b83ac5 4419 spin_lock(&root->fs_info->super_lock);
a8bfd4ab 4420 strcpy(super_block->label, label);
a1b83ac5 4421 spin_unlock(&root->fs_info->super_lock);
a8bfd4ab 4422 ret = btrfs_end_transaction(trans, root);
4423
4424out_unlock:
a8bfd4ab 4425 mnt_drop_write_file(file);
4426 return ret;
4427}
4428
f46b5a66
CH		 4429long btrfs_ioctl(struct file *file, unsigned int
4430 cmd, unsigned long arg)
4431{
496ad9aa 4432 struct btrfs_root *root = BTRFS_I(file_inode(file))->root;
4bcabaa3 4433 void __user *argp = (void __user *)arg;
f46b5a66
CH		 4434
4435 switch (cmd) {
6cbff00f
CH		 4436 case FS_IOC_GETFLAGS:
4437 return btrfs_ioctl_getflags(file, argp);
4438 case FS_IOC_SETFLAGS:
4439 return btrfs_ioctl_setflags(file, argp);
4440 case FS_IOC_GETVERSION:
4441 return btrfs_ioctl_getversion(file, argp);
f7039b1d
LD		 4442 case FITRIM:
4443 return btrfs_ioctl_fitrim(file, argp);
f46b5a66 4444 case BTRFS_IOC_SNAP_CREATE:
fa0d2b9b 4445 return btrfs_ioctl_snap_create(file, argp, 0);
fdfb1e4f 4446 case BTRFS_IOC_SNAP_CREATE_V2:
fa0d2b9b 4447 return btrfs_ioctl_snap_create_v2(file, argp, 0);
3de4586c 4448 case BTRFS_IOC_SUBVOL_CREATE:
fa0d2b9b 4449 return btrfs_ioctl_snap_create(file, argp, 1);
6f72c7e2
AJ		 4450 case BTRFS_IOC_SUBVOL_CREATE_V2:
4451 return btrfs_ioctl_snap_create_v2(file, argp, 1);
76dda93c
YZ		 4452 case BTRFS_IOC_SNAP_DESTROY:
4453 return btrfs_ioctl_snap_destroy(file, argp);
0caa102d
LZ		 4454 case BTRFS_IOC_SUBVOL_GETFLAGS:
4455 return btrfs_ioctl_subvol_getflags(file, argp);
4456 case BTRFS_IOC_SUBVOL_SETFLAGS:
4457 return btrfs_ioctl_subvol_setflags(file, argp);
6ef5ed0d
JB		 4458 case BTRFS_IOC_DEFAULT_SUBVOL:
4459 return btrfs_ioctl_default_subvol(file, argp);
f46b5a66 4460 case BTRFS_IOC_DEFRAG:
1e701a32
CM		 4461 return btrfs_ioctl_defrag(file, NULL);
4462 case BTRFS_IOC_DEFRAG_RANGE:
4463 return btrfs_ioctl_defrag(file, argp);
f46b5a66 4464 case BTRFS_IOC_RESIZE:
198605a8 4465 return btrfs_ioctl_resize(file, argp);
f46b5a66 4466 case BTRFS_IOC_ADD_DEV:
4bcabaa3 4467 return btrfs_ioctl_add_dev(root, argp);
f46b5a66 4468 case BTRFS_IOC_RM_DEV:
da24927b 4469 return btrfs_ioctl_rm_dev(file, argp);
475f6387
JS		 4470 case BTRFS_IOC_FS_INFO:
4471 return btrfs_ioctl_fs_info(root, argp);
4472 case BTRFS_IOC_DEV_INFO:
4473 return btrfs_ioctl_dev_info(root, argp);
f46b5a66 4474 case BTRFS_IOC_BALANCE:
9ba1f6e4 4475 return btrfs_ioctl_balance(file, NULL);
f46b5a66 4476 case BTRFS_IOC_CLONE:
c5c9cd4d
SW		 4477 return btrfs_ioctl_clone(file, arg, 0, 0, 0);
4478 case BTRFS_IOC_CLONE_RANGE:
7a865e8a 4479 return btrfs_ioctl_clone_range(file, argp);
f46b5a66
CH		 4480 case BTRFS_IOC_TRANS_START:
4481 return btrfs_ioctl_trans_start(file);
4482 case BTRFS_IOC_TRANS_END:
4483 return btrfs_ioctl_trans_end(file);
ac8e9819
CM		 4484 case BTRFS_IOC_TREE_SEARCH:
4485 return btrfs_ioctl_tree_search(file, argp);
4486 case BTRFS_IOC_INO_LOOKUP:
4487 return btrfs_ioctl_ino_lookup(file, argp);
d7728c96
JS		 4488 case BTRFS_IOC_INO_PATHS:
4489 return btrfs_ioctl_ino_to_path(root, argp);
4490 case BTRFS_IOC_LOGICAL_INO:
4491 return btrfs_ioctl_logical_to_ino(root, argp);
1406e432
JB		 4492 case BTRFS_IOC_SPACE_INFO:
4493 return btrfs_ioctl_space_info(root, argp);
f46b5a66
CH		 4494 case BTRFS_IOC_SYNC:
4495 btrfs_sync_fs(file->f_dentry->d_sb, 1);
4496 return 0;
46204592 4497 case BTRFS_IOC_START_SYNC:
9a8c28be 4498 return btrfs_ioctl_start_sync(root, argp);
46204592 4499 case BTRFS_IOC_WAIT_SYNC:
9a8c28be 4500 return btrfs_ioctl_wait_sync(root, argp);
475f6387 4501 case BTRFS_IOC_SCRUB:
b8e95489 4502 return btrfs_ioctl_scrub(file, argp);
475f6387
JS		 4503 case BTRFS_IOC_SCRUB_CANCEL:
4504 return btrfs_ioctl_scrub_cancel(root, argp);
4505 case BTRFS_IOC_SCRUB_PROGRESS:
4506 return btrfs_ioctl_scrub_progress(root, argp);
c9e9f97b 4507 case BTRFS_IOC_BALANCE_V2:
9ba1f6e4 4508 return btrfs_ioctl_balance(file, argp);
837d5b6e
ID		 4509 case BTRFS_IOC_BALANCE_CTL:
4510 return btrfs_ioctl_balance_ctl(root, arg);
19a39dce
ID		 4511 case BTRFS_IOC_BALANCE_PROGRESS:
4512 return btrfs_ioctl_balance_progress(root, argp);
8ea05e3a
AB		 4513 case BTRFS_IOC_SET_RECEIVED_SUBVOL:
4514 return btrfs_ioctl_set_received_subvol(file, argp);
31db9f7c
AB		 4515 case BTRFS_IOC_SEND:
4516 return btrfs_ioctl_send(file, argp);
c11d2c23 4517 case BTRFS_IOC_GET_DEV_STATS:
b27f7c0c 4518 return btrfs_ioctl_get_dev_stats(root, argp);
5d13a37b 4519 case BTRFS_IOC_QUOTA_CTL:
905b0dda 4520 return btrfs_ioctl_quota_ctl(file, argp);
5d13a37b 4521 case BTRFS_IOC_QGROUP_ASSIGN:
905b0dda 4522 return btrfs_ioctl_qgroup_assign(file, argp);
5d13a37b 4523 case BTRFS_IOC_QGROUP_CREATE:
905b0dda 4524 return btrfs_ioctl_qgroup_create(file, argp);
5d13a37b 4525 case BTRFS_IOC_QGROUP_LIMIT:
905b0dda 4526 return btrfs_ioctl_qgroup_limit(file, argp);
2f232036
JS		 4527 case BTRFS_IOC_QUOTA_RESCAN:
4528 return btrfs_ioctl_quota_rescan(file, argp);
4529 case BTRFS_IOC_QUOTA_RESCAN_STATUS:
4530 return btrfs_ioctl_quota_rescan_status(file, argp);
57254b6e
JS		 4531 case BTRFS_IOC_QUOTA_RESCAN_WAIT:
4532 return btrfs_ioctl_quota_rescan_wait(file, argp);
3f6bcfbd
SB		 4533 case BTRFS_IOC_DEV_REPLACE:
4534 return btrfs_ioctl_dev_replace(root, argp);
867ab667 4535 case BTRFS_IOC_GET_FSLABEL:
4536 return btrfs_ioctl_get_fslabel(file, argp);
a8bfd4ab 4537 case BTRFS_IOC_SET_FSLABEL:
4538 return btrfs_ioctl_set_fslabel(file, argp);
416161db
MF		 4539 case BTRFS_IOC_FILE_EXTENT_SAME:
4540 return btrfs_ioctl_file_extent_same(file, argp);
f46b5a66
CH		 4541 }
4542
4543 return -ENOTTY;
4544}
Linux 6.x block layer and io_uring tree(s)